The following changes since commit 39a6e4f87e7b75a45b08d6dc8b8b7c2954c87440: Merge tag 'pull-qapi-2024-02-03' of https://repo.or.cz/qemu/armbru into staging (2024-02-03 13:31:58 +0000) are available in the Git repository at: https://gitlab.com/rth7680/qemu.git tags/pull-tcg-20240205 for you to fetch changes up to 867db6870a6f5d4e0915822d6a84e665bec1f22e: tcg/tci: Support TCG_COND_TST{EQ,NE} (2024-02-03 23:53:49 +0000) ---------------------------------------------------------------- tcg: Introduce TCG_COND_TST{EQ,NE} target/alpha: Use TCG_COND_TST{EQ,NE} target/m68k: Use TCG_COND_TST{EQ,NE} in gen_fcc_cond target/sparc: Use TCG_COND_TSTEQ in gen_op_mulscc target/s390x: Use TCG_COND_TSTNE for CC_OP_{TM,ICM} target/s390x: Improve general case of disas_jcc ---------------------------------------------------------------- Paolo Bonzini (1): tcg/i386: Use TEST r,r to test 8/16/32 bits Philippe Mathieu-Daudé (1): tcg/aarch64: Massage tcg_out_brcond() Richard Henderson (37): tcg: Introduce TCG_COND_TST{EQ,NE} tcg: Introduce TCG_TARGET_HAS_tst tcg/optimize: Split out arg_is_const_val tcg/optimize: Split out do_constant_folding_cond1 tcg/optimize: Do swap_commutative2 in do_constant_folding_cond2 tcg/optimize: Handle TCG_COND_TST{EQ,NE} tcg/optimize: Lower TCG_COND_TST{EQ,NE} if unsupported target/alpha: Pass immediate value to gen_bcond_internal() target/alpha: Use TCG_COND_TST{EQ,NE} for BLB{C,S} target/alpha: Use TCG_COND_TST{EQ,NE} for CMOVLB{C,S} target/alpha: Use TCG_COND_TSTNE for gen_fold_mzero target/m68k: Use TCG_COND_TST{EQ,NE} in gen_fcc_cond target/sparc: Use TCG_COND_TSTEQ in gen_op_mulscc target/s390x: Use TCG_COND_TSTNE for CC_OP_{TM,ICM} target/s390x: Improve general case of disas_jcc tcg: Add TCGConst argument to tcg_target_const_match tcg/aarch64: Support TCG_COND_TST{EQ,NE} tcg/aarch64: Generate TBZ, TBNZ tcg/aarch64: Generate CBNZ for TSTNE of UINT32_MAX tcg/arm: Split out tcg_out_cmp() tcg/arm: Support TCG_COND_TST{EQ,NE} tcg/i386: Pass x86 condition codes to tcg_out_cmov tcg/i386: Move tcg_cond_to_jcc[] into tcg_out_cmp tcg/i386: Support TCG_COND_TST{EQ,NE} tcg/i386: Improve TSTNE/TESTEQ vs powers of two tcg/sparc64: Hoist read of tcg_cond_to_rcond tcg/sparc64: Pass TCGCond to tcg_out_cmp tcg/sparc64: Support TCG_COND_TST{EQ,NE} tcg/ppc: Sink tcg_to_bc usage into tcg_out_bc tcg/ppc: Use cr0 in tcg_to_bc and tcg_to_isel tcg/ppc: Tidy up tcg_target_const_match tcg/ppc: Add TCG_CT_CONST_CMP tcg/ppc: Support TCG_COND_TST{EQ,NE} tcg/s390x: Split constraint A into J+U tcg/s390x: Add TCG_CT_CONST_CMP tcg/s390x: Support TCG_COND_TST{EQ,NE} tcg/tci: Support TCG_COND_TST{EQ,NE} include/tcg/tcg-cond.h | 74 ++++-- tcg/aarch64/tcg-target-con-set.h | 5 +- tcg/aarch64/tcg-target-con-str.h | 1 + tcg/aarch64/tcg-target.h | 2 + tcg/arm/tcg-target.h | 2 + tcg/i386/tcg-target-con-set.h | 6 +- tcg/i386/tcg-target-con-str.h | 1 + tcg/i386/tcg-target.h | 2 + tcg/loongarch64/tcg-target.h | 2 + tcg/mips/tcg-target.h | 2 + tcg/ppc/tcg-target-con-set.h | 5 +- tcg/ppc/tcg-target-con-str.h | 1 + tcg/ppc/tcg-target.h | 2 + tcg/riscv/tcg-target.h | 2 + tcg/s390x/tcg-target-con-set.h | 8 +- tcg/s390x/tcg-target-con-str.h | 3 +- tcg/s390x/tcg-target.h | 2 + tcg/sparc64/tcg-target.h | 2 + tcg/tcg-internal.h | 2 + tcg/tci/tcg-target.h | 2 + target/alpha/translate.c | 94 ++++---- target/m68k/translate.c | 74 +++--- target/s390x/tcg/translate.c | 100 +++------ target/sparc/translate.c | 4 +- tcg/optimize.c | 474 ++++++++++++++++++++++++++++++--------- tcg/tcg.c | 40 +++- tcg/tci.c | 14 ++ docs/devel/tcg-ops.rst | 2 + tcg/aarch64/tcg-target.c.inc | 166 +++++++++++--- tcg/arm/tcg-target.c.inc | 62 +++-- tcg/i386/tcg-target.c.inc | 201 ++++++++++++----- tcg/loongarch64/tcg-target.c.inc | 3 +- tcg/mips/tcg-target.c.inc | 3 +- tcg/ppc/tcg-target.c.inc | 294 ++++++++++++++++++------ tcg/riscv/tcg-target.c.inc | 3 +- tcg/s390x/tcg-target.c.inc | 246 +++++++++++++------- tcg/sparc64/tcg-target.c.inc | 65 ++++-- tcg/tci/tcg-target.c.inc | 3 +- 38 files changed, 1379 insertions(+), 595 deletions(-) Add the enumerators, adjust the helpers to match, and dump. Not supported anywhere else just yet. Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- include/tcg/tcg-cond.h | 74 ++++++++++++++++++++++++++++++------------ tcg/tcg.c | 4 ++- docs/devel/tcg-ops.rst | 2 ++ 3 files changed, 58 insertions(+), 22 deletions(-) diff --git a/include/tcg/tcg-cond.h b/include/tcg/tcg-cond.h index XXXXXXX..XXXXXXX 100644 --- a/include/tcg/tcg-cond.h +++ b/include/tcg/tcg-cond.h @@ -XXX,XX +XXX,XX @@ * Conditions. Note that these are laid out for easy manipulation by * the functions below: * bit 0 is used for inverting; - * bit 1 is signed, - * bit 2 is unsigned, - * bit 3 is used with bit 0 for swapping signed/unsigned. + * bit 1 is used for conditions that need swapping (signed/unsigned). + * bit 2 is used with bit 1 for swapping. + * bit 3 is used for unsigned conditions. */ typedef enum { /* non-signed */ TCG_COND_NEVER = 0 | 0 | 0 | 0, TCG_COND_ALWAYS = 0 | 0 | 0 | 1, + + /* equality */ TCG_COND_EQ = 8 | 0 | 0 | 0, TCG_COND_NE = 8 | 0 | 0 | 1, + + /* "test" i.e. and then compare vs 0 */ + TCG_COND_TSTEQ = 8 | 4 | 0 | 0, + TCG_COND_TSTNE = 8 | 4 | 0 | 1, + /* signed */ TCG_COND_LT = 0 | 0 | 2 | 0, TCG_COND_GE = 0 | 0 | 2 | 1, - TCG_COND_LE = 8 | 0 | 2 | 0, - TCG_COND_GT = 8 | 0 | 2 | 1, + TCG_COND_GT = 0 | 4 | 2 | 0, + TCG_COND_LE = 0 | 4 | 2 | 1, + /* unsigned */ - TCG_COND_LTU = 0 | 4 | 0 | 0, - TCG_COND_GEU = 0 | 4 | 0 | 1, - TCG_COND_LEU = 8 | 4 | 0 | 0, - TCG_COND_GTU = 8 | 4 | 0 | 1, + TCG_COND_LTU = 8 | 0 | 2 | 0, + TCG_COND_GEU = 8 | 0 | 2 | 1, + TCG_COND_GTU = 8 | 4 | 2 | 0, + TCG_COND_LEU = 8 | 4 | 2 | 1, } TCGCond; /* Invert the sense of the comparison. */ @@ -XXX,XX +XXX,XX @@ static inline TCGCond tcg_invert_cond(TCGCond c) /* Swap the operands in a comparison. */ static inline TCGCond tcg_swap_cond(TCGCond c) { - return c & 6 ? (TCGCond)(c ^ 9) : c; + return (TCGCond)(c ^ ((c & 2) << 1)); } -/* Create an "unsigned" version of a "signed" comparison. */ -static inline TCGCond tcg_unsigned_cond(TCGCond c) +/* Must a comparison be considered signed? */ +static inline bool is_signed_cond(TCGCond c) { - return c & 2 ? (TCGCond)(c ^ 6) : c; -} - -/* Create a "signed" version of an "unsigned" comparison. */ -static inline TCGCond tcg_signed_cond(TCGCond c) -{ - return c & 4 ? (TCGCond)(c ^ 6) : c; + return (c & (8 | 2)) == 2; } /* Must a comparison be considered unsigned? */ static inline bool is_unsigned_cond(TCGCond c) { - return (c & 4) != 0; + return (c & (8 | 2)) == (8 | 2); +} + +/* Must a comparison be considered a test? */ +static inline bool is_tst_cond(TCGCond c) +{ + return (c | 1) == TCG_COND_TSTNE; +} + +/* Create an "unsigned" version of a "signed" comparison. */ +static inline TCGCond tcg_unsigned_cond(TCGCond c) +{ + return is_signed_cond(c) ? (TCGCond)(c + 8) : c; +} + +/* Create a "signed" version of an "unsigned" comparison. */ +static inline TCGCond tcg_signed_cond(TCGCond c) +{ + return is_unsigned_cond(c) ? (TCGCond)(c - 8) : c; +} + +/* Create the eq/ne version of a tsteq/tstne comparison. */ +static inline TCGCond tcg_tst_eqne_cond(TCGCond c) +{ + return is_tst_cond(c) ? (TCGCond)(c - 4) : c; +} + +/* Create the lt/ge version of a tstne/tsteq comparison of the sign. */ +static inline TCGCond tcg_tst_ltge_cond(TCGCond c) +{ + return is_tst_cond(c) ? (TCGCond)(c ^ 0xf) : c; } /* @@ -XXX,XX +XXX,XX @@ static inline TCGCond tcg_high_cond(TCGCond c) case TCG_COND_LE: case TCG_COND_GEU: case TCG_COND_LEU: - return (TCGCond)(c ^ 8); + return (TCGCond)(c ^ (4 | 1)); default: return c; } diff --git a/tcg/tcg.c b/tcg/tcg.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/tcg.c +++ b/tcg/tcg.c @@ -XXX,XX +XXX,XX @@ static const char * const cond_name[] = [TCG_COND_LTU] = "ltu", [TCG_COND_GEU] = "geu", [TCG_COND_LEU] = "leu", - [TCG_COND_GTU] = "gtu" + [TCG_COND_GTU] = "gtu", + [TCG_COND_TSTEQ] = "tsteq", + [TCG_COND_TSTNE] = "tstne", }; static const char * const ldst_name[(MO_BSWAP | MO_SSIZE) + 1] = diff --git a/docs/devel/tcg-ops.rst b/docs/devel/tcg-ops.rst index XXXXXXX..XXXXXXX 100644 --- a/docs/devel/tcg-ops.rst +++ b/docs/devel/tcg-ops.rst @@ -XXX,XX +XXX,XX @@ Jumps/Labels | ``TCG_COND_GEU /* unsigned */`` | ``TCG_COND_LEU /* unsigned */`` | ``TCG_COND_GTU /* unsigned */`` + | ``TCG_COND_TSTEQ /* t1 & t2 == 0 */`` + | ``TCG_COND_TSTNE /* t1 & t2 != 0 */`` Arithmetic ---------- -- 2.34.1 Define as 0 for all tcg backends. Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/aarch64/tcg-target.h | 2 ++ tcg/arm/tcg-target.h | 2 ++ tcg/i386/tcg-target.h | 2 ++ tcg/loongarch64/tcg-target.h | 2 ++ tcg/mips/tcg-target.h | 2 ++ tcg/ppc/tcg-target.h | 2 ++ tcg/riscv/tcg-target.h | 2 ++ tcg/s390x/tcg-target.h | 2 ++ tcg/sparc64/tcg-target.h | 2 ++ tcg/tci/tcg-target.h | 2 ++ 10 files changed, 20 insertions(+) diff --git a/tcg/aarch64/tcg-target.h b/tcg/aarch64/tcg-target.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/aarch64/tcg-target.h +++ b/tcg/aarch64/tcg-target.h @@ -XXX,XX +XXX,XX @@ typedef enum { #define TCG_TARGET_HAS_qemu_ldst_i128 1 #endif +#define TCG_TARGET_HAS_tst 0 + #define TCG_TARGET_HAS_v64 1 #define TCG_TARGET_HAS_v128 1 #define TCG_TARGET_HAS_v256 0 diff --git a/tcg/arm/tcg-target.h b/tcg/arm/tcg-target.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/arm/tcg-target.h +++ b/tcg/arm/tcg-target.h @@ -XXX,XX +XXX,XX @@ extern bool use_neon_instructions; #define TCG_TARGET_HAS_qemu_ldst_i128 0 +#define TCG_TARGET_HAS_tst 0 + #define TCG_TARGET_HAS_v64 use_neon_instructions #define TCG_TARGET_HAS_v128 use_neon_instructions #define TCG_TARGET_HAS_v256 0 diff --git a/tcg/i386/tcg-target.h b/tcg/i386/tcg-target.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/i386/tcg-target.h +++ b/tcg/i386/tcg-target.h @@ -XXX,XX +XXX,XX @@ typedef enum { #define TCG_TARGET_HAS_qemu_ldst_i128 \ (TCG_TARGET_REG_BITS == 64 && (cpuinfo & CPUINFO_ATOMIC_VMOVDQA)) +#define TCG_TARGET_HAS_tst 0 + /* We do not support older SSE systems, only beginning with AVX1. */ #define TCG_TARGET_HAS_v64 have_avx1 #define TCG_TARGET_HAS_v128 have_avx1 diff --git a/tcg/loongarch64/tcg-target.h b/tcg/loongarch64/tcg-target.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/loongarch64/tcg-target.h +++ b/tcg/loongarch64/tcg-target.h @@ -XXX,XX +XXX,XX @@ typedef enum { #define TCG_TARGET_HAS_qemu_ldst_i128 (cpuinfo & CPUINFO_LSX) +#define TCG_TARGET_HAS_tst 0 + #define TCG_TARGET_HAS_v64 0 #define TCG_TARGET_HAS_v128 (cpuinfo & CPUINFO_LSX) #define TCG_TARGET_HAS_v256 0 diff --git a/tcg/mips/tcg-target.h b/tcg/mips/tcg-target.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/mips/tcg-target.h +++ b/tcg/mips/tcg-target.h @@ -XXX,XX +XXX,XX @@ extern bool use_mips32r2_instructions; #define TCG_TARGET_HAS_qemu_ldst_i128 0 +#define TCG_TARGET_HAS_tst 0 + #define TCG_TARGET_DEFAULT_MO 0 #define TCG_TARGET_NEED_LDST_LABELS #define TCG_TARGET_NEED_POOL_LABELS diff --git a/tcg/ppc/tcg-target.h b/tcg/ppc/tcg-target.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/ppc/tcg-target.h +++ b/tcg/ppc/tcg-target.h @@ -XXX,XX +XXX,XX @@ typedef enum { #define TCG_TARGET_HAS_qemu_ldst_i128 \ (TCG_TARGET_REG_BITS == 64 && have_isa_2_07) +#define TCG_TARGET_HAS_tst 0 + /* * While technically Altivec could support V64, it has no 64-bit store * instruction and substituting two 32-bit stores makes the generated diff --git a/tcg/riscv/tcg-target.h b/tcg/riscv/tcg-target.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/riscv/tcg-target.h +++ b/tcg/riscv/tcg-target.h @@ -XXX,XX +XXX,XX @@ extern bool have_zbb; #define TCG_TARGET_HAS_qemu_ldst_i128 0 +#define TCG_TARGET_HAS_tst 0 + #define TCG_TARGET_DEFAULT_MO (0) #define TCG_TARGET_NEED_LDST_LABELS diff --git a/tcg/s390x/tcg-target.h b/tcg/s390x/tcg-target.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/s390x/tcg-target.h +++ b/tcg/s390x/tcg-target.h @@ -XXX,XX +XXX,XX @@ extern uint64_t s390_facilities[3]; #define TCG_TARGET_HAS_qemu_ldst_i128 1 +#define TCG_TARGET_HAS_tst 0 + #define TCG_TARGET_HAS_v64 HAVE_FACILITY(VECTOR) #define TCG_TARGET_HAS_v128 HAVE_FACILITY(VECTOR) #define TCG_TARGET_HAS_v256 0 diff --git a/tcg/sparc64/tcg-target.h b/tcg/sparc64/tcg-target.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/sparc64/tcg-target.h +++ b/tcg/sparc64/tcg-target.h @@ -XXX,XX +XXX,XX @@ extern bool use_vis3_instructions; #define TCG_TARGET_HAS_qemu_ldst_i128 0 +#define TCG_TARGET_HAS_tst 0 + #define TCG_AREG0 TCG_REG_I0 #define TCG_TARGET_DEFAULT_MO (0) diff --git a/tcg/tci/tcg-target.h b/tcg/tci/tcg-target.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/tci/tcg-target.h +++ b/tcg/tci/tcg-target.h @@ -XXX,XX +XXX,XX @@ #define TCG_TARGET_HAS_qemu_ldst_i128 0 +#define TCG_TARGET_HAS_tst 0 + /* Number of registers available. */ #define TCG_TARGET_NB_REGS 16 -- 2.34.1 Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 38 +++++++++++++++++++++++--------------- 1 file changed, 23 insertions(+), 15 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static inline bool ts_is_const(TCGTemp *ts) return ts_info(ts)->is_const; } +static inline bool ts_is_const_val(TCGTemp *ts, uint64_t val) +{ + TempOptInfo *ti = ts_info(ts); + return ti->is_const && ti->val == val; +} + static inline bool arg_is_const(TCGArg arg) { return ts_is_const(arg_temp(arg)); } +static inline bool arg_is_const_val(TCGArg arg, uint64_t val) +{ + return ts_is_const_val(arg_temp(arg), val); +} + static inline bool ts_is_copy(TCGTemp *ts) { return ts_info(ts)->next_copy != ts; @@ -XXX,XX +XXX,XX @@ static int do_constant_folding_cond(TCGType type, TCGArg x, } } else if (args_are_copies(x, y)) { return do_constant_folding_cond_eq(c); - } else if (arg_is_const(y) && arg_info(y)->val == 0) { + } else if (arg_is_const_val(y, 0)) { switch (c) { case TCG_COND_LTU: return 0; @@ -XXX,XX +XXX,XX @@ static bool fold_to_not(OptContext *ctx, TCGOp *op, int idx) /* If the binary operation has first argument @i, fold to @i. */ static bool fold_ix_to_i(OptContext *ctx, TCGOp *op, uint64_t i) { - if (arg_is_const(op->args[1]) && arg_info(op->args[1])->val == i) { + if (arg_is_const_val(op->args[1], i)) { return tcg_opt_gen_movi(ctx, op, op->args[0], i); } return false; @@ -XXX,XX +XXX,XX @@ static bool fold_ix_to_i(OptContext *ctx, TCGOp *op, uint64_t i) /* If the binary operation has first argument @i, fold to NOT. */ static bool fold_ix_to_not(OptContext *ctx, TCGOp *op, uint64_t i) { - if (arg_is_const(op->args[1]) && arg_info(op->args[1])->val == i) { + if (arg_is_const_val(op->args[1], i)) { return fold_to_not(ctx, op, 2); } return false; @@ -XXX,XX +XXX,XX @@ static bool fold_ix_to_not(OptContext *ctx, TCGOp *op, uint64_t i) /* If the binary operation has second argument @i, fold to @i. */ static bool fold_xi_to_i(OptContext *ctx, TCGOp *op, uint64_t i) { - if (arg_is_const(op->args[2]) && arg_info(op->args[2])->val == i) { + if (arg_is_const_val(op->args[2], i)) { return tcg_opt_gen_movi(ctx, op, op->args[0], i); } return false; @@ -XXX,XX +XXX,XX @@ static bool fold_xi_to_i(OptContext *ctx, TCGOp *op, uint64_t i) /* If the binary operation has second argument @i, fold to identity. */ static bool fold_xi_to_x(OptContext *ctx, TCGOp *op, uint64_t i) { - if (arg_is_const(op->args[2]) && arg_info(op->args[2])->val == i) { + if (arg_is_const_val(op->args[2], i)) { return tcg_opt_gen_mov(ctx, op, op->args[0], op->args[1]); } return false; @@ -XXX,XX +XXX,XX @@ static bool fold_xi_to_x(OptContext *ctx, TCGOp *op, uint64_t i) /* If the binary operation has second argument @i, fold to NOT. */ static bool fold_xi_to_not(OptContext *ctx, TCGOp *op, uint64_t i) { - if (arg_is_const(op->args[2]) && arg_info(op->args[2])->val == i) { + if (arg_is_const_val(op->args[2], i)) { return fold_to_not(ctx, op, 1); } return false; @@ -XXX,XX +XXX,XX @@ static bool fold_brcond2(OptContext *ctx, TCGOp *op) * Simplify LT/GE comparisons vs zero to a single compare * vs the high word of the input. */ - if (arg_is_const(op->args[2]) && arg_info(op->args[2])->val == 0 && - arg_is_const(op->args[3]) && arg_info(op->args[3])->val == 0) { + if (arg_is_const_val(op->args[2], 0) && + arg_is_const_val(op->args[3], 0)) { goto do_brcond_high; } break; @@ -XXX,XX +XXX,XX @@ static bool fold_deposit(OptContext *ctx, TCGOp *op) } /* Inserting a value into zero at offset 0. */ - if (arg_is_const(op->args[1]) - && arg_info(op->args[1])->val == 0 - && op->args[3] == 0) { + if (arg_is_const_val(op->args[1], 0) && op->args[3] == 0) { uint64_t mask = MAKE_64BIT_MASK(0, op->args[4]); op->opc = and_opc; @@ -XXX,XX +XXX,XX @@ static bool fold_deposit(OptContext *ctx, TCGOp *op) } /* Inserting zero into a value. */ - if (arg_is_const(op->args[2]) - && arg_info(op->args[2])->val == 0) { + if (arg_is_const_val(op->args[2], 0)) { uint64_t mask = deposit64(-1, op->args[3], op->args[4], 0); op->opc = and_opc; @@ -XXX,XX +XXX,XX @@ static bool fold_setcond2(OptContext *ctx, TCGOp *op) * Simplify LT/GE comparisons vs zero to a single compare * vs the high word of the input. */ - if (arg_is_const(op->args[3]) && arg_info(op->args[3])->val == 0 && - arg_is_const(op->args[4]) && arg_info(op->args[4])->val == 0) { + if (arg_is_const_val(op->args[3], 0) && + arg_is_const_val(op->args[4], 0)) { goto do_setcond_high; } break; -- 2.34.1 Handle modifications to the arguments and condition in a single place. Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 57 ++++++++++++++++++++++++-------------------------- 1 file changed, 27 insertions(+), 30 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool swap_commutative2(TCGArg *p1, TCGArg *p2) return false; } +static int do_constant_folding_cond1(OptContext *ctx, TCGArg dest, + TCGArg *p1, TCGArg *p2, TCGArg *pcond) +{ + TCGCond cond; + bool swap; + int r; + + swap = swap_commutative(dest, p1, p2); + cond = *pcond; + if (swap) { + *pcond = cond = tcg_swap_cond(cond); + } + + r = do_constant_folding_cond(ctx->type, *p1, *p2, cond); + return r; +} + static void init_arguments(OptContext *ctx, TCGOp *op, int nb_args) { for (int i = 0; i < nb_args; i++) { @@ -XXX,XX +XXX,XX @@ static bool fold_andc(OptContext *ctx, TCGOp *op) static bool fold_brcond(OptContext *ctx, TCGOp *op) { - TCGCond cond = op->args[2]; - int i; - - if (swap_commutative(NO_DEST, &op->args[0], &op->args[1])) { - op->args[2] = cond = tcg_swap_cond(cond); - } - - i = do_constant_folding_cond(ctx->type, op->args[0], op->args[1], cond); + int i = do_constant_folding_cond1(ctx, NO_DEST, &op->args[0], + &op->args[1], &op->args[2]); if (i == 0) { tcg_op_remove(ctx->tcg, op); return true; @@ -XXX,XX +XXX,XX @@ static bool fold_mov(OptContext *ctx, TCGOp *op) static bool fold_movcond(OptContext *ctx, TCGOp *op) { - TCGCond cond = op->args[5]; int i; - if (swap_commutative(NO_DEST, &op->args[1], &op->args[2])) { - op->args[5] = cond = tcg_swap_cond(cond); - } /* * Canonicalize the "false" input reg to match the destination reg so * that the tcg backend can implement a "move if true" operation. */ if (swap_commutative(op->args[0], &op->args[4], &op->args[3])) { - op->args[5] = cond = tcg_invert_cond(cond); + op->args[5] = tcg_invert_cond(op->args[5]); } - i = do_constant_folding_cond(ctx->type, op->args[1], op->args[2], cond); + i = do_constant_folding_cond1(ctx, NO_DEST, &op->args[1], + &op->args[2], &op->args[5]); if (i >= 0) { return tcg_opt_gen_mov(ctx, op, op->args[0], op->args[4 - i]); } @@ -XXX,XX +XXX,XX @@ static bool fold_movcond(OptContext *ctx, TCGOp *op) uint64_t tv = arg_info(op->args[3])->val; uint64_t fv = arg_info(op->args[4])->val; TCGOpcode opc, negopc = 0; + TCGCond cond = op->args[5]; switch (ctx->type) { case TCG_TYPE_I32: @@ -XXX,XX +XXX,XX @@ static bool fold_remainder(OptContext *ctx, TCGOp *op) static bool fold_setcond(OptContext *ctx, TCGOp *op) { - TCGCond cond = op->args[3]; - int i; - - if (swap_commutative(op->args[0], &op->args[1], &op->args[2])) { - op->args[3] = cond = tcg_swap_cond(cond); - } - - i = do_constant_folding_cond(ctx->type, op->args[1], op->args[2], cond); + int i = do_constant_folding_cond1(ctx, op->args[0], &op->args[1], + &op->args[2], &op->args[3]); if (i >= 0) { return tcg_opt_gen_movi(ctx, op, op->args[0], i); } @@ -XXX,XX +XXX,XX @@ static bool fold_setcond(OptContext *ctx, TCGOp *op) static bool fold_negsetcond(OptContext *ctx, TCGOp *op) { - TCGCond cond = op->args[3]; - int i; - - if (swap_commutative(op->args[0], &op->args[1], &op->args[2])) { - op->args[3] = cond = tcg_swap_cond(cond); - } - - i = do_constant_folding_cond(ctx->type, op->args[1], op->args[2], cond); + int i = do_constant_folding_cond1(ctx, op->args[0], &op->args[1], + &op->args[2], &op->args[3]); if (i >= 0) { return tcg_opt_gen_movi(ctx, op, op->args[0], -i); } -- 2.34.1 Mirror the new do_constant_folding_cond1 by doing all argument and condition adjustment within one helper. Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 107 ++++++++++++++++++++++++++----------------------- 1 file changed, 57 insertions(+), 50 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static int do_constant_folding_cond(TCGType type, TCGArg x, return -1; } -/* - * Return -1 if the condition can't be simplified, - * and the result of the condition (0 or 1) if it can. - */ -static int do_constant_folding_cond2(TCGArg *p1, TCGArg *p2, TCGCond c) -{ - TCGArg al = p1[0], ah = p1[1]; - TCGArg bl = p2[0], bh = p2[1]; - - if (arg_is_const(bl) && arg_is_const(bh)) { - tcg_target_ulong blv = arg_info(bl)->val; - tcg_target_ulong bhv = arg_info(bh)->val; - uint64_t b = deposit64(blv, 32, 32, bhv); - - if (arg_is_const(al) && arg_is_const(ah)) { - tcg_target_ulong alv = arg_info(al)->val; - tcg_target_ulong ahv = arg_info(ah)->val; - uint64_t a = deposit64(alv, 32, 32, ahv); - return do_constant_folding_cond_64(a, b, c); - } - if (b == 0) { - switch (c) { - case TCG_COND_LTU: - return 0; - case TCG_COND_GEU: - return 1; - default: - break; - } - } - } - if (args_are_copies(al, bl) && args_are_copies(ah, bh)) { - return do_constant_folding_cond_eq(c); - } - return -1; -} - /** * swap_commutative: * @dest: TCGArg of the destination argument, or NO_DEST. @@ -XXX,XX +XXX,XX @@ static bool swap_commutative2(TCGArg *p1, TCGArg *p2) return false; } +/* + * Return -1 if the condition can't be simplified, + * and the result of the condition (0 or 1) if it can. + */ static int do_constant_folding_cond1(OptContext *ctx, TCGArg dest, TCGArg *p1, TCGArg *p2, TCGArg *pcond) { @@ -XXX,XX +XXX,XX @@ static int do_constant_folding_cond1(OptContext *ctx, TCGArg dest, return r; } +static int do_constant_folding_cond2(OptContext *ctx, TCGArg *args) +{ + TCGArg al, ah, bl, bh; + TCGCond c; + bool swap; + + swap = swap_commutative2(args, args + 2); + c = args[4]; + if (swap) { + args[4] = c = tcg_swap_cond(c); + } + + al = args[0]; + ah = args[1]; + bl = args[2]; + bh = args[3]; + + if (arg_is_const(bl) && arg_is_const(bh)) { + tcg_target_ulong blv = arg_info(bl)->val; + tcg_target_ulong bhv = arg_info(bh)->val; + uint64_t b = deposit64(blv, 32, 32, bhv); + + if (arg_is_const(al) && arg_is_const(ah)) { + tcg_target_ulong alv = arg_info(al)->val; + tcg_target_ulong ahv = arg_info(ah)->val; + uint64_t a = deposit64(alv, 32, 32, ahv); + return do_constant_folding_cond_64(a, b, c); + } + if (b == 0) { + switch (c) { + case TCG_COND_LTU: + return 0; + case TCG_COND_GEU: + return 1; + default: + break; + } + } + } + if (args_are_copies(al, bl) && args_are_copies(ah, bh)) { + return do_constant_folding_cond_eq(c); + } + return -1; +} + static void init_arguments(OptContext *ctx, TCGOp *op, int nb_args) { for (int i = 0; i < nb_args; i++) { @@ -XXX,XX +XXX,XX @@ static bool fold_brcond(OptContext *ctx, TCGOp *op) static bool fold_brcond2(OptContext *ctx, TCGOp *op) { - TCGCond cond = op->args[4]; - TCGArg label = op->args[5]; + TCGCond cond; + TCGArg label; int i, inv = 0; - if (swap_commutative2(&op->args[0], &op->args[2])) { - op->args[4] = cond = tcg_swap_cond(cond); - } - - i = do_constant_folding_cond2(&op->args[0], &op->args[2], cond); + i = do_constant_folding_cond2(ctx, &op->args[0]); + cond = op->args[4]; + label = op->args[5]; if (i >= 0) { goto do_brcond_const; } @@ -XXX,XX +XXX,XX @@ static bool fold_negsetcond(OptContext *ctx, TCGOp *op) static bool fold_setcond2(OptContext *ctx, TCGOp *op) { - TCGCond cond = op->args[5]; + TCGCond cond; int i, inv = 0; - if (swap_commutative2(&op->args[1], &op->args[3])) { - op->args[5] = cond = tcg_swap_cond(cond); - } - - i = do_constant_folding_cond2(&op->args[1], &op->args[3], cond); + i = do_constant_folding_cond2(ctx, &op->args[1]); + cond = op->args[5]; if (i >= 0) { goto do_setcond_const; } -- 2.34.1 Fold constant comparisons. Canonicalize "tst x,x" to equality vs zero. Canonicalize "tst x,sign" to sign test vs zero. Fold double-word comparisons with zero parts. Fold setcond of "tst x,pow2" to a bit extract. Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 240 ++++++++++++++++++++++++++++++++++++++++++++----- 1 file changed, 218 insertions(+), 22 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool do_constant_folding_cond_32(uint32_t x, uint32_t y, TCGCond c) return x <= y; case TCG_COND_GTU: return x > y; - default: - g_assert_not_reached(); + case TCG_COND_TSTEQ: + return (x & y) == 0; + case TCG_COND_TSTNE: + return (x & y) != 0; + case TCG_COND_ALWAYS: + case TCG_COND_NEVER: + break; } + g_assert_not_reached(); } static bool do_constant_folding_cond_64(uint64_t x, uint64_t y, TCGCond c) @@ -XXX,XX +XXX,XX @@ static bool do_constant_folding_cond_64(uint64_t x, uint64_t y, TCGCond c) return x <= y; case TCG_COND_GTU: return x > y; - default: - g_assert_not_reached(); + case TCG_COND_TSTEQ: + return (x & y) == 0; + case TCG_COND_TSTNE: + return (x & y) != 0; + case TCG_COND_ALWAYS: + case TCG_COND_NEVER: + break; } + g_assert_not_reached(); } -static bool do_constant_folding_cond_eq(TCGCond c) +static int do_constant_folding_cond_eq(TCGCond c) { switch (c) { case TCG_COND_GT: @@ -XXX,XX +XXX,XX @@ static bool do_constant_folding_cond_eq(TCGCond c) case TCG_COND_LEU: case TCG_COND_EQ: return 1; - default: - g_assert_not_reached(); + case TCG_COND_TSTEQ: + case TCG_COND_TSTNE: + return -1; + case TCG_COND_ALWAYS: + case TCG_COND_NEVER: + break; } + g_assert_not_reached(); } /* @@ -XXX,XX +XXX,XX @@ static int do_constant_folding_cond(TCGType type, TCGArg x, } else if (arg_is_const_val(y, 0)) { switch (c) { case TCG_COND_LTU: + case TCG_COND_TSTNE: return 0; case TCG_COND_GEU: + case TCG_COND_TSTEQ: return 1; default: return -1; @@ -XXX,XX +XXX,XX @@ static int do_constant_folding_cond1(OptContext *ctx, TCGArg dest, } r = do_constant_folding_cond(ctx->type, *p1, *p2, cond); - return r; + if (r >= 0) { + return r; + } + if (!is_tst_cond(cond)) { + return -1; + } + + /* + * TSTNE x,x -> NE x,0 + * TSTNE x,-1 -> NE x,0 + */ + if (args_are_copies(*p1, *p2) || arg_is_const_val(*p2, -1)) { + *p2 = arg_new_constant(ctx, 0); + *pcond = tcg_tst_eqne_cond(cond); + return -1; + } + + /* TSTNE x,sign -> LT x,0 */ + if (arg_is_const_val(*p2, (ctx->type == TCG_TYPE_I32 + ? INT32_MIN : INT64_MIN))) { + *p2 = arg_new_constant(ctx, 0); + *pcond = tcg_tst_ltge_cond(cond); + } + return -1; } static int do_constant_folding_cond2(OptContext *ctx, TCGArg *args) @@ -XXX,XX +XXX,XX @@ static int do_constant_folding_cond2(OptContext *ctx, TCGArg *args) TCGArg al, ah, bl, bh; TCGCond c; bool swap; + int r; swap = swap_commutative2(args, args + 2); c = args[4]; @@ -XXX,XX +XXX,XX @@ static int do_constant_folding_cond2(OptContext *ctx, TCGArg *args) tcg_target_ulong alv = arg_info(al)->val; tcg_target_ulong ahv = arg_info(ah)->val; uint64_t a = deposit64(alv, 32, 32, ahv); - return do_constant_folding_cond_64(a, b, c); + + r = do_constant_folding_cond_64(a, b, c); + if (r >= 0) { + return r; + } } + if (b == 0) { switch (c) { case TCG_COND_LTU: + case TCG_COND_TSTNE: return 0; case TCG_COND_GEU: + case TCG_COND_TSTEQ: return 1; default: break; } } + + /* TSTNE x,-1 -> NE x,0 */ + if (b == -1 && is_tst_cond(c)) { + args[3] = args[2] = arg_new_constant(ctx, 0); + args[4] = tcg_tst_eqne_cond(c); + return -1; + } + + /* TSTNE x,sign -> LT x,0 */ + if (b == INT64_MIN && is_tst_cond(c)) { + /* bl must be 0, so copy that to bh */ + args[3] = bl; + args[4] = tcg_tst_ltge_cond(c); + return -1; + } } + if (args_are_copies(al, bl) && args_are_copies(ah, bh)) { - return do_constant_folding_cond_eq(c); + r = do_constant_folding_cond_eq(c); + if (r >= 0) { + return r; + } + + /* TSTNE x,x -> NE x,0 */ + if (is_tst_cond(c)) { + args[3] = args[2] = arg_new_constant(ctx, 0); + args[4] = tcg_tst_eqne_cond(c); + return -1; + } } return -1; } @@ -XXX,XX +XXX,XX @@ static bool fold_brcond2(OptContext *ctx, TCGOp *op) case 0: goto do_brcond_const; case 1: - op->opc = INDEX_op_brcond_i32; - op->args[1] = op->args[2]; - op->args[2] = cond; - op->args[3] = label; - break; + goto do_brcond_low; + } + break; + + case TCG_COND_TSTEQ: + case TCG_COND_TSTNE: + if (arg_is_const_val(op->args[2], 0)) { + goto do_brcond_high; + } + if (arg_is_const_val(op->args[3], 0)) { + goto do_brcond_low; } break; default: break; + do_brcond_low: + op->opc = INDEX_op_brcond_i32; + op->args[1] = op->args[2]; + op->args[2] = cond; + op->args[3] = label; + return fold_brcond(ctx, op); + do_brcond_high: op->opc = INDEX_op_brcond_i32; op->args[0] = op->args[1]; op->args[1] = op->args[3]; op->args[2] = cond; op->args[3] = label; - break; + return fold_brcond(ctx, op); do_brcond_const: if (i == 0) { @@ -XXX,XX +XXX,XX @@ static bool fold_remainder(OptContext *ctx, TCGOp *op) return false; } +static void fold_setcond_tst_pow2(OptContext *ctx, TCGOp *op, bool neg) +{ + TCGOpcode and_opc, sub_opc, xor_opc, neg_opc, shr_opc, uext_opc, sext_opc; + TCGCond cond = op->args[3]; + TCGArg ret, src1, src2; + TCGOp *op2; + uint64_t val; + int sh; + bool inv; + + if (!is_tst_cond(cond) || !arg_is_const(op->args[2])) { + return; + } + + src2 = op->args[2]; + val = arg_info(src2)->val; + if (!is_power_of_2(val)) { + return; + } + sh = ctz64(val); + + switch (ctx->type) { + case TCG_TYPE_I32: + and_opc = INDEX_op_and_i32; + sub_opc = INDEX_op_sub_i32; + xor_opc = INDEX_op_xor_i32; + shr_opc = INDEX_op_shr_i32; + neg_opc = INDEX_op_neg_i32; + if (TCG_TARGET_extract_i32_valid(sh, 1)) { + uext_opc = TCG_TARGET_HAS_extract_i32 ? INDEX_op_extract_i32 : 0; + sext_opc = TCG_TARGET_HAS_sextract_i32 ? INDEX_op_sextract_i32 : 0; + } + break; + case TCG_TYPE_I64: + and_opc = INDEX_op_and_i64; + sub_opc = INDEX_op_sub_i64; + xor_opc = INDEX_op_xor_i64; + shr_opc = INDEX_op_shr_i64; + neg_opc = INDEX_op_neg_i64; + if (TCG_TARGET_extract_i64_valid(sh, 1)) { + uext_opc = TCG_TARGET_HAS_extract_i64 ? INDEX_op_extract_i64 : 0; + sext_opc = TCG_TARGET_HAS_sextract_i64 ? INDEX_op_sextract_i64 : 0; + } + break; + default: + g_assert_not_reached(); + } + + ret = op->args[0]; + src1 = op->args[1]; + inv = cond == TCG_COND_TSTEQ; + + if (sh && sext_opc && neg && !inv) { + op->opc = sext_opc; + op->args[1] = src1; + op->args[2] = sh; + op->args[3] = 1; + return; + } else if (sh && uext_opc) { + op->opc = uext_opc; + op->args[1] = src1; + op->args[2] = sh; + op->args[3] = 1; + } else { + if (sh) { + op2 = tcg_op_insert_before(ctx->tcg, op, shr_opc, 3); + op2->args[0] = ret; + op2->args[1] = src1; + op2->args[2] = arg_new_constant(ctx, sh); + src1 = ret; + } + op->opc = and_opc; + op->args[1] = src1; + op->args[2] = arg_new_constant(ctx, 1); + } + + if (neg && inv) { + op2 = tcg_op_insert_after(ctx->tcg, op, sub_opc, 3); + op2->args[0] = ret; + op2->args[1] = ret; + op2->args[2] = arg_new_constant(ctx, 1); + } else if (inv) { + op2 = tcg_op_insert_after(ctx->tcg, op, xor_opc, 3); + op2->args[0] = ret; + op2->args[1] = ret; + op2->args[2] = arg_new_constant(ctx, 1); + } else if (neg) { + op2 = tcg_op_insert_after(ctx->tcg, op, neg_opc, 2); + op2->args[0] = ret; + op2->args[1] = ret; + } +} + static bool fold_setcond(OptContext *ctx, TCGOp *op) { int i = do_constant_folding_cond1(ctx, op->args[0], &op->args[1], @@ -XXX,XX +XXX,XX @@ static bool fold_setcond(OptContext *ctx, TCGOp *op) if (i >= 0) { return tcg_opt_gen_movi(ctx, op, op->args[0], i); } + fold_setcond_tst_pow2(ctx, op, false); ctx->z_mask = 1; ctx->s_mask = smask_from_zmask(1); @@ -XXX,XX +XXX,XX @@ static bool fold_negsetcond(OptContext *ctx, TCGOp *op) if (i >= 0) { return tcg_opt_gen_movi(ctx, op, op->args[0], -i); } + fold_setcond_tst_pow2(ctx, op, true); /* Value is {0,-1} so all bits are repetitions of the sign. */ ctx->s_mask = -1; return false; } - static bool fold_setcond2(OptContext *ctx, TCGOp *op) { TCGCond cond; @@ -XXX,XX +XXX,XX @@ static bool fold_setcond2(OptContext *ctx, TCGOp *op) case 0: goto do_setcond_const; case 1: - op->args[2] = op->args[3]; - op->args[3] = cond; - op->opc = INDEX_op_setcond_i32; - break; + goto do_setcond_low; + } + break; + + case TCG_COND_TSTEQ: + case TCG_COND_TSTNE: + if (arg_is_const_val(op->args[2], 0)) { + goto do_setcond_high; + } + if (arg_is_const_val(op->args[4], 0)) { + goto do_setcond_low; } break; default: break; + do_setcond_low: + op->args[2] = op->args[3]; + op->args[3] = cond; + op->opc = INDEX_op_setcond_i32; + return fold_setcond(ctx, op); + do_setcond_high: op->args[1] = op->args[2]; op->args[2] = op->args[4]; op->args[3] = cond; op->opc = INDEX_op_setcond_i32; - break; + return fold_setcond(ctx, op); } ctx->z_mask = 1; -- 2.34.1 After having performed other simplifications, lower any remaining test comparisons with AND. Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/tcg-internal.h | 2 ++ tcg/optimize.c | 60 +++++++++++++++++++++++++++++++++++++++------- tcg/tcg.c | 2 +- 3 files changed, 55 insertions(+), 9 deletions(-) diff --git a/tcg/tcg-internal.h b/tcg/tcg-internal.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/tcg-internal.h +++ b/tcg/tcg-internal.h @@ -XXX,XX +XXX,XX @@ static inline TCGv_i64 TCGV128_HIGH(TCGv_i128 t) bool tcg_target_has_memory_bswap(MemOp memop); +TCGTemp *tcg_temp_new_internal(TCGType type, TCGTempKind kind); + /* * Locate or create a read-only temporary that is a constant. * This kind of temporary need not be freed, but for convenience diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static TCGArg arg_new_constant(OptContext *ctx, uint64_t val) return temp_arg(ts); } +static TCGArg arg_new_temp(OptContext *ctx) +{ + TCGTemp *ts = tcg_temp_new_internal(ctx->type, TEMP_EBB); + init_ts_info(ctx, ts); + return temp_arg(ts); +} + static bool tcg_opt_gen_mov(OptContext *ctx, TCGOp *op, TCGArg dst, TCGArg src) { TCGTemp *dst_ts = arg_temp(dst); @@ -XXX,XX +XXX,XX @@ static bool swap_commutative2(TCGArg *p1, TCGArg *p2) * Return -1 if the condition can't be simplified, * and the result of the condition (0 or 1) if it can. */ -static int do_constant_folding_cond1(OptContext *ctx, TCGArg dest, +static int do_constant_folding_cond1(OptContext *ctx, TCGOp *op, TCGArg dest, TCGArg *p1, TCGArg *p2, TCGArg *pcond) { TCGCond cond; @@ -XXX,XX +XXX,XX @@ static int do_constant_folding_cond1(OptContext *ctx, TCGArg dest, ? INT32_MIN : INT64_MIN))) { *p2 = arg_new_constant(ctx, 0); *pcond = tcg_tst_ltge_cond(cond); + return -1; + } + + /* Expand to AND with a temporary if no backend support. */ + if (!TCG_TARGET_HAS_tst) { + TCGOpcode and_opc = (ctx->type == TCG_TYPE_I32 + ? INDEX_op_and_i32 : INDEX_op_and_i64); + TCGOp *op2 = tcg_op_insert_before(ctx->tcg, op, and_opc, 3); + TCGArg tmp = arg_new_temp(ctx); + + op2->args[0] = tmp; + op2->args[1] = *p1; + op2->args[2] = *p2; + + *p1 = tmp; + *p2 = arg_new_constant(ctx, 0); + *pcond = tcg_tst_eqne_cond(cond); } return -1; } -static int do_constant_folding_cond2(OptContext *ctx, TCGArg *args) +static int do_constant_folding_cond2(OptContext *ctx, TCGOp *op, TCGArg *args) { TCGArg al, ah, bl, bh; TCGCond c; @@ -XXX,XX +XXX,XX @@ static int do_constant_folding_cond2(OptContext *ctx, TCGArg *args) return -1; } } + + /* Expand to AND with a temporary if no backend support. */ + if (!TCG_TARGET_HAS_tst && is_tst_cond(c)) { + TCGOp *op1 = tcg_op_insert_before(ctx->tcg, op, INDEX_op_and_i32, 3); + TCGOp *op2 = tcg_op_insert_before(ctx->tcg, op, INDEX_op_and_i32, 3); + TCGArg t1 = arg_new_temp(ctx); + TCGArg t2 = arg_new_temp(ctx); + + op1->args[0] = t1; + op1->args[1] = al; + op1->args[2] = bl; + op2->args[0] = t2; + op2->args[1] = ah; + op2->args[2] = bh; + + args[0] = t1; + args[1] = t2; + args[3] = args[2] = arg_new_constant(ctx, 0); + args[4] = tcg_tst_eqne_cond(c); + } return -1; } @@ -XXX,XX +XXX,XX @@ static bool fold_andc(OptContext *ctx, TCGOp *op) static bool fold_brcond(OptContext *ctx, TCGOp *op) { - int i = do_constant_folding_cond1(ctx, NO_DEST, &op->args[0], + int i = do_constant_folding_cond1(ctx, op, NO_DEST, &op->args[0], &op->args[1], &op->args[2]); if (i == 0) { tcg_op_remove(ctx->tcg, op); @@ -XXX,XX +XXX,XX @@ static bool fold_brcond2(OptContext *ctx, TCGOp *op) TCGArg label; int i, inv = 0; - i = do_constant_folding_cond2(ctx, &op->args[0]); + i = do_constant_folding_cond2(ctx, op, &op->args[0]); cond = op->args[4]; label = op->args[5]; if (i >= 0) { @@ -XXX,XX +XXX,XX @@ static bool fold_movcond(OptContext *ctx, TCGOp *op) op->args[5] = tcg_invert_cond(op->args[5]); } - i = do_constant_folding_cond1(ctx, NO_DEST, &op->args[1], + i = do_constant_folding_cond1(ctx, op, NO_DEST, &op->args[1], &op->args[2], &op->args[5]); if (i >= 0) { return tcg_opt_gen_mov(ctx, op, op->args[0], op->args[4 - i]); @@ -XXX,XX +XXX,XX @@ static void fold_setcond_tst_pow2(OptContext *ctx, TCGOp *op, bool neg) static bool fold_setcond(OptContext *ctx, TCGOp *op) { - int i = do_constant_folding_cond1(ctx, op->args[0], &op->args[1], + int i = do_constant_folding_cond1(ctx, op, op->args[0], &op->args[1], &op->args[2], &op->args[3]); if (i >= 0) { return tcg_opt_gen_movi(ctx, op, op->args[0], i); @@ -XXX,XX +XXX,XX @@ static bool fold_setcond(OptContext *ctx, TCGOp *op) static bool fold_negsetcond(OptContext *ctx, TCGOp *op) { - int i = do_constant_folding_cond1(ctx, op->args[0], &op->args[1], + int i = do_constant_folding_cond1(ctx, op, op->args[0], &op->args[1], &op->args[2], &op->args[3]); if (i >= 0) { return tcg_opt_gen_movi(ctx, op, op->args[0], -i); @@ -XXX,XX +XXX,XX @@ static bool fold_setcond2(OptContext *ctx, TCGOp *op) TCGCond cond; int i, inv = 0; - i = do_constant_folding_cond2(ctx, &op->args[1]); + i = do_constant_folding_cond2(ctx, op, &op->args[1]); cond = op->args[5]; if (i >= 0) { goto do_setcond_const; diff --git a/tcg/tcg.c b/tcg/tcg.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/tcg.c +++ b/tcg/tcg.c @@ -XXX,XX +XXX,XX @@ TCGv_ptr tcg_global_mem_new_ptr(TCGv_ptr reg, intptr_t off, const char *name) return temp_tcgv_ptr(ts); } -static TCGTemp *tcg_temp_new_internal(TCGType type, TCGTempKind kind) +TCGTemp *tcg_temp_new_internal(TCGType type, TCGTempKind kind) { TCGContext *s = tcg_ctx; TCGTemp *ts; -- 2.34.1 Simplify gen_bcond() by passing an immediate value. Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-Id: <20231028194522.245170-33-richard.henderson@linaro.org> [PMD: Split from bigger patch, part 1/2] Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org> Message-Id: <20231108205247.83234-1-philmd@linaro.org> --- target/alpha/translate.c | 21 +++++++-------------- 1 file changed, 7 insertions(+), 14 deletions(-) diff --git a/target/alpha/translate.c b/target/alpha/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/alpha/translate.c +++ b/target/alpha/translate.c @@ -XXX,XX +XXX,XX @@ static DisasJumpType gen_bdirect(DisasContext *ctx, int ra, int32_t disp) } static DisasJumpType gen_bcond_internal(DisasContext *ctx, TCGCond cond, - TCGv cmp, int32_t disp) + TCGv cmp, uint64_t imm, int32_t disp) { uint64_t dest = ctx->base.pc_next + (disp << 2); TCGLabel *lab_true = gen_new_label(); if (use_goto_tb(ctx, dest)) { - tcg_gen_brcondi_i64(cond, cmp, 0, lab_true); + tcg_gen_brcondi_i64(cond, cmp, imm, lab_true); tcg_gen_goto_tb(0); tcg_gen_movi_i64(cpu_pc, ctx->base.pc_next); @@ -XXX,XX +XXX,XX @@ static DisasJumpType gen_bcond_internal(DisasContext *ctx, TCGCond cond, return DISAS_NORETURN; } else { - TCGv_i64 z = load_zero(ctx); + TCGv_i64 i = tcg_constant_i64(imm); TCGv_i64 d = tcg_constant_i64(dest); TCGv_i64 p = tcg_constant_i64(ctx->base.pc_next); - tcg_gen_movcond_i64(cond, cpu_pc, cmp, z, d, p); + tcg_gen_movcond_i64(cond, cpu_pc, cmp, i, d, p); return DISAS_PC_UPDATED; } } @@ -XXX,XX +XXX,XX @@ static DisasJumpType gen_bcond_internal(DisasContext *ctx, TCGCond cond, static DisasJumpType gen_bcond(DisasContext *ctx, TCGCond cond, int ra, int32_t disp, int mask) { - if (mask) { - TCGv tmp = tcg_temp_new(); - DisasJumpType ret; - - tcg_gen_andi_i64(tmp, load_gpr(ctx, ra), 1); - ret = gen_bcond_internal(ctx, cond, tmp, disp); - return ret; - } - return gen_bcond_internal(ctx, cond, load_gpr(ctx, ra), disp); + return gen_bcond_internal(ctx, cond, load_gpr(ctx, ra), + mask, disp); } /* Fold -0.0 for comparison with COND. */ @@ -XXX,XX +XXX,XX @@ static DisasJumpType gen_fbcond(DisasContext *ctx, TCGCond cond, int ra, DisasJumpType ret; gen_fold_mzero(cond, cmp_tmp, load_fpr(ctx, ra)); - ret = gen_bcond_internal(ctx, cond, cmp_tmp, disp); + ret = gen_bcond_internal(ctx, cond, cmp_tmp, 0, disp); return ret; } -- 2.34.1 Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-Id: <20231028194522.245170-33-richard.henderson@linaro.org> [PMD: Split from bigger patch, part 2/2] Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org> Message-Id: <20231108205247.83234-2-philmd@linaro.org> --- target/alpha/translate.c | 20 ++++++++++---------- 1 file changed, 10 insertions(+), 10 deletions(-) diff --git a/target/alpha/translate.c b/target/alpha/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/alpha/translate.c +++ b/target/alpha/translate.c @@ -XXX,XX +XXX,XX @@ static DisasJumpType gen_bcond_internal(DisasContext *ctx, TCGCond cond, } static DisasJumpType gen_bcond(DisasContext *ctx, TCGCond cond, int ra, - int32_t disp, int mask) + int32_t disp) { return gen_bcond_internal(ctx, cond, load_gpr(ctx, ra), - mask, disp); + is_tst_cond(cond), disp); } /* Fold -0.0 for comparison with COND. */ @@ -XXX,XX +XXX,XX @@ static DisasJumpType translate_one(DisasContext *ctx, uint32_t insn) break; case 0x38: /* BLBC */ - ret = gen_bcond(ctx, TCG_COND_EQ, ra, disp21, 1); + ret = gen_bcond(ctx, TCG_COND_TSTEQ, ra, disp21); break; case 0x39: /* BEQ */ - ret = gen_bcond(ctx, TCG_COND_EQ, ra, disp21, 0); + ret = gen_bcond(ctx, TCG_COND_EQ, ra, disp21); break; case 0x3A: /* BLT */ - ret = gen_bcond(ctx, TCG_COND_LT, ra, disp21, 0); + ret = gen_bcond(ctx, TCG_COND_LT, ra, disp21); break; case 0x3B: /* BLE */ - ret = gen_bcond(ctx, TCG_COND_LE, ra, disp21, 0); + ret = gen_bcond(ctx, TCG_COND_LE, ra, disp21); break; case 0x3C: /* BLBS */ - ret = gen_bcond(ctx, TCG_COND_NE, ra, disp21, 1); + ret = gen_bcond(ctx, TCG_COND_TSTNE, ra, disp21); break; case 0x3D: /* BNE */ - ret = gen_bcond(ctx, TCG_COND_NE, ra, disp21, 0); + ret = gen_bcond(ctx, TCG_COND_NE, ra, disp21); break; case 0x3E: /* BGE */ - ret = gen_bcond(ctx, TCG_COND_GE, ra, disp21, 0); + ret = gen_bcond(ctx, TCG_COND_GE, ra, disp21); break; case 0x3F: /* BGT */ - ret = gen_bcond(ctx, TCG_COND_GT, ra, disp21, 0); + ret = gen_bcond(ctx, TCG_COND_GT, ra, disp21); break; invalid_opc: ret = gen_invalid(ctx); -- 2.34.1 Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- target/alpha/translate.c | 8 ++------ 1 file changed, 2 insertions(+), 6 deletions(-) diff --git a/target/alpha/translate.c b/target/alpha/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/alpha/translate.c +++ b/target/alpha/translate.c @@ -XXX,XX +XXX,XX @@ static DisasJumpType translate_one(DisasContext *ctx, uint32_t insn) break; case 0x14: /* CMOVLBS */ - tmp = tcg_temp_new(); - tcg_gen_andi_i64(tmp, va, 1); - tcg_gen_movcond_i64(TCG_COND_NE, vc, tmp, load_zero(ctx), + tcg_gen_movcond_i64(TCG_COND_TSTNE, vc, va, tcg_constant_i64(1), vb, load_gpr(ctx, rc)); break; case 0x16: /* CMOVLBC */ - tmp = tcg_temp_new(); - tcg_gen_andi_i64(tmp, va, 1); - tcg_gen_movcond_i64(TCG_COND_EQ, vc, tmp, load_zero(ctx), + tcg_gen_movcond_i64(TCG_COND_TSTEQ, vc, va, tcg_constant_i64(1), vb, load_gpr(ctx, rc)); break; case 0x20: -- 2.34.1 Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- target/alpha/translate.c | 49 +++++++++++++++++++--------------------- 1 file changed, 23 insertions(+), 26 deletions(-) diff --git a/target/alpha/translate.c b/target/alpha/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/alpha/translate.c +++ b/target/alpha/translate.c @@ -XXX,XX +XXX,XX @@ static DisasJumpType gen_bcond(DisasContext *ctx, TCGCond cond, int ra, /* Fold -0.0 for comparison with COND. */ -static void gen_fold_mzero(TCGCond cond, TCGv dest, TCGv src) +static TCGv_i64 gen_fold_mzero(TCGCond *pcond, uint64_t *pimm, TCGv_i64 src) { - uint64_t mzero = 1ull << 63; + TCGv_i64 tmp; - switch (cond) { + *pimm = 0; + switch (*pcond) { case TCG_COND_LE: case TCG_COND_GT: /* For <= or >, the -0.0 value directly compares the way we want. */ - tcg_gen_mov_i64(dest, src); - break; + return src; case TCG_COND_EQ: case TCG_COND_NE: - /* For == or !=, we can simply mask off the sign bit and compare. */ - tcg_gen_andi_i64(dest, src, mzero - 1); - break; + /* For == or !=, we can compare without the sign bit. */ + *pcond = *pcond == TCG_COND_EQ ? TCG_COND_TSTEQ : TCG_COND_TSTNE; + *pimm = INT64_MAX; + return src; case TCG_COND_GE: case TCG_COND_LT: /* For >= or <, map -0.0 to +0.0. */ - tcg_gen_movcond_i64(TCG_COND_NE, dest, src, tcg_constant_i64(mzero), - src, tcg_constant_i64(0)); - break; + tmp = tcg_temp_new_i64(); + tcg_gen_movcond_i64(TCG_COND_EQ, tmp, + src, tcg_constant_i64(INT64_MIN), + tcg_constant_i64(0), src); + return tmp; default: - abort(); + g_assert_not_reached(); } } static DisasJumpType gen_fbcond(DisasContext *ctx, TCGCond cond, int ra, int32_t disp) { - TCGv cmp_tmp = tcg_temp_new(); - DisasJumpType ret; - - gen_fold_mzero(cond, cmp_tmp, load_fpr(ctx, ra)); - ret = gen_bcond_internal(ctx, cond, cmp_tmp, 0, disp); - return ret; + uint64_t imm; + TCGv_i64 tmp = gen_fold_mzero(&cond, &imm, load_fpr(ctx, ra)); + return gen_bcond_internal(ctx, cond, tmp, imm, disp); } static void gen_fcmov(DisasContext *ctx, TCGCond cond, int ra, int rb, int rc) { - TCGv_i64 va, vb, z; - - z = load_zero(ctx); - vb = load_fpr(ctx, rb); - va = tcg_temp_new(); - gen_fold_mzero(cond, va, load_fpr(ctx, ra)); - - tcg_gen_movcond_i64(cond, dest_fpr(ctx, rc), va, z, vb, load_fpr(ctx, rc)); + uint64_t imm; + TCGv_i64 tmp = gen_fold_mzero(&cond, &imm, load_fpr(ctx, ra)); + tcg_gen_movcond_i64(cond, dest_fpr(ctx, rc), + tmp, tcg_constant_i64(imm), + load_fpr(ctx, rb), load_fpr(ctx, rc)); } #define QUAL_RM_N 0x080 /* Round mode nearest even */ -- 2.34.1 Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- target/m68k/translate.c | 74 ++++++++++++++++++----------------------- 1 file changed, 33 insertions(+), 41 deletions(-) diff --git a/target/m68k/translate.c b/target/m68k/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/m68k/translate.c +++ b/target/m68k/translate.c @@ -XXX,XX +XXX,XX @@ undef: static void gen_fcc_cond(DisasCompare *c, DisasContext *s, int cond) { TCGv fpsr; + int imm = 0; - c->v2 = tcg_constant_i32(0); /* TODO: Raise BSUN exception. */ fpsr = tcg_temp_new(); gen_load_fcr(s, fpsr, M68K_FPSR); + c->v1 = fpsr; + switch (cond) { case 0: /* False */ case 16: /* Signaling False */ - c->v1 = c->v2; c->tcond = TCG_COND_NEVER; break; case 1: /* EQual Z */ case 17: /* Signaling EQual Z */ - c->v1 = tcg_temp_new(); - tcg_gen_andi_i32(c->v1, fpsr, FPSR_CC_Z); - c->tcond = TCG_COND_NE; + imm = FPSR_CC_Z; + c->tcond = TCG_COND_TSTNE; break; case 2: /* Ordered Greater Than !(A || Z || N) */ case 18: /* Greater Than !(A || Z || N) */ - c->v1 = tcg_temp_new(); - tcg_gen_andi_i32(c->v1, fpsr, - FPSR_CC_A | FPSR_CC_Z | FPSR_CC_N); - c->tcond = TCG_COND_EQ; + imm = FPSR_CC_A | FPSR_CC_Z | FPSR_CC_N; + c->tcond = TCG_COND_TSTEQ; break; case 3: /* Ordered Greater than or Equal Z || !(A || N) */ case 19: /* Greater than or Equal Z || !(A || N) */ c->v1 = tcg_temp_new(); tcg_gen_andi_i32(c->v1, fpsr, FPSR_CC_A); tcg_gen_shli_i32(c->v1, c->v1, ctz32(FPSR_CC_N) - ctz32(FPSR_CC_A)); - tcg_gen_andi_i32(fpsr, fpsr, FPSR_CC_Z | FPSR_CC_N); tcg_gen_or_i32(c->v1, c->v1, fpsr); tcg_gen_xori_i32(c->v1, c->v1, FPSR_CC_N); - c->tcond = TCG_COND_NE; + imm = FPSR_CC_Z | FPSR_CC_N; + c->tcond = TCG_COND_TSTNE; break; case 4: /* Ordered Less Than !(!N || A || Z); */ case 20: /* Less Than !(!N || A || Z); */ c->v1 = tcg_temp_new(); tcg_gen_xori_i32(c->v1, fpsr, FPSR_CC_N); - tcg_gen_andi_i32(c->v1, c->v1, FPSR_CC_N | FPSR_CC_A | FPSR_CC_Z); - c->tcond = TCG_COND_EQ; + imm = FPSR_CC_N | FPSR_CC_A | FPSR_CC_Z; + c->tcond = TCG_COND_TSTEQ; break; case 5: /* Ordered Less than or Equal Z || (N && !A) */ case 21: /* Less than or Equal Z || (N && !A) */ @@ -XXX,XX +XXX,XX @@ static void gen_fcc_cond(DisasCompare *c, DisasContext *s, int cond) tcg_gen_andi_i32(c->v1, fpsr, FPSR_CC_A); tcg_gen_shli_i32(c->v1, c->v1, ctz32(FPSR_CC_N) - ctz32(FPSR_CC_A)); tcg_gen_andc_i32(c->v1, fpsr, c->v1); - tcg_gen_andi_i32(c->v1, c->v1, FPSR_CC_Z | FPSR_CC_N); - c->tcond = TCG_COND_NE; + imm = FPSR_CC_Z | FPSR_CC_N; + c->tcond = TCG_COND_TSTNE; break; case 6: /* Ordered Greater or Less than !(A || Z) */ case 22: /* Greater or Less than !(A || Z) */ - c->v1 = tcg_temp_new(); - tcg_gen_andi_i32(c->v1, fpsr, FPSR_CC_A | FPSR_CC_Z); - c->tcond = TCG_COND_EQ; + imm = FPSR_CC_A | FPSR_CC_Z; + c->tcond = TCG_COND_TSTEQ; break; case 7: /* Ordered !A */ case 23: /* Greater, Less or Equal !A */ - c->v1 = tcg_temp_new(); - tcg_gen_andi_i32(c->v1, fpsr, FPSR_CC_A); - c->tcond = TCG_COND_EQ; + imm = FPSR_CC_A; + c->tcond = TCG_COND_TSTEQ; break; case 8: /* Unordered A */ case 24: /* Not Greater, Less or Equal A */ - c->v1 = tcg_temp_new(); - tcg_gen_andi_i32(c->v1, fpsr, FPSR_CC_A); - c->tcond = TCG_COND_NE; + imm = FPSR_CC_A; + c->tcond = TCG_COND_TSTNE; break; case 9: /* Unordered or Equal A || Z */ case 25: /* Not Greater or Less then A || Z */ - c->v1 = tcg_temp_new(); - tcg_gen_andi_i32(c->v1, fpsr, FPSR_CC_A | FPSR_CC_Z); - c->tcond = TCG_COND_NE; + imm = FPSR_CC_A | FPSR_CC_Z; + c->tcond = TCG_COND_TSTNE; break; case 10: /* Unordered or Greater Than A || !(N || Z)) */ case 26: /* Not Less or Equal A || !(N || Z)) */ c->v1 = tcg_temp_new(); tcg_gen_andi_i32(c->v1, fpsr, FPSR_CC_Z); tcg_gen_shli_i32(c->v1, c->v1, ctz32(FPSR_CC_N) - ctz32(FPSR_CC_Z)); - tcg_gen_andi_i32(fpsr, fpsr, FPSR_CC_A | FPSR_CC_N); tcg_gen_or_i32(c->v1, c->v1, fpsr); tcg_gen_xori_i32(c->v1, c->v1, FPSR_CC_N); - c->tcond = TCG_COND_NE; + imm = FPSR_CC_A | FPSR_CC_N; + c->tcond = TCG_COND_TSTNE; break; case 11: /* Unordered or Greater or Equal A || Z || !N */ case 27: /* Not Less Than A || Z || !N */ c->v1 = tcg_temp_new(); - tcg_gen_andi_i32(c->v1, fpsr, FPSR_CC_A | FPSR_CC_Z | FPSR_CC_N); - tcg_gen_xori_i32(c->v1, c->v1, FPSR_CC_N); - c->tcond = TCG_COND_NE; + tcg_gen_xori_i32(c->v1, fpsr, FPSR_CC_N); + imm = FPSR_CC_A | FPSR_CC_Z | FPSR_CC_N; + c->tcond = TCG_COND_TSTNE; break; case 12: /* Unordered or Less Than A || (N && !Z) */ case 28: /* Not Greater than or Equal A || (N && !Z) */ @@ -XXX,XX +XXX,XX @@ static void gen_fcc_cond(DisasCompare *c, DisasContext *s, int cond) tcg_gen_andi_i32(c->v1, fpsr, FPSR_CC_Z); tcg_gen_shli_i32(c->v1, c->v1, ctz32(FPSR_CC_N) - ctz32(FPSR_CC_Z)); tcg_gen_andc_i32(c->v1, fpsr, c->v1); - tcg_gen_andi_i32(c->v1, c->v1, FPSR_CC_A | FPSR_CC_N); - c->tcond = TCG_COND_NE; + imm = FPSR_CC_A | FPSR_CC_N; + c->tcond = TCG_COND_TSTNE; break; case 13: /* Unordered or Less or Equal A || Z || N */ case 29: /* Not Greater Than A || Z || N */ - c->v1 = tcg_temp_new(); - tcg_gen_andi_i32(c->v1, fpsr, FPSR_CC_A | FPSR_CC_Z | FPSR_CC_N); - c->tcond = TCG_COND_NE; + imm = FPSR_CC_A | FPSR_CC_Z | FPSR_CC_N; + c->tcond = TCG_COND_TSTNE; break; case 14: /* Not Equal !Z */ case 30: /* Signaling Not Equal !Z */ - c->v1 = tcg_temp_new(); - tcg_gen_andi_i32(c->v1, fpsr, FPSR_CC_Z); - c->tcond = TCG_COND_EQ; + imm = FPSR_CC_Z; + c->tcond = TCG_COND_TSTEQ; break; case 15: /* True */ case 31: /* Signaling True */ - c->v1 = c->v2; c->tcond = TCG_COND_ALWAYS; break; } + c->v2 = tcg_constant_i32(imm); } static void gen_fjmpcc(DisasContext *s, int cond, TCGLabel *l1) -- 2.34.1 Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- target/sparc/translate.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/target/sparc/translate.c b/target/sparc/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/sparc/translate.c +++ b/target/sparc/translate.c @@ -XXX,XX +XXX,XX @@ static void gen_op_subccc(TCGv dst, TCGv src1, TCGv src2) static void gen_op_mulscc(TCGv dst, TCGv src1, TCGv src2) { TCGv zero = tcg_constant_tl(0); + TCGv one = tcg_constant_tl(1); TCGv t_src1 = tcg_temp_new(); TCGv t_src2 = tcg_temp_new(); TCGv t0 = tcg_temp_new(); @@ -XXX,XX +XXX,XX @@ static void gen_op_mulscc(TCGv dst, TCGv src1, TCGv src2) * if (!(env->y & 1)) * src2 = 0; */ - tcg_gen_andi_tl(t0, cpu_y, 0x1); - tcg_gen_movcond_tl(TCG_COND_EQ, t_src2, t0, zero, zero, t_src2); + tcg_gen_movcond_tl(TCG_COND_TSTEQ, t_src2, cpu_y, one, zero, t_src2); /* * b2 = src1 & 1; -- 2.34.1 These are all test-and-compare type instructions. Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- target/s390x/tcg/translate.c | 18 +++++++----------- 1 file changed, 7 insertions(+), 11 deletions(-) diff --git a/target/s390x/tcg/translate.c b/target/s390x/tcg/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/s390x/tcg/translate.c +++ b/target/s390x/tcg/translate.c @@ -XXX,XX +XXX,XX @@ static void disas_jcc(DisasContext *s, DisasCompare *c, uint32_t mask) case CC_OP_TM_64: switch (mask) { case 8: - cond = TCG_COND_EQ; + cond = TCG_COND_TSTEQ; break; case 4 | 2 | 1: - cond = TCG_COND_NE; + cond = TCG_COND_TSTNE; break; default: goto do_dynamic; @@ -XXX,XX +XXX,XX @@ static void disas_jcc(DisasContext *s, DisasCompare *c, uint32_t mask) case CC_OP_ICM: switch (mask) { case 8: - cond = TCG_COND_EQ; + cond = TCG_COND_TSTEQ; break; case 4 | 2 | 1: case 4 | 2: - cond = TCG_COND_NE; + cond = TCG_COND_TSTNE; break; default: goto do_dynamic; @@ -XXX,XX +XXX,XX @@ static void disas_jcc(DisasContext *s, DisasCompare *c, uint32_t mask) c->u.s64.a = cc_dst; c->u.s64.b = tcg_constant_i64(0); break; + case CC_OP_LTGT_64: case CC_OP_LTUGTU_64: - c->u.s64.a = cc_src; - c->u.s64.b = cc_dst; - break; - case CC_OP_TM_32: case CC_OP_TM_64: case CC_OP_ICM: - c->u.s64.a = tcg_temp_new_i64(); - c->u.s64.b = tcg_constant_i64(0); - tcg_gen_and_i64(c->u.s64.a, cc_src, cc_dst); + c->u.s64.a = cc_src; + c->u.s64.b = cc_dst; break; case CC_OP_ADDU: -- 2.34.1 Avoid code duplication by handling 7 of the 14 cases by inverting the test for the other 7 cases. Use TCG_COND_TSTNE for cc in {1,3}. Use (cc - 1) <= 1 for cc in {1,2}. Acked-by: Ilya Leoshkevich <iii@linux.ibm.com> Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- target/s390x/tcg/translate.c | 82 +++++++++++++----------------------- 1 file changed, 30 insertions(+), 52 deletions(-) diff --git a/target/s390x/tcg/translate.c b/target/s390x/tcg/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/s390x/tcg/translate.c +++ b/target/s390x/tcg/translate.c @@ -XXX,XX +XXX,XX @@ static void disas_jcc(DisasContext *s, DisasCompare *c, uint32_t mask) case CC_OP_STATIC: c->is_64 = false; c->u.s32.a = cc_op; - switch (mask) { - case 0x8 | 0x4 | 0x2: /* cc != 3 */ - cond = TCG_COND_NE; + + /* Fold half of the cases using bit 3 to invert. */ + switch (mask & 8 ? mask ^ 0xf : mask) { + case 0x1: /* cc == 3 */ + cond = TCG_COND_EQ; c->u.s32.b = tcg_constant_i32(3); break; - case 0x8 | 0x4 | 0x1: /* cc != 2 */ - cond = TCG_COND_NE; - c->u.s32.b = tcg_constant_i32(2); - break; - case 0x8 | 0x2 | 0x1: /* cc != 1 */ - cond = TCG_COND_NE; - c->u.s32.b = tcg_constant_i32(1); - break; - case 0x8 | 0x2: /* cc == 0 || cc == 2 => (cc & 1) == 0 */ - cond = TCG_COND_EQ; - c->u.s32.a = tcg_temp_new_i32(); - c->u.s32.b = tcg_constant_i32(0); - tcg_gen_andi_i32(c->u.s32.a, cc_op, 1); - break; - case 0x8 | 0x4: /* cc < 2 */ - cond = TCG_COND_LTU; - c->u.s32.b = tcg_constant_i32(2); - break; - case 0x8: /* cc == 0 */ - cond = TCG_COND_EQ; - c->u.s32.b = tcg_constant_i32(0); - break; - case 0x4 | 0x2 | 0x1: /* cc != 0 */ - cond = TCG_COND_NE; - c->u.s32.b = tcg_constant_i32(0); - break; - case 0x4 | 0x1: /* cc == 1 || cc == 3 => (cc & 1) != 0 */ - cond = TCG_COND_NE; - c->u.s32.a = tcg_temp_new_i32(); - c->u.s32.b = tcg_constant_i32(0); - tcg_gen_andi_i32(c->u.s32.a, cc_op, 1); - break; - case 0x4: /* cc == 1 */ - cond = TCG_COND_EQ; - c->u.s32.b = tcg_constant_i32(1); - break; - case 0x2 | 0x1: /* cc > 1 */ - cond = TCG_COND_GTU; - c->u.s32.b = tcg_constant_i32(1); - break; case 0x2: /* cc == 2 */ cond = TCG_COND_EQ; c->u.s32.b = tcg_constant_i32(2); break; - case 0x1: /* cc == 3 */ + case 0x4: /* cc == 1 */ cond = TCG_COND_EQ; - c->u.s32.b = tcg_constant_i32(3); + c->u.s32.b = tcg_constant_i32(1); + break; + case 0x2 | 0x1: /* cc == 2 || cc == 3 => cc > 1 */ + cond = TCG_COND_GTU; + c->u.s32.b = tcg_constant_i32(1); + break; + case 0x4 | 0x1: /* cc == 1 || cc == 3 => (cc & 1) != 0 */ + cond = TCG_COND_TSTNE; + c->u.s32.b = tcg_constant_i32(1); + break; + case 0x4 | 0x2: /* cc == 1 || cc == 2 => (cc - 1) <= 1 */ + cond = TCG_COND_LEU; + c->u.s32.a = tcg_temp_new_i32(); + c->u.s32.b = tcg_constant_i32(1); + tcg_gen_addi_i32(c->u.s32.a, cc_op, -1); + break; + case 0x4 | 0x2 | 0x1: /* cc != 0 */ + cond = TCG_COND_NE; + c->u.s32.b = tcg_constant_i32(0); break; default: - /* CC is masked by something else: (8 >> cc) & mask. */ - cond = TCG_COND_NE; - c->u.s32.a = tcg_temp_new_i32(); - c->u.s32.b = tcg_constant_i32(0); - tcg_gen_shr_i32(c->u.s32.a, tcg_constant_i32(8), cc_op); - tcg_gen_andi_i32(c->u.s32.a, c->u.s32.a, mask); - break; + /* case 0: never, handled above. */ + g_assert_not_reached(); + } + if (mask & 8) { + cond = tcg_invert_cond(cond); } break; -- 2.34.1 Fill the new argument from any condition within the opcode. Not yet used within any backend. Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/tcg.c | 34 ++++++++++++++++++++++++++++++-- tcg/aarch64/tcg-target.c.inc | 3 ++- tcg/arm/tcg-target.c.inc | 3 ++- tcg/i386/tcg-target.c.inc | 3 ++- tcg/loongarch64/tcg-target.c.inc | 3 ++- tcg/mips/tcg-target.c.inc | 3 ++- tcg/ppc/tcg-target.c.inc | 3 ++- tcg/riscv/tcg-target.c.inc | 3 ++- tcg/s390x/tcg-target.c.inc | 3 ++- tcg/sparc64/tcg-target.c.inc | 3 ++- tcg/tci/tcg-target.c.inc | 3 ++- 11 files changed, 52 insertions(+), 12 deletions(-) diff --git a/tcg/tcg.c b/tcg/tcg.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/tcg.c +++ b/tcg/tcg.c @@ -XXX,XX +XXX,XX @@ static bool tcg_out_sti(TCGContext *s, TCGType type, TCGArg val, static void tcg_out_call(TCGContext *s, const tcg_insn_unit *target, const TCGHelperInfo *info); static TCGReg tcg_target_call_oarg_reg(TCGCallReturnKind kind, int slot); -static bool tcg_target_const_match(int64_t val, TCGType type, int ct, int vece); +static bool tcg_target_const_match(int64_t val, int ct, + TCGType type, TCGCond cond, int vece); #ifdef TCG_TARGET_NEED_LDST_LABELS static int tcg_out_ldst_finalize(TCGContext *s); #endif @@ -XXX,XX +XXX,XX @@ static void tcg_reg_alloc_op(TCGContext *s, const TCGOp *op) TCGTemp *ts; TCGArg new_args[TCG_MAX_OP_ARGS]; int const_args[TCG_MAX_OP_ARGS]; + TCGCond op_cond; nb_oargs = def->nb_oargs; nb_iargs = def->nb_iargs; @@ -XXX,XX +XXX,XX @@ static void tcg_reg_alloc_op(TCGContext *s, const TCGOp *op) i_allocated_regs = s->reserved_regs; o_allocated_regs = s->reserved_regs; + switch (op->opc) { + case INDEX_op_brcond_i32: + case INDEX_op_brcond_i64: + op_cond = op->args[2]; + break; + case INDEX_op_setcond_i32: + case INDEX_op_setcond_i64: + case INDEX_op_negsetcond_i32: + case INDEX_op_negsetcond_i64: + case INDEX_op_cmp_vec: + op_cond = op->args[3]; + break; + case INDEX_op_brcond2_i32: + op_cond = op->args[4]; + break; + case INDEX_op_movcond_i32: + case INDEX_op_movcond_i64: + case INDEX_op_setcond2_i32: + case INDEX_op_cmpsel_vec: + op_cond = op->args[5]; + break; + default: + /* No condition within opcode. */ + op_cond = TCG_COND_ALWAYS; + break; + } + /* satisfy input constraints */ for (k = 0; k < nb_iargs; k++) { TCGRegSet i_preferred_regs, i_required_regs; @@ -XXX,XX +XXX,XX @@ static void tcg_reg_alloc_op(TCGContext *s, const TCGOp *op) ts = arg_temp(arg); if (ts->val_type == TEMP_VAL_CONST - && tcg_target_const_match(ts->val, ts->type, arg_ct->ct, TCGOP_VECE(op))) { + && tcg_target_const_match(ts->val, arg_ct->ct, ts->type, + op_cond, TCGOP_VECE(op))) { /* constant is OK for instruction */ const_args[i] = 1; new_args[i] = ts->val; diff --git a/tcg/aarch64/tcg-target.c.inc b/tcg/aarch64/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/aarch64/tcg-target.c.inc +++ b/tcg/aarch64/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ static bool is_shimm1632(uint32_t v32, int *cmode, int *imm8) } } -static bool tcg_target_const_match(int64_t val, TCGType type, int ct, int vece) +static bool tcg_target_const_match(int64_t val, int ct, + TCGType type, TCGCond cond, int vece) { if (ct & TCG_CT_CONST) { return 1; diff --git a/tcg/arm/tcg-target.c.inc b/tcg/arm/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/arm/tcg-target.c.inc +++ b/tcg/arm/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ static bool is_shimm1632(uint32_t v32, int *cmode, int *imm8) * mov operand2: values represented with x << (2 * y), x < 0x100 * add, sub, eor...: ditto */ -static bool tcg_target_const_match(int64_t val, TCGType type, int ct, int vece) +static bool tcg_target_const_match(int64_t val, int ct, + TCGType type, TCGCond cond, int vece) { if (ct & TCG_CT_CONST) { return 1; diff --git a/tcg/i386/tcg-target.c.inc b/tcg/i386/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/i386/tcg-target.c.inc +++ b/tcg/i386/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ static bool patch_reloc(tcg_insn_unit *code_ptr, int type, } /* test if a constant matches the constraint */ -static bool tcg_target_const_match(int64_t val, TCGType type, int ct, int vece) +static bool tcg_target_const_match(int64_t val, int ct, + TCGType type, TCGCond cond, int vece) { if (ct & TCG_CT_CONST) { return 1; diff --git a/tcg/loongarch64/tcg-target.c.inc b/tcg/loongarch64/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/loongarch64/tcg-target.c.inc +++ b/tcg/loongarch64/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ static inline tcg_target_long sextreg(tcg_target_long val, int pos, int len) } /* test if a constant matches the constraint */ -static bool tcg_target_const_match(int64_t val, TCGType type, int ct, int vece) +static bool tcg_target_const_match(int64_t val, int ct, + TCGType type, TCGCond cond, int vece) { if (ct & TCG_CT_CONST) { return true; diff --git a/tcg/mips/tcg-target.c.inc b/tcg/mips/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/mips/tcg-target.c.inc +++ b/tcg/mips/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ static bool is_p2m1(tcg_target_long val) } /* test if a constant matches the constraint */ -static bool tcg_target_const_match(int64_t val, TCGType type, int ct, int vece) +static bool tcg_target_const_match(int64_t val, int ct, + TCGType type, TCGCond cond, int vece) { if (ct & TCG_CT_CONST) { return 1; diff --git a/tcg/ppc/tcg-target.c.inc b/tcg/ppc/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/ppc/tcg-target.c.inc +++ b/tcg/ppc/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ static bool reloc_pc34(tcg_insn_unit *src_rw, const tcg_insn_unit *target) } /* test if a constant matches the constraint */ -static bool tcg_target_const_match(int64_t val, TCGType type, int ct, int vece) +static bool tcg_target_const_match(int64_t val, int ct, + TCGType type, TCGCond cond, int vece) { if (ct & TCG_CT_CONST) { return 1; diff --git a/tcg/riscv/tcg-target.c.inc b/tcg/riscv/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/riscv/tcg-target.c.inc +++ b/tcg/riscv/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ static TCGReg tcg_target_call_oarg_reg(TCGCallReturnKind kind, int slot) #define sextreg sextract64 /* test if a constant matches the constraint */ -static bool tcg_target_const_match(int64_t val, TCGType type, int ct, int vece) +static bool tcg_target_const_match(int64_t val, int ct, + TCGType type, TCGCond cond, int vece) { if (ct & TCG_CT_CONST) { return 1; diff --git a/tcg/s390x/tcg-target.c.inc b/tcg/s390x/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/s390x/tcg-target.c.inc +++ b/tcg/s390x/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ static bool risbg_mask(uint64_t c) } /* Test if a constant matches the constraint. */ -static bool tcg_target_const_match(int64_t val, TCGType type, int ct, int vece) +static bool tcg_target_const_match(int64_t val, int ct, + TCGType type, TCGCond cond, int vece) { if (ct & TCG_CT_CONST) { return 1; diff --git a/tcg/sparc64/tcg-target.c.inc b/tcg/sparc64/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/sparc64/tcg-target.c.inc +++ b/tcg/sparc64/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ static bool patch_reloc(tcg_insn_unit *src_rw, int type, } /* test if a constant matches the constraint */ -static bool tcg_target_const_match(int64_t val, TCGType type, int ct, int vece) +static bool tcg_target_const_match(int64_t val, int ct, + TCGType type, TCGCond cond, int vece) { if (ct & TCG_CT_CONST) { return 1; diff --git a/tcg/tci/tcg-target.c.inc b/tcg/tci/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/tci/tcg-target.c.inc +++ b/tcg/tci/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ static inline bool tcg_out_sti(TCGContext *s, TCGType type, TCGArg val, } /* Test if a constant matches the constraint. */ -static bool tcg_target_const_match(int64_t val, TCGType type, int ct, int vece) +static bool tcg_target_const_match(int64_t val, int ct, + TCGType type, TCGCond cond, int vece) { return ct & TCG_CT_CONST; } -- 2.34.1 Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/aarch64/tcg-target-con-set.h | 5 +-- tcg/aarch64/tcg-target-con-str.h | 1 + tcg/aarch64/tcg-target.h | 2 +- tcg/aarch64/tcg-target.c.inc | 56 ++++++++++++++++++++++---------- 4 files changed, 44 insertions(+), 20 deletions(-) diff --git a/tcg/aarch64/tcg-target-con-set.h b/tcg/aarch64/tcg-target-con-set.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/aarch64/tcg-target-con-set.h +++ b/tcg/aarch64/tcg-target-con-set.h @@ -XXX,XX +XXX,XX @@ * tcg-target-con-str.h; the constraint combination is inclusive or. */ C_O0_I1(r) -C_O0_I2(r, rA) +C_O0_I2(r, rC) C_O0_I2(rZ, r) C_O0_I2(w, r) C_O0_I3(rZ, rZ, r) @@ -XXX,XX +XXX,XX @@ C_O1_I2(r, 0, rZ) C_O1_I2(r, r, r) C_O1_I2(r, r, rA) C_O1_I2(r, r, rAL) +C_O1_I2(r, r, rC) C_O1_I2(r, r, ri) C_O1_I2(r, r, rL) C_O1_I2(r, rZ, rZ) @@ -XXX,XX +XXX,XX @@ C_O1_I2(w, w, wN) C_O1_I2(w, w, wO) C_O1_I2(w, w, wZ) C_O1_I3(w, w, w, w) -C_O1_I4(r, r, rA, rZ, rZ) +C_O1_I4(r, r, rC, rZ, rZ) C_O2_I1(r, r, r) C_O2_I4(r, r, rZ, rZ, rA, rMZ) diff --git a/tcg/aarch64/tcg-target-con-str.h b/tcg/aarch64/tcg-target-con-str.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/aarch64/tcg-target-con-str.h +++ b/tcg/aarch64/tcg-target-con-str.h @@ -XXX,XX +XXX,XX @@ REGS('w', ALL_VECTOR_REGS) * CONST(letter, TCG_CT_CONST_* bit set) */ CONST('A', TCG_CT_CONST_AIMM) +CONST('C', TCG_CT_CONST_CMP) CONST('L', TCG_CT_CONST_LIMM) CONST('M', TCG_CT_CONST_MONE) CONST('O', TCG_CT_CONST_ORRI) diff --git a/tcg/aarch64/tcg-target.h b/tcg/aarch64/tcg-target.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/aarch64/tcg-target.h +++ b/tcg/aarch64/tcg-target.h @@ -XXX,XX +XXX,XX @@ typedef enum { #define TCG_TARGET_HAS_qemu_ldst_i128 1 #endif -#define TCG_TARGET_HAS_tst 0 +#define TCG_TARGET_HAS_tst 1 #define TCG_TARGET_HAS_v64 1 #define TCG_TARGET_HAS_v128 1 diff --git a/tcg/aarch64/tcg-target.c.inc b/tcg/aarch64/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/aarch64/tcg-target.c.inc +++ b/tcg/aarch64/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ static bool patch_reloc(tcg_insn_unit *code_ptr, int type, #define TCG_CT_CONST_MONE 0x800 #define TCG_CT_CONST_ORRI 0x1000 #define TCG_CT_CONST_ANDI 0x2000 +#define TCG_CT_CONST_CMP 0x4000 #define ALL_GENERAL_REGS 0xffffffffu #define ALL_VECTOR_REGS 0xffffffff00000000ull @@ -XXX,XX +XXX,XX @@ static bool tcg_target_const_match(int64_t val, int ct, if (type == TCG_TYPE_I32) { val = (int32_t)val; } + + if (ct & TCG_CT_CONST_CMP) { + if (is_tst_cond(cond)) { + ct |= TCG_CT_CONST_LIMM; + } else { + ct |= TCG_CT_CONST_AIMM; + } + } + if ((ct & TCG_CT_CONST_AIMM) && (is_aimm(val) || is_aimm(-val))) { return 1; } @@ -XXX,XX +XXX,XX @@ static const enum aarch64_cond_code tcg_cond_to_aarch64[] = { [TCG_COND_GTU] = COND_HI, [TCG_COND_GEU] = COND_HS, [TCG_COND_LEU] = COND_LS, + /* bit test */ + [TCG_COND_TSTEQ] = COND_EQ, + [TCG_COND_TSTNE] = COND_NE, }; typedef enum { @@ -XXX,XX +XXX,XX @@ static inline void tcg_out_dep(TCGContext *s, TCGType ext, TCGReg rd, tcg_out_bfm(s, ext, rd, rn, a, b); } -static void tcg_out_cmp(TCGContext *s, TCGType ext, TCGReg a, +static void tcg_out_cmp(TCGContext *s, TCGType ext, TCGCond cond, TCGReg a, tcg_target_long b, bool const_b) { - if (const_b) { - /* Using CMP or CMN aliases. */ - if (b >= 0) { - tcg_out_insn(s, 3401, SUBSI, ext, TCG_REG_XZR, a, b); + if (is_tst_cond(cond)) { + if (!const_b) { + tcg_out_insn(s, 3510, ANDS, ext, TCG_REG_XZR, a, b); } else { - tcg_out_insn(s, 3401, ADDSI, ext, TCG_REG_XZR, a, -b); + tcg_debug_assert(is_limm(b)); + tcg_out_logicali(s, I3404_ANDSI, 0, TCG_REG_XZR, a, b); } } else { - /* Using CMP alias SUBS wzr, Wn, Wm */ - tcg_out_insn(s, 3502, SUBS, ext, TCG_REG_XZR, a, b); + if (!const_b) { + tcg_out_insn(s, 3502, SUBS, ext, TCG_REG_XZR, a, b); + } else if (b >= 0) { + tcg_debug_assert(is_aimm(b)); + tcg_out_insn(s, 3401, SUBSI, ext, TCG_REG_XZR, a, b); + } else { + tcg_debug_assert(is_aimm(-b)); + tcg_out_insn(s, 3401, ADDSI, ext, TCG_REG_XZR, a, -b); + } } } @@ -XXX,XX +XXX,XX @@ static void tcg_out_brcond(TCGContext *s, TCGType ext, TCGCond c, TCGArg a, need_cmp = false; } else { need_cmp = true; - tcg_out_cmp(s, ext, a, b, b_const); + tcg_out_cmp(s, ext, c, a, b, b_const); } if (!l->has_value) { @@ -XXX,XX +XXX,XX @@ static void tcg_out_cltz(TCGContext *s, TCGType ext, TCGReg d, } else { AArch64Insn sel = I3506_CSEL; - tcg_out_cmp(s, ext, a0, 0, 1); + tcg_out_cmp(s, ext, TCG_COND_NE, a0, 0, 1); tcg_out_insn(s, 3507, CLZ, ext, TCG_REG_TMP0, a1); if (const_b) { @@ -XXX,XX +XXX,XX @@ static TCGLabelQemuLdst *prepare_host_addr(TCGContext *s, HostAddress *h, addr_adj, compare_mask); /* Perform the address comparison. */ - tcg_out_cmp(s, addr_type, TCG_REG_TMP0, TCG_REG_TMP2, 0); + tcg_out_cmp(s, addr_type, TCG_COND_NE, TCG_REG_TMP0, TCG_REG_TMP2, 0); /* If not equal, we jump to the slow path. */ ldst->label_ptr[0] = s->code_ptr; @@ -XXX,XX +XXX,XX @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc, a2 = (int32_t)a2; /* FALLTHRU */ case INDEX_op_setcond_i64: - tcg_out_cmp(s, ext, a1, a2, c2); + tcg_out_cmp(s, ext, args[3], a1, a2, c2); /* Use CSET alias of CSINC Wd, WZR, WZR, invert(cond). */ tcg_out_insn(s, 3506, CSINC, TCG_TYPE_I32, a0, TCG_REG_XZR, TCG_REG_XZR, tcg_invert_cond(args[3])); @@ -XXX,XX +XXX,XX @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc, a2 = (int32_t)a2; /* FALLTHRU */ case INDEX_op_negsetcond_i64: - tcg_out_cmp(s, ext, a1, a2, c2); + tcg_out_cmp(s, ext, args[3], a1, a2, c2); /* Use CSETM alias of CSINV Wd, WZR, WZR, invert(cond). */ tcg_out_insn(s, 3506, CSINV, ext, a0, TCG_REG_XZR, TCG_REG_XZR, tcg_invert_cond(args[3])); @@ -XXX,XX +XXX,XX @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc, a2 = (int32_t)a2; /* FALLTHRU */ case INDEX_op_movcond_i64: - tcg_out_cmp(s, ext, a1, a2, c2); + tcg_out_cmp(s, ext, args[5], a1, a2, c2); tcg_out_insn(s, 3506, CSEL, ext, a0, REG0(3), REG0(4), args[5]); break; @@ -XXX,XX +XXX,XX @@ static TCGConstraintSetIndex tcg_target_op_def(TCGOpcode op) case INDEX_op_add_i64: case INDEX_op_sub_i32: case INDEX_op_sub_i64: + return C_O1_I2(r, r, rA); + case INDEX_op_setcond_i32: case INDEX_op_setcond_i64: case INDEX_op_negsetcond_i32: case INDEX_op_negsetcond_i64: - return C_O1_I2(r, r, rA); + return C_O1_I2(r, r, rC); case INDEX_op_mul_i32: case INDEX_op_mul_i64: @@ -XXX,XX +XXX,XX @@ static TCGConstraintSetIndex tcg_target_op_def(TCGOpcode op) case INDEX_op_brcond_i32: case INDEX_op_brcond_i64: - return C_O0_I2(r, rA); + return C_O0_I2(r, rC); case INDEX_op_movcond_i32: case INDEX_op_movcond_i64: - return C_O1_I4(r, r, rA, rZ, rZ); + return C_O1_I4(r, r, rC, rZ, rZ); case INDEX_op_qemu_ld_a32_i32: case INDEX_op_qemu_ld_a64_i32: -- 2.34.1 From: Philippe Mathieu-Daudé <philmd@linaro.org> In order to ease next commit review, modify tcg_out_brcond() to switch over TCGCond. No logical change intended. Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org> Message-Id: <20240119224737.48943-1-philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/aarch64/tcg-target.c.inc | 31 +++++++++++++++++++++++-------- 1 file changed, 23 insertions(+), 8 deletions(-) diff --git a/tcg/aarch64/tcg-target.c.inc b/tcg/aarch64/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/aarch64/tcg-target.c.inc +++ b/tcg/aarch64/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ static void tcg_out_brcond(TCGContext *s, TCGType ext, TCGCond c, TCGArg a, TCGArg b, bool b_const, TCGLabel *l) { intptr_t offset; - bool need_cmp; + bool need_cmp = true; - if (b_const && b == 0 && (c == TCG_COND_EQ || c == TCG_COND_NE)) { - need_cmp = false; - } else { - need_cmp = true; + switch (c) { + case TCG_COND_EQ: + case TCG_COND_NE: + if (b_const && b == 0) { + need_cmp = false; + } + break; + default: + break; + } + + if (need_cmp) { tcg_out_cmp(s, ext, c, a, b, b_const); } @@ -XXX,XX +XXX,XX @@ static void tcg_out_brcond(TCGContext *s, TCGType ext, TCGCond c, TCGArg a, if (need_cmp) { tcg_out_insn(s, 3202, B_C, c, offset); - } else if (c == TCG_COND_EQ) { - tcg_out_insn(s, 3201, CBZ, ext, a, offset); } else { - tcg_out_insn(s, 3201, CBNZ, ext, a, offset); + switch (c) { + case TCG_COND_EQ: + tcg_out_insn(s, 3201, CBZ, ext, a, offset); + break; + case TCG_COND_NE: + tcg_out_insn(s, 3201, CBNZ, ext, a, offset); + break; + default: + g_assert_not_reached(); + } } } -- 2.34.1 Test the sign bit for LT/GE vs 0, and TSTNE/EQ vs a power of 2. Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org> Message-Id: <20240119224737.48943-2-philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/aarch64/tcg-target.c.inc | 74 ++++++++++++++++++++++++++++++------ 1 file changed, 62 insertions(+), 12 deletions(-) diff --git a/tcg/aarch64/tcg-target.c.inc b/tcg/aarch64/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/aarch64/tcg-target.c.inc +++ b/tcg/aarch64/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ static bool reloc_pc19(tcg_insn_unit *src_rw, const tcg_insn_unit *target) return false; } +static bool reloc_pc14(tcg_insn_unit *src_rw, const tcg_insn_unit *target) +{ + const tcg_insn_unit *src_rx = tcg_splitwx_to_rx(src_rw); + ptrdiff_t offset = target - src_rx; + + if (offset == sextract64(offset, 0, 14)) { + *src_rw = deposit32(*src_rw, 5, 14, offset); + return true; + } + return false; +} + static bool patch_reloc(tcg_insn_unit *code_ptr, int type, intptr_t value, intptr_t addend) { @@ -XXX,XX +XXX,XX @@ static bool patch_reloc(tcg_insn_unit *code_ptr, int type, return reloc_pc26(code_ptr, (const tcg_insn_unit *)value); case R_AARCH64_CONDBR19: return reloc_pc19(code_ptr, (const tcg_insn_unit *)value); + case R_AARCH64_TSTBR14: + return reloc_pc14(code_ptr, (const tcg_insn_unit *)value); default: g_assert_not_reached(); } @@ -XXX,XX +XXX,XX @@ typedef enum { /* Conditional branch (immediate). */ I3202_B_C = 0x54000000, + /* Test and branch (immediate). */ + I3205_TBZ = 0x36000000, + I3205_TBNZ = 0x37000000, + /* Unconditional branch (immediate). */ I3206_B = 0x14000000, I3206_BL = 0x94000000, @@ -XXX,XX +XXX,XX @@ static void tcg_out_insn_3202(TCGContext *s, AArch64Insn insn, tcg_out32(s, insn | tcg_cond_to_aarch64[c] | (imm19 & 0x7ffff) << 5); } +static void tcg_out_insn_3205(TCGContext *s, AArch64Insn insn, + TCGReg rt, int imm6, int imm14) +{ + insn |= (imm6 & 0x20) << (31 - 5); + insn |= (imm6 & 0x1f) << 19; + tcg_out32(s, insn | (imm14 & 0x3fff) << 5 | rt); +} + static void tcg_out_insn_3206(TCGContext *s, AArch64Insn insn, int imm26) { tcg_out32(s, insn | (imm26 & 0x03ffffff)); @@ -XXX,XX +XXX,XX @@ static inline void tcg_out_goto_label(TCGContext *s, TCGLabel *l) static void tcg_out_brcond(TCGContext *s, TCGType ext, TCGCond c, TCGArg a, TCGArg b, bool b_const, TCGLabel *l) { - intptr_t offset; + int tbit = -1; bool need_cmp = true; switch (c) { case TCG_COND_EQ: case TCG_COND_NE: + /* cmp xN,0; b.ne L -> cbnz xN,L */ if (b_const && b == 0) { need_cmp = false; } break; + case TCG_COND_LT: + case TCG_COND_GE: + /* cmp xN,0; b.mi L -> tbnz xN,63,L */ + if (b_const && b == 0) { + c = (c == TCG_COND_LT ? TCG_COND_TSTNE : TCG_COND_TSTEQ); + tbit = ext ? 63 : 31; + need_cmp = false; + } + break; + case TCG_COND_TSTEQ: + case TCG_COND_TSTNE: + /* tst xN,1<<B; b.ne L -> tbnz xN,B,L */ + if (b_const && is_power_of_2(b)) { + tbit = ctz64(b); + need_cmp = false; + } + break; default: break; } if (need_cmp) { tcg_out_cmp(s, ext, c, a, b, b_const); - } - - if (!l->has_value) { tcg_out_reloc(s, s->code_ptr, R_AARCH64_CONDBR19, l, 0); - offset = tcg_in32(s) >> 5; - } else { - offset = tcg_pcrel_diff(s, l->u.value_ptr) >> 2; - tcg_debug_assert(offset == sextract64(offset, 0, 19)); + tcg_out_insn(s, 3202, B_C, c, 0); + return; } - if (need_cmp) { - tcg_out_insn(s, 3202, B_C, c, offset); + if (tbit >= 0) { + tcg_out_reloc(s, s->code_ptr, R_AARCH64_TSTBR14, l, 0); + switch (c) { + case TCG_COND_TSTEQ: + tcg_out_insn(s, 3205, TBZ, a, tbit, 0); + break; + case TCG_COND_TSTNE: + tcg_out_insn(s, 3205, TBNZ, a, tbit, 0); + break; + default: + g_assert_not_reached(); + } } else { + tcg_out_reloc(s, s->code_ptr, R_AARCH64_CONDBR19, l, 0); switch (c) { case TCG_COND_EQ: - tcg_out_insn(s, 3201, CBZ, ext, a, offset); + tcg_out_insn(s, 3201, CBZ, ext, a, 0); break; case TCG_COND_NE: - tcg_out_insn(s, 3201, CBNZ, ext, a, offset); + tcg_out_insn(s, 3201, CBNZ, ext, a, 0); break; default: g_assert_not_reached(); -- 2.34.1 ... and the inverse, CBZ for TSTEQ. Suggested-by: Paolo Bonzini <pbonzini@redhat.com> Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/aarch64/tcg-target.c.inc | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/tcg/aarch64/tcg-target.c.inc b/tcg/aarch64/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/aarch64/tcg-target.c.inc +++ b/tcg/aarch64/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ static void tcg_out_brcond(TCGContext *s, TCGType ext, TCGCond c, TCGArg a, break; case TCG_COND_TSTEQ: case TCG_COND_TSTNE: + /* tst xN,0xffffffff; b.ne L -> cbnz wN,L */ + if (b_const && b == UINT32_MAX) { + ext = TCG_TYPE_I32; + need_cmp = false; + break; + } /* tst xN,1<<B; b.ne L -> tbnz xN,B,L */ if (b_const && is_power_of_2(b)) { tbit = ctz64(b); -- 2.34.1 Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-Id: <20231028194522.245170-12-richard.henderson@linaro.org> [PMD: Split from bigger patch, part 1/2] Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org> Message-Id: <20231108145244.72421-1-philmd@linaro.org> --- tcg/arm/tcg-target.c.inc | 32 +++++++++++++++++--------------- 1 file changed, 17 insertions(+), 15 deletions(-) diff --git a/tcg/arm/tcg-target.c.inc b/tcg/arm/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/arm/tcg-target.c.inc +++ b/tcg/arm/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ static void tcg_out_mb(TCGContext *s, TCGArg a0) } } +static TCGCond tcg_out_cmp(TCGContext *s, TCGCond cond, TCGReg a, + TCGArg b, int b_const) +{ + tcg_out_dat_rIN(s, COND_AL, ARITH_CMP, ARITH_CMN, 0, a, b, b_const); + return cond; +} + static TCGCond tcg_out_cmp2(TCGContext *s, const TCGArg *args, const int *const_args) { @@ -XXX,XX +XXX,XX @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc, /* Constraints mean that v2 is always in the same register as dest, * so we only need to do "if condition passed, move v1 to dest". */ - tcg_out_dat_rIN(s, COND_AL, ARITH_CMP, ARITH_CMN, 0, - args[1], args[2], const_args[2]); - tcg_out_dat_rIK(s, tcg_cond_to_arm_cond[args[5]], ARITH_MOV, + c = tcg_out_cmp(s, args[5], args[1], args[2], const_args[2]); + tcg_out_dat_rIK(s, tcg_cond_to_arm_cond[c], ARITH_MOV, ARITH_MVN, args[0], 0, args[3], const_args[3]); break; case INDEX_op_add_i32: @@ -XXX,XX +XXX,XX @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc, break; case INDEX_op_brcond_i32: - tcg_out_dat_rIN(s, COND_AL, ARITH_CMP, ARITH_CMN, 0, - args[0], args[1], const_args[1]); - tcg_out_goto_label(s, tcg_cond_to_arm_cond[args[2]], - arg_label(args[3])); + c = tcg_out_cmp(s, args[2], args[0], args[1], const_args[1]); + tcg_out_goto_label(s, tcg_cond_to_arm_cond[c], arg_label(args[3])); break; case INDEX_op_setcond_i32: - tcg_out_dat_rIN(s, COND_AL, ARITH_CMP, ARITH_CMN, 0, - args[1], args[2], const_args[2]); - tcg_out_dat_imm(s, tcg_cond_to_arm_cond[args[3]], + c = tcg_out_cmp(s, args[3], args[1], args[2], const_args[2]); + tcg_out_dat_imm(s, tcg_cond_to_arm_cond[c], ARITH_MOV, args[0], 0, 1); - tcg_out_dat_imm(s, tcg_cond_to_arm_cond[tcg_invert_cond(args[3])], + tcg_out_dat_imm(s, tcg_cond_to_arm_cond[tcg_invert_cond(c)], ARITH_MOV, args[0], 0, 0); break; case INDEX_op_negsetcond_i32: - tcg_out_dat_rIN(s, COND_AL, ARITH_CMP, ARITH_CMN, 0, - args[1], args[2], const_args[2]); - tcg_out_dat_imm(s, tcg_cond_to_arm_cond[args[3]], + c = tcg_out_cmp(s, args[3], args[1], args[2], const_args[2]); + tcg_out_dat_imm(s, tcg_cond_to_arm_cond[c], ARITH_MVN, args[0], 0, 0); - tcg_out_dat_imm(s, tcg_cond_to_arm_cond[tcg_invert_cond(args[3])], + tcg_out_dat_imm(s, tcg_cond_to_arm_cond[tcg_invert_cond(c)], ARITH_MOV, args[0], 0, 0); break; -- 2.34.1 Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-Id: <20231028194522.245170-12-richard.henderson@linaro.org> [PMD: Split from bigger patch, part 2/2] Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org> Message-Id: <20231108145244.72421-2-philmd@linaro.org> --- tcg/arm/tcg-target.h | 2 +- tcg/arm/tcg-target.c.inc | 29 ++++++++++++++++++++++++++++- 2 files changed, 29 insertions(+), 2 deletions(-) diff --git a/tcg/arm/tcg-target.h b/tcg/arm/tcg-target.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/arm/tcg-target.h +++ b/tcg/arm/tcg-target.h @@ -XXX,XX +XXX,XX @@ extern bool use_neon_instructions; #define TCG_TARGET_HAS_qemu_ldst_i128 0 -#define TCG_TARGET_HAS_tst 0 +#define TCG_TARGET_HAS_tst 1 #define TCG_TARGET_HAS_v64 use_neon_instructions #define TCG_TARGET_HAS_v128 use_neon_instructions diff --git a/tcg/arm/tcg-target.c.inc b/tcg/arm/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/arm/tcg-target.c.inc +++ b/tcg/arm/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ static void tcg_out_mb(TCGContext *s, TCGArg a0) static TCGCond tcg_out_cmp(TCGContext *s, TCGCond cond, TCGReg a, TCGArg b, int b_const) { - tcg_out_dat_rIN(s, COND_AL, ARITH_CMP, ARITH_CMN, 0, a, b, b_const); + if (!is_tst_cond(cond)) { + tcg_out_dat_rIN(s, COND_AL, ARITH_CMP, ARITH_CMN, 0, a, b, b_const); + return cond; + } + + cond = tcg_tst_eqne_cond(cond); + if (b_const) { + int imm12 = encode_imm(b); + + /* + * The compare constraints allow rIN, but TST does not support N. + * Be prepared to load the constant into a scratch register. + */ + if (imm12 >= 0) { + tcg_out_dat_imm(s, COND_AL, ARITH_TST, 0, a, imm12); + return cond; + } + tcg_out_movi32(s, COND_AL, TCG_REG_TMP, b); + b = TCG_REG_TMP; + } + tcg_out_dat_reg(s, COND_AL, ARITH_TST, 0, a, b, SHIFT_IMM_LSL(0)); return cond; } @@ -XXX,XX +XXX,XX @@ static TCGCond tcg_out_cmp2(TCGContext *s, const TCGArg *args, tcg_out_dat_rI(s, COND_EQ, ARITH_CMP, 0, al, bl, const_bl); return cond; + case TCG_COND_TSTEQ: + case TCG_COND_TSTNE: + /* Similar, but with TST instead of CMP. */ + tcg_out_dat_rI(s, COND_AL, ARITH_TST, 0, ah, bh, const_bh); + tcg_out_dat_rI(s, COND_EQ, ARITH_TST, 0, al, bl, const_bl); + return tcg_tst_eqne_cond(cond); + case TCG_COND_LT: case TCG_COND_GE: /* We perform a double-word subtraction and examine the result. -- 2.34.1 Hoist the tcg_cond_to_jcc index outside the function. Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/i386/tcg-target.c.inc | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) diff --git a/tcg/i386/tcg-target.c.inc b/tcg/i386/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/i386/tcg-target.c.inc +++ b/tcg/i386/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ static void tcg_out_setcond2(TCGContext *s, const TCGArg *args, } #endif -static void tcg_out_cmov(TCGContext *s, TCGCond cond, int rexw, +static void tcg_out_cmov(TCGContext *s, int jcc, int rexw, TCGReg dest, TCGReg v1) { if (have_cmov) { - tcg_out_modrm(s, OPC_CMOVCC | tcg_cond_to_jcc[cond] | rexw, dest, v1); + tcg_out_modrm(s, OPC_CMOVCC | jcc | rexw, dest, v1); } else { TCGLabel *over = gen_new_label(); - tcg_out_jxx(s, tcg_cond_to_jcc[tcg_invert_cond(cond)], over, 1); + tcg_out_jxx(s, jcc ^ 1, over, 1); tcg_out_mov(s, TCG_TYPE_I32, dest, v1); tcg_out_label(s, over); } @@ -XXX,XX +XXX,XX @@ static void tcg_out_movcond(TCGContext *s, int rexw, TCGCond cond, TCGReg v1) { tcg_out_cmp(s, c1, c2, const_c2, rexw); - tcg_out_cmov(s, cond, rexw, dest, v1); + tcg_out_cmov(s, tcg_cond_to_jcc[cond], rexw, dest, v1); } static void tcg_out_ctz(TCGContext *s, int rexw, TCGReg dest, TCGReg arg1, @@ -XXX,XX +XXX,XX @@ static void tcg_out_ctz(TCGContext *s, int rexw, TCGReg dest, TCGReg arg1, tcg_debug_assert(arg2 == (rexw ? 64 : 32)); } else { tcg_debug_assert(dest != arg2); - tcg_out_cmov(s, TCG_COND_LTU, rexw, dest, arg2); + tcg_out_cmov(s, JCC_JB, rexw, dest, arg2); } } else { tcg_debug_assert(dest != arg2); tcg_out_modrm(s, OPC_BSF + rexw, dest, arg1); - tcg_out_cmov(s, TCG_COND_EQ, rexw, dest, arg2); + tcg_out_cmov(s, JCC_JE, rexw, dest, arg2); } } @@ -XXX,XX +XXX,XX @@ static void tcg_out_clz(TCGContext *s, int rexw, TCGReg dest, TCGReg arg1, tcg_debug_assert(arg2 == (rexw ? 64 : 32)); } else { tcg_debug_assert(dest != arg2); - tcg_out_cmov(s, TCG_COND_LTU, rexw, dest, arg2); + tcg_out_cmov(s, JCC_JB, rexw, dest, arg2); } } else { tcg_debug_assert(!const_a2); @@ -XXX,XX +XXX,XX @@ static void tcg_out_clz(TCGContext *s, int rexw, TCGReg dest, TCGReg arg1, /* Since we have destroyed the flags from BSR, we have to re-test. */ tcg_out_cmp(s, arg1, 0, 1, rexw); - tcg_out_cmov(s, TCG_COND_EQ, rexw, dest, arg2); + tcg_out_cmov(s, JCC_JE, rexw, dest, arg2); } } -- 2.34.1 Return the x86 condition codes to use after the compare. Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/i386/tcg-target.c.inc | 24 +++++++++++++----------- 1 file changed, 13 insertions(+), 11 deletions(-) diff --git a/tcg/i386/tcg-target.c.inc b/tcg/i386/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/i386/tcg-target.c.inc +++ b/tcg/i386/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ static void tcg_out_jxx(TCGContext *s, int opc, TCGLabel *l, bool small) } } -static void tcg_out_cmp(TCGContext *s, TCGArg arg1, TCGArg arg2, - int const_arg2, int rexw) +static int tcg_out_cmp(TCGContext *s, TCGCond cond, TCGArg arg1, + TCGArg arg2, int const_arg2, int rexw) { if (const_arg2) { if (arg2 == 0) { @@ -XXX,XX +XXX,XX @@ static void tcg_out_cmp(TCGContext *s, TCGArg arg1, TCGArg arg2, } else { tgen_arithr(s, ARITH_CMP + rexw, arg1, arg2); } + return tcg_cond_to_jcc[cond]; } static void tcg_out_brcond(TCGContext *s, int rexw, TCGCond cond, TCGArg arg1, TCGArg arg2, int const_arg2, TCGLabel *label, bool small) { - tcg_out_cmp(s, arg1, arg2, const_arg2, rexw); - tcg_out_jxx(s, tcg_cond_to_jcc[cond], label, small); + int jcc = tcg_out_cmp(s, cond, arg1, arg2, const_arg2, rexw); + tcg_out_jxx(s, jcc, label, small); } #if TCG_TARGET_REG_BITS == 32 @@ -XXX,XX +XXX,XX @@ static void tcg_out_setcond(TCGContext *s, int rexw, TCGCond cond, { bool inv = false; bool cleared; + int jcc; switch (cond) { case TCG_COND_NE: @@ -XXX,XX +XXX,XX @@ static void tcg_out_setcond(TCGContext *s, int rexw, TCGCond cond, * We can then use NEG or INC to produce the desired result. * This is always smaller than the SETCC expansion. */ - tcg_out_cmp(s, arg1, arg2, const_arg2, rexw); + tcg_out_cmp(s, TCG_COND_LTU, arg1, arg2, const_arg2, rexw); /* X - X - C = -C = (C ? -1 : 0) */ tgen_arithr(s, ARITH_SBB + (neg ? rexw : 0), dest, dest); @@ -XXX,XX +XXX,XX @@ static void tcg_out_setcond(TCGContext *s, int rexw, TCGCond cond, cleared = true; } - tcg_out_cmp(s, arg1, arg2, const_arg2, rexw); - tcg_out_modrm(s, OPC_SETCC | tcg_cond_to_jcc[cond], 0, dest); + jcc = tcg_out_cmp(s, cond, arg1, arg2, const_arg2, rexw); + tcg_out_modrm(s, OPC_SETCC | jcc, 0, dest); if (!cleared) { tcg_out_ext8u(s, dest, dest); @@ -XXX,XX +XXX,XX @@ static void tcg_out_movcond(TCGContext *s, int rexw, TCGCond cond, TCGReg dest, TCGReg c1, TCGArg c2, int const_c2, TCGReg v1) { - tcg_out_cmp(s, c1, c2, const_c2, rexw); - tcg_out_cmov(s, tcg_cond_to_jcc[cond], rexw, dest, v1); + int jcc = tcg_out_cmp(s, cond, c1, c2, const_c2, rexw); + tcg_out_cmov(s, jcc, rexw, dest, v1); } static void tcg_out_ctz(TCGContext *s, int rexw, TCGReg dest, TCGReg arg1, @@ -XXX,XX +XXX,XX @@ static void tcg_out_clz(TCGContext *s, int rexw, TCGReg dest, TCGReg arg1, tgen_arithi(s, ARITH_XOR + rexw, dest, rexw ? 63 : 31, 0); /* Since we have destroyed the flags from BSR, we have to re-test. */ - tcg_out_cmp(s, arg1, 0, 1, rexw); - tcg_out_cmov(s, JCC_JE, rexw, dest, arg2); + int jcc = tcg_out_cmp(s, TCG_COND_EQ, arg1, 0, 1, rexw); + tcg_out_cmov(s, jcc, rexw, dest, arg2); } } -- 2.34.1 Merge tcg_out_testi into tcg_out_cmp and adjust the two uses. Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/i386/tcg-target.h | 2 +- tcg/i386/tcg-target.c.inc | 95 ++++++++++++++++++++++++--------------- 2 files changed, 60 insertions(+), 37 deletions(-) diff --git a/tcg/i386/tcg-target.h b/tcg/i386/tcg-target.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/i386/tcg-target.h +++ b/tcg/i386/tcg-target.h @@ -XXX,XX +XXX,XX @@ typedef enum { #define TCG_TARGET_HAS_qemu_ldst_i128 \ (TCG_TARGET_REG_BITS == 64 && (cpuinfo & CPUINFO_ATOMIC_VMOVDQA)) -#define TCG_TARGET_HAS_tst 0 +#define TCG_TARGET_HAS_tst 1 /* We do not support older SSE systems, only beginning with AVX1. */ #define TCG_TARGET_HAS_v64 have_avx1 diff --git a/tcg/i386/tcg-target.c.inc b/tcg/i386/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/i386/tcg-target.c.inc +++ b/tcg/i386/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ static const uint8_t tcg_cond_to_jcc[] = { [TCG_COND_GEU] = JCC_JAE, [TCG_COND_LEU] = JCC_JBE, [TCG_COND_GTU] = JCC_JA, + [TCG_COND_TSTEQ] = JCC_JE, + [TCG_COND_TSTNE] = JCC_JNE, }; #if TCG_TARGET_REG_BITS == 64 @@ -XXX,XX +XXX,XX @@ static void tcg_out_jxx(TCGContext *s, int opc, TCGLabel *l, bool small) static int tcg_out_cmp(TCGContext *s, TCGCond cond, TCGArg arg1, TCGArg arg2, int const_arg2, int rexw) { - if (const_arg2) { - if (arg2 == 0) { - /* test r, r */ + int jz; + + if (!is_tst_cond(cond)) { + if (!const_arg2) { + tgen_arithr(s, ARITH_CMP + rexw, arg1, arg2); + } else if (arg2 == 0) { tcg_out_modrm(s, OPC_TESTL + rexw, arg1, arg1); } else { + tcg_debug_assert(!rexw || arg2 == (int32_t)arg2); tgen_arithi(s, ARITH_CMP + rexw, arg1, arg2, 0); } - } else { - tgen_arithr(s, ARITH_CMP + rexw, arg1, arg2); + return tcg_cond_to_jcc[cond]; } - return tcg_cond_to_jcc[cond]; + + jz = tcg_cond_to_jcc[cond]; + + if (!const_arg2) { + tcg_out_modrm(s, OPC_TESTL + rexw, arg1, arg2); + return jz; + } + + if (arg2 <= 0xff && (TCG_TARGET_REG_BITS == 64 || arg1 < 4)) { + tcg_out_modrm(s, OPC_GRP3_Eb | P_REXB_RM, EXT3_TESTi, arg1); + tcg_out8(s, arg2); + return jz; + } + + if ((arg2 & ~0xff00) == 0 && arg1 < 4) { + tcg_out_modrm(s, OPC_GRP3_Eb, EXT3_TESTi, arg1 + 4); + tcg_out8(s, arg2 >> 8); + return jz; + } + + if (rexw) { + if (arg2 == (uint32_t)arg2) { + rexw = 0; + } else { + tcg_debug_assert(arg2 == (int32_t)arg2); + } + } + tcg_out_modrm(s, OPC_GRP3_Ev + rexw, EXT3_TESTi, arg1); + tcg_out32(s, arg2); + return jz; } static void tcg_out_brcond(TCGContext *s, int rexw, TCGCond cond, @@ -XXX,XX +XXX,XX @@ static void tcg_out_brcond2(TCGContext *s, const TCGArg *args, { TCGLabel *label_next = gen_new_label(); TCGLabel *label_this = arg_label(args[5]); + TCGCond cond = args[4]; - switch(args[4]) { + switch (cond) { case TCG_COND_EQ: - tcg_out_brcond(s, 0, TCG_COND_NE, args[0], args[2], const_args[2], - label_next, 1); - tcg_out_brcond(s, 0, TCG_COND_EQ, args[1], args[3], const_args[3], + case TCG_COND_TSTEQ: + tcg_out_brcond(s, 0, tcg_invert_cond(cond), + args[0], args[2], const_args[2], label_next, 1); + tcg_out_brcond(s, 0, cond, args[1], args[3], const_args[3], label_this, small); break; case TCG_COND_NE: - tcg_out_brcond(s, 0, TCG_COND_NE, args[0], args[2], const_args[2], + case TCG_COND_TSTNE: + tcg_out_brcond(s, 0, cond, args[0], args[2], const_args[2], label_this, small); - tcg_out_brcond(s, 0, TCG_COND_NE, args[1], args[3], const_args[3], + tcg_out_brcond(s, 0, cond, args[1], args[3], const_args[3], label_this, small); break; case TCG_COND_LT: @@ -XXX,XX +XXX,XX @@ static void tcg_out_nopn(TCGContext *s, int n) tcg_out8(s, 0x90); } -/* Test register R vs immediate bits I, setting Z flag for EQ/NE. */ -static void __attribute__((unused)) -tcg_out_testi(TCGContext *s, TCGReg r, uint32_t i) -{ - /* - * This is used for testing alignment, so we can usually use testb. - * For i686, we have to use testl for %esi/%edi. - */ - if (i <= 0xff && (TCG_TARGET_REG_BITS == 64 || r < 4)) { - tcg_out_modrm(s, OPC_GRP3_Eb | P_REXB_RM, EXT3_TESTi, r); - tcg_out8(s, i); - } else { - tcg_out_modrm(s, OPC_GRP3_Ev, EXT3_TESTi, r); - tcg_out32(s, i); - } -} - typedef struct { TCGReg base; int index; @@ -XXX,XX +XXX,XX @@ static TCGLabelQemuLdst *prepare_host_addr(TCGContext *s, HostAddress *h, tcg_out_ld(s, TCG_TYPE_PTR, TCG_REG_L0, TCG_REG_L0, offsetof(CPUTLBEntry, addend)); } else if (a_mask) { - ldst = new_ldst_label(s); + int jcc; + ldst = new_ldst_label(s); ldst->is_ld = is_ld; ldst->oi = oi; ldst->addrlo_reg = addrlo; ldst->addrhi_reg = addrhi; - tcg_out_testi(s, addrlo, a_mask); /* jne slow_path */ - tcg_out_opc(s, OPC_JCC_long + JCC_JNE, 0, 0, 0); + jcc = tcg_out_cmp(s, TCG_COND_TSTNE, addrlo, a_mask, true, false); + tcg_out_opc(s, OPC_JCC_long + jcc, 0, 0, 0); ldst->label_ptr[0] = s->code_ptr; s->code_ptr += 4; } @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_ld_direct(TCGContext *s, TCGReg datalo, TCGReg datahi, } else { TCGLabel *l1 = gen_new_label(); TCGLabel *l2 = gen_new_label(); + int jcc; - tcg_out_testi(s, h.base, 15); - tcg_out_jxx(s, JCC_JNE, l1, true); + jcc = tcg_out_cmp(s, TCG_COND_TSTNE, h.base, 15, true, false); + tcg_out_jxx(s, jcc, l1, true); tcg_out_vex_modrm_sib_offset(s, OPC_MOVDQA_VxWx + h.seg, TCG_TMP_VEC, 0, @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_st_direct(TCGContext *s, TCGReg datalo, TCGReg datahi, } else { TCGLabel *l1 = gen_new_label(); TCGLabel *l2 = gen_new_label(); + int jcc; - tcg_out_testi(s, h.base, 15); - tcg_out_jxx(s, JCC_JNE, l1, true); + jcc = tcg_out_cmp(s, TCG_COND_TSTNE, h.base, 15, true, false); + tcg_out_jxx(s, jcc, l1, true); tcg_out_vex_modrm_sib_offset(s, OPC_MOVDQA_WxVx + h.seg, TCG_TMP_VEC, 0, -- 2.34.1 Use "test x,x" when the bit is one of the 4 sign bits. Use "bt imm,x" otherwise. Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/i386/tcg-target-con-set.h | 6 ++-- tcg/i386/tcg-target-con-str.h | 1 + tcg/i386/tcg-target.c.inc | 54 +++++++++++++++++++++++++++++++---- 3 files changed, 53 insertions(+), 8 deletions(-) diff --git a/tcg/i386/tcg-target-con-set.h b/tcg/i386/tcg-target-con-set.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/i386/tcg-target-con-set.h +++ b/tcg/i386/tcg-target-con-set.h @@ -XXX,XX +XXX,XX @@ C_O0_I2(L, L) C_O0_I2(qi, r) C_O0_I2(re, r) C_O0_I2(ri, r) -C_O0_I2(r, re) +C_O0_I2(r, reT) C_O0_I2(s, L) C_O0_I2(x, r) C_O0_I3(L, L, L) @@ -XXX,XX +XXX,XX @@ C_O1_I1(r, r) C_O1_I1(x, r) C_O1_I1(x, x) C_O1_I2(q, 0, qi) -C_O1_I2(q, r, re) +C_O1_I2(q, r, reT) C_O1_I2(r, 0, ci) C_O1_I2(r, 0, r) C_O1_I2(r, 0, re) @@ -XXX,XX +XXX,XX @@ C_N1_I2(r, r, r) C_N1_I2(r, r, rW) C_O1_I3(x, 0, x, x) C_O1_I3(x, x, x, x) -C_O1_I4(r, r, re, r, 0) +C_O1_I4(r, r, reT, r, 0) C_O1_I4(r, r, r, ri, ri) C_O2_I1(r, r, L) C_O2_I2(a, d, a, r) diff --git a/tcg/i386/tcg-target-con-str.h b/tcg/i386/tcg-target-con-str.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/i386/tcg-target-con-str.h +++ b/tcg/i386/tcg-target-con-str.h @@ -XXX,XX +XXX,XX @@ REGS('s', ALL_BYTEL_REGS & ~SOFTMMU_RESERVE_REGS) /* qemu_st8_i32 data */ */ CONST('e', TCG_CT_CONST_S32) CONST('I', TCG_CT_CONST_I32) +CONST('T', TCG_CT_CONST_TST) CONST('W', TCG_CT_CONST_WSZ) CONST('Z', TCG_CT_CONST_U32) diff --git a/tcg/i386/tcg-target.c.inc b/tcg/i386/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/i386/tcg-target.c.inc +++ b/tcg/i386/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ static TCGReg tcg_target_call_oarg_reg(TCGCallReturnKind kind, int slot) #define TCG_CT_CONST_U32 0x200 #define TCG_CT_CONST_I32 0x400 #define TCG_CT_CONST_WSZ 0x800 +#define TCG_CT_CONST_TST 0x1000 /* Registers used with L constraint, which are the first argument registers on x86_64, and two random call clobbered registers on @@ -XXX,XX +XXX,XX @@ static bool tcg_target_const_match(int64_t val, int ct, return 1; } if (type == TCG_TYPE_I32) { - if (ct & (TCG_CT_CONST_S32 | TCG_CT_CONST_U32 | TCG_CT_CONST_I32)) { + if (ct & (TCG_CT_CONST_S32 | TCG_CT_CONST_U32 | + TCG_CT_CONST_I32 | TCG_CT_CONST_TST)) { return 1; } } else { @@ -XXX,XX +XXX,XX @@ static bool tcg_target_const_match(int64_t val, int ct, if ((ct & TCG_CT_CONST_I32) && ~val == (int32_t)~val) { return 1; } + /* + * This will be used in combination with TCG_CT_CONST_S32, + * so "normal" TESTQ is already matched. Also accept: + * TESTQ -> TESTL (uint32_t) + * TESTQ -> BT (is_power_of_2) + */ + if ((ct & TCG_CT_CONST_TST) + && is_tst_cond(cond) + && (val == (uint32_t)val || is_power_of_2(val))) { + return 1; + } } if ((ct & TCG_CT_CONST_WSZ) && val == (type == TCG_TYPE_I32 ? 32 : 64)) { return 1; @@ -XXX,XX +XXX,XX @@ static bool tcg_target_const_match(int64_t val, int ct, #define OPC_SHLX (0xf7 | P_EXT38 | P_DATA16) #define OPC_SHRX (0xf7 | P_EXT38 | P_SIMDF2) #define OPC_SHRD_Ib (0xac | P_EXT) +#define OPC_TESTB (0x84) #define OPC_TESTL (0x85) #define OPC_TZCNT (0xbc | P_EXT | P_SIMDF3) #define OPC_UD2 (0x0b | P_EXT) @@ -XXX,XX +XXX,XX @@ static bool tcg_target_const_match(int64_t val, int ct, #define OPC_GRP3_Ev (0xf7) #define OPC_GRP5 (0xff) #define OPC_GRP14 (0x73 | P_EXT | P_DATA16) +#define OPC_GRPBT (0xba | P_EXT) + +#define OPC_GRPBT_BT 4 +#define OPC_GRPBT_BTS 5 +#define OPC_GRPBT_BTR 6 +#define OPC_GRPBT_BTC 7 /* Group 1 opcode extensions for 0x80-0x83. These are also used as modifiers for OPC_ARITH. */ @@ -XXX,XX +XXX,XX @@ static void tcg_out_jxx(TCGContext *s, int opc, TCGLabel *l, bool small) static int tcg_out_cmp(TCGContext *s, TCGCond cond, TCGArg arg1, TCGArg arg2, int const_arg2, int rexw) { - int jz; + int jz, js; if (!is_tst_cond(cond)) { if (!const_arg2) { @@ -XXX,XX +XXX,XX @@ static int tcg_out_cmp(TCGContext *s, TCGCond cond, TCGArg arg1, } jz = tcg_cond_to_jcc[cond]; + js = (cond == TCG_COND_TSTNE ? JCC_JS : JCC_JNS); if (!const_arg2) { tcg_out_modrm(s, OPC_TESTL + rexw, arg1, arg2); @@ -XXX,XX +XXX,XX @@ static int tcg_out_cmp(TCGContext *s, TCGCond cond, TCGArg arg1, } if (arg2 <= 0xff && (TCG_TARGET_REG_BITS == 64 || arg1 < 4)) { + if (arg2 == 0x80) { + tcg_out_modrm(s, OPC_TESTB | P_REXB_R, arg1, arg1); + return js; + } tcg_out_modrm(s, OPC_GRP3_Eb | P_REXB_RM, EXT3_TESTi, arg1); tcg_out8(s, arg2); return jz; } if ((arg2 & ~0xff00) == 0 && arg1 < 4) { + if (arg2 == 0x8000) { + tcg_out_modrm(s, OPC_TESTB, arg1 + 4, arg1 + 4); + return js; + } tcg_out_modrm(s, OPC_GRP3_Eb, EXT3_TESTi, arg1 + 4); tcg_out8(s, arg2 >> 8); return jz; } + if (is_power_of_2(rexw ? arg2 : (uint32_t)arg2)) { + int jc = (cond == TCG_COND_TSTNE ? JCC_JB : JCC_JAE); + int sh = ctz64(arg2); + + rexw = (sh & 32 ? P_REXW : 0); + if ((sh & 31) == 31) { + tcg_out_modrm(s, OPC_TESTL | rexw, arg1, arg1); + return js; + } else { + tcg_out_modrm(s, OPC_GRPBT | rexw, OPC_GRPBT_BT, arg1); + tcg_out8(s, sh); + return jc; + } + } + if (rexw) { if (arg2 == (uint32_t)arg2) { rexw = 0; @@ -XXX,XX +XXX,XX @@ static TCGConstraintSetIndex tcg_target_op_def(TCGOpcode op) case INDEX_op_brcond_i32: case INDEX_op_brcond_i64: - return C_O0_I2(r, re); + return C_O0_I2(r, reT); case INDEX_op_bswap16_i32: case INDEX_op_bswap16_i64: @@ -XXX,XX +XXX,XX @@ static TCGConstraintSetIndex tcg_target_op_def(TCGOpcode op) case INDEX_op_setcond_i64: case INDEX_op_negsetcond_i32: case INDEX_op_negsetcond_i64: - return C_O1_I2(q, r, re); + return C_O1_I2(q, r, reT); case INDEX_op_movcond_i32: case INDEX_op_movcond_i64: - return C_O1_I4(r, r, re, r, 0); + return C_O1_I4(r, r, reT, r, 0); case INDEX_op_div2_i32: case INDEX_op_div2_i64: -- 2.34.1 From: Paolo Bonzini <pbonzini@redhat.com> Just like when testing against the sign bits, TEST r,r can be used when the immediate is 0xff, 0xff00, 0xffff, 0xffffffff. Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/i386/tcg-target.c.inc | 17 +++++++++++++++++ 1 file changed, 17 insertions(+) diff --git a/tcg/i386/tcg-target.c.inc b/tcg/i386/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/i386/tcg-target.c.inc +++ b/tcg/i386/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ static int tcg_out_cmp(TCGContext *s, TCGCond cond, TCGArg arg1, tcg_out_modrm(s, OPC_TESTB | P_REXB_R, arg1, arg1); return js; } + if (arg2 == 0xff) { + tcg_out_modrm(s, OPC_TESTB | P_REXB_R, arg1, arg1); + return jz; + } tcg_out_modrm(s, OPC_GRP3_Eb | P_REXB_RM, EXT3_TESTi, arg1); tcg_out8(s, arg2); return jz; @@ -XXX,XX +XXX,XX @@ static int tcg_out_cmp(TCGContext *s, TCGCond cond, TCGArg arg1, tcg_out_modrm(s, OPC_TESTB, arg1 + 4, arg1 + 4); return js; } + if (arg2 == 0xff00) { + tcg_out_modrm(s, OPC_TESTB, arg1 + 4, arg1 + 4); + return jz; + } tcg_out_modrm(s, OPC_GRP3_Eb, EXT3_TESTi, arg1 + 4); tcg_out8(s, arg2 >> 8); return jz; } + if (arg2 == 0xffff) { + tcg_out_modrm(s, OPC_TESTL | P_DATA16, arg1, arg1); + return jz; + } + if (arg2 == 0xffffffffu) { + tcg_out_modrm(s, OPC_TESTL, arg1, arg1); + return jz; + } + if (is_power_of_2(rexw ? arg2 : (uint32_t)arg2)) { int jc = (cond == TCG_COND_TSTNE ? JCC_JB : JCC_JAE); int sh = ctz64(arg2); -- 2.34.1 Use a non-zero value here (an illegal encoding) as a better condition than is_unsigned_cond for when MOVR/BPR is usable. Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/sparc64/tcg-target.c.inc | 25 ++++++++++++++----------- 1 file changed, 14 insertions(+), 11 deletions(-) diff --git a/tcg/sparc64/tcg-target.c.inc b/tcg/sparc64/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/sparc64/tcg-target.c.inc +++ b/tcg/sparc64/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ static const uint8_t tcg_cond_to_bcond[] = { [TCG_COND_GTU] = COND_GU, }; -static const uint8_t tcg_cond_to_rcond[] = { +static const uint8_t tcg_cond_to_rcond[16] = { [TCG_COND_EQ] = RCOND_Z, [TCG_COND_NE] = RCOND_NZ, [TCG_COND_LT] = RCOND_LZ, @@ -XXX,XX +XXX,XX @@ static void tcg_out_brcond_i64(TCGContext *s, TCGCond cond, TCGReg arg1, int32_t arg2, int const_arg2, TCGLabel *l) { /* For 64-bit signed comparisons vs zero, we can avoid the compare. */ - if (arg2 == 0 && !is_unsigned_cond(cond)) { + int rcond = tcg_cond_to_rcond[cond]; + if (arg2 == 0 && rcond) { int off16 = 0; if (l->has_value) { @@ -XXX,XX +XXX,XX @@ static void tcg_out_brcond_i64(TCGContext *s, TCGCond cond, TCGReg arg1, tcg_out_reloc(s, s->code_ptr, R_SPARC_WDISP16, l, 0); } tcg_out32(s, INSN_OP(0) | INSN_OP2(3) | BPR_PT | INSN_RS1(arg1) - | INSN_COND(tcg_cond_to_rcond[cond]) | off16); + | INSN_COND(rcond) | off16); } else { tcg_out_cmp(s, arg1, arg2, const_arg2); tcg_out_bpcc(s, tcg_cond_to_bcond[cond], BPCC_XCC | BPCC_PT, l); @@ -XXX,XX +XXX,XX @@ static void tcg_out_brcond_i64(TCGContext *s, TCGCond cond, TCGReg arg1, tcg_out_nop(s); } -static void tcg_out_movr(TCGContext *s, TCGCond cond, TCGReg ret, TCGReg c1, +static void tcg_out_movr(TCGContext *s, int rcond, TCGReg ret, TCGReg c1, int32_t v1, int v1const) { - tcg_out32(s, ARITH_MOVR | INSN_RD(ret) | INSN_RS1(c1) - | (tcg_cond_to_rcond[cond] << 10) + tcg_out32(s, ARITH_MOVR | INSN_RD(ret) | INSN_RS1(c1) | (rcond << 10) | (v1const ? INSN_IMM10(v1) : INSN_RS2(v1))); } @@ -XXX,XX +XXX,XX @@ static void tcg_out_movcond_i64(TCGContext *s, TCGCond cond, TCGReg ret, /* For 64-bit signed comparisons vs zero, we can avoid the compare. Note that the immediate range is one bit smaller, so we must check for that as well. */ - if (c2 == 0 && !is_unsigned_cond(cond) - && (!v1const || check_fit_i32(v1, 10))) { - tcg_out_movr(s, cond, ret, c1, v1, v1const); + int rcond = tcg_cond_to_rcond[cond]; + if (c2 == 0 && rcond && (!v1const || check_fit_i32(v1, 10))) { + tcg_out_movr(s, rcond, ret, c1, v1, v1const); } else { tcg_out_cmp(s, c1, c2, c2const); tcg_out_movcc(s, cond, MOVCC_XCC, ret, v1, v1const); @@ -XXX,XX +XXX,XX @@ static void tcg_out_setcond_i32(TCGContext *s, TCGCond cond, TCGReg ret, static void tcg_out_setcond_i64(TCGContext *s, TCGCond cond, TCGReg ret, TCGReg c1, int32_t c2, int c2const, bool neg) { + int rcond; + if (use_vis3_instructions && !neg) { switch (cond) { case TCG_COND_NE: @@ -XXX,XX +XXX,XX @@ static void tcg_out_setcond_i64(TCGContext *s, TCGCond cond, TCGReg ret, /* For 64-bit signed comparisons vs zero, we can avoid the compare if the input does not overlap the output. */ - if (c2 == 0 && !is_unsigned_cond(cond) && c1 != ret) { + rcond = tcg_cond_to_rcond[cond]; + if (c2 == 0 && rcond && c1 != ret) { tcg_out_movi_s13(s, ret, 0); - tcg_out_movr(s, cond, ret, c1, neg ? -1 : 1, 1); + tcg_out_movr(s, rcond, ret, c1, neg ? -1 : 1, 1); } else { tcg_out_cmp(s, c1, c2, c2const); tcg_out_movi_s13(s, ret, 0); -- 2.34.1 Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/sparc64/tcg-target.c.inc | 21 +++++++++++---------- 1 file changed, 11 insertions(+), 10 deletions(-) diff --git a/tcg/sparc64/tcg-target.c.inc b/tcg/sparc64/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/sparc64/tcg-target.c.inc +++ b/tcg/sparc64/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ static void tcg_out_bpcc(TCGContext *s, int scond, int flags, TCGLabel *l) tcg_out_bpcc0(s, scond, flags, off19); } -static void tcg_out_cmp(TCGContext *s, TCGReg c1, int32_t c2, int c2const) +static void tcg_out_cmp(TCGContext *s, TCGCond cond, + TCGReg c1, int32_t c2, int c2const) { tcg_out_arithc(s, TCG_REG_G0, c1, c2, c2const, ARITH_SUBCC); } @@ -XXX,XX +XXX,XX @@ static void tcg_out_cmp(TCGContext *s, TCGReg c1, int32_t c2, int c2const) static void tcg_out_brcond_i32(TCGContext *s, TCGCond cond, TCGReg arg1, int32_t arg2, int const_arg2, TCGLabel *l) { - tcg_out_cmp(s, arg1, arg2, const_arg2); + tcg_out_cmp(s, cond, arg1, arg2, const_arg2); tcg_out_bpcc(s, tcg_cond_to_bcond[cond], BPCC_ICC | BPCC_PT, l); tcg_out_nop(s); } @@ -XXX,XX +XXX,XX @@ static void tcg_out_movcond_i32(TCGContext *s, TCGCond cond, TCGReg ret, TCGReg c1, int32_t c2, int c2const, int32_t v1, int v1const) { - tcg_out_cmp(s, c1, c2, c2const); + tcg_out_cmp(s, cond, c1, c2, c2const); tcg_out_movcc(s, cond, MOVCC_ICC, ret, v1, v1const); } @@ -XXX,XX +XXX,XX @@ static void tcg_out_brcond_i64(TCGContext *s, TCGCond cond, TCGReg arg1, tcg_out32(s, INSN_OP(0) | INSN_OP2(3) | BPR_PT | INSN_RS1(arg1) | INSN_COND(rcond) | off16); } else { - tcg_out_cmp(s, arg1, arg2, const_arg2); + tcg_out_cmp(s, cond, arg1, arg2, const_arg2); tcg_out_bpcc(s, tcg_cond_to_bcond[cond], BPCC_XCC | BPCC_PT, l); } tcg_out_nop(s); @@ -XXX,XX +XXX,XX @@ static void tcg_out_movcond_i64(TCGContext *s, TCGCond cond, TCGReg ret, if (c2 == 0 && rcond && (!v1const || check_fit_i32(v1, 10))) { tcg_out_movr(s, rcond, ret, c1, v1, v1const); } else { - tcg_out_cmp(s, c1, c2, c2const); + tcg_out_cmp(s, cond, c1, c2, c2const); tcg_out_movcc(s, cond, MOVCC_XCC, ret, v1, v1const); } } @@ -XXX,XX +XXX,XX @@ static void tcg_out_setcond_i32(TCGContext *s, TCGCond cond, TCGReg ret, /* FALLTHRU */ default: - tcg_out_cmp(s, c1, c2, c2const); + tcg_out_cmp(s, cond, c1, c2, c2const); tcg_out_movi_s13(s, ret, 0); tcg_out_movcc(s, cond, MOVCC_ICC, ret, neg ? -1 : 1, 1); return; } - tcg_out_cmp(s, c1, c2, c2const); + tcg_out_cmp(s, cond, c1, c2, c2const); if (cond == TCG_COND_LTU) { if (neg) { /* 0 - 0 - C = -C = (C ? -1 : 0) */ @@ -XXX,XX +XXX,XX @@ static void tcg_out_setcond_i64(TCGContext *s, TCGCond cond, TCGReg ret, c2 = c1, c2const = 0, c1 = TCG_REG_G0; /* FALLTHRU */ case TCG_COND_LTU: - tcg_out_cmp(s, c1, c2, c2const); + tcg_out_cmp(s, cond, c1, c2, c2const); tcg_out_arith(s, ret, TCG_REG_G0, TCG_REG_G0, ARITH_ADDXC); return; default: @@ -XXX,XX +XXX,XX @@ static void tcg_out_setcond_i64(TCGContext *s, TCGCond cond, TCGReg ret, tcg_out_movi_s13(s, ret, 0); tcg_out_movr(s, rcond, ret, c1, neg ? -1 : 1, 1); } else { - tcg_out_cmp(s, c1, c2, c2const); + tcg_out_cmp(s, cond, c1, c2, c2const); tcg_out_movi_s13(s, ret, 0); tcg_out_movcc(s, cond, MOVCC_XCC, ret, neg ? -1 : 1, 1); } @@ -XXX,XX +XXX,XX @@ static TCGLabelQemuLdst *prepare_host_addr(TCGContext *s, HostAddress *h, tcg_out_movi_s32(s, TCG_REG_T3, compare_mask); tcg_out_arith(s, TCG_REG_T3, addr_reg, TCG_REG_T3, ARITH_AND); } - tcg_out_cmp(s, TCG_REG_T2, TCG_REG_T3, 0); + tcg_out_cmp(s, TCG_COND_NE, TCG_REG_T2, TCG_REG_T3, 0); ldst = new_ldst_label(s); ldst->is_ld = is_ld; -- 2.34.1 Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/sparc64/tcg-target.h | 2 +- tcg/sparc64/tcg-target.c.inc | 16 ++++++++++++++-- 2 files changed, 15 insertions(+), 3 deletions(-) diff --git a/tcg/sparc64/tcg-target.h b/tcg/sparc64/tcg-target.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/sparc64/tcg-target.h +++ b/tcg/sparc64/tcg-target.h @@ -XXX,XX +XXX,XX @@ extern bool use_vis3_instructions; #define TCG_TARGET_HAS_qemu_ldst_i128 0 -#define TCG_TARGET_HAS_tst 0 +#define TCG_TARGET_HAS_tst 1 #define TCG_AREG0 TCG_REG_I0 diff --git a/tcg/sparc64/tcg-target.c.inc b/tcg/sparc64/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/sparc64/tcg-target.c.inc +++ b/tcg/sparc64/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ static void tcg_out_div32(TCGContext *s, TCGReg rd, TCGReg rs1, uns ? ARITH_UDIV : ARITH_SDIV); } -static const uint8_t tcg_cond_to_bcond[] = { +static const uint8_t tcg_cond_to_bcond[16] = { [TCG_COND_EQ] = COND_E, [TCG_COND_NE] = COND_NE, + [TCG_COND_TSTEQ] = COND_E, + [TCG_COND_TSTNE] = COND_NE, [TCG_COND_LT] = COND_L, [TCG_COND_GE] = COND_GE, [TCG_COND_LE] = COND_LE, @@ -XXX,XX +XXX,XX @@ static void tcg_out_bpcc(TCGContext *s, int scond, int flags, TCGLabel *l) static void tcg_out_cmp(TCGContext *s, TCGCond cond, TCGReg c1, int32_t c2, int c2const) { - tcg_out_arithc(s, TCG_REG_G0, c1, c2, c2const, ARITH_SUBCC); + tcg_out_arithc(s, TCG_REG_G0, c1, c2, c2const, + is_tst_cond(cond) ? ARITH_ANDCC : ARITH_SUBCC); } static void tcg_out_brcond_i32(TCGContext *s, TCGCond cond, TCGReg arg1, @@ -XXX,XX +XXX,XX @@ static void tcg_out_setcond_i32(TCGContext *s, TCGCond cond, TCGReg ret, cond = (cond == TCG_COND_EQ ? TCG_COND_GEU : TCG_COND_LTU); break; + case TCG_COND_TSTEQ: + case TCG_COND_TSTNE: + /* Transform to inequality vs zero. */ + tcg_out_arithc(s, TCG_REG_T1, c1, c2, c2const, ARITH_AND); + c1 = TCG_REG_G0; + c2 = TCG_REG_T1, c2const = 0; + cond = (cond == TCG_COND_TSTEQ ? TCG_COND_GEU : TCG_COND_LTU); + break; + case TCG_COND_GTU: case TCG_COND_LEU: /* If we don't need to load a constant into a register, we can -- 2.34.1 Rename the current tcg_out_bc function to tcg_out_bc_lab, and create a new function that takes an integer displacement + link. Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/ppc/tcg-target.c.inc | 28 +++++++++++++++++----------- 1 file changed, 17 insertions(+), 11 deletions(-) diff --git a/tcg/ppc/tcg-target.c.inc b/tcg/ppc/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/ppc/tcg-target.c.inc +++ b/tcg/ppc/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ static void tcg_out_setcond(TCGContext *s, TCGType type, TCGCond cond, } } -static void tcg_out_bc(TCGContext *s, int bc, TCGLabel *l) +static void tcg_out_bc(TCGContext *s, TCGCond cond, int bd) { + tcg_out32(s, tcg_to_bc[cond] | bd); +} + +static void tcg_out_bc_lab(TCGContext *s, TCGCond cond, TCGLabel *l) +{ + int bd = 0; if (l->has_value) { - bc |= reloc_pc14_val(tcg_splitwx_to_rx(s->code_ptr), l->u.value_ptr); + bd = reloc_pc14_val(tcg_splitwx_to_rx(s->code_ptr), l->u.value_ptr); } else { tcg_out_reloc(s, s->code_ptr, R_PPC_REL14, l, 0); } - tcg_out32(s, bc); + tcg_out_bc(s, cond, bd); } static void tcg_out_brcond(TCGContext *s, TCGCond cond, @@ -XXX,XX +XXX,XX @@ static void tcg_out_brcond(TCGContext *s, TCGCond cond, TCGLabel *l, TCGType type) { tcg_out_cmp(s, cond, arg1, arg2, const_arg2, 7, type); - tcg_out_bc(s, tcg_to_bc[cond], l); + tcg_out_bc_lab(s, cond, l); } static void tcg_out_movcond(TCGContext *s, TCGType type, TCGCond cond, @@ -XXX,XX +XXX,XX @@ static void tcg_out_movcond(TCGContext *s, TCGType type, TCGCond cond, } } /* Branch forward over one insn */ - tcg_out32(s, tcg_to_bc[cond] | 8); + tcg_out_bc(s, cond, 8); if (v2 == 0) { tcg_out_movi(s, type, dest, 0); } else { @@ -XXX,XX +XXX,XX @@ static void tcg_out_cntxz(TCGContext *s, TCGType type, uint32_t opc, tcg_out32(s, opc | RA(TCG_REG_R0) | RS(a1)); tcg_out32(s, tcg_to_isel[TCG_COND_EQ] | TAB(a0, a2, TCG_REG_R0)); } else if (!const_a2 && a0 == a2) { - tcg_out32(s, tcg_to_bc[TCG_COND_EQ] | 8); + tcg_out_bc(s, TCG_COND_EQ, 8); tcg_out32(s, opc | RA(a0) | RS(a1)); } else { tcg_out32(s, opc | RA(a0) | RS(a1)); - tcg_out32(s, tcg_to_bc[TCG_COND_NE] | 8); + tcg_out_bc(s, TCG_COND_NE, 8); if (const_a2) { tcg_out_movi(s, type, a0, 0); } else { @@ -XXX,XX +XXX,XX @@ static void tcg_out_setcond2(TCGContext *s, const TCGArg *args, tcg_out_rlw(s, RLWINM, args[0], TCG_REG_R0, 31, 31, 31); } -static void tcg_out_brcond2 (TCGContext *s, const TCGArg *args, - const int *const_args) +static void tcg_out_brcond2(TCGContext *s, const TCGArg *args, + const int *const_args) { tcg_out_cmp2(s, args, const_args); - tcg_out_bc(s, BC | BI(7, CR_EQ) | BO_COND_TRUE, arg_label(args[5])); + tcg_out_bc_lab(s, TCG_COND_EQ, arg_label(args[5])); } static void tcg_out_mb(TCGContext *s, TCGArg a0) @@ -XXX,XX +XXX,XX @@ static TCGLabelQemuLdst *prepare_host_addr(TCGContext *s, HostAddress *h, /* Load a pointer into the current opcode w/conditional branch-link. */ ldst->label_ptr[0] = s->code_ptr; - tcg_out32(s, BC | BI(7, CR_EQ) | BO_COND_FALSE | LK); + tcg_out_bc(s, TCG_COND_NE, LK); h->base = TCG_REG_TMP1; } else { -- 2.34.1 Using cr0 means we could choose to use rc=1 to compute the condition. Adjust the tables and tcg_out_cmp that feeds them. Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/ppc/tcg-target.c.inc | 68 ++++++++++++++++++++-------------------- 1 file changed, 34 insertions(+), 34 deletions(-) diff --git a/tcg/ppc/tcg-target.c.inc b/tcg/ppc/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/ppc/tcg-target.c.inc +++ b/tcg/ppc/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ enum { }; static const uint32_t tcg_to_bc[] = { - [TCG_COND_EQ] = BC | BI(7, CR_EQ) | BO_COND_TRUE, - [TCG_COND_NE] = BC | BI(7, CR_EQ) | BO_COND_FALSE, - [TCG_COND_LT] = BC | BI(7, CR_LT) | BO_COND_TRUE, - [TCG_COND_GE] = BC | BI(7, CR_LT) | BO_COND_FALSE, - [TCG_COND_LE] = BC | BI(7, CR_GT) | BO_COND_FALSE, - [TCG_COND_GT] = BC | BI(7, CR_GT) | BO_COND_TRUE, - [TCG_COND_LTU] = BC | BI(7, CR_LT) | BO_COND_TRUE, - [TCG_COND_GEU] = BC | BI(7, CR_LT) | BO_COND_FALSE, - [TCG_COND_LEU] = BC | BI(7, CR_GT) | BO_COND_FALSE, - [TCG_COND_GTU] = BC | BI(7, CR_GT) | BO_COND_TRUE, + [TCG_COND_EQ] = BC | BI(0, CR_EQ) | BO_COND_TRUE, + [TCG_COND_NE] = BC | BI(0, CR_EQ) | BO_COND_FALSE, + [TCG_COND_LT] = BC | BI(0, CR_LT) | BO_COND_TRUE, + [TCG_COND_GE] = BC | BI(0, CR_LT) | BO_COND_FALSE, + [TCG_COND_LE] = BC | BI(0, CR_GT) | BO_COND_FALSE, + [TCG_COND_GT] = BC | BI(0, CR_GT) | BO_COND_TRUE, + [TCG_COND_LTU] = BC | BI(0, CR_LT) | BO_COND_TRUE, + [TCG_COND_GEU] = BC | BI(0, CR_LT) | BO_COND_FALSE, + [TCG_COND_LEU] = BC | BI(0, CR_GT) | BO_COND_FALSE, + [TCG_COND_GTU] = BC | BI(0, CR_GT) | BO_COND_TRUE, }; /* The low bit here is set if the RA and RB fields must be inverted. */ static const uint32_t tcg_to_isel[] = { - [TCG_COND_EQ] = ISEL | BC_(7, CR_EQ), - [TCG_COND_NE] = ISEL | BC_(7, CR_EQ) | 1, - [TCG_COND_LT] = ISEL | BC_(7, CR_LT), - [TCG_COND_GE] = ISEL | BC_(7, CR_LT) | 1, - [TCG_COND_LE] = ISEL | BC_(7, CR_GT) | 1, - [TCG_COND_GT] = ISEL | BC_(7, CR_GT), - [TCG_COND_LTU] = ISEL | BC_(7, CR_LT), - [TCG_COND_GEU] = ISEL | BC_(7, CR_LT) | 1, - [TCG_COND_LEU] = ISEL | BC_(7, CR_GT) | 1, - [TCG_COND_GTU] = ISEL | BC_(7, CR_GT), + [TCG_COND_EQ] = ISEL | BC_(0, CR_EQ), + [TCG_COND_NE] = ISEL | BC_(0, CR_EQ) | 1, + [TCG_COND_LT] = ISEL | BC_(0, CR_LT), + [TCG_COND_GE] = ISEL | BC_(0, CR_LT) | 1, + [TCG_COND_LE] = ISEL | BC_(0, CR_GT) | 1, + [TCG_COND_GT] = ISEL | BC_(0, CR_GT), + [TCG_COND_LTU] = ISEL | BC_(0, CR_LT), + [TCG_COND_GEU] = ISEL | BC_(0, CR_LT) | 1, + [TCG_COND_LEU] = ISEL | BC_(0, CR_GT) | 1, + [TCG_COND_GTU] = ISEL | BC_(0, CR_GT), }; static bool patch_reloc(tcg_insn_unit *code_ptr, int type, @@ -XXX,XX +XXX,XX @@ static void tcg_out_setcond(TCGContext *s, TCGType type, TCGCond cond, if (have_isa_3_10) { tcg_insn_unit bi, opc; - tcg_out_cmp(s, cond, arg1, arg2, const_arg2, 7, type); + tcg_out_cmp(s, cond, arg1, arg2, const_arg2, 0, type); /* Re-use tcg_to_bc for BI and BO_COND_{TRUE,FALSE}. */ bi = tcg_to_bc[cond] & (0x1f << 16); @@ -XXX,XX +XXX,XX @@ static void tcg_out_setcond(TCGContext *s, TCGType type, TCGCond cond, if (have_isel) { int isel, tab; - tcg_out_cmp(s, cond, arg1, arg2, const_arg2, 7, type); + tcg_out_cmp(s, cond, arg1, arg2, const_arg2, 0, type); isel = tcg_to_isel[cond]; @@ -XXX,XX +XXX,XX @@ static void tcg_out_brcond(TCGContext *s, TCGCond cond, TCGArg arg1, TCGArg arg2, int const_arg2, TCGLabel *l, TCGType type) { - tcg_out_cmp(s, cond, arg1, arg2, const_arg2, 7, type); + tcg_out_cmp(s, cond, arg1, arg2, const_arg2, 0, type); tcg_out_bc_lab(s, cond, l); } @@ -XXX,XX +XXX,XX @@ static void tcg_out_movcond(TCGContext *s, TCGType type, TCGCond cond, return; } - tcg_out_cmp(s, cond, c1, c2, const_c2, 7, type); + tcg_out_cmp(s, cond, c1, c2, const_c2, 0, type); if (have_isel) { int isel = tcg_to_isel[cond]; @@ -XXX,XX +XXX,XX @@ static void tcg_out_cntxz(TCGContext *s, TCGType type, uint32_t opc, if (const_a2 && a2 == (type == TCG_TYPE_I32 ? 32 : 64)) { tcg_out32(s, opc | RA(a0) | RS(a1)); } else { - tcg_out_cmp(s, TCG_COND_EQ, a1, 0, 1, 7, type); + tcg_out_cmp(s, TCG_COND_EQ, a1, 0, 1, 0, type); /* Note that the only other valid constant for a2 is 0. */ if (have_isel) { tcg_out32(s, opc | RA(TCG_REG_R0) | RS(a1)); @@ -XXX,XX +XXX,XX @@ static void tcg_out_cmp2(TCGContext *s, const TCGArg *args, do_equality: tcg_out_cmp(s, cond, al, bl, blconst, 6, TCG_TYPE_I32); tcg_out_cmp(s, cond, ah, bh, bhconst, 7, TCG_TYPE_I32); - tcg_out32(s, op | BT(7, CR_EQ) | BA(6, CR_EQ) | BB(7, CR_EQ)); + tcg_out32(s, op | BT(0, CR_EQ) | BA(6, CR_EQ) | BB(7, CR_EQ)); break; case TCG_COND_LT: @@ -XXX,XX +XXX,XX @@ static void tcg_out_cmp2(TCGContext *s, const TCGArg *args, tcg_out_cmp(s, cond, ah, bh, bhconst, 6, TCG_TYPE_I32); tcg_out_cmp(s, cond2, al, bl, blconst, 7, TCG_TYPE_I32); - tcg_out32(s, op | BT(7, CR_EQ) | BA(6, CR_EQ) | BB(7, bit2)); - tcg_out32(s, CROR | BT(7, CR_EQ) | BA(6, bit1) | BB(7, CR_EQ)); + tcg_out32(s, op | BT(0, CR_EQ) | BA(6, CR_EQ) | BB(7, bit2)); + tcg_out32(s, CROR | BT(0, CR_EQ) | BA(6, bit1) | BB(0, CR_EQ)); break; default: @@ -XXX,XX +XXX,XX @@ static void tcg_out_setcond2(TCGContext *s, const TCGArg *args, const int *const_args) { tcg_out_cmp2(s, args + 1, const_args + 1); - tcg_out32(s, MFOCRF | RT(TCG_REG_R0) | FXM(7)); - tcg_out_rlw(s, RLWINM, args[0], TCG_REG_R0, 31, 31, 31); + tcg_out32(s, MFOCRF | RT(TCG_REG_R0) | FXM(0)); + tcg_out_rlw(s, RLWINM, args[0], TCG_REG_R0, CR_EQ + 0*4 + 1, 31, 31); } static void tcg_out_brcond2(TCGContext *s, const TCGArg *args, @@ -XXX,XX +XXX,XX @@ static TCGLabelQemuLdst *prepare_host_addr(TCGContext *s, HostAddress *h, tcg_out_cmp(s, TCG_COND_EQ, addrhi, TCG_REG_TMP2, 0, 6, TCG_TYPE_I32); - /* Combine comparisons into cr7. */ - tcg_out32(s, CRAND | BT(7, CR_EQ) | BA(6, CR_EQ) | BB(7, CR_EQ)); + /* Combine comparisons into cr0. */ + tcg_out32(s, CRAND | BT(0, CR_EQ) | BA(6, CR_EQ) | BB(7, CR_EQ)); } else { - /* Full comparison into cr7. */ + /* Full comparison into cr0. */ tcg_out_cmp(s, TCG_COND_EQ, TCG_REG_R0, TCG_REG_TMP2, - 0, 7, addr_type); + 0, 0, addr_type); } /* Load a pointer into the current opcode w/conditional branch-link. */ -- 2.34.1 Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/ppc/tcg-target.c.inc | 27 ++++++++++++++++----------- 1 file changed, 16 insertions(+), 11 deletions(-) diff --git a/tcg/ppc/tcg-target.c.inc b/tcg/ppc/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/ppc/tcg-target.c.inc +++ b/tcg/ppc/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ static bool reloc_pc34(tcg_insn_unit *src_rw, const tcg_insn_unit *target) } /* test if a constant matches the constraint */ -static bool tcg_target_const_match(int64_t val, int ct, +static bool tcg_target_const_match(int64_t sval, int ct, TCGType type, TCGCond cond, int vece) { + uint64_t uval = sval; + if (ct & TCG_CT_CONST) { return 1; } - /* The only 32-bit constraint we use aside from - TCG_CT_CONST is TCG_CT_CONST_S16. */ if (type == TCG_TYPE_I32) { - val = (int32_t)val; + uval = (uint32_t)sval; + sval = (int32_t)sval; } - if ((ct & TCG_CT_CONST_S16) && val == (int16_t)val) { + if ((ct & TCG_CT_CONST_S16) && sval == (int16_t)sval) { return 1; - } else if ((ct & TCG_CT_CONST_S32) && val == (int32_t)val) { + } + if ((ct & TCG_CT_CONST_S32) && sval == (int32_t)sval) { return 1; - } else if ((ct & TCG_CT_CONST_U32) && val == (uint32_t)val) { + } + if ((ct & TCG_CT_CONST_U32) && uval == (uint32_t)uval) { return 1; - } else if ((ct & TCG_CT_CONST_ZERO) && val == 0) { + } + if ((ct & TCG_CT_CONST_ZERO) && sval == 0) { return 1; - } else if ((ct & TCG_CT_CONST_MONE) && val == -1) { + } + if ((ct & TCG_CT_CONST_MONE) && sval == -1) { return 1; - } else if ((ct & TCG_CT_CONST_WSZ) - && val == (type == TCG_TYPE_I32 ? 32 : 64)) { + } + if ((ct & TCG_CT_CONST_WSZ) && sval == (type == TCG_TYPE_I32 ? 32 : 64)) { return 1; } return 0; -- 2.34.1 Better constraint for tcg_out_cmp, based on the comparison. We can't yet remove the fallback to load constants into a scratch because of tcg_out_cmp2, but that path should not be as frequent. Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/ppc/tcg-target-con-set.h | 5 ++-- tcg/ppc/tcg-target-con-str.h | 1 + tcg/ppc/tcg-target.c.inc | 48 ++++++++++++++++++++++++++++++------ 3 files changed, 44 insertions(+), 10 deletions(-) diff --git a/tcg/ppc/tcg-target-con-set.h b/tcg/ppc/tcg-target-con-set.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/ppc/tcg-target-con-set.h +++ b/tcg/ppc/tcg-target-con-set.h @@ -XXX,XX +XXX,XX @@ */ C_O0_I1(r) C_O0_I2(r, r) -C_O0_I2(r, ri) +C_O0_I2(r, rC) C_O0_I2(v, r) C_O0_I3(r, r, r) C_O0_I3(o, m, r) @@ -XXX,XX +XXX,XX @@ C_O1_I2(r, rI, ri) C_O1_I2(r, rI, rT) C_O1_I2(r, r, r) C_O1_I2(r, r, ri) +C_O1_I2(r, r, rC) C_O1_I2(r, r, rI) C_O1_I2(r, r, rT) C_O1_I2(r, r, rU) C_O1_I2(r, r, rZW) C_O1_I2(v, v, v) C_O1_I3(v, v, v, v) -C_O1_I4(r, r, ri, rZ, rZ) +C_O1_I4(r, r, rC, rZ, rZ) C_O1_I4(r, r, r, ri, ri) C_O2_I1(r, r, r) C_N1O1_I1(o, m, r) diff --git a/tcg/ppc/tcg-target-con-str.h b/tcg/ppc/tcg-target-con-str.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/ppc/tcg-target-con-str.h +++ b/tcg/ppc/tcg-target-con-str.h @@ -XXX,XX +XXX,XX @@ REGS('v', ALL_VECTOR_REGS) * Define constraint letters for constants: * CONST(letter, TCG_CT_CONST_* bit set) */ +CONST('C', TCG_CT_CONST_CMP) CONST('I', TCG_CT_CONST_S16) CONST('M', TCG_CT_CONST_MONE) CONST('T', TCG_CT_CONST_S32) diff --git a/tcg/ppc/tcg-target.c.inc b/tcg/ppc/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/ppc/tcg-target.c.inc +++ b/tcg/ppc/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ #define SZR (TCG_TARGET_REG_BITS / 8) #define TCG_CT_CONST_S16 0x100 +#define TCG_CT_CONST_U16 0x200 #define TCG_CT_CONST_S32 0x400 #define TCG_CT_CONST_U32 0x800 #define TCG_CT_CONST_ZERO 0x1000 #define TCG_CT_CONST_MONE 0x2000 #define TCG_CT_CONST_WSZ 0x4000 +#define TCG_CT_CONST_CMP 0x8000 #define ALL_GENERAL_REGS 0xffffffffu #define ALL_VECTOR_REGS 0xffffffff00000000ull @@ -XXX,XX +XXX,XX @@ static bool tcg_target_const_match(int64_t sval, int ct, sval = (int32_t)sval; } + if (ct & TCG_CT_CONST_CMP) { + switch (cond) { + case TCG_COND_EQ: + case TCG_COND_NE: + ct |= TCG_CT_CONST_S16 | TCG_CT_CONST_U16; + break; + case TCG_COND_LT: + case TCG_COND_GE: + case TCG_COND_LE: + case TCG_COND_GT: + ct |= TCG_CT_CONST_S16; + break; + case TCG_COND_LTU: + case TCG_COND_GEU: + case TCG_COND_LEU: + case TCG_COND_GTU: + ct |= TCG_CT_CONST_U16; + break; + default: + g_assert_not_reached(); + } + } + if ((ct & TCG_CT_CONST_S16) && sval == (int16_t)sval) { return 1; } + if ((ct & TCG_CT_CONST_U16) && uval == (uint16_t)uval) { + return 1; + } if ((ct & TCG_CT_CONST_S32) && sval == (int32_t)sval) { return 1; } @@ -XXX,XX +XXX,XX @@ static void tcg_out_cmp(TCGContext *s, int cond, TCGArg arg1, TCGArg arg2, tcg_debug_assert(TCG_TARGET_REG_BITS == 64 || type == TCG_TYPE_I32); - /* Simplify the comparisons below wrt CMPI. */ + /* + * Simplify the comparisons below wrt CMPI. + * All of the tests are 16-bit, so a 32-bit sign extend always works. + */ if (type == TCG_TYPE_I32) { arg2 = (int32_t)arg2; } @@ -XXX,XX +XXX,XX @@ static TCGConstraintSetIndex tcg_target_op_def(TCGOpcode op) case INDEX_op_sar_i32: case INDEX_op_rotl_i32: case INDEX_op_rotr_i32: - case INDEX_op_setcond_i32: - case INDEX_op_negsetcond_i32: case INDEX_op_and_i64: case INDEX_op_andc_i64: case INDEX_op_shl_i64: @@ -XXX,XX +XXX,XX @@ static TCGConstraintSetIndex tcg_target_op_def(TCGOpcode op) case INDEX_op_sar_i64: case INDEX_op_rotl_i64: case INDEX_op_rotr_i64: - case INDEX_op_setcond_i64: - case INDEX_op_negsetcond_i64: return C_O1_I2(r, r, ri); case INDEX_op_mul_i32: @@ -XXX,XX +XXX,XX @@ static TCGConstraintSetIndex tcg_target_op_def(TCGOpcode op) case INDEX_op_brcond_i32: case INDEX_op_brcond_i64: - return C_O0_I2(r, ri); - + return C_O0_I2(r, rC); + case INDEX_op_setcond_i32: + case INDEX_op_setcond_i64: + case INDEX_op_negsetcond_i32: + case INDEX_op_negsetcond_i64: + return C_O1_I2(r, r, rC); case INDEX_op_movcond_i32: case INDEX_op_movcond_i64: - return C_O1_I4(r, r, ri, rZ, rZ); + return C_O1_I4(r, r, rC, rZ, rZ); + case INDEX_op_deposit_i32: case INDEX_op_deposit_i64: return C_O1_I2(r, 0, rZ); -- 2.34.1 Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/ppc/tcg-target.h | 2 +- tcg/ppc/tcg-target.c.inc | 122 ++++++++++++++++++++++++++++++++++++--- 2 files changed, 115 insertions(+), 9 deletions(-) diff --git a/tcg/ppc/tcg-target.h b/tcg/ppc/tcg-target.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/ppc/tcg-target.h +++ b/tcg/ppc/tcg-target.h @@ -XXX,XX +XXX,XX @@ typedef enum { #define TCG_TARGET_HAS_qemu_ldst_i128 \ (TCG_TARGET_REG_BITS == 64 && have_isa_2_07) -#define TCG_TARGET_HAS_tst 0 +#define TCG_TARGET_HAS_tst 1 /* * While technically Altivec could support V64, it has no 64-bit store diff --git a/tcg/ppc/tcg-target.c.inc b/tcg/ppc/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/ppc/tcg-target.c.inc +++ b/tcg/ppc/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ static bool reloc_pc34(tcg_insn_unit *src_rw, const tcg_insn_unit *target) return false; } +static bool mask_operand(uint32_t c, int *mb, int *me); +static bool mask64_operand(uint64_t c, int *mb, int *me); + /* test if a constant matches the constraint */ static bool tcg_target_const_match(int64_t sval, int ct, TCGType type, TCGCond cond, int vece) { uint64_t uval = sval; + int mb, me; if (ct & TCG_CT_CONST) { return 1; @@ -XXX,XX +XXX,XX @@ static bool tcg_target_const_match(int64_t sval, int ct, case TCG_COND_GTU: ct |= TCG_CT_CONST_U16; break; + case TCG_COND_TSTEQ: + case TCG_COND_TSTNE: + if ((uval & ~0xffff) == 0 || (uval & ~0xffff0000ull) == 0) { + return 1; + } + if (TCG_TARGET_REG_BITS == 32 || type == TCG_TYPE_I32 + ? mask_operand(uval, &mb, &me) + : mask64_operand(uval << clz64(uval), &mb, &me)) { + return 1; + } + return 0; default: g_assert_not_reached(); } @@ -XXX,XX +XXX,XX @@ enum { CR_SO }; -static const uint32_t tcg_to_bc[] = { +static const uint32_t tcg_to_bc[16] = { [TCG_COND_EQ] = BC | BI(0, CR_EQ) | BO_COND_TRUE, [TCG_COND_NE] = BC | BI(0, CR_EQ) | BO_COND_FALSE, + [TCG_COND_TSTEQ] = BC | BI(0, CR_EQ) | BO_COND_TRUE, + [TCG_COND_TSTNE] = BC | BI(0, CR_EQ) | BO_COND_FALSE, [TCG_COND_LT] = BC | BI(0, CR_LT) | BO_COND_TRUE, [TCG_COND_GE] = BC | BI(0, CR_LT) | BO_COND_FALSE, [TCG_COND_LE] = BC | BI(0, CR_GT) | BO_COND_FALSE, @@ -XXX,XX +XXX,XX @@ static const uint32_t tcg_to_bc[] = { }; /* The low bit here is set if the RA and RB fields must be inverted. */ -static const uint32_t tcg_to_isel[] = { +static const uint32_t tcg_to_isel[16] = { [TCG_COND_EQ] = ISEL | BC_(0, CR_EQ), [TCG_COND_NE] = ISEL | BC_(0, CR_EQ) | 1, + [TCG_COND_TSTEQ] = ISEL | BC_(0, CR_EQ), + [TCG_COND_TSTNE] = ISEL | BC_(0, CR_EQ) | 1, [TCG_COND_LT] = ISEL | BC_(0, CR_LT), [TCG_COND_GE] = ISEL | BC_(0, CR_LT) | 1, [TCG_COND_LE] = ISEL | BC_(0, CR_GT) | 1, @@ -XXX,XX +XXX,XX @@ static bool tcg_out_mov(TCGContext *s, TCGType type, TCGReg ret, TCGReg arg) return true; } -static inline void tcg_out_rld(TCGContext *s, int op, TCGReg ra, TCGReg rs, - int sh, int mb) +static void tcg_out_rld_rc(TCGContext *s, int op, TCGReg ra, TCGReg rs, + int sh, int mb, bool rc) { tcg_debug_assert(TCG_TARGET_REG_BITS == 64); sh = SH(sh & 0x1f) | (((sh >> 5) & 1) << 1); mb = MB64((mb >> 5) | ((mb << 1) & 0x3f)); - tcg_out32(s, op | RA(ra) | RS(rs) | sh | mb); + tcg_out32(s, op | RA(ra) | RS(rs) | sh | mb | rc); } -static inline void tcg_out_rlw(TCGContext *s, int op, TCGReg ra, TCGReg rs, - int sh, int mb, int me) +static void tcg_out_rld(TCGContext *s, int op, TCGReg ra, TCGReg rs, + int sh, int mb) { - tcg_out32(s, op | RA(ra) | RS(rs) | SH(sh) | MB(mb) | ME(me)); + tcg_out_rld_rc(s, op, ra, rs, sh, mb, false); +} + +static void tcg_out_rlw_rc(TCGContext *s, int op, TCGReg ra, TCGReg rs, + int sh, int mb, int me, bool rc) +{ + tcg_out32(s, op | RA(ra) | RS(rs) | SH(sh) | MB(mb) | ME(me) | rc); +} + +static void tcg_out_rlw(TCGContext *s, int op, TCGReg ra, TCGReg rs, + int sh, int mb, int me) +{ + tcg_out_rlw_rc(s, op, ra, rs, sh, mb, me, false); } static void tcg_out_ext8s(TCGContext *s, TCGType type, TCGReg dst, TCGReg src) @@ -XXX,XX +XXX,XX @@ static inline bool tcg_out_sti(TCGContext *s, TCGType type, TCGArg val, return false; } +/* + * Set dest non-zero if and only if (arg1 & arg2) is non-zero. + * If RC, then also set RC0. + */ +static void tcg_out_test(TCGContext *s, TCGReg dest, TCGReg arg1, TCGArg arg2, + bool const_arg2, TCGType type, bool rc) +{ + int mb, me; + + if (!const_arg2) { + tcg_out32(s, AND | SAB(arg1, dest, arg2) | rc); + return; + } + + if (type == TCG_TYPE_I32) { + arg2 = (uint32_t)arg2; + } else if (arg2 == (uint32_t)arg2) { + type = TCG_TYPE_I32; + } + + if ((arg2 & ~0xffff) == 0) { + tcg_out32(s, ANDI | SAI(arg1, dest, arg2)); + return; + } + if ((arg2 & ~0xffff0000ull) == 0) { + tcg_out32(s, ANDIS | SAI(arg1, dest, arg2 >> 16)); + return; + } + if (TCG_TARGET_REG_BITS == 32 || type == TCG_TYPE_I32) { + if (mask_operand(arg2, &mb, &me)) { + tcg_out_rlw_rc(s, RLWINM, dest, arg1, 0, mb, me, rc); + return; + } + } else { + int sh = clz64(arg2); + if (mask64_operand(arg2 << sh, &mb, &me)) { + tcg_out_rld_rc(s, RLDICR, dest, arg1, sh, me, rc); + return; + } + } + /* Constraints should satisfy this. */ + g_assert_not_reached(); +} + static void tcg_out_cmp(TCGContext *s, int cond, TCGArg arg1, TCGArg arg2, int const_arg2, int cr, TCGType type) { @@ -XXX,XX +XXX,XX @@ static void tcg_out_cmp(TCGContext *s, int cond, TCGArg arg1, TCGArg arg2, imm = 0; break; + case TCG_COND_TSTEQ: + case TCG_COND_TSTNE: + tcg_debug_assert(cr == 0); + tcg_out_test(s, TCG_REG_R0, arg1, arg2, const_arg2, type, true); + return; + case TCG_COND_LT: case TCG_COND_GE: case TCG_COND_LE: @@ -XXX,XX +XXX,XX @@ static void tcg_out_setcond(TCGContext *s, TCGType type, TCGCond cond, tcg_out_setcond_ne0(s, type, arg0, arg1, neg); break; + case TCG_COND_TSTEQ: + tcg_out_test(s, TCG_REG_R0, arg1, arg2, const_arg2, type, false); + tcg_out_setcond_eq0(s, type, arg0, TCG_REG_R0, neg); + break; + + case TCG_COND_TSTNE: + tcg_out_test(s, TCG_REG_R0, arg1, arg2, const_arg2, type, false); + tcg_out_setcond_ne0(s, type, arg0, TCG_REG_R0, neg); + break; + case TCG_COND_LE: case TCG_COND_LEU: inv = true; @@ -XXX,XX +XXX,XX @@ static void tcg_out_cmp2(TCGContext *s, const TCGArg *args, tcg_out32(s, op | BT(0, CR_EQ) | BA(6, CR_EQ) | BB(7, CR_EQ)); break; + case TCG_COND_TSTEQ: + case TCG_COND_TSTNE: + if (blconst) { + tcg_out_andi32(s, TCG_REG_R0, al, bl); + } else { + tcg_out32(s, AND | SAB(al, TCG_REG_R0, bl)); + } + if (bhconst) { + tcg_out_andi32(s, TCG_REG_TMP1, ah, bh); + } else { + tcg_out32(s, AND | SAB(ah, TCG_REG_TMP1, bh)); + } + tcg_out32(s, OR | SAB(TCG_REG_R0, TCG_REG_R0, TCG_REG_TMP1) | 1); + break; + case TCG_COND_LT: case TCG_COND_LE: case TCG_COND_GT: -- 2.34.1 Signed 33-bit == signed 32-bit + unsigned 32-bit. Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/s390x/tcg-target-con-set.h | 8 ++++---- tcg/s390x/tcg-target-con-str.h | 2 +- tcg/s390x/tcg-target.c.inc | 36 +++++++++++++++++----------------- 3 files changed, 23 insertions(+), 23 deletions(-) diff --git a/tcg/s390x/tcg-target-con-set.h b/tcg/s390x/tcg-target-con-set.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/s390x/tcg-target-con-set.h +++ b/tcg/s390x/tcg-target-con-set.h @@ -XXX,XX +XXX,XX @@ C_O0_I1(r) C_O0_I2(r, r) C_O0_I2(r, ri) -C_O0_I2(r, rA) +C_O0_I2(r, rJU) C_O0_I2(v, r) C_O0_I3(o, m, r) C_O1_I1(r, r) @@ -XXX,XX +XXX,XX @@ C_O1_I2(r, 0, rI) C_O1_I2(r, 0, rJ) C_O1_I2(r, r, r) C_O1_I2(r, r, ri) -C_O1_I2(r, r, rA) +C_O1_I2(r, r, rJU) C_O1_I2(r, r, rI) C_O1_I2(r, r, rJ) C_O1_I2(r, r, rK) @@ -XXX,XX +XXX,XX @@ C_O1_I2(v, v, r) C_O1_I2(v, v, v) C_O1_I3(v, v, v, v) C_O1_I4(r, r, ri, rI, r) -C_O1_I4(r, r, rA, rI, r) +C_O1_I4(r, r, rJU, rI, r) C_O2_I1(o, m, r) C_O2_I2(o, m, 0, r) C_O2_I2(o, m, r, r) C_O2_I3(o, m, 0, 1, r) C_N1_O1_I4(r, r, 0, 1, ri, r) -C_N1_O1_I4(r, r, 0, 1, rA, r) +C_N1_O1_I4(r, r, 0, 1, rJU, r) diff --git a/tcg/s390x/tcg-target-con-str.h b/tcg/s390x/tcg-target-con-str.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/s390x/tcg-target-con-str.h +++ b/tcg/s390x/tcg-target-con-str.h @@ -XXX,XX +XXX,XX @@ REGS('o', 0xaaaa) /* odd numbered general regs */ * Define constraint letters for constants: * CONST(letter, TCG_CT_CONST_* bit set) */ -CONST('A', TCG_CT_CONST_S33) CONST('I', TCG_CT_CONST_S16) CONST('J', TCG_CT_CONST_S32) CONST('K', TCG_CT_CONST_P32) CONST('N', TCG_CT_CONST_INV) CONST('R', TCG_CT_CONST_INVRISBG) +CONST('U', TCG_CT_CONST_U32) CONST('Z', TCG_CT_CONST_ZERO) diff --git a/tcg/s390x/tcg-target.c.inc b/tcg/s390x/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/s390x/tcg-target.c.inc +++ b/tcg/s390x/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ #define TCG_CT_CONST_S16 (1 << 8) #define TCG_CT_CONST_S32 (1 << 9) -#define TCG_CT_CONST_S33 (1 << 10) +#define TCG_CT_CONST_U32 (1 << 10) #define TCG_CT_CONST_ZERO (1 << 11) #define TCG_CT_CONST_P32 (1 << 12) #define TCG_CT_CONST_INV (1 << 13) @@ -XXX,XX +XXX,XX @@ static bool tcg_target_const_match(int64_t val, int ct, TCGType type, TCGCond cond, int vece) { if (ct & TCG_CT_CONST) { - return 1; + return true; } - if (type == TCG_TYPE_I32) { val = (int32_t)val; } - /* The following are mutually exclusive. */ - if (ct & TCG_CT_CONST_S16) { - return val == (int16_t)val; - } else if (ct & TCG_CT_CONST_S32) { - return val == (int32_t)val; - } else if (ct & TCG_CT_CONST_S33) { - return val >= -0xffffffffll && val <= 0xffffffffll; - } else if (ct & TCG_CT_CONST_ZERO) { - return val == 0; + if ((ct & TCG_CT_CONST_S32) && val == (int32_t)val) { + return true; + } + if ((ct & TCG_CT_CONST_U32) && val == (uint32_t)val) { + return true; + } + if ((ct & TCG_CT_CONST_S16) && val == (int16_t)val) { + return true; + } + if ((ct & TCG_CT_CONST_ZERO) && val == 0) { + return true; } if (ct & TCG_CT_CONST_INV) { @@ -XXX,XX +XXX,XX @@ static bool tcg_target_const_match(int64_t val, int ct, if ((ct & TCG_CT_CONST_INVRISBG) && risbg_mask(~val)) { return true; } - - return 0; + return false; } /* Emit instructions according to the given instruction format. */ @@ -XXX,XX +XXX,XX @@ static TCGConstraintSetIndex tcg_target_op_def(TCGOpcode op) return C_O1_I2(r, r, ri); case INDEX_op_setcond_i64: case INDEX_op_negsetcond_i64: - return C_O1_I2(r, r, rA); + return C_O1_I2(r, r, rJU); case INDEX_op_clz_i64: return C_O1_I2(r, r, rI); @@ -XXX,XX +XXX,XX @@ static TCGConstraintSetIndex tcg_target_op_def(TCGOpcode op) case INDEX_op_brcond_i32: return C_O0_I2(r, ri); case INDEX_op_brcond_i64: - return C_O0_I2(r, rA); + return C_O0_I2(r, rJU); case INDEX_op_bswap16_i32: case INDEX_op_bswap16_i64: @@ -XXX,XX +XXX,XX @@ static TCGConstraintSetIndex tcg_target_op_def(TCGOpcode op) case INDEX_op_movcond_i32: return C_O1_I4(r, r, ri, rI, r); case INDEX_op_movcond_i64: - return C_O1_I4(r, r, rA, rI, r); + return C_O1_I4(r, r, rJU, rI, r); case INDEX_op_div2_i32: case INDEX_op_div2_i64: @@ -XXX,XX +XXX,XX @@ static TCGConstraintSetIndex tcg_target_op_def(TCGOpcode op) case INDEX_op_add2_i64: case INDEX_op_sub2_i64: - return C_N1_O1_I4(r, r, 0, 1, rA, r); + return C_N1_O1_I4(r, r, 0, 1, rJU, r); case INDEX_op_st_vec: return C_O0_I2(v, r); -- 2.34.1 Better constraint for tcg_out_cmp, based on the comparison. Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/s390x/tcg-target-con-set.h | 6 +-- tcg/s390x/tcg-target-con-str.h | 1 + tcg/s390x/tcg-target.c.inc | 72 +++++++++++++++++++++++++--------- 3 files changed, 58 insertions(+), 21 deletions(-) diff --git a/tcg/s390x/tcg-target-con-set.h b/tcg/s390x/tcg-target-con-set.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/s390x/tcg-target-con-set.h +++ b/tcg/s390x/tcg-target-con-set.h @@ -XXX,XX +XXX,XX @@ C_O0_I1(r) C_O0_I2(r, r) C_O0_I2(r, ri) -C_O0_I2(r, rJU) +C_O0_I2(r, rC) C_O0_I2(v, r) C_O0_I3(o, m, r) C_O1_I1(r, r) @@ -XXX,XX +XXX,XX @@ C_O1_I2(r, 0, rI) C_O1_I2(r, 0, rJ) C_O1_I2(r, r, r) C_O1_I2(r, r, ri) -C_O1_I2(r, r, rJU) +C_O1_I2(r, r, rC) C_O1_I2(r, r, rI) C_O1_I2(r, r, rJ) C_O1_I2(r, r, rK) @@ -XXX,XX +XXX,XX @@ C_O1_I2(v, v, r) C_O1_I2(v, v, v) C_O1_I3(v, v, v, v) C_O1_I4(r, r, ri, rI, r) -C_O1_I4(r, r, rJU, rI, r) +C_O1_I4(r, r, rC, rI, r) C_O2_I1(o, m, r) C_O2_I2(o, m, 0, r) C_O2_I2(o, m, r, r) diff --git a/tcg/s390x/tcg-target-con-str.h b/tcg/s390x/tcg-target-con-str.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/s390x/tcg-target-con-str.h +++ b/tcg/s390x/tcg-target-con-str.h @@ -XXX,XX +XXX,XX @@ REGS('o', 0xaaaa) /* odd numbered general regs */ * Define constraint letters for constants: * CONST(letter, TCG_CT_CONST_* bit set) */ +CONST('C', TCG_CT_CONST_CMP) CONST('I', TCG_CT_CONST_S16) CONST('J', TCG_CT_CONST_S32) CONST('K', TCG_CT_CONST_P32) diff --git a/tcg/s390x/tcg-target.c.inc b/tcg/s390x/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/s390x/tcg-target.c.inc +++ b/tcg/s390x/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ #define TCG_CT_CONST_P32 (1 << 12) #define TCG_CT_CONST_INV (1 << 13) #define TCG_CT_CONST_INVRISBG (1 << 14) +#define TCG_CT_CONST_CMP (1 << 15) #define ALL_GENERAL_REGS MAKE_64BIT_MASK(0, 16) #define ALL_VECTOR_REGS MAKE_64BIT_MASK(32, 32) @@ -XXX,XX +XXX,XX @@ static bool tcg_target_const_match(int64_t val, int ct, val = (int32_t)val; } + if (ct & TCG_CT_CONST_CMP) { + switch (cond) { + case TCG_COND_EQ: + case TCG_COND_NE: + ct |= TCG_CT_CONST_S32 | TCG_CT_CONST_U32; /* CGFI or CLGFI */ + break; + case TCG_COND_LT: + case TCG_COND_GE: + case TCG_COND_LE: + case TCG_COND_GT: + ct |= TCG_CT_CONST_S32; /* CGFI */ + break; + case TCG_COND_LTU: + case TCG_COND_GEU: + case TCG_COND_LEU: + case TCG_COND_GTU: + ct |= TCG_CT_CONST_U32; /* CLGFI */ + break; + default: + g_assert_not_reached(); + } + } + if ((ct & TCG_CT_CONST_S32) && val == (int32_t)val) { return true; } @@ -XXX,XX +XXX,XX @@ static int tgen_cmp2(TCGContext *s, TCGType type, TCGCond c, TCGReg r1, goto exit; } - /* - * Constraints are for a signed 33-bit operand, which is a - * convenient superset of this signed/unsigned test. - */ - if (c2 == (is_unsigned ? (TCGArg)(uint32_t)c2 : (TCGArg)(int32_t)c2)) { - op = (is_unsigned ? RIL_CLGFI : RIL_CGFI); - tcg_out_insn_RIL(s, op, r1, c2); - goto exit; + /* Should match TCG_CT_CONST_CMP. */ + switch (c) { + case TCG_COND_LT: + case TCG_COND_GE: + case TCG_COND_LE: + case TCG_COND_GT: + tcg_debug_assert(c2 == (int32_t)c2); + op = RIL_CGFI; + break; + case TCG_COND_EQ: + case TCG_COND_NE: + if (c2 == (int32_t)c2) { + op = RIL_CGFI; + break; + } + /* fall through */ + case TCG_COND_LTU: + case TCG_COND_GEU: + case TCG_COND_LEU: + case TCG_COND_GTU: + tcg_debug_assert(c2 == (uint32_t)c2); + op = RIL_CLGFI; + break; + default: + g_assert_not_reached(); } - - /* Load everything else into a register. */ - tcg_out_movi(s, TCG_TYPE_I64, TCG_TMP0, c2); - c2 = TCG_TMP0; - } - - if (type == TCG_TYPE_I32) { + tcg_out_insn_RIL(s, op, r1, c2); + } else if (type == TCG_TYPE_I32) { op = (is_unsigned ? RR_CLR : RR_CR); tcg_out_insn_RR(s, op, r1, c2); } else { @@ -XXX,XX +XXX,XX @@ static TCGConstraintSetIndex tcg_target_op_def(TCGOpcode op) return C_O1_I2(r, r, ri); case INDEX_op_setcond_i64: case INDEX_op_negsetcond_i64: - return C_O1_I2(r, r, rJU); + return C_O1_I2(r, r, rC); case INDEX_op_clz_i64: return C_O1_I2(r, r, rI); @@ -XXX,XX +XXX,XX @@ static TCGConstraintSetIndex tcg_target_op_def(TCGOpcode op) case INDEX_op_brcond_i32: return C_O0_I2(r, ri); case INDEX_op_brcond_i64: - return C_O0_I2(r, rJU); + return C_O0_I2(r, rC); case INDEX_op_bswap16_i32: case INDEX_op_bswap16_i64: @@ -XXX,XX +XXX,XX @@ static TCGConstraintSetIndex tcg_target_op_def(TCGOpcode op) case INDEX_op_movcond_i32: return C_O1_I4(r, r, ri, rI, r); case INDEX_op_movcond_i64: - return C_O1_I4(r, r, rJU, rI, r); + return C_O1_I4(r, r, rC, rI, r); case INDEX_op_div2_i32: case INDEX_op_div2_i64: -- 2.34.1 Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/s390x/tcg-target.h | 2 +- tcg/s390x/tcg-target.c.inc | 139 +++++++++++++++++++++++++------------ 2 files changed, 97 insertions(+), 44 deletions(-) diff --git a/tcg/s390x/tcg-target.h b/tcg/s390x/tcg-target.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/s390x/tcg-target.h +++ b/tcg/s390x/tcg-target.h @@ -XXX,XX +XXX,XX @@ extern uint64_t s390_facilities[3]; #define TCG_TARGET_HAS_qemu_ldst_i128 1 -#define TCG_TARGET_HAS_tst 0 +#define TCG_TARGET_HAS_tst 1 #define TCG_TARGET_HAS_v64 HAVE_FACILITY(VECTOR) #define TCG_TARGET_HAS_v128 HAVE_FACILITY(VECTOR) diff --git a/tcg/s390x/tcg-target.c.inc b/tcg/s390x/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/s390x/tcg-target.c.inc +++ b/tcg/s390x/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ typedef enum S390Opcode { RI_OILH = 0xa50a, RI_OILL = 0xa50b, RI_TMLL = 0xa701, + RI_TMLH = 0xa700, + RI_TMHL = 0xa703, + RI_TMHH = 0xa702, RIEb_CGRJ = 0xec64, RIEb_CLGRJ = 0xec65, @@ -XXX,XX +XXX,XX @@ static TCGReg tcg_target_call_oarg_reg(TCGCallReturnKind kind, int slot) #define S390_CC_NEVER 0 #define S390_CC_ALWAYS 15 +#define S390_TM_EQ 8 /* CC == 0 */ +#define S390_TM_NE 7 /* CC in {1,2,3} */ + /* Condition codes that result from a COMPARE and COMPARE LOGICAL. */ -static const uint8_t tcg_cond_to_s390_cond[] = { +static const uint8_t tcg_cond_to_s390_cond[16] = { [TCG_COND_EQ] = S390_CC_EQ, [TCG_COND_NE] = S390_CC_NE, + [TCG_COND_TSTEQ] = S390_CC_EQ, + [TCG_COND_TSTNE] = S390_CC_NE, [TCG_COND_LT] = S390_CC_LT, [TCG_COND_LE] = S390_CC_LE, [TCG_COND_GT] = S390_CC_GT, @@ -XXX,XX +XXX,XX @@ static const uint8_t tcg_cond_to_s390_cond[] = { /* Condition codes that result from a LOAD AND TEST. Here, we have no unsigned instruction variation, however since the test is vs zero we can re-map the outcomes appropriately. */ -static const uint8_t tcg_cond_to_ltr_cond[] = { +static const uint8_t tcg_cond_to_ltr_cond[16] = { [TCG_COND_EQ] = S390_CC_EQ, [TCG_COND_NE] = S390_CC_NE, + [TCG_COND_TSTEQ] = S390_CC_ALWAYS, + [TCG_COND_TSTNE] = S390_CC_NEVER, [TCG_COND_LT] = S390_CC_LT, [TCG_COND_LE] = S390_CC_LE, [TCG_COND_GT] = S390_CC_GT, @@ -XXX,XX +XXX,XX @@ static bool risbg_mask(uint64_t c) static bool tcg_target_const_match(int64_t val, int ct, TCGType type, TCGCond cond, int vece) { + uint64_t uval = val; + if (ct & TCG_CT_CONST) { return true; } if (type == TCG_TYPE_I32) { + uval = (uint32_t)val; val = (int32_t)val; } @@ -XXX,XX +XXX,XX @@ static bool tcg_target_const_match(int64_t val, int ct, case TCG_COND_GTU: ct |= TCG_CT_CONST_U32; /* CLGFI */ break; + case TCG_COND_TSTNE: + case TCG_COND_TSTEQ: + if (is_const_p16(uval) >= 0) { + return true; /* TMxx */ + } + if (risbg_mask(uval)) { + return true; /* RISBG */ + } + break; default: g_assert_not_reached(); } @@ -XXX,XX +XXX,XX @@ static bool tcg_target_const_match(int64_t val, int ct, if (ct & TCG_CT_CONST_INV) { val = ~val; } - /* - * Note that is_const_p16 is a subset of is_const_p32, - * so we don't need both constraints. - */ if ((ct & TCG_CT_CONST_P32) && is_const_p32(val) >= 0) { return true; } @@ -XXX,XX +XXX,XX @@ static const S390Opcode oi_insns[4] = { static const S390Opcode lif_insns[2] = { RIL_LLILF, RIL_LLIHF, }; +static const S390Opcode tm_insns[4] = { + RI_TMLL, RI_TMLH, RI_TMHL, RI_TMHH +}; /* load a register with an immediate value */ static void tcg_out_movi(TCGContext *s, TCGType type, @@ -XXX,XX +XXX,XX @@ static int tgen_cmp2(TCGContext *s, TCGType type, TCGCond c, TCGReg r1, TCGCond inv_c = tcg_invert_cond(c); S390Opcode op; + if (is_tst_cond(c)) { + tcg_debug_assert(!need_carry); + + if (!c2const) { + if (type == TCG_TYPE_I32) { + tcg_out_insn(s, RRFa, NRK, TCG_REG_R0, r1, c2); + } else { + tcg_out_insn(s, RRFa, NGRK, TCG_REG_R0, r1, c2); + } + goto exit; + } + + if (type == TCG_TYPE_I32) { + c2 = (uint32_t)c2; + } + + int i = is_const_p16(c2); + if (i >= 0) { + tcg_out_insn_RI(s, tm_insns[i], r1, c2 >> (i * 16)); + *inv_cc = TCG_COND_TSTEQ ? S390_TM_NE : S390_TM_EQ; + return *inv_cc ^ 15; + } + + if (risbg_mask(c2)) { + tgen_andi_risbg(s, TCG_REG_R0, r1, c2); + goto exit; + } + g_assert_not_reached(); + } + if (c2const) { if (c2 == 0) { if (!(is_unsigned && need_carry)) { @@ -XXX,XX +XXX,XX @@ static void tgen_brcond(TCGContext *s, TCGType type, TCGCond c, TCGReg r1, TCGArg c2, int c2const, TCGLabel *l) { int cc; - bool is_unsigned = is_unsigned_cond(c); - bool in_range; - S390Opcode opc; - cc = tcg_cond_to_s390_cond[c]; + if (!is_tst_cond(c)) { + bool is_unsigned = is_unsigned_cond(c); + bool in_range; + S390Opcode opc; - if (!c2const) { - opc = (type == TCG_TYPE_I32 - ? (is_unsigned ? RIEb_CLRJ : RIEb_CRJ) - : (is_unsigned ? RIEb_CLGRJ : RIEb_CGRJ)); - tgen_compare_branch(s, opc, cc, r1, c2, l); - return; - } + cc = tcg_cond_to_s390_cond[c]; - /* - * COMPARE IMMEDIATE AND BRANCH RELATIVE has an 8-bit immediate field. - * If the immediate we've been given does not fit that range, we'll - * fall back to separate compare and branch instructions using the - * larger comparison range afforded by COMPARE IMMEDIATE. - */ - if (type == TCG_TYPE_I32) { - if (is_unsigned) { - opc = RIEc_CLIJ; - in_range = (uint32_t)c2 == (uint8_t)c2; - } else { - opc = RIEc_CIJ; - in_range = (int32_t)c2 == (int8_t)c2; + if (!c2const) { + opc = (type == TCG_TYPE_I32 + ? (is_unsigned ? RIEb_CLRJ : RIEb_CRJ) + : (is_unsigned ? RIEb_CLGRJ : RIEb_CGRJ)); + tgen_compare_branch(s, opc, cc, r1, c2, l); + return; } - } else { - if (is_unsigned) { - opc = RIEc_CLGIJ; - in_range = (uint64_t)c2 == (uint8_t)c2; + + /* + * COMPARE IMMEDIATE AND BRANCH RELATIVE has an 8-bit immediate field. + * If the immediate we've been given does not fit that range, we'll + * fall back to separate compare and branch instructions using the + * larger comparison range afforded by COMPARE IMMEDIATE. + */ + if (type == TCG_TYPE_I32) { + if (is_unsigned) { + opc = RIEc_CLIJ; + in_range = (uint32_t)c2 == (uint8_t)c2; + } else { + opc = RIEc_CIJ; + in_range = (int32_t)c2 == (int8_t)c2; + } } else { - opc = RIEc_CGIJ; - in_range = (int64_t)c2 == (int8_t)c2; + if (is_unsigned) { + opc = RIEc_CLGIJ; + in_range = (uint64_t)c2 == (uint8_t)c2; + } else { + opc = RIEc_CGIJ; + in_range = (int64_t)c2 == (int8_t)c2; + } + } + if (in_range) { + tgen_compare_imm_branch(s, opc, cc, r1, c2, l); + return; } - } - if (in_range) { - tgen_compare_imm_branch(s, opc, cc, r1, c2, l); - return; } cc = tgen_cmp(s, type, c, r1, c2, c2const, false); @@ -XXX,XX +XXX,XX @@ static TCGLabelQemuLdst *prepare_host_addr(TCGContext *s, HostAddress *h, ldst->oi = oi; ldst->addrlo_reg = addr_reg; - /* We are expecting a_bits to max out at 7, much lower than TMLL. */ tcg_debug_assert(a_mask <= 0xffff); tcg_out_insn(s, RI, TMLL, addr_reg, a_mask); - tcg_out16(s, RI_BRC | (7 << 4)); /* CC in {1,2,3} */ + tcg_out16(s, RI_BRC | (S390_TM_NE << 4)); ldst->label_ptr[0] = s->code_ptr++; } @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_ldst_i128(TCGContext *s, TCGReg datalo, TCGReg datahi, l2 = gen_new_label(); tcg_out_insn(s, RI, TMLL, addr_reg, 15); - tgen_branch(s, 7, l1); /* CC in {1,2,3} */ + tgen_branch(s, S390_TM_NE, l1); } tcg_debug_assert(!need_bswap); -- 2.34.1 Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/tci/tcg-target.h | 2 +- tcg/tci.c | 14 ++++++++++++++ 2 files changed, 15 insertions(+), 1 deletion(-) diff --git a/tcg/tci/tcg-target.h b/tcg/tci/tcg-target.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/tci/tcg-target.h +++ b/tcg/tci/tcg-target.h @@ -XXX,XX +XXX,XX @@ #define TCG_TARGET_HAS_qemu_ldst_i128 0 -#define TCG_TARGET_HAS_tst 0 +#define TCG_TARGET_HAS_tst 1 /* Number of registers available. */ #define TCG_TARGET_NB_REGS 16 diff --git a/tcg/tci.c b/tcg/tci.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/tci.c +++ b/tcg/tci.c @@ -XXX,XX +XXX,XX @@ static bool tci_compare32(uint32_t u0, uint32_t u1, TCGCond condition) case TCG_COND_GTU: result = (u0 > u1); break; + case TCG_COND_TSTEQ: + result = (u0 & u1) == 0; + break; + case TCG_COND_TSTNE: + result = (u0 & u1) != 0; + break; default: g_assert_not_reached(); } @@ -XXX,XX +XXX,XX @@ static bool tci_compare64(uint64_t u0, uint64_t u1, TCGCond condition) case TCG_COND_GTU: result = (u0 > u1); break; + case TCG_COND_TSTEQ: + result = (u0 & u1) == 0; + break; + case TCG_COND_TSTNE: + result = (u0 & u1) != 0; + break; default: g_assert_not_reached(); } @@ -XXX,XX +XXX,XX @@ static const char *str_c(TCGCond c) [TCG_COND_GEU] = "geu", [TCG_COND_LEU] = "leu", [TCG_COND_GTU] = "gtu", + [TCG_COND_TSTEQ] = "tsteq", + [TCG_COND_TSTNE] = "tstne", }; assert((unsigned)c < ARRAY_SIZE(cond)); -- 2.34.1 The following changes since commit aa3a285b5bc56a4208b3b57d4a55291e9c260107: Merge tag 'mem-2024-12-21' of https://github.com/davidhildenbrand/qemu into staging (2024-12-22 14:33:27 -0500) are available in the Git repository at: https://gitlab.com/rth7680/qemu.git tags/pull-tcg-20241224 for you to fetch changes up to e4a8e093dc74be049f4829831dce76e5edab0003: accel/tcg: Move gen_intermediate_code to TCGCPUOps.translate_core (2024-12-24 08:32:15 -0800) ---------------------------------------------------------------- tcg/optimize: Remove in-flight mask data from OptContext fpu: Add float*_muladd_scalbn fpu: Remove float_muladd_halve_result fpu: Add float_round_nearest_even_max fpu: Add float_muladd_suppress_add_product_zero target/hexagon: Use float32_muladd accel/tcg: Move gen_intermediate_code to TCGCPUOps.translate_core ---------------------------------------------------------------- Ilya Leoshkevich (1): tests/tcg: Do not use inttypes.h in multiarch/system/memory.c Pierrick Bouvier (1): plugins: optimize cpu_index code generation Richard Henderson (70): tcg/optimize: Split out finish_bb, finish_ebb tcg/optimize: Split out fold_affected_mask tcg/optimize: Copy mask writeback to fold_masks tcg/optimize: Split out fold_masks_zs tcg/optimize: Augment s_mask from z_mask in fold_masks_zs tcg/optimize: Change representation of s_mask tcg/optimize: Use finish_folding in fold_add, fold_add_vec, fold_addsub2 tcg/optimize: Introduce const value accessors for TempOptInfo tcg/optimize: Use fold_masks_zs in fold_and tcg/optimize: Use fold_masks_zs in fold_andc tcg/optimize: Use fold_masks_zs in fold_bswap tcg/optimize: Use fold_masks_zs in fold_count_zeros tcg/optimize: Use fold_masks_z in fold_ctpop tcg/optimize: Use fold_and and fold_masks_z in fold_deposit tcg/optimize: Compute sign mask in fold_deposit tcg/optimize: Use finish_folding in fold_divide tcg/optimize: Use finish_folding in fold_dup, fold_dup2 tcg/optimize: Use fold_masks_s in fold_eqv tcg/optimize: Use fold_masks_z in fold_extract tcg/optimize: Use finish_folding in fold_extract2 tcg/optimize: Use fold_masks_zs in fold_exts tcg/optimize: Use fold_masks_z in fold_extu tcg/optimize: Use fold_masks_zs in fold_movcond tcg/optimize: Use finish_folding in fold_mul* tcg/optimize: Use fold_masks_s in fold_nand tcg/optimize: Use fold_masks_z in fold_neg_no_const tcg/optimize: Use fold_masks_s in fold_nor tcg/optimize: Use fold_masks_s in fold_not tcg/optimize: Use fold_masks_zs in fold_or tcg/optimize: Use fold_masks_zs in fold_orc tcg/optimize: Use fold_masks_zs in fold_qemu_ld tcg/optimize: Return true from fold_qemu_st, fold_tcg_st tcg/optimize: Use finish_folding in fold_remainder tcg/optimize: Distinguish simplification in fold_setcond_zmask tcg/optimize: Use fold_masks_z in fold_setcond tcg/optimize: Use fold_masks_s in fold_negsetcond tcg/optimize: Use fold_masks_z in fold_setcond2 tcg/optimize: Use finish_folding in fold_cmp_vec tcg/optimize: Use finish_folding in fold_cmpsel_vec tcg/optimize: Use fold_masks_zs in fold_sextract tcg/optimize: Use fold_masks_zs, fold_masks_s in fold_shift tcg/optimize: Simplify sign bit test in fold_shift tcg/optimize: Use finish_folding in fold_sub, fold_sub_vec tcg/optimize: Use fold_masks_zs in fold_tcg_ld tcg/optimize: Use finish_folding in fold_tcg_ld_memcopy tcg/optimize: Use fold_masks_zs in fold_xor tcg/optimize: Use finish_folding in fold_bitsel_vec tcg/optimize: Use finish_folding as default in tcg_optimize tcg/optimize: Remove z_mask, s_mask from OptContext tcg/optimize: Re-enable sign-mask optimizations tcg/optimize: Move fold_bitsel_vec into alphabetic sort tcg/optimize: Move fold_cmp_vec, fold_cmpsel_vec into alphabetic sort softfloat: Add float{16,32,64}_muladd_scalbn target/arm: Use float*_muladd_scalbn target/sparc: Use float*_muladd_scalbn softfloat: Remove float_muladd_halve_result softfloat: Add float_round_nearest_even_max softfloat: Add float_muladd_suppress_add_product_zero target/hexagon: Use float32_mul in helper_sfmpy target/hexagon: Use float32_muladd for helper_sffma target/hexagon: Use float32_muladd for helper_sffms target/hexagon: Use float32_muladd_scalbn for helper_sffma_sc target/hexagon: Use float32_muladd for helper_sffm[as]_lib target/hexagon: Remove internal_fmafx target/hexagon: Expand GEN_XF_ROUND target/hexagon: Remove Float target/hexagon: Remove Double target/hexagon: Use mulu64 for int128_mul_6464 target/hexagon: Simplify internal_mpyhh setup accel/tcg: Move gen_intermediate_code to TCGCPUOps.translate_core include/exec/translator.h | 14 - include/fpu/softfloat-types.h | 2 + include/fpu/softfloat.h | 14 +- include/hw/core/tcg-cpu-ops.h | 13 + target/alpha/cpu.h | 2 + target/arm/internals.h | 2 + target/avr/cpu.h | 2 + target/hexagon/cpu.h | 2 + target/hexagon/fma_emu.h | 3 - target/hppa/cpu.h | 2 + target/i386/tcg/helper-tcg.h | 2 + target/loongarch/internals.h | 2 + target/m68k/cpu.h | 2 + target/microblaze/cpu.h | 2 + target/mips/tcg/tcg-internal.h | 2 + target/openrisc/cpu.h | 2 + target/ppc/cpu.h | 2 + target/riscv/cpu.h | 3 + target/rx/cpu.h | 2 + target/s390x/s390x-internal.h | 2 + target/sh4/cpu.h | 2 + target/sparc/cpu.h | 2 + target/sparc/helper.h | 4 +- target/tricore/cpu.h | 2 + target/xtensa/cpu.h | 2 + accel/tcg/cpu-exec.c | 8 +- accel/tcg/plugin-gen.c | 9 + accel/tcg/translate-all.c | 8 +- fpu/softfloat.c | 63 +-- target/alpha/cpu.c | 1 + target/alpha/translate.c | 4 +- target/arm/cpu.c | 1 + target/arm/tcg/cpu-v7m.c | 1 + target/arm/tcg/helper-a64.c | 6 +- target/arm/tcg/translate.c | 5 +- target/avr/cpu.c | 1 + target/avr/translate.c | 6 +- target/hexagon/cpu.c | 1 + target/hexagon/fma_emu.c | 496 ++++++--------------- target/hexagon/op_helper.c | 125 ++---- target/hexagon/translate.c | 4 +- target/hppa/cpu.c | 1 + target/hppa/translate.c | 4 +- target/i386/tcg/tcg-cpu.c | 1 + target/i386/tcg/translate.c | 5 +- target/loongarch/cpu.c | 1 + target/loongarch/tcg/translate.c | 4 +- target/m68k/cpu.c | 1 + target/m68k/translate.c | 4 +- target/microblaze/cpu.c | 1 + target/microblaze/translate.c | 4 +- target/mips/cpu.c | 1 + target/mips/tcg/translate.c | 4 +- target/openrisc/cpu.c | 1 + target/openrisc/translate.c | 4 +- target/ppc/cpu_init.c | 1 + target/ppc/translate.c | 4 +- target/riscv/tcg/tcg-cpu.c | 1 + target/riscv/translate.c | 4 +- target/rx/cpu.c | 1 + target/rx/translate.c | 4 +- target/s390x/cpu.c | 1 + target/s390x/tcg/translate.c | 4 +- target/sh4/cpu.c | 1 + target/sh4/translate.c | 4 +- target/sparc/cpu.c | 1 + target/sparc/fop_helper.c | 8 +- target/sparc/translate.c | 84 ++-- target/tricore/cpu.c | 1 + target/tricore/translate.c | 5 +- target/xtensa/cpu.c | 1 + target/xtensa/translate.c | 4 +- tcg/optimize.c | 857 +++++++++++++++++++----------------- tests/tcg/multiarch/system/memory.c | 9 +- fpu/softfloat-parts.c.inc | 16 +- 75 files changed, 866 insertions(+), 1009 deletions(-) From: Ilya Leoshkevich <iii@linux.ibm.com> make check-tcg fails on Fedora with the following error message: alpha-linux-gnu-gcc [...] qemu/tests/tcg/multiarch/system/memory.c -o memory [...] qemu/tests/tcg/multiarch/system/memory.c:17:10: fatal error: inttypes.h: No such file or directory 17 | #include <inttypes.h> | ^~~~~~~~~~~~ compilation terminated. The reason is that Fedora has cross-compilers, but no cross-glibc headers. Fix by hardcoding the format specifiers and dropping the include. An alternative fix would be to introduce a configure check for inttypes.h. But this would make it impossible to use Fedora cross-compilers for softmmu tests, which used to work so far. Fixes: ecbcc9ead2f8 ("tests/tcg: add a system test to check memory instrumentation") Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com> Reviewed-by: Paolo Bonzini <pbonzini@redhat.com> Message-ID: <20241010085906.226249-1-iii@linux.ibm.com> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tests/tcg/multiarch/system/memory.c | 9 ++++----- 1 file changed, 4 insertions(+), 5 deletions(-) diff --git a/tests/tcg/multiarch/system/memory.c b/tests/tcg/multiarch/system/memory.c index XXXXXXX..XXXXXXX 100644 --- a/tests/tcg/multiarch/system/memory.c +++ b/tests/tcg/multiarch/system/memory.c @@ -XXX,XX +XXX,XX @@ #include <stdint.h> #include <stdbool.h> -#include <inttypes.h> #include <minilib.h> #ifndef CHECK_UNALIGNED @@ -XXX,XX +XXX,XX @@ int main(void) int i; bool ok = true; - ml_printf("Test data start: 0x%"PRIxPTR"\n", &test_data[0]); - ml_printf("Test data end: 0x%"PRIxPTR"\n", &test_data[TEST_SIZE]); + ml_printf("Test data start: 0x%lx\n", (unsigned long)&test_data[0]); + ml_printf("Test data end: 0x%lx\n", (unsigned long)&test_data[TEST_SIZE]); /* Run through the unsigned tests first */ for (i = 0; i < ARRAY_SIZE(init_ufns) && ok; i++) { @@ -XXX,XX +XXX,XX @@ int main(void) ok = do_signed_reads(true); } - ml_printf("Test data read: %"PRId32"\n", test_read_count); - ml_printf("Test data write: %"PRId32"\n", test_write_count); + ml_printf("Test data read: %lu\n", (unsigned long)test_read_count); + ml_printf("Test data write: %lu\n", (unsigned long)test_write_count); ml_printf("Test complete: %s\n", ok ? "PASSED" : "FAILED"); return ok ? 0 : -1; } -- 2.43.0 From: Pierrick Bouvier <pierrick.bouvier@linaro.org> When running with a single vcpu, we can return a constant instead of a load when accessing cpu_index. A side effect is that all tcg operations using it are optimized, most notably scoreboard access. When running a simple loop in user-mode, the speedup is around 20%. Signed-off-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-ID: <20241128213843.1023080-1-pierrick.bouvier@linaro.org> --- accel/tcg/plugin-gen.c | 9 +++++++++ 1 file changed, 9 insertions(+) diff --git a/accel/tcg/plugin-gen.c b/accel/tcg/plugin-gen.c index XXXXXXX..XXXXXXX 100644 --- a/accel/tcg/plugin-gen.c +++ b/accel/tcg/plugin-gen.c @@ -XXX,XX +XXX,XX @@ static void gen_disable_mem_helper(void) static TCGv_i32 gen_cpu_index(void) { + /* + * Optimize when we run with a single vcpu. All values using cpu_index, + * including scoreboard index, will be optimized out. + * User-mode calls tb_flush when setting this flag. In system-mode, all + * vcpus are created before generating code. + */ + if (!tcg_cflags_has(current_cpu, CF_PARALLEL)) { + return tcg_constant_i32(current_cpu->cpu_index); + } TCGv_i32 cpu_index = tcg_temp_ebb_new_i32(); tcg_gen_ld_i32(cpu_index, tcg_env, -offsetof(ArchCPU, env) + offsetof(CPUState, cpu_index)); -- 2.43.0 Call them directly from the opcode switch statement in tcg_optimize, rather than in finish_folding based on opcode flags. Adjust folding of conditional branches to match. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 47 +++++++++++++++++++++++++++++++---------------- 1 file changed, 31 insertions(+), 16 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static void copy_propagate(OptContext *ctx, TCGOp *op, } } +static void finish_bb(OptContext *ctx) +{ + /* We only optimize memory barriers across basic blocks. */ + ctx->prev_mb = NULL; +} + +static void finish_ebb(OptContext *ctx) +{ + finish_bb(ctx); + /* We only optimize across extended basic blocks. */ + memset(&ctx->temps_used, 0, sizeof(ctx->temps_used)); + remove_mem_copy_all(ctx); +} + static void finish_folding(OptContext *ctx, TCGOp *op) { const TCGOpDef *def = &tcg_op_defs[op->opc]; int i, nb_oargs; - /* - * We only optimize extended basic blocks. If the opcode ends a BB - * and is not a conditional branch, reset all temp data. - */ - if (def->flags & TCG_OPF_BB_END) { - ctx->prev_mb = NULL; - if (!(def->flags & TCG_OPF_COND_BRANCH)) { - memset(&ctx->temps_used, 0, sizeof(ctx->temps_used)); - remove_mem_copy_all(ctx); - } - return; - } - nb_oargs = def->nb_oargs; for (i = 0; i < nb_oargs; i++) { TCGTemp *ts = arg_temp(op->args[i]); @@ -XXX,XX +XXX,XX @@ static bool fold_brcond(OptContext *ctx, TCGOp *op) if (i > 0) { op->opc = INDEX_op_br; op->args[0] = op->args[3]; + finish_ebb(ctx); + } else { + finish_bb(ctx); } - return false; + return true; } static bool fold_brcond2(OptContext *ctx, TCGOp *op) @@ -XXX,XX +XXX,XX @@ static bool fold_brcond2(OptContext *ctx, TCGOp *op) } op->opc = INDEX_op_br; op->args[0] = label; - break; + finish_ebb(ctx); + return true; } - return false; + + finish_bb(ctx); + return true; } static bool fold_bswap(OptContext *ctx, TCGOp *op) @@ -XXX,XX +XXX,XX @@ void tcg_optimize(TCGContext *s) CASE_OP_32_64_VEC(xor): done = fold_xor(&ctx, op); break; + case INDEX_op_set_label: + case INDEX_op_br: + case INDEX_op_exit_tb: + case INDEX_op_goto_tb: + case INDEX_op_goto_ptr: + finish_ebb(&ctx); + done = true; + break; default: break; } -- 2.43.0 There are only a few logical operations which can compute an "affected" mask. Split out handling of this optimization to a separate function, only to be called when applicable. Remove the a_mask field from OptContext, as the mask is no longer stored anywhere. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 42 +++++++++++++++++++++++++++--------------- 1 file changed, 27 insertions(+), 15 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ typedef struct OptContext { QSIMPLEQ_HEAD(, MemCopyInfo) mem_free; /* In flight values from optimization. */ - uint64_t a_mask; /* mask bit is 0 iff value identical to first input */ uint64_t z_mask; /* mask bit is 0 iff value bit is 0 */ uint64_t s_mask; /* mask of clrsb(value) bits */ TCGType type; @@ -XXX,XX +XXX,XX @@ static bool fold_const2_commutative(OptContext *ctx, TCGOp *op) static bool fold_masks(OptContext *ctx, TCGOp *op) { - uint64_t a_mask = ctx->a_mask; uint64_t z_mask = ctx->z_mask; uint64_t s_mask = ctx->s_mask; @@ -XXX,XX +XXX,XX @@ static bool fold_masks(OptContext *ctx, TCGOp *op) * type changing opcodes. */ if (ctx->type == TCG_TYPE_I32) { - a_mask = (int32_t)a_mask; z_mask = (int32_t)z_mask; s_mask |= MAKE_64BIT_MASK(32, 32); ctx->z_mask = z_mask; @@ -XXX,XX +XXX,XX @@ static bool fold_masks(OptContext *ctx, TCGOp *op) if (z_mask == 0) { return tcg_opt_gen_movi(ctx, op, op->args[0], 0); } + return false; +} + +/* + * An "affected" mask bit is 0 if and only if the result is identical + * to the first input. Thus if the entire mask is 0, the operation + * is equivalent to a copy. + */ +static bool fold_affected_mask(OptContext *ctx, TCGOp *op, uint64_t a_mask) +{ + if (ctx->type == TCG_TYPE_I32) { + a_mask = (uint32_t)a_mask; + } if (a_mask == 0) { return tcg_opt_gen_mov(ctx, op, op->args[0], op->args[1]); } @@ -XXX,XX +XXX,XX @@ static bool fold_and(OptContext *ctx, TCGOp *op) * Known-zeros does not imply known-ones. Therefore unless * arg2 is constant, we can't infer affected bits from it. */ - if (arg_is_const(op->args[2])) { - ctx->a_mask = z1 & ~z2; + if (arg_is_const(op->args[2]) && + fold_affected_mask(ctx, op, z1 & ~z2)) { + return true; } return fold_masks(ctx, op); @@ -XXX,XX +XXX,XX @@ static bool fold_andc(OptContext *ctx, TCGOp *op) */ if (arg_is_const(op->args[2])) { uint64_t z2 = ~arg_info(op->args[2])->z_mask; - ctx->a_mask = z1 & ~z2; + if (fold_affected_mask(ctx, op, z1 & ~z2)) { + return true; + } z1 &= z2; } ctx->z_mask = z1; @@ -XXX,XX +XXX,XX @@ static bool fold_extract(OptContext *ctx, TCGOp *op) z_mask_old = arg_info(op->args[1])->z_mask; z_mask = extract64(z_mask_old, pos, len); - if (pos == 0) { - ctx->a_mask = z_mask_old ^ z_mask; + if (pos == 0 && fold_affected_mask(ctx, op, z_mask_old ^ z_mask)) { + return true; } ctx->z_mask = z_mask; ctx->s_mask = smask_from_zmask(z_mask); @@ -XXX,XX +XXX,XX @@ static bool fold_exts(OptContext *ctx, TCGOp *op) ctx->z_mask = z_mask; ctx->s_mask = s_mask; - if (!type_change) { - ctx->a_mask = s_mask & ~s_mask_old; + if (!type_change && fold_affected_mask(ctx, op, s_mask & ~s_mask_old)) { + return true; } return fold_masks(ctx, op); @@ -XXX,XX +XXX,XX @@ static bool fold_extu(OptContext *ctx, TCGOp *op) ctx->z_mask = z_mask; ctx->s_mask = smask_from_zmask(z_mask); - if (!type_change) { - ctx->a_mask = z_mask_old ^ z_mask; + if (!type_change && fold_affected_mask(ctx, op, z_mask_old ^ z_mask)) { + return true; } return fold_masks(ctx, op); } @@ -XXX,XX +XXX,XX @@ static bool fold_sextract(OptContext *ctx, TCGOp *op) s_mask |= MAKE_64BIT_MASK(len, 64 - len); ctx->s_mask = s_mask; - if (pos == 0) { - ctx->a_mask = s_mask & ~s_mask_old; + if (pos == 0 && fold_affected_mask(ctx, op, s_mask & ~s_mask_old)) { + return true; } return fold_masks(ctx, op); @@ -XXX,XX +XXX,XX @@ void tcg_optimize(TCGContext *s) } /* Assume all bits affected, no bits known zero, no sign reps. */ - ctx.a_mask = -1; ctx.z_mask = -1; ctx.s_mask = 0; -- 2.43.0 Use of fold_masks should be restricted to those opcodes that can reliably make use of it -- those with a single output, and from higher-level folders that set up the masks. Prepare for conversion of each folder in turn. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 17 ++++++++++++++--- 1 file changed, 14 insertions(+), 3 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_masks(OptContext *ctx, TCGOp *op) { uint64_t z_mask = ctx->z_mask; uint64_t s_mask = ctx->s_mask; + const TCGOpDef *def = &tcg_op_defs[op->opc]; + TCGTemp *ts; + TempOptInfo *ti; + + /* Only single-output opcodes are supported here. */ + tcg_debug_assert(def->nb_oargs == 1); /* * 32-bit ops generate 32-bit results, which for the purpose of @@ -XXX,XX +XXX,XX @@ static bool fold_masks(OptContext *ctx, TCGOp *op) if (ctx->type == TCG_TYPE_I32) { z_mask = (int32_t)z_mask; s_mask |= MAKE_64BIT_MASK(32, 32); - ctx->z_mask = z_mask; - ctx->s_mask = s_mask; } if (z_mask == 0) { return tcg_opt_gen_movi(ctx, op, op->args[0], 0); } - return false; + + ts = arg_temp(op->args[0]); + reset_ts(ctx, ts); + + ti = ts_info(ts); + ti->z_mask = z_mask; + ti->s_mask = s_mask; + return true; } /* -- 2.43.0 Add a routine to which masks can be passed directly, rather than storing them into OptContext. To be used in upcoming patches. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 15 ++++++++++++--- 1 file changed, 12 insertions(+), 3 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_const2_commutative(OptContext *ctx, TCGOp *op) return fold_const2(ctx, op); } -static bool fold_masks(OptContext *ctx, TCGOp *op) +/* + * Record "zero" and "sign" masks for the single output of @op. + * See TempOptInfo definition of z_mask and s_mask. + * If z_mask allows, fold the output to constant zero. + */ +static bool fold_masks_zs(OptContext *ctx, TCGOp *op, + uint64_t z_mask, uint64_t s_mask) { - uint64_t z_mask = ctx->z_mask; - uint64_t s_mask = ctx->s_mask; const TCGOpDef *def = &tcg_op_defs[op->opc]; TCGTemp *ts; TempOptInfo *ti; @@ -XXX,XX +XXX,XX @@ static bool fold_masks(OptContext *ctx, TCGOp *op) return true; } +static bool fold_masks(OptContext *ctx, TCGOp *op) +{ + return fold_masks_zs(ctx, op, ctx->z_mask, ctx->s_mask); +} + /* * An "affected" mask bit is 0 if and only if the result is identical * to the first input. Thus if the entire mask is 0, the operation -- 2.43.0 Consider the passed s_mask to be a minimum deduced from either existing s_mask or from a sign-extension operation. We may be able to deduce more from the set of known zeros. Remove identical logic from several opcode folders. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 21 ++++++--------------- 1 file changed, 6 insertions(+), 15 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_const2_commutative(OptContext *ctx, TCGOp *op) * Record "zero" and "sign" masks for the single output of @op. * See TempOptInfo definition of z_mask and s_mask. * If z_mask allows, fold the output to constant zero. + * The passed s_mask may be augmented by z_mask. */ static bool fold_masks_zs(OptContext *ctx, TCGOp *op, uint64_t z_mask, uint64_t s_mask) @@ -XXX,XX +XXX,XX @@ static bool fold_masks_zs(OptContext *ctx, TCGOp *op, ti = ts_info(ts); ti->z_mask = z_mask; - ti->s_mask = s_mask; + ti->s_mask = s_mask | smask_from_zmask(z_mask); return true; } @@ -XXX,XX +XXX,XX @@ static bool fold_bswap(OptContext *ctx, TCGOp *op) default: g_assert_not_reached(); } - s_mask = smask_from_zmask(z_mask); + s_mask = 0; switch (op->args[2] & (TCG_BSWAP_OZ | TCG_BSWAP_OS)) { case TCG_BSWAP_OZ: break; @@ -XXX,XX +XXX,XX @@ static bool fold_bswap(OptContext *ctx, TCGOp *op) default: /* The high bits are undefined: force all bits above the sign to 1. */ z_mask |= sign << 1; - s_mask = 0; break; } ctx->z_mask = z_mask; @@ -XXX,XX +XXX,XX @@ static bool fold_count_zeros(OptContext *ctx, TCGOp *op) g_assert_not_reached(); } ctx->z_mask = arg_info(op->args[2])->z_mask | z_mask; - ctx->s_mask = smask_from_zmask(ctx->z_mask); return false; } @@ -XXX,XX +XXX,XX @@ static bool fold_ctpop(OptContext *ctx, TCGOp *op) default: g_assert_not_reached(); } - ctx->s_mask = smask_from_zmask(ctx->z_mask); return false; } @@ -XXX,XX +XXX,XX @@ static bool fold_extract(OptContext *ctx, TCGOp *op) return true; } ctx->z_mask = z_mask; - ctx->s_mask = smask_from_zmask(z_mask); return fold_masks(ctx, op); } @@ -XXX,XX +XXX,XX @@ static bool fold_extu(OptContext *ctx, TCGOp *op) } ctx->z_mask = z_mask; - ctx->s_mask = smask_from_zmask(z_mask); if (!type_change && fold_affected_mask(ctx, op, z_mask_old ^ z_mask)) { return true; } @@ -XXX,XX +XXX,XX @@ static bool fold_qemu_ld(OptContext *ctx, TCGOp *op) int width = 8 * memop_size(mop); if (width < 64) { - ctx->s_mask = MAKE_64BIT_MASK(width, 64 - width); - if (!(mop & MO_SIGN)) { + if (mop & MO_SIGN) { + ctx->s_mask = MAKE_64BIT_MASK(width, 64 - width); + } else { ctx->z_mask = MAKE_64BIT_MASK(0, width); - ctx->s_mask <<= 1; } } @@ -XXX,XX +XXX,XX @@ static bool fold_setcond(OptContext *ctx, TCGOp *op) fold_setcond_tst_pow2(ctx, op, false); ctx->z_mask = 1; - ctx->s_mask = smask_from_zmask(1); return false; } @@ -XXX,XX +XXX,XX @@ static bool fold_setcond2(OptContext *ctx, TCGOp *op) } ctx->z_mask = 1; - ctx->s_mask = smask_from_zmask(1); return false; do_setcond_const: @@ -XXX,XX +XXX,XX @@ static bool fold_tcg_ld(OptContext *ctx, TCGOp *op) break; CASE_OP_32_64(ld8u): ctx->z_mask = MAKE_64BIT_MASK(0, 8); - ctx->s_mask = MAKE_64BIT_MASK(9, 55); break; CASE_OP_32_64(ld16s): ctx->s_mask = MAKE_64BIT_MASK(16, 48); break; CASE_OP_32_64(ld16u): ctx->z_mask = MAKE_64BIT_MASK(0, 16); - ctx->s_mask = MAKE_64BIT_MASK(17, 47); break; case INDEX_op_ld32s_i64: ctx->s_mask = MAKE_64BIT_MASK(32, 32); break; case INDEX_op_ld32u_i64: ctx->z_mask = MAKE_64BIT_MASK(0, 32); - ctx->s_mask = MAKE_64BIT_MASK(33, 31); break; default: g_assert_not_reached(); -- 2.43.0 Change the representation from sign bit repetitions to all bits equal to the sign bit, including the sign bit itself. The previous format has a problem in that it is difficult to recreate a valid sign mask after a shift operation: the "repetitions" part of the previous format meant that applying the same shift as for the value lead to an off-by-one value. The new format, including the sign bit itself, means that the sign mask can be manipulated in exactly the same way as the value, canonicalization is easier. Canonicalize the s_mask in fold_masks_zs, rather than requiring callers to do so. Treat 0 as a non-canonical but typeless input for no sign information, which will be reset as appropriate for the data type. We can easily fold in the data from z_mask while canonicalizing. Temporarily disable optimizations using s_mask while each operation is converted to use fold_masks_zs and to the new form. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 64 ++++++++++++-------------------------------------- 1 file changed, 15 insertions(+), 49 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ typedef struct TempOptInfo { QSIMPLEQ_HEAD(, MemCopyInfo) mem_copy; uint64_t val; uint64_t z_mask; /* mask bit is 0 if and only if value bit is 0 */ - uint64_t s_mask; /* a left-aligned mask of clrsb(value) bits. */ + uint64_t s_mask; /* mask bit is 1 if value bit matches msb */ } TempOptInfo; typedef struct OptContext { @@ -XXX,XX +XXX,XX @@ typedef struct OptContext { /* In flight values from optimization. */ uint64_t z_mask; /* mask bit is 0 iff value bit is 0 */ - uint64_t s_mask; /* mask of clrsb(value) bits */ + uint64_t s_mask; /* mask bit is 1 if value bit matches msb */ TCGType type; } OptContext; -/* Calculate the smask for a specific value. */ -static uint64_t smask_from_value(uint64_t value) -{ - int rep = clrsb64(value); - return ~(~0ull >> rep); -} - -/* - * Calculate the smask for a given set of known-zeros. - * If there are lots of zeros on the left, we can consider the remainder - * an unsigned field, and thus the corresponding signed field is one bit - * larger. - */ -static uint64_t smask_from_zmask(uint64_t zmask) -{ - /* - * Only the 0 bits are significant for zmask, thus the msb itself - * must be zero, else we have no sign information. - */ - int rep = clz64(zmask); - if (rep == 0) { - return 0; - } - rep -= 1; - return ~(~0ull >> rep); -} - -/* - * Recreate a properly left-aligned smask after manipulation. - * Some bit-shuffling, particularly shifts and rotates, may - * retain sign bits on the left, but may scatter disconnected - * sign bits on the right. Retain only what remains to the left. - */ -static uint64_t smask_from_smask(int64_t smask) -{ - /* Only the 1 bits are significant for smask */ - return smask_from_zmask(~smask); -} - static inline TempOptInfo *ts_info(TCGTemp *ts) { return ts->state_ptr; @@ -XXX,XX +XXX,XX @@ static void init_ts_info(OptContext *ctx, TCGTemp *ts) ti->is_const = true; ti->val = ts->val; ti->z_mask = ts->val; - ti->s_mask = smask_from_value(ts->val); + ti->s_mask = INT64_MIN >> clrsb64(ts->val); } else { ti->is_const = false; ti->z_mask = -1; @@ -XXX,XX +XXX,XX @@ static void finish_folding(OptContext *ctx, TCGOp *op) */ if (i == 0) { ts_info(ts)->z_mask = ctx->z_mask; - ts_info(ts)->s_mask = ctx->s_mask; } } } @@ -XXX,XX +XXX,XX @@ static bool fold_const2_commutative(OptContext *ctx, TCGOp *op) * The passed s_mask may be augmented by z_mask. */ static bool fold_masks_zs(OptContext *ctx, TCGOp *op, - uint64_t z_mask, uint64_t s_mask) + uint64_t z_mask, int64_t s_mask) { const TCGOpDef *def = &tcg_op_defs[op->opc]; TCGTemp *ts; TempOptInfo *ti; + int rep; /* Only single-output opcodes are supported here. */ tcg_debug_assert(def->nb_oargs == 1); @@ -XXX,XX +XXX,XX @@ static bool fold_masks_zs(OptContext *ctx, TCGOp *op, */ if (ctx->type == TCG_TYPE_I32) { z_mask = (int32_t)z_mask; - s_mask |= MAKE_64BIT_MASK(32, 32); + s_mask |= INT32_MIN; } if (z_mask == 0) { @@ -XXX,XX +XXX,XX @@ static bool fold_masks_zs(OptContext *ctx, TCGOp *op, ti = ts_info(ts); ti->z_mask = z_mask; - ti->s_mask = s_mask | smask_from_zmask(z_mask); + + /* Canonicalize s_mask and incorporate data from z_mask. */ + rep = clz64(~s_mask); + rep = MAX(rep, clz64(z_mask)); + rep = MAX(rep - 1, 0); + ti->s_mask = INT64_MIN >> rep; + return true; } @@ -XXX,XX +XXX,XX @@ static bool fold_exts(OptContext *ctx, TCGOp *op) ctx->z_mask = z_mask; ctx->s_mask = s_mask; - if (!type_change && fold_affected_mask(ctx, op, s_mask & ~s_mask_old)) { + if (0 && !type_change && fold_affected_mask(ctx, op, s_mask & ~s_mask_old)) { return true; } @@ -XXX,XX +XXX,XX @@ static bool fold_sextract(OptContext *ctx, TCGOp *op) s_mask |= MAKE_64BIT_MASK(len, 64 - len); ctx->s_mask = s_mask; - if (pos == 0 && fold_affected_mask(ctx, op, s_mask & ~s_mask_old)) { + if (0 && pos == 0 && fold_affected_mask(ctx, op, s_mask & ~s_mask_old)) { return true; } @@ -XXX,XX +XXX,XX @@ static bool fold_shift(OptContext *ctx, TCGOp *op) ctx->z_mask = do_constant_folding(op->opc, ctx->type, z_mask, sh); s_mask = do_constant_folding(op->opc, ctx->type, s_mask, sh); - ctx->s_mask = smask_from_smask(s_mask); return fold_masks(ctx, op); } -- 2.43.0 Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static void finish_ebb(OptContext *ctx) remove_mem_copy_all(ctx); } -static void finish_folding(OptContext *ctx, TCGOp *op) +static bool finish_folding(OptContext *ctx, TCGOp *op) { const TCGOpDef *def = &tcg_op_defs[op->opc]; int i, nb_oargs; @@ -XXX,XX +XXX,XX @@ static void finish_folding(OptContext *ctx, TCGOp *op) ts_info(ts)->z_mask = ctx->z_mask; } } + return true; } /* @@ -XXX,XX +XXX,XX @@ static bool fold_add(OptContext *ctx, TCGOp *op) fold_xi_to_x(ctx, op, 0)) { return true; } - return false; + return finish_folding(ctx, op); } /* We cannot as yet do_constant_folding with vectors. */ @@ -XXX,XX +XXX,XX @@ static bool fold_add_vec(OptContext *ctx, TCGOp *op) fold_xi_to_x(ctx, op, 0)) { return true; } - return false; + return finish_folding(ctx, op); } static bool fold_addsub2(OptContext *ctx, TCGOp *op, bool add) @@ -XXX,XX +XXX,XX @@ static bool fold_addsub2(OptContext *ctx, TCGOp *op, bool add) op->args[4] = arg_new_constant(ctx, bl); op->args[5] = arg_new_constant(ctx, bh); } - return false; + return finish_folding(ctx, op); } static bool fold_add2(OptContext *ctx, TCGOp *op) -- 2.43.0 Introduce ti_is_const, ti_const_val, ti_is_const_val. Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 20 +++++++++++++++++--- 1 file changed, 17 insertions(+), 3 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static inline TempOptInfo *arg_info(TCGArg arg) return ts_info(arg_temp(arg)); } +static inline bool ti_is_const(TempOptInfo *ti) +{ + return ti->is_const; +} + +static inline uint64_t ti_const_val(TempOptInfo *ti) +{ + return ti->val; +} + +static inline bool ti_is_const_val(TempOptInfo *ti, uint64_t val) +{ + return ti_is_const(ti) && ti_const_val(ti) == val; +} + static inline bool ts_is_const(TCGTemp *ts) { - return ts_info(ts)->is_const; + return ti_is_const(ts_info(ts)); } static inline bool ts_is_const_val(TCGTemp *ts, uint64_t val) { - TempOptInfo *ti = ts_info(ts); - return ti->is_const && ti->val == val; + return ti_is_const_val(ts_info(ts), val); } static inline bool arg_is_const(TCGArg arg) -- 2.43.0 Avoid the use of the OptContext slots. Find TempOptInfo once. Sink mask computation below fold_affected_mask early exit. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 30 ++++++++++++++++-------------- 1 file changed, 16 insertions(+), 14 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_add2(OptContext *ctx, TCGOp *op) static bool fold_and(OptContext *ctx, TCGOp *op) { - uint64_t z1, z2; + uint64_t z1, z2, z_mask, s_mask; + TempOptInfo *t1, *t2; if (fold_const2_commutative(ctx, op) || fold_xi_to_i(ctx, op, 0) || @@ -XXX,XX +XXX,XX @@ static bool fold_and(OptContext *ctx, TCGOp *op) return true; } - z1 = arg_info(op->args[1])->z_mask; - z2 = arg_info(op->args[2])->z_mask; - ctx->z_mask = z1 & z2; - - /* - * Sign repetitions are perforce all identical, whether they are 1 or 0. - * Bitwise operations preserve the relative quantity of the repetitions. - */ - ctx->s_mask = arg_info(op->args[1])->s_mask - & arg_info(op->args[2])->s_mask; + t1 = arg_info(op->args[1]); + t2 = arg_info(op->args[2]); + z1 = t1->z_mask; + z2 = t2->z_mask; /* * Known-zeros does not imply known-ones. Therefore unless * arg2 is constant, we can't infer affected bits from it. */ - if (arg_is_const(op->args[2]) && - fold_affected_mask(ctx, op, z1 & ~z2)) { + if (ti_is_const(t2) && fold_affected_mask(ctx, op, z1 & ~z2)) { return true; } - return fold_masks(ctx, op); + z_mask = z1 & z2; + + /* + * Sign repetitions are perforce all identical, whether they are 1 or 0. + * Bitwise operations preserve the relative quantity of the repetitions. + */ + s_mask = t1->s_mask & t2->s_mask; + + return fold_masks_zs(ctx, op, z_mask, s_mask); } static bool fold_andc(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Find TempOptInfo once. Avoid double inversion of the value of second const operand. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 21 +++++++++++---------- 1 file changed, 11 insertions(+), 10 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_and(OptContext *ctx, TCGOp *op) static bool fold_andc(OptContext *ctx, TCGOp *op) { - uint64_t z1; + uint64_t z_mask, s_mask; + TempOptInfo *t1, *t2; if (fold_const2(ctx, op) || fold_xx_to_i(ctx, op, 0) || @@ -XXX,XX +XXX,XX @@ static bool fold_andc(OptContext *ctx, TCGOp *op) return true; } - z1 = arg_info(op->args[1])->z_mask; + t1 = arg_info(op->args[1]); + t2 = arg_info(op->args[2]); + z_mask = t1->z_mask; /* * Known-zeros does not imply known-ones. Therefore unless * arg2 is constant, we can't infer anything from it. */ - if (arg_is_const(op->args[2])) { - uint64_t z2 = ~arg_info(op->args[2])->z_mask; - if (fold_affected_mask(ctx, op, z1 & ~z2)) { + if (ti_is_const(t2)) { + uint64_t v2 = ti_const_val(t2); + if (fold_affected_mask(ctx, op, z_mask & v2)) { return true; } - z1 &= z2; + z_mask &= ~v2; } - ctx->z_mask = z1; - ctx->s_mask = arg_info(op->args[1])->s_mask - & arg_info(op->args[2])->s_mask; - return fold_masks(ctx, op); + s_mask = t1->s_mask & t2->s_mask; + return fold_masks_zs(ctx, op, z_mask, s_mask); } static bool fold_brcond(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Find TempOptInfo once. Always set s_mask along the BSWAP_OS path, since the result is being explicitly sign-extended. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 21 ++++++++++----------- 1 file changed, 10 insertions(+), 11 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_brcond2(OptContext *ctx, TCGOp *op) static bool fold_bswap(OptContext *ctx, TCGOp *op) { uint64_t z_mask, s_mask, sign; + TempOptInfo *t1 = arg_info(op->args[1]); - if (arg_is_const(op->args[1])) { - uint64_t t = arg_info(op->args[1])->val; - - t = do_constant_folding(op->opc, ctx->type, t, op->args[2]); - return tcg_opt_gen_movi(ctx, op, op->args[0], t); + if (ti_is_const(t1)) { + return tcg_opt_gen_movi(ctx, op, op->args[0], + do_constant_folding(op->opc, ctx->type, + ti_const_val(t1), + op->args[2])); } - z_mask = arg_info(op->args[1])->z_mask; - + z_mask = t1->z_mask; switch (op->opc) { case INDEX_op_bswap16_i32: case INDEX_op_bswap16_i64: @@ -XXX,XX +XXX,XX @@ static bool fold_bswap(OptContext *ctx, TCGOp *op) /* If the sign bit may be 1, force all the bits above to 1. */ if (z_mask & sign) { z_mask |= sign; - s_mask = sign << 1; } + /* The value and therefore s_mask is explicitly sign-extended. */ + s_mask = sign; break; default: /* The high bits are undefined: force all bits above the sign to 1. */ z_mask |= sign << 1; break; } - ctx->z_mask = z_mask; - ctx->s_mask = s_mask; - return fold_masks(ctx, op); + return fold_masks_zs(ctx, op, z_mask, s_mask); } static bool fold_call(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Find TempOptInfo once. Compute s_mask from the union of the maximum count and the op2 fallback for op1 being zero. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 15 ++++++++++----- 1 file changed, 10 insertions(+), 5 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_call(OptContext *ctx, TCGOp *op) static bool fold_count_zeros(OptContext *ctx, TCGOp *op) { - uint64_t z_mask; + uint64_t z_mask, s_mask; + TempOptInfo *t1 = arg_info(op->args[1]); + TempOptInfo *t2 = arg_info(op->args[2]); - if (arg_is_const(op->args[1])) { - uint64_t t = arg_info(op->args[1])->val; + if (ti_is_const(t1)) { + uint64_t t = ti_const_val(t1); if (t != 0) { t = do_constant_folding(op->opc, ctx->type, t, 0); @@ -XXX,XX +XXX,XX @@ static bool fold_count_zeros(OptContext *ctx, TCGOp *op) default: g_assert_not_reached(); } - ctx->z_mask = arg_info(op->args[2])->z_mask | z_mask; - return false; + s_mask = ~z_mask; + z_mask |= t2->z_mask; + s_mask &= t2->s_mask; + + return fold_masks_zs(ctx, op, z_mask, s_mask); } static bool fold_ctpop(OptContext *ctx, TCGOp *op) -- 2.43.0 Add fold_masks_z as a trivial wrapper around fold_masks_zs. Avoid the use of the OptContext slots. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 13 ++++++++++--- 1 file changed, 10 insertions(+), 3 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_masks_zs(OptContext *ctx, TCGOp *op, return true; } +static bool fold_masks_z(OptContext *ctx, TCGOp *op, uint64_t z_mask) +{ + return fold_masks_zs(ctx, op, z_mask, 0); +} + static bool fold_masks(OptContext *ctx, TCGOp *op) { return fold_masks_zs(ctx, op, ctx->z_mask, ctx->s_mask); @@ -XXX,XX +XXX,XX @@ static bool fold_count_zeros(OptContext *ctx, TCGOp *op) static bool fold_ctpop(OptContext *ctx, TCGOp *op) { + uint64_t z_mask; + if (fold_const1(ctx, op)) { return true; } switch (ctx->type) { case TCG_TYPE_I32: - ctx->z_mask = 32 | 31; + z_mask = 32 | 31; break; case TCG_TYPE_I64: - ctx->z_mask = 64 | 63; + z_mask = 64 | 63; break; default: g_assert_not_reached(); } - return false; + return fold_masks_z(ctx, op, z_mask); } static bool fold_deposit(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Find TempOptInfo once. When we fold to and, use fold_and. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 35 +++++++++++++++++------------------ 1 file changed, 17 insertions(+), 18 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_ctpop(OptContext *ctx, TCGOp *op) static bool fold_deposit(OptContext *ctx, TCGOp *op) { + TempOptInfo *t1 = arg_info(op->args[1]); + TempOptInfo *t2 = arg_info(op->args[2]); + int ofs = op->args[3]; + int len = op->args[4]; TCGOpcode and_opc; + uint64_t z_mask; - if (arg_is_const(op->args[1]) && arg_is_const(op->args[2])) { - uint64_t t1 = arg_info(op->args[1])->val; - uint64_t t2 = arg_info(op->args[2])->val; - - t1 = deposit64(t1, op->args[3], op->args[4], t2); - return tcg_opt_gen_movi(ctx, op, op->args[0], t1); + if (ti_is_const(t1) && ti_is_const(t2)) { + return tcg_opt_gen_movi(ctx, op, op->args[0], + deposit64(ti_const_val(t1), ofs, len, + ti_const_val(t2))); } switch (ctx->type) { @@ -XXX,XX +XXX,XX @@ static bool fold_deposit(OptContext *ctx, TCGOp *op) } /* Inserting a value into zero at offset 0. */ - if (arg_is_const_val(op->args[1], 0) && op->args[3] == 0) { - uint64_t mask = MAKE_64BIT_MASK(0, op->args[4]); + if (ti_is_const_val(t1, 0) && ofs == 0) { + uint64_t mask = MAKE_64BIT_MASK(0, len); op->opc = and_opc; op->args[1] = op->args[2]; op->args[2] = arg_new_constant(ctx, mask); - ctx->z_mask = mask & arg_info(op->args[1])->z_mask; - return false; + return fold_and(ctx, op); } /* Inserting zero into a value. */ - if (arg_is_const_val(op->args[2], 0)) { - uint64_t mask = deposit64(-1, op->args[3], op->args[4], 0); + if (ti_is_const_val(t2, 0)) { + uint64_t mask = deposit64(-1, ofs, len, 0); op->opc = and_opc; op->args[2] = arg_new_constant(ctx, mask); - ctx->z_mask = mask & arg_info(op->args[1])->z_mask; - return false; + return fold_and(ctx, op); } - ctx->z_mask = deposit64(arg_info(op->args[1])->z_mask, - op->args[3], op->args[4], - arg_info(op->args[2])->z_mask); - return false; + z_mask = deposit64(t1->z_mask, ofs, len, t2->z_mask); + return fold_masks_z(ctx, op, z_mask); } static bool fold_divide(OptContext *ctx, TCGOp *op) -- 2.43.0 The input which overlaps the sign bit of the output can have its input s_mask propagated to the output s_mask. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 14 ++++++++++++-- 1 file changed, 12 insertions(+), 2 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_deposit(OptContext *ctx, TCGOp *op) TempOptInfo *t2 = arg_info(op->args[2]); int ofs = op->args[3]; int len = op->args[4]; + int width; TCGOpcode and_opc; - uint64_t z_mask; + uint64_t z_mask, s_mask; if (ti_is_const(t1) && ti_is_const(t2)) { return tcg_opt_gen_movi(ctx, op, op->args[0], @@ -XXX,XX +XXX,XX @@ static bool fold_deposit(OptContext *ctx, TCGOp *op) switch (ctx->type) { case TCG_TYPE_I32: and_opc = INDEX_op_and_i32; + width = 32; break; case TCG_TYPE_I64: and_opc = INDEX_op_and_i64; + width = 64; break; default: g_assert_not_reached(); @@ -XXX,XX +XXX,XX @@ static bool fold_deposit(OptContext *ctx, TCGOp *op) return fold_and(ctx, op); } + /* The s_mask from the top portion of the deposit is still valid. */ + if (ofs + len == width) { + s_mask = t2->s_mask << ofs; + } else { + s_mask = t1->s_mask & ~MAKE_64BIT_MASK(0, ofs + len); + } + z_mask = deposit64(t1->z_mask, ofs, len, t2->z_mask); - return fold_masks_z(ctx, op, z_mask); + return fold_masks_zs(ctx, op, z_mask, s_mask); } static bool fold_divide(OptContext *ctx, TCGOp *op) -- 2.43.0 Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_divide(OptContext *ctx, TCGOp *op) fold_xi_to_x(ctx, op, 1)) { return true; } - return false; + return finish_folding(ctx, op); } static bool fold_dup(OptContext *ctx, TCGOp *op) -- 2.43.0 Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_dup(OptContext *ctx, TCGOp *op) t = dup_const(TCGOP_VECE(op), t); return tcg_opt_gen_movi(ctx, op, op->args[0], t); } - return false; + return finish_folding(ctx, op); } static bool fold_dup2(OptContext *ctx, TCGOp *op) @@ -XXX,XX +XXX,XX @@ static bool fold_dup2(OptContext *ctx, TCGOp *op) op->opc = INDEX_op_dup_vec; TCGOP_VECE(op) = MO_32; } - return false; + return finish_folding(ctx, op); } static bool fold_eqv(OptContext *ctx, TCGOp *op) -- 2.43.0 Add fold_masks_s as a trivial wrapper around fold_masks_zs. Avoid the use of the OptContext slots. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 13 ++++++++++--- 1 file changed, 10 insertions(+), 3 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_masks_z(OptContext *ctx, TCGOp *op, uint64_t z_mask) return fold_masks_zs(ctx, op, z_mask, 0); } +static bool fold_masks_s(OptContext *ctx, TCGOp *op, uint64_t s_mask) +{ + return fold_masks_zs(ctx, op, -1, s_mask); +} + static bool fold_masks(OptContext *ctx, TCGOp *op) { return fold_masks_zs(ctx, op, ctx->z_mask, ctx->s_mask); @@ -XXX,XX +XXX,XX @@ static bool fold_dup2(OptContext *ctx, TCGOp *op) static bool fold_eqv(OptContext *ctx, TCGOp *op) { + uint64_t s_mask; + if (fold_const2_commutative(ctx, op) || fold_xi_to_x(ctx, op, -1) || fold_xi_to_not(ctx, op, 0)) { return true; } - ctx->s_mask = arg_info(op->args[1])->s_mask - & arg_info(op->args[2])->s_mask; - return false; + s_mask = arg_info(op->args[1])->s_mask + & arg_info(op->args[2])->s_mask; + return fold_masks_s(ctx, op, s_mask); } static bool fold_extract(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Find TempOptInfo once. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 15 ++++++--------- 1 file changed, 6 insertions(+), 9 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_eqv(OptContext *ctx, TCGOp *op) static bool fold_extract(OptContext *ctx, TCGOp *op) { uint64_t z_mask_old, z_mask; + TempOptInfo *t1 = arg_info(op->args[1]); int pos = op->args[2]; int len = op->args[3]; - if (arg_is_const(op->args[1])) { - uint64_t t; - - t = arg_info(op->args[1])->val; - t = extract64(t, pos, len); - return tcg_opt_gen_movi(ctx, op, op->args[0], t); + if (ti_is_const(t1)) { + return tcg_opt_gen_movi(ctx, op, op->args[0], + extract64(ti_const_val(t1), pos, len)); } - z_mask_old = arg_info(op->args[1])->z_mask; + z_mask_old = t1->z_mask; z_mask = extract64(z_mask_old, pos, len); if (pos == 0 && fold_affected_mask(ctx, op, z_mask_old ^ z_mask)) { return true; } - ctx->z_mask = z_mask; - return fold_masks(ctx, op); + return fold_masks_z(ctx, op, z_mask); } static bool fold_extract2(OptContext *ctx, TCGOp *op) -- 2.43.0 Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_extract2(OptContext *ctx, TCGOp *op) } return tcg_opt_gen_movi(ctx, op, op->args[0], v1 | v2); } - return false; + return finish_folding(ctx, op); } static bool fold_exts(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Find TempOptInfo once. Explicitly sign-extend z_mask instead of doing that manually. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 29 ++++++++++++----------------- 1 file changed, 12 insertions(+), 17 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_extract2(OptContext *ctx, TCGOp *op) static bool fold_exts(OptContext *ctx, TCGOp *op) { - uint64_t s_mask_old, s_mask, z_mask, sign; + uint64_t s_mask_old, s_mask, z_mask; bool type_change = false; + TempOptInfo *t1; if (fold_const1(ctx, op)) { return true; } - z_mask = arg_info(op->args[1])->z_mask; - s_mask = arg_info(op->args[1])->s_mask; + t1 = arg_info(op->args[1]); + z_mask = t1->z_mask; + s_mask = t1->s_mask; s_mask_old = s_mask; switch (op->opc) { CASE_OP_32_64(ext8s): - sign = INT8_MIN; - z_mask = (uint8_t)z_mask; + s_mask |= INT8_MIN; + z_mask = (int8_t)z_mask; break; CASE_OP_32_64(ext16s): - sign = INT16_MIN; - z_mask = (uint16_t)z_mask; + s_mask |= INT16_MIN; + z_mask = (int16_t)z_mask; break; case INDEX_op_ext_i32_i64: type_change = true; QEMU_FALLTHROUGH; case INDEX_op_ext32s_i64: - sign = INT32_MIN; - z_mask = (uint32_t)z_mask; + s_mask |= INT32_MIN; + z_mask = (int32_t)z_mask; break; default: g_assert_not_reached(); } - if (z_mask & sign) { - z_mask |= sign; - } - s_mask |= sign << 1; - - ctx->z_mask = z_mask; - ctx->s_mask = s_mask; if (0 && !type_change && fold_affected_mask(ctx, op, s_mask & ~s_mask_old)) { return true; } - return fold_masks(ctx, op); + return fold_masks_zs(ctx, op, z_mask, s_mask); } static bool fold_extu(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_extu(OptContext *ctx, TCGOp *op) g_assert_not_reached(); } - ctx->z_mask = z_mask; if (!type_change && fold_affected_mask(ctx, op, z_mask_old ^ z_mask)) { return true; } - return fold_masks(ctx, op); + + return fold_masks_z(ctx, op, z_mask); } static bool fold_mb(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Find TempOptInfo once. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 19 +++++++++++-------- 1 file changed, 11 insertions(+), 8 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_mov(OptContext *ctx, TCGOp *op) static bool fold_movcond(OptContext *ctx, TCGOp *op) { + uint64_t z_mask, s_mask; + TempOptInfo *tt, *ft; int i; /* If true and false values are the same, eliminate the cmp. */ @@ -XXX,XX +XXX,XX @@ static bool fold_movcond(OptContext *ctx, TCGOp *op) return tcg_opt_gen_mov(ctx, op, op->args[0], op->args[4 - i]); } - ctx->z_mask = arg_info(op->args[3])->z_mask - | arg_info(op->args[4])->z_mask; - ctx->s_mask = arg_info(op->args[3])->s_mask - & arg_info(op->args[4])->s_mask; + tt = arg_info(op->args[3]); + ft = arg_info(op->args[4]); + z_mask = tt->z_mask | ft->z_mask; + s_mask = tt->s_mask & ft->s_mask; - if (arg_is_const(op->args[3]) && arg_is_const(op->args[4])) { - uint64_t tv = arg_info(op->args[3])->val; - uint64_t fv = arg_info(op->args[4])->val; + if (ti_is_const(tt) && ti_is_const(ft)) { + uint64_t tv = ti_const_val(tt); + uint64_t fv = ti_const_val(ft); TCGOpcode opc, negopc = 0; TCGCond cond = op->args[5]; @@ -XXX,XX +XXX,XX @@ static bool fold_movcond(OptContext *ctx, TCGOp *op) } } } - return false; + + return fold_masks_zs(ctx, op, z_mask, s_mask); } static bool fold_mul(OptContext *ctx, TCGOp *op) -- 2.43.0 Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_mul(OptContext *ctx, TCGOp *op) fold_xi_to_x(ctx, op, 1)) { return true; } - return false; + return finish_folding(ctx, op); } static bool fold_mul_highpart(OptContext *ctx, TCGOp *op) @@ -XXX,XX +XXX,XX @@ static bool fold_mul_highpart(OptContext *ctx, TCGOp *op) fold_xi_to_i(ctx, op, 0)) { return true; } - return false; + return finish_folding(ctx, op); } static bool fold_multiply2(OptContext *ctx, TCGOp *op) @@ -XXX,XX +XXX,XX @@ static bool fold_multiply2(OptContext *ctx, TCGOp *op) tcg_opt_gen_movi(ctx, op2, rh, h); return true; } - return false; + return finish_folding(ctx, op); } static bool fold_nand(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_multiply2(OptContext *ctx, TCGOp *op) static bool fold_nand(OptContext *ctx, TCGOp *op) { + uint64_t s_mask; + if (fold_const2_commutative(ctx, op) || fold_xi_to_not(ctx, op, -1)) { return true; } - ctx->s_mask = arg_info(op->args[1])->s_mask - & arg_info(op->args[2])->s_mask; - return false; + s_mask = arg_info(op->args[1])->s_mask + & arg_info(op->args[2])->s_mask; + return fold_masks_s(ctx, op, s_mask); } static bool fold_neg_no_const(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 9 ++------- 1 file changed, 2 insertions(+), 7 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_neg_no_const(OptContext *ctx, TCGOp *op) { /* Set to 1 all bits to the left of the rightmost. */ uint64_t z_mask = arg_info(op->args[1])->z_mask; - ctx->z_mask = -(z_mask & -z_mask); + z_mask = -(z_mask & -z_mask); - /* - * Because of fold_sub_to_neg, we want to always return true, - * via finish_folding. - */ - finish_folding(ctx, op); - return true; + return fold_masks_z(ctx, op, z_mask); } static bool fold_neg(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_neg(OptContext *ctx, TCGOp *op) static bool fold_nor(OptContext *ctx, TCGOp *op) { + uint64_t s_mask; + if (fold_const2_commutative(ctx, op) || fold_xi_to_not(ctx, op, 0)) { return true; } - ctx->s_mask = arg_info(op->args[1])->s_mask - & arg_info(op->args[2])->s_mask; - return false; + s_mask = arg_info(op->args[1])->s_mask + & arg_info(op->args[2])->s_mask; + return fold_masks_s(ctx, op, s_mask); } static bool fold_not(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 7 +------ 1 file changed, 1 insertion(+), 6 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_not(OptContext *ctx, TCGOp *op) if (fold_const1(ctx, op)) { return true; } - - ctx->s_mask = arg_info(op->args[1])->s_mask; - - /* Because of fold_to_not, we want to always return true, via finish. */ - finish_folding(ctx, op); - return true; + return fold_masks_s(ctx, op, arg_info(op->args[1])->s_mask); } static bool fold_or(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Find TempOptInfo once. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 13 ++++++++----- 1 file changed, 8 insertions(+), 5 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_not(OptContext *ctx, TCGOp *op) static bool fold_or(OptContext *ctx, TCGOp *op) { + uint64_t z_mask, s_mask; + TempOptInfo *t1, *t2; + if (fold_const2_commutative(ctx, op) || fold_xi_to_x(ctx, op, 0) || fold_xx_to_x(ctx, op)) { return true; } - ctx->z_mask = arg_info(op->args[1])->z_mask - | arg_info(op->args[2])->z_mask; - ctx->s_mask = arg_info(op->args[1])->s_mask - & arg_info(op->args[2])->s_mask; - return fold_masks(ctx, op); + t1 = arg_info(op->args[1]); + t2 = arg_info(op->args[2]); + z_mask = t1->z_mask | t2->z_mask; + s_mask = t1->s_mask & t2->s_mask; + return fold_masks_zs(ctx, op, z_mask, s_mask); } static bool fold_orc(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_or(OptContext *ctx, TCGOp *op) static bool fold_orc(OptContext *ctx, TCGOp *op) { + uint64_t s_mask; + if (fold_const2(ctx, op) || fold_xx_to_i(ctx, op, -1) || fold_xi_to_x(ctx, op, -1) || @@ -XXX,XX +XXX,XX @@ static bool fold_orc(OptContext *ctx, TCGOp *op) return true; } - ctx->s_mask = arg_info(op->args[1])->s_mask - & arg_info(op->args[2])->s_mask; - return false; + s_mask = arg_info(op->args[1])->s_mask + & arg_info(op->args[2])->s_mask; + return fold_masks_s(ctx, op, s_mask); } static bool fold_qemu_ld(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Be careful not to call fold_masks_zs when the memory operation is wide enough to require multiple outputs, so split into two functions: fold_qemu_ld_1reg and fold_qemu_ld_2reg. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 26 +++++++++++++++++++++----- 1 file changed, 21 insertions(+), 5 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_orc(OptContext *ctx, TCGOp *op) return fold_masks_s(ctx, op, s_mask); } -static bool fold_qemu_ld(OptContext *ctx, TCGOp *op) +static bool fold_qemu_ld_1reg(OptContext *ctx, TCGOp *op) { const TCGOpDef *def = &tcg_op_defs[op->opc]; MemOpIdx oi = op->args[def->nb_oargs + def->nb_iargs]; MemOp mop = get_memop(oi); int width = 8 * memop_size(mop); + uint64_t z_mask = -1, s_mask = 0; if (width < 64) { if (mop & MO_SIGN) { - ctx->s_mask = MAKE_64BIT_MASK(width, 64 - width); + s_mask = MAKE_64BIT_MASK(width - 1, 64 - (width - 1)); } else { - ctx->z_mask = MAKE_64BIT_MASK(0, width); + z_mask = MAKE_64BIT_MASK(0, width); } } /* Opcodes that touch guest memory stop the mb optimization. */ ctx->prev_mb = NULL; - return false; + + return fold_masks_zs(ctx, op, z_mask, s_mask); +} + +static bool fold_qemu_ld_2reg(OptContext *ctx, TCGOp *op) +{ + /* Opcodes that touch guest memory stop the mb optimization. */ + ctx->prev_mb = NULL; + return finish_folding(ctx, op); } static bool fold_qemu_st(OptContext *ctx, TCGOp *op) @@ -XXX,XX +XXX,XX @@ void tcg_optimize(TCGContext *s) break; case INDEX_op_qemu_ld_a32_i32: case INDEX_op_qemu_ld_a64_i32: + done = fold_qemu_ld_1reg(&ctx, op); + break; case INDEX_op_qemu_ld_a32_i64: case INDEX_op_qemu_ld_a64_i64: + if (TCG_TARGET_REG_BITS == 64) { + done = fold_qemu_ld_1reg(&ctx, op); + break; + } + QEMU_FALLTHROUGH; case INDEX_op_qemu_ld_a32_i128: case INDEX_op_qemu_ld_a64_i128: - done = fold_qemu_ld(&ctx, op); + done = fold_qemu_ld_2reg(&ctx, op); break; case INDEX_op_qemu_st8_a32_i32: case INDEX_op_qemu_st8_a64_i32: -- 2.43.0 Stores have no output operands, and so need no further work. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 11 +++++------ 1 file changed, 5 insertions(+), 6 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_qemu_st(OptContext *ctx, TCGOp *op) { /* Opcodes that touch guest memory stop the mb optimization. */ ctx->prev_mb = NULL; - return false; + return true; } static bool fold_remainder(OptContext *ctx, TCGOp *op) @@ -XXX,XX +XXX,XX @@ static bool fold_tcg_st(OptContext *ctx, TCGOp *op) if (op->args[1] != tcgv_ptr_arg(tcg_env)) { remove_mem_copy_all(ctx); - return false; + return true; } switch (op->opc) { @@ -XXX,XX +XXX,XX @@ static bool fold_tcg_st(OptContext *ctx, TCGOp *op) g_assert_not_reached(); } remove_mem_copy_in(ctx, ofs, ofs + lm1); - return false; + return true; } static bool fold_tcg_st_memcopy(OptContext *ctx, TCGOp *op) @@ -XXX,XX +XXX,XX @@ static bool fold_tcg_st_memcopy(OptContext *ctx, TCGOp *op) TCGType type; if (op->args[1] != tcgv_ptr_arg(tcg_env)) { - fold_tcg_st(ctx, op); - return false; + return fold_tcg_st(ctx, op); } src = arg_temp(op->args[0]); @@ -XXX,XX +XXX,XX @@ static bool fold_tcg_st_memcopy(OptContext *ctx, TCGOp *op) last = ofs + tcg_type_size(type) - 1; remove_mem_copy_in(ctx, ofs, last); record_mem_copy(ctx, type, src, ofs, last); - return false; + return true; } static bool fold_xor(OptContext *ctx, TCGOp *op) -- 2.43.0 Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_remainder(OptContext *ctx, TCGOp *op) fold_xx_to_i(ctx, op, 0)) { return true; } - return false; + return finish_folding(ctx, op); } static bool fold_setcond_zmask(OptContext *ctx, TCGOp *op, bool neg) -- 2.43.0 Change return from bool to int; distinguish between complete folding, simplification, and no change. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 22 ++++++++++++++-------- 1 file changed, 14 insertions(+), 8 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_remainder(OptContext *ctx, TCGOp *op) return finish_folding(ctx, op); } -static bool fold_setcond_zmask(OptContext *ctx, TCGOp *op, bool neg) +/* Return 1 if finished, -1 if simplified, 0 if unchanged. */ +static int fold_setcond_zmask(OptContext *ctx, TCGOp *op, bool neg) { uint64_t a_zmask, b_val; TCGCond cond; @@ -XXX,XX +XXX,XX @@ static bool fold_setcond_zmask(OptContext *ctx, TCGOp *op, bool neg) op->opc = xor_opc; op->args[2] = arg_new_constant(ctx, 1); } - return false; + return -1; } } - - return false; + return 0; } static void fold_setcond_tst_pow2(OptContext *ctx, TCGOp *op, bool neg) @@ -XXX,XX +XXX,XX @@ static bool fold_setcond(OptContext *ctx, TCGOp *op) return tcg_opt_gen_movi(ctx, op, op->args[0], i); } - if (fold_setcond_zmask(ctx, op, false)) { + i = fold_setcond_zmask(ctx, op, false); + if (i > 0) { return true; } - fold_setcond_tst_pow2(ctx, op, false); + if (i == 0) { + fold_setcond_tst_pow2(ctx, op, false); + } ctx->z_mask = 1; return false; @@ -XXX,XX +XXX,XX @@ static bool fold_negsetcond(OptContext *ctx, TCGOp *op) return tcg_opt_gen_movi(ctx, op, op->args[0], -i); } - if (fold_setcond_zmask(ctx, op, true)) { + i = fold_setcond_zmask(ctx, op, true); + if (i > 0) { return true; } - fold_setcond_tst_pow2(ctx, op, true); + if (i == 0) { + fold_setcond_tst_pow2(ctx, op, true); + } /* Value is {0,-1} so all bits are repetitions of the sign. */ ctx->s_mask = -1; -- 2.43.0 Avoid the use of the OptContext slots. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_setcond(OptContext *ctx, TCGOp *op) fold_setcond_tst_pow2(ctx, op, false); } - ctx->z_mask = 1; - return false; + return fold_masks_z(ctx, op, 1); } static bool fold_negsetcond(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_negsetcond(OptContext *ctx, TCGOp *op) } /* Value is {0,-1} so all bits are repetitions of the sign. */ - ctx->s_mask = -1; - return false; + return fold_masks_s(ctx, op, -1); } static bool fold_setcond2(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_setcond2(OptContext *ctx, TCGOp *op) return fold_setcond(ctx, op); } - ctx->z_mask = 1; - return false; + return fold_masks_z(ctx, op, 1); do_setcond_const: return tcg_opt_gen_movi(ctx, op, op->args[0], i); -- 2.43.0 Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_cmp_vec(OptContext *ctx, TCGOp *op) if (swap_commutative(NO_DEST, &op->args[1], &op->args[2])) { op->args[3] = tcg_swap_cond(op->args[3]); } - return false; + return finish_folding(ctx, op); } static bool fold_cmpsel_vec(OptContext *ctx, TCGOp *op) -- 2.43.0 Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_cmpsel_vec(OptContext *ctx, TCGOp *op) if (swap_commutative(op->args[0], &op->args[4], &op->args[3])) { op->args[5] = tcg_invert_cond(op->args[5]); } - return false; + return finish_folding(ctx, op); } static bool fold_sextract(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Find TempOptInfo once. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 24 +++++++++--------------- 1 file changed, 9 insertions(+), 15 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_cmpsel_vec(OptContext *ctx, TCGOp *op) static bool fold_sextract(OptContext *ctx, TCGOp *op) { uint64_t z_mask, s_mask, s_mask_old; + TempOptInfo *t1 = arg_info(op->args[1]); int pos = op->args[2]; int len = op->args[3]; - if (arg_is_const(op->args[1])) { - uint64_t t; - - t = arg_info(op->args[1])->val; - t = sextract64(t, pos, len); - return tcg_opt_gen_movi(ctx, op, op->args[0], t); + if (ti_is_const(t1)) { + return tcg_opt_gen_movi(ctx, op, op->args[0], + sextract64(ti_const_val(t1), pos, len)); } - z_mask = arg_info(op->args[1])->z_mask; - z_mask = sextract64(z_mask, pos, len); - ctx->z_mask = z_mask; - - s_mask_old = arg_info(op->args[1])->s_mask; - s_mask = sextract64(s_mask_old, pos, len); - s_mask |= MAKE_64BIT_MASK(len, 64 - len); - ctx->s_mask = s_mask; + s_mask_old = t1->s_mask; + s_mask = s_mask_old >> pos; + s_mask |= -1ull << (len - 1); if (0 && pos == 0 && fold_affected_mask(ctx, op, s_mask & ~s_mask_old)) { return true; } - return fold_masks(ctx, op); + z_mask = sextract64(t1->z_mask, pos, len); + return fold_masks_zs(ctx, op, z_mask, s_mask); } static bool fold_shift(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Find TempOptInfo once. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 27 ++++++++++++++------------- 1 file changed, 14 insertions(+), 13 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_sextract(OptContext *ctx, TCGOp *op) static bool fold_shift(OptContext *ctx, TCGOp *op) { uint64_t s_mask, z_mask, sign; + TempOptInfo *t1, *t2; if (fold_const2(ctx, op) || fold_ix_to_i(ctx, op, 0) || @@ -XXX,XX +XXX,XX @@ static bool fold_shift(OptContext *ctx, TCGOp *op) return true; } - s_mask = arg_info(op->args[1])->s_mask; - z_mask = arg_info(op->args[1])->z_mask; + t1 = arg_info(op->args[1]); + t2 = arg_info(op->args[2]); + s_mask = t1->s_mask; + z_mask = t1->z_mask; - if (arg_is_const(op->args[2])) { - int sh = arg_info(op->args[2])->val; - - ctx->z_mask = do_constant_folding(op->opc, ctx->type, z_mask, sh); + if (ti_is_const(t2)) { + int sh = ti_const_val(t2); + z_mask = do_constant_folding(op->opc, ctx->type, z_mask, sh); s_mask = do_constant_folding(op->opc, ctx->type, s_mask, sh); - return fold_masks(ctx, op); + return fold_masks_zs(ctx, op, z_mask, s_mask); } switch (op->opc) { @@ -XXX,XX +XXX,XX @@ static bool fold_shift(OptContext *ctx, TCGOp *op) * Arithmetic right shift will not reduce the number of * input sign repetitions. */ - ctx->s_mask = s_mask; - break; + return fold_masks_s(ctx, op, s_mask); CASE_OP_32_64(shr): /* * If the sign bit is known zero, then logical right shift - * will not reduced the number of input sign repetitions. + * will not reduce the number of input sign repetitions. */ - sign = (s_mask & -s_mask) >> 1; + sign = -s_mask; if (sign && !(z_mask & sign)) { - ctx->s_mask = s_mask; + return fold_masks_s(ctx, op, s_mask); } break; default: break; } - return false; + return finish_folding(ctx, op); } static bool fold_sub_to_neg(OptContext *ctx, TCGOp *op) -- 2.43.0 Merge the two conditions, sign != 0 && !(z_mask & sign), by testing ~z_mask & sign. If sign == 0, the logical and will produce false. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_sextract(OptContext *ctx, TCGOp *op) static bool fold_shift(OptContext *ctx, TCGOp *op) { - uint64_t s_mask, z_mask, sign; + uint64_t s_mask, z_mask; TempOptInfo *t1, *t2; if (fold_const2(ctx, op) || @@ -XXX,XX +XXX,XX @@ static bool fold_shift(OptContext *ctx, TCGOp *op) * If the sign bit is known zero, then logical right shift * will not reduce the number of input sign repetitions. */ - sign = -s_mask; - if (sign && !(z_mask & sign)) { + if (~z_mask & -s_mask) { return fold_masks_s(ctx, op, s_mask); } break; -- 2.43.0 Duplicate fold_sub_vec into fold_sub instead of calling it, now that fold_sub_vec always returns true. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_sub_vec(OptContext *ctx, TCGOp *op) fold_sub_to_neg(ctx, op)) { return true; } - return false; + return finish_folding(ctx, op); } static bool fold_sub(OptContext *ctx, TCGOp *op) { - if (fold_const2(ctx, op) || fold_sub_vec(ctx, op)) { + if (fold_const2(ctx, op) || + fold_xx_to_i(ctx, op, 0) || + fold_xi_to_x(ctx, op, 0) || + fold_sub_to_neg(ctx, op)) { return true; } @@ -XXX,XX +XXX,XX @@ static bool fold_sub(OptContext *ctx, TCGOp *op) ? INDEX_op_add_i32 : INDEX_op_add_i64); op->args[2] = arg_new_constant(ctx, -val); } - return false; + return finish_folding(ctx, op); } static bool fold_sub2(OptContext *ctx, TCGOp *op) -- 2.43.0 Avoid the use of the OptContext slots. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 16 +++++++++------- 1 file changed, 9 insertions(+), 7 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_sub2(OptContext *ctx, TCGOp *op) static bool fold_tcg_ld(OptContext *ctx, TCGOp *op) { + uint64_t z_mask = -1, s_mask = 0; + /* We can't do any folding with a load, but we can record bits. */ switch (op->opc) { CASE_OP_32_64(ld8s): - ctx->s_mask = MAKE_64BIT_MASK(8, 56); + s_mask = INT8_MIN; break; CASE_OP_32_64(ld8u): - ctx->z_mask = MAKE_64BIT_MASK(0, 8); + z_mask = MAKE_64BIT_MASK(0, 8); break; CASE_OP_32_64(ld16s): - ctx->s_mask = MAKE_64BIT_MASK(16, 48); + s_mask = INT16_MIN; break; CASE_OP_32_64(ld16u): - ctx->z_mask = MAKE_64BIT_MASK(0, 16); + z_mask = MAKE_64BIT_MASK(0, 16); break; case INDEX_op_ld32s_i64: - ctx->s_mask = MAKE_64BIT_MASK(32, 32); + s_mask = INT32_MIN; break; case INDEX_op_ld32u_i64: - ctx->z_mask = MAKE_64BIT_MASK(0, 32); + z_mask = MAKE_64BIT_MASK(0, 32); break; default: g_assert_not_reached(); } - return false; + return fold_masks_zs(ctx, op, z_mask, s_mask); } static bool fold_tcg_ld_memcopy(OptContext *ctx, TCGOp *op) -- 2.43.0 Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_tcg_ld_memcopy(OptContext *ctx, TCGOp *op) TCGType type; if (op->args[1] != tcgv_ptr_arg(tcg_env)) { - return false; + return finish_folding(ctx, op); } type = ctx->type; -- 2.43.0 Avoid the use of the OptContext slots. Find TempOptInfo once. Remove fold_masks as the function becomes unused. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 18 ++++++++---------- 1 file changed, 8 insertions(+), 10 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_masks_s(OptContext *ctx, TCGOp *op, uint64_t s_mask) return fold_masks_zs(ctx, op, -1, s_mask); } -static bool fold_masks(OptContext *ctx, TCGOp *op) -{ - return fold_masks_zs(ctx, op, ctx->z_mask, ctx->s_mask); -} - /* * An "affected" mask bit is 0 if and only if the result is identical * to the first input. Thus if the entire mask is 0, the operation @@ -XXX,XX +XXX,XX @@ static bool fold_tcg_st_memcopy(OptContext *ctx, TCGOp *op) static bool fold_xor(OptContext *ctx, TCGOp *op) { + uint64_t z_mask, s_mask; + TempOptInfo *t1, *t2; + if (fold_const2_commutative(ctx, op) || fold_xx_to_i(ctx, op, 0) || fold_xi_to_x(ctx, op, 0) || @@ -XXX,XX +XXX,XX @@ static bool fold_xor(OptContext *ctx, TCGOp *op) return true; } - ctx->z_mask = arg_info(op->args[1])->z_mask - | arg_info(op->args[2])->z_mask; - ctx->s_mask = arg_info(op->args[1])->s_mask - & arg_info(op->args[2])->s_mask; - return fold_masks(ctx, op); + t1 = arg_info(op->args[1]); + t2 = arg_info(op->args[2]); + z_mask = t1->z_mask | t2->z_mask; + s_mask = t1->s_mask & t2->s_mask; + return fold_masks_zs(ctx, op, z_mask, s_mask); } static bool fold_bitsel_vec(OptContext *ctx, TCGOp *op) -- 2.43.0 Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_bitsel_vec(OptContext *ctx, TCGOp *op) return fold_orc(ctx, op); } } - return false; + return finish_folding(ctx, op); } /* Propagate constants and copies, fold constant expressions. */ -- 2.43.0 All non-default cases now finish folding within each function. Do the same with the default case and assert it is done after. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ void tcg_optimize(TCGContext *s) done = true; break; default: + done = finish_folding(&ctx, op); break; } - - if (!done) { - finish_folding(&ctx, op); - } + tcg_debug_assert(done); } } -- 2.43.0 All mask setting is now done with parameters via fold_masks_*. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 13 ------------- 1 file changed, 13 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ typedef struct OptContext { QSIMPLEQ_HEAD(, MemCopyInfo) mem_free; /* In flight values from optimization. */ - uint64_t z_mask; /* mask bit is 0 iff value bit is 0 */ - uint64_t s_mask; /* mask bit is 1 if value bit matches msb */ TCGType type; } OptContext; @@ -XXX,XX +XXX,XX @@ static bool finish_folding(OptContext *ctx, TCGOp *op) for (i = 0; i < nb_oargs; i++) { TCGTemp *ts = arg_temp(op->args[i]); reset_ts(ctx, ts); - /* - * Save the corresponding known-zero/sign bits mask for the - * first output argument (only one supported so far). - */ - if (i == 0) { - ts_info(ts)->z_mask = ctx->z_mask; - } } return true; } @@ -XXX,XX +XXX,XX @@ void tcg_optimize(TCGContext *s) ctx.type = TCG_TYPE_I32; } - /* Assume all bits affected, no bits known zero, no sign reps. */ - ctx.z_mask = -1; - ctx.s_mask = 0; - /* * Process each opcode. * Sorted alphabetically by opcode as much as possible. -- 2.43.0 All instances of s_mask have been converted to the new representation. We can now re-enable usage. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_exts(OptContext *ctx, TCGOp *op) g_assert_not_reached(); } - if (0 && !type_change && fold_affected_mask(ctx, op, s_mask & ~s_mask_old)) { + if (!type_change && fold_affected_mask(ctx, op, s_mask & ~s_mask_old)) { return true; } @@ -XXX,XX +XXX,XX @@ static bool fold_sextract(OptContext *ctx, TCGOp *op) s_mask = s_mask_old >> pos; s_mask |= -1ull << (len - 1); - if (0 && pos == 0 && fold_affected_mask(ctx, op, s_mask & ~s_mask_old)) { + if (pos == 0 && fold_affected_mask(ctx, op, s_mask & ~s_mask_old)) { return true; } -- 2.43.0 The big comment just above says functions should be sorted. Add forward declarations as needed. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 114 +++++++++++++++++++++++++------------------------ 1 file changed, 59 insertions(+), 55 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_xx_to_x(OptContext *ctx, TCGOp *op) * 3) those that produce information about the result value. */ +static bool fold_or(OptContext *ctx, TCGOp *op); +static bool fold_orc(OptContext *ctx, TCGOp *op); +static bool fold_xor(OptContext *ctx, TCGOp *op); + static bool fold_add(OptContext *ctx, TCGOp *op) { if (fold_const2_commutative(ctx, op) || @@ -XXX,XX +XXX,XX @@ static bool fold_andc(OptContext *ctx, TCGOp *op) return fold_masks_zs(ctx, op, z_mask, s_mask); } +static bool fold_bitsel_vec(OptContext *ctx, TCGOp *op) +{ + /* If true and false values are the same, eliminate the cmp. */ + if (args_are_copies(op->args[2], op->args[3])) { + return tcg_opt_gen_mov(ctx, op, op->args[0], op->args[2]); + } + + if (arg_is_const(op->args[2]) && arg_is_const(op->args[3])) { + uint64_t tv = arg_info(op->args[2])->val; + uint64_t fv = arg_info(op->args[3])->val; + + if (tv == -1 && fv == 0) { + return tcg_opt_gen_mov(ctx, op, op->args[0], op->args[1]); + } + if (tv == 0 && fv == -1) { + if (TCG_TARGET_HAS_not_vec) { + op->opc = INDEX_op_not_vec; + return fold_not(ctx, op); + } else { + op->opc = INDEX_op_xor_vec; + op->args[2] = arg_new_constant(ctx, -1); + return fold_xor(ctx, op); + } + } + } + if (arg_is_const(op->args[2])) { + uint64_t tv = arg_info(op->args[2])->val; + if (tv == -1) { + op->opc = INDEX_op_or_vec; + op->args[2] = op->args[3]; + return fold_or(ctx, op); + } + if (tv == 0 && TCG_TARGET_HAS_andc_vec) { + op->opc = INDEX_op_andc_vec; + op->args[2] = op->args[1]; + op->args[1] = op->args[3]; + return fold_andc(ctx, op); + } + } + if (arg_is_const(op->args[3])) { + uint64_t fv = arg_info(op->args[3])->val; + if (fv == 0) { + op->opc = INDEX_op_and_vec; + return fold_and(ctx, op); + } + if (fv == -1 && TCG_TARGET_HAS_orc_vec) { + op->opc = INDEX_op_orc_vec; + op->args[2] = op->args[1]; + op->args[1] = op->args[3]; + return fold_orc(ctx, op); + } + } + return finish_folding(ctx, op); +} + static bool fold_brcond(OptContext *ctx, TCGOp *op) { int i = do_constant_folding_cond1(ctx, op, NO_DEST, &op->args[0], @@ -XXX,XX +XXX,XX @@ static bool fold_xor(OptContext *ctx, TCGOp *op) return fold_masks_zs(ctx, op, z_mask, s_mask); } -static bool fold_bitsel_vec(OptContext *ctx, TCGOp *op) -{ - /* If true and false values are the same, eliminate the cmp. */ - if (args_are_copies(op->args[2], op->args[3])) { - return tcg_opt_gen_mov(ctx, op, op->args[0], op->args[2]); - } - - if (arg_is_const(op->args[2]) && arg_is_const(op->args[3])) { - uint64_t tv = arg_info(op->args[2])->val; - uint64_t fv = arg_info(op->args[3])->val; - - if (tv == -1 && fv == 0) { - return tcg_opt_gen_mov(ctx, op, op->args[0], op->args[1]); - } - if (tv == 0 && fv == -1) { - if (TCG_TARGET_HAS_not_vec) { - op->opc = INDEX_op_not_vec; - return fold_not(ctx, op); - } else { - op->opc = INDEX_op_xor_vec; - op->args[2] = arg_new_constant(ctx, -1); - return fold_xor(ctx, op); - } - } - } - if (arg_is_const(op->args[2])) { - uint64_t tv = arg_info(op->args[2])->val; - if (tv == -1) { - op->opc = INDEX_op_or_vec; - op->args[2] = op->args[3]; - return fold_or(ctx, op); - } - if (tv == 0 && TCG_TARGET_HAS_andc_vec) { - op->opc = INDEX_op_andc_vec; - op->args[2] = op->args[1]; - op->args[1] = op->args[3]; - return fold_andc(ctx, op); - } - } - if (arg_is_const(op->args[3])) { - uint64_t fv = arg_info(op->args[3])->val; - if (fv == 0) { - op->opc = INDEX_op_and_vec; - return fold_and(ctx, op); - } - if (fv == -1 && TCG_TARGET_HAS_orc_vec) { - op->opc = INDEX_op_orc_vec; - op->args[2] = op->args[1]; - op->args[1] = op->args[3]; - return fold_orc(ctx, op); - } - } - return finish_folding(ctx, op); -} - /* Propagate constants and copies, fold constant expressions. */ void tcg_optimize(TCGContext *s) { -- 2.43.0 The big comment just above says functions should be sorted. Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/optimize.c | 60 +++++++++++++++++++++++++------------------------- 1 file changed, 30 insertions(+), 30 deletions(-) diff --git a/tcg/optimize.c b/tcg/optimize.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/optimize.c +++ b/tcg/optimize.c @@ -XXX,XX +XXX,XX @@ static bool fold_call(OptContext *ctx, TCGOp *op) return true; } +static bool fold_cmp_vec(OptContext *ctx, TCGOp *op) +{ + /* Canonicalize the comparison to put immediate second. */ + if (swap_commutative(NO_DEST, &op->args[1], &op->args[2])) { + op->args[3] = tcg_swap_cond(op->args[3]); + } + return finish_folding(ctx, op); +} + +static bool fold_cmpsel_vec(OptContext *ctx, TCGOp *op) +{ + /* If true and false values are the same, eliminate the cmp. */ + if (args_are_copies(op->args[3], op->args[4])) { + return tcg_opt_gen_mov(ctx, op, op->args[0], op->args[3]); + } + + /* Canonicalize the comparison to put immediate second. */ + if (swap_commutative(NO_DEST, &op->args[1], &op->args[2])) { + op->args[5] = tcg_swap_cond(op->args[5]); + } + /* + * Canonicalize the "false" input reg to match the destination, + * so that the tcg backend can implement "move if true". + */ + if (swap_commutative(op->args[0], &op->args[4], &op->args[3])) { + op->args[5] = tcg_invert_cond(op->args[5]); + } + return finish_folding(ctx, op); +} + static bool fold_count_zeros(OptContext *ctx, TCGOp *op) { uint64_t z_mask, s_mask; @@ -XXX,XX +XXX,XX @@ static bool fold_setcond2(OptContext *ctx, TCGOp *op) return tcg_opt_gen_movi(ctx, op, op->args[0], i); } -static bool fold_cmp_vec(OptContext *ctx, TCGOp *op) -{ - /* Canonicalize the comparison to put immediate second. */ - if (swap_commutative(NO_DEST, &op->args[1], &op->args[2])) { - op->args[3] = tcg_swap_cond(op->args[3]); - } - return finish_folding(ctx, op); -} - -static bool fold_cmpsel_vec(OptContext *ctx, TCGOp *op) -{ - /* If true and false values are the same, eliminate the cmp. */ - if (args_are_copies(op->args[3], op->args[4])) { - return tcg_opt_gen_mov(ctx, op, op->args[0], op->args[3]); - } - - /* Canonicalize the comparison to put immediate second. */ - if (swap_commutative(NO_DEST, &op->args[1], &op->args[2])) { - op->args[5] = tcg_swap_cond(op->args[5]); - } - /* - * Canonicalize the "false" input reg to match the destination, - * so that the tcg backend can implement "move if true". - */ - if (swap_commutative(op->args[0], &op->args[4], &op->args[3])) { - op->args[5] = tcg_invert_cond(op->args[5]); - } - return finish_folding(ctx, op); -} - static bool fold_sextract(OptContext *ctx, TCGOp *op) { uint64_t z_mask, s_mask, s_mask_old; -- 2.43.0 We currently have a flag, float_muladd_halve_result, to scale the result by 2**-1. Extend this to handle arbitrary scaling. Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- include/fpu/softfloat.h | 6 ++++ fpu/softfloat.c | 58 ++++++++++++++++++++++----------------- fpu/softfloat-parts.c.inc | 7 +++-- 3 files changed, 44 insertions(+), 27 deletions(-) diff --git a/include/fpu/softfloat.h b/include/fpu/softfloat.h index XXXXXXX..XXXXXXX 100644 --- a/include/fpu/softfloat.h +++ b/include/fpu/softfloat.h @@ -XXX,XX +XXX,XX @@ float16 float16_add(float16, float16, float_status *status); float16 float16_sub(float16, float16, float_status *status); float16 float16_mul(float16, float16, float_status *status); float16 float16_muladd(float16, float16, float16, int, float_status *status); +float16 float16_muladd_scalbn(float16, float16, float16, + int, int, float_status *status); float16 float16_div(float16, float16, float_status *status); float16 float16_scalbn(float16, int, float_status *status); float16 float16_min(float16, float16, float_status *status); @@ -XXX,XX +XXX,XX @@ float32 float32_mul(float32, float32, float_status *status); float32 float32_div(float32, float32, float_status *status); float32 float32_rem(float32, float32, float_status *status); float32 float32_muladd(float32, float32, float32, int, float_status *status); +float32 float32_muladd_scalbn(float32, float32, float32, + int, int, float_status *status); float32 float32_sqrt(float32, float_status *status); float32 float32_exp2(float32, float_status *status); float32 float32_log2(float32, float_status *status); @@ -XXX,XX +XXX,XX @@ float64 float64_mul(float64, float64, float_status *status); float64 float64_div(float64, float64, float_status *status); float64 float64_rem(float64, float64, float_status *status); float64 float64_muladd(float64, float64, float64, int, float_status *status); +float64 float64_muladd_scalbn(float64, float64, float64, + int, int, float_status *status); float64 float64_sqrt(float64, float_status *status); float64 float64_log2(float64, float_status *status); FloatRelation float64_compare(float64, float64, float_status *status); diff --git a/fpu/softfloat.c b/fpu/softfloat.c index XXXXXXX..XXXXXXX 100644 --- a/fpu/softfloat.c +++ b/fpu/softfloat.c @@ -XXX,XX +XXX,XX @@ static FloatParts128 *parts128_mul(FloatParts128 *a, FloatParts128 *b, #define parts_mul(A, B, S) \ PARTS_GENERIC_64_128(mul, A)(A, B, S) -static FloatParts64 *parts64_muladd(FloatParts64 *a, FloatParts64 *b, - FloatParts64 *c, int flags, - float_status *s); -static FloatParts128 *parts128_muladd(FloatParts128 *a, FloatParts128 *b, - FloatParts128 *c, int flags, - float_status *s); +static FloatParts64 *parts64_muladd_scalbn(FloatParts64 *a, FloatParts64 *b, + FloatParts64 *c, int scale, + int flags, float_status *s); +static FloatParts128 *parts128_muladd_scalbn(FloatParts128 *a, FloatParts128 *b, + FloatParts128 *c, int scale, + int flags, float_status *s); -#define parts_muladd(A, B, C, Z, S) \ - PARTS_GENERIC_64_128(muladd, A)(A, B, C, Z, S) +#define parts_muladd_scalbn(A, B, C, Z, Y, S) \ + PARTS_GENERIC_64_128(muladd_scalbn, A)(A, B, C, Z, Y, S) static FloatParts64 *parts64_div(FloatParts64 *a, FloatParts64 *b, float_status *s); @@ -XXX,XX +XXX,XX @@ floatx80_mul(floatx80 a, floatx80 b, float_status *status) * Fused multiply-add */ -float16 QEMU_FLATTEN float16_muladd(float16 a, float16 b, float16 c, - int flags, float_status *status) +float16 QEMU_FLATTEN +float16_muladd_scalbn(float16 a, float16 b, float16 c, + int scale, int flags, float_status *status) { FloatParts64 pa, pb, pc, *pr; float16_unpack_canonical(&pa, a, status); float16_unpack_canonical(&pb, b, status); float16_unpack_canonical(&pc, c, status); - pr = parts_muladd(&pa, &pb, &pc, flags, status); + pr = parts_muladd_scalbn(&pa, &pb, &pc, scale, flags, status); return float16_round_pack_canonical(pr, status); } -static float32 QEMU_SOFTFLOAT_ATTR -soft_f32_muladd(float32 a, float32 b, float32 c, int flags, - float_status *status) +float16 float16_muladd(float16 a, float16 b, float16 c, + int flags, float_status *status) +{ + return float16_muladd_scalbn(a, b, c, 0, flags, status); +} + +float32 QEMU_SOFTFLOAT_ATTR +float32_muladd_scalbn(float32 a, float32 b, float32 c, + int scale, int flags, float_status *status) { FloatParts64 pa, pb, pc, *pr; float32_unpack_canonical(&pa, a, status); float32_unpack_canonical(&pb, b, status); float32_unpack_canonical(&pc, c, status); - pr = parts_muladd(&pa, &pb, &pc, flags, status); + pr = parts_muladd_scalbn(&pa, &pb, &pc, scale, flags, status); return float32_round_pack_canonical(pr, status); } -static float64 QEMU_SOFTFLOAT_ATTR -soft_f64_muladd(float64 a, float64 b, float64 c, int flags, - float_status *status) +float64 QEMU_SOFTFLOAT_ATTR +float64_muladd_scalbn(float64 a, float64 b, float64 c, + int scale, int flags, float_status *status) { FloatParts64 pa, pb, pc, *pr; float64_unpack_canonical(&pa, a, status); float64_unpack_canonical(&pb, b, status); float64_unpack_canonical(&pc, c, status); - pr = parts_muladd(&pa, &pb, &pc, flags, status); + pr = parts_muladd_scalbn(&pa, &pb, &pc, scale, flags, status); return float64_round_pack_canonical(pr, status); } @@ -XXX,XX +XXX,XX @@ float32_muladd(float32 xa, float32 xb, float32 xc, int flags, float_status *s) return ur.s; soft: - return soft_f32_muladd(ua.s, ub.s, uc.s, flags, s); + return float32_muladd_scalbn(ua.s, ub.s, uc.s, 0, flags, s); } float64 QEMU_FLATTEN @@ -XXX,XX +XXX,XX @@ float64_muladd(float64 xa, float64 xb, float64 xc, int flags, float_status *s) return ur.s; soft: - return soft_f64_muladd(ua.s, ub.s, uc.s, flags, s); + return float64_muladd_scalbn(ua.s, ub.s, uc.s, 0, flags, s); } float64 float64r32_muladd(float64 a, float64 b, float64 c, @@ -XXX,XX +XXX,XX @@ float64 float64r32_muladd(float64 a, float64 b, float64 c, float64_unpack_canonical(&pa, a, status); float64_unpack_canonical(&pb, b, status); float64_unpack_canonical(&pc, c, status); - pr = parts_muladd(&pa, &pb, &pc, flags, status); + pr = parts_muladd_scalbn(&pa, &pb, &pc, 0, flags, status); return float64r32_round_pack_canonical(pr, status); } @@ -XXX,XX +XXX,XX @@ bfloat16 QEMU_FLATTEN bfloat16_muladd(bfloat16 a, bfloat16 b, bfloat16 c, bfloat16_unpack_canonical(&pa, a, status); bfloat16_unpack_canonical(&pb, b, status); bfloat16_unpack_canonical(&pc, c, status); - pr = parts_muladd(&pa, &pb, &pc, flags, status); + pr = parts_muladd_scalbn(&pa, &pb, &pc, 0, flags, status); return bfloat16_round_pack_canonical(pr, status); } @@ -XXX,XX +XXX,XX @@ float128 QEMU_FLATTEN float128_muladd(float128 a, float128 b, float128 c, float128_unpack_canonical(&pa, a, status); float128_unpack_canonical(&pb, b, status); float128_unpack_canonical(&pc, c, status); - pr = parts_muladd(&pa, &pb, &pc, flags, status); + pr = parts_muladd_scalbn(&pa, &pb, &pc, 0, flags, status); return float128_round_pack_canonical(pr, status); } @@ -XXX,XX +XXX,XX @@ float32 float32_exp2(float32 a, float_status *status) float64_unpack_canonical(&rp, float64_one, status); for (i = 0 ; i < 15 ; i++) { + float64_unpack_canonical(&tp, float32_exp2_coefficients[i], status); - rp = *parts_muladd(&tp, &xnp, &rp, 0, status); + rp = *parts_muladd_scalbn(&tp, &xnp, &rp, 0, 0, status); xnp = *parts_mul(&xnp, &xp, status); } diff --git a/fpu/softfloat-parts.c.inc b/fpu/softfloat-parts.c.inc index XXXXXXX..XXXXXXX 100644 --- a/fpu/softfloat-parts.c.inc +++ b/fpu/softfloat-parts.c.inc @@ -XXX,XX +XXX,XX @@ static FloatPartsN *partsN(mul)(FloatPartsN *a, FloatPartsN *b, * Requires A and C extracted into a double-sized structure to provide the * extra space for the widening multiply. */ -static FloatPartsN *partsN(muladd)(FloatPartsN *a, FloatPartsN *b, - FloatPartsN *c, int flags, float_status *s) +static FloatPartsN *partsN(muladd_scalbn)(FloatPartsN *a, FloatPartsN *b, + FloatPartsN *c, int scale, + int flags, float_status *s) { int ab_mask, abc_mask; FloatPartsW p_widen, c_widen; @@ -XXX,XX +XXX,XX @@ static FloatPartsN *partsN(muladd)(FloatPartsN *a, FloatPartsN *b, a->exp = p_widen.exp; return_normal: + /* TODO: Replace all use of float_muladd_halve_result with scale. */ if (flags & float_muladd_halve_result) { a->exp -= 1; } + a->exp += scale; finish_sign: if (flags & float_muladd_negate_result) { a->sign ^= 1; -- 2.43.0 Use the scalbn interface instead of float_muladd_halve_result. Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- target/arm/tcg/helper-a64.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/target/arm/tcg/helper-a64.c b/target/arm/tcg/helper-a64.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/tcg/helper-a64.c +++ b/target/arm/tcg/helper-a64.c @@ -XXX,XX +XXX,XX @@ uint32_t HELPER(rsqrtsf_f16)(uint32_t a, uint32_t b, float_status *fpst) (float16_is_infinity(b) && float16_is_zero(a))) { return float16_one_point_five; } - return float16_muladd(a, b, float16_three, float_muladd_halve_result, fpst); + return float16_muladd_scalbn(a, b, float16_three, -1, 0, fpst); } float32 HELPER(rsqrtsf_f32)(float32 a, float32 b, float_status *fpst) @@ -XXX,XX +XXX,XX @@ float32 HELPER(rsqrtsf_f32)(float32 a, float32 b, float_status *fpst) (float32_is_infinity(b) && float32_is_zero(a))) { return float32_one_point_five; } - return float32_muladd(a, b, float32_three, float_muladd_halve_result, fpst); + return float32_muladd_scalbn(a, b, float32_three, -1, 0, fpst); } float64 HELPER(rsqrtsf_f64)(float64 a, float64 b, float_status *fpst) @@ -XXX,XX +XXX,XX @@ float64 HELPER(rsqrtsf_f64)(float64 a, float64 b, float_status *fpst) (float64_is_infinity(b) && float64_is_zero(a))) { return float64_one_point_five; } - return float64_muladd(a, b, float64_three, float_muladd_halve_result, fpst); + return float64_muladd_scalbn(a, b, float64_three, -1, 0, fpst); } /* Floating-point reciprocal exponent - see FPRecpX in ARM ARM */ -- 2.43.0 Use the scalbn interface instead of float_muladd_halve_result. Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- target/sparc/helper.h | 4 +- target/sparc/fop_helper.c | 8 ++-- target/sparc/translate.c | 80 +++++++++++++++++++++++---------------- 3 files changed, 54 insertions(+), 38 deletions(-) diff --git a/target/sparc/helper.h b/target/sparc/helper.h index XXXXXXX..XXXXXXX 100644 --- a/target/sparc/helper.h +++ b/target/sparc/helper.h @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_3(faddd, TCG_CALL_NO_WG, f64, env, f64, f64) DEF_HELPER_FLAGS_3(fsubd, TCG_CALL_NO_WG, f64, env, f64, f64) DEF_HELPER_FLAGS_3(fmuld, TCG_CALL_NO_WG, f64, env, f64, f64) DEF_HELPER_FLAGS_3(fdivd, TCG_CALL_NO_WG, f64, env, f64, f64) -DEF_HELPER_FLAGS_5(fmaddd, TCG_CALL_NO_WG, f64, env, f64, f64, f64, i32) +DEF_HELPER_FLAGS_6(fmaddd, TCG_CALL_NO_WG, f64, env, f64, f64, f64, s32, i32) DEF_HELPER_FLAGS_3(fnaddd, TCG_CALL_NO_WG, f64, env, f64, f64) DEF_HELPER_FLAGS_3(fnmuld, TCG_CALL_NO_WG, f64, env, f64, f64) @@ -XXX,XX +XXX,XX @@ DEF_HELPER_FLAGS_3(fadds, TCG_CALL_NO_WG, f32, env, f32, f32) DEF_HELPER_FLAGS_3(fsubs, TCG_CALL_NO_WG, f32, env, f32, f32) DEF_HELPER_FLAGS_3(fmuls, TCG_CALL_NO_WG, f32, env, f32, f32) DEF_HELPER_FLAGS_3(fdivs, TCG_CALL_NO_WG, f32, env, f32, f32) -DEF_HELPER_FLAGS_5(fmadds, TCG_CALL_NO_WG, f32, env, f32, f32, f32, i32) +DEF_HELPER_FLAGS_6(fmadds, TCG_CALL_NO_WG, f32, env, f32, f32, f32, s32, i32) DEF_HELPER_FLAGS_3(fnadds, TCG_CALL_NO_WG, f32, env, f32, f32) DEF_HELPER_FLAGS_3(fnmuls, TCG_CALL_NO_WG, f32, env, f32, f32) diff --git a/target/sparc/fop_helper.c b/target/sparc/fop_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/sparc/fop_helper.c +++ b/target/sparc/fop_helper.c @@ -XXX,XX +XXX,XX @@ Int128 helper_fsqrtq(CPUSPARCState *env, Int128 src) } float32 helper_fmadds(CPUSPARCState *env, float32 s1, - float32 s2, float32 s3, uint32_t op) + float32 s2, float32 s3, int32_t sc, uint32_t op) { - float32 ret = float32_muladd(s1, s2, s3, op, &env->fp_status); + float32 ret = float32_muladd_scalbn(s1, s2, s3, sc, op, &env->fp_status); check_ieee_exceptions(env, GETPC()); return ret; } float64 helper_fmaddd(CPUSPARCState *env, float64 s1, - float64 s2, float64 s3, uint32_t op) + float64 s2, float64 s3, int32_t sc, uint32_t op) { - float64 ret = float64_muladd(s1, s2, s3, op, &env->fp_status); + float64 ret = float64_muladd_scalbn(s1, s2, s3, sc, op, &env->fp_status); check_ieee_exceptions(env, GETPC()); return ret; } diff --git a/target/sparc/translate.c b/target/sparc/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/sparc/translate.c +++ b/target/sparc/translate.c @@ -XXX,XX +XXX,XX @@ static void gen_op_fabsq(TCGv_i128 dst, TCGv_i128 src) static void gen_op_fmadds(TCGv_i32 d, TCGv_i32 s1, TCGv_i32 s2, TCGv_i32 s3) { - gen_helper_fmadds(d, tcg_env, s1, s2, s3, tcg_constant_i32(0)); + TCGv_i32 z = tcg_constant_i32(0); + gen_helper_fmadds(d, tcg_env, s1, s2, s3, z, z); } static void gen_op_fmaddd(TCGv_i64 d, TCGv_i64 s1, TCGv_i64 s2, TCGv_i64 s3) { - gen_helper_fmaddd(d, tcg_env, s1, s2, s3, tcg_constant_i32(0)); + TCGv_i32 z = tcg_constant_i32(0); + gen_helper_fmaddd(d, tcg_env, s1, s2, s3, z, z); } static void gen_op_fmsubs(TCGv_i32 d, TCGv_i32 s1, TCGv_i32 s2, TCGv_i32 s3) { - int op = float_muladd_negate_c; - gen_helper_fmadds(d, tcg_env, s1, s2, s3, tcg_constant_i32(op)); + TCGv_i32 z = tcg_constant_i32(0); + TCGv_i32 op = tcg_constant_i32(float_muladd_negate_c); + gen_helper_fmadds(d, tcg_env, s1, s2, s3, z, op); } static void gen_op_fmsubd(TCGv_i64 d, TCGv_i64 s1, TCGv_i64 s2, TCGv_i64 s3) { - int op = float_muladd_negate_c; - gen_helper_fmaddd(d, tcg_env, s1, s2, s3, tcg_constant_i32(op)); + TCGv_i32 z = tcg_constant_i32(0); + TCGv_i32 op = tcg_constant_i32(float_muladd_negate_c); + gen_helper_fmaddd(d, tcg_env, s1, s2, s3, z, op); } static void gen_op_fnmsubs(TCGv_i32 d, TCGv_i32 s1, TCGv_i32 s2, TCGv_i32 s3) { - int op = float_muladd_negate_c | float_muladd_negate_result; - gen_helper_fmadds(d, tcg_env, s1, s2, s3, tcg_constant_i32(op)); + TCGv_i32 z = tcg_constant_i32(0); + TCGv_i32 op = tcg_constant_i32(float_muladd_negate_c | + float_muladd_negate_result); + gen_helper_fmadds(d, tcg_env, s1, s2, s3, z, op); } static void gen_op_fnmsubd(TCGv_i64 d, TCGv_i64 s1, TCGv_i64 s2, TCGv_i64 s3) { - int op = float_muladd_negate_c | float_muladd_negate_result; - gen_helper_fmaddd(d, tcg_env, s1, s2, s3, tcg_constant_i32(op)); + TCGv_i32 z = tcg_constant_i32(0); + TCGv_i32 op = tcg_constant_i32(float_muladd_negate_c | + float_muladd_negate_result); + gen_helper_fmaddd(d, tcg_env, s1, s2, s3, z, op); } static void gen_op_fnmadds(TCGv_i32 d, TCGv_i32 s1, TCGv_i32 s2, TCGv_i32 s3) { - int op = float_muladd_negate_result; - gen_helper_fmadds(d, tcg_env, s1, s2, s3, tcg_constant_i32(op)); + TCGv_i32 z = tcg_constant_i32(0); + TCGv_i32 op = tcg_constant_i32(float_muladd_negate_result); + gen_helper_fmadds(d, tcg_env, s1, s2, s3, z, op); } static void gen_op_fnmaddd(TCGv_i64 d, TCGv_i64 s1, TCGv_i64 s2, TCGv_i64 s3) { - int op = float_muladd_negate_result; - gen_helper_fmaddd(d, tcg_env, s1, s2, s3, tcg_constant_i32(op)); + TCGv_i32 z = tcg_constant_i32(0); + TCGv_i32 op = tcg_constant_i32(float_muladd_negate_result); + gen_helper_fmaddd(d, tcg_env, s1, s2, s3, z, op); } /* Use muladd to compute (1 * src1) + src2 / 2 with one rounding. */ static void gen_op_fhadds(TCGv_i32 d, TCGv_i32 s1, TCGv_i32 s2) { - TCGv_i32 one = tcg_constant_i32(float32_one); - int op = float_muladd_halve_result; - gen_helper_fmadds(d, tcg_env, one, s1, s2, tcg_constant_i32(op)); + TCGv_i32 fone = tcg_constant_i32(float32_one); + TCGv_i32 mone = tcg_constant_i32(-1); + TCGv_i32 op = tcg_constant_i32(0); + gen_helper_fmadds(d, tcg_env, fone, s1, s2, mone, op); } static void gen_op_fhaddd(TCGv_i64 d, TCGv_i64 s1, TCGv_i64 s2) { - TCGv_i64 one = tcg_constant_i64(float64_one); - int op = float_muladd_halve_result; - gen_helper_fmaddd(d, tcg_env, one, s1, s2, tcg_constant_i32(op)); + TCGv_i64 fone = tcg_constant_i64(float64_one); + TCGv_i32 mone = tcg_constant_i32(-1); + TCGv_i32 op = tcg_constant_i32(0); + gen_helper_fmaddd(d, tcg_env, fone, s1, s2, mone, op); } /* Use muladd to compute (1 * src1) - src2 / 2 with one rounding. */ static void gen_op_fhsubs(TCGv_i32 d, TCGv_i32 s1, TCGv_i32 s2) { - TCGv_i32 one = tcg_constant_i32(float32_one); - int op = float_muladd_negate_c | float_muladd_halve_result; - gen_helper_fmadds(d, tcg_env, one, s1, s2, tcg_constant_i32(op)); + TCGv_i32 fone = tcg_constant_i32(float32_one); + TCGv_i32 mone = tcg_constant_i32(-1); + TCGv_i32 op = tcg_constant_i32(float_muladd_negate_c); + gen_helper_fmadds(d, tcg_env, fone, s1, s2, mone, op); } static void gen_op_fhsubd(TCGv_i64 d, TCGv_i64 s1, TCGv_i64 s2) { - TCGv_i64 one = tcg_constant_i64(float64_one); - int op = float_muladd_negate_c | float_muladd_halve_result; - gen_helper_fmaddd(d, tcg_env, one, s1, s2, tcg_constant_i32(op)); + TCGv_i64 fone = tcg_constant_i64(float64_one); + TCGv_i32 mone = tcg_constant_i32(-1); + TCGv_i32 op = tcg_constant_i32(float_muladd_negate_c); + gen_helper_fmaddd(d, tcg_env, fone, s1, s2, mone, op); } /* Use muladd to compute -((1 * src1) + src2 / 2) with one rounding. */ static void gen_op_fnhadds(TCGv_i32 d, TCGv_i32 s1, TCGv_i32 s2) { - TCGv_i32 one = tcg_constant_i32(float32_one); - int op = float_muladd_negate_result | float_muladd_halve_result; - gen_helper_fmadds(d, tcg_env, one, s1, s2, tcg_constant_i32(op)); + TCGv_i32 fone = tcg_constant_i32(float32_one); + TCGv_i32 mone = tcg_constant_i32(-1); + TCGv_i32 op = tcg_constant_i32(float_muladd_negate_result); + gen_helper_fmadds(d, tcg_env, fone, s1, s2, mone, op); } static void gen_op_fnhaddd(TCGv_i64 d, TCGv_i64 s1, TCGv_i64 s2) { - TCGv_i64 one = tcg_constant_i64(float64_one); - int op = float_muladd_negate_result | float_muladd_halve_result; - gen_helper_fmaddd(d, tcg_env, one, s1, s2, tcg_constant_i32(op)); + TCGv_i64 fone = tcg_constant_i64(float64_one); + TCGv_i32 mone = tcg_constant_i32(-1); + TCGv_i32 op = tcg_constant_i32(float_muladd_negate_result); + gen_helper_fmaddd(d, tcg_env, fone, s1, s2, mone, op); } static void gen_op_fpexception_im(DisasContext *dc, int ftt) -- 2.43.0 All uses have been convered to float*_muladd_scalbn. Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- include/fpu/softfloat.h | 3 --- fpu/softfloat.c | 6 ------ fpu/softfloat-parts.c.inc | 4 ---- 3 files changed, 13 deletions(-) diff --git a/include/fpu/softfloat.h b/include/fpu/softfloat.h index XXXXXXX..XXXXXXX 100644 --- a/include/fpu/softfloat.h +++ b/include/fpu/softfloat.h @@ -XXX,XX +XXX,XX @@ bfloat16 bfloat16_squash_input_denormal(bfloat16 a, float_status *status); | Using these differs from negating an input or output before calling | the muladd function in that this means that a NaN doesn't have its | sign bit inverted before it is propagated. -| We also support halving the result before rounding, as a special -| case to support the ARM fused-sqrt-step instruction FRSQRTS. *----------------------------------------------------------------------------*/ enum { float_muladd_negate_c = 1, float_muladd_negate_product = 2, float_muladd_negate_result = 4, - float_muladd_halve_result = 8, }; /*---------------------------------------------------------------------------- diff --git a/fpu/softfloat.c b/fpu/softfloat.c index XXXXXXX..XXXXXXX 100644 --- a/fpu/softfloat.c +++ b/fpu/softfloat.c @@ -XXX,XX +XXX,XX @@ float32_muladd(float32 xa, float32 xb, float32 xc, int flags, float_status *s) if (unlikely(!can_use_fpu(s))) { goto soft; } - if (unlikely(flags & float_muladd_halve_result)) { - goto soft; - } float32_input_flush3(&ua.s, &ub.s, &uc.s, s); if (unlikely(!f32_is_zon3(ua, ub, uc))) { @@ -XXX,XX +XXX,XX @@ float64_muladd(float64 xa, float64 xb, float64 xc, int flags, float_status *s) if (unlikely(!can_use_fpu(s))) { goto soft; } - if (unlikely(flags & float_muladd_halve_result)) { - goto soft; - } float64_input_flush3(&ua.s, &ub.s, &uc.s, s); if (unlikely(!f64_is_zon3(ua, ub, uc))) { diff --git a/fpu/softfloat-parts.c.inc b/fpu/softfloat-parts.c.inc index XXXXXXX..XXXXXXX 100644 --- a/fpu/softfloat-parts.c.inc +++ b/fpu/softfloat-parts.c.inc @@ -XXX,XX +XXX,XX @@ static FloatPartsN *partsN(muladd_scalbn)(FloatPartsN *a, FloatPartsN *b, a->exp = p_widen.exp; return_normal: - /* TODO: Replace all use of float_muladd_halve_result with scale. */ - if (flags & float_muladd_halve_result) { - a->exp -= 1; - } a->exp += scale; finish_sign: if (flags & float_muladd_negate_result) { -- 2.43.0 This rounding mode is used by Hexagon. Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- include/fpu/softfloat-types.h | 2 ++ fpu/softfloat-parts.c.inc | 3 +++ 2 files changed, 5 insertions(+) diff --git a/include/fpu/softfloat-types.h b/include/fpu/softfloat-types.h index XXXXXXX..XXXXXXX 100644 --- a/include/fpu/softfloat-types.h +++ b/include/fpu/softfloat-types.h @@ -XXX,XX +XXX,XX @@ typedef enum __attribute__((__packed__)) { float_round_to_odd = 5, /* Not an IEEE rounding mode: round to closest odd, overflow to inf */ float_round_to_odd_inf = 6, + /* Not an IEEE rounding mode: round to nearest even, overflow to max */ + float_round_nearest_even_max = 7, } FloatRoundMode; /* diff --git a/fpu/softfloat-parts.c.inc b/fpu/softfloat-parts.c.inc index XXXXXXX..XXXXXXX 100644 --- a/fpu/softfloat-parts.c.inc +++ b/fpu/softfloat-parts.c.inc @@ -XXX,XX +XXX,XX @@ static void partsN(uncanon_normal)(FloatPartsN *p, float_status *s, int exp, flags = 0; switch (s->float_rounding_mode) { + case float_round_nearest_even_max: + overflow_norm = true; + /* fall through */ case float_round_nearest_even: if (N > 64 && frac_lsb == 0) { inc = ((p->frac_hi & 1) || (p->frac_lo & round_mask) != frac_lsbm1 -- 2.43.0 Certain Hexagon instructions suppress changes to the result when the product of fma() is a true zero. Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- include/fpu/softfloat.h | 5 +++++ fpu/softfloat.c | 3 +++ fpu/softfloat-parts.c.inc | 4 +++- 3 files changed, 11 insertions(+), 1 deletion(-) diff --git a/include/fpu/softfloat.h b/include/fpu/softfloat.h index XXXXXXX..XXXXXXX 100644 --- a/include/fpu/softfloat.h +++ b/include/fpu/softfloat.h @@ -XXX,XX +XXX,XX @@ bfloat16 bfloat16_squash_input_denormal(bfloat16 a, float_status *status); | Using these differs from negating an input or output before calling | the muladd function in that this means that a NaN doesn't have its | sign bit inverted before it is propagated. +| +| With float_muladd_suppress_add_product_zero, if A or B is zero +| such that the product is a true zero, then return C without addition. +| This preserves the sign of C when C is +/- 0. Used for Hexagon. *----------------------------------------------------------------------------*/ enum { float_muladd_negate_c = 1, float_muladd_negate_product = 2, float_muladd_negate_result = 4, + float_muladd_suppress_add_product_zero = 8, }; /*---------------------------------------------------------------------------- diff --git a/fpu/softfloat.c b/fpu/softfloat.c index XXXXXXX..XXXXXXX 100644 --- a/fpu/softfloat.c +++ b/fpu/softfloat.c @@ -XXX,XX +XXX,XX @@ float32_muladd(float32 xa, float32 xb, float32 xc, int flags, float_status *s) if (unlikely(!can_use_fpu(s))) { goto soft; } + if (unlikely(flags & float_muladd_suppress_add_product_zero)) { + goto soft; + } float32_input_flush3(&ua.s, &ub.s, &uc.s, s); if (unlikely(!f32_is_zon3(ua, ub, uc))) { diff --git a/fpu/softfloat-parts.c.inc b/fpu/softfloat-parts.c.inc index XXXXXXX..XXXXXXX 100644 --- a/fpu/softfloat-parts.c.inc +++ b/fpu/softfloat-parts.c.inc @@ -XXX,XX +XXX,XX @@ static FloatPartsN *partsN(muladd_scalbn)(FloatPartsN *a, FloatPartsN *b, goto return_normal; } if (c->cls == float_class_zero) { - if (a->sign != c->sign) { + if (flags & float_muladd_suppress_add_product_zero) { + a->sign = c->sign; + } else if (a->sign != c->sign) { goto return_sub_zero; } goto return_zero; -- 2.43.0 There are no special cases for this instruction. Remove internal_mpyf as unused. Reviewed-by: Brian Cain <brian.cain@oss.qualcomm.com> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- target/hexagon/fma_emu.h | 1 - target/hexagon/fma_emu.c | 8 -------- target/hexagon/op_helper.c | 2 +- 3 files changed, 1 insertion(+), 10 deletions(-) diff --git a/target/hexagon/fma_emu.h b/target/hexagon/fma_emu.h index XXXXXXX..XXXXXXX 100644 --- a/target/hexagon/fma_emu.h +++ b/target/hexagon/fma_emu.h @@ -XXX,XX +XXX,XX @@ int32_t float32_getexp(float32 f32); float32 infinite_float32(uint8_t sign); float32 internal_fmafx(float32 a, float32 b, float32 c, int scale, float_status *fp_status); -float32 internal_mpyf(float32 a, float32 b, float_status *fp_status); float64 internal_mpyhh(float64 a, float64 b, unsigned long long int accumulated, float_status *fp_status); diff --git a/target/hexagon/fma_emu.c b/target/hexagon/fma_emu.c index XXXXXXX..XXXXXXX 100644 --- a/target/hexagon/fma_emu.c +++ b/target/hexagon/fma_emu.c @@ -XXX,XX +XXX,XX @@ float32 internal_fmafx(float32 a, float32 b, float32 c, int scale, return accum_round_float32(result, fp_status); } -float32 internal_mpyf(float32 a, float32 b, float_status *fp_status) -{ - if (float32_is_zero(a) || float32_is_zero(b)) { - return float32_mul(a, b, fp_status); - } - return internal_fmafx(a, b, float32_zero, 0, fp_status); -} - float64 internal_mpyhh(float64 a, float64 b, unsigned long long int accumulated, float_status *fp_status) diff --git a/target/hexagon/op_helper.c b/target/hexagon/op_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/hexagon/op_helper.c +++ b/target/hexagon/op_helper.c @@ -XXX,XX +XXX,XX @@ float32 HELPER(sfmpy)(CPUHexagonState *env, float32 RsV, float32 RtV) { float32 RdV; arch_fpop_start(env); - RdV = internal_mpyf(RsV, RtV, &env->fp_status); + RdV = float32_mul(RsV, RtV, &env->fp_status); arch_fpop_end(env); return RdV; } -- 2.43.0 There are no special cases for this instruction. Reviewed-by: Brian Cain <brian.cain@oss.qualcomm.com> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- target/hexagon/op_helper.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/target/hexagon/op_helper.c b/target/hexagon/op_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/hexagon/op_helper.c +++ b/target/hexagon/op_helper.c @@ -XXX,XX +XXX,XX @@ float32 HELPER(sffma)(CPUHexagonState *env, float32 RxV, float32 RsV, float32 RtV) { arch_fpop_start(env); - RxV = internal_fmafx(RsV, RtV, RxV, 0, &env->fp_status); + RxV = float32_muladd(RsV, RtV, RxV, 0, &env->fp_status); arch_fpop_end(env); return RxV; } -- 2.43.0 There are no special cases for this instruction. Since hexagon always uses default-nan mode, explicitly negating the first input is unnecessary. Use float_muladd_negate_product instead. Reviewed-by: Brian Cain <brian.cain@oss.qualcomm.com> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- target/hexagon/op_helper.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) diff --git a/target/hexagon/op_helper.c b/target/hexagon/op_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/hexagon/op_helper.c +++ b/target/hexagon/op_helper.c @@ -XXX,XX +XXX,XX @@ float32 HELPER(sffma_sc)(CPUHexagonState *env, float32 RxV, float32 HELPER(sffms)(CPUHexagonState *env, float32 RxV, float32 RsV, float32 RtV) { - float32 neg_RsV; arch_fpop_start(env); - neg_RsV = float32_set_sign(RsV, float32_is_neg(RsV) ? 0 : 1); - RxV = internal_fmafx(neg_RsV, RtV, RxV, 0, &env->fp_status); + RxV = float32_muladd(RsV, RtV, RxV, float_muladd_negate_product, + &env->fp_status); arch_fpop_end(env); return RxV; } -- 2.43.0 This instruction has a special case that 0 * x + c returns c without the normal sign folding that comes with 0 + -0. Use the new float_muladd_suppress_add_product_zero to describe this. Reviewed-by: Brian Cain <brian.cain@oss.qualcomm.com> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- target/hexagon/op_helper.c | 11 +++-------- 1 file changed, 3 insertions(+), 8 deletions(-) diff --git a/target/hexagon/op_helper.c b/target/hexagon/op_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/hexagon/op_helper.c +++ b/target/hexagon/op_helper.c @@ -XXX,XX +XXX,XX @@ static float32 check_nan(float32 dst, float32 x, float_status *fp_status) float32 HELPER(sffma_sc)(CPUHexagonState *env, float32 RxV, float32 RsV, float32 RtV, float32 PuV) { - size4s_t tmp; arch_fpop_start(env); - RxV = check_nan(RxV, RxV, &env->fp_status); - RxV = check_nan(RxV, RsV, &env->fp_status); - RxV = check_nan(RxV, RtV, &env->fp_status); - tmp = internal_fmafx(RsV, RtV, RxV, fSXTN(8, 64, PuV), &env->fp_status); - if (!(float32_is_zero(RxV) && is_zero_prod(RsV, RtV))) { - RxV = tmp; - } + RxV = float32_muladd_scalbn(RsV, RtV, RxV, fSXTN(8, 64, PuV), + float_muladd_suppress_add_product_zero, + &env->fp_status); arch_fpop_end(env); return RxV; } -- 2.43.0 There are multiple special cases for this instruction. (1) The saturate to normal maximum instead of overflow to infinity is handled by the new float_round_nearest_even_max rounding mode. (2) The 0 * n + c special case is handled by the new float_muladd_suppress_add_product_zero flag. (3) The Inf - Inf -> 0 special case can be detected after the fact by examining float_flag_invalid_isi. Reviewed-by: Brian Cain <brian.cain@oss.qualcomm.com> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- target/hexagon/op_helper.c | 105 +++++++++---------------------------- 1 file changed, 26 insertions(+), 79 deletions(-) diff --git a/target/hexagon/op_helper.c b/target/hexagon/op_helper.c index XXXXXXX..XXXXXXX 100644 --- a/target/hexagon/op_helper.c +++ b/target/hexagon/op_helper.c @@ -XXX,XX +XXX,XX @@ float32 HELPER(sffma)(CPUHexagonState *env, float32 RxV, return RxV; } -static bool is_zero_prod(float32 a, float32 b) -{ - return ((float32_is_zero(a) && is_finite(b)) || - (float32_is_zero(b) && is_finite(a))); -} - -static float32 check_nan(float32 dst, float32 x, float_status *fp_status) -{ - float32 ret = dst; - if (float32_is_any_nan(x)) { - if (extract32(x, 22, 1) == 0) { - float_raise(float_flag_invalid, fp_status); - } - ret = make_float32(0xffffffff); /* nan */ - } - return ret; -} - float32 HELPER(sffma_sc)(CPUHexagonState *env, float32 RxV, float32 RsV, float32 RtV, float32 PuV) { @@ -XXX,XX +XXX,XX @@ float32 HELPER(sffms)(CPUHexagonState *env, float32 RxV, return RxV; } -static bool is_inf_prod(int32_t a, int32_t b) +static float32 do_sffma_lib(CPUHexagonState *env, float32 RxV, + float32 RsV, float32 RtV, int negate) { - return (float32_is_infinity(a) && float32_is_infinity(b)) || - (float32_is_infinity(a) && is_finite(b) && !float32_is_zero(b)) || - (float32_is_infinity(b) && is_finite(a) && !float32_is_zero(a)); + int flags; + + arch_fpop_start(env); + + set_float_rounding_mode(float_round_nearest_even_max, &env->fp_status); + RxV = float32_muladd(RsV, RtV, RxV, + negate | float_muladd_suppress_add_product_zero, + &env->fp_status); + + flags = get_float_exception_flags(&env->fp_status); + if (flags) { + /* Flags are suppressed by this instruction. */ + set_float_exception_flags(0, &env->fp_status); + + /* Return 0 for Inf - Inf. */ + if (flags & float_flag_invalid_isi) { + RxV = 0; + } + } + + arch_fpop_end(env); + return RxV; } float32 HELPER(sffma_lib)(CPUHexagonState *env, float32 RxV, float32 RsV, float32 RtV) { - bool infinp; - bool infminusinf; - float32 tmp; - - arch_fpop_start(env); - set_float_rounding_mode(float_round_nearest_even, &env->fp_status); - infminusinf = float32_is_infinity(RxV) && - is_inf_prod(RsV, RtV) && - (fGETBIT(31, RsV ^ RxV ^ RtV) != 0); - infinp = float32_is_infinity(RxV) || - float32_is_infinity(RtV) || - float32_is_infinity(RsV); - RxV = check_nan(RxV, RxV, &env->fp_status); - RxV = check_nan(RxV, RsV, &env->fp_status); - RxV = check_nan(RxV, RtV, &env->fp_status); - tmp = internal_fmafx(RsV, RtV, RxV, 0, &env->fp_status); - if (!(float32_is_zero(RxV) && is_zero_prod(RsV, RtV))) { - RxV = tmp; - } - set_float_exception_flags(0, &env->fp_status); - if (float32_is_infinity(RxV) && !infinp) { - RxV = RxV - 1; - } - if (infminusinf) { - RxV = 0; - } - arch_fpop_end(env); - return RxV; + return do_sffma_lib(env, RxV, RsV, RtV, 0); } float32 HELPER(sffms_lib)(CPUHexagonState *env, float32 RxV, float32 RsV, float32 RtV) { - bool infinp; - bool infminusinf; - float32 tmp; - - arch_fpop_start(env); - set_float_rounding_mode(float_round_nearest_even, &env->fp_status); - infminusinf = float32_is_infinity(RxV) && - is_inf_prod(RsV, RtV) && - (fGETBIT(31, RsV ^ RxV ^ RtV) == 0); - infinp = float32_is_infinity(RxV) || - float32_is_infinity(RtV) || - float32_is_infinity(RsV); - RxV = check_nan(RxV, RxV, &env->fp_status); - RxV = check_nan(RxV, RsV, &env->fp_status); - RxV = check_nan(RxV, RtV, &env->fp_status); - float32 minus_RsV = float32_sub(float32_zero, RsV, &env->fp_status); - tmp = internal_fmafx(minus_RsV, RtV, RxV, 0, &env->fp_status); - if (!(float32_is_zero(RxV) && is_zero_prod(RsV, RtV))) { - RxV = tmp; - } - set_float_exception_flags(0, &env->fp_status); - if (float32_is_infinity(RxV) && !infinp) { - RxV = RxV - 1; - } - if (infminusinf) { - RxV = 0; - } - arch_fpop_end(env); - return RxV; + return do_sffma_lib(env, RxV, RsV, RtV, float_muladd_negate_product); } float64 HELPER(dfmpyfix)(CPUHexagonState *env, float64 RssV, float64 RttV) -- 2.43.0 The function is now unused. Reviewed-by: Brian Cain <brian.cain@oss.qualcomm.com> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- target/hexagon/fma_emu.h | 2 - target/hexagon/fma_emu.c | 171 --------------------------------------- 2 files changed, 173 deletions(-) diff --git a/target/hexagon/fma_emu.h b/target/hexagon/fma_emu.h index XXXXXXX..XXXXXXX 100644 --- a/target/hexagon/fma_emu.h +++ b/target/hexagon/fma_emu.h @@ -XXX,XX +XXX,XX @@ static inline uint32_t float32_getexp_raw(float32 f32) } int32_t float32_getexp(float32 f32); float32 infinite_float32(uint8_t sign); -float32 internal_fmafx(float32 a, float32 b, float32 c, - int scale, float_status *fp_status); float64 internal_mpyhh(float64 a, float64 b, unsigned long long int accumulated, float_status *fp_status); diff --git a/target/hexagon/fma_emu.c b/target/hexagon/fma_emu.c index XXXXXXX..XXXXXXX 100644 --- a/target/hexagon/fma_emu.c +++ b/target/hexagon/fma_emu.c @@ -XXX,XX +XXX,XX @@ int32_t float64_getexp(float64 f64) return -1; } -static uint64_t float32_getmant(float32 f32) -{ - Float a = { .i = f32 }; - if (float32_is_normal(f32)) { - return a.mant | 1ULL << 23; - } - if (float32_is_zero(f32)) { - return 0; - } - if (float32_is_denormal(f32)) { - return a.mant; - } - return ~0ULL; -} - int32_t float32_getexp(float32 f32) { Float a = { .i = f32 }; @@ -XXX,XX +XXX,XX @@ float32 infinite_float32(uint8_t sign) } /* Return a maximum finite value with the requested sign */ -static float32 maxfinite_float32(uint8_t sign) -{ - if (sign) { - return make_float32(SF_MINUS_MAXF); - } else { - return make_float32(SF_MAXF); - } -} - -/* Return a zero value with requested sign */ -static float32 zero_float32(uint8_t sign) -{ - if (sign) { - return make_float32(0x80000000); - } else { - return float32_zero; - } -} - #define GEN_XF_ROUND(SUFFIX, MANTBITS, INF_EXP, INTERNAL_TYPE) \ static SUFFIX accum_round_##SUFFIX(Accum a, float_status * fp_status) \ { \ @@ -XXX,XX +XXX,XX @@ static SUFFIX accum_round_##SUFFIX(Accum a, float_status * fp_status) \ } GEN_XF_ROUND(float64, DF_MANTBITS, DF_INF_EXP, Double) -GEN_XF_ROUND(float32, SF_MANTBITS, SF_INF_EXP, Float) - -static bool is_inf_prod(float64 a, float64 b) -{ - return ((float64_is_infinity(a) && float64_is_infinity(b)) || - (float64_is_infinity(a) && is_finite(b) && (!float64_is_zero(b))) || - (float64_is_infinity(b) && is_finite(a) && (!float64_is_zero(a)))); -} - -static float64 special_fma(float64 a, float64 b, float64 c, - float_status *fp_status) -{ - float64 ret = make_float64(0); - - /* - * If A multiplied by B is an exact infinity and C is also an infinity - * but with the opposite sign, FMA returns NaN and raises invalid. - */ - uint8_t a_sign = float64_is_neg(a); - uint8_t b_sign = float64_is_neg(b); - uint8_t c_sign = float64_is_neg(c); - if (is_inf_prod(a, b) && float64_is_infinity(c)) { - if ((a_sign ^ b_sign) != c_sign) { - ret = make_float64(DF_NAN); - float_raise(float_flag_invalid, fp_status); - return ret; - } - } - if ((float64_is_infinity(a) && float64_is_zero(b)) || - (float64_is_zero(a) && float64_is_infinity(b))) { - ret = make_float64(DF_NAN); - float_raise(float_flag_invalid, fp_status); - return ret; - } - /* - * If none of the above checks are true and C is a NaN, - * a NaN shall be returned - * If A or B are NaN, a NAN shall be returned. - */ - if (float64_is_any_nan(a) || - float64_is_any_nan(b) || - float64_is_any_nan(c)) { - if (float64_is_any_nan(a) && (fGETBIT(51, a) == 0)) { - float_raise(float_flag_invalid, fp_status); - } - if (float64_is_any_nan(b) && (fGETBIT(51, b) == 0)) { - float_raise(float_flag_invalid, fp_status); - } - if (float64_is_any_nan(c) && (fGETBIT(51, c) == 0)) { - float_raise(float_flag_invalid, fp_status); - } - ret = make_float64(DF_NAN); - return ret; - } - /* - * We have checked for adding opposite-signed infinities. - * Other infinities return infinity with the correct sign - */ - if (float64_is_infinity(c)) { - ret = infinite_float64(c_sign); - return ret; - } - if (float64_is_infinity(a) || float64_is_infinity(b)) { - ret = infinite_float64(a_sign ^ b_sign); - return ret; - } - g_assert_not_reached(); -} - -static float32 special_fmaf(float32 a, float32 b, float32 c, - float_status *fp_status) -{ - float64 aa, bb, cc; - aa = float32_to_float64(a, fp_status); - bb = float32_to_float64(b, fp_status); - cc = float32_to_float64(c, fp_status); - return float64_to_float32(special_fma(aa, bb, cc, fp_status), fp_status); -} - -float32 internal_fmafx(float32 a, float32 b, float32 c, int scale, - float_status *fp_status) -{ - Accum prod; - Accum acc; - Accum result; - accum_init(&prod); - accum_init(&acc); - accum_init(&result); - - uint8_t a_sign = float32_is_neg(a); - uint8_t b_sign = float32_is_neg(b); - uint8_t c_sign = float32_is_neg(c); - if (float32_is_infinity(a) || - float32_is_infinity(b) || - float32_is_infinity(c)) { - return special_fmaf(a, b, c, fp_status); - } - if (float32_is_any_nan(a) || - float32_is_any_nan(b) || - float32_is_any_nan(c)) { - return special_fmaf(a, b, c, fp_status); - } - if ((scale == 0) && (float32_is_zero(a) || float32_is_zero(b))) { - float32 tmp = float32_mul(a, b, fp_status); - tmp = float32_add(tmp, c, fp_status); - return tmp; - } - - /* (a * 2**b) * (c * 2**d) == a*c * 2**(b+d) */ - prod.mant = int128_mul_6464(float32_getmant(a), float32_getmant(b)); - - /* - * Note: extracting the mantissa into an int is multiplying by - * 2**23, so adjust here - */ - prod.exp = float32_getexp(a) + float32_getexp(b) - SF_BIAS - 23; - prod.sign = a_sign ^ b_sign; - if (float32_is_zero(a) || float32_is_zero(b)) { - prod.exp = -2 * WAY_BIG_EXP; - } - if ((scale > 0) && float32_is_denormal(c)) { - acc.mant = int128_mul_6464(0, 0); - acc.exp = -WAY_BIG_EXP; - acc.sign = c_sign; - acc.sticky = 1; - result = accum_add(prod, acc); - } else if (!float32_is_zero(c)) { - acc.mant = int128_mul_6464(float32_getmant(c), 1); - acc.exp = float32_getexp(c); - acc.sign = c_sign; - result = accum_add(prod, acc); - } else { - result = prod; - } - result.exp += scale; - return accum_round_float32(result, fp_status); -} float64 internal_mpyhh(float64 a, float64 b, unsigned long long int accumulated, -- 2.43.0 This massive macro is now only used once. Expand it for use only by float64. Reviewed-by: Brian Cain <brian.cain@oss.qualcomm.com> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- target/hexagon/fma_emu.c | 255 +++++++++++++++++++-------------------- 1 file changed, 127 insertions(+), 128 deletions(-) diff --git a/target/hexagon/fma_emu.c b/target/hexagon/fma_emu.c index XXXXXXX..XXXXXXX 100644 --- a/target/hexagon/fma_emu.c +++ b/target/hexagon/fma_emu.c @@ -XXX,XX +XXX,XX @@ float32 infinite_float32(uint8_t sign) } /* Return a maximum finite value with the requested sign */ -#define GEN_XF_ROUND(SUFFIX, MANTBITS, INF_EXP, INTERNAL_TYPE) \ -static SUFFIX accum_round_##SUFFIX(Accum a, float_status * fp_status) \ -{ \ - if ((int128_gethi(a.mant) == 0) && (int128_getlo(a.mant) == 0) \ - && ((a.guard | a.round | a.sticky) == 0)) { \ - /* result zero */ \ - switch (fp_status->float_rounding_mode) { \ - case float_round_down: \ - return zero_##SUFFIX(1); \ - default: \ - return zero_##SUFFIX(0); \ - } \ - } \ - /* Normalize right */ \ - /* We want MANTBITS bits of mantissa plus the leading one. */ \ - /* That means that we want MANTBITS+1 bits, or 0x000000000000FF_FFFF */ \ - /* So we need to normalize right while the high word is non-zero and \ - * while the low word is nonzero when masked with 0xffe0_0000_0000_0000 */ \ - while ((int128_gethi(a.mant) != 0) || \ - ((int128_getlo(a.mant) >> (MANTBITS + 1)) != 0)) { \ - a = accum_norm_right(a, 1); \ - } \ - /* \ - * OK, now normalize left \ - * We want to normalize left until we have a leading one in bit 24 \ - * Theoretically, we only need to shift a maximum of one to the left if we \ - * shifted out lots of bits from B, or if we had no shift / 1 shift sticky \ - * should be 0 \ - */ \ - while ((int128_getlo(a.mant) & (1ULL << MANTBITS)) == 0) { \ - a = accum_norm_left(a); \ - } \ - /* \ - * OK, now we might need to denormalize because of potential underflow. \ - * We need to do this before rounding, and rounding might make us normal \ - * again \ - */ \ - while (a.exp <= 0) { \ - a = accum_norm_right(a, 1 - a.exp); \ - /* \ - * Do we have underflow? \ - * That's when we get an inexact answer because we ran out of bits \ - * in a denormal. \ - */ \ - if (a.guard || a.round || a.sticky) { \ - float_raise(float_flag_underflow, fp_status); \ - } \ - } \ - /* OK, we're relatively canonical... now we need to round */ \ - if (a.guard || a.round || a.sticky) { \ - float_raise(float_flag_inexact, fp_status); \ - switch (fp_status->float_rounding_mode) { \ - case float_round_to_zero: \ - /* Chop and we're done */ \ - break; \ - case float_round_up: \ - if (a.sign == 0) { \ - a.mant = int128_add(a.mant, int128_one()); \ - } \ - break; \ - case float_round_down: \ - if (a.sign != 0) { \ - a.mant = int128_add(a.mant, int128_one()); \ - } \ - break; \ - default: \ - if (a.round || a.sticky) { \ - /* round up if guard is 1, down if guard is zero */ \ - a.mant = int128_add(a.mant, int128_make64(a.guard)); \ - } else if (a.guard) { \ - /* exactly .5, round up if odd */ \ - a.mant = int128_add(a.mant, int128_and(a.mant, int128_one())); \ - } \ - break; \ - } \ - } \ - /* \ - * OK, now we might have carried all the way up. \ - * So we might need to shr once \ - * at least we know that the lsb should be zero if we rounded and \ - * got a carry out... \ - */ \ - if ((int128_getlo(a.mant) >> (MANTBITS + 1)) != 0) { \ - a = accum_norm_right(a, 1); \ - } \ - /* Overflow? */ \ - if (a.exp >= INF_EXP) { \ - /* Yep, inf result */ \ - float_raise(float_flag_overflow, fp_status); \ - float_raise(float_flag_inexact, fp_status); \ - switch (fp_status->float_rounding_mode) { \ - case float_round_to_zero: \ - return maxfinite_##SUFFIX(a.sign); \ - case float_round_up: \ - if (a.sign == 0) { \ - return infinite_##SUFFIX(a.sign); \ - } else { \ - return maxfinite_##SUFFIX(a.sign); \ - } \ - case float_round_down: \ - if (a.sign != 0) { \ - return infinite_##SUFFIX(a.sign); \ - } else { \ - return maxfinite_##SUFFIX(a.sign); \ - } \ - default: \ - return infinite_##SUFFIX(a.sign); \ - } \ - } \ - /* Underflow? */ \ - if (int128_getlo(a.mant) & (1ULL << MANTBITS)) { \ - /* Leading one means: No, we're normal. So, we should be done... */ \ - INTERNAL_TYPE ret; \ - ret.i = 0; \ - ret.sign = a.sign; \ - ret.exp = a.exp; \ - ret.mant = int128_getlo(a.mant); \ - return ret.i; \ - } \ - assert(a.exp == 1); \ - INTERNAL_TYPE ret; \ - ret.i = 0; \ - ret.sign = a.sign; \ - ret.exp = 0; \ - ret.mant = int128_getlo(a.mant); \ - return ret.i; \ +static float64 accum_round_float64(Accum a, float_status *fp_status) +{ + if ((int128_gethi(a.mant) == 0) && (int128_getlo(a.mant) == 0) + && ((a.guard | a.round | a.sticky) == 0)) { + /* result zero */ + switch (fp_status->float_rounding_mode) { + case float_round_down: + return zero_float64(1); + default: + return zero_float64(0); + } + } + /* + * Normalize right + * We want DF_MANTBITS bits of mantissa plus the leading one. + * That means that we want DF_MANTBITS+1 bits, or 0x000000000000FF_FFFF + * So we need to normalize right while the high word is non-zero and + * while the low word is nonzero when masked with 0xffe0_0000_0000_0000 + */ + while ((int128_gethi(a.mant) != 0) || + ((int128_getlo(a.mant) >> (DF_MANTBITS + 1)) != 0)) { + a = accum_norm_right(a, 1); + } + /* + * OK, now normalize left + * We want to normalize left until we have a leading one in bit 24 + * Theoretically, we only need to shift a maximum of one to the left if we + * shifted out lots of bits from B, or if we had no shift / 1 shift sticky + * should be 0 + */ + while ((int128_getlo(a.mant) & (1ULL << DF_MANTBITS)) == 0) { + a = accum_norm_left(a); + } + /* + * OK, now we might need to denormalize because of potential underflow. + * We need to do this before rounding, and rounding might make us normal + * again + */ + while (a.exp <= 0) { + a = accum_norm_right(a, 1 - a.exp); + /* + * Do we have underflow? + * That's when we get an inexact answer because we ran out of bits + * in a denormal. + */ + if (a.guard || a.round || a.sticky) { + float_raise(float_flag_underflow, fp_status); + } + } + /* OK, we're relatively canonical... now we need to round */ + if (a.guard || a.round || a.sticky) { + float_raise(float_flag_inexact, fp_status); + switch (fp_status->float_rounding_mode) { + case float_round_to_zero: + /* Chop and we're done */ + break; + case float_round_up: + if (a.sign == 0) { + a.mant = int128_add(a.mant, int128_one()); + } + break; + case float_round_down: + if (a.sign != 0) { + a.mant = int128_add(a.mant, int128_one()); + } + break; + default: + if (a.round || a.sticky) { + /* round up if guard is 1, down if guard is zero */ + a.mant = int128_add(a.mant, int128_make64(a.guard)); + } else if (a.guard) { + /* exactly .5, round up if odd */ + a.mant = int128_add(a.mant, int128_and(a.mant, int128_one())); + } + break; + } + } + /* + * OK, now we might have carried all the way up. + * So we might need to shr once + * at least we know that the lsb should be zero if we rounded and + * got a carry out... + */ + if ((int128_getlo(a.mant) >> (DF_MANTBITS + 1)) != 0) { + a = accum_norm_right(a, 1); + } + /* Overflow? */ + if (a.exp >= DF_INF_EXP) { + /* Yep, inf result */ + float_raise(float_flag_overflow, fp_status); + float_raise(float_flag_inexact, fp_status); + switch (fp_status->float_rounding_mode) { + case float_round_to_zero: + return maxfinite_float64(a.sign); + case float_round_up: + if (a.sign == 0) { + return infinite_float64(a.sign); + } else { + return maxfinite_float64(a.sign); + } + case float_round_down: + if (a.sign != 0) { + return infinite_float64(a.sign); + } else { + return maxfinite_float64(a.sign); + } + default: + return infinite_float64(a.sign); + } + } + /* Underflow? */ + if (int128_getlo(a.mant) & (1ULL << DF_MANTBITS)) { + /* Leading one means: No, we're normal. So, we should be done... */ + Double ret; + ret.i = 0; + ret.sign = a.sign; + ret.exp = a.exp; + ret.mant = int128_getlo(a.mant); + return ret.i; + } + assert(a.exp == 1); + Double ret; + ret.i = 0; + ret.sign = a.sign; + ret.exp = 0; + ret.mant = int128_getlo(a.mant); + return ret.i; } -GEN_XF_ROUND(float64, DF_MANTBITS, DF_INF_EXP, Double) - float64 internal_mpyhh(float64 a, float64 b, unsigned long long int accumulated, float_status *fp_status) -- 2.43.0 This structure, with bitfields, is incorrect for big-endian. Use the existing float32_getexp_raw which uses extract32. Reviewed-by: Brian Cain <brian.cain@oss.qualcomm.com> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- target/hexagon/fma_emu.c | 16 +++------------- 1 file changed, 3 insertions(+), 13 deletions(-) diff --git a/target/hexagon/fma_emu.c b/target/hexagon/fma_emu.c index XXXXXXX..XXXXXXX 100644 --- a/target/hexagon/fma_emu.c +++ b/target/hexagon/fma_emu.c @@ -XXX,XX +XXX,XX @@ typedef union { }; } Double; -typedef union { - float f; - uint32_t i; - struct { - uint32_t mant:23; - uint32_t exp:8; - uint32_t sign:1; - }; -} Float; - static uint64_t float64_getmant(float64 f64) { Double a = { .i = f64 }; @@ -XXX,XX +XXX,XX @@ int32_t float64_getexp(float64 f64) int32_t float32_getexp(float32 f32) { - Float a = { .i = f32 }; + int exp = float32_getexp_raw(f32); if (float32_is_normal(f32)) { - return a.exp; + return exp; } if (float32_is_denormal(f32)) { - return a.exp + 1; + return exp + 1; } return -1; } -- 2.43.0 This structure, with bitfields, is incorrect for big-endian. Use extract64 and deposit64 instead. Reviewed-by: Brian Cain <brian.cain@oss.qualcomm.com> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- target/hexagon/fma_emu.c | 46 ++++++++++++++-------------------------- 1 file changed, 16 insertions(+), 30 deletions(-) diff --git a/target/hexagon/fma_emu.c b/target/hexagon/fma_emu.c index XXXXXXX..XXXXXXX 100644 --- a/target/hexagon/fma_emu.c +++ b/target/hexagon/fma_emu.c @@ -XXX,XX +XXX,XX @@ #define WAY_BIG_EXP 4096 -typedef union { - double f; - uint64_t i; - struct { - uint64_t mant:52; - uint64_t exp:11; - uint64_t sign:1; - }; -} Double; - static uint64_t float64_getmant(float64 f64) { - Double a = { .i = f64 }; + uint64_t mant = extract64(f64, 0, 52); if (float64_is_normal(f64)) { - return a.mant | 1ULL << 52; + return mant | 1ULL << 52; } if (float64_is_zero(f64)) { return 0; } if (float64_is_denormal(f64)) { - return a.mant; + return mant; } return ~0ULL; } int32_t float64_getexp(float64 f64) { - Double a = { .i = f64 }; + int exp = extract64(f64, 52, 11); if (float64_is_normal(f64)) { - return a.exp; + return exp; } if (float64_is_denormal(f64)) { - return a.exp + 1; + return exp + 1; } return -1; } @@ -XXX,XX +XXX,XX @@ float32 infinite_float32(uint8_t sign) /* Return a maximum finite value with the requested sign */ static float64 accum_round_float64(Accum a, float_status *fp_status) { + uint64_t ret; + if ((int128_gethi(a.mant) == 0) && (int128_getlo(a.mant) == 0) && ((a.guard | a.round | a.sticky) == 0)) { /* result zero */ @@ -XXX,XX +XXX,XX @@ static float64 accum_round_float64(Accum a, float_status *fp_status) } } /* Underflow? */ - if (int128_getlo(a.mant) & (1ULL << DF_MANTBITS)) { + ret = int128_getlo(a.mant); + if (ret & (1ULL << DF_MANTBITS)) { /* Leading one means: No, we're normal. So, we should be done... */ - Double ret; - ret.i = 0; - ret.sign = a.sign; - ret.exp = a.exp; - ret.mant = int128_getlo(a.mant); - return ret.i; + ret = deposit64(ret, 52, 11, a.exp); + } else { + assert(a.exp == 1); + ret = deposit64(ret, 52, 11, 0); } - assert(a.exp == 1); - Double ret; - ret.i = 0; - ret.sign = a.sign; - ret.exp = 0; - ret.mant = int128_getlo(a.mant); - return ret.i; + ret = deposit64(ret, 63, 1, a.sign); + return ret; } float64 internal_mpyhh(float64 a, float64 b, -- 2.43.0 No need to open-code 64x64->128-bit multiplication. Reviewed-by: Brian Cain <brian.cain@oss.qualcomm.com> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- target/hexagon/fma_emu.c | 32 +++----------------------------- 1 file changed, 3 insertions(+), 29 deletions(-) diff --git a/target/hexagon/fma_emu.c b/target/hexagon/fma_emu.c index XXXXXXX..XXXXXXX 100644 --- a/target/hexagon/fma_emu.c +++ b/target/hexagon/fma_emu.c @@ -XXX,XX +XXX,XX @@ int32_t float32_getexp(float32 f32) return -1; } -static uint32_t int128_getw0(Int128 x) -{ - return int128_getlo(x); -} - -static uint32_t int128_getw1(Int128 x) -{ - return int128_getlo(x) >> 32; -} - static Int128 int128_mul_6464(uint64_t ai, uint64_t bi) { - Int128 a, b; - uint64_t pp0, pp1a, pp1b, pp1s, pp2; + uint64_t l, h; - a = int128_make64(ai); - b = int128_make64(bi); - pp0 = (uint64_t)int128_getw0(a) * (uint64_t)int128_getw0(b); - pp1a = (uint64_t)int128_getw1(a) * (uint64_t)int128_getw0(b); - pp1b = (uint64_t)int128_getw1(b) * (uint64_t)int128_getw0(a); - pp2 = (uint64_t)int128_getw1(a) * (uint64_t)int128_getw1(b); - - pp1s = pp1a + pp1b; - if ((pp1s < pp1a) || (pp1s < pp1b)) { - pp2 += (1ULL << 32); - } - uint64_t ret_low = pp0 + (pp1s << 32); - if ((ret_low < pp0) || (ret_low < (pp1s << 32))) { - pp2 += 1; - } - - return int128_make128(ret_low, pp2 + (pp1s >> 32)); + mulu64(&l, &h, ai, bi); + return int128_make128(l, h); } static Int128 int128_sub_borrow(Int128 a, Int128 b, int borrow) -- 2.43.0 Initialize x with accumulated via direct assignment, rather than multiplying by 1. Reviewed-by: Brian Cain <brian.cain@oss.qualcomm.com> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- target/hexagon/fma_emu.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/target/hexagon/fma_emu.c b/target/hexagon/fma_emu.c index XXXXXXX..XXXXXXX 100644 --- a/target/hexagon/fma_emu.c +++ b/target/hexagon/fma_emu.c @@ -XXX,XX +XXX,XX @@ float64 internal_mpyhh(float64 a, float64 b, float64_is_infinity(b)) { return float64_mul(a, b, fp_status); } - x.mant = int128_mul_6464(accumulated, 1); + x.mant = int128_make64(accumulated); x.sticky = sticky; prod = fGETUWORD(1, float64_getmant(a)) * fGETUWORD(1, float64_getmant(b)); x.mant = int128_add(x.mant, int128_mul_6464(prod, 0x100000000ULL)); -- 2.43.0 Convert all targets simultaneously, as the gen_intermediate_code function disappears from the target. While there are possible workarounds, they're larger than simply performing the conversion. Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- include/exec/translator.h | 14 -------------- include/hw/core/tcg-cpu-ops.h | 13 +++++++++++++ target/alpha/cpu.h | 2 ++ target/arm/internals.h | 2 ++ target/avr/cpu.h | 2 ++ target/hexagon/cpu.h | 2 ++ target/hppa/cpu.h | 2 ++ target/i386/tcg/helper-tcg.h | 2 ++ target/loongarch/internals.h | 2 ++ target/m68k/cpu.h | 2 ++ target/microblaze/cpu.h | 2 ++ target/mips/tcg/tcg-internal.h | 2 ++ target/openrisc/cpu.h | 2 ++ target/ppc/cpu.h | 2 ++ target/riscv/cpu.h | 3 +++ target/rx/cpu.h | 2 ++ target/s390x/s390x-internal.h | 2 ++ target/sh4/cpu.h | 2 ++ target/sparc/cpu.h | 2 ++ target/tricore/cpu.h | 2 ++ target/xtensa/cpu.h | 2 ++ accel/tcg/cpu-exec.c | 8 +++++--- accel/tcg/translate-all.c | 8 +++++--- target/alpha/cpu.c | 1 + target/alpha/translate.c | 4 ++-- target/arm/cpu.c | 1 + target/arm/tcg/cpu-v7m.c | 1 + target/arm/tcg/translate.c | 5 ++--- target/avr/cpu.c | 1 + target/avr/translate.c | 6 +++--- target/hexagon/cpu.c | 1 + target/hexagon/translate.c | 4 ++-- target/hppa/cpu.c | 1 + target/hppa/translate.c | 4 ++-- target/i386/tcg/tcg-cpu.c | 1 + target/i386/tcg/translate.c | 5 ++--- target/loongarch/cpu.c | 1 + target/loongarch/tcg/translate.c | 4 ++-- target/m68k/cpu.c | 1 + target/m68k/translate.c | 4 ++-- target/microblaze/cpu.c | 1 + target/microblaze/translate.c | 4 ++-- target/mips/cpu.c | 1 + target/mips/tcg/translate.c | 4 ++-- target/openrisc/cpu.c | 1 + target/openrisc/translate.c | 4 ++-- target/ppc/cpu_init.c | 1 + target/ppc/translate.c | 4 ++-- target/riscv/tcg/tcg-cpu.c | 1 + target/riscv/translate.c | 4 ++-- target/rx/cpu.c | 1 + target/rx/translate.c | 4 ++-- target/s390x/cpu.c | 1 + target/s390x/tcg/translate.c | 4 ++-- target/sh4/cpu.c | 1 + target/sh4/translate.c | 4 ++-- target/sparc/cpu.c | 1 + target/sparc/translate.c | 4 ++-- target/tricore/cpu.c | 1 + target/tricore/translate.c | 5 ++--- target/xtensa/cpu.c | 1 + target/xtensa/translate.c | 4 ++-- 62 files changed, 121 insertions(+), 62 deletions(-) diff --git a/include/exec/translator.h b/include/exec/translator.h index XXXXXXX..XXXXXXX 100644 --- a/include/exec/translator.h +++ b/include/exec/translator.h @@ -XXX,XX +XXX,XX @@ #include "qemu/bswap.h" #include "exec/vaddr.h" -/** - * gen_intermediate_code - * @cpu: cpu context - * @tb: translation block - * @max_insns: max number of instructions to translate - * @pc: guest virtual program counter address - * @host_pc: host physical program counter address - * - * This function must be provided by the target, which should create - * the target-specific DisasContext, and then invoke translator_loop. - */ -void gen_intermediate_code(CPUState *cpu, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc); - /** * DisasJumpType: * @DISAS_NEXT: Next instruction in program order. diff --git a/include/hw/core/tcg-cpu-ops.h b/include/hw/core/tcg-cpu-ops.h index XXXXXXX..XXXXXXX 100644 --- a/include/hw/core/tcg-cpu-ops.h +++ b/include/hw/core/tcg-cpu-ops.h @@ -XXX,XX +XXX,XX @@ struct TCGCPUOps { * Called when the first CPU is realized. */ void (*initialize)(void); + /** + * @translate_code: Translate guest instructions to TCGOps + * @cpu: cpu context + * @tb: translation block + * @max_insns: max number of instructions to translate + * @pc: guest virtual program counter address + * @host_pc: host physical program counter address + * + * This function must be provided by the target, which should create + * the target-specific DisasContext, and then invoke translator_loop. + */ + void (*translate_code)(CPUState *cpu, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); /** * @synchronize_from_tb: Synchronize state from a TCG #TranslationBlock * diff --git a/target/alpha/cpu.h b/target/alpha/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/alpha/cpu.h +++ b/target/alpha/cpu.h @@ -XXX,XX +XXX,XX @@ enum { }; void alpha_translate_init(void); +void alpha_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); #define CPU_RESOLVING_TYPE TYPE_ALPHA_CPU diff --git a/target/arm/internals.h b/target/arm/internals.h index XXXXXXX..XXXXXXX 100644 --- a/target/arm/internals.h +++ b/target/arm/internals.h @@ -XXX,XX +XXX,XX @@ void init_cpreg_list(ARMCPU *cpu); void arm_cpu_register_gdb_regs_for_features(ARMCPU *cpu); void arm_translate_init(void); +void arm_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); void arm_cpu_register_gdb_commands(ARMCPU *cpu); void aarch64_cpu_register_gdb_commands(ARMCPU *cpu, GString *, diff --git a/target/avr/cpu.h b/target/avr/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/avr/cpu.h +++ b/target/avr/cpu.h @@ -XXX,XX +XXX,XX @@ static inline void set_avr_feature(CPUAVRState *env, int feature) } void avr_cpu_tcg_init(void); +void avr_cpu_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); int cpu_avr_exec(CPUState *cpu); diff --git a/target/hexagon/cpu.h b/target/hexagon/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/hexagon/cpu.h +++ b/target/hexagon/cpu.h @@ -XXX,XX +XXX,XX @@ static inline void cpu_get_tb_cpu_state(CPUHexagonState *env, vaddr *pc, typedef HexagonCPU ArchCPU; void hexagon_translate_init(void); +void hexagon_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); #include "exec/cpu-all.h" diff --git a/target/hppa/cpu.h b/target/hppa/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/hppa/cpu.h +++ b/target/hppa/cpu.h @@ -XXX,XX +XXX,XX @@ static inline int HPPA_BTLB_ENTRIES(CPUHPPAState *env) } void hppa_translate_init(void); +void hppa_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); #define CPU_RESOLVING_TYPE TYPE_HPPA_CPU diff --git a/target/i386/tcg/helper-tcg.h b/target/i386/tcg/helper-tcg.h index XXXXXXX..XXXXXXX 100644 --- a/target/i386/tcg/helper-tcg.h +++ b/target/i386/tcg/helper-tcg.h @@ -XXX,XX +XXX,XX @@ static inline target_long lshift(target_long x, int n) /* translate.c */ void tcg_x86_init(void); +void x86_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); /* excp_helper.c */ G_NORETURN void raise_exception(CPUX86State *env, int exception_index); diff --git a/target/loongarch/internals.h b/target/loongarch/internals.h index XXXXXXX..XXXXXXX 100644 --- a/target/loongarch/internals.h +++ b/target/loongarch/internals.h @@ -XXX,XX +XXX,XX @@ #define TARGET_VIRT_MASK MAKE_64BIT_MASK(0, TARGET_VIRT_ADDR_SPACE_BITS) void loongarch_translate_init(void); +void loongarch_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); void G_NORETURN do_raise_exception(CPULoongArchState *env, uint32_t exception, diff --git a/target/m68k/cpu.h b/target/m68k/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/m68k/cpu.h +++ b/target/m68k/cpu.h @@ -XXX,XX +XXX,XX @@ int m68k_cpu_gdb_read_register(CPUState *cpu, GByteArray *buf, int reg); int m68k_cpu_gdb_write_register(CPUState *cpu, uint8_t *buf, int reg); void m68k_tcg_init(void); +void m68k_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); void m68k_cpu_init_gdb(M68kCPU *cpu); uint32_t cpu_m68k_get_ccr(CPUM68KState *env); void cpu_m68k_set_ccr(CPUM68KState *env, uint32_t); diff --git a/target/microblaze/cpu.h b/target/microblaze/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/microblaze/cpu.h +++ b/target/microblaze/cpu.h @@ -XXX,XX +XXX,XX @@ static inline void mb_cpu_write_msr(CPUMBState *env, uint32_t val) } void mb_tcg_init(void); +void mb_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); #define CPU_RESOLVING_TYPE TYPE_MICROBLAZE_CPU diff --git a/target/mips/tcg/tcg-internal.h b/target/mips/tcg/tcg-internal.h index XXXXXXX..XXXXXXX 100644 --- a/target/mips/tcg/tcg-internal.h +++ b/target/mips/tcg/tcg-internal.h @@ -XXX,XX +XXX,XX @@ #include "cpu.h" void mips_tcg_init(void); +void mips_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); void mips_cpu_synchronize_from_tb(CPUState *cs, const TranslationBlock *tb); G_NORETURN void mips_cpu_do_unaligned_access(CPUState *cpu, vaddr addr, diff --git a/target/openrisc/cpu.h b/target/openrisc/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/openrisc/cpu.h +++ b/target/openrisc/cpu.h @@ -XXX,XX +XXX,XX @@ void openrisc_cpu_dump_state(CPUState *cpu, FILE *f, int flags); int openrisc_cpu_gdb_read_register(CPUState *cpu, GByteArray *buf, int reg); int openrisc_cpu_gdb_write_register(CPUState *cpu, uint8_t *buf, int reg); void openrisc_translate_init(void); +void openrisc_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); int print_insn_or1k(bfd_vma addr, disassemble_info *info); #ifndef CONFIG_USER_ONLY diff --git a/target/ppc/cpu.h b/target/ppc/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/ppc/cpu.h +++ b/target/ppc/cpu.h @@ -XXX,XX +XXX,XX @@ extern const VMStateDescription vmstate_ppc_cpu; /*****************************************************************************/ void ppc_translate_init(void); +void ppc_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); #if !defined(CONFIG_USER_ONLY) void ppc_store_sdr1(CPUPPCState *env, target_ulong value); diff --git a/target/riscv/cpu.h b/target/riscv/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/riscv/cpu.h +++ b/target/riscv/cpu.h @@ -XXX,XX +XXX,XX @@ RISCVException smstateen_acc_ok(CPURISCVState *env, int index, uint64_t bit); void riscv_cpu_set_mode(CPURISCVState *env, target_ulong newpriv, bool virt_en); void riscv_translate_init(void); +void riscv_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); + G_NORETURN void riscv_raise_exception(CPURISCVState *env, uint32_t exception, uintptr_t pc); diff --git a/target/rx/cpu.h b/target/rx/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/rx/cpu.h +++ b/target/rx/cpu.h @@ -XXX,XX +XXX,XX @@ int rx_cpu_gdb_read_register(CPUState *cpu, GByteArray *buf, int reg); int rx_cpu_gdb_write_register(CPUState *cpu, uint8_t *buf, int reg); void rx_translate_init(void); +void rx_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); void rx_cpu_unpack_psw(CPURXState *env, uint32_t psw, int rte); #include "exec/cpu-all.h" diff --git a/target/s390x/s390x-internal.h b/target/s390x/s390x-internal.h index XXXXXXX..XXXXXXX 100644 --- a/target/s390x/s390x-internal.h +++ b/target/s390x/s390x-internal.h @@ -XXX,XX +XXX,XX @@ void handle_diag_308(CPUS390XState *env, uint64_t r1, uint64_t r3, /* translate.c */ void s390x_translate_init(void); +void s390x_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); void s390x_restore_state_to_opc(CPUState *cs, const TranslationBlock *tb, const uint64_t *data); diff --git a/target/sh4/cpu.h b/target/sh4/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/sh4/cpu.h +++ b/target/sh4/cpu.h @@ -XXX,XX +XXX,XX @@ G_NORETURN void superh_cpu_do_unaligned_access(CPUState *cpu, vaddr addr, uintptr_t retaddr); void sh4_translate_init(void); +void sh4_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); #if !defined(CONFIG_USER_ONLY) hwaddr superh_cpu_get_phys_page_debug(CPUState *cpu, vaddr addr); diff --git a/target/sparc/cpu.h b/target/sparc/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/sparc/cpu.h +++ b/target/sparc/cpu.h @@ -XXX,XX +XXX,XX @@ int sparc_cpu_memory_rw_debug(CPUState *cpu, vaddr addr, /* translate.c */ void sparc_tcg_init(void); +void sparc_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); /* fop_helper.c */ target_ulong cpu_get_fsr(CPUSPARCState *); diff --git a/target/tricore/cpu.h b/target/tricore/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/tricore/cpu.h +++ b/target/tricore/cpu.h @@ -XXX,XX +XXX,XX @@ FIELD(TB_FLAGS, PRIV, 0, 2) void cpu_state_reset(CPUTriCoreState *s); void tricore_tcg_init(void); +void tricore_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); static inline void cpu_get_tb_cpu_state(CPUTriCoreState *env, vaddr *pc, uint64_t *cs_base, uint32_t *flags) diff --git a/target/xtensa/cpu.h b/target/xtensa/cpu.h index XXXXXXX..XXXXXXX 100644 --- a/target/xtensa/cpu.h +++ b/target/xtensa/cpu.h @@ -XXX,XX +XXX,XX @@ G_NORETURN void xtensa_cpu_do_unaligned_access(CPUState *cpu, vaddr addr, void xtensa_collect_sr_names(const XtensaConfig *config); void xtensa_translate_init(void); +void xtensa_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc); void **xtensa_get_regfile_by_name(const char *name, int entries, int bits); void xtensa_breakpoint_handler(CPUState *cs); void xtensa_register_core(XtensaConfigList *node); diff --git a/accel/tcg/cpu-exec.c b/accel/tcg/cpu-exec.c index XXXXXXX..XXXXXXX 100644 --- a/accel/tcg/cpu-exec.c +++ b/accel/tcg/cpu-exec.c @@ -XXX,XX +XXX,XX @@ bool tcg_exec_realizefn(CPUState *cpu, Error **errp) if (!tcg_target_initialized) { /* Check mandatory TCGCPUOps handlers */ + const TCGCPUOps *tcg_ops = cpu->cc->tcg_ops; #ifndef CONFIG_USER_ONLY - assert(cpu->cc->tcg_ops->cpu_exec_halt); - assert(cpu->cc->tcg_ops->cpu_exec_interrupt); + assert(tcg_ops->cpu_exec_halt); + assert(tcg_ops->cpu_exec_interrupt); #endif /* !CONFIG_USER_ONLY */ - cpu->cc->tcg_ops->initialize(); + assert(tcg_ops->translate_code); + tcg_ops->initialize(); tcg_target_initialized = true; } diff --git a/accel/tcg/translate-all.c b/accel/tcg/translate-all.c index XXXXXXX..XXXXXXX 100644 --- a/accel/tcg/translate-all.c +++ b/accel/tcg/translate-all.c @@ -XXX,XX +XXX,XX @@ static int setjmp_gen_code(CPUArchState *env, TranslationBlock *tb, tcg_func_start(tcg_ctx); - tcg_ctx->cpu = env_cpu(env); - gen_intermediate_code(env_cpu(env), tb, max_insns, pc, host_pc); + CPUState *cs = env_cpu(env); + tcg_ctx->cpu = cs; + cs->cc->tcg_ops->translate_code(cs, tb, max_insns, pc, host_pc); + assert(tb->size != 0); tcg_ctx->cpu = NULL; *max_insns = tb->icount; @@ -XXX,XX +XXX,XX @@ TranslationBlock *tb_gen_code(CPUState *cpu, /* * Overflow of code_gen_buffer, or the current slice of it. * - * TODO: We don't need to re-do gen_intermediate_code, nor + * TODO: We don't need to re-do tcg_ops->translate_code, nor * should we re-do the tcg optimization currently hidden * inside tcg_gen_code. All that should be required is to * flush the TBs, allocate a new TB, re-initialize it per diff --git a/target/alpha/cpu.c b/target/alpha/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/alpha/cpu.c +++ b/target/alpha/cpu.c @@ -XXX,XX +XXX,XX @@ static const struct SysemuCPUOps alpha_sysemu_ops = { static const TCGCPUOps alpha_tcg_ops = { .initialize = alpha_translate_init, + .translate_code = alpha_translate_code, .synchronize_from_tb = alpha_cpu_synchronize_from_tb, .restore_state_to_opc = alpha_restore_state_to_opc, diff --git a/target/alpha/translate.c b/target/alpha/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/alpha/translate.c +++ b/target/alpha/translate.c @@ -XXX,XX +XXX,XX @@ static const TranslatorOps alpha_tr_ops = { .tb_stop = alpha_tr_tb_stop, }; -void gen_intermediate_code(CPUState *cpu, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void alpha_translate_code(CPUState *cpu, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext dc; translator_loop(cpu, tb, max_insns, pc, host_pc, &alpha_tr_ops, &dc.base); diff --git a/target/arm/cpu.c b/target/arm/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/cpu.c +++ b/target/arm/cpu.c @@ -XXX,XX +XXX,XX @@ static const struct SysemuCPUOps arm_sysemu_ops = { #ifdef CONFIG_TCG static const TCGCPUOps arm_tcg_ops = { .initialize = arm_translate_init, + .translate_code = arm_translate_code, .synchronize_from_tb = arm_cpu_synchronize_from_tb, .debug_excp_handler = arm_debug_excp_handler, .restore_state_to_opc = arm_restore_state_to_opc, diff --git a/target/arm/tcg/cpu-v7m.c b/target/arm/tcg/cpu-v7m.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/tcg/cpu-v7m.c +++ b/target/arm/tcg/cpu-v7m.c @@ -XXX,XX +XXX,XX @@ static void cortex_m55_initfn(Object *obj) static const TCGCPUOps arm_v7m_tcg_ops = { .initialize = arm_translate_init, + .translate_code = arm_translate_code, .synchronize_from_tb = arm_cpu_synchronize_from_tb, .debug_excp_handler = arm_debug_excp_handler, .restore_state_to_opc = arm_restore_state_to_opc, diff --git a/target/arm/tcg/translate.c b/target/arm/tcg/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/arm/tcg/translate.c +++ b/target/arm/tcg/translate.c @@ -XXX,XX +XXX,XX @@ static const TranslatorOps thumb_translator_ops = { .tb_stop = arm_tr_tb_stop, }; -/* generate intermediate code for basic block 'tb'. */ -void gen_intermediate_code(CPUState *cpu, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void arm_translate_code(CPUState *cpu, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext dc = { }; const TranslatorOps *ops = &arm_translator_ops; diff --git a/target/avr/cpu.c b/target/avr/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/avr/cpu.c +++ b/target/avr/cpu.c @@ -XXX,XX +XXX,XX @@ static const struct SysemuCPUOps avr_sysemu_ops = { static const TCGCPUOps avr_tcg_ops = { .initialize = avr_cpu_tcg_init, + .translate_code = avr_cpu_translate_code, .synchronize_from_tb = avr_cpu_synchronize_from_tb, .restore_state_to_opc = avr_restore_state_to_opc, .cpu_exec_interrupt = avr_cpu_exec_interrupt, diff --git a/target/avr/translate.c b/target/avr/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/avr/translate.c +++ b/target/avr/translate.c @@ -XXX,XX +XXX,XX @@ static bool trans_WDR(DisasContext *ctx, arg_WDR *a) * * - translate() * - canonicalize_skip() - * - gen_intermediate_code() + * - translate_code() * - restore_state_to_opc() * */ @@ -XXX,XX +XXX,XX @@ static const TranslatorOps avr_tr_ops = { .tb_stop = avr_tr_tb_stop, }; -void gen_intermediate_code(CPUState *cs, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void avr_cpu_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext dc = { }; translator_loop(cs, tb, max_insns, pc, host_pc, &avr_tr_ops, &dc.base); diff --git a/target/hexagon/cpu.c b/target/hexagon/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/hexagon/cpu.c +++ b/target/hexagon/cpu.c @@ -XXX,XX +XXX,XX @@ static void hexagon_cpu_init(Object *obj) static const TCGCPUOps hexagon_tcg_ops = { .initialize = hexagon_translate_init, + .translate_code = hexagon_translate_code, .synchronize_from_tb = hexagon_cpu_synchronize_from_tb, .restore_state_to_opc = hexagon_restore_state_to_opc, }; diff --git a/target/hexagon/translate.c b/target/hexagon/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/hexagon/translate.c +++ b/target/hexagon/translate.c @@ -XXX,XX +XXX,XX @@ static const TranslatorOps hexagon_tr_ops = { .tb_stop = hexagon_tr_tb_stop, }; -void gen_intermediate_code(CPUState *cs, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void hexagon_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext ctx; diff --git a/target/hppa/cpu.c b/target/hppa/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/hppa/cpu.c +++ b/target/hppa/cpu.c @@ -XXX,XX +XXX,XX @@ static const struct SysemuCPUOps hppa_sysemu_ops = { static const TCGCPUOps hppa_tcg_ops = { .initialize = hppa_translate_init, + .translate_code = hppa_translate_code, .synchronize_from_tb = hppa_cpu_synchronize_from_tb, .restore_state_to_opc = hppa_restore_state_to_opc, diff --git a/target/hppa/translate.c b/target/hppa/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/hppa/translate.c +++ b/target/hppa/translate.c @@ -XXX,XX +XXX,XX @@ static const TranslatorOps hppa_tr_ops = { #endif }; -void gen_intermediate_code(CPUState *cs, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void hppa_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext ctx = { }; translator_loop(cs, tb, max_insns, pc, host_pc, &hppa_tr_ops, &ctx.base); diff --git a/target/i386/tcg/tcg-cpu.c b/target/i386/tcg/tcg-cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/i386/tcg/tcg-cpu.c +++ b/target/i386/tcg/tcg-cpu.c @@ -XXX,XX +XXX,XX @@ static bool x86_debug_check_breakpoint(CPUState *cs) static const TCGCPUOps x86_tcg_ops = { .initialize = tcg_x86_init, + .translate_code = x86_translate_code, .synchronize_from_tb = x86_cpu_synchronize_from_tb, .restore_state_to_opc = x86_restore_state_to_opc, .cpu_exec_enter = x86_cpu_exec_enter, diff --git a/target/i386/tcg/translate.c b/target/i386/tcg/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/i386/tcg/translate.c +++ b/target/i386/tcg/translate.c @@ -XXX,XX +XXX,XX @@ static const TranslatorOps i386_tr_ops = { .tb_stop = i386_tr_tb_stop, }; -/* generate intermediate code for basic block 'tb'. */ -void gen_intermediate_code(CPUState *cpu, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void x86_translate_code(CPUState *cpu, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext dc; diff --git a/target/loongarch/cpu.c b/target/loongarch/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/loongarch/cpu.c +++ b/target/loongarch/cpu.c @@ -XXX,XX +XXX,XX @@ static void loongarch_cpu_dump_state(CPUState *cs, FILE *f, int flags) static const TCGCPUOps loongarch_tcg_ops = { .initialize = loongarch_translate_init, + .translate_code = loongarch_translate_code, .synchronize_from_tb = loongarch_cpu_synchronize_from_tb, .restore_state_to_opc = loongarch_restore_state_to_opc, diff --git a/target/loongarch/tcg/translate.c b/target/loongarch/tcg/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/loongarch/tcg/translate.c +++ b/target/loongarch/tcg/translate.c @@ -XXX,XX +XXX,XX @@ static const TranslatorOps loongarch_tr_ops = { .tb_stop = loongarch_tr_tb_stop, }; -void gen_intermediate_code(CPUState *cs, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void loongarch_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext ctx; diff --git a/target/m68k/cpu.c b/target/m68k/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/m68k/cpu.c +++ b/target/m68k/cpu.c @@ -XXX,XX +XXX,XX @@ static const struct SysemuCPUOps m68k_sysemu_ops = { static const TCGCPUOps m68k_tcg_ops = { .initialize = m68k_tcg_init, + .translate_code = m68k_translate_code, .restore_state_to_opc = m68k_restore_state_to_opc, #ifndef CONFIG_USER_ONLY diff --git a/target/m68k/translate.c b/target/m68k/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/m68k/translate.c +++ b/target/m68k/translate.c @@ -XXX,XX +XXX,XX @@ static const TranslatorOps m68k_tr_ops = { .tb_stop = m68k_tr_tb_stop, }; -void gen_intermediate_code(CPUState *cpu, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void m68k_translate_code(CPUState *cpu, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext dc; translator_loop(cpu, tb, max_insns, pc, host_pc, &m68k_tr_ops, &dc.base); diff --git a/target/microblaze/cpu.c b/target/microblaze/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/microblaze/cpu.c +++ b/target/microblaze/cpu.c @@ -XXX,XX +XXX,XX @@ static const struct SysemuCPUOps mb_sysemu_ops = { static const TCGCPUOps mb_tcg_ops = { .initialize = mb_tcg_init, + .translate_code = mb_translate_code, .synchronize_from_tb = mb_cpu_synchronize_from_tb, .restore_state_to_opc = mb_restore_state_to_opc, diff --git a/target/microblaze/translate.c b/target/microblaze/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/microblaze/translate.c +++ b/target/microblaze/translate.c @@ -XXX,XX +XXX,XX @@ static const TranslatorOps mb_tr_ops = { .tb_stop = mb_tr_tb_stop, }; -void gen_intermediate_code(CPUState *cpu, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void mb_translate_code(CPUState *cpu, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext dc; translator_loop(cpu, tb, max_insns, pc, host_pc, &mb_tr_ops, &dc.base); diff --git a/target/mips/cpu.c b/target/mips/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/mips/cpu.c +++ b/target/mips/cpu.c @@ -XXX,XX +XXX,XX @@ static const Property mips_cpu_properties[] = { #include "hw/core/tcg-cpu-ops.h" static const TCGCPUOps mips_tcg_ops = { .initialize = mips_tcg_init, + .translate_code = mips_translate_code, .synchronize_from_tb = mips_cpu_synchronize_from_tb, .restore_state_to_opc = mips_restore_state_to_opc, diff --git a/target/mips/tcg/translate.c b/target/mips/tcg/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/mips/tcg/translate.c +++ b/target/mips/tcg/translate.c @@ -XXX,XX +XXX,XX @@ static const TranslatorOps mips_tr_ops = { .tb_stop = mips_tr_tb_stop, }; -void gen_intermediate_code(CPUState *cs, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void mips_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext ctx; diff --git a/target/openrisc/cpu.c b/target/openrisc/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/openrisc/cpu.c +++ b/target/openrisc/cpu.c @@ -XXX,XX +XXX,XX @@ static const struct SysemuCPUOps openrisc_sysemu_ops = { static const TCGCPUOps openrisc_tcg_ops = { .initialize = openrisc_translate_init, + .translate_code = openrisc_translate_code, .synchronize_from_tb = openrisc_cpu_synchronize_from_tb, .restore_state_to_opc = openrisc_restore_state_to_opc, diff --git a/target/openrisc/translate.c b/target/openrisc/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/openrisc/translate.c +++ b/target/openrisc/translate.c @@ -XXX,XX +XXX,XX @@ static const TranslatorOps openrisc_tr_ops = { .tb_stop = openrisc_tr_tb_stop, }; -void gen_intermediate_code(CPUState *cs, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void openrisc_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext ctx; diff --git a/target/ppc/cpu_init.c b/target/ppc/cpu_init.c index XXXXXXX..XXXXXXX 100644 --- a/target/ppc/cpu_init.c +++ b/target/ppc/cpu_init.c @@ -XXX,XX +XXX,XX @@ static const struct SysemuCPUOps ppc_sysemu_ops = { static const TCGCPUOps ppc_tcg_ops = { .initialize = ppc_translate_init, + .translate_code = ppc_translate_code, .restore_state_to_opc = ppc_restore_state_to_opc, #ifdef CONFIG_USER_ONLY diff --git a/target/ppc/translate.c b/target/ppc/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/ppc/translate.c +++ b/target/ppc/translate.c @@ -XXX,XX +XXX,XX @@ static const TranslatorOps ppc_tr_ops = { .tb_stop = ppc_tr_tb_stop, }; -void gen_intermediate_code(CPUState *cs, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void ppc_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext ctx; diff --git a/target/riscv/tcg/tcg-cpu.c b/target/riscv/tcg/tcg-cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/riscv/tcg/tcg-cpu.c +++ b/target/riscv/tcg/tcg-cpu.c @@ -XXX,XX +XXX,XX @@ static void riscv_restore_state_to_opc(CPUState *cs, static const TCGCPUOps riscv_tcg_ops = { .initialize = riscv_translate_init, + .translate_code = riscv_translate_code, .synchronize_from_tb = riscv_cpu_synchronize_from_tb, .restore_state_to_opc = riscv_restore_state_to_opc, diff --git a/target/riscv/translate.c b/target/riscv/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/riscv/translate.c +++ b/target/riscv/translate.c @@ -XXX,XX +XXX,XX @@ static const TranslatorOps riscv_tr_ops = { .tb_stop = riscv_tr_tb_stop, }; -void gen_intermediate_code(CPUState *cs, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void riscv_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext ctx; diff --git a/target/rx/cpu.c b/target/rx/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/rx/cpu.c +++ b/target/rx/cpu.c @@ -XXX,XX +XXX,XX @@ static const struct SysemuCPUOps rx_sysemu_ops = { static const TCGCPUOps rx_tcg_ops = { .initialize = rx_translate_init, + .translate_code = rx_translate_code, .synchronize_from_tb = rx_cpu_synchronize_from_tb, .restore_state_to_opc = rx_restore_state_to_opc, .tlb_fill = rx_cpu_tlb_fill, diff --git a/target/rx/translate.c b/target/rx/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/rx/translate.c +++ b/target/rx/translate.c @@ -XXX,XX +XXX,XX @@ static const TranslatorOps rx_tr_ops = { .tb_stop = rx_tr_tb_stop, }; -void gen_intermediate_code(CPUState *cs, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void rx_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext dc; diff --git a/target/s390x/cpu.c b/target/s390x/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/s390x/cpu.c +++ b/target/s390x/cpu.c @@ -XXX,XX +XXX,XX @@ void cpu_get_tb_cpu_state(CPUS390XState *env, vaddr *pc, static const TCGCPUOps s390_tcg_ops = { .initialize = s390x_translate_init, + .translate_code = s390x_translate_code, .restore_state_to_opc = s390x_restore_state_to_opc, #ifdef CONFIG_USER_ONLY diff --git a/target/s390x/tcg/translate.c b/target/s390x/tcg/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/s390x/tcg/translate.c +++ b/target/s390x/tcg/translate.c @@ -XXX,XX +XXX,XX @@ static const TranslatorOps s390x_tr_ops = { .disas_log = s390x_tr_disas_log, }; -void gen_intermediate_code(CPUState *cs, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void s390x_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext dc; diff --git a/target/sh4/cpu.c b/target/sh4/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/sh4/cpu.c +++ b/target/sh4/cpu.c @@ -XXX,XX +XXX,XX @@ static const struct SysemuCPUOps sh4_sysemu_ops = { static const TCGCPUOps superh_tcg_ops = { .initialize = sh4_translate_init, + .translate_code = sh4_translate_code, .synchronize_from_tb = superh_cpu_synchronize_from_tb, .restore_state_to_opc = superh_restore_state_to_opc, diff --git a/target/sh4/translate.c b/target/sh4/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/sh4/translate.c +++ b/target/sh4/translate.c @@ -XXX,XX +XXX,XX @@ static const TranslatorOps sh4_tr_ops = { .tb_stop = sh4_tr_tb_stop, }; -void gen_intermediate_code(CPUState *cs, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void sh4_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext ctx; diff --git a/target/sparc/cpu.c b/target/sparc/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/sparc/cpu.c +++ b/target/sparc/cpu.c @@ -XXX,XX +XXX,XX @@ static const struct SysemuCPUOps sparc_sysemu_ops = { static const TCGCPUOps sparc_tcg_ops = { .initialize = sparc_tcg_init, + .translate_code = sparc_translate_code, .synchronize_from_tb = sparc_cpu_synchronize_from_tb, .restore_state_to_opc = sparc_restore_state_to_opc, diff --git a/target/sparc/translate.c b/target/sparc/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/sparc/translate.c +++ b/target/sparc/translate.c @@ -XXX,XX +XXX,XX @@ static const TranslatorOps sparc_tr_ops = { .tb_stop = sparc_tr_tb_stop, }; -void gen_intermediate_code(CPUState *cs, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void sparc_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext dc = {}; diff --git a/target/tricore/cpu.c b/target/tricore/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/tricore/cpu.c +++ b/target/tricore/cpu.c @@ -XXX,XX +XXX,XX @@ static const struct SysemuCPUOps tricore_sysemu_ops = { static const TCGCPUOps tricore_tcg_ops = { .initialize = tricore_tcg_init, + .translate_code = tricore_translate_code, .synchronize_from_tb = tricore_cpu_synchronize_from_tb, .restore_state_to_opc = tricore_restore_state_to_opc, .tlb_fill = tricore_cpu_tlb_fill, diff --git a/target/tricore/translate.c b/target/tricore/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/tricore/translate.c +++ b/target/tricore/translate.c @@ -XXX,XX +XXX,XX @@ static const TranslatorOps tricore_tr_ops = { .tb_stop = tricore_tr_tb_stop, }; - -void gen_intermediate_code(CPUState *cs, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void tricore_translate_code(CPUState *cs, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext ctx; translator_loop(cs, tb, max_insns, pc, host_pc, diff --git a/target/xtensa/cpu.c b/target/xtensa/cpu.c index XXXXXXX..XXXXXXX 100644 --- a/target/xtensa/cpu.c +++ b/target/xtensa/cpu.c @@ -XXX,XX +XXX,XX @@ static const struct SysemuCPUOps xtensa_sysemu_ops = { static const TCGCPUOps xtensa_tcg_ops = { .initialize = xtensa_translate_init, + .translate_code = xtensa_translate_code, .debug_excp_handler = xtensa_breakpoint_handler, .restore_state_to_opc = xtensa_restore_state_to_opc, diff --git a/target/xtensa/translate.c b/target/xtensa/translate.c index XXXXXXX..XXXXXXX 100644 --- a/target/xtensa/translate.c +++ b/target/xtensa/translate.c @@ -XXX,XX +XXX,XX @@ static const TranslatorOps xtensa_translator_ops = { .tb_stop = xtensa_tr_tb_stop, }; -void gen_intermediate_code(CPUState *cpu, TranslationBlock *tb, int *max_insns, - vaddr pc, void *host_pc) +void xtensa_translate_code(CPUState *cpu, TranslationBlock *tb, + int *max_insns, vaddr pc, void *host_pc) { DisasContext dc = {}; translator_loop(cpu, tb, max_insns, pc, host_pc, -- 2.43.0