The following changes since commit 0a301624c2f4ced3331ffd5bce85b4274fe132af: Merge remote-tracking branch 'remotes/pmaydell/tags/pull-target-arm-20220208' into staging (2022-02-08 11:40:08 +0000) are available in the Git repository at: https://gitlab.com/rth7680/qemu.git tags/pull-tcg-20220211 for you to fetch changes up to 5c1a101ef6b85537a4ade93c39ea81cadd5c246e: tests/tcg/multiarch: Add sigbus.c (2022-02-09 09:00:01 +1100) ---------------------------------------------------------------- Fix safe_syscall_base for sparc64. Fix host signal handling for sparc64-linux. Speedups for jump cache and work list probing. Fix for exception replays. Raise guest SIGBUS for user-only misaligned accesses. ---------------------------------------------------------------- Idan Horowitz (2): accel/tcg: Optimize jump cache flush during tlb range flush softmmu/cpus: Check if the cpu work list is empty atomically Pavel Dovgalyuk (1): replay: use CF_NOIRQ for special exception-replaying TB Richard Henderson (29): common-user/host/sparc64: Fix safe_syscall_base linux-user: Introduce host_signal_mask linux-user: Introduce host_sigcontext linux-user: Move sparc/host-signal.h to sparc64/host-signal.h linux-user/include/host/sparc64: Fix host_sigcontext tcg/i386: Support raising sigbus for user-only tcg/aarch64: Support raising sigbus for user-only tcg/ppc: Support raising sigbus for user-only tcg/riscv: Support raising sigbus for user-only tcg/s390x: Support raising sigbus for user-only tcg/tci: Support raising sigbus for user-only tcg/arm: Drop support for armv4 and armv5 hosts tcg/arm: Remove use_armv5t_instructions tcg/arm: Remove use_armv6_instructions tcg/arm: Check alignment for ldrd and strd tcg/arm: Support unaligned access for softmmu tcg/arm: Reserve a register for guest_base tcg/arm: Support raising sigbus for user-only tcg/mips: Support unaligned access for user-only tcg/mips: Support unaligned access for softmmu tcg/sparc: Use tcg_out_movi_imm13 in tcg_out_addsub2_i64 tcg/sparc: Split out tcg_out_movi_imm32 tcg/sparc: Add scratch argument to tcg_out_movi_int tcg/sparc: Improve code gen for shifted 32-bit constants tcg/sparc: Convert patch_reloc to return bool tcg/sparc: Use the constant pool for 64-bit constants tcg/sparc: Add tcg_out_jmpl_const for better tail calls tcg/sparc: Support unaligned access for user-only tests/tcg/multiarch: Add sigbus.c WANG Xuerui (2): tcg/loongarch64: Fix fallout from recent MO_Q renaming tcg/loongarch64: Support raising sigbus for user-only linux-user/include/host/aarch64/host-signal.h | 16 +- linux-user/include/host/alpha/host-signal.h | 14 +- linux-user/include/host/arm/host-signal.h | 14 +- linux-user/include/host/i386/host-signal.h | 14 +- linux-user/include/host/loongarch64/host-signal.h | 14 +- linux-user/include/host/mips/host-signal.h | 14 +- linux-user/include/host/ppc/host-signal.h | 14 +- linux-user/include/host/riscv/host-signal.h | 14 +- linux-user/include/host/s390/host-signal.h | 14 +- linux-user/include/host/sparc/host-signal.h | 63 ---- linux-user/include/host/sparc64/host-signal.h | 65 +++- linux-user/include/host/x86_64/host-signal.h | 14 +- tcg/aarch64/tcg-target.h | 2 - tcg/arm/tcg-target.h | 6 +- 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 - accel/tcg/cpu-exec.c | 3 +- accel/tcg/cputlb.c | 9 + linux-user/signal.c | 22 +- softmmu/cpus.c | 7 +- tcg/tci.c | 20 +- tests/tcg/multiarch/sigbus.c | 68 ++++ tcg/aarch64/tcg-target.c.inc | 91 ++++- tcg/arm/tcg-target.c.inc | 410 +++++++++------------- tcg/i386/tcg-target.c.inc | 103 +++++- tcg/loongarch64/tcg-target.c.inc | 73 +++- tcg/mips/tcg-target.c.inc | 387 ++++++++++++++++++-- tcg/ppc/tcg-target.c.inc | 98 +++++- tcg/riscv/tcg-target.c.inc | 63 +++- tcg/s390x/tcg-target.c.inc | 59 +++- tcg/sparc/tcg-target.c.inc | 348 +++++++++++++++--- common-user/host/sparc64/safe-syscall.inc.S | 5 +- 36 files changed, 1561 insertions(+), 495 deletions(-) delete mode 100644 linux-user/include/host/sparc/host-signal.h create mode 100644 tests/tcg/multiarch/sigbus.c Use the "retl" instead of "ret" instruction alias, since we do not allocate a register window in this function. Fix the offset to the first stacked parameter, which lies beyond the register window save area. Fixes: 95c021dac835 ("linux-user/host/sparc64: Add safe-syscall.inc.S") Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- common-user/host/sparc64/safe-syscall.inc.S | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/common-user/host/sparc64/safe-syscall.inc.S b/common-user/host/sparc64/safe-syscall.inc.S index XXXXXXX..XXXXXXX 100644 --- a/common-user/host/sparc64/safe-syscall.inc.S +++ b/common-user/host/sparc64/safe-syscall.inc.S @@ -XXX,XX +XXX,XX @@ .type safe_syscall_end, @function #define STACK_BIAS 2047 -#define PARAM(N) STACK_BIAS + N*8 +#define WINDOW_SIZE 16 * 8 +#define PARAM(N) STACK_BIAS + WINDOW_SIZE + N * 8 /* * This is the entry point for making a system call. The calling @@ -XXX,XX +XXX,XX @@ safe_syscall_end: /* code path for having successfully executed the syscall */ bcs,pn %xcc, 1f nop - ret + retl nop /* code path when we didn't execute the syscall */ -- 2.25.1 Do not directly access the uc_sigmask member. This is preparation for a sparc64 fix. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- linux-user/include/host/aarch64/host-signal.h | 5 +++++ linux-user/include/host/alpha/host-signal.h | 5 +++++ linux-user/include/host/arm/host-signal.h | 5 +++++ linux-user/include/host/i386/host-signal.h | 5 +++++ .../include/host/loongarch64/host-signal.h | 5 +++++ linux-user/include/host/mips/host-signal.h | 5 +++++ linux-user/include/host/ppc/host-signal.h | 5 +++++ linux-user/include/host/riscv/host-signal.h | 5 +++++ linux-user/include/host/s390/host-signal.h | 5 +++++ linux-user/include/host/sparc/host-signal.h | 5 +++++ linux-user/include/host/x86_64/host-signal.h | 5 +++++ linux-user/signal.c | 18 ++++++++---------- 12 files changed, 63 insertions(+), 10 deletions(-) diff --git a/linux-user/include/host/aarch64/host-signal.h b/linux-user/include/host/aarch64/host-signal.h index XXXXXXX..XXXXXXX 100644 --- a/linux-user/include/host/aarch64/host-signal.h +++ b/linux-user/include/host/aarch64/host-signal.h @@ -XXX,XX +XXX,XX @@ static inline void host_signal_set_pc(ucontext_t *uc, uintptr_t pc) uc->uc_mcontext.pc = pc; } +static inline void *host_signal_mask(ucontext_t *uc) +{ + return &uc->uc_sigmask; +} + static inline bool host_signal_write(siginfo_t *info, ucontext_t *uc) { struct _aarch64_ctx *hdr; diff --git a/linux-user/include/host/alpha/host-signal.h b/linux-user/include/host/alpha/host-signal.h index XXXXXXX..XXXXXXX 100644 --- a/linux-user/include/host/alpha/host-signal.h +++ b/linux-user/include/host/alpha/host-signal.h @@ -XXX,XX +XXX,XX @@ static inline void host_signal_set_pc(ucontext_t *uc, uintptr_t pc) uc->uc_mcontext.sc_pc = pc; } +static inline void *host_signal_mask(ucontext_t *uc) +{ + return &uc->uc_sigmask; +} + static inline bool host_signal_write(siginfo_t *info, ucontext_t *uc) { uint32_t *pc = (uint32_t *)host_signal_pc(uc); diff --git a/linux-user/include/host/arm/host-signal.h b/linux-user/include/host/arm/host-signal.h index XXXXXXX..XXXXXXX 100644 --- a/linux-user/include/host/arm/host-signal.h +++ b/linux-user/include/host/arm/host-signal.h @@ -XXX,XX +XXX,XX @@ static inline void host_signal_set_pc(ucontext_t *uc, uintptr_t pc) uc->uc_mcontext.arm_pc = pc; } +static inline void *host_signal_mask(ucontext_t *uc) +{ + return &uc->uc_sigmask; +} + static inline bool host_signal_write(siginfo_t *info, ucontext_t *uc) { /* diff --git a/linux-user/include/host/i386/host-signal.h b/linux-user/include/host/i386/host-signal.h index XXXXXXX..XXXXXXX 100644 --- a/linux-user/include/host/i386/host-signal.h +++ b/linux-user/include/host/i386/host-signal.h @@ -XXX,XX +XXX,XX @@ static inline void host_signal_set_pc(ucontext_t *uc, uintptr_t pc) uc->uc_mcontext.gregs[REG_EIP] = pc; } +static inline void *host_signal_mask(ucontext_t *uc) +{ + return &uc->uc_sigmask; +} + static inline bool host_signal_write(siginfo_t *info, ucontext_t *uc) { return uc->uc_mcontext.gregs[REG_TRAPNO] == 0xe diff --git a/linux-user/include/host/loongarch64/host-signal.h b/linux-user/include/host/loongarch64/host-signal.h index XXXXXXX..XXXXXXX 100644 --- a/linux-user/include/host/loongarch64/host-signal.h +++ b/linux-user/include/host/loongarch64/host-signal.h @@ -XXX,XX +XXX,XX @@ static inline void host_signal_set_pc(ucontext_t *uc, uintptr_t pc) uc->uc_mcontext.__pc = pc; } +static inline void *host_signal_mask(ucontext_t *uc) +{ + return &uc->uc_sigmask; +} + static inline bool host_signal_write(siginfo_t *info, ucontext_t *uc) { const uint32_t *pinsn = (const uint32_t *)host_signal_pc(uc); diff --git a/linux-user/include/host/mips/host-signal.h b/linux-user/include/host/mips/host-signal.h index XXXXXXX..XXXXXXX 100644 --- a/linux-user/include/host/mips/host-signal.h +++ b/linux-user/include/host/mips/host-signal.h @@ -XXX,XX +XXX,XX @@ static inline void host_signal_set_pc(ucontext_t *uc, uintptr_t pc) uc->uc_mcontext.pc = pc; } +static inline void *host_signal_mask(ucontext_t *uc) +{ + return &uc->uc_sigmask; +} + #if defined(__misp16) || defined(__mips_micromips) #error "Unsupported encoding" #endif diff --git a/linux-user/include/host/ppc/host-signal.h b/linux-user/include/host/ppc/host-signal.h index XXXXXXX..XXXXXXX 100644 --- a/linux-user/include/host/ppc/host-signal.h +++ b/linux-user/include/host/ppc/host-signal.h @@ -XXX,XX +XXX,XX @@ static inline void host_signal_set_pc(ucontext_t *uc, uintptr_t pc) uc->uc_mcontext.regs->nip = pc; } +static inline void *host_signal_mask(ucontext_t *uc) +{ + return &uc->uc_sigmask; +} + static inline bool host_signal_write(siginfo_t *info, ucontext_t *uc) { return uc->uc_mcontext.regs->trap != 0x400 diff --git a/linux-user/include/host/riscv/host-signal.h b/linux-user/include/host/riscv/host-signal.h index XXXXXXX..XXXXXXX 100644 --- a/linux-user/include/host/riscv/host-signal.h +++ b/linux-user/include/host/riscv/host-signal.h @@ -XXX,XX +XXX,XX @@ static inline void host_signal_set_pc(ucontext_t *uc, uintptr_t pc) uc->uc_mcontext.__gregs[REG_PC] = pc; } +static inline void *host_signal_mask(ucontext_t *uc) +{ + return &uc->uc_sigmask; +} + static inline bool host_signal_write(siginfo_t *info, ucontext_t *uc) { /* diff --git a/linux-user/include/host/s390/host-signal.h b/linux-user/include/host/s390/host-signal.h index XXXXXXX..XXXXXXX 100644 --- a/linux-user/include/host/s390/host-signal.h +++ b/linux-user/include/host/s390/host-signal.h @@ -XXX,XX +XXX,XX @@ static inline void host_signal_set_pc(ucontext_t *uc, uintptr_t pc) uc->uc_mcontext.psw.addr = pc; } +static inline void *host_signal_mask(ucontext_t *uc) +{ + return &uc->uc_sigmask; +} + static inline bool host_signal_write(siginfo_t *info, ucontext_t *uc) { uint16_t *pinsn = (uint16_t *)host_signal_pc(uc); diff --git a/linux-user/include/host/sparc/host-signal.h b/linux-user/include/host/sparc/host-signal.h index XXXXXXX..XXXXXXX 100644 --- a/linux-user/include/host/sparc/host-signal.h +++ b/linux-user/include/host/sparc/host-signal.h @@ -XXX,XX +XXX,XX @@ static inline void host_signal_set_pc(ucontext_t *uc, uintptr_t pc) #endif } +static inline void *host_signal_mask(ucontext_t *uc) +{ + return &uc->uc_sigmask; +} + static inline bool host_signal_write(siginfo_t *info, ucontext_t *uc) { uint32_t insn = *(uint32_t *)host_signal_pc(uc); diff --git a/linux-user/include/host/x86_64/host-signal.h b/linux-user/include/host/x86_64/host-signal.h index XXXXXXX..XXXXXXX 100644 --- a/linux-user/include/host/x86_64/host-signal.h +++ b/linux-user/include/host/x86_64/host-signal.h @@ -XXX,XX +XXX,XX @@ static inline void host_signal_set_pc(ucontext_t *uc, uintptr_t pc) uc->uc_mcontext.gregs[REG_RIP] = pc; } +static inline void *host_signal_mask(ucontext_t *uc) +{ + return &uc->uc_sigmask; +} + static inline bool host_signal_write(siginfo_t *info, ucontext_t *uc) { return uc->uc_mcontext.gregs[REG_TRAPNO] == 0xe diff --git a/linux-user/signal.c b/linux-user/signal.c index XXXXXXX..XXXXXXX 100644 --- a/linux-user/signal.c +++ b/linux-user/signal.c @@ -XXX,XX +XXX,XX @@ static void host_signal_handler(int host_sig, siginfo_t *info, void *puc) int guest_sig; uintptr_t pc = 0; bool sync_sig = false; + void *sigmask = host_signal_mask(uc); /* * Non-spoofed SIGSEGV and SIGBUS are synchronous, and need special @@ -XXX,XX +XXX,XX @@ static void host_signal_handler(int host_sig, siginfo_t *info, void *puc) if (info->si_code == SEGV_ACCERR && h2g_valid(host_addr)) { /* If this was a write to a TB protected page, restart. */ if (is_write && - handle_sigsegv_accerr_write(cpu, &uc->uc_sigmask, - pc, guest_addr)) { + handle_sigsegv_accerr_write(cpu, sigmask, pc, guest_addr)) { return; } @@ -XXX,XX +XXX,XX @@ static void host_signal_handler(int host_sig, siginfo_t *info, void *puc) } } - sigprocmask(SIG_SETMASK, &uc->uc_sigmask, NULL); + sigprocmask(SIG_SETMASK, sigmask, NULL); cpu_loop_exit_sigsegv(cpu, guest_addr, access_type, maperr, pc); } else { - sigprocmask(SIG_SETMASK, &uc->uc_sigmask, NULL); + sigprocmask(SIG_SETMASK, sigmask, NULL); if (info->si_code == BUS_ADRALN) { cpu_loop_exit_sigbus(cpu, guest_addr, access_type, pc); } @@ -XXX,XX +XXX,XX @@ static void host_signal_handler(int host_sig, siginfo_t *info, void *puc) * now and it getting out to the main loop. Signals will be * unblocked again in process_pending_signals(). * - * WARNING: we cannot use sigfillset() here because the uc_sigmask + * WARNING: we cannot use sigfillset() here because the sigmask * field is a kernel sigset_t, which is much smaller than the * libc sigset_t which sigfillset() operates on. Using sigfillset() * would write 0xff bytes off the end of the structure and trash * data on the struct. - * We can't use sizeof(uc->uc_sigmask) either, because the libc - * headers define the struct field with the wrong (too large) type. */ - memset(&uc->uc_sigmask, 0xff, SIGSET_T_SIZE); - sigdelset(&uc->uc_sigmask, SIGSEGV); - sigdelset(&uc->uc_sigmask, SIGBUS); + memset(sigmask, 0xff, SIGSET_T_SIZE); + sigdelset(sigmask, SIGSEGV); + sigdelset(sigmask, SIGBUS); /* interrupt the virtual CPU as soon as possible */ cpu_exit(thread_cpu); -- 2.25.1 Do not directly access ucontext_t as the third signal parameter. This is preparation for a sparc64 fix. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- linux-user/include/host/aarch64/host-signal.h | 13 ++++++++----- linux-user/include/host/alpha/host-signal.h | 11 +++++++---- linux-user/include/host/arm/host-signal.h | 11 +++++++---- linux-user/include/host/i386/host-signal.h | 11 +++++++---- linux-user/include/host/loongarch64/host-signal.h | 11 +++++++---- linux-user/include/host/mips/host-signal.h | 11 +++++++---- linux-user/include/host/ppc/host-signal.h | 11 +++++++---- linux-user/include/host/riscv/host-signal.h | 11 +++++++---- linux-user/include/host/s390/host-signal.h | 11 +++++++---- linux-user/include/host/sparc/host-signal.h | 11 +++++++---- linux-user/include/host/x86_64/host-signal.h | 11 +++++++---- linux-user/signal.c | 4 ++-- 12 files changed, 80 insertions(+), 47 deletions(-) diff --git a/linux-user/include/host/aarch64/host-signal.h b/linux-user/include/host/aarch64/host-signal.h index XXXXXXX..XXXXXXX 100644 --- a/linux-user/include/host/aarch64/host-signal.h +++ b/linux-user/include/host/aarch64/host-signal.h @@ -XXX,XX +XXX,XX @@ #ifndef AARCH64_HOST_SIGNAL_H #define AARCH64_HOST_SIGNAL_H +/* The third argument to a SA_SIGINFO handler is ucontext_t. */ +typedef ucontext_t host_sigcontext; + /* Pre-3.16 kernel headers don't have these, so provide fallback definitions */ #ifndef ESR_MAGIC #define ESR_MAGIC 0x45535201 @@ -XXX,XX +XXX,XX @@ struct esr_context { }; #endif -static inline struct _aarch64_ctx *first_ctx(ucontext_t *uc) +static inline struct _aarch64_ctx *first_ctx(host_sigcontext *uc) { return (struct _aarch64_ctx *)&uc->uc_mcontext.__reserved; } @@ -XXX,XX +XXX,XX @@ static inline struct _aarch64_ctx *next_ctx(struct _aarch64_ctx *hdr) return (struct _aarch64_ctx *)((char *)hdr + hdr->size); } -static inline uintptr_t host_signal_pc(ucontext_t *uc) +static inline uintptr_t host_signal_pc(host_sigcontext *uc) { return uc->uc_mcontext.pc; } -static inline void host_signal_set_pc(ucontext_t *uc, uintptr_t pc) +static inline void host_signal_set_pc(host_sigcontext *uc, uintptr_t pc) { uc->uc_mcontext.pc = pc; } -static inline void *host_signal_mask(ucontext_t *uc) +static inline void *host_signal_mask(host_sigcontext *uc) { return &uc->uc_sigmask; } -static inline bool host_signal_write(siginfo_t *info, ucontext_t *uc) +static inline bool host_signal_write(siginfo_t *info, host_sigcontext *uc) { struct _aarch64_ctx *hdr; uint32_t insn; diff --git a/linux-user/include/host/alpha/host-signal.h b/linux-user/include/host/alpha/host-signal.h index XXXXXXX..XXXXXXX 100644 --- a/linux-user/include/host/alpha/host-signal.h +++ b/linux-user/include/host/alpha/host-signal.h @@ -XXX,XX +XXX,XX @@ #ifndef ALPHA_HOST_SIGNAL_H #define ALPHA_HOST_SIGNAL_H -static inline uintptr_t host_signal_pc(ucontext_t *uc) +/* The third argument to a SA_SIGINFO handler is ucontext_t. */ +typedef ucontext_t host_sigcontext; + +static inline uintptr_t host_signal_pc(host_sigcontext *uc) { return uc->uc_mcontext.sc_pc; } -static inline void host_signal_set_pc(ucontext_t *uc, uintptr_t pc) +static inline void host_signal_set_pc(host_sigcontext *uc, uintptr_t pc) { uc->uc_mcontext.sc_pc = pc; } -static inline void *host_signal_mask(ucontext_t *uc) +static inline void *host_signal_mask(host_sigcontext *uc) { return &uc->uc_sigmask; } -static inline bool host_signal_write(siginfo_t *info, ucontext_t *uc) +static inline bool host_signal_write(siginfo_t *info, host_sigcontext *uc) { uint32_t *pc = (uint32_t *)host_signal_pc(uc); uint32_t insn = *pc; diff --git a/linux-user/include/host/arm/host-signal.h b/linux-user/include/host/arm/host-signal.h index XXXXXXX..XXXXXXX 100644 --- a/linux-user/include/host/arm/host-signal.h +++ b/linux-user/include/host/arm/host-signal.h @@ -XXX,XX +XXX,XX @@ #ifndef ARM_HOST_SIGNAL_H #define ARM_HOST_SIGNAL_H -static inline uintptr_t host_signal_pc(ucontext_t *uc) +/* The third argument to a SA_SIGINFO handler is ucontext_t. */ +typedef ucontext_t host_sigcontext; + +static inline uintptr_t host_signal_pc(host_sigcontext *uc) { return uc->uc_mcontext.arm_pc; } -static inline void host_signal_set_pc(ucontext_t *uc, uintptr_t pc) +static inline void host_signal_set_pc(host_sigcontext *uc, uintptr_t pc) { uc->uc_mcontext.arm_pc = pc; } -static inline void *host_signal_mask(ucontext_t *uc) +static inline void *host_signal_mask(host_sigcontext *uc) { return &uc->uc_sigmask; } -static inline bool host_signal_write(siginfo_t *info, ucontext_t *uc) +static inline bool host_signal_write(siginfo_t *info, host_sigcontext *uc) { /* * In the FSR, bit 11 is WnR, assuming a v6 or diff --git a/linux-user/include/host/i386/host-signal.h b/linux-user/include/host/i386/host-signal.h index XXXXXXX..XXXXXXX 100644 --- a/linux-user/include/host/i386/host-signal.h +++ b/linux-user/include/host/i386/host-signal.h @@ -XXX,XX +XXX,XX @@ #ifndef I386_HOST_SIGNAL_H #define I386_HOST_SIGNAL_H -static inline uintptr_t host_signal_pc(ucontext_t *uc) +/* The third argument to a SA_SIGINFO handler is ucontext_t. */ +typedef ucontext_t host_sigcontext; + +static inline uintptr_t host_signal_pc(host_sigcontext *uc) { return uc->uc_mcontext.gregs[REG_EIP]; } -static inline void host_signal_set_pc(ucontext_t *uc, uintptr_t pc) +static inline void host_signal_set_pc(host_sigcontext *uc, uintptr_t pc) { uc->uc_mcontext.gregs[REG_EIP] = pc; } -static inline void *host_signal_mask(ucontext_t *uc) +static inline void *host_signal_mask(host_sigcontext *uc) { return &uc->uc_sigmask; } -static inline bool host_signal_write(siginfo_t *info, ucontext_t *uc) +static inline bool host_signal_write(siginfo_t *info, host_sigcontext *uc) { return uc->uc_mcontext.gregs[REG_TRAPNO] == 0xe && (uc->uc_mcontext.gregs[REG_ERR] & 0x2); diff --git a/linux-user/include/host/loongarch64/host-signal.h b/linux-user/include/host/loongarch64/host-signal.h index XXXXXXX..XXXXXXX 100644 --- a/linux-user/include/host/loongarch64/host-signal.h +++ b/linux-user/include/host/loongarch64/host-signal.h @@ -XXX,XX +XXX,XX @@ #ifndef LOONGARCH64_HOST_SIGNAL_H #define LOONGARCH64_HOST_SIGNAL_H -static inline uintptr_t host_signal_pc(ucontext_t *uc) +/* The third argument to a SA_SIGINFO handler is ucontext_t. */ +typedef ucontext_t host_sigcontext; + +static inline uintptr_t host_signal_pc(host_sigcontext *uc) { return uc->uc_mcontext.__pc; } -static inline void host_signal_set_pc(ucontext_t *uc, uintptr_t pc) +static inline void host_signal_set_pc(host_sigcontext *uc, uintptr_t pc) { uc->uc_mcontext.__pc = pc; } -static inline void *host_signal_mask(ucontext_t *uc) +static inline void *host_signal_mask(host_sigcontext *uc) { return &uc->uc_sigmask; } -static inline bool host_signal_write(siginfo_t *info, ucontext_t *uc) +static inline bool host_signal_write(siginfo_t *info, host_sigcontext *uc) { const uint32_t *pinsn = (const uint32_t *)host_signal_pc(uc); uint32_t insn = pinsn[0]; diff --git a/linux-user/include/host/mips/host-signal.h b/linux-user/include/host/mips/host-signal.h index XXXXXXX..XXXXXXX 100644 --- a/linux-user/include/host/mips/host-signal.h +++ b/linux-user/include/host/mips/host-signal.h @@ -XXX,XX +XXX,XX @@ #ifndef MIPS_HOST_SIGNAL_H #define MIPS_HOST_SIGNAL_H -static inline uintptr_t host_signal_pc(ucontext_t *uc) +/* The third argument to a SA_SIGINFO handler is ucontext_t. */ +typedef ucontext_t host_sigcontext; + +static inline uintptr_t host_signal_pc(host_sigcontext *uc) { return uc->uc_mcontext.pc; } -static inline void host_signal_set_pc(ucontext_t *uc, uintptr_t pc) +static inline void host_signal_set_pc(host_sigcontext *uc, uintptr_t pc) { uc->uc_mcontext.pc = pc; } -static inline void *host_signal_mask(ucontext_t *uc) +static inline void *host_signal_mask(host_sigcontext *uc) { return &uc->uc_sigmask; } @@ -XXX,XX +XXX,XX @@ static inline void *host_signal_mask(ucontext_t *uc) #error "Unsupported encoding" #endif -static inline bool host_signal_write(siginfo_t *info, ucontext_t *uc) +static inline bool host_signal_write(siginfo_t *info, host_sigcontext *uc) { uint32_t insn = *(uint32_t *)host_signal_pc(uc); diff --git a/linux-user/include/host/ppc/host-signal.h b/linux-user/include/host/ppc/host-signal.h index XXXXXXX..XXXXXXX 100644 --- a/linux-user/include/host/ppc/host-signal.h +++ b/linux-user/include/host/ppc/host-signal.h @@ -XXX,XX +XXX,XX @@ #ifndef PPC_HOST_SIGNAL_H #define PPC_HOST_SIGNAL_H -static inline uintptr_t host_signal_pc(ucontext_t *uc) +/* The third argument to a SA_SIGINFO handler is ucontext_t. */ +typedef ucontext_t host_sigcontext; + +static inline uintptr_t host_signal_pc(host_sigcontext *uc) { return uc->uc_mcontext.regs->nip; } -static inline void host_signal_set_pc(ucontext_t *uc, uintptr_t pc) +static inline void host_signal_set_pc(host_sigcontext *uc, uintptr_t pc) { uc->uc_mcontext.regs->nip = pc; } -static inline void *host_signal_mask(ucontext_t *uc) +static inline void *host_signal_mask(host_sigcontext *uc) { return &uc->uc_sigmask; } -static inline bool host_signal_write(siginfo_t *info, ucontext_t *uc) +static inline bool host_signal_write(siginfo_t *info, host_sigcontext *uc) { return uc->uc_mcontext.regs->trap != 0x400 && (uc->uc_mcontext.regs->dsisr & 0x02000000); diff --git a/linux-user/include/host/riscv/host-signal.h b/linux-user/include/host/riscv/host-signal.h index XXXXXXX..XXXXXXX 100644 --- a/linux-user/include/host/riscv/host-signal.h +++ b/linux-user/include/host/riscv/host-signal.h @@ -XXX,XX +XXX,XX @@ #ifndef RISCV_HOST_SIGNAL_H #define RISCV_HOST_SIGNAL_H -static inline uintptr_t host_signal_pc(ucontext_t *uc) +/* The third argument to a SA_SIGINFO handler is ucontext_t. */ +typedef ucontext_t host_sigcontext; + +static inline uintptr_t host_signal_pc(host_sigcontext *uc) { return uc->uc_mcontext.__gregs[REG_PC]; } -static inline void host_signal_set_pc(ucontext_t *uc, uintptr_t pc) +static inline void host_signal_set_pc(host_sigcontext *uc, uintptr_t pc) { uc->uc_mcontext.__gregs[REG_PC] = pc; } -static inline void *host_signal_mask(ucontext_t *uc) +static inline void *host_signal_mask(host_sigcontext *uc) { return &uc->uc_sigmask; } -static inline bool host_signal_write(siginfo_t *info, ucontext_t *uc) +static inline bool host_signal_write(siginfo_t *info, host_sigcontext *uc) { /* * Detect store by reading the instruction at the program counter. diff --git a/linux-user/include/host/s390/host-signal.h b/linux-user/include/host/s390/host-signal.h index XXXXXXX..XXXXXXX 100644 --- a/linux-user/include/host/s390/host-signal.h +++ b/linux-user/include/host/s390/host-signal.h @@ -XXX,XX +XXX,XX @@ #ifndef S390_HOST_SIGNAL_H #define S390_HOST_SIGNAL_H -static inline uintptr_t host_signal_pc(ucontext_t *uc) +/* The third argument to a SA_SIGINFO handler is ucontext_t. */ +typedef ucontext_t host_sigcontext; + +static inline uintptr_t host_signal_pc(host_sigcontext *uc) { return uc->uc_mcontext.psw.addr; } -static inline void host_signal_set_pc(ucontext_t *uc, uintptr_t pc) +static inline void host_signal_set_pc(host_sigcontext *uc, uintptr_t pc) { uc->uc_mcontext.psw.addr = pc; } -static inline void *host_signal_mask(ucontext_t *uc) +static inline void *host_signal_mask(host_sigcontext *uc) { return &uc->uc_sigmask; } -static inline bool host_signal_write(siginfo_t *info, ucontext_t *uc) +static inline bool host_signal_write(siginfo_t *info, host_sigcontext *uc) { uint16_t *pinsn = (uint16_t *)host_signal_pc(uc); diff --git a/linux-user/include/host/sparc/host-signal.h b/linux-user/include/host/sparc/host-signal.h index XXXXXXX..XXXXXXX 100644 --- a/linux-user/include/host/sparc/host-signal.h +++ b/linux-user/include/host/sparc/host-signal.h @@ -XXX,XX +XXX,XX @@ #ifndef SPARC_HOST_SIGNAL_H #define SPARC_HOST_SIGNAL_H -static inline uintptr_t host_signal_pc(ucontext_t *uc) +/* FIXME: the third argument to a SA_SIGINFO handler is *not* ucontext_t. */ +typedef ucontext_t host_sigcontext; + +static inline uintptr_t host_signal_pc(host_sigcontext *uc) { #ifdef __arch64__ return uc->uc_mcontext.mc_gregs[MC_PC]; @@ -XXX,XX +XXX,XX @@ static inline uintptr_t host_signal_pc(ucontext_t *uc) #endif } -static inline void host_signal_set_pc(ucontext_t *uc, uintptr_t pc) +static inline void host_signal_set_pc(host_sigcontext *uc, uintptr_t pc) { #ifdef __arch64__ uc->uc_mcontext.mc_gregs[MC_PC] = pc; @@ -XXX,XX +XXX,XX @@ static inline void host_signal_set_pc(ucontext_t *uc, uintptr_t pc) #endif } -static inline void *host_signal_mask(ucontext_t *uc) +static inline void *host_signal_mask(host_sigcontext *uc) { return &uc->uc_sigmask; } -static inline bool host_signal_write(siginfo_t *info, ucontext_t *uc) +static inline bool host_signal_write(siginfo_t *info, host_sigcontext *uc) { uint32_t insn = *(uint32_t *)host_signal_pc(uc); diff --git a/linux-user/include/host/x86_64/host-signal.h b/linux-user/include/host/x86_64/host-signal.h index XXXXXXX..XXXXXXX 100644 --- a/linux-user/include/host/x86_64/host-signal.h +++ b/linux-user/include/host/x86_64/host-signal.h @@ -XXX,XX +XXX,XX @@ #ifndef X86_64_HOST_SIGNAL_H #define X86_64_HOST_SIGNAL_H -static inline uintptr_t host_signal_pc(ucontext_t *uc) +/* The third argument to a SA_SIGINFO handler is ucontext_t. */ +typedef ucontext_t host_sigcontext; + +static inline uintptr_t host_signal_pc(host_sigcontext *uc) { return uc->uc_mcontext.gregs[REG_RIP]; } -static inline void host_signal_set_pc(ucontext_t *uc, uintptr_t pc) +static inline void host_signal_set_pc(host_sigcontext *uc, uintptr_t pc) { uc->uc_mcontext.gregs[REG_RIP] = pc; } -static inline void *host_signal_mask(ucontext_t *uc) +static inline void *host_signal_mask(host_sigcontext *uc) { return &uc->uc_sigmask; } -static inline bool host_signal_write(siginfo_t *info, ucontext_t *uc) +static inline bool host_signal_write(siginfo_t *info, host_sigcontext *uc) { return uc->uc_mcontext.gregs[REG_TRAPNO] == 0xe && (uc->uc_mcontext.gregs[REG_ERR] & 0x2); diff --git a/linux-user/signal.c b/linux-user/signal.c index XXXXXXX..XXXXXXX 100644 --- a/linux-user/signal.c +++ b/linux-user/signal.c @@ -XXX,XX +XXX,XX @@ void queue_signal(CPUArchState *env, int sig, int si_type, /* Adjust the signal context to rewind out of safe-syscall if we're in it */ static inline void rewind_if_in_safe_syscall(void *puc) { - ucontext_t *uc = (ucontext_t *)puc; + host_sigcontext *uc = (host_sigcontext *)puc; uintptr_t pcreg = host_signal_pc(uc); if (pcreg > (uintptr_t)safe_syscall_start @@ -XXX,XX +XXX,XX @@ static void host_signal_handler(int host_sig, siginfo_t *info, void *puc) CPUState *cpu = env_cpu(env); TaskState *ts = cpu->opaque; target_siginfo_t tinfo; - ucontext_t *uc = puc; + host_sigcontext *uc = puc; struct emulated_sigtable *k; int guest_sig; uintptr_t pc = 0; -- 2.25.1 We do not support sparc32 as a host, so there's no point in sparc64 redirecting to sparc. Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- linux-user/include/host/sparc/host-signal.h | 71 ------------------- linux-user/include/host/sparc64/host-signal.h | 64 ++++++++++++++++- 2 files changed, 63 insertions(+), 72 deletions(-) delete mode 100644 linux-user/include/host/sparc/host-signal.h diff --git a/linux-user/include/host/sparc/host-signal.h b/linux-user/include/host/sparc/host-signal.h deleted file mode 100644 index XXXXXXX..XXXXXXX --- a/linux-user/include/host/sparc/host-signal.h +++ /dev/null @@ -XXX,XX +XXX,XX @@ -/* - * host-signal.h: signal info dependent on the host architecture - * - * Copyright (c) 2003-2005 Fabrice Bellard - * Copyright (c) 2021 Linaro Limited - * - * This work is licensed under the terms of the GNU LGPL, version 2.1 or later. - * See the COPYING file in the top-level directory. - */ - -#ifndef SPARC_HOST_SIGNAL_H -#define SPARC_HOST_SIGNAL_H - -/* FIXME: the third argument to a SA_SIGINFO handler is *not* ucontext_t. */ -typedef ucontext_t host_sigcontext; - -static inline uintptr_t host_signal_pc(host_sigcontext *uc) -{ -#ifdef __arch64__ - return uc->uc_mcontext.mc_gregs[MC_PC]; -#else - return uc->uc_mcontext.gregs[REG_PC]; -#endif -} - -static inline void host_signal_set_pc(host_sigcontext *uc, uintptr_t pc) -{ -#ifdef __arch64__ - uc->uc_mcontext.mc_gregs[MC_PC] = pc; -#else - uc->uc_mcontext.gregs[REG_PC] = pc; -#endif -} - -static inline void *host_signal_mask(host_sigcontext *uc) -{ - return &uc->uc_sigmask; -} - -static inline bool host_signal_write(siginfo_t *info, host_sigcontext *uc) -{ - uint32_t insn = *(uint32_t *)host_signal_pc(uc); - - if ((insn >> 30) == 3) { - switch ((insn >> 19) & 0x3f) { - case 0x05: /* stb */ - case 0x15: /* stba */ - case 0x06: /* sth */ - case 0x16: /* stha */ - case 0x04: /* st */ - case 0x14: /* sta */ - case 0x07: /* std */ - case 0x17: /* stda */ - case 0x0e: /* stx */ - case 0x1e: /* stxa */ - case 0x24: /* stf */ - case 0x34: /* stfa */ - case 0x27: /* stdf */ - case 0x37: /* stdfa */ - case 0x26: /* stqf */ - case 0x36: /* stqfa */ - case 0x25: /* stfsr */ - case 0x3c: /* casa */ - case 0x3e: /* casxa */ - return true; - } - } - return false; -} - -#endif diff --git a/linux-user/include/host/sparc64/host-signal.h b/linux-user/include/host/sparc64/host-signal.h index XXXXXXX..XXXXXXX 100644 --- a/linux-user/include/host/sparc64/host-signal.h +++ b/linux-user/include/host/sparc64/host-signal.h @@ -1 +1,63 @@ -#include "../sparc/host-signal.h" +/* + * host-signal.h: signal info dependent on the host architecture + * + * Copyright (c) 2003-2005 Fabrice Bellard + * Copyright (c) 2021 Linaro Limited + * + * This work is licensed under the terms of the GNU LGPL, version 2.1 or later. + * See the COPYING file in the top-level directory. + */ + +#ifndef SPARC64_HOST_SIGNAL_H +#define SPARC64_HOST_SIGNAL_H + +/* FIXME: the third argument to a SA_SIGINFO handler is *not* ucontext_t. */ +typedef ucontext_t host_sigcontext; + +static inline uintptr_t host_signal_pc(host_sigcontext *uc) +{ + return uc->uc_mcontext.mc_gregs[MC_PC]; +} + +static inline void host_signal_set_pc(host_sigcontext *uc, uintptr_t pc) +{ + uc->uc_mcontext.mc_gregs[MC_PC] = pc; +} + +static inline void *host_signal_mask(host_sigcontext *uc) +{ + return &uc->uc_sigmask; +} + +static inline bool host_signal_write(siginfo_t *info, host_sigcontext *uc) +{ + uint32_t insn = *(uint32_t *)host_signal_pc(uc); + + if ((insn >> 30) == 3) { + switch ((insn >> 19) & 0x3f) { + case 0x05: /* stb */ + case 0x15: /* stba */ + case 0x06: /* sth */ + case 0x16: /* stha */ + case 0x04: /* st */ + case 0x14: /* sta */ + case 0x07: /* std */ + case 0x17: /* stda */ + case 0x0e: /* stx */ + case 0x1e: /* stxa */ + case 0x24: /* stf */ + case 0x34: /* stfa */ + case 0x27: /* stdf */ + case 0x37: /* stdfa */ + case 0x26: /* stqf */ + case 0x36: /* stqfa */ + case 0x25: /* stfsr */ + case 0x3c: /* casa */ + case 0x3e: /* casxa */ + return true; + } + } + return false; +} + +#endif -- 2.25.1 Sparc64 is unique on linux in *not* passing ucontext_t as the third argument to a SA_SIGINFO handler. It passes the old struct sigcontext instead. Set both pc and npc in host_signal_set_pc. Fixes: 8b5bd461935b ("linux-user/host/sparc: Populate host_signal.h") Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- linux-user/include/host/sparc64/host-signal.h | 17 +++++++++-------- 1 file changed, 9 insertions(+), 8 deletions(-) diff --git a/linux-user/include/host/sparc64/host-signal.h b/linux-user/include/host/sparc64/host-signal.h index XXXXXXX..XXXXXXX 100644 --- a/linux-user/include/host/sparc64/host-signal.h +++ b/linux-user/include/host/sparc64/host-signal.h @@ -XXX,XX +XXX,XX @@ #ifndef SPARC64_HOST_SIGNAL_H #define SPARC64_HOST_SIGNAL_H -/* FIXME: the third argument to a SA_SIGINFO handler is *not* ucontext_t. */ -typedef ucontext_t host_sigcontext; +/* The third argument to a SA_SIGINFO handler is struct sigcontext. */ +typedef struct sigcontext host_sigcontext; -static inline uintptr_t host_signal_pc(host_sigcontext *uc) +static inline uintptr_t host_signal_pc(host_sigcontext *sc) { - return uc->uc_mcontext.mc_gregs[MC_PC]; + return sc->sigc_regs.tpc; } -static inline void host_signal_set_pc(host_sigcontext *uc, uintptr_t pc) +static inline void host_signal_set_pc(host_sigcontext *sc, uintptr_t pc) { - uc->uc_mcontext.mc_gregs[MC_PC] = pc; + sc->sigc_regs.tpc = pc; + sc->sigc_regs.tnpc = pc + 4; } -static inline void *host_signal_mask(host_sigcontext *uc) +static inline void *host_signal_mask(host_sigcontext *sc) { - return &uc->uc_sigmask; + return &sc->sigc_mask; } static inline bool host_signal_write(siginfo_t *info, host_sigcontext *uc) -- 2.25.1 From: Idan Horowitz <idan.horowitz@gmail.com> When the length of the range is large enough, clearing the whole cache is faster than iterating over the (possibly extremely large) set of pages contained in the range. This mimics the pre-existing similar optimization done on the flush of the tlb itself. Signed-off-by: Idan Horowitz <idan.horowitz@gmail.com> Message-Id: <20220110164754.1066025-1-idan.horowitz@gmail.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- accel/tcg/cputlb.c | 9 +++++++++ 1 file changed, 9 insertions(+) diff --git a/accel/tcg/cputlb.c b/accel/tcg/cputlb.c index XXXXXXX..XXXXXXX 100644 --- a/accel/tcg/cputlb.c +++ b/accel/tcg/cputlb.c @@ -XXX,XX +XXX,XX @@ static void tlb_flush_range_by_mmuidx_async_0(CPUState *cpu, } qemu_spin_unlock(&env_tlb(env)->c.lock); + /* + * If the length is larger than the jump cache size, then it will take + * longer to clear each entry individually than it will to clear it all. + */ + if (d.len >= (TARGET_PAGE_SIZE * TB_JMP_CACHE_SIZE)) { + cpu_tb_jmp_cache_clear(cpu); + return; + } + for (target_ulong i = 0; i < d.len; i += TARGET_PAGE_SIZE) { tb_flush_jmp_cache(cpu, d.addr + i); } -- 2.25.1 From: Idan Horowitz <idan.horowitz@gmail.com> Instead of taking the lock of the cpu work list in order to check if it's empty, we can just read the head pointer atomically. This decreases cpu_work_list_empty's share from 5% to 1.3% in a profile of icount-enabled aarch64-softmmu. Signed-off-by: Idan Horowitz <idan.horowitz@gmail.com> Message-Id: <20220114004358.299534-1-idan.horowitz@gmail.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- softmmu/cpus.c | 7 +------ 1 file changed, 1 insertion(+), 6 deletions(-) diff --git a/softmmu/cpus.c b/softmmu/cpus.c index XXXXXXX..XXXXXXX 100644 --- a/softmmu/cpus.c +++ b/softmmu/cpus.c @@ -XXX,XX +XXX,XX @@ bool cpu_is_stopped(CPUState *cpu) bool cpu_work_list_empty(CPUState *cpu) { - bool ret; - - qemu_mutex_lock(&cpu->work_mutex); - ret = QSIMPLEQ_EMPTY(&cpu->work_list); - qemu_mutex_unlock(&cpu->work_mutex); - return ret; + return QSIMPLEQ_EMPTY_ATOMIC(&cpu->work_list); } bool cpu_thread_is_idle(CPUState *cpu) -- 2.25.1 From: Pavel Dovgalyuk <pavel.dovgalyuk@ispras.ru> Commit aff0e204cb1f1c036a496c94c15f5dfafcd9b4b4 introduced CF_NOIRQ usage, but one case was forgotten. Record/replay uses one special TB which is not really executed, but used to cause a correct exception in replay mode. This patch adds CF_NOIRQ flag for such block. Signed-off-by: Pavel Dovgalyuk <Pavel.Dovgalyuk@ispras.ru> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-Id: <164362834054.1754532.7678416881159817273.stgit@pasha-ThinkPad-X280> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- accel/tcg/cpu-exec.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) 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 @@ static inline bool cpu_handle_exception(CPUState *cpu, int *ret) if (replay_has_exception() && cpu_neg(cpu)->icount_decr.u16.low + cpu->icount_extra == 0) { /* Execute just one insn to trigger exception pending in the log */ - cpu->cflags_next_tb = (curr_cflags(cpu) & ~CF_USE_ICOUNT) | 1; + cpu->cflags_next_tb = (curr_cflags(cpu) & ~CF_USE_ICOUNT) + | CF_NOIRQ | 1; } #endif return false; -- 2.25.1 From: WANG Xuerui <git@xen0n.name> Apparently we were left behind; just renaming MO_Q to MO_UQ is enough. Fixes: fc313c64345453c7 ("exec/memop: Adding signedness to quad definitions") Signed-off-by: WANG Xuerui <git@xen0n.name> Message-Id: <20220206162106.1092364-1-i.qemu@xen0n.name> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/loongarch64/tcg-target.c.inc | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) 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 void tcg_out_qemu_ld_indexed(TCGContext *s, TCGReg rd, TCGReg rj, case MO_SL: tcg_out_opc_ldx_w(s, rd, rj, rk); break; - case MO_Q: + case MO_UQ: tcg_out_opc_ldx_d(s, rd, rj, rk); break; default: -- 2.25.1 Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/i386/tcg-target.h | 2 - tcg/i386/tcg-target.c.inc | 103 ++++++++++++++++++++++++++++++++++++-- 2 files changed, 98 insertions(+), 7 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 @@ static inline void tb_target_set_jmp_target(uintptr_t tc_ptr, uintptr_t jmp_rx, #define TCG_TARGET_HAS_MEMORY_BSWAP have_movbe -#ifdef CONFIG_SOFTMMU #define TCG_TARGET_NEED_LDST_LABELS -#endif #define TCG_TARGET_NEED_POOL_LABELS #endif 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 @@ * THE SOFTWARE. */ +#include "../tcg-ldst.c.inc" #include "../tcg-pool.c.inc" #ifdef CONFIG_DEBUG_TCG @@ -XXX,XX +XXX,XX @@ static bool tcg_target_const_match(int64_t val, TCGType type, int ct) #define OPC_VZEROUPPER (0x77 | P_EXT) #define OPC_XCHG_ax_r32 (0x90) -#define OPC_GRP3_Ev (0xf7) -#define OPC_GRP5 (0xff) +#define OPC_GRP3_Eb (0xf6) +#define OPC_GRP3_Ev (0xf7) +#define OPC_GRP5 (0xff) #define OPC_GRP14 (0x73 | P_EXT | P_DATA16) /* Group 1 opcode extensions for 0x80-0x83. @@ -XXX,XX +XXX,XX @@ static bool tcg_target_const_match(int64_t val, TCGType type, int ct) #define SHIFT_SAR 7 /* Group 3 opcode extensions for 0xf6, 0xf7. To be used with OPC_GRP3. */ +#define EXT3_TESTi 0 #define EXT3_NOT 2 #define EXT3_NEG 3 #define EXT3_MUL 4 @@ -XXX,XX +XXX,XX @@ static void tcg_out_nopn(TCGContext *s, int n) } #if defined(CONFIG_SOFTMMU) -#include "../tcg-ldst.c.inc" - /* helper signature: helper_ret_ld_mmu(CPUState *env, target_ulong addr, * int mmu_idx, uintptr_t ra) */ @@ -XXX,XX +XXX,XX @@ static bool tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *l) tcg_out_jmp(s, qemu_st_helpers[opc & (MO_BSWAP | MO_SIZE)]); return true; } -#elif TCG_TARGET_REG_BITS == 32 +#else + +static void tcg_out_test_alignment(TCGContext *s, bool is_ld, TCGReg addrlo, + TCGReg addrhi, unsigned a_bits) +{ + unsigned a_mask = (1 << a_bits) - 1; + TCGLabelQemuLdst *label; + + /* + * We are expecting a_bits to max out at 7, so we can usually use testb. + * For i686, we have to use testl for %esi/%edi. + */ + if (a_mask <= 0xff && (TCG_TARGET_REG_BITS == 64 || addrlo < 4)) { + tcg_out_modrm(s, OPC_GRP3_Eb | P_REXB_RM, EXT3_TESTi, addrlo); + tcg_out8(s, a_mask); + } else { + tcg_out_modrm(s, OPC_GRP3_Ev, EXT3_TESTi, addrlo); + tcg_out32(s, a_mask); + } + + /* jne slow_path */ + tcg_out_opc(s, OPC_JCC_long + JCC_JNE, 0, 0, 0); + + label = new_ldst_label(s); + label->is_ld = is_ld; + label->addrlo_reg = addrlo; + label->addrhi_reg = addrhi; + label->raddr = tcg_splitwx_to_rx(s->code_ptr + 4); + label->label_ptr[0] = s->code_ptr; + + s->code_ptr += 4; +} + +static bool tcg_out_fail_alignment(TCGContext *s, TCGLabelQemuLdst *l) +{ + /* resolve label address */ + tcg_patch32(l->label_ptr[0], s->code_ptr - l->label_ptr[0] - 4); + + if (TCG_TARGET_REG_BITS == 32) { + int ofs = 0; + + tcg_out_st(s, TCG_TYPE_PTR, TCG_AREG0, TCG_REG_ESP, ofs); + ofs += 4; + + tcg_out_st(s, TCG_TYPE_I32, l->addrlo_reg, TCG_REG_ESP, ofs); + ofs += 4; + if (TARGET_LONG_BITS == 64) { + tcg_out_st(s, TCG_TYPE_I32, l->addrhi_reg, TCG_REG_ESP, ofs); + ofs += 4; + } + + tcg_out_pushi(s, (uintptr_t)l->raddr); + } else { + tcg_out_mov(s, TCG_TYPE_TL, tcg_target_call_iarg_regs[1], + l->addrlo_reg); + tcg_out_mov(s, TCG_TYPE_PTR, tcg_target_call_iarg_regs[0], TCG_AREG0); + + tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_RAX, (uintptr_t)l->raddr); + tcg_out_push(s, TCG_REG_RAX); + } + + /* "Tail call" to the helper, with the return address back inline. */ + tcg_out_jmp(s, (const void *)(l->is_ld ? helper_unaligned_ld + : helper_unaligned_st)); + return true; +} + +static bool tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *l) +{ + return tcg_out_fail_alignment(s, l); +} + +static bool tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *l) +{ + return tcg_out_fail_alignment(s, l); +} + +#if TCG_TARGET_REG_BITS == 32 # define x86_guest_base_seg 0 # define x86_guest_base_index -1 # define x86_guest_base_offset guest_base @@ -XXX,XX +XXX,XX @@ static inline int setup_guest_base_seg(void) return 0; } # endif +#endif #endif /* SOFTMMU */ static void tcg_out_qemu_ld_direct(TCGContext *s, TCGReg datalo, TCGReg datahi, @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, bool is64) #if defined(CONFIG_SOFTMMU) int mem_index; tcg_insn_unit *label_ptr[2]; +#else + unsigned a_bits; #endif datalo = *args++; @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, bool is64) add_qemu_ldst_label(s, true, is64, oi, datalo, datahi, addrlo, addrhi, s->code_ptr, label_ptr); #else + a_bits = get_alignment_bits(opc); + if (a_bits) { + tcg_out_test_alignment(s, true, addrlo, addrhi, a_bits); + } + tcg_out_qemu_ld_direct(s, datalo, datahi, addrlo, x86_guest_base_index, x86_guest_base_offset, x86_guest_base_seg, is64, opc); @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, bool is64) #if defined(CONFIG_SOFTMMU) int mem_index; tcg_insn_unit *label_ptr[2]; +#else + unsigned a_bits; #endif datalo = *args++; @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, bool is64) add_qemu_ldst_label(s, false, is64, oi, datalo, datahi, addrlo, addrhi, s->code_ptr, label_ptr); #else + a_bits = get_alignment_bits(opc); + if (a_bits) { + tcg_out_test_alignment(s, false, addrlo, addrhi, a_bits); + } + tcg_out_qemu_st_direct(s, datalo, datahi, addrlo, x86_guest_base_index, x86_guest_base_offset, x86_guest_base_seg, opc); #endif -- 2.25.1 Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/aarch64/tcg-target.h | 2 - tcg/aarch64/tcg-target.c.inc | 91 +++++++++++++++++++++++++++++------- 2 files changed, 74 insertions(+), 19 deletions(-) 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 { void tb_target_set_jmp_target(uintptr_t, uintptr_t, uintptr_t, uintptr_t); -#ifdef CONFIG_SOFTMMU #define TCG_TARGET_NEED_LDST_LABELS -#endif #define TCG_TARGET_NEED_POOL_LABELS #endif /* AARCH64_TCG_TARGET_H */ 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 @@ * See the COPYING file in the top-level directory for details. */ +#include "../tcg-ldst.c.inc" #include "../tcg-pool.c.inc" #include "qemu/bitops.h" @@ -XXX,XX +XXX,XX @@ typedef enum { I3404_ANDI = 0x12000000, I3404_ORRI = 0x32000000, I3404_EORI = 0x52000000, + I3404_ANDSI = 0x72000000, /* Move wide immediate instructions. */ I3405_MOVN = 0x12800000, @@ -XXX,XX +XXX,XX @@ static void tcg_out_goto_long(TCGContext *s, const tcg_insn_unit *target) if (offset == sextract64(offset, 0, 26)) { tcg_out_insn(s, 3206, B, offset); } else { - tcg_out_movi(s, TCG_TYPE_I64, TCG_REG_TMP, (intptr_t)target); - tcg_out_insn(s, 3207, BR, TCG_REG_TMP); + /* Choose X9 as a call-clobbered non-LR temporary. */ + tcg_out_movi(s, TCG_TYPE_I64, TCG_REG_X9, (intptr_t)target); + tcg_out_insn(s, 3207, BR, TCG_REG_X9); } } @@ -XXX,XX +XXX,XX @@ static void tcg_out_cltz(TCGContext *s, TCGType ext, TCGReg d, } } -#ifdef CONFIG_SOFTMMU -#include "../tcg-ldst.c.inc" +static void tcg_out_adr(TCGContext *s, TCGReg rd, const void *target) +{ + ptrdiff_t offset = tcg_pcrel_diff(s, target); + tcg_debug_assert(offset == sextract64(offset, 0, 21)); + tcg_out_insn(s, 3406, ADR, rd, offset); +} +#ifdef CONFIG_SOFTMMU /* helper signature: helper_ret_ld_mmu(CPUState *env, target_ulong addr, * MemOpIdx oi, uintptr_t ra) */ @@ -XXX,XX +XXX,XX @@ static void * const qemu_st_helpers[MO_SIZE + 1] = { #endif }; -static inline void tcg_out_adr(TCGContext *s, TCGReg rd, const void *target) -{ - ptrdiff_t offset = tcg_pcrel_diff(s, target); - tcg_debug_assert(offset == sextract64(offset, 0, 21)); - tcg_out_insn(s, 3406, ADR, rd, offset); -} - static bool tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *lb) { MemOpIdx oi = lb->oi; @@ -XXX,XX +XXX,XX @@ static void tcg_out_tlb_read(TCGContext *s, TCGReg addr_reg, MemOp opc, tcg_out_insn(s, 3202, B_C, TCG_COND_NE, 0); } +#else +static void tcg_out_test_alignment(TCGContext *s, bool is_ld, TCGReg addr_reg, + unsigned a_bits) +{ + unsigned a_mask = (1 << a_bits) - 1; + TCGLabelQemuLdst *label = new_ldst_label(s); + + label->is_ld = is_ld; + label->addrlo_reg = addr_reg; + + /* tst addr, #mask */ + tcg_out_logicali(s, I3404_ANDSI, 0, TCG_REG_XZR, addr_reg, a_mask); + + label->label_ptr[0] = s->code_ptr; + + /* b.ne slow_path */ + tcg_out_insn(s, 3202, B_C, TCG_COND_NE, 0); + + label->raddr = tcg_splitwx_to_rx(s->code_ptr); +} + +static bool tcg_out_fail_alignment(TCGContext *s, TCGLabelQemuLdst *l) +{ + if (!reloc_pc19(l->label_ptr[0], tcg_splitwx_to_rx(s->code_ptr))) { + return false; + } + + tcg_out_mov(s, TCG_TYPE_TL, TCG_REG_X1, l->addrlo_reg); + tcg_out_mov(s, TCG_TYPE_PTR, TCG_REG_X0, TCG_AREG0); + + /* "Tail call" to the helper, with the return address back inline. */ + tcg_out_adr(s, TCG_REG_LR, l->raddr); + tcg_out_goto_long(s, (const void *)(l->is_ld ? helper_unaligned_ld + : helper_unaligned_st)); + return true; +} + +static bool tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *l) +{ + return tcg_out_fail_alignment(s, l); +} + +static bool tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *l) +{ + return tcg_out_fail_alignment(s, l); +} #endif /* CONFIG_SOFTMMU */ static void tcg_out_qemu_ld_direct(TCGContext *s, MemOp memop, TCGType ext, TCGReg data_r, TCGReg addr_r, TCGType otype, TCGReg off_r) { - /* Byte swapping is left to middle-end expansion. */ - tcg_debug_assert((memop & MO_BSWAP) == 0); - switch (memop & MO_SSIZE) { case MO_UB: tcg_out_ldst_r(s, I3312_LDRB, data_r, addr_r, otype, off_r); @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_st_direct(TCGContext *s, MemOp memop, TCGReg data_r, TCGReg addr_r, TCGType otype, TCGReg off_r) { - /* Byte swapping is left to middle-end expansion. */ - tcg_debug_assert((memop & MO_BSWAP) == 0); - switch (memop & MO_SIZE) { case MO_8: tcg_out_ldst_r(s, I3312_STRB, data_r, addr_r, otype, off_r); @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_ld(TCGContext *s, TCGReg data_reg, TCGReg addr_reg, { MemOp memop = get_memop(oi); const TCGType otype = TARGET_LONG_BITS == 64 ? TCG_TYPE_I64 : TCG_TYPE_I32; + + /* Byte swapping is left to middle-end expansion. */ + tcg_debug_assert((memop & MO_BSWAP) == 0); + #ifdef CONFIG_SOFTMMU unsigned mem_index = get_mmuidx(oi); tcg_insn_unit *label_ptr; @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_ld(TCGContext *s, TCGReg data_reg, TCGReg addr_reg, add_qemu_ldst_label(s, true, oi, ext, data_reg, addr_reg, s->code_ptr, label_ptr); #else /* !CONFIG_SOFTMMU */ + unsigned a_bits = get_alignment_bits(memop); + if (a_bits) { + tcg_out_test_alignment(s, true, addr_reg, a_bits); + } if (USE_GUEST_BASE) { tcg_out_qemu_ld_direct(s, memop, ext, data_reg, TCG_REG_GUEST_BASE, otype, addr_reg); @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_st(TCGContext *s, TCGReg data_reg, TCGReg addr_reg, { MemOp memop = get_memop(oi); const TCGType otype = TARGET_LONG_BITS == 64 ? TCG_TYPE_I64 : TCG_TYPE_I32; + + /* Byte swapping is left to middle-end expansion. */ + tcg_debug_assert((memop & MO_BSWAP) == 0); + #ifdef CONFIG_SOFTMMU unsigned mem_index = get_mmuidx(oi); tcg_insn_unit *label_ptr; @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_st(TCGContext *s, TCGReg data_reg, TCGReg addr_reg, add_qemu_ldst_label(s, false, oi, (memop & MO_SIZE)== MO_64, data_reg, addr_reg, s->code_ptr, label_ptr); #else /* !CONFIG_SOFTMMU */ + unsigned a_bits = get_alignment_bits(memop); + if (a_bits) { + tcg_out_test_alignment(s, false, addr_reg, a_bits); + } if (USE_GUEST_BASE) { tcg_out_qemu_st_direct(s, memop, data_reg, TCG_REG_GUEST_BASE, otype, addr_reg); -- 2.25.1 Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/ppc/tcg-target.h | 2 - tcg/ppc/tcg-target.c.inc | 98 ++++++++++++++++++++++++++++++++++++---- 2 files changed, 90 insertions(+), 10 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 @@ void tb_target_set_jmp_target(uintptr_t, uintptr_t, uintptr_t, uintptr_t); #define TCG_TARGET_DEFAULT_MO (0) #define TCG_TARGET_HAS_MEMORY_BSWAP 1 -#ifdef CONFIG_SOFTMMU #define TCG_TARGET_NEED_LDST_LABELS -#endif #define TCG_TARGET_NEED_POOL_LABELS #endif 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 @@ #include "elf.h" #include "../tcg-pool.c.inc" +#include "../tcg-ldst.c.inc" /* * Standardize on the _CALL_FOO symbols used by GCC: @@ -XXX,XX +XXX,XX @@ void tb_target_set_jmp_target(uintptr_t tc_ptr, uintptr_t jmp_rx, } } -static void tcg_out_call(TCGContext *s, const tcg_insn_unit *target) +static void tcg_out_call_int(TCGContext *s, int lk, + const tcg_insn_unit *target) { #ifdef _CALL_AIX /* Look through the descriptor. If the branch is in range, and we @@ -XXX,XX +XXX,XX @@ static void tcg_out_call(TCGContext *s, const tcg_insn_unit *target) if (in_range_b(diff) && toc == (uint32_t)toc) { tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_TMP1, toc); - tcg_out_b(s, LK, tgt); + tcg_out_b(s, lk, tgt); } else { /* Fold the low bits of the constant into the addresses below. */ intptr_t arg = (intptr_t)target; @@ -XXX,XX +XXX,XX @@ static void tcg_out_call(TCGContext *s, const tcg_insn_unit *target) tcg_out_ld(s, TCG_TYPE_PTR, TCG_REG_R0, TCG_REG_TMP1, ofs); tcg_out32(s, MTSPR | RA(TCG_REG_R0) | CTR); tcg_out_ld(s, TCG_TYPE_PTR, TCG_REG_R2, TCG_REG_TMP1, ofs + SZP); - tcg_out32(s, BCCTR | BO_ALWAYS | LK); + tcg_out32(s, BCCTR | BO_ALWAYS | lk); } #elif defined(_CALL_ELF) && _CALL_ELF == 2 intptr_t diff; @@ -XXX,XX +XXX,XX @@ static void tcg_out_call(TCGContext *s, const tcg_insn_unit *target) diff = tcg_pcrel_diff(s, target); if (in_range_b(diff)) { - tcg_out_b(s, LK, target); + tcg_out_b(s, lk, target); } else { tcg_out32(s, MTSPR | RS(TCG_REG_R12) | CTR); - tcg_out32(s, BCCTR | BO_ALWAYS | LK); + tcg_out32(s, BCCTR | BO_ALWAYS | lk); } #else - tcg_out_b(s, LK, target); + tcg_out_b(s, lk, target); #endif } +static void tcg_out_call(TCGContext *s, const tcg_insn_unit *target) +{ + tcg_out_call_int(s, LK, target); +} + static const uint32_t qemu_ldx_opc[(MO_SSIZE + MO_BSWAP) + 1] = { [MO_UB] = LBZX, [MO_UW] = LHZX, @@ -XXX,XX +XXX,XX @@ static const uint32_t qemu_exts_opc[4] = { }; #if defined (CONFIG_SOFTMMU) -#include "../tcg-ldst.c.inc" - /* helper signature: helper_ld_mmu(CPUState *env, target_ulong addr, * int mmu_idx, uintptr_t ra) */ @@ -XXX,XX +XXX,XX @@ static bool tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *lb) tcg_out_b(s, 0, lb->raddr); return true; } +#else + +static void tcg_out_test_alignment(TCGContext *s, bool is_ld, TCGReg addrlo, + TCGReg addrhi, unsigned a_bits) +{ + unsigned a_mask = (1 << a_bits) - 1; + TCGLabelQemuLdst *label = new_ldst_label(s); + + label->is_ld = is_ld; + label->addrlo_reg = addrlo; + label->addrhi_reg = addrhi; + + /* We are expecting a_bits to max out at 7, much lower than ANDI. */ + tcg_debug_assert(a_bits < 16); + tcg_out32(s, ANDI | SAI(addrlo, TCG_REG_R0, a_mask)); + + label->label_ptr[0] = s->code_ptr; + tcg_out32(s, BC | BI(0, CR_EQ) | BO_COND_FALSE | LK); + + label->raddr = tcg_splitwx_to_rx(s->code_ptr); +} + +static bool tcg_out_fail_alignment(TCGContext *s, TCGLabelQemuLdst *l) +{ + if (!reloc_pc14(l->label_ptr[0], tcg_splitwx_to_rx(s->code_ptr))) { + return false; + } + + if (TCG_TARGET_REG_BITS < TARGET_LONG_BITS) { + TCGReg arg = TCG_REG_R4; +#ifdef TCG_TARGET_CALL_ALIGN_ARGS + arg |= 1; +#endif + if (l->addrlo_reg != arg) { + tcg_out_mov(s, TCG_TYPE_I32, arg, l->addrhi_reg); + tcg_out_mov(s, TCG_TYPE_I32, arg + 1, l->addrlo_reg); + } else if (l->addrhi_reg != arg + 1) { + tcg_out_mov(s, TCG_TYPE_I32, arg + 1, l->addrlo_reg); + tcg_out_mov(s, TCG_TYPE_I32, arg, l->addrhi_reg); + } else { + tcg_out_mov(s, TCG_TYPE_I32, TCG_REG_R0, arg); + tcg_out_mov(s, TCG_TYPE_I32, arg, arg + 1); + tcg_out_mov(s, TCG_TYPE_I32, arg + 1, TCG_REG_R0); + } + } else { + tcg_out_mov(s, TCG_TYPE_TL, TCG_REG_R4, l->addrlo_reg); + } + tcg_out_mov(s, TCG_TYPE_TL, TCG_REG_R3, TCG_AREG0); + + /* "Tail call" to the helper, with the return address back inline. */ + tcg_out_call_int(s, 0, (const void *)(l->is_ld ? helper_unaligned_ld + : helper_unaligned_st)); + return true; +} + +static bool tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *l) +{ + return tcg_out_fail_alignment(s, l); +} + +static bool tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *l) +{ + return tcg_out_fail_alignment(s, l); +} + #endif /* SOFTMMU */ static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, bool is_64) @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, bool is_64) #ifdef CONFIG_SOFTMMU int mem_index; tcg_insn_unit *label_ptr; +#else + unsigned a_bits; #endif datalo = *args++; @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, bool is_64) rbase = TCG_REG_R3; #else /* !CONFIG_SOFTMMU */ + a_bits = get_alignment_bits(opc); + if (a_bits) { + tcg_out_test_alignment(s, true, addrlo, addrhi, a_bits); + } rbase = guest_base ? TCG_GUEST_BASE_REG : 0; if (TCG_TARGET_REG_BITS > TARGET_LONG_BITS) { tcg_out_ext32u(s, TCG_REG_TMP1, addrlo); @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, bool is_64) #ifdef CONFIG_SOFTMMU int mem_index; tcg_insn_unit *label_ptr; +#else + unsigned a_bits; #endif datalo = *args++; @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, bool is_64) rbase = TCG_REG_R3; #else /* !CONFIG_SOFTMMU */ + a_bits = get_alignment_bits(opc); + if (a_bits) { + tcg_out_test_alignment(s, false, addrlo, addrhi, a_bits); + } rbase = guest_base ? TCG_GUEST_BASE_REG : 0; if (TCG_TARGET_REG_BITS > TARGET_LONG_BITS) { tcg_out_ext32u(s, TCG_REG_TMP1, addrlo); -- 2.25.1 Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/riscv/tcg-target.h | 2 -- tcg/riscv/tcg-target.c.inc | 63 ++++++++++++++++++++++++++++++++++++-- 2 files changed, 61 insertions(+), 4 deletions(-) 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 @@ void tb_target_set_jmp_target(uintptr_t, uintptr_t, uintptr_t, uintptr_t); #define TCG_TARGET_DEFAULT_MO (0) -#ifdef CONFIG_SOFTMMU #define TCG_TARGET_NEED_LDST_LABELS -#endif #define TCG_TARGET_NEED_POOL_LABELS #define TCG_TARGET_HAS_MEMORY_BSWAP 0 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 @@ * THE SOFTWARE. */ +#include "../tcg-ldst.c.inc" #include "../tcg-pool.c.inc" #ifdef CONFIG_DEBUG_TCG @@ -XXX,XX +XXX,XX @@ static void tcg_out_mb(TCGContext *s, TCGArg a0) */ #if defined(CONFIG_SOFTMMU) -#include "../tcg-ldst.c.inc" - /* helper signature: helper_ret_ld_mmu(CPUState *env, target_ulong addr, * MemOpIdx oi, uintptr_t ra) */ @@ -XXX,XX +XXX,XX @@ static bool tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *l) tcg_out_goto(s, l->raddr); return true; } +#else + +static void tcg_out_test_alignment(TCGContext *s, bool is_ld, TCGReg addr_reg, + unsigned a_bits) +{ + unsigned a_mask = (1 << a_bits) - 1; + TCGLabelQemuLdst *l = new_ldst_label(s); + + l->is_ld = is_ld; + l->addrlo_reg = addr_reg; + + /* We are expecting a_bits to max out at 7, so we can always use andi. */ + tcg_debug_assert(a_bits < 12); + tcg_out_opc_imm(s, OPC_ANDI, TCG_REG_TMP1, addr_reg, a_mask); + + l->label_ptr[0] = s->code_ptr; + tcg_out_opc_branch(s, OPC_BNE, TCG_REG_TMP1, TCG_REG_ZERO, 0); + + l->raddr = tcg_splitwx_to_rx(s->code_ptr); +} + +static bool tcg_out_fail_alignment(TCGContext *s, TCGLabelQemuLdst *l) +{ + /* resolve label address */ + if (!reloc_sbimm12(l->label_ptr[0], tcg_splitwx_to_rx(s->code_ptr))) { + return false; + } + + tcg_out_mov(s, TCG_TYPE_TL, TCG_REG_A1, l->addrlo_reg); + tcg_out_mov(s, TCG_TYPE_PTR, TCG_REG_A0, TCG_AREG0); + + /* tail call, with the return address back inline. */ + tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_RA, (uintptr_t)l->raddr); + tcg_out_call_int(s, (const void *)(l->is_ld ? helper_unaligned_ld + : helper_unaligned_st), true); + return true; +} + +static bool tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *l) +{ + return tcg_out_fail_alignment(s, l); +} + +static bool tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *l) +{ + return tcg_out_fail_alignment(s, l); +} + #endif /* CONFIG_SOFTMMU */ static void tcg_out_qemu_ld_direct(TCGContext *s, TCGReg lo, TCGReg hi, @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, bool is_64) MemOp opc; #if defined(CONFIG_SOFTMMU) tcg_insn_unit *label_ptr[1]; +#else + unsigned a_bits; #endif TCGReg base = TCG_REG_TMP0; @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, bool is_64) tcg_out_ext32u(s, base, addr_regl); addr_regl = base; } + a_bits = get_alignment_bits(opc); + if (a_bits) { + tcg_out_test_alignment(s, true, addr_regl, a_bits); + } if (guest_base != 0) { tcg_out_opc_reg(s, OPC_ADD, base, TCG_GUEST_BASE_REG, addr_regl); } @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, bool is_64) MemOp opc; #if defined(CONFIG_SOFTMMU) tcg_insn_unit *label_ptr[1]; +#else + unsigned a_bits; #endif TCGReg base = TCG_REG_TMP0; @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, bool is_64) tcg_out_ext32u(s, base, addr_regl); addr_regl = base; } + a_bits = get_alignment_bits(opc); + if (a_bits) { + tcg_out_test_alignment(s, false, addr_regl, a_bits); + } if (guest_base != 0) { tcg_out_opc_reg(s, OPC_ADD, base, TCG_GUEST_BASE_REG, addr_regl); } -- 2.25.1 Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/s390x/tcg-target.h | 2 -- tcg/s390x/tcg-target.c.inc | 59 ++++++++++++++++++++++++++++++++++++-- 2 files changed, 57 insertions(+), 4 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 @@ static inline void tb_target_set_jmp_target(uintptr_t tc_ptr, uintptr_t jmp_rx, /* no need to flush icache explicitly */ } -#ifdef CONFIG_SOFTMMU #define TCG_TARGET_NEED_LDST_LABELS -#endif #define TCG_TARGET_NEED_POOL_LABELS #endif 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 @@ #error "unsupported code generation mode" #endif +#include "../tcg-ldst.c.inc" #include "../tcg-pool.c.inc" #include "elf.h" @@ -XXX,XX +XXX,XX @@ typedef enum S390Opcode { RI_OIHL = 0xa509, RI_OILH = 0xa50a, RI_OILL = 0xa50b, + RI_TMLL = 0xa701, RIE_CGIJ = 0xec7c, RIE_CGRJ = 0xec64, @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_st_direct(TCGContext *s, MemOp opc, TCGReg data, } #if defined(CONFIG_SOFTMMU) -#include "../tcg-ldst.c.inc" - /* We're expecting to use a 20-bit negative offset on the tlb memory ops. */ QEMU_BUILD_BUG_ON(TLB_MASK_TABLE_OFS(0) > 0); QEMU_BUILD_BUG_ON(TLB_MASK_TABLE_OFS(0) < -(1 << 19)); @@ -XXX,XX +XXX,XX @@ static bool tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *lb) return true; } #else +static void tcg_out_test_alignment(TCGContext *s, bool is_ld, + TCGReg addrlo, unsigned a_bits) +{ + unsigned a_mask = (1 << a_bits) - 1; + TCGLabelQemuLdst *l = new_ldst_label(s); + + l->is_ld = is_ld; + l->addrlo_reg = addrlo; + + /* We are expecting a_bits to max out at 7, much lower than TMLL. */ + tcg_debug_assert(a_bits < 16); + tcg_out_insn(s, RI, TMLL, addrlo, a_mask); + + tcg_out16(s, RI_BRC | (7 << 4)); /* CC in {1,2,3} */ + l->label_ptr[0] = s->code_ptr; + s->code_ptr += 1; + + l->raddr = tcg_splitwx_to_rx(s->code_ptr); +} + +static bool tcg_out_fail_alignment(TCGContext *s, TCGLabelQemuLdst *l) +{ + if (!patch_reloc(l->label_ptr[0], R_390_PC16DBL, + (intptr_t)tcg_splitwx_to_rx(s->code_ptr), 2)) { + return false; + } + + tcg_out_mov(s, TCG_TYPE_TL, TCG_REG_R3, l->addrlo_reg); + tcg_out_mov(s, TCG_TYPE_PTR, TCG_REG_R2, TCG_AREG0); + + /* "Tail call" to the helper, with the return address back inline. */ + tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_R14, (uintptr_t)l->raddr); + tgen_gotoi(s, S390_CC_ALWAYS, (const void *)(l->is_ld ? helper_unaligned_ld + : helper_unaligned_st)); + return true; +} + +static bool tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *l) +{ + return tcg_out_fail_alignment(s, l); +} + +static bool tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *l) +{ + return tcg_out_fail_alignment(s, l); +} + static void tcg_prepare_user_ldst(TCGContext *s, TCGReg *addr_reg, TCGReg *index_reg, tcg_target_long *disp) { @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_ld(TCGContext* s, TCGReg data_reg, TCGReg addr_reg, #else TCGReg index_reg; tcg_target_long disp; + unsigned a_bits = get_alignment_bits(opc); + if (a_bits) { + tcg_out_test_alignment(s, true, addr_reg, a_bits); + } tcg_prepare_user_ldst(s, &addr_reg, &index_reg, &disp); tcg_out_qemu_ld_direct(s, opc, data_reg, addr_reg, index_reg, disp); #endif @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_st(TCGContext* s, TCGReg data_reg, TCGReg addr_reg, #else TCGReg index_reg; tcg_target_long disp; + unsigned a_bits = get_alignment_bits(opc); + if (a_bits) { + tcg_out_test_alignment(s, false, addr_reg, a_bits); + } tcg_prepare_user_ldst(s, &addr_reg, &index_reg, &disp); tcg_out_qemu_st_direct(s, opc, data_reg, addr_reg, index_reg, disp); #endif -- 2.25.1 Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/tci.c | 20 ++++++++++++++------ 1 file changed, 14 insertions(+), 6 deletions(-) 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_compare64(uint64_t u0, uint64_t u1, TCGCond condition) static uint64_t tci_qemu_ld(CPUArchState *env, target_ulong taddr, MemOpIdx oi, const void *tb_ptr) { - MemOp mop = get_memop(oi) & (MO_BSWAP | MO_SSIZE); + MemOp mop = get_memop(oi); uintptr_t ra = (uintptr_t)tb_ptr; #ifdef CONFIG_SOFTMMU - switch (mop) { + switch (mop & (MO_BSWAP | MO_SSIZE)) { case MO_UB: return helper_ret_ldub_mmu(env, taddr, oi, ra); case MO_SB: @@ -XXX,XX +XXX,XX @@ static uint64_t tci_qemu_ld(CPUArchState *env, target_ulong taddr, } #else void *haddr = g2h(env_cpu(env), taddr); + unsigned a_mask = (1u << get_alignment_bits(mop)) - 1; uint64_t ret; set_helper_retaddr(ra); - switch (mop) { + if (taddr & a_mask) { + helper_unaligned_ld(env, taddr); + } + switch (mop & (MO_BSWAP | MO_SSIZE)) { case MO_UB: ret = ldub_p(haddr); break; @@ -XXX,XX +XXX,XX @@ static uint64_t tci_qemu_ld(CPUArchState *env, target_ulong taddr, static void tci_qemu_st(CPUArchState *env, target_ulong taddr, uint64_t val, MemOpIdx oi, const void *tb_ptr) { - MemOp mop = get_memop(oi) & (MO_BSWAP | MO_SSIZE); + MemOp mop = get_memop(oi); uintptr_t ra = (uintptr_t)tb_ptr; #ifdef CONFIG_SOFTMMU - switch (mop) { + switch (mop & (MO_BSWAP | MO_SIZE)) { case MO_UB: helper_ret_stb_mmu(env, taddr, val, oi, ra); break; @@ -XXX,XX +XXX,XX @@ static void tci_qemu_st(CPUArchState *env, target_ulong taddr, uint64_t val, } #else void *haddr = g2h(env_cpu(env), taddr); + unsigned a_mask = (1u << get_alignment_bits(mop)) - 1; set_helper_retaddr(ra); - switch (mop) { + if (taddr & a_mask) { + helper_unaligned_st(env, taddr); + } + switch (mop & (MO_BSWAP | MO_SIZE)) { case MO_UB: stb_p(haddr, val); break; -- 2.25.1 From: WANG Xuerui <git@xen0n.name> Signed-off-by: WANG Xuerui <git@xen0n.name> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-Id: <20220106134238.3936163-1-git@xen0n.name> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/loongarch64/tcg-target.h | 2 - tcg/loongarch64/tcg-target.c.inc | 71 +++++++++++++++++++++++++++++++- 2 files changed, 69 insertions(+), 4 deletions(-) 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 @@ void tb_target_set_jmp_target(uintptr_t, uintptr_t, uintptr_t, uintptr_t); #define TCG_TARGET_DEFAULT_MO (0) -#ifdef CONFIG_SOFTMMU #define TCG_TARGET_NEED_LDST_LABELS -#endif #define TCG_TARGET_HAS_MEMORY_BSWAP 0 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 @@ * THE SOFTWARE. */ +#include "../tcg-ldst.c.inc" + #ifdef CONFIG_DEBUG_TCG static const char * const tcg_target_reg_names[TCG_TARGET_NB_REGS] = { "zero", @@ -XXX,XX +XXX,XX @@ static bool tcg_out_sti(TCGContext *s, TCGType type, TCGArg val, */ #if defined(CONFIG_SOFTMMU) -#include "../tcg-ldst.c.inc" - /* * helper signature: helper_ret_ld_mmu(CPUState *env, target_ulong addr, * MemOpIdx oi, uintptr_t ra) @@ -XXX,XX +XXX,XX @@ static bool tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *l) return tcg_out_goto(s, l->raddr); } +#else + +/* + * Alignment helpers for user-mode emulation + */ + +static void tcg_out_test_alignment(TCGContext *s, bool is_ld, TCGReg addr_reg, + unsigned a_bits) +{ + TCGLabelQemuLdst *l = new_ldst_label(s); + + l->is_ld = is_ld; + l->addrlo_reg = addr_reg; + + /* + * Without micro-architecture details, we don't know which of bstrpick or + * andi is faster, so use bstrpick as it's not constrained by imm field + * width. (Not to say alignments >= 2^12 are going to happen any time + * soon, though) + */ + tcg_out_opc_bstrpick_d(s, TCG_REG_TMP1, addr_reg, 0, a_bits - 1); + + l->label_ptr[0] = s->code_ptr; + tcg_out_opc_bne(s, TCG_REG_TMP1, TCG_REG_ZERO, 0); + + l->raddr = tcg_splitwx_to_rx(s->code_ptr); +} + +static bool tcg_out_fail_alignment(TCGContext *s, TCGLabelQemuLdst *l) +{ + /* resolve label address */ + if (!reloc_br_sk16(l->label_ptr[0], tcg_splitwx_to_rx(s->code_ptr))) { + return false; + } + + tcg_out_mov(s, TCG_TYPE_TL, TCG_REG_A1, l->addrlo_reg); + tcg_out_mov(s, TCG_TYPE_PTR, TCG_REG_A0, TCG_AREG0); + + /* tail call, with the return address back inline. */ + tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_RA, (uintptr_t)l->raddr); + tcg_out_call_int(s, (const void *)(l->is_ld ? helper_unaligned_ld + : helper_unaligned_st), true); + return true; +} + +static bool tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *l) +{ + return tcg_out_fail_alignment(s, l); +} + +static bool tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *l) +{ + return tcg_out_fail_alignment(s, l); +} + #endif /* CONFIG_SOFTMMU */ /* @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, TCGType type) MemOp opc; #if defined(CONFIG_SOFTMMU) tcg_insn_unit *label_ptr[1]; +#else + unsigned a_bits; #endif TCGReg base; @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, TCGType type) data_regl, addr_regl, s->code_ptr, label_ptr); #else + a_bits = get_alignment_bits(opc); + if (a_bits) { + tcg_out_test_alignment(s, true, addr_regl, a_bits); + } base = tcg_out_zext_addr_if_32_bit(s, addr_regl, TCG_REG_TMP0); TCGReg guest_base_reg = USE_GUEST_BASE ? TCG_GUEST_BASE_REG : TCG_REG_ZERO; tcg_out_qemu_ld_indexed(s, data_regl, base, guest_base_reg, opc, type); @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args) MemOp opc; #if defined(CONFIG_SOFTMMU) tcg_insn_unit *label_ptr[1]; +#else + unsigned a_bits; #endif TCGReg base; @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args) data_regl, addr_regl, s->code_ptr, label_ptr); #else + a_bits = get_alignment_bits(opc); + if (a_bits) { + tcg_out_test_alignment(s, false, addr_regl, a_bits); + } base = tcg_out_zext_addr_if_32_bit(s, addr_regl, TCG_REG_TMP0); TCGReg guest_base_reg = USE_GUEST_BASE ? TCG_GUEST_BASE_REG : TCG_REG_ZERO; tcg_out_qemu_st_indexed(s, data_regl, base, guest_base_reg, opc); -- 2.25.1 Support for unaligned accesses is difficult for pre-v6 hosts. While debian still builds for armv4, we cannot use a compile time test, so test the architecture at runtime and error out. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/arm/tcg-target.c.inc | 5 +++++ 1 file changed, 5 insertions(+) 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_target_init(TCGContext *s) if (pl != NULL && pl[0] == 'v' && pl[1] >= '4' && pl[1] <= '9') { arm_arch = pl[1] - '0'; } + + if (arm_arch < 6) { + error_report("TCG: ARMv%d is unsupported; exiting", arm_arch); + exit(EXIT_FAILURE); + } } tcg_target_available_regs[TCG_TYPE_I32] = ALL_GENERAL_REGS; -- 2.25.1 This is now always true, since we require armv6. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/arm/tcg-target.h | 3 +-- tcg/arm/tcg-target.c.inc | 35 ++++++----------------------------- 2 files changed, 7 insertions(+), 31 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 int arm_arch; -#define use_armv5t_instructions (__ARM_ARCH >= 5 || arm_arch >= 5) #define use_armv6_instructions (__ARM_ARCH >= 6 || arm_arch >= 6) #define use_armv7_instructions (__ARM_ARCH >= 7 || arm_arch >= 7) @@ -XXX,XX +XXX,XX @@ extern bool use_neon_instructions; #define TCG_TARGET_HAS_eqv_i32 0 #define TCG_TARGET_HAS_nand_i32 0 #define TCG_TARGET_HAS_nor_i32 0 -#define TCG_TARGET_HAS_clz_i32 use_armv5t_instructions +#define TCG_TARGET_HAS_clz_i32 1 #define TCG_TARGET_HAS_ctz_i32 use_armv7_instructions #define TCG_TARGET_HAS_ctpop_i32 0 #define TCG_TARGET_HAS_deposit_i32 use_armv7_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_b_reg(TCGContext *s, ARMCond cond, TCGReg rn) * Unless the C portion of QEMU is compiled as thumb, we don't need * true BX semantics; merely a branch to an address held in a register. */ - if (use_armv5t_instructions) { - tcg_out_bx_reg(s, cond, rn); - } else { - tcg_out_mov_reg(s, cond, TCG_REG_PC, rn); - } + tcg_out_bx_reg(s, cond, rn); } static void tcg_out_dat_imm(TCGContext *s, ARMCond cond, ARMInsn opc, @@ -XXX,XX +XXX,XX @@ static void tcg_out_goto(TCGContext *s, ARMCond cond, const tcg_insn_unit *addr) } /* LDR is interworking from v5t. */ - if (arm_mode || use_armv5t_instructions) { - tcg_out_movi_pool(s, cond, TCG_REG_PC, addri); - return; - } - - /* else v4t */ - tcg_out_movi32(s, COND_AL, TCG_REG_TMP, addri); - tcg_out_bx_reg(s, COND_AL, TCG_REG_TMP); + tcg_out_movi_pool(s, cond, TCG_REG_PC, addri); } /* @@ -XXX,XX +XXX,XX @@ static void tcg_out_call(TCGContext *s, const tcg_insn_unit *addr) if (disp - 8 < 0x02000000 && disp - 8 >= -0x02000000) { if (arm_mode) { tcg_out_bl_imm(s, COND_AL, disp); - return; - } - if (use_armv5t_instructions) { + } else { tcg_out_blx_imm(s, disp); - return; } + return; } - if (use_armv5t_instructions) { - tcg_out_movi32(s, COND_AL, TCG_REG_TMP, addri); - tcg_out_blx_reg(s, COND_AL, TCG_REG_TMP); - } else if (arm_mode) { - /* ??? Know that movi_pool emits exactly 1 insn. */ - tcg_out_mov_reg(s, COND_AL, TCG_REG_R14, TCG_REG_PC); - tcg_out_movi_pool(s, COND_AL, TCG_REG_PC, addri); - } else { - tcg_out_movi32(s, COND_AL, TCG_REG_TMP, addri); - tcg_out_mov_reg(s, COND_AL, TCG_REG_R14, TCG_REG_PC); - tcg_out_bx_reg(s, COND_AL, TCG_REG_TMP); - } + tcg_out_movi32(s, COND_AL, TCG_REG_TMP, addri); + tcg_out_blx_reg(s, COND_AL, TCG_REG_TMP); } static void tcg_out_goto_label(TCGContext *s, ARMCond cond, TCGLabel *l) -- 2.25.1 This is now always true, since we require armv6. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/arm/tcg-target.h | 1 - tcg/arm/tcg-target.c.inc | 192 ++++++--------------------------------- 2 files changed, 27 insertions(+), 166 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 int arm_arch; -#define use_armv6_instructions (__ARM_ARCH >= 6 || arm_arch >= 6) #define use_armv7_instructions (__ARM_ARCH >= 7 || arm_arch >= 7) #undef TCG_TARGET_STACK_GROWSUP 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_dat_rIN(TCGContext *s, ARMCond cond, ARMInsn opc, static void tcg_out_mul32(TCGContext *s, ARMCond cond, TCGReg rd, TCGReg rn, TCGReg rm) { - /* if ArchVersion() < 6 && d == n then UNPREDICTABLE; */ - if (!use_armv6_instructions && rd == rn) { - if (rd == rm) { - /* rd == rn == rm; copy an input to tmp first. */ - tcg_out_mov_reg(s, cond, TCG_REG_TMP, rn); - rm = rn = TCG_REG_TMP; - } else { - rn = rm; - rm = rd; - } - } /* mul */ tcg_out32(s, (cond << 28) | 0x90 | (rd << 16) | (rm << 8) | rn); } @@ -XXX,XX +XXX,XX @@ static void tcg_out_mul32(TCGContext *s, ARMCond cond, TCGReg rd, static void tcg_out_umull32(TCGContext *s, ARMCond cond, TCGReg rd0, TCGReg rd1, TCGReg rn, TCGReg rm) { - /* if ArchVersion() < 6 && (dHi == n || dLo == n) then UNPREDICTABLE; */ - if (!use_armv6_instructions && (rd0 == rn || rd1 == rn)) { - if (rd0 == rm || rd1 == rm) { - tcg_out_mov_reg(s, cond, TCG_REG_TMP, rn); - rn = TCG_REG_TMP; - } else { - TCGReg t = rn; - rn = rm; - rm = t; - } - } /* umull */ tcg_out32(s, (cond << 28) | 0x00800090 | (rd1 << 16) | (rd0 << 12) | (rm << 8) | rn); @@ -XXX,XX +XXX,XX @@ static void tcg_out_umull32(TCGContext *s, ARMCond cond, TCGReg rd0, static void tcg_out_smull32(TCGContext *s, ARMCond cond, TCGReg rd0, TCGReg rd1, TCGReg rn, TCGReg rm) { - /* if ArchVersion() < 6 && (dHi == n || dLo == n) then UNPREDICTABLE; */ - if (!use_armv6_instructions && (rd0 == rn || rd1 == rn)) { - if (rd0 == rm || rd1 == rm) { - tcg_out_mov_reg(s, cond, TCG_REG_TMP, rn); - rn = TCG_REG_TMP; - } else { - TCGReg t = rn; - rn = rm; - rm = t; - } - } /* smull */ tcg_out32(s, (cond << 28) | 0x00c00090 | (rd1 << 16) | (rd0 << 12) | (rm << 8) | rn); @@ -XXX,XX +XXX,XX @@ static void tcg_out_udiv(TCGContext *s, ARMCond cond, static void tcg_out_ext8s(TCGContext *s, ARMCond cond, TCGReg rd, TCGReg rn) { - if (use_armv6_instructions) { - /* sxtb */ - tcg_out32(s, 0x06af0070 | (cond << 28) | (rd << 12) | rn); - } else { - tcg_out_dat_reg(s, cond, ARITH_MOV, - rd, 0, rn, SHIFT_IMM_LSL(24)); - tcg_out_dat_reg(s, cond, ARITH_MOV, - rd, 0, rd, SHIFT_IMM_ASR(24)); - } + /* sxtb */ + tcg_out32(s, 0x06af0070 | (cond << 28) | (rd << 12) | rn); } static void __attribute__((unused)) @@ -XXX,XX +XXX,XX @@ tcg_out_ext8u(TCGContext *s, ARMCond cond, TCGReg rd, TCGReg rn) static void tcg_out_ext16s(TCGContext *s, ARMCond cond, TCGReg rd, TCGReg rn) { - if (use_armv6_instructions) { - /* sxth */ - tcg_out32(s, 0x06bf0070 | (cond << 28) | (rd << 12) | rn); - } else { - tcg_out_dat_reg(s, cond, ARITH_MOV, - rd, 0, rn, SHIFT_IMM_LSL(16)); - tcg_out_dat_reg(s, cond, ARITH_MOV, - rd, 0, rd, SHIFT_IMM_ASR(16)); - } + /* sxth */ + tcg_out32(s, 0x06bf0070 | (cond << 28) | (rd << 12) | rn); } static void tcg_out_ext16u(TCGContext *s, ARMCond cond, TCGReg rd, TCGReg rn) { - if (use_armv6_instructions) { - /* uxth */ - tcg_out32(s, 0x06ff0070 | (cond << 28) | (rd << 12) | rn); - } else { - tcg_out_dat_reg(s, cond, ARITH_MOV, - rd, 0, rn, SHIFT_IMM_LSL(16)); - tcg_out_dat_reg(s, cond, ARITH_MOV, - rd, 0, rd, SHIFT_IMM_LSR(16)); - } + /* uxth */ + tcg_out32(s, 0x06ff0070 | (cond << 28) | (rd << 12) | rn); } static void tcg_out_bswap16(TCGContext *s, ARMCond cond, TCGReg rd, TCGReg rn, int flags) { - if (use_armv6_instructions) { - if (flags & TCG_BSWAP_OS) { - /* revsh */ - tcg_out32(s, 0x06ff0fb0 | (cond << 28) | (rd << 12) | rn); - return; - } - - /* rev16 */ - tcg_out32(s, 0x06bf0fb0 | (cond << 28) | (rd << 12) | rn); - if ((flags & (TCG_BSWAP_IZ | TCG_BSWAP_OZ)) == TCG_BSWAP_OZ) { - /* uxth */ - tcg_out32(s, 0x06ff0070 | (cond << 28) | (rd << 12) | rd); - } + if (flags & TCG_BSWAP_OS) { + /* revsh */ + tcg_out32(s, 0x06ff0fb0 | (cond << 28) | (rd << 12) | rn); return; } - if (flags == 0) { - /* - * For stores, no input or output extension: - * rn = xxAB - * lsr tmp, rn, #8 tmp = 0xxA - * and tmp, tmp, #0xff tmp = 000A - * orr rd, tmp, rn, lsl #8 rd = xABA - */ - tcg_out_dat_reg(s, cond, ARITH_MOV, - TCG_REG_TMP, 0, rn, SHIFT_IMM_LSR(8)); - tcg_out_dat_imm(s, cond, ARITH_AND, TCG_REG_TMP, TCG_REG_TMP, 0xff); - tcg_out_dat_reg(s, cond, ARITH_ORR, - rd, TCG_REG_TMP, rn, SHIFT_IMM_LSL(8)); - return; + /* rev16 */ + tcg_out32(s, 0x06bf0fb0 | (cond << 28) | (rd << 12) | rn); + if ((flags & (TCG_BSWAP_IZ | TCG_BSWAP_OZ)) == TCG_BSWAP_OZ) { + /* uxth */ + tcg_out32(s, 0x06ff0070 | (cond << 28) | (rd << 12) | rd); } - - /* - * Byte swap, leaving the result at the top of the register. - * We will then shift down, zero or sign-extending. - */ - if (flags & TCG_BSWAP_IZ) { - /* - * rn = 00AB - * ror tmp, rn, #8 tmp = B00A - * orr tmp, tmp, tmp, lsl #16 tmp = BA00 - */ - tcg_out_dat_reg(s, cond, ARITH_MOV, - TCG_REG_TMP, 0, rn, SHIFT_IMM_ROR(8)); - tcg_out_dat_reg(s, cond, ARITH_ORR, - TCG_REG_TMP, TCG_REG_TMP, TCG_REG_TMP, - SHIFT_IMM_LSL(16)); - } else { - /* - * rn = xxAB - * and tmp, rn, #0xff00 tmp = 00A0 - * lsl tmp, tmp, #8 tmp = 0A00 - * orr tmp, tmp, rn, lsl #24 tmp = BA00 - */ - tcg_out_dat_rI(s, cond, ARITH_AND, TCG_REG_TMP, rn, 0xff00, 1); - tcg_out_dat_reg(s, cond, ARITH_MOV, - TCG_REG_TMP, 0, TCG_REG_TMP, SHIFT_IMM_LSL(8)); - tcg_out_dat_reg(s, cond, ARITH_ORR, - TCG_REG_TMP, TCG_REG_TMP, rn, SHIFT_IMM_LSL(24)); - } - tcg_out_dat_reg(s, cond, ARITH_MOV, rd, 0, TCG_REG_TMP, - (flags & TCG_BSWAP_OS - ? SHIFT_IMM_ASR(8) : SHIFT_IMM_LSR(8))); } static void tcg_out_bswap32(TCGContext *s, ARMCond cond, TCGReg rd, TCGReg rn) { - if (use_armv6_instructions) { - /* rev */ - tcg_out32(s, 0x06bf0f30 | (cond << 28) | (rd << 12) | rn); - } else { - tcg_out_dat_reg(s, cond, ARITH_EOR, - TCG_REG_TMP, rn, rn, SHIFT_IMM_ROR(16)); - tcg_out_dat_imm(s, cond, ARITH_BIC, - TCG_REG_TMP, TCG_REG_TMP, 0xff | 0x800); - tcg_out_dat_reg(s, cond, ARITH_MOV, - rd, 0, rn, SHIFT_IMM_ROR(8)); - tcg_out_dat_reg(s, cond, ARITH_EOR, - rd, rd, TCG_REG_TMP, SHIFT_IMM_LSR(8)); - } + /* rev */ + tcg_out32(s, 0x06bf0f30 | (cond << 28) | (rd << 12) | rn); } static void tcg_out_deposit(TCGContext *s, ARMCond cond, TCGReg rd, @@ -XXX,XX +XXX,XX @@ static void tcg_out_mb(TCGContext *s, TCGArg a0) { if (use_armv7_instructions) { tcg_out32(s, INSN_DMB_ISH); - } else if (use_armv6_instructions) { + } else { tcg_out32(s, INSN_DMB_MCR); } } @@ -XXX,XX +XXX,XX @@ static TCGReg tcg_out_arg_reg64(TCGContext *s, TCGReg argreg, if (argreg & 1) { argreg++; } - if (use_armv6_instructions && argreg >= 4 - && (arglo & 1) == 0 && arghi == arglo + 1) { + if (argreg >= 4 && (arglo & 1) == 0 && arghi == arglo + 1) { tcg_out_strd_8(s, COND_AL, arglo, TCG_REG_CALL_STACK, (argreg - 4) * 4); return argreg + 2; @@ -XXX,XX +XXX,XX @@ static TCGReg tcg_out_tlb_read(TCGContext *s, TCGReg addrlo, TCGReg addrhi, int cmp_off = (is_load ? offsetof(CPUTLBEntry, addr_read) : offsetof(CPUTLBEntry, addr_write)); int fast_off = TLB_MASK_TABLE_OFS(mem_index); - int mask_off = fast_off + offsetof(CPUTLBDescFast, mask); - int table_off = fast_off + offsetof(CPUTLBDescFast, table); unsigned s_bits = opc & MO_SIZE; unsigned a_bits = get_alignment_bits(opc); @@ -XXX,XX +XXX,XX @@ static TCGReg tcg_out_tlb_read(TCGContext *s, TCGReg addrlo, TCGReg addrhi, } /* Load env_tlb(env)->f[mmu_idx].{mask,table} into {r0,r1}. */ - if (use_armv6_instructions) { - tcg_out_ldrd_8(s, COND_AL, TCG_REG_R0, TCG_AREG0, fast_off); - } else { - tcg_out_ld(s, TCG_TYPE_I32, TCG_REG_R0, TCG_AREG0, mask_off); - tcg_out_ld(s, TCG_TYPE_I32, TCG_REG_R1, TCG_AREG0, table_off); - } + tcg_out_ldrd_8(s, COND_AL, TCG_REG_R0, TCG_AREG0, fast_off); /* Extract the tlb index from the address into R0. */ tcg_out_dat_reg(s, COND_AL, ARITH_AND, TCG_REG_R0, TCG_REG_R0, addrlo, @@ -XXX,XX +XXX,XX @@ static TCGReg tcg_out_tlb_read(TCGContext *s, TCGReg addrlo, TCGReg addrhi, * Load the tlb comparator into R2/R3 and the fast path addend into R1. */ if (cmp_off == 0) { - if (use_armv6_instructions && TARGET_LONG_BITS == 64) { + if (TARGET_LONG_BITS == 64) { tcg_out_ldrd_rwb(s, COND_AL, TCG_REG_R2, TCG_REG_R1, TCG_REG_R0); } else { tcg_out_ld32_rwb(s, COND_AL, TCG_REG_R2, TCG_REG_R1, TCG_REG_R0); @@ -XXX,XX +XXX,XX @@ static TCGReg tcg_out_tlb_read(TCGContext *s, TCGReg addrlo, TCGReg addrhi, } else { tcg_out_dat_reg(s, COND_AL, ARITH_ADD, TCG_REG_R1, TCG_REG_R1, TCG_REG_R0, 0); - if (use_armv6_instructions && TARGET_LONG_BITS == 64) { + if (TARGET_LONG_BITS == 64) { tcg_out_ldrd_8(s, COND_AL, TCG_REG_R2, TCG_REG_R1, cmp_off); } else { tcg_out_ld32_12(s, COND_AL, TCG_REG_R2, TCG_REG_R1, cmp_off); } } - if (!use_armv6_instructions && TARGET_LONG_BITS == 64) { - tcg_out_ld32_12(s, COND_AL, TCG_REG_R3, TCG_REG_R1, cmp_off + 4); - } /* Load the tlb addend. */ tcg_out_ld32_12(s, COND_AL, TCG_REG_R1, TCG_REG_R1, @@ -XXX,XX +XXX,XX @@ static bool tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *lb) TCGReg argreg, datalo, datahi; MemOpIdx oi = lb->oi; MemOp opc = get_memop(oi); - void *func; if (!reloc_pc24(lb->label_ptr[0], tcg_splitwx_to_rx(s->code_ptr))) { return false; @@ -XXX,XX +XXX,XX @@ static bool tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *lb) argreg = tcg_out_arg_imm32(s, argreg, oi); argreg = tcg_out_arg_reg32(s, argreg, TCG_REG_R14); - /* For armv6 we can use the canonical unsigned helpers and minimize - icache usage. For pre-armv6, use the signed helpers since we do - not have a single insn sign-extend. */ - if (use_armv6_instructions) { - func = qemu_ld_helpers[opc & MO_SIZE]; - } else { - func = qemu_ld_helpers[opc & MO_SSIZE]; - if (opc & MO_SIGN) { - opc = MO_UL; - } - } - tcg_out_call(s, func); + /* Use the canonical unsigned helpers and minimize icache usage. */ + tcg_out_call(s, qemu_ld_helpers[opc & MO_SIZE]); datalo = lb->datalo_reg; datahi = lb->datahi_reg; @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_ld_index(TCGContext *s, MemOp opc, break; case MO_UQ: /* Avoid ldrd for user-only emulation, to handle unaligned. */ - if (USING_SOFTMMU && use_armv6_instructions + if (USING_SOFTMMU && (datalo & 1) == 0 && datahi == datalo + 1) { tcg_out_ldrd_r(s, COND_AL, datalo, addrlo, addend); } else if (datalo != addend) { @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_ld_direct(TCGContext *s, MemOp opc, TCGReg datalo, break; case MO_UQ: /* Avoid ldrd for user-only emulation, to handle unaligned. */ - if (USING_SOFTMMU && use_armv6_instructions + if (USING_SOFTMMU && (datalo & 1) == 0 && datahi == datalo + 1) { tcg_out_ldrd_8(s, COND_AL, datalo, addrlo, 0); } else if (datalo == addrlo) { @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_st_index(TCGContext *s, ARMCond cond, MemOp opc, break; case MO_64: /* Avoid strd for user-only emulation, to handle unaligned. */ - if (USING_SOFTMMU && use_armv6_instructions + if (USING_SOFTMMU && (datalo & 1) == 0 && datahi == datalo + 1) { tcg_out_strd_r(s, cond, datalo, addrlo, addend); } else { @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_st_direct(TCGContext *s, MemOp opc, TCGReg datalo, break; case MO_64: /* Avoid strd for user-only emulation, to handle unaligned. */ - if (USING_SOFTMMU && use_armv6_instructions + if (USING_SOFTMMU && (datalo & 1) == 0 && datahi == datalo + 1) { tcg_out_strd_8(s, COND_AL, datalo, addrlo, 0); } else { -- 2.25.1 We will shortly allow the use of unaligned memory accesses, and these require proper alignment. Use get_alignment_bits to verify and remove USING_SOFTMMU. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/arm/tcg-target.c.inc | 23 ++++++++--------------- 1 file changed, 8 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 @@ bool use_idiv_instructions; bool use_neon_instructions; #endif -/* ??? Ought to think about changing CONFIG_SOFTMMU to always defined. */ -#ifdef CONFIG_SOFTMMU -# define USING_SOFTMMU 1 -#else -# define USING_SOFTMMU 0 -#endif - #ifdef CONFIG_DEBUG_TCG static const char * const tcg_target_reg_names[TCG_TARGET_NB_REGS] = { "%r0", "%r1", "%r2", "%r3", "%r4", "%r5", "%r6", "%r7", @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_ld_index(TCGContext *s, MemOp opc, tcg_out_ld32_r(s, COND_AL, datalo, addrlo, addend); break; case MO_UQ: - /* Avoid ldrd for user-only emulation, to handle unaligned. */ - if (USING_SOFTMMU + /* LDRD requires alignment; double-check that. */ + if (get_alignment_bits(opc) >= MO_64 && (datalo & 1) == 0 && datahi == datalo + 1) { tcg_out_ldrd_r(s, COND_AL, datalo, addrlo, addend); } else if (datalo != addend) { @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_ld_direct(TCGContext *s, MemOp opc, TCGReg datalo, tcg_out_ld32_12(s, COND_AL, datalo, addrlo, 0); break; case MO_UQ: - /* Avoid ldrd for user-only emulation, to handle unaligned. */ - if (USING_SOFTMMU + /* LDRD requires alignment; double-check that. */ + if (get_alignment_bits(opc) >= MO_64 && (datalo & 1) == 0 && datahi == datalo + 1) { tcg_out_ldrd_8(s, COND_AL, datalo, addrlo, 0); } else if (datalo == addrlo) { @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_st_index(TCGContext *s, ARMCond cond, MemOp opc, tcg_out_st32_r(s, cond, datalo, addrlo, addend); break; case MO_64: - /* Avoid strd for user-only emulation, to handle unaligned. */ - if (USING_SOFTMMU + /* STRD requires alignment; double-check that. */ + if (get_alignment_bits(opc) >= MO_64 && (datalo & 1) == 0 && datahi == datalo + 1) { tcg_out_strd_r(s, cond, datalo, addrlo, addend); } else { @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_st_direct(TCGContext *s, MemOp opc, TCGReg datalo, tcg_out_st32_12(s, COND_AL, datalo, addrlo, 0); break; case MO_64: - /* Avoid strd for user-only emulation, to handle unaligned. */ - if (USING_SOFTMMU + /* STRD requires alignment; double-check that. */ + if (get_alignment_bits(opc) >= MO_64 && (datalo & 1) == 0 && datahi == datalo + 1) { tcg_out_strd_8(s, COND_AL, datalo, addrlo, 0); } else { -- 2.25.1 From armv6, the architecture supports unaligned accesses. All we need to do is perform the correct alignment check in tcg_out_tlb_read. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/arm/tcg-target.c.inc | 41 ++++++++++++++++++++-------------------- 1 file changed, 21 insertions(+), 20 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 TCGReg tcg_out_tlb_read(TCGContext *s, TCGReg addrlo, TCGReg addrhi, int cmp_off = (is_load ? offsetof(CPUTLBEntry, addr_read) : offsetof(CPUTLBEntry, addr_write)); int fast_off = TLB_MASK_TABLE_OFS(mem_index); - unsigned s_bits = opc & MO_SIZE; - unsigned a_bits = get_alignment_bits(opc); - - /* - * We don't support inline unaligned acceses, but we can easily - * support overalignment checks. - */ - if (a_bits < s_bits) { - a_bits = s_bits; - } + unsigned s_mask = (1 << (opc & MO_SIZE)) - 1; + unsigned a_mask = (1 << get_alignment_bits(opc)) - 1; + TCGReg t_addr; /* Load env_tlb(env)->f[mmu_idx].{mask,table} into {r0,r1}. */ tcg_out_ldrd_8(s, COND_AL, TCG_REG_R0, TCG_AREG0, fast_off); @@ -XXX,XX +XXX,XX @@ static TCGReg tcg_out_tlb_read(TCGContext *s, TCGReg addrlo, TCGReg addrhi, /* * Check alignment, check comparators. - * Do this in no more than 3 insns. Use MOVW for v7, if possible, + * Do this in 2-4 insns. Use MOVW for v7, if possible, * to reduce the number of sequential conditional instructions. * Almost all guests have at least 4k pages, which means that we need * to clear at least 9 bits even for an 8-byte memory, which means it * isn't worth checking for an immediate operand for BIC. + * + * For unaligned accesses, test the page of the last unit of alignment. + * This leaves the least significant alignment bits unchanged, and of + * course must be zero. */ + t_addr = addrlo; + if (a_mask < s_mask) { + t_addr = TCG_REG_R0; + tcg_out_dat_imm(s, COND_AL, ARITH_ADD, t_addr, + addrlo, s_mask - a_mask); + } if (use_armv7_instructions && TARGET_PAGE_BITS <= 16) { - tcg_target_ulong mask = ~(TARGET_PAGE_MASK | ((1 << a_bits) - 1)); - - tcg_out_movi32(s, COND_AL, TCG_REG_TMP, mask); + tcg_out_movi32(s, COND_AL, TCG_REG_TMP, ~(TARGET_PAGE_MASK | a_mask)); tcg_out_dat_reg(s, COND_AL, ARITH_BIC, TCG_REG_TMP, - addrlo, TCG_REG_TMP, 0); + t_addr, TCG_REG_TMP, 0); tcg_out_dat_reg(s, COND_AL, ARITH_CMP, 0, TCG_REG_R2, TCG_REG_TMP, 0); } else { - if (a_bits) { - tcg_out_dat_imm(s, COND_AL, ARITH_TST, 0, addrlo, - (1 << a_bits) - 1); + if (a_mask) { + tcg_debug_assert(a_mask <= 0xff); + tcg_out_dat_imm(s, COND_AL, ARITH_TST, 0, addrlo, a_mask); } - tcg_out_dat_reg(s, COND_AL, ARITH_MOV, TCG_REG_TMP, 0, addrlo, + tcg_out_dat_reg(s, COND_AL, ARITH_MOV, TCG_REG_TMP, 0, t_addr, SHIFT_IMM_LSR(TARGET_PAGE_BITS)); - tcg_out_dat_reg(s, (a_bits ? COND_EQ : COND_AL), ARITH_CMP, + tcg_out_dat_reg(s, (a_mask ? COND_EQ : COND_AL), ARITH_CMP, 0, TCG_REG_R2, TCG_REG_TMP, SHIFT_IMM_LSL(TARGET_PAGE_BITS)); } -- 2.25.1 Reserve a register for the guest_base using aarch64 for reference. By doing so, we do not have to recompute it for every memory load. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/arm/tcg-target.c.inc | 39 ++++++++++++++++++++++++++++----------- 1 file changed, 28 insertions(+), 11 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 const int tcg_target_call_oarg_regs[2] = { #define TCG_REG_TMP TCG_REG_R12 #define TCG_VEC_TMP TCG_REG_Q15 +#ifndef CONFIG_SOFTMMU +#define TCG_REG_GUEST_BASE TCG_REG_R11 +#endif typedef enum { COND_EQ = 0x0, @@ -XXX,XX +XXX,XX @@ static bool tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *lb) static void tcg_out_qemu_ld_index(TCGContext *s, MemOp opc, TCGReg datalo, TCGReg datahi, - TCGReg addrlo, TCGReg addend) + TCGReg addrlo, TCGReg addend, + bool scratch_addend) { /* Byte swapping is left to middle-end expansion. */ tcg_debug_assert((opc & MO_BSWAP) == 0); @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_ld_index(TCGContext *s, MemOp opc, if (get_alignment_bits(opc) >= MO_64 && (datalo & 1) == 0 && datahi == datalo + 1) { tcg_out_ldrd_r(s, COND_AL, datalo, addrlo, addend); - } else if (datalo != addend) { + } else if (scratch_addend) { tcg_out_ld32_rwb(s, COND_AL, datalo, addend, addrlo); tcg_out_ld32_12(s, COND_AL, datahi, addend, 4); } else { @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, bool is64) label_ptr = s->code_ptr; tcg_out_bl_imm(s, COND_NE, 0); - tcg_out_qemu_ld_index(s, opc, datalo, datahi, addrlo, addend); + tcg_out_qemu_ld_index(s, opc, datalo, datahi, addrlo, addend, true); add_qemu_ldst_label(s, true, oi, datalo, datahi, addrlo, addrhi, s->code_ptr, label_ptr); #else /* !CONFIG_SOFTMMU */ if (guest_base) { - tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_TMP, guest_base); - tcg_out_qemu_ld_index(s, opc, datalo, datahi, addrlo, TCG_REG_TMP); + tcg_out_qemu_ld_index(s, opc, datalo, datahi, + addrlo, TCG_REG_GUEST_BASE, false); } else { tcg_out_qemu_ld_direct(s, opc, datalo, datahi, addrlo); } @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, bool is64) static void tcg_out_qemu_st_index(TCGContext *s, ARMCond cond, MemOp opc, TCGReg datalo, TCGReg datahi, - TCGReg addrlo, TCGReg addend) + TCGReg addrlo, TCGReg addend, + bool scratch_addend) { /* Byte swapping is left to middle-end expansion. */ tcg_debug_assert((opc & MO_BSWAP) == 0); @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_st_index(TCGContext *s, ARMCond cond, MemOp opc, if (get_alignment_bits(opc) >= MO_64 && (datalo & 1) == 0 && datahi == datalo + 1) { tcg_out_strd_r(s, cond, datalo, addrlo, addend); - } else { + } else if (scratch_addend) { tcg_out_st32_rwb(s, cond, datalo, addend, addrlo); tcg_out_st32_12(s, cond, datahi, addend, 4); + } else { + tcg_out_dat_reg(s, cond, ARITH_ADD, TCG_REG_TMP, + addend, addrlo, SHIFT_IMM_LSL(0)); + tcg_out_st32_12(s, cond, datalo, TCG_REG_TMP, 0); + tcg_out_st32_12(s, cond, datahi, TCG_REG_TMP, 4); } break; default: @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, bool is64) mem_index = get_mmuidx(oi); addend = tcg_out_tlb_read(s, addrlo, addrhi, opc, mem_index, 0); - tcg_out_qemu_st_index(s, COND_EQ, opc, datalo, datahi, addrlo, addend); + tcg_out_qemu_st_index(s, COND_EQ, opc, datalo, datahi, + addrlo, addend, true); /* The conditional call must come last, as we're going to return here. */ label_ptr = s->code_ptr; @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, bool is64) s->code_ptr, label_ptr); #else /* !CONFIG_SOFTMMU */ if (guest_base) { - tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_TMP, guest_base); - tcg_out_qemu_st_index(s, COND_AL, opc, datalo, - datahi, addrlo, TCG_REG_TMP); + tcg_out_qemu_st_index(s, COND_AL, opc, datalo, datahi, + addrlo, TCG_REG_GUEST_BASE, false); } else { tcg_out_qemu_st_direct(s, opc, datalo, datahi, addrlo); } @@ -XXX,XX +XXX,XX @@ static void tcg_target_qemu_prologue(TCGContext *s) tcg_out_mov(s, TCG_TYPE_PTR, TCG_AREG0, tcg_target_call_iarg_regs[0]); +#ifndef CONFIG_SOFTMMU + if (guest_base) { + tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_GUEST_BASE, guest_base); + tcg_regset_set_reg(s->reserved_regs, TCG_REG_GUEST_BASE); + } +#endif + tcg_out_b_reg(s, COND_AL, tcg_target_call_iarg_regs[1]); /* -- 2.25.1 Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/arm/tcg-target.h | 2 - tcg/arm/tcg-target.c.inc | 83 +++++++++++++++++++++++++++++++++++++++- 2 files changed, 81 insertions(+), 4 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; /* not defined -- call should be eliminated at compile time */ void tb_target_set_jmp_target(uintptr_t, uintptr_t, uintptr_t, uintptr_t); -#ifdef CONFIG_SOFTMMU #define TCG_TARGET_NEED_LDST_LABELS -#endif #define TCG_TARGET_NEED_POOL_LABELS #endif 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 @@ */ #include "elf.h" +#include "../tcg-ldst.c.inc" #include "../tcg-pool.c.inc" int arm_arch = __ARM_ARCH; @@ -XXX,XX +XXX,XX @@ static void tcg_out_vldst(TCGContext *s, ARMInsn insn, } #ifdef CONFIG_SOFTMMU -#include "../tcg-ldst.c.inc" - /* helper signature: helper_ret_ld_mmu(CPUState *env, target_ulong addr, * int mmu_idx, uintptr_t ra) */ @@ -XXX,XX +XXX,XX @@ static bool tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *lb) tcg_out_goto(s, COND_AL, qemu_st_helpers[opc & MO_SIZE]); return true; } +#else + +static void tcg_out_test_alignment(TCGContext *s, bool is_ld, TCGReg addrlo, + TCGReg addrhi, unsigned a_bits) +{ + unsigned a_mask = (1 << a_bits) - 1; + TCGLabelQemuLdst *label = new_ldst_label(s); + + label->is_ld = is_ld; + label->addrlo_reg = addrlo; + label->addrhi_reg = addrhi; + + /* We are expecting a_bits to max out at 7, and can easily support 8. */ + tcg_debug_assert(a_mask <= 0xff); + /* tst addr, #mask */ + tcg_out_dat_imm(s, COND_AL, ARITH_TST, 0, addrlo, a_mask); + + /* blne slow_path */ + label->label_ptr[0] = s->code_ptr; + tcg_out_bl_imm(s, COND_NE, 0); + + label->raddr = tcg_splitwx_to_rx(s->code_ptr); +} + +static bool tcg_out_fail_alignment(TCGContext *s, TCGLabelQemuLdst *l) +{ + if (!reloc_pc24(l->label_ptr[0], tcg_splitwx_to_rx(s->code_ptr))) { + return false; + } + + if (TARGET_LONG_BITS == 64) { + /* 64-bit target address is aligned into R2:R3. */ + if (l->addrhi_reg != TCG_REG_R2) { + tcg_out_mov(s, TCG_TYPE_I32, TCG_REG_R2, l->addrlo_reg); + tcg_out_mov(s, TCG_TYPE_I32, TCG_REG_R3, l->addrhi_reg); + } else if (l->addrlo_reg != TCG_REG_R3) { + tcg_out_mov(s, TCG_TYPE_I32, TCG_REG_R3, l->addrhi_reg); + tcg_out_mov(s, TCG_TYPE_I32, TCG_REG_R2, l->addrlo_reg); + } else { + tcg_out_mov(s, TCG_TYPE_I32, TCG_REG_R1, TCG_REG_R2); + tcg_out_mov(s, TCG_TYPE_I32, TCG_REG_R2, TCG_REG_R3); + tcg_out_mov(s, TCG_TYPE_I32, TCG_REG_R3, TCG_REG_R1); + } + } else { + tcg_out_mov(s, TCG_TYPE_I32, TCG_REG_R1, l->addrlo_reg); + } + tcg_out_mov(s, TCG_TYPE_PTR, TCG_REG_R0, TCG_AREG0); + + /* + * Tail call to the helper, with the return address back inline, + * just for the clarity of the debugging traceback -- the helper + * cannot return. We have used BLNE to arrive here, so LR is + * already set. + */ + tcg_out_goto(s, COND_AL, (const void *) + (l->is_ld ? helper_unaligned_ld : helper_unaligned_st)); + return true; +} + +static bool tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *l) +{ + return tcg_out_fail_alignment(s, l); +} + +static bool tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *l) +{ + return tcg_out_fail_alignment(s, l); +} #endif /* SOFTMMU */ static void tcg_out_qemu_ld_index(TCGContext *s, MemOp opc, @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, bool is64) int mem_index; TCGReg addend; tcg_insn_unit *label_ptr; +#else + unsigned a_bits; #endif datalo = *args++; @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, bool is64) add_qemu_ldst_label(s, true, oi, datalo, datahi, addrlo, addrhi, s->code_ptr, label_ptr); #else /* !CONFIG_SOFTMMU */ + a_bits = get_alignment_bits(opc); + if (a_bits) { + tcg_out_test_alignment(s, true, addrlo, addrhi, a_bits); + } if (guest_base) { tcg_out_qemu_ld_index(s, opc, datalo, datahi, addrlo, TCG_REG_GUEST_BASE, false); @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, bool is64) int mem_index; TCGReg addend; tcg_insn_unit *label_ptr; +#else + unsigned a_bits; #endif datalo = *args++; @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, bool is64) add_qemu_ldst_label(s, false, oi, datalo, datahi, addrlo, addrhi, s->code_ptr, label_ptr); #else /* !CONFIG_SOFTMMU */ + a_bits = get_alignment_bits(opc); + if (a_bits) { + tcg_out_test_alignment(s, false, addrlo, addrhi, a_bits); + } if (guest_base) { tcg_out_qemu_st_index(s, COND_AL, opc, datalo, datahi, addrlo, TCG_REG_GUEST_BASE, false); -- 2.25.1 This is kinda sorta the opposite of the other tcg hosts, where we get (normal) alignment checks for free with host SIGBUS and need to add code to support unaligned accesses. Fortunately, the ISA contains pairs of instructions that are used to implement unaligned memory accesses. Use them. Tested-by: Jiaxun Yang <jiaxun.yang@flygoat.com> Reviewed-by: Jiaxun Yang <jiaxun.yang@flygoat.com> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/mips/tcg-target.h | 2 - tcg/mips/tcg-target.c.inc | 334 +++++++++++++++++++++++++++++++++++++- 2 files changed, 328 insertions(+), 8 deletions(-) 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; void tb_target_set_jmp_target(uintptr_t, uintptr_t, uintptr_t, uintptr_t) QEMU_ERROR("code path is reachable"); -#ifdef CONFIG_SOFTMMU #define TCG_TARGET_NEED_LDST_LABELS -#endif #endif 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 @@ * THE SOFTWARE. */ +#include "../tcg-ldst.c.inc" + #ifdef HOST_WORDS_BIGENDIAN # define MIPS_BE 1 #else @@ -XXX,XX +XXX,XX @@ typedef enum { OPC_ORI = 015 << 26, OPC_XORI = 016 << 26, OPC_LUI = 017 << 26, + OPC_BNEL = 025 << 26, + OPC_BNEZALC_R6 = 030 << 26, OPC_DADDIU = 031 << 26, + OPC_LDL = 032 << 26, + OPC_LDR = 033 << 26, OPC_LB = 040 << 26, OPC_LH = 041 << 26, + OPC_LWL = 042 << 26, OPC_LW = 043 << 26, OPC_LBU = 044 << 26, OPC_LHU = 045 << 26, + OPC_LWR = 046 << 26, OPC_LWU = 047 << 26, OPC_SB = 050 << 26, OPC_SH = 051 << 26, + OPC_SWL = 052 << 26, OPC_SW = 053 << 26, + OPC_SDL = 054 << 26, + OPC_SDR = 055 << 26, + OPC_SWR = 056 << 26, OPC_LD = 067 << 26, OPC_SD = 077 << 26, @@ -XXX,XX +XXX,XX @@ static void tcg_out_call(TCGContext *s, const tcg_insn_unit *arg) } #if defined(CONFIG_SOFTMMU) -#include "../tcg-ldst.c.inc" - static void * const qemu_ld_helpers[(MO_SSIZE | MO_BSWAP) + 1] = { [MO_UB] = helper_ret_ldub_mmu, [MO_SB] = helper_ret_ldsb_mmu, @@ -XXX,XX +XXX,XX @@ static bool tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *l) tcg_out_mov(s, TCG_TYPE_PTR, tcg_target_call_iarg_regs[0], TCG_AREG0); return true; } -#endif + +#else + +static void tcg_out_test_alignment(TCGContext *s, bool is_ld, TCGReg addrlo, + TCGReg addrhi, unsigned a_bits) +{ + unsigned a_mask = (1 << a_bits) - 1; + TCGLabelQemuLdst *l = new_ldst_label(s); + + l->is_ld = is_ld; + l->addrlo_reg = addrlo; + l->addrhi_reg = addrhi; + + /* We are expecting a_bits to max out at 7, much lower than ANDI. */ + tcg_debug_assert(a_bits < 16); + tcg_out_opc_imm(s, OPC_ANDI, TCG_TMP0, addrlo, a_mask); + + l->label_ptr[0] = s->code_ptr; + if (use_mips32r6_instructions) { + tcg_out_opc_br(s, OPC_BNEZALC_R6, TCG_REG_ZERO, TCG_TMP0); + } else { + tcg_out_opc_br(s, OPC_BNEL, TCG_TMP0, TCG_REG_ZERO); + tcg_out_nop(s); + } + + l->raddr = tcg_splitwx_to_rx(s->code_ptr); +} + +static bool tcg_out_fail_alignment(TCGContext *s, TCGLabelQemuLdst *l) +{ + void *target; + + if (!reloc_pc16(l->label_ptr[0], tcg_splitwx_to_rx(s->code_ptr))) { + return false; + } + + if (TCG_TARGET_REG_BITS < TARGET_LONG_BITS) { + /* A0 is env, A1 is skipped, A2:A3 is the uint64_t address. */ + TCGReg a2 = MIPS_BE ? l->addrhi_reg : l->addrlo_reg; + TCGReg a3 = MIPS_BE ? l->addrlo_reg : l->addrhi_reg; + + if (a3 != TCG_REG_A2) { + tcg_out_mov(s, TCG_TYPE_I32, TCG_REG_A2, a2); + tcg_out_mov(s, TCG_TYPE_I32, TCG_REG_A3, a3); + } else if (a2 != TCG_REG_A3) { + tcg_out_mov(s, TCG_TYPE_I32, TCG_REG_A3, a3); + tcg_out_mov(s, TCG_TYPE_I32, TCG_REG_A2, a2); + } else { + tcg_out_mov(s, TCG_TYPE_I32, TCG_TMP0, TCG_REG_A2); + tcg_out_mov(s, TCG_TYPE_I32, TCG_REG_A2, TCG_REG_A3); + tcg_out_mov(s, TCG_TYPE_I32, TCG_REG_A3, TCG_TMP0); + } + } else { + tcg_out_mov(s, TCG_TYPE_TL, TCG_REG_A1, l->addrlo_reg); + } + tcg_out_mov(s, TCG_TYPE_PTR, TCG_REG_A0, TCG_AREG0); + + /* + * Tail call to the helper, with the return address back inline. + * We have arrived here via BNEL, so $31 is already set. + */ + target = (l->is_ld ? helper_unaligned_ld : helper_unaligned_st); + tcg_out_call_int(s, target, true); + return true; +} + +static bool tcg_out_qemu_ld_slow_path(TCGContext *s, TCGLabelQemuLdst *l) +{ + return tcg_out_fail_alignment(s, l); +} + +static bool tcg_out_qemu_st_slow_path(TCGContext *s, TCGLabelQemuLdst *l) +{ + return tcg_out_fail_alignment(s, l); +} +#endif /* SOFTMMU */ static void tcg_out_qemu_ld_direct(TCGContext *s, TCGReg lo, TCGReg hi, TCGReg base, MemOp opc, bool is_64) @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_ld_direct(TCGContext *s, TCGReg lo, TCGReg hi, } } +static void __attribute__((unused)) +tcg_out_qemu_ld_unalign(TCGContext *s, TCGReg lo, TCGReg hi, + TCGReg base, MemOp opc, bool is_64) +{ + const MIPSInsn lw1 = MIPS_BE ? OPC_LWL : OPC_LWR; + const MIPSInsn lw2 = MIPS_BE ? OPC_LWR : OPC_LWL; + const MIPSInsn ld1 = MIPS_BE ? OPC_LDL : OPC_LDR; + const MIPSInsn ld2 = MIPS_BE ? OPC_LDR : OPC_LDL; + + bool sgn = (opc & MO_SIGN); + + switch (opc & (MO_SSIZE | MO_BSWAP)) { + case MO_SW | MO_BE: + case MO_UW | MO_BE: + tcg_out_opc_imm(s, sgn ? OPC_LB : OPC_LBU, TCG_TMP0, base, 0); + tcg_out_opc_imm(s, OPC_LBU, lo, base, 1); + if (use_mips32r2_instructions) { + tcg_out_opc_bf(s, OPC_INS, lo, TCG_TMP0, 31, 8); + } else { + tcg_out_opc_sa(s, OPC_SLL, TCG_TMP0, TCG_TMP0, 8); + tcg_out_opc_reg(s, OPC_OR, lo, TCG_TMP0, TCG_TMP1); + } + break; + + case MO_SW | MO_LE: + case MO_UW | MO_LE: + if (use_mips32r2_instructions && lo != base) { + tcg_out_opc_imm(s, OPC_LBU, lo, base, 0); + tcg_out_opc_imm(s, sgn ? OPC_LB : OPC_LBU, TCG_TMP0, base, 1); + tcg_out_opc_bf(s, OPC_INS, lo, TCG_TMP0, 31, 8); + } else { + tcg_out_opc_imm(s, OPC_LBU, TCG_TMP0, base, 0); + tcg_out_opc_imm(s, sgn ? OPC_LB : OPC_LBU, TCG_TMP1, base, 1); + tcg_out_opc_sa(s, OPC_SLL, TCG_TMP1, TCG_TMP1, 8); + tcg_out_opc_reg(s, OPC_OR, lo, TCG_TMP0, TCG_TMP1); + } + break; + + case MO_SL: + case MO_UL: + tcg_out_opc_imm(s, lw1, lo, base, 0); + tcg_out_opc_imm(s, lw2, lo, base, 3); + if (TCG_TARGET_REG_BITS == 64 && is_64 && !sgn) { + tcg_out_ext32u(s, lo, lo); + } + break; + + case MO_UL | MO_BSWAP: + case MO_SL | MO_BSWAP: + if (use_mips32r2_instructions) { + tcg_out_opc_imm(s, lw1, lo, base, 0); + tcg_out_opc_imm(s, lw2, lo, base, 3); + tcg_out_bswap32(s, lo, lo, + TCG_TARGET_REG_BITS == 64 && is_64 + ? (sgn ? TCG_BSWAP_OS : TCG_BSWAP_OZ) : 0); + } else { + const tcg_insn_unit *subr = + (TCG_TARGET_REG_BITS == 64 && is_64 && !sgn + ? bswap32u_addr : bswap32_addr); + + tcg_out_opc_imm(s, lw1, TCG_TMP0, base, 0); + tcg_out_bswap_subr(s, subr); + /* delay slot */ + tcg_out_opc_imm(s, lw2, TCG_TMP0, base, 3); + tcg_out_mov(s, is_64 ? TCG_TYPE_I64 : TCG_TYPE_I32, lo, TCG_TMP3); + } + break; + + case MO_UQ: + if (TCG_TARGET_REG_BITS == 64) { + tcg_out_opc_imm(s, ld1, lo, base, 0); + tcg_out_opc_imm(s, ld2, lo, base, 7); + } else { + tcg_out_opc_imm(s, lw1, MIPS_BE ? hi : lo, base, 0 + 0); + tcg_out_opc_imm(s, lw2, MIPS_BE ? hi : lo, base, 0 + 3); + tcg_out_opc_imm(s, lw1, MIPS_BE ? lo : hi, base, 4 + 0); + tcg_out_opc_imm(s, lw2, MIPS_BE ? lo : hi, base, 4 + 3); + } + break; + + case MO_UQ | MO_BSWAP: + if (TCG_TARGET_REG_BITS == 64) { + if (use_mips32r2_instructions) { + tcg_out_opc_imm(s, ld1, lo, base, 0); + tcg_out_opc_imm(s, ld2, lo, base, 7); + tcg_out_bswap64(s, lo, lo); + } else { + tcg_out_opc_imm(s, ld1, TCG_TMP0, base, 0); + tcg_out_bswap_subr(s, bswap64_addr); + /* delay slot */ + tcg_out_opc_imm(s, ld2, TCG_TMP0, base, 7); + tcg_out_mov(s, TCG_TYPE_I64, lo, TCG_TMP3); + } + } else if (use_mips32r2_instructions) { + tcg_out_opc_imm(s, lw1, TCG_TMP0, base, 0 + 0); + tcg_out_opc_imm(s, lw2, TCG_TMP0, base, 0 + 3); + tcg_out_opc_imm(s, lw1, TCG_TMP1, base, 4 + 0); + tcg_out_opc_imm(s, lw2, TCG_TMP1, base, 4 + 3); + tcg_out_opc_reg(s, OPC_WSBH, TCG_TMP0, 0, TCG_TMP0); + tcg_out_opc_reg(s, OPC_WSBH, TCG_TMP1, 0, TCG_TMP1); + tcg_out_opc_sa(s, OPC_ROTR, MIPS_BE ? lo : hi, TCG_TMP0, 16); + tcg_out_opc_sa(s, OPC_ROTR, MIPS_BE ? hi : lo, TCG_TMP1, 16); + } else { + tcg_out_opc_imm(s, lw1, TCG_TMP0, base, 0 + 0); + tcg_out_bswap_subr(s, bswap32_addr); + /* delay slot */ + tcg_out_opc_imm(s, lw2, TCG_TMP0, base, 0 + 3); + tcg_out_opc_imm(s, lw1, TCG_TMP0, base, 4 + 0); + tcg_out_mov(s, TCG_TYPE_I32, MIPS_BE ? lo : hi, TCG_TMP3); + tcg_out_bswap_subr(s, bswap32_addr); + /* delay slot */ + tcg_out_opc_imm(s, lw2, TCG_TMP0, base, 4 + 3); + tcg_out_mov(s, TCG_TYPE_I32, MIPS_BE ? hi : lo, TCG_TMP3); + } + break; + + default: + g_assert_not_reached(); + } +} + static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, bool is_64) { TCGReg addr_regl, addr_regh __attribute__((unused)); @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, bool is_64) MemOp opc; #if defined(CONFIG_SOFTMMU) tcg_insn_unit *label_ptr[2]; +#else + unsigned a_bits, s_bits; #endif TCGReg base = TCG_REG_A0; @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, bool is_64) } else { tcg_out_opc_reg(s, ALIAS_PADD, base, TCG_GUEST_BASE_REG, addr_regl); } - tcg_out_qemu_ld_direct(s, data_regl, data_regh, base, opc, is_64); + a_bits = get_alignment_bits(opc); + s_bits = opc & MO_SIZE; + /* + * R6 removes the left/right instructions but requires the + * system to support misaligned memory accesses. + */ + if (use_mips32r6_instructions) { + if (a_bits) { + tcg_out_test_alignment(s, true, addr_regl, addr_regh, a_bits); + } + tcg_out_qemu_ld_direct(s, data_regl, data_regh, base, opc, is_64); + } else { + if (a_bits && a_bits != s_bits) { + tcg_out_test_alignment(s, true, addr_regl, addr_regh, a_bits); + } + if (a_bits >= s_bits) { + tcg_out_qemu_ld_direct(s, data_regl, data_regh, base, opc, is_64); + } else { + tcg_out_qemu_ld_unalign(s, data_regl, data_regh, base, opc, is_64); + } + } #endif } @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_st_direct(TCGContext *s, TCGReg lo, TCGReg hi, } } +static void __attribute__((unused)) +tcg_out_qemu_st_unalign(TCGContext *s, TCGReg lo, TCGReg hi, + TCGReg base, MemOp opc) +{ + const MIPSInsn sw1 = MIPS_BE ? OPC_SWL : OPC_SWR; + const MIPSInsn sw2 = MIPS_BE ? OPC_SWR : OPC_SWL; + const MIPSInsn sd1 = MIPS_BE ? OPC_SDL : OPC_SDR; + const MIPSInsn sd2 = MIPS_BE ? OPC_SDR : OPC_SDL; + + /* Don't clutter the code below with checks to avoid bswapping ZERO. */ + if ((lo | hi) == 0) { + opc &= ~MO_BSWAP; + } + + switch (opc & (MO_SIZE | MO_BSWAP)) { + case MO_16 | MO_BE: + tcg_out_opc_sa(s, OPC_SRL, TCG_TMP0, lo, 8); + tcg_out_opc_imm(s, OPC_SB, TCG_TMP0, base, 0); + tcg_out_opc_imm(s, OPC_SB, lo, base, 1); + break; + + case MO_16 | MO_LE: + tcg_out_opc_sa(s, OPC_SRL, TCG_TMP0, lo, 8); + tcg_out_opc_imm(s, OPC_SB, lo, base, 0); + tcg_out_opc_imm(s, OPC_SB, TCG_TMP0, base, 1); + break; + + case MO_32 | MO_BSWAP: + tcg_out_bswap32(s, TCG_TMP3, lo, 0); + lo = TCG_TMP3; + /* fall through */ + case MO_32: + tcg_out_opc_imm(s, sw1, lo, base, 0); + tcg_out_opc_imm(s, sw2, lo, base, 3); + break; + + case MO_64 | MO_BSWAP: + if (TCG_TARGET_REG_BITS == 64) { + tcg_out_bswap64(s, TCG_TMP3, lo); + lo = TCG_TMP3; + } else if (use_mips32r2_instructions) { + tcg_out_opc_reg(s, OPC_WSBH, TCG_TMP0, 0, MIPS_BE ? hi : lo); + tcg_out_opc_reg(s, OPC_WSBH, TCG_TMP1, 0, MIPS_BE ? lo : hi); + tcg_out_opc_sa(s, OPC_ROTR, TCG_TMP0, TCG_TMP0, 16); + tcg_out_opc_sa(s, OPC_ROTR, TCG_TMP1, TCG_TMP1, 16); + hi = MIPS_BE ? TCG_TMP0 : TCG_TMP1; + lo = MIPS_BE ? TCG_TMP1 : TCG_TMP0; + } else { + tcg_out_bswap32(s, TCG_TMP3, MIPS_BE ? lo : hi, 0); + tcg_out_opc_imm(s, sw1, TCG_TMP3, base, 0 + 0); + tcg_out_opc_imm(s, sw2, TCG_TMP3, base, 0 + 3); + tcg_out_bswap32(s, TCG_TMP3, MIPS_BE ? hi : lo, 0); + tcg_out_opc_imm(s, sw1, TCG_TMP3, base, 4 + 0); + tcg_out_opc_imm(s, sw2, TCG_TMP3, base, 4 + 3); + break; + } + /* fall through */ + case MO_64: + if (TCG_TARGET_REG_BITS == 64) { + tcg_out_opc_imm(s, sd1, lo, base, 0); + tcg_out_opc_imm(s, sd2, lo, base, 7); + } else { + tcg_out_opc_imm(s, sw1, MIPS_BE ? hi : lo, base, 0 + 0); + tcg_out_opc_imm(s, sw2, MIPS_BE ? hi : lo, base, 0 + 3); + tcg_out_opc_imm(s, sw1, MIPS_BE ? lo : hi, base, 4 + 0); + tcg_out_opc_imm(s, sw2, MIPS_BE ? lo : hi, base, 4 + 3); + } + break; + + default: + tcg_abort(); + } +} static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, bool is_64) { TCGReg addr_regl, addr_regh __attribute__((unused)); @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, bool is_64) MemOp opc; #if defined(CONFIG_SOFTMMU) tcg_insn_unit *label_ptr[2]; +#else + unsigned a_bits, s_bits; #endif TCGReg base = TCG_REG_A0; @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, bool is_64) data_regl, data_regh, addr_regl, addr_regh, s->code_ptr, label_ptr); #else - base = TCG_REG_A0; if (TCG_TARGET_REG_BITS > TARGET_LONG_BITS) { tcg_out_ext32u(s, base, addr_regl); addr_regl = base; @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, bool is_64) } else { tcg_out_opc_reg(s, ALIAS_PADD, base, TCG_GUEST_BASE_REG, addr_regl); } - tcg_out_qemu_st_direct(s, data_regl, data_regh, base, opc); + a_bits = get_alignment_bits(opc); + s_bits = opc & MO_SIZE; + /* + * R6 removes the left/right instructions but requires the + * system to support misaligned memory accesses. + */ + if (use_mips32r6_instructions) { + if (a_bits) { + tcg_out_test_alignment(s, true, addr_regl, addr_regh, a_bits); + } + tcg_out_qemu_st_direct(s, data_regl, data_regh, base, opc); + } else { + if (a_bits && a_bits != s_bits) { + tcg_out_test_alignment(s, true, addr_regl, addr_regh, a_bits); + } + if (a_bits >= s_bits) { + tcg_out_qemu_st_direct(s, data_regl, data_regh, base, opc); + } else { + tcg_out_qemu_st_unalign(s, data_regl, data_regh, base, opc); + } + } #endif } -- 2.25.1 We can use the routines just added for user-only to emit unaligned accesses in softmmu mode too. Tested-by: Jiaxun Yang <jiaxun.yang@flygoat.com> Reviewed-by: Jiaxun Yang <jiaxun.yang@flygoat.com> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/mips/tcg-target.c.inc | 91 ++++++++++++++++++++++----------------- 1 file changed, 51 insertions(+), 40 deletions(-) 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 void tcg_out_tlb_load(TCGContext *s, TCGReg base, TCGReg addrl, tcg_insn_unit *label_ptr[2], bool is_load) { MemOp opc = get_memop(oi); - unsigned s_bits = opc & MO_SIZE; unsigned a_bits = get_alignment_bits(opc); + unsigned s_bits = opc & MO_SIZE; + unsigned a_mask = (1 << a_bits) - 1; + unsigned s_mask = (1 << s_bits) - 1; int mem_index = get_mmuidx(oi); int fast_off = TLB_MASK_TABLE_OFS(mem_index); int mask_off = fast_off + offsetof(CPUTLBDescFast, mask); @@ -XXX,XX +XXX,XX @@ static void tcg_out_tlb_load(TCGContext *s, TCGReg base, TCGReg addrl, int add_off = offsetof(CPUTLBEntry, addend); int cmp_off = (is_load ? offsetof(CPUTLBEntry, addr_read) : offsetof(CPUTLBEntry, addr_write)); - target_ulong mask; + target_ulong tlb_mask; /* Load tlb_mask[mmu_idx] and tlb_table[mmu_idx]. */ tcg_out_ld(s, TCG_TYPE_PTR, TCG_TMP0, TCG_AREG0, mask_off); @@ -XXX,XX +XXX,XX @@ static void tcg_out_tlb_load(TCGContext *s, TCGReg base, TCGReg addrl, /* Add the tlb_table pointer, creating the CPUTLBEntry address in TMP3. */ tcg_out_opc_reg(s, ALIAS_PADD, TCG_TMP3, TCG_TMP3, TCG_TMP1); - /* We don't currently support unaligned accesses. - We could do so with mips32r6. */ - if (a_bits < s_bits) { - a_bits = s_bits; - } - - /* Mask the page bits, keeping the alignment bits to compare against. */ - mask = (target_ulong)TARGET_PAGE_MASK | ((1 << a_bits) - 1); - /* Load the (low-half) tlb comparator. */ if (TCG_TARGET_REG_BITS < TARGET_LONG_BITS) { - tcg_out_ld(s, TCG_TYPE_I32, TCG_TMP0, TCG_TMP3, cmp_off + LO_OFF); - tcg_out_movi(s, TCG_TYPE_I32, TCG_TMP1, mask); + tcg_out_ldst(s, OPC_LW, TCG_TMP0, TCG_TMP3, cmp_off + LO_OFF); } else { tcg_out_ldst(s, (TARGET_LONG_BITS == 64 ? OPC_LD : TCG_TARGET_REG_BITS == 64 ? OPC_LWU : OPC_LW), TCG_TMP0, TCG_TMP3, cmp_off); - tcg_out_movi(s, TCG_TYPE_TL, TCG_TMP1, mask); - /* No second compare is required here; - load the tlb addend for the fast path. */ - tcg_out_ld(s, TCG_TYPE_PTR, TCG_TMP2, TCG_TMP3, add_off); } /* Zero extend a 32-bit guest address for a 64-bit host. */ @@ -XXX,XX +XXX,XX @@ static void tcg_out_tlb_load(TCGContext *s, TCGReg base, TCGReg addrl, tcg_out_ext32u(s, base, addrl); addrl = base; } - tcg_out_opc_reg(s, OPC_AND, TCG_TMP1, TCG_TMP1, addrl); + + /* + * Mask the page bits, keeping the alignment bits to compare against. + * For unaligned accesses, compare against the end of the access to + * verify that it does not cross a page boundary. + */ + tlb_mask = (target_ulong)TARGET_PAGE_MASK | a_mask; + tcg_out_movi(s, TCG_TYPE_I32, TCG_TMP1, tlb_mask); + if (a_mask >= s_mask) { + tcg_out_opc_reg(s, OPC_AND, TCG_TMP1, TCG_TMP1, addrl); + } else { + tcg_out_opc_imm(s, ALIAS_PADDI, TCG_TMP2, addrl, s_mask - a_mask); + tcg_out_opc_reg(s, OPC_AND, TCG_TMP1, TCG_TMP1, TCG_TMP2); + } + + if (TCG_TARGET_REG_BITS >= TARGET_LONG_BITS) { + /* Load the tlb addend for the fast path. */ + tcg_out_ld(s, TCG_TYPE_PTR, TCG_TMP2, TCG_TMP3, add_off); + } label_ptr[0] = s->code_ptr; tcg_out_opc_br(s, OPC_BNE, TCG_TMP1, TCG_TMP0); @@ -XXX,XX +XXX,XX @@ static void tcg_out_tlb_load(TCGContext *s, TCGReg base, TCGReg addrl, /* Load and test the high half tlb comparator. */ if (TCG_TARGET_REG_BITS < TARGET_LONG_BITS) { /* delay slot */ - tcg_out_ld(s, TCG_TYPE_I32, TCG_TMP0, TCG_TMP3, cmp_off + HI_OFF); + tcg_out_ldst(s, OPC_LW, TCG_TMP0, TCG_TMP3, cmp_off + HI_OFF); /* Load the tlb addend for the fast path. */ tcg_out_ld(s, TCG_TYPE_PTR, TCG_TMP2, TCG_TMP3, add_off); @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_ld_direct(TCGContext *s, TCGReg lo, TCGReg hi, } } -static void __attribute__((unused)) -tcg_out_qemu_ld_unalign(TCGContext *s, TCGReg lo, TCGReg hi, +static void tcg_out_qemu_ld_unalign(TCGContext *s, TCGReg lo, TCGReg hi, TCGReg base, MemOp opc, bool is_64) { const MIPSInsn lw1 = MIPS_BE ? OPC_LWL : OPC_LWR; @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, bool is_64) #if defined(CONFIG_SOFTMMU) tcg_insn_unit *label_ptr[2]; #else - unsigned a_bits, s_bits; #endif + unsigned a_bits, s_bits; TCGReg base = TCG_REG_A0; data_regl = *args++; @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, bool is_64) addr_regh = (TCG_TARGET_REG_BITS < TARGET_LONG_BITS ? *args++ : 0); oi = *args++; opc = get_memop(oi); + a_bits = get_alignment_bits(opc); + s_bits = opc & MO_SIZE; + /* + * R6 removes the left/right instructions but requires the + * system to support misaligned memory accesses. + */ #if defined(CONFIG_SOFTMMU) tcg_out_tlb_load(s, base, addr_regl, addr_regh, oi, label_ptr, 1); - tcg_out_qemu_ld_direct(s, data_regl, data_regh, base, opc, is_64); + if (use_mips32r6_instructions || a_bits >= s_bits) { + tcg_out_qemu_ld_direct(s, data_regl, data_regh, base, opc, is_64); + } else { + tcg_out_qemu_ld_unalign(s, data_regl, data_regh, base, opc, is_64); + } add_qemu_ldst_label(s, 1, oi, (is_64 ? TCG_TYPE_I64 : TCG_TYPE_I32), data_regl, data_regh, addr_regl, addr_regh, @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, bool is_64) } else { tcg_out_opc_reg(s, ALIAS_PADD, base, TCG_GUEST_BASE_REG, addr_regl); } - a_bits = get_alignment_bits(opc); - s_bits = opc & MO_SIZE; - /* - * R6 removes the left/right instructions but requires the - * system to support misaligned memory accesses. - */ if (use_mips32r6_instructions) { if (a_bits) { tcg_out_test_alignment(s, true, addr_regl, addr_regh, a_bits); @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_st_direct(TCGContext *s, TCGReg lo, TCGReg hi, } } -static void __attribute__((unused)) -tcg_out_qemu_st_unalign(TCGContext *s, TCGReg lo, TCGReg hi, +static void tcg_out_qemu_st_unalign(TCGContext *s, TCGReg lo, TCGReg hi, TCGReg base, MemOp opc) { const MIPSInsn sw1 = MIPS_BE ? OPC_SWL : OPC_SWR; @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, bool is_64) MemOp opc; #if defined(CONFIG_SOFTMMU) tcg_insn_unit *label_ptr[2]; -#else - unsigned a_bits, s_bits; #endif + unsigned a_bits, s_bits; TCGReg base = TCG_REG_A0; data_regl = *args++; @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, bool is_64) addr_regh = (TCG_TARGET_REG_BITS < TARGET_LONG_BITS ? *args++ : 0); oi = *args++; opc = get_memop(oi); + a_bits = get_alignment_bits(opc); + s_bits = opc & MO_SIZE; + /* + * R6 removes the left/right instructions but requires the + * system to support misaligned memory accesses. + */ #if defined(CONFIG_SOFTMMU) tcg_out_tlb_load(s, base, addr_regl, addr_regh, oi, label_ptr, 0); - tcg_out_qemu_st_direct(s, data_regl, data_regh, base, opc); + if (use_mips32r6_instructions || a_bits >= s_bits) { + tcg_out_qemu_st_direct(s, data_regl, data_regh, base, opc); + } else { + tcg_out_qemu_st_unalign(s, data_regl, data_regh, base, opc); + } add_qemu_ldst_label(s, 0, oi, (is_64 ? TCG_TYPE_I64 : TCG_TYPE_I32), data_regl, data_regh, addr_regl, addr_regh, @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, bool is_64) } else { tcg_out_opc_reg(s, ALIAS_PADD, base, TCG_GUEST_BASE_REG, addr_regl); } - a_bits = get_alignment_bits(opc); - s_bits = opc & MO_SIZE; - /* - * R6 removes the left/right instructions but requires the - * system to support misaligned memory accesses. - */ if (use_mips32r6_instructions) { if (a_bits) { tcg_out_test_alignment(s, true, addr_regl, addr_regh, a_bits); -- 2.25.1 When BH is constant, it is constrained to 11 bits for use in MOVCC. For the cases in which we must load the constant BH into a register, we do not need the full logic of tcg_out_movi; we can use the simpler function for emitting a 13 bit constant. This eliminates the only case in which TCG_REG_T2 was passed to tcg_out_movi, which will shortly become invalid. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/sparc/tcg-target.c.inc | 10 +++++++--- 1 file changed, 7 insertions(+), 3 deletions(-) diff --git a/tcg/sparc/tcg-target.c.inc b/tcg/sparc/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/sparc/tcg-target.c.inc +++ b/tcg/sparc/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ static void tcg_out_addsub2_i64(TCGContext *s, TCGReg rl, TCGReg rh, if (use_vis3_instructions && !is_sub) { /* Note that ADDXC doesn't accept immediates. */ if (bhconst && bh != 0) { - tcg_out_movi(s, TCG_TYPE_I64, TCG_REG_T2, bh); + tcg_out_movi_imm13(s, TCG_REG_T2, bh); bh = TCG_REG_T2; } tcg_out_arith(s, rh, ah, bh, ARITH_ADDXC); @@ -XXX,XX +XXX,XX @@ static void tcg_out_addsub2_i64(TCGContext *s, TCGReg rl, TCGReg rh, tcg_out_movcc(s, TCG_COND_GEU, MOVCC_XCC, rh, ah, 0); } } else { - /* Otherwise adjust BH as if there is carry into T2 ... */ + /* + * Otherwise adjust BH as if there is carry into T2. + * Note that constant BH is constrained to 11 bits for the MOVCC, + * so the adjustment fits 12 bits. + */ if (bhconst) { - tcg_out_movi(s, TCG_TYPE_I64, TCG_REG_T2, bh + (is_sub ? -1 : 1)); + tcg_out_movi_imm13(s, TCG_REG_T2, bh + (is_sub ? -1 : 1)); } else { tcg_out_arithi(s, TCG_REG_T2, bh, 1, is_sub ? ARITH_SUB : ARITH_ADD); -- 2.25.1 Handle 32-bit constants with a separate function, so that tcg_out_movi_int does not need to recurse. This slightly rearranges the order of tests for small constants, but produces the same output. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/sparc/tcg-target.c.inc | 36 +++++++++++++++++++++--------------- 1 file changed, 21 insertions(+), 15 deletions(-) diff --git a/tcg/sparc/tcg-target.c.inc b/tcg/sparc/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/sparc/tcg-target.c.inc +++ b/tcg/sparc/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ static void tcg_out_movi_imm13(TCGContext *s, TCGReg ret, int32_t arg) tcg_out_arithi(s, ret, TCG_REG_G0, arg, ARITH_OR); } +static void tcg_out_movi_imm32(TCGContext *s, TCGReg ret, int32_t arg) +{ + if (check_fit_i32(arg, 13)) { + /* A 13-bit constant sign-extended to 64-bits. */ + tcg_out_movi_imm13(s, ret, arg); + } else { + /* A 32-bit constant zero-extended to 64 bits. */ + tcg_out_sethi(s, ret, arg); + if (arg & 0x3ff) { + tcg_out_arithi(s, ret, ret, arg & 0x3ff, ARITH_OR); + } + } +} + static void tcg_out_movi_int(TCGContext *s, TCGType type, TCGReg ret, tcg_target_long arg, bool in_prologue) { tcg_target_long hi, lo = (int32_t)arg; tcg_target_long test, lsb; - /* Make sure we test 32-bit constants for imm13 properly. */ - if (type == TCG_TYPE_I32) { - arg = lo; + /* A 32-bit constant, or 32-bit zero-extended to 64-bits. */ + if (type == TCG_TYPE_I32 || arg == (uint32_t)arg) { + tcg_out_movi_imm32(s, ret, arg); + return; } /* A 13-bit constant sign-extended to 64-bits. */ @@ -XXX,XX +XXX,XX @@ static void tcg_out_movi_int(TCGContext *s, TCGType type, TCGReg ret, } } - /* A 32-bit constant, or 32-bit zero-extended to 64-bits. */ - if (type == TCG_TYPE_I32 || arg == (uint32_t)arg) { - tcg_out_sethi(s, ret, arg); - if (arg & 0x3ff) { - tcg_out_arithi(s, ret, ret, arg & 0x3ff, ARITH_OR); - } - return; - } - /* A 32-bit constant sign-extended to 64-bits. */ if (arg == lo) { tcg_out_sethi(s, ret, ~arg); @@ -XXX,XX +XXX,XX @@ static void tcg_out_movi_int(TCGContext *s, TCGType type, TCGReg ret, /* A 64-bit constant decomposed into 2 32-bit pieces. */ if (check_fit_i32(lo, 13)) { hi = (arg - lo) >> 32; - tcg_out_movi(s, TCG_TYPE_I32, ret, hi); + tcg_out_movi_imm32(s, ret, hi); tcg_out_arithi(s, ret, ret, 32, SHIFT_SLLX); tcg_out_arithi(s, ret, ret, lo, ARITH_ADD); } else { hi = arg >> 32; - tcg_out_movi(s, TCG_TYPE_I32, ret, hi); - tcg_out_movi(s, TCG_TYPE_I32, TCG_REG_T2, lo); + tcg_out_movi_imm32(s, ret, hi); + tcg_out_movi_imm32(s, TCG_REG_T2, lo); tcg_out_arithi(s, ret, ret, 32, SHIFT_SLLX); tcg_out_arith(s, ret, ret, TCG_REG_T2, ARITH_OR); } -- 2.25.1 This will allow us to control exactly what scratch register is used for loading the constant. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/sparc/tcg-target.c.inc | 15 +++++++++------ 1 file changed, 9 insertions(+), 6 deletions(-) diff --git a/tcg/sparc/tcg-target.c.inc b/tcg/sparc/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/sparc/tcg-target.c.inc +++ b/tcg/sparc/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ static void tcg_out_movi_imm32(TCGContext *s, TCGReg ret, int32_t arg) } static void tcg_out_movi_int(TCGContext *s, TCGType type, TCGReg ret, - tcg_target_long arg, bool in_prologue) + tcg_target_long arg, bool in_prologue, + TCGReg scratch) { tcg_target_long hi, lo = (int32_t)arg; tcg_target_long test, lsb; @@ -XXX,XX +XXX,XX @@ static void tcg_out_movi_int(TCGContext *s, TCGType type, TCGReg ret, } else { hi = arg >> 32; tcg_out_movi_imm32(s, ret, hi); - tcg_out_movi_imm32(s, TCG_REG_T2, lo); + tcg_out_movi_imm32(s, scratch, lo); tcg_out_arithi(s, ret, ret, 32, SHIFT_SLLX); - tcg_out_arith(s, ret, ret, TCG_REG_T2, ARITH_OR); + tcg_out_arith(s, ret, ret, scratch, ARITH_OR); } } static void tcg_out_movi(TCGContext *s, TCGType type, TCGReg ret, tcg_target_long arg) { - tcg_out_movi_int(s, type, ret, arg, false); + tcg_debug_assert(ret != TCG_REG_T2); + tcg_out_movi_int(s, type, ret, arg, false, TCG_REG_T2); } static void tcg_out_ldst_rr(TCGContext *s, TCGReg data, TCGReg a1, @@ -XXX,XX +XXX,XX @@ static void tcg_out_call_nodelay(TCGContext *s, const tcg_insn_unit *dest, } else { uintptr_t desti = (uintptr_t)dest; tcg_out_movi_int(s, TCG_TYPE_PTR, TCG_REG_T1, - desti & ~0xfff, in_prologue); + desti & ~0xfff, in_prologue, TCG_REG_O7); tcg_out_arithi(s, TCG_REG_O7, TCG_REG_T1, desti & 0xfff, JMPL); } } @@ -XXX,XX +XXX,XX @@ static void tcg_target_qemu_prologue(TCGContext *s) #ifndef CONFIG_SOFTMMU if (guest_base != 0) { - tcg_out_movi_int(s, TCG_TYPE_PTR, TCG_GUEST_BASE_REG, guest_base, true); + tcg_out_movi_int(s, TCG_TYPE_PTR, TCG_GUEST_BASE_REG, + guest_base, true, TCG_REG_T1); tcg_regset_set_reg(s->reserved_regs, TCG_GUEST_BASE_REG); } #endif -- 2.25.1 We had code for checking for 13 and 21-bit shifted constants, but we can do better and allow 32-bit shifted constants. This is still 2 insns shorter than the full 64-bit sequence. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/sparc/tcg-target.c.inc | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) diff --git a/tcg/sparc/tcg-target.c.inc b/tcg/sparc/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/sparc/tcg-target.c.inc +++ b/tcg/sparc/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ static void tcg_out_movi_int(TCGContext *s, TCGType type, TCGReg ret, return; } - /* A 21-bit constant, shifted. */ + /* A 32-bit constant, shifted. */ lsb = ctz64(arg); test = (tcg_target_long)arg >> lsb; - if (check_fit_tl(test, 13)) { - tcg_out_movi_imm13(s, ret, test); - tcg_out_arithi(s, ret, ret, lsb, SHIFT_SLLX); - return; - } else if (lsb > 10 && test == extract64(test, 0, 21)) { + if (lsb > 10 && test == extract64(test, 0, 21)) { tcg_out_sethi(s, ret, test << 10); tcg_out_arithi(s, ret, ret, lsb - 10, SHIFT_SLLX); return; + } else if (test == (uint32_t)test || test == (int32_t)test) { + tcg_out_movi_int(s, TCG_TYPE_I64, ret, test, in_prologue, scratch); + tcg_out_arithi(s, ret, ret, lsb, SHIFT_SLLX); + return; } /* A 64-bit constant decomposed into 2 32-bit pieces. */ -- 2.25.1 Since 7ecd02a06f8, if patch_reloc fails we restart translation with a smaller TB. SPARC had its function signature changed, but not the logic. Replace assert with return false. Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/sparc/tcg-target.c.inc | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) diff --git a/tcg/sparc/tcg-target.c.inc b/tcg/sparc/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/sparc/tcg-target.c.inc +++ b/tcg/sparc/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ static bool patch_reloc(tcg_insn_unit *src_rw, int type, switch (type) { case R_SPARC_WDISP16: - assert(check_fit_ptr(pcrel >> 2, 16)); + if (!check_fit_ptr(pcrel >> 2, 16)) { + return false; + } insn &= ~INSN_OFF16(-1); insn |= INSN_OFF16(pcrel); break; case R_SPARC_WDISP19: - assert(check_fit_ptr(pcrel >> 2, 19)); + if (!check_fit_ptr(pcrel >> 2, 19)) { + return false; + } insn &= ~INSN_OFF19(-1); insn |= INSN_OFF19(pcrel); break; -- 2.25.1 Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/sparc/tcg-target.c.inc | 15 +++++++++++++++ 1 file changed, 15 insertions(+) diff --git a/tcg/sparc/tcg-target.c.inc b/tcg/sparc/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/sparc/tcg-target.c.inc +++ b/tcg/sparc/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ static bool patch_reloc(tcg_insn_unit *src_rw, int type, insn &= ~INSN_OFF19(-1); insn |= INSN_OFF19(pcrel); break; + case R_SPARC_13: + if (!check_fit_ptr(value, 13)) { + return false; + } + insn &= ~INSN_IMM13(-1); + insn |= INSN_IMM13(value); + break; default: g_assert_not_reached(); } @@ -XXX,XX +XXX,XX @@ static void tcg_out_movi_int(TCGContext *s, TCGType type, TCGReg ret, return; } + /* Use the constant pool, if possible. */ + if (!in_prologue && USE_REG_TB) { + new_pool_label(s, arg, R_SPARC_13, s->code_ptr, + tcg_tbrel_diff(s, NULL)); + tcg_out32(s, LDX | INSN_RD(ret) | INSN_RS1(TCG_REG_TB)); + return; + } + /* A 64-bit constant decomposed into 2 32-bit pieces. */ if (check_fit_i32(lo, 13)) { hi = (arg - lo) >> 32; -- 2.25.1 Due to mapping changes, we now rarely place the code_gen_buffer near the main executable. Which means that direct calls will now rarely be in range. So, always use indirect calls for tail calls, which allows us to avoid clobbering %o7, and therefore we need not save and restore it. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/sparc/tcg-target.c.inc | 37 +++++++++++++++++++++++-------------- 1 file changed, 23 insertions(+), 14 deletions(-) diff --git a/tcg/sparc/tcg-target.c.inc b/tcg/sparc/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/sparc/tcg-target.c.inc +++ b/tcg/sparc/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ static void tcg_out_addsub2_i64(TCGContext *s, TCGReg rl, TCGReg rh, tcg_out_mov(s, TCG_TYPE_I64, rl, tmp); } +static void tcg_out_jmpl_const(TCGContext *s, const tcg_insn_unit *dest, + bool in_prologue, bool tail_call) +{ + uintptr_t desti = (uintptr_t)dest; + + /* Be careful not to clobber %o7 for a tail call. */ + tcg_out_movi_int(s, TCG_TYPE_PTR, TCG_REG_T1, + desti & ~0xfff, in_prologue, + tail_call ? TCG_REG_G2 : TCG_REG_O7); + tcg_out_arithi(s, tail_call ? TCG_REG_G0 : TCG_REG_O7, + TCG_REG_T1, desti & 0xfff, JMPL); +} + static void tcg_out_call_nodelay(TCGContext *s, const tcg_insn_unit *dest, bool in_prologue) { @@ -XXX,XX +XXX,XX @@ static void tcg_out_call_nodelay(TCGContext *s, const tcg_insn_unit *dest, if (disp == (int32_t)disp) { tcg_out32(s, CALL | (uint32_t)disp >> 2); } else { - uintptr_t desti = (uintptr_t)dest; - tcg_out_movi_int(s, TCG_TYPE_PTR, TCG_REG_T1, - desti & ~0xfff, in_prologue, TCG_REG_O7); - tcg_out_arithi(s, TCG_REG_O7, TCG_REG_T1, desti & 0xfff, JMPL); + tcg_out_jmpl_const(s, dest, in_prologue, false); } } @@ -XXX,XX +XXX,XX @@ static void build_trampolines(TCGContext *s) /* Set the retaddr operand. */ tcg_out_mov(s, TCG_TYPE_PTR, ra, TCG_REG_O7); - /* Set the env operand. */ - tcg_out_mov(s, TCG_TYPE_PTR, TCG_REG_O0, TCG_AREG0); /* Tail call. */ - tcg_out_call_nodelay(s, qemu_ld_helpers[i], true); - tcg_out_mov(s, TCG_TYPE_PTR, TCG_REG_O7, ra); + tcg_out_jmpl_const(s, qemu_ld_helpers[i], true, true); + /* delay slot -- set the env argument */ + tcg_out_mov_delay(s, TCG_REG_O0, TCG_AREG0); } for (i = 0; i < ARRAY_SIZE(qemu_st_helpers); ++i) { @@ -XXX,XX +XXX,XX @@ static void build_trampolines(TCGContext *s) if (ra >= TCG_REG_O6) { tcg_out_st(s, TCG_TYPE_PTR, TCG_REG_O7, TCG_REG_CALL_STACK, TCG_TARGET_CALL_STACK_OFFSET); - ra = TCG_REG_G1; + } else { + tcg_out_mov(s, TCG_TYPE_PTR, ra, TCG_REG_O7); } - tcg_out_mov(s, TCG_TYPE_PTR, ra, TCG_REG_O7); - /* Set the env operand. */ - tcg_out_mov(s, TCG_TYPE_PTR, TCG_REG_O0, TCG_AREG0); + /* Tail call. */ - tcg_out_call_nodelay(s, qemu_st_helpers[i], true); - tcg_out_mov(s, TCG_TYPE_PTR, TCG_REG_O7, ra); + tcg_out_jmpl_const(s, qemu_st_helpers[i], true, true); + /* delay slot -- set the env argument */ + tcg_out_mov_delay(s, TCG_REG_O0, TCG_AREG0); } } #endif -- 2.25.1 This is kinda sorta the opposite of the other tcg hosts, where we get (normal) alignment checks for free with host SIGBUS and need to add code to support unaligned accesses. This inline code expansion is somewhat large, but it takes quite a few instructions to make a function call to a helper anyway. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/sparc/tcg-target.c.inc | 219 +++++++++++++++++++++++++++++++++++-- 1 file changed, 211 insertions(+), 8 deletions(-) diff --git a/tcg/sparc/tcg-target.c.inc b/tcg/sparc/tcg-target.c.inc index XXXXXXX..XXXXXXX 100644 --- a/tcg/sparc/tcg-target.c.inc +++ b/tcg/sparc/tcg-target.c.inc @@ -XXX,XX +XXX,XX @@ static const int tcg_target_call_oarg_regs[] = { #define ARITH_ADD (INSN_OP(2) | INSN_OP3(0x00)) #define ARITH_ADDCC (INSN_OP(2) | INSN_OP3(0x10)) #define ARITH_AND (INSN_OP(2) | INSN_OP3(0x01)) +#define ARITH_ANDCC (INSN_OP(2) | INSN_OP3(0x11)) #define ARITH_ANDN (INSN_OP(2) | INSN_OP3(0x05)) #define ARITH_OR (INSN_OP(2) | INSN_OP3(0x02)) #define ARITH_ORCC (INSN_OP(2) | INSN_OP3(0x12)) @@ -XXX,XX +XXX,XX @@ static void build_trampolines(TCGContext *s) tcg_out_mov_delay(s, TCG_REG_O0, TCG_AREG0); } } +#else +static const tcg_insn_unit *qemu_unalign_ld_trampoline; +static const tcg_insn_unit *qemu_unalign_st_trampoline; + +static void build_trampolines(TCGContext *s) +{ + for (int ld = 0; ld < 2; ++ld) { + void *helper; + + while ((uintptr_t)s->code_ptr & 15) { + tcg_out_nop(s); + } + + if (ld) { + helper = helper_unaligned_ld; + qemu_unalign_ld_trampoline = tcg_splitwx_to_rx(s->code_ptr); + } else { + helper = helper_unaligned_st; + qemu_unalign_st_trampoline = tcg_splitwx_to_rx(s->code_ptr); + } + + if (!SPARC64 && TARGET_LONG_BITS == 64) { + /* Install the high part of the address. */ + tcg_out_arithi(s, TCG_REG_O1, TCG_REG_O2, 32, SHIFT_SRLX); + } + + /* Tail call. */ + tcg_out_jmpl_const(s, helper, true, true); + /* delay slot -- set the env argument */ + tcg_out_mov_delay(s, TCG_REG_O0, TCG_AREG0); + } +} #endif /* Generate global QEMU prologue and epilogue code */ @@ -XXX,XX +XXX,XX @@ static void tcg_target_qemu_prologue(TCGContext *s) /* delay slot */ tcg_out_movi_imm13(s, TCG_REG_O0, 0); -#ifdef CONFIG_SOFTMMU build_trampolines(s); -#endif } static void tcg_out_nop_fill(tcg_insn_unit *p, int count) @@ -XXX,XX +XXX,XX @@ static TCGReg tcg_out_tlb_load(TCGContext *s, TCGReg addr, int mem_index, static const int qemu_ld_opc[(MO_SSIZE | MO_BSWAP) + 1] = { [MO_UB] = LDUB, [MO_SB] = LDSB, + [MO_UB | MO_LE] = LDUB, + [MO_SB | MO_LE] = LDSB, [MO_BEUW] = LDUH, [MO_BESW] = LDSH, [MO_BEUL] = LDUW, [MO_BESL] = LDSW, [MO_BEUQ] = LDX, + [MO_BESQ] = LDX, [MO_LEUW] = LDUH_LE, [MO_LESW] = LDSH_LE, [MO_LEUL] = LDUW_LE, [MO_LESL] = LDSW_LE, [MO_LEUQ] = LDX_LE, + [MO_LESQ] = LDX_LE, }; static const int qemu_st_opc[(MO_SIZE | MO_BSWAP) + 1] = { @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_ld(TCGContext *s, TCGReg data, TCGReg addr, MemOpIdx oi, bool is_64) { MemOp memop = get_memop(oi); + tcg_insn_unit *label_ptr; + #ifdef CONFIG_SOFTMMU unsigned memi = get_mmuidx(oi); TCGReg addrz, param; const tcg_insn_unit *func; - tcg_insn_unit *label_ptr; addrz = tcg_out_tlb_load(s, addr, memi, memop, offsetof(CPUTLBEntry, addr_read)); @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_ld(TCGContext *s, TCGReg data, TCGReg addr, *label_ptr |= INSN_OFF19(tcg_ptr_byte_diff(s->code_ptr, label_ptr)); #else + TCGReg index = (guest_base ? TCG_GUEST_BASE_REG : TCG_REG_G0); + unsigned a_bits = get_alignment_bits(memop); + unsigned s_bits = memop & MO_SIZE; + unsigned t_bits; + if (SPARC64 && TARGET_LONG_BITS == 32) { tcg_out_arithi(s, TCG_REG_T1, addr, 0, SHIFT_SRL); addr = TCG_REG_T1; } - tcg_out_ldst_rr(s, data, addr, - (guest_base ? TCG_GUEST_BASE_REG : TCG_REG_G0), + + /* + * Normal case: alignment equal to access size. + */ + if (a_bits == s_bits) { + tcg_out_ldst_rr(s, data, addr, index, + qemu_ld_opc[memop & (MO_BSWAP | MO_SSIZE)]); + return; + } + + /* + * Test for at least natural alignment, and assume most accesses + * will be aligned -- perform a straight load in the delay slot. + * This is required to preserve atomicity for aligned accesses. + */ + t_bits = MAX(a_bits, s_bits); + tcg_debug_assert(t_bits < 13); + tcg_out_arithi(s, TCG_REG_G0, addr, (1u << t_bits) - 1, ARITH_ANDCC); + + /* beq,a,pt %icc, label */ + label_ptr = s->code_ptr; + tcg_out_bpcc0(s, COND_E, BPCC_A | BPCC_PT | BPCC_ICC, 0); + /* delay slot */ + tcg_out_ldst_rr(s, data, addr, index, qemu_ld_opc[memop & (MO_BSWAP | MO_SSIZE)]); + + if (a_bits >= s_bits) { + /* + * Overalignment: A successful alignment test will perform the memory + * operation in the delay slot, and failure need only invoke the + * handler for SIGBUS. + */ + TCGReg arg_low = TCG_REG_O1 + (!SPARC64 && TARGET_LONG_BITS == 64); + tcg_out_call_nodelay(s, qemu_unalign_ld_trampoline, false); + /* delay slot -- move to low part of argument reg */ + tcg_out_mov_delay(s, arg_low, addr); + } else { + /* Underalignment: load by pieces of minimum alignment. */ + int ld_opc, a_size, s_size, i; + + /* + * Force full address into T1 early; avoids problems with + * overlap between @addr and @data. + */ + tcg_out_arith(s, TCG_REG_T1, addr, index, ARITH_ADD); + + a_size = 1 << a_bits; + s_size = 1 << s_bits; + if ((memop & MO_BSWAP) == MO_BE) { + ld_opc = qemu_ld_opc[a_bits | MO_BE | (memop & MO_SIGN)]; + tcg_out_ldst(s, data, TCG_REG_T1, 0, ld_opc); + ld_opc = qemu_ld_opc[a_bits | MO_BE]; + for (i = a_size; i < s_size; i += a_size) { + tcg_out_ldst(s, TCG_REG_T2, TCG_REG_T1, i, ld_opc); + tcg_out_arithi(s, data, data, a_size, SHIFT_SLLX); + tcg_out_arith(s, data, data, TCG_REG_T2, ARITH_OR); + } + } else if (a_bits == 0) { + ld_opc = LDUB; + tcg_out_ldst(s, data, TCG_REG_T1, 0, ld_opc); + for (i = a_size; i < s_size; i += a_size) { + if ((memop & MO_SIGN) && i == s_size - a_size) { + ld_opc = LDSB; + } + tcg_out_ldst(s, TCG_REG_T2, TCG_REG_T1, i, ld_opc); + tcg_out_arithi(s, TCG_REG_T2, TCG_REG_T2, i * 8, SHIFT_SLLX); + tcg_out_arith(s, data, data, TCG_REG_T2, ARITH_OR); + } + } else { + ld_opc = qemu_ld_opc[a_bits | MO_LE]; + tcg_out_ldst_rr(s, data, TCG_REG_T1, TCG_REG_G0, ld_opc); + for (i = a_size; i < s_size; i += a_size) { + tcg_out_arithi(s, TCG_REG_T1, TCG_REG_T1, a_size, ARITH_ADD); + if ((memop & MO_SIGN) && i == s_size - a_size) { + ld_opc = qemu_ld_opc[a_bits | MO_LE | MO_SIGN]; + } + tcg_out_ldst_rr(s, TCG_REG_T2, TCG_REG_T1, TCG_REG_G0, ld_opc); + tcg_out_arithi(s, TCG_REG_T2, TCG_REG_T2, i * 8, SHIFT_SLLX); + tcg_out_arith(s, data, data, TCG_REG_T2, ARITH_OR); + } + } + } + + *label_ptr |= INSN_OFF19(tcg_ptr_byte_diff(s->code_ptr, label_ptr)); #endif /* CONFIG_SOFTMMU */ } @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_st(TCGContext *s, TCGReg data, TCGReg addr, MemOpIdx oi) { MemOp memop = get_memop(oi); + tcg_insn_unit *label_ptr; + #ifdef CONFIG_SOFTMMU unsigned memi = get_mmuidx(oi); TCGReg addrz, param; const tcg_insn_unit *func; - tcg_insn_unit *label_ptr; addrz = tcg_out_tlb_load(s, addr, memi, memop, offsetof(CPUTLBEntry, addr_write)); @@ -XXX,XX +XXX,XX @@ static void tcg_out_qemu_st(TCGContext *s, TCGReg data, TCGReg addr, *label_ptr |= INSN_OFF19(tcg_ptr_byte_diff(s->code_ptr, label_ptr)); #else + TCGReg index = (guest_base ? TCG_GUEST_BASE_REG : TCG_REG_G0); + unsigned a_bits = get_alignment_bits(memop); + unsigned s_bits = memop & MO_SIZE; + unsigned t_bits; + if (SPARC64 && TARGET_LONG_BITS == 32) { tcg_out_arithi(s, TCG_REG_T1, addr, 0, SHIFT_SRL); addr = TCG_REG_T1; } - tcg_out_ldst_rr(s, data, addr, - (guest_base ? TCG_GUEST_BASE_REG : TCG_REG_G0), + + /* + * Normal case: alignment equal to access size. + */ + if (a_bits == s_bits) { + tcg_out_ldst_rr(s, data, addr, index, + qemu_st_opc[memop & (MO_BSWAP | MO_SIZE)]); + return; + } + + /* + * Test for at least natural alignment, and assume most accesses + * will be aligned -- perform a straight store in the delay slot. + * This is required to preserve atomicity for aligned accesses. + */ + t_bits = MAX(a_bits, s_bits); + tcg_debug_assert(t_bits < 13); + tcg_out_arithi(s, TCG_REG_G0, addr, (1u << t_bits) - 1, ARITH_ANDCC); + + /* beq,a,pt %icc, label */ + label_ptr = s->code_ptr; + tcg_out_bpcc0(s, COND_E, BPCC_A | BPCC_PT | BPCC_ICC, 0); + /* delay slot */ + tcg_out_ldst_rr(s, data, addr, index, qemu_st_opc[memop & (MO_BSWAP | MO_SIZE)]); + + if (a_bits >= s_bits) { + /* + * Overalignment: A successful alignment test will perform the memory + * operation in the delay slot, and failure need only invoke the + * handler for SIGBUS. + */ + TCGReg arg_low = TCG_REG_O1 + (!SPARC64 && TARGET_LONG_BITS == 64); + tcg_out_call_nodelay(s, qemu_unalign_st_trampoline, false); + /* delay slot -- move to low part of argument reg */ + tcg_out_mov_delay(s, arg_low, addr); + } else { + /* Underalignment: store by pieces of minimum alignment. */ + int st_opc, a_size, s_size, i; + + /* + * Force full address into T1 early; avoids problems with + * overlap between @addr and @data. + */ + tcg_out_arith(s, TCG_REG_T1, addr, index, ARITH_ADD); + + a_size = 1 << a_bits; + s_size = 1 << s_bits; + if ((memop & MO_BSWAP) == MO_BE) { + st_opc = qemu_st_opc[a_bits | MO_BE]; + for (i = 0; i < s_size; i += a_size) { + TCGReg d = data; + int shift = (s_size - a_size - i) * 8; + if (shift) { + d = TCG_REG_T2; + tcg_out_arithi(s, d, data, shift, SHIFT_SRLX); + } + tcg_out_ldst(s, d, TCG_REG_T1, i, st_opc); + } + } else if (a_bits == 0) { + tcg_out_ldst(s, data, TCG_REG_T1, 0, STB); + for (i = 1; i < s_size; i++) { + tcg_out_arithi(s, TCG_REG_T2, data, i * 8, SHIFT_SRLX); + tcg_out_ldst(s, TCG_REG_T2, TCG_REG_T1, i, STB); + } + } else { + /* Note that ST*A with immediate asi must use indexed address. */ + st_opc = qemu_st_opc[a_bits + MO_LE]; + tcg_out_ldst_rr(s, data, TCG_REG_T1, TCG_REG_G0, st_opc); + for (i = a_size; i < s_size; i += a_size) { + tcg_out_arithi(s, TCG_REG_T2, data, i * 8, SHIFT_SRLX); + tcg_out_arithi(s, TCG_REG_T1, TCG_REG_T1, a_size, ARITH_ADD); + tcg_out_ldst_rr(s, TCG_REG_T2, TCG_REG_T1, TCG_REG_G0, st_opc); + } + } + } + + *label_ptr |= INSN_OFF19(tcg_ptr_byte_diff(s->code_ptr, label_ptr)); #endif /* CONFIG_SOFTMMU */ } -- 2.25.1 A mostly generic test for unaligned access raising SIGBUS. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tests/tcg/multiarch/sigbus.c | 68 ++++++++++++++++++++++++++++++++++++ 1 file changed, 68 insertions(+) create mode 100644 tests/tcg/multiarch/sigbus.c diff --git a/tests/tcg/multiarch/sigbus.c b/tests/tcg/multiarch/sigbus.c new file mode 100644 index XXXXXXX..XXXXXXX --- /dev/null +++ b/tests/tcg/multiarch/sigbus.c @@ -XXX,XX +XXX,XX @@ +#define _GNU_SOURCE 1 + +#include <assert.h> +#include <stdlib.h> +#include <signal.h> +#include <endian.h> + + +unsigned long long x = 0x8877665544332211ull; +void * volatile p = (void *)&x + 1; + +void sigbus(int sig, siginfo_t *info, void *uc) +{ + assert(sig == SIGBUS); + assert(info->si_signo == SIGBUS); +#ifdef BUS_ADRALN + assert(info->si_code == BUS_ADRALN); +#endif + assert(info->si_addr == p); + exit(EXIT_SUCCESS); +} + +int main() +{ + struct sigaction sa = { + .sa_sigaction = sigbus, + .sa_flags = SA_SIGINFO + }; + int allow_fail = 0; + int tmp; + + tmp = sigaction(SIGBUS, &sa, NULL); + assert(tmp == 0); + + /* + * Select an operation that's likely to enforce alignment. + * On many guests that support unaligned accesses by default, + * this is often an atomic operation. + */ +#if defined(__aarch64__) + asm volatile("ldxr %w0,[%1]" : "=r"(tmp) : "r"(p) : "memory"); +#elif defined(__alpha__) + asm volatile("ldl_l %0,0(%1)" : "=r"(tmp) : "r"(p) : "memory"); +#elif defined(__arm__) + asm volatile("ldrex %0,[%1]" : "=r"(tmp) : "r"(p) : "memory"); +#elif defined(__powerpc__) + asm volatile("lwarx %0,0,%1" : "=r"(tmp) : "r"(p) : "memory"); +#elif defined(__riscv_atomic) + asm volatile("lr.w %0,(%1)" : "=r"(tmp) : "r"(p) : "memory"); +#else + /* No insn known to fault unaligned -- try for a straight load. */ + allow_fail = 1; + tmp = *(volatile int *)p; +#endif + + assert(allow_fail); + + /* + * We didn't see a signal. + * We might as well validate the unaligned load worked. + */ + if (BYTE_ORDER == LITTLE_ENDIAN) { + assert(tmp == 0x55443322); + } else { + assert(tmp == 0x77665544); + } + return EXIT_SUCCESS; +} -- 2.25.1 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