This is mostly my code_gen_buffer cleanup, plus a few other random changes thrown in. Including a fix for a recent float32_exp2 bug. r~ The following changes since commit 894fc4fd670aaf04a67dc7507739f914ff4bacf2: Merge remote-tracking branch 'remotes/jasowang/tags/net-pull-request' into staging (2021-06-11 09:21:48 +0100) are available in the Git repository at: https://gitlab.com/rth7680/qemu.git tags/pull-tcg-20210611 for you to fetch changes up to 60afaddc208d34f6dc86dd974f6e02724fba6eb6: docs/devel: Explain in more detail the TB chaining mechanisms (2021-06-11 09:41:25 -0700) ---------------------------------------------------------------- Clean up code_gen_buffer allocation. Add tcg_remove_ops_after. Fix tcg_constant_* documentation. Improve TB chaining documentation. Fix float32_exp2. ---------------------------------------------------------------- Jose R. Ziviani (1): tcg/arm: Fix tcg_out_op function signature Luis Pires (1): docs/devel: Explain in more detail the TB chaining mechanisms Richard Henderson (32): meson: Split out tcg/meson.build meson: Split out fpu/meson.build tcg: Re-order tcg_region_init vs tcg_prologue_init tcg: Remove error return from tcg_region_initial_alloc__locked tcg: Split out tcg_region_initial_alloc tcg: Split out tcg_region_prologue_set tcg: Split out region.c accel/tcg: Inline cpu_gen_init accel/tcg: Move alloc_code_gen_buffer to tcg/region.c accel/tcg: Rename tcg_init to tcg_init_machine tcg: Create tcg_init accel/tcg: Merge tcg_exec_init into tcg_init_machine accel/tcg: Use MiB in tcg_init_machine accel/tcg: Pass down max_cpus to tcg_init tcg: Introduce tcg_max_ctxs tcg: Move MAX_CODE_GEN_BUFFER_SIZE to tcg-target.h tcg: Replace region.end with region.total_size tcg: Rename region.start to region.after_prologue tcg: Tidy tcg_n_regions tcg: Tidy split_cross_256mb tcg: Move in_code_gen_buffer and tests to region.c tcg: Allocate code_gen_buffer into struct tcg_region_state tcg: Return the map protection from alloc_code_gen_buffer tcg: Sink qemu_madvise call to common code util/osdep: Add qemu_mprotect_rw tcg: Round the tb_size default from qemu_get_host_physmem tcg: Merge buffer protection and guard page protection tcg: When allocating for !splitwx, begin with PROT_NONE tcg: Move tcg_init_ctx and tcg_ctx from accel/tcg/ tcg: Introduce tcg_remove_ops_after tcg: Fix documentation for tcg_constant_* vs tcg_temp_free_* softfloat: Fix tp init in float32_exp2 docs/devel/tcg.rst | 101 ++++- meson.build | 12 +- accel/tcg/internal.h | 2 + include/qemu/osdep.h | 1 + include/sysemu/tcg.h | 2 - include/tcg/tcg.h | 28 +- tcg/aarch64/tcg-target.h | 1 + tcg/arm/tcg-target.h | 1 + tcg/i386/tcg-target.h | 2 + tcg/mips/tcg-target.h | 6 + tcg/ppc/tcg-target.h | 2 + tcg/riscv/tcg-target.h | 1 + tcg/s390/tcg-target.h | 3 + tcg/sparc/tcg-target.h | 1 + tcg/tcg-internal.h | 40 ++ tcg/tci/tcg-target.h | 1 + accel/tcg/tcg-all.c | 32 +- accel/tcg/translate-all.c | 439 +------------------- bsd-user/main.c | 3 +- fpu/softfloat.c | 2 +- linux-user/main.c | 1 - tcg/region.c | 999 ++++++++++++++++++++++++++++++++++++++++++++++ tcg/tcg.c | 649 +++--------------------------- util/osdep.c | 9 + tcg/arm/tcg-target.c.inc | 3 +- fpu/meson.build | 1 + tcg/meson.build | 14 + 27 files changed, 1266 insertions(+), 1090 deletions(-) create mode 100644 tcg/tcg-internal.h create mode 100644 tcg/region.c create mode 100644 fpu/meson.build create mode 100644 tcg/meson.build Reviewed-by: Luis Pires <luis.pires@eldorado.org.br> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- meson.build | 8 +------- tcg/meson.build | 13 +++++++++++++ 2 files changed, 14 insertions(+), 7 deletions(-) create mode 100644 tcg/meson.build diff --git a/meson.build b/meson.build index XXXXXXX..XXXXXXX 100644 --- a/meson.build +++ b/meson.build @@ -XXX,XX +XXX,XX @@ common_ss.add(capstone) specific_ss.add(files('cpu.c', 'disas.c', 'gdbstub.c'), capstone) specific_ss.add(when: 'CONFIG_TCG', if_true: files( 'fpu/softfloat.c', - 'tcg/optimize.c', - 'tcg/tcg-common.c', - 'tcg/tcg-op-gvec.c', - 'tcg/tcg-op-vec.c', - 'tcg/tcg-op.c', - 'tcg/tcg.c', )) -specific_ss.add(when: 'CONFIG_TCG_INTERPRETER', if_true: files('tcg/tci.c')) # Work around a gcc bug/misfeature wherein constant propagation looks # through an alias: @@ -XXX,XX +XXX,XX @@ subdir('net') subdir('replay') subdir('semihosting') subdir('hw') +subdir('tcg') subdir('accel') subdir('plugins') subdir('bsd-user') diff --git a/tcg/meson.build b/tcg/meson.build new file mode 100644 index XXXXXXX..XXXXXXX --- /dev/null +++ b/tcg/meson.build @@ -XXX,XX +XXX,XX @@ +tcg_ss = ss.source_set() + +tcg_ss.add(files( + 'optimize.c', + 'tcg.c', + 'tcg-common.c', + 'tcg-op.c', + 'tcg-op-gvec.c', + 'tcg-op-vec.c', +)) +tcg_ss.add(when: 'CONFIG_TCG_INTERPRETER', if_true: files('tci.c')) + +specific_ss.add_all(when: 'CONFIG_TCG', if_true: tcg_ss) -- 2.25.1 Reviewed-by: Luis Pires <luis.pires@eldorado.org.br> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- meson.build | 4 +--- fpu/meson.build | 1 + 2 files changed, 2 insertions(+), 3 deletions(-) create mode 100644 fpu/meson.build diff --git a/meson.build b/meson.build index XXXXXXX..XXXXXXX 100644 --- a/meson.build +++ b/meson.build @@ -XXX,XX +XXX,XX @@ subdir('softmmu') common_ss.add(capstone) specific_ss.add(files('cpu.c', 'disas.c', 'gdbstub.c'), capstone) -specific_ss.add(when: 'CONFIG_TCG', if_true: files( - 'fpu/softfloat.c', -)) # Work around a gcc bug/misfeature wherein constant propagation looks # through an alias: @@ -XXX,XX +XXX,XX @@ subdir('replay') subdir('semihosting') subdir('hw') subdir('tcg') +subdir('fpu') subdir('accel') subdir('plugins') subdir('bsd-user') diff --git a/fpu/meson.build b/fpu/meson.build new file mode 100644 index XXXXXXX..XXXXXXX --- /dev/null +++ b/fpu/meson.build @@ -0,0 +1 @@ +specific_ss.add(when: 'CONFIG_TCG', if_true: files('softfloat.c')) -- 2.25.1 Instead of delaying tcg_region_init until after tcg_prologue_init is complete, do tcg_region_init first and let tcg_prologue_init shrink the first region by the size of the generated prologue. Reviewed-by: Luis Pires <luis.pires@eldorado.org.br> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- accel/tcg/tcg-all.c | 11 --------- accel/tcg/translate-all.c | 3 +++ bsd-user/main.c | 1 - linux-user/main.c | 1 - tcg/tcg.c | 52 ++++++++++++++------------------------- 5 files changed, 22 insertions(+), 46 deletions(-) diff --git a/accel/tcg/tcg-all.c b/accel/tcg/tcg-all.c index XXXXXXX..XXXXXXX 100644 --- a/accel/tcg/tcg-all.c +++ b/accel/tcg/tcg-all.c @@ -XXX,XX +XXX,XX @@ static int tcg_init(MachineState *ms) tcg_exec_init(s->tb_size * 1024 * 1024, s->splitwx_enabled); mttcg_enabled = s->mttcg_enabled; - - /* - * Initialize TCG regions only for softmmu. - * - * This needs to be done later for user mode, because the prologue - * generation needs to be delayed so that GUEST_BASE is already set. - */ -#ifndef CONFIG_USER_ONLY - tcg_region_init(); -#endif /* !CONFIG_USER_ONLY */ - return 0; } 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 @@ void tcg_exec_init(unsigned long tb_size, int splitwx) splitwx, &error_fatal); assert(ok); + /* TODO: allocating regions is hand-in-glove with code_gen_buffer. */ + tcg_region_init(); + #if defined(CONFIG_SOFTMMU) /* There's no guest base to take into account, so go ahead and initialize the prologue now. */ diff --git a/bsd-user/main.c b/bsd-user/main.c index XXXXXXX..XXXXXXX 100644 --- a/bsd-user/main.c +++ b/bsd-user/main.c @@ -XXX,XX +XXX,XX @@ int main(int argc, char **argv) * the real value of GUEST_BASE into account. */ tcg_prologue_init(tcg_ctx); - tcg_region_init(); /* build Task State */ memset(ts, 0, sizeof(TaskState)); diff --git a/linux-user/main.c b/linux-user/main.c index XXXXXXX..XXXXXXX 100644 --- a/linux-user/main.c +++ b/linux-user/main.c @@ -XXX,XX +XXX,XX @@ int main(int argc, char **argv, char **envp) generating the prologue until now so that the prologue can take the real value of GUEST_BASE into account. */ tcg_prologue_init(tcg_ctx); - tcg_region_init(); target_cpu_copy_regs(env, regs); diff --git a/tcg/tcg.c b/tcg/tcg.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/tcg.c +++ b/tcg/tcg.c @@ -XXX,XX +XXX,XX @@ TranslationBlock *tcg_tb_alloc(TCGContext *s) void tcg_prologue_init(TCGContext *s) { - size_t prologue_size, total_size; - void *buf0, *buf1; + size_t prologue_size; /* Put the prologue at the beginning of code_gen_buffer. */ - buf0 = s->code_gen_buffer; - total_size = s->code_gen_buffer_size; - s->code_ptr = buf0; - s->code_buf = buf0; + tcg_region_assign(s, 0); + s->code_ptr = s->code_gen_ptr; + s->code_buf = s->code_gen_ptr; s->data_gen_ptr = NULL; - /* - * The region trees are not yet configured, but tcg_splitwx_to_rx - * needs the bounds for an assert. - */ - region.start = buf0; - region.end = buf0 + total_size; - #ifndef CONFIG_TCG_INTERPRETER - tcg_qemu_tb_exec = (tcg_prologue_fn *)tcg_splitwx_to_rx(buf0); + tcg_qemu_tb_exec = (tcg_prologue_fn *)tcg_splitwx_to_rx(s->code_ptr); #endif - /* Compute a high-water mark, at which we voluntarily flush the buffer - and start over. The size here is arbitrary, significantly larger - than we expect the code generation for any one opcode to require. */ - s->code_gen_highwater = s->code_gen_buffer + (total_size - TCG_HIGHWATER); - #ifdef TCG_TARGET_NEED_POOL_LABELS s->pool_labels = NULL; #endif @@ -XXX,XX +XXX,XX @@ void tcg_prologue_init(TCGContext *s) } #endif - buf1 = s->code_ptr; + prologue_size = tcg_current_code_size(s); + #ifndef CONFIG_TCG_INTERPRETER - flush_idcache_range((uintptr_t)tcg_splitwx_to_rx(buf0), (uintptr_t)buf0, - tcg_ptr_byte_diff(buf1, buf0)); + flush_idcache_range((uintptr_t)tcg_splitwx_to_rx(s->code_buf), + (uintptr_t)s->code_buf, prologue_size); #endif - /* Deduct the prologue from the buffer. */ - prologue_size = tcg_current_code_size(s); - s->code_gen_ptr = buf1; - s->code_gen_buffer = buf1; - s->code_buf = buf1; - total_size -= prologue_size; - s->code_gen_buffer_size = total_size; + /* Deduct the prologue from the first region. */ + region.start = s->code_ptr; - tcg_register_jit(tcg_splitwx_to_rx(s->code_gen_buffer), total_size); + /* Recompute boundaries of the first region. */ + tcg_region_assign(s, 0); + + tcg_register_jit(tcg_splitwx_to_rx(region.start), + region.end - region.start); #ifdef DEBUG_DISAS if (qemu_loglevel_mask(CPU_LOG_TB_OUT_ASM)) { FILE *logfile = qemu_log_lock(); qemu_log("PROLOGUE: [size=%zu]\n", prologue_size); if (s->data_gen_ptr) { - size_t code_size = s->data_gen_ptr - buf0; + size_t code_size = s->data_gen_ptr - s->code_gen_ptr; size_t data_size = prologue_size - code_size; size_t i; - log_disas(buf0, code_size); + log_disas(s->code_gen_ptr, code_size); for (i = 0; i < data_size; i += sizeof(tcg_target_ulong)) { if (sizeof(tcg_target_ulong) == 8) { @@ -XXX,XX +XXX,XX @@ void tcg_prologue_init(TCGContext *s) } } } else { - log_disas(buf0, prologue_size); + log_disas(s->code_gen_ptr, prologue_size); } qemu_log("\n"); qemu_log_flush(); -- 2.25.1 All callers immediately assert on error, so move the assert into the function itself. Reviewed-by: Luis Pires <luis.pires@eldorado.org.br> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/tcg.c | 19 ++++++------------- 1 file changed, 6 insertions(+), 13 deletions(-) diff --git a/tcg/tcg.c b/tcg/tcg.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/tcg.c +++ b/tcg/tcg.c @@ -XXX,XX +XXX,XX @@ static bool tcg_region_alloc(TCGContext *s) * Perform a context's first region allocation. * This function does _not_ increment region.agg_size_full. */ -static inline bool tcg_region_initial_alloc__locked(TCGContext *s) +static void tcg_region_initial_alloc__locked(TCGContext *s) { - return tcg_region_alloc__locked(s); + bool err = tcg_region_alloc__locked(s); + g_assert(!err); } /* Call from a safe-work context */ @@ -XXX,XX +XXX,XX @@ void tcg_region_reset_all(void) for (i = 0; i < n_ctxs; i++) { TCGContext *s = qatomic_read(&tcg_ctxs[i]); - bool err = tcg_region_initial_alloc__locked(s); - - g_assert(!err); + tcg_region_initial_alloc__locked(s); } qemu_mutex_unlock(&region.lock); @@ -XXX,XX +XXX,XX @@ void tcg_region_init(void) /* In user-mode we support only one ctx, so do the initial allocation now */ #ifdef CONFIG_USER_ONLY - { - bool err = tcg_region_initial_alloc__locked(tcg_ctx); - - g_assert(!err); - } + tcg_region_initial_alloc__locked(tcg_ctx); #endif } @@ -XXX,XX +XXX,XX @@ void tcg_register_thread(void) MachineState *ms = MACHINE(qdev_get_machine()); TCGContext *s = g_malloc(sizeof(*s)); unsigned int i, n; - bool err; *s = tcg_init_ctx; @@ -XXX,XX +XXX,XX @@ void tcg_register_thread(void) tcg_ctx = s; qemu_mutex_lock(&region.lock); - err = tcg_region_initial_alloc__locked(tcg_ctx); - g_assert(!err); + tcg_region_initial_alloc__locked(s); qemu_mutex_unlock(&region.lock); } #endif /* !CONFIG_USER_ONLY */ -- 2.25.1 This has only one user, and currently needs an ifdef, but will make more sense after some code motion. Reviewed-by: Luis Pires <luis.pires@eldorado.org.br> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/tcg.c | 13 ++++++++++--- 1 file changed, 10 insertions(+), 3 deletions(-) diff --git a/tcg/tcg.c b/tcg/tcg.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/tcg.c +++ b/tcg/tcg.c @@ -XXX,XX +XXX,XX @@ static void tcg_region_initial_alloc__locked(TCGContext *s) g_assert(!err); } +#ifndef CONFIG_USER_ONLY +static void tcg_region_initial_alloc(TCGContext *s) +{ + qemu_mutex_lock(&region.lock); + tcg_region_initial_alloc__locked(s); + qemu_mutex_unlock(&region.lock); +} +#endif + /* Call from a safe-work context */ void tcg_region_reset_all(void) { @@ -XXX,XX +XXX,XX @@ void tcg_register_thread(void) } tcg_ctx = s; - qemu_mutex_lock(&region.lock); - tcg_region_initial_alloc__locked(s); - qemu_mutex_unlock(&region.lock); + tcg_region_initial_alloc(s); } #endif /* !CONFIG_USER_ONLY */ -- 2.25.1 This has only one user, but will make more sense after some code motion. Always leave the tcg_init_ctx initialized to the first region, in preparation for tcg_prologue_init(). This also requires that we don't re-allocate the region for the first cpu, lest we hit the assertion for total number of regions allocated . Reviewed-by: Luis Pires <luis.pires@eldorado.org.br> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/tcg.c | 37 ++++++++++++++++++++++--------------- 1 file changed, 22 insertions(+), 15 deletions(-) diff --git a/tcg/tcg.c b/tcg/tcg.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/tcg.c +++ b/tcg/tcg.c @@ -XXX,XX +XXX,XX @@ void tcg_region_init(void) tcg_region_trees_init(); - /* In user-mode we support only one ctx, so do the initial allocation now */ -#ifdef CONFIG_USER_ONLY - tcg_region_initial_alloc__locked(tcg_ctx); -#endif + /* + * Leave the initial context initialized to the first region. + * This will be the context into which we generate the prologue. + * It is also the only context for CONFIG_USER_ONLY. + */ + tcg_region_initial_alloc__locked(&tcg_init_ctx); +} + +static void tcg_region_prologue_set(TCGContext *s) +{ + /* Deduct the prologue from the first region. */ + g_assert(region.start == s->code_gen_buffer); + region.start = s->code_ptr; + + /* Recompute boundaries of the first region. */ + tcg_region_assign(s, 0); + + /* Register the balance of the buffer with gdb. */ + tcg_register_jit(tcg_splitwx_to_rx(region.start), + region.end - region.start); } #ifdef CONFIG_DEBUG_TCG @@ -XXX,XX +XXX,XX @@ void tcg_register_thread(void) if (n > 0) { alloc_tcg_plugin_context(s); + tcg_region_initial_alloc(s); } tcg_ctx = s; - tcg_region_initial_alloc(s); } #endif /* !CONFIG_USER_ONLY */ @@ -XXX,XX +XXX,XX @@ void tcg_prologue_init(TCGContext *s) { size_t prologue_size; - /* Put the prologue at the beginning of code_gen_buffer. */ - tcg_region_assign(s, 0); s->code_ptr = s->code_gen_ptr; s->code_buf = s->code_gen_ptr; s->data_gen_ptr = NULL; @@ -XXX,XX +XXX,XX @@ void tcg_prologue_init(TCGContext *s) (uintptr_t)s->code_buf, prologue_size); #endif - /* Deduct the prologue from the first region. */ - region.start = s->code_ptr; - - /* Recompute boundaries of the first region. */ - tcg_region_assign(s, 0); - - tcg_register_jit(tcg_splitwx_to_rx(region.start), - region.end - region.start); + tcg_region_prologue_set(s); #ifdef DEBUG_DISAS if (qemu_loglevel_mask(CPU_LOG_TB_OUT_ASM)) { -- 2.25.1 Reviewed-by: Luis Pires <luis.pires@eldorado.org.br> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/tcg-internal.h | 37 +++ tcg/region.c | 572 +++++++++++++++++++++++++++++++++++++++++++++ tcg/tcg.c | 547 +------------------------------------------ tcg/meson.build | 1 + 4 files changed, 613 insertions(+), 544 deletions(-) create mode 100644 tcg/tcg-internal.h create mode 100644 tcg/region.c diff --git a/tcg/tcg-internal.h b/tcg/tcg-internal.h new file mode 100644 index XXXXXXX..XXXXXXX --- /dev/null +++ b/tcg/tcg-internal.h @@ -XXX,XX +XXX,XX @@ +/* + * Internal declarations for Tiny Code Generator for QEMU + * + * Copyright (c) 2008 Fabrice Bellard + * + * Permission is hereby granted, free of charge, to any person obtaining a copy + * of this software and associated documentation files (the "Software"), to deal + * in the Software without restriction, including without limitation the rights + * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell + * copies of the Software, and to permit persons to whom the Software is + * furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN + * THE SOFTWARE. + */ + +#ifndef TCG_INTERNAL_H +#define TCG_INTERNAL_H 1 + +#define TCG_HIGHWATER 1024 + +extern TCGContext **tcg_ctxs; +extern unsigned int n_tcg_ctxs; + +bool tcg_region_alloc(TCGContext *s); +void tcg_region_initial_alloc(TCGContext *s); +void tcg_region_prologue_set(TCGContext *s); + +#endif /* TCG_INTERNAL_H */ diff --git a/tcg/region.c b/tcg/region.c new file mode 100644 index XXXXXXX..XXXXXXX --- /dev/null +++ b/tcg/region.c @@ -XXX,XX +XXX,XX @@ +/* + * Memory region management for Tiny Code Generator for QEMU + * + * Copyright (c) 2008 Fabrice Bellard + * + * Permission is hereby granted, free of charge, to any person obtaining a copy + * of this software and associated documentation files (the "Software"), to deal + * in the Software without restriction, including without limitation the rights + * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell + * copies of the Software, and to permit persons to whom the Software is + * furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN + * THE SOFTWARE. + */ + +#include "qemu/osdep.h" +#include "exec/exec-all.h" +#include "tcg/tcg.h" +#if !defined(CONFIG_USER_ONLY) +#include "hw/boards.h" +#endif +#include "tcg-internal.h" + + +struct tcg_region_tree { + QemuMutex lock; + GTree *tree; + /* padding to avoid false sharing is computed at run-time */ +}; + +/* + * We divide code_gen_buffer into equally-sized "regions" that TCG threads + * dynamically allocate from as demand dictates. Given appropriate region + * sizing, this minimizes flushes even when some TCG threads generate a lot + * more code than others. + */ +struct tcg_region_state { + QemuMutex lock; + + /* fields set at init time */ + void *start; + void *start_aligned; + void *end; + size_t n; + size_t size; /* size of one region */ + size_t stride; /* .size + guard size */ + + /* fields protected by the lock */ + size_t current; /* current region index */ + size_t agg_size_full; /* aggregate size of full regions */ +}; + +static struct tcg_region_state region; + +/* + * This is an array of struct tcg_region_tree's, with padding. + * We use void * to simplify the computation of region_trees[i]; each + * struct is found every tree_size bytes. + */ +static void *region_trees; +static size_t tree_size; + +/* compare a pointer @ptr and a tb_tc @s */ +static int ptr_cmp_tb_tc(const void *ptr, const struct tb_tc *s) +{ + if (ptr >= s->ptr + s->size) { + return 1; + } else if (ptr < s->ptr) { + return -1; + } + return 0; +} + +static gint tb_tc_cmp(gconstpointer ap, gconstpointer bp) +{ + const struct tb_tc *a = ap; + const struct tb_tc *b = bp; + + /* + * When both sizes are set, we know this isn't a lookup. + * This is the most likely case: every TB must be inserted; lookups + * are a lot less frequent. + */ + if (likely(a->size && b->size)) { + if (a->ptr > b->ptr) { + return 1; + } else if (a->ptr < b->ptr) { + return -1; + } + /* a->ptr == b->ptr should happen only on deletions */ + g_assert(a->size == b->size); + return 0; + } + /* + * All lookups have either .size field set to 0. + * From the glib sources we see that @ap is always the lookup key. However + * the docs provide no guarantee, so we just mark this case as likely. + */ + if (likely(a->size == 0)) { + return ptr_cmp_tb_tc(a->ptr, b); + } + return ptr_cmp_tb_tc(b->ptr, a); +} + +static void tcg_region_trees_init(void) +{ + size_t i; + + tree_size = ROUND_UP(sizeof(struct tcg_region_tree), qemu_dcache_linesize); + region_trees = qemu_memalign(qemu_dcache_linesize, region.n * tree_size); + for (i = 0; i < region.n; i++) { + struct tcg_region_tree *rt = region_trees + i * tree_size; + + qemu_mutex_init(&rt->lock); + rt->tree = g_tree_new(tb_tc_cmp); + } +} + +static struct tcg_region_tree *tc_ptr_to_region_tree(const void *p) +{ + size_t region_idx; + + /* + * Like tcg_splitwx_to_rw, with no assert. The pc may come from + * a signal handler over which the caller has no control. + */ + if (!in_code_gen_buffer(p)) { + p -= tcg_splitwx_diff; + if (!in_code_gen_buffer(p)) { + return NULL; + } + } + + if (p < region.start_aligned) { + region_idx = 0; + } else { + ptrdiff_t offset = p - region.start_aligned; + + if (offset > region.stride * (region.n - 1)) { + region_idx = region.n - 1; + } else { + region_idx = offset / region.stride; + } + } + return region_trees + region_idx * tree_size; +} + +void tcg_tb_insert(TranslationBlock *tb) +{ + struct tcg_region_tree *rt = tc_ptr_to_region_tree(tb->tc.ptr); + + g_assert(rt != NULL); + qemu_mutex_lock(&rt->lock); + g_tree_insert(rt->tree, &tb->tc, tb); + qemu_mutex_unlock(&rt->lock); +} + +void tcg_tb_remove(TranslationBlock *tb) +{ + struct tcg_region_tree *rt = tc_ptr_to_region_tree(tb->tc.ptr); + + g_assert(rt != NULL); + qemu_mutex_lock(&rt->lock); + g_tree_remove(rt->tree, &tb->tc); + qemu_mutex_unlock(&rt->lock); +} + +/* + * Find the TB 'tb' such that + * tb->tc.ptr <= tc_ptr < tb->tc.ptr + tb->tc.size + * Return NULL if not found. + */ +TranslationBlock *tcg_tb_lookup(uintptr_t tc_ptr) +{ + struct tcg_region_tree *rt = tc_ptr_to_region_tree((void *)tc_ptr); + TranslationBlock *tb; + struct tb_tc s = { .ptr = (void *)tc_ptr }; + + if (rt == NULL) { + return NULL; + } + + qemu_mutex_lock(&rt->lock); + tb = g_tree_lookup(rt->tree, &s); + qemu_mutex_unlock(&rt->lock); + return tb; +} + +static void tcg_region_tree_lock_all(void) +{ + size_t i; + + for (i = 0; i < region.n; i++) { + struct tcg_region_tree *rt = region_trees + i * tree_size; + + qemu_mutex_lock(&rt->lock); + } +} + +static void tcg_region_tree_unlock_all(void) +{ + size_t i; + + for (i = 0; i < region.n; i++) { + struct tcg_region_tree *rt = region_trees + i * tree_size; + + qemu_mutex_unlock(&rt->lock); + } +} + +void tcg_tb_foreach(GTraverseFunc func, gpointer user_data) +{ + size_t i; + + tcg_region_tree_lock_all(); + for (i = 0; i < region.n; i++) { + struct tcg_region_tree *rt = region_trees + i * tree_size; + + g_tree_foreach(rt->tree, func, user_data); + } + tcg_region_tree_unlock_all(); +} + +size_t tcg_nb_tbs(void) +{ + size_t nb_tbs = 0; + size_t i; + + tcg_region_tree_lock_all(); + for (i = 0; i < region.n; i++) { + struct tcg_region_tree *rt = region_trees + i * tree_size; + + nb_tbs += g_tree_nnodes(rt->tree); + } + tcg_region_tree_unlock_all(); + return nb_tbs; +} + +static gboolean tcg_region_tree_traverse(gpointer k, gpointer v, gpointer data) +{ + TranslationBlock *tb = v; + + tb_destroy(tb); + return FALSE; +} + +static void tcg_region_tree_reset_all(void) +{ + size_t i; + + tcg_region_tree_lock_all(); + for (i = 0; i < region.n; i++) { + struct tcg_region_tree *rt = region_trees + i * tree_size; + + g_tree_foreach(rt->tree, tcg_region_tree_traverse, NULL); + /* Increment the refcount first so that destroy acts as a reset */ + g_tree_ref(rt->tree); + g_tree_destroy(rt->tree); + } + tcg_region_tree_unlock_all(); +} + +static void tcg_region_bounds(size_t curr_region, void **pstart, void **pend) +{ + void *start, *end; + + start = region.start_aligned + curr_region * region.stride; + end = start + region.size; + + if (curr_region == 0) { + start = region.start; + } + if (curr_region == region.n - 1) { + end = region.end; + } + + *pstart = start; + *pend = end; +} + +static void tcg_region_assign(TCGContext *s, size_t curr_region) +{ + void *start, *end; + + tcg_region_bounds(curr_region, &start, &end); + + s->code_gen_buffer = start; + s->code_gen_ptr = start; + s->code_gen_buffer_size = end - start; + s->code_gen_highwater = end - TCG_HIGHWATER; +} + +static bool tcg_region_alloc__locked(TCGContext *s) +{ + if (region.current == region.n) { + return true; + } + tcg_region_assign(s, region.current); + region.current++; + return false; +} + +/* + * Request a new region once the one in use has filled up. + * Returns true on error. + */ +bool tcg_region_alloc(TCGContext *s) +{ + bool err; + /* read the region size now; alloc__locked will overwrite it on success */ + size_t size_full = s->code_gen_buffer_size; + + qemu_mutex_lock(&region.lock); + err = tcg_region_alloc__locked(s); + if (!err) { + region.agg_size_full += size_full - TCG_HIGHWATER; + } + qemu_mutex_unlock(&region.lock); + return err; +} + +/* + * Perform a context's first region allocation. + * This function does _not_ increment region.agg_size_full. + */ +static void tcg_region_initial_alloc__locked(TCGContext *s) +{ + bool err = tcg_region_alloc__locked(s); + g_assert(!err); +} + +void tcg_region_initial_alloc(TCGContext *s) +{ + qemu_mutex_lock(&region.lock); + tcg_region_initial_alloc__locked(s); + qemu_mutex_unlock(&region.lock); +} + +/* Call from a safe-work context */ +void tcg_region_reset_all(void) +{ + unsigned int n_ctxs = qatomic_read(&n_tcg_ctxs); + unsigned int i; + + qemu_mutex_lock(&region.lock); + region.current = 0; + region.agg_size_full = 0; + + for (i = 0; i < n_ctxs; i++) { + TCGContext *s = qatomic_read(&tcg_ctxs[i]); + tcg_region_initial_alloc__locked(s); + } + qemu_mutex_unlock(&region.lock); + + tcg_region_tree_reset_all(); +} + +#ifdef CONFIG_USER_ONLY +static size_t tcg_n_regions(void) +{ + return 1; +} +#else +/* + * It is likely that some vCPUs will translate more code than others, so we + * first try to set more regions than max_cpus, with those regions being of + * reasonable size. If that's not possible we make do by evenly dividing + * the code_gen_buffer among the vCPUs. + */ +static size_t tcg_n_regions(void) +{ + size_t i; + + /* Use a single region if all we have is one vCPU thread */ +#if !defined(CONFIG_USER_ONLY) + MachineState *ms = MACHINE(qdev_get_machine()); + unsigned int max_cpus = ms->smp.max_cpus; +#endif + if (max_cpus == 1 || !qemu_tcg_mttcg_enabled()) { + return 1; + } + + /* Try to have more regions than max_cpus, with each region being >= 2 MB */ + for (i = 8; i > 0; i--) { + size_t regions_per_thread = i; + size_t region_size; + + region_size = tcg_init_ctx.code_gen_buffer_size; + region_size /= max_cpus * regions_per_thread; + + if (region_size >= 2 * 1024u * 1024) { + return max_cpus * regions_per_thread; + } + } + /* If we can't, then just allocate one region per vCPU thread */ + return max_cpus; +} +#endif + +/* + * Initializes region partitioning. + * + * Called at init time from the parent thread (i.e. the one calling + * tcg_context_init), after the target's TCG globals have been set. + * + * Region partitioning works by splitting code_gen_buffer into separate regions, + * and then assigning regions to TCG threads so that the threads can translate + * code in parallel without synchronization. + * + * In softmmu the number of TCG threads is bounded by max_cpus, so we use at + * least max_cpus regions in MTTCG. In !MTTCG we use a single region. + * Note that the TCG options from the command-line (i.e. -accel accel=tcg,[...]) + * must have been parsed before calling this function, since it calls + * qemu_tcg_mttcg_enabled(). + * + * In user-mode we use a single region. Having multiple regions in user-mode + * is not supported, because the number of vCPU threads (recall that each thread + * spawned by the guest corresponds to a vCPU thread) is only bounded by the + * OS, and usually this number is huge (tens of thousands is not uncommon). + * Thus, given this large bound on the number of vCPU threads and the fact + * that code_gen_buffer is allocated at compile-time, we cannot guarantee + * that the availability of at least one region per vCPU thread. + * + * However, this user-mode limitation is unlikely to be a significant problem + * in practice. Multi-threaded guests share most if not all of their translated + * code, which makes parallel code generation less appealing than in softmmu. + */ +void tcg_region_init(void) +{ + void *buf = tcg_init_ctx.code_gen_buffer; + void *aligned; + size_t size = tcg_init_ctx.code_gen_buffer_size; + size_t page_size = qemu_real_host_page_size; + size_t region_size; + size_t n_regions; + size_t i; + + n_regions = tcg_n_regions(); + + /* The first region will be 'aligned - buf' bytes larger than the others */ + aligned = QEMU_ALIGN_PTR_UP(buf, page_size); + g_assert(aligned < tcg_init_ctx.code_gen_buffer + size); + /* + * Make region_size a multiple of page_size, using aligned as the start. + * As a result of this we might end up with a few extra pages at the end of + * the buffer; we will assign those to the last region. + */ + region_size = (size - (aligned - buf)) / n_regions; + region_size = QEMU_ALIGN_DOWN(region_size, page_size); + + /* A region must have at least 2 pages; one code, one guard */ + g_assert(region_size >= 2 * page_size); + + /* init the region struct */ + qemu_mutex_init(&region.lock); + region.n = n_regions; + region.size = region_size - page_size; + region.stride = region_size; + region.start = buf; + region.start_aligned = aligned; + /* page-align the end, since its last page will be a guard page */ + region.end = QEMU_ALIGN_PTR_DOWN(buf + size, page_size); + /* account for that last guard page */ + region.end -= page_size; + + /* + * Set guard pages in the rw buffer, as that's the one into which + * buffer overruns could occur. Do not set guard pages in the rx + * buffer -- let that one use hugepages throughout. + */ + for (i = 0; i < region.n; i++) { + void *start, *end; + + tcg_region_bounds(i, &start, &end); + + /* + * macOS 11.2 has a bug (Apple Feedback FB8994773) in which mprotect + * rejects a permission change from RWX -> NONE. Guard pages are + * nice for bug detection but are not essential; ignore any failure. + */ + (void)qemu_mprotect_none(end, page_size); + } + + tcg_region_trees_init(); + + /* + * Leave the initial context initialized to the first region. + * This will be the context into which we generate the prologue. + * It is also the only context for CONFIG_USER_ONLY. + */ + tcg_region_initial_alloc__locked(&tcg_init_ctx); +} + +void tcg_region_prologue_set(TCGContext *s) +{ + /* Deduct the prologue from the first region. */ + g_assert(region.start == s->code_gen_buffer); + region.start = s->code_ptr; + + /* Recompute boundaries of the first region. */ + tcg_region_assign(s, 0); + + /* Register the balance of the buffer with gdb. */ + tcg_register_jit(tcg_splitwx_to_rx(region.start), + region.end - region.start); +} + +/* + * Returns the size (in bytes) of all translated code (i.e. from all regions) + * currently in the cache. + * See also: tcg_code_capacity() + * Do not confuse with tcg_current_code_size(); that one applies to a single + * TCG context. + */ +size_t tcg_code_size(void) +{ + unsigned int n_ctxs = qatomic_read(&n_tcg_ctxs); + unsigned int i; + size_t total; + + qemu_mutex_lock(&region.lock); + total = region.agg_size_full; + for (i = 0; i < n_ctxs; i++) { + const TCGContext *s = qatomic_read(&tcg_ctxs[i]); + size_t size; + + size = qatomic_read(&s->code_gen_ptr) - s->code_gen_buffer; + g_assert(size <= s->code_gen_buffer_size); + total += size; + } + qemu_mutex_unlock(&region.lock); + return total; +} + +/* + * Returns the code capacity (in bytes) of the entire cache, i.e. including all + * regions. + * See also: tcg_code_size() + */ +size_t tcg_code_capacity(void) +{ + size_t guard_size, capacity; + + /* no need for synchronization; these variables are set at init time */ + guard_size = region.stride - region.size; + capacity = region.end + guard_size - region.start; + capacity -= region.n * (guard_size + TCG_HIGHWATER); + return capacity; +} + +size_t tcg_tb_phys_invalidate_count(void) +{ + unsigned int n_ctxs = qatomic_read(&n_tcg_ctxs); + unsigned int i; + size_t total = 0; + + for (i = 0; i < n_ctxs; i++) { + const TCGContext *s = qatomic_read(&tcg_ctxs[i]); + + total += qatomic_read(&s->tb_phys_invalidate_count); + } + return total; +} diff --git a/tcg/tcg.c b/tcg/tcg.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/tcg.c +++ b/tcg/tcg.c @@ -XXX,XX +XXX,XX @@ #include "elf.h" #include "exec/log.h" +#include "tcg-internal.h" /* Forward declarations for functions declared in tcg-target.c.inc and used here. */ @@ -XXX,XX +XXX,XX @@ static bool tcg_target_const_match(int64_t val, TCGType type, int ct); static int tcg_out_ldst_finalize(TCGContext *s); #endif -#define TCG_HIGHWATER 1024 - -static TCGContext **tcg_ctxs; -static unsigned int n_tcg_ctxs; +TCGContext **tcg_ctxs; +unsigned int n_tcg_ctxs; TCGv_env cpu_env = 0; const void *tcg_code_gen_epilogue; uintptr_t tcg_splitwx_diff; @@ -XXX,XX +XXX,XX @@ uintptr_t tcg_splitwx_diff; tcg_prologue_fn *tcg_qemu_tb_exec; #endif -struct tcg_region_tree { - QemuMutex lock; - GTree *tree; - /* padding to avoid false sharing is computed at run-time */ -}; - -/* - * We divide code_gen_buffer into equally-sized "regions" that TCG threads - * dynamically allocate from as demand dictates. Given appropriate region - * sizing, this minimizes flushes even when some TCG threads generate a lot - * more code than others. - */ -struct tcg_region_state { - QemuMutex lock; - - /* fields set at init time */ - void *start; - void *start_aligned; - void *end; - size_t n; - size_t size; /* size of one region */ - size_t stride; /* .size + guard size */ - - /* fields protected by the lock */ - size_t current; /* current region index */ - size_t agg_size_full; /* aggregate size of full regions */ -}; - -static struct tcg_region_state region; -/* - * This is an array of struct tcg_region_tree's, with padding. - * We use void * to simplify the computation of region_trees[i]; each - * struct is found every tree_size bytes. - */ -static void *region_trees; -static size_t tree_size; static TCGRegSet tcg_target_available_regs[TCG_TYPE_COUNT]; static TCGRegSet tcg_target_call_clobber_regs; @@ -XXX,XX +XXX,XX @@ static const TCGTargetOpDef constraint_sets[] = { #include "tcg-target.c.inc" -/* compare a pointer @ptr and a tb_tc @s */ -static int ptr_cmp_tb_tc(const void *ptr, const struct tb_tc *s) -{ - if (ptr >= s->ptr + s->size) { - return 1; - } else if (ptr < s->ptr) { - return -1; - } - return 0; -} - -static gint tb_tc_cmp(gconstpointer ap, gconstpointer bp) -{ - const struct tb_tc *a = ap; - const struct tb_tc *b = bp; - - /* - * When both sizes are set, we know this isn't a lookup. - * This is the most likely case: every TB must be inserted; lookups - * are a lot less frequent. - */ - if (likely(a->size && b->size)) { - if (a->ptr > b->ptr) { - return 1; - } else if (a->ptr < b->ptr) { - return -1; - } - /* a->ptr == b->ptr should happen only on deletions */ - g_assert(a->size == b->size); - return 0; - } - /* - * All lookups have either .size field set to 0. - * From the glib sources we see that @ap is always the lookup key. However - * the docs provide no guarantee, so we just mark this case as likely. - */ - if (likely(a->size == 0)) { - return ptr_cmp_tb_tc(a->ptr, b); - } - return ptr_cmp_tb_tc(b->ptr, a); -} - -static void tcg_region_trees_init(void) -{ - size_t i; - - tree_size = ROUND_UP(sizeof(struct tcg_region_tree), qemu_dcache_linesize); - region_trees = qemu_memalign(qemu_dcache_linesize, region.n * tree_size); - for (i = 0; i < region.n; i++) { - struct tcg_region_tree *rt = region_trees + i * tree_size; - - qemu_mutex_init(&rt->lock); - rt->tree = g_tree_new(tb_tc_cmp); - } -} - -static struct tcg_region_tree *tc_ptr_to_region_tree(const void *p) -{ - size_t region_idx; - - /* - * Like tcg_splitwx_to_rw, with no assert. The pc may come from - * a signal handler over which the caller has no control. - */ - if (!in_code_gen_buffer(p)) { - p -= tcg_splitwx_diff; - if (!in_code_gen_buffer(p)) { - return NULL; - } - } - - if (p < region.start_aligned) { - region_idx = 0; - } else { - ptrdiff_t offset = p - region.start_aligned; - - if (offset > region.stride * (region.n - 1)) { - region_idx = region.n - 1; - } else { - region_idx = offset / region.stride; - } - } - return region_trees + region_idx * tree_size; -} - -void tcg_tb_insert(TranslationBlock *tb) -{ - struct tcg_region_tree *rt = tc_ptr_to_region_tree(tb->tc.ptr); - - g_assert(rt != NULL); - qemu_mutex_lock(&rt->lock); - g_tree_insert(rt->tree, &tb->tc, tb); - qemu_mutex_unlock(&rt->lock); -} - -void tcg_tb_remove(TranslationBlock *tb) -{ - struct tcg_region_tree *rt = tc_ptr_to_region_tree(tb->tc.ptr); - - g_assert(rt != NULL); - qemu_mutex_lock(&rt->lock); - g_tree_remove(rt->tree, &tb->tc); - qemu_mutex_unlock(&rt->lock); -} - -/* - * Find the TB 'tb' such that - * tb->tc.ptr <= tc_ptr < tb->tc.ptr + tb->tc.size - * Return NULL if not found. - */ -TranslationBlock *tcg_tb_lookup(uintptr_t tc_ptr) -{ - struct tcg_region_tree *rt = tc_ptr_to_region_tree((void *)tc_ptr); - TranslationBlock *tb; - struct tb_tc s = { .ptr = (void *)tc_ptr }; - - if (rt == NULL) { - return NULL; - } - - qemu_mutex_lock(&rt->lock); - tb = g_tree_lookup(rt->tree, &s); - qemu_mutex_unlock(&rt->lock); - return tb; -} - -static void tcg_region_tree_lock_all(void) -{ - size_t i; - - for (i = 0; i < region.n; i++) { - struct tcg_region_tree *rt = region_trees + i * tree_size; - - qemu_mutex_lock(&rt->lock); - } -} - -static void tcg_region_tree_unlock_all(void) -{ - size_t i; - - for (i = 0; i < region.n; i++) { - struct tcg_region_tree *rt = region_trees + i * tree_size; - - qemu_mutex_unlock(&rt->lock); - } -} - -void tcg_tb_foreach(GTraverseFunc func, gpointer user_data) -{ - size_t i; - - tcg_region_tree_lock_all(); - for (i = 0; i < region.n; i++) { - struct tcg_region_tree *rt = region_trees + i * tree_size; - - g_tree_foreach(rt->tree, func, user_data); - } - tcg_region_tree_unlock_all(); -} - -size_t tcg_nb_tbs(void) -{ - size_t nb_tbs = 0; - size_t i; - - tcg_region_tree_lock_all(); - for (i = 0; i < region.n; i++) { - struct tcg_region_tree *rt = region_trees + i * tree_size; - - nb_tbs += g_tree_nnodes(rt->tree); - } - tcg_region_tree_unlock_all(); - return nb_tbs; -} - -static gboolean tcg_region_tree_traverse(gpointer k, gpointer v, gpointer data) -{ - TranslationBlock *tb = v; - - tb_destroy(tb); - return FALSE; -} - -static void tcg_region_tree_reset_all(void) -{ - size_t i; - - tcg_region_tree_lock_all(); - for (i = 0; i < region.n; i++) { - struct tcg_region_tree *rt = region_trees + i * tree_size; - - g_tree_foreach(rt->tree, tcg_region_tree_traverse, NULL); - /* Increment the refcount first so that destroy acts as a reset */ - g_tree_ref(rt->tree); - g_tree_destroy(rt->tree); - } - tcg_region_tree_unlock_all(); -} - -static void tcg_region_bounds(size_t curr_region, void **pstart, void **pend) -{ - void *start, *end; - - start = region.start_aligned + curr_region * region.stride; - end = start + region.size; - - if (curr_region == 0) { - start = region.start; - } - if (curr_region == region.n - 1) { - end = region.end; - } - - *pstart = start; - *pend = end; -} - -static void tcg_region_assign(TCGContext *s, size_t curr_region) -{ - void *start, *end; - - tcg_region_bounds(curr_region, &start, &end); - - s->code_gen_buffer = start; - s->code_gen_ptr = start; - s->code_gen_buffer_size = end - start; - s->code_gen_highwater = end - TCG_HIGHWATER; -} - -static bool tcg_region_alloc__locked(TCGContext *s) -{ - if (region.current == region.n) { - return true; - } - tcg_region_assign(s, region.current); - region.current++; - return false; -} - -/* - * Request a new region once the one in use has filled up. - * Returns true on error. - */ -static bool tcg_region_alloc(TCGContext *s) -{ - bool err; - /* read the region size now; alloc__locked will overwrite it on success */ - size_t size_full = s->code_gen_buffer_size; - - qemu_mutex_lock(&region.lock); - err = tcg_region_alloc__locked(s); - if (!err) { - region.agg_size_full += size_full - TCG_HIGHWATER; - } - qemu_mutex_unlock(&region.lock); - return err; -} - -/* - * Perform a context's first region allocation. - * This function does _not_ increment region.agg_size_full. - */ -static void tcg_region_initial_alloc__locked(TCGContext *s) -{ - bool err = tcg_region_alloc__locked(s); - g_assert(!err); -} - -#ifndef CONFIG_USER_ONLY -static void tcg_region_initial_alloc(TCGContext *s) -{ - qemu_mutex_lock(&region.lock); - tcg_region_initial_alloc__locked(s); - qemu_mutex_unlock(&region.lock); -} -#endif - -/* Call from a safe-work context */ -void tcg_region_reset_all(void) -{ - unsigned int n_ctxs = qatomic_read(&n_tcg_ctxs); - unsigned int i; - - qemu_mutex_lock(&region.lock); - region.current = 0; - region.agg_size_full = 0; - - for (i = 0; i < n_ctxs; i++) { - TCGContext *s = qatomic_read(&tcg_ctxs[i]); - tcg_region_initial_alloc__locked(s); - } - qemu_mutex_unlock(&region.lock); - - tcg_region_tree_reset_all(); -} - -#ifdef CONFIG_USER_ONLY -static size_t tcg_n_regions(void) -{ - return 1; -} -#else -/* - * It is likely that some vCPUs will translate more code than others, so we - * first try to set more regions than max_cpus, with those regions being of - * reasonable size. If that's not possible we make do by evenly dividing - * the code_gen_buffer among the vCPUs. - */ -static size_t tcg_n_regions(void) -{ - size_t i; - - /* Use a single region if all we have is one vCPU thread */ -#if !defined(CONFIG_USER_ONLY) - MachineState *ms = MACHINE(qdev_get_machine()); - unsigned int max_cpus = ms->smp.max_cpus; -#endif - if (max_cpus == 1 || !qemu_tcg_mttcg_enabled()) { - return 1; - } - - /* Try to have more regions than max_cpus, with each region being >= 2 MB */ - for (i = 8; i > 0; i--) { - size_t regions_per_thread = i; - size_t region_size; - - region_size = tcg_init_ctx.code_gen_buffer_size; - region_size /= max_cpus * regions_per_thread; - - if (region_size >= 2 * 1024u * 1024) { - return max_cpus * regions_per_thread; - } - } - /* If we can't, then just allocate one region per vCPU thread */ - return max_cpus; -} -#endif - -/* - * Initializes region partitioning. - * - * Called at init time from the parent thread (i.e. the one calling - * tcg_context_init), after the target's TCG globals have been set. - * - * Region partitioning works by splitting code_gen_buffer into separate regions, - * and then assigning regions to TCG threads so that the threads can translate - * code in parallel without synchronization. - * - * In softmmu the number of TCG threads is bounded by max_cpus, so we use at - * least max_cpus regions in MTTCG. In !MTTCG we use a single region. - * Note that the TCG options from the command-line (i.e. -accel accel=tcg,[...]) - * must have been parsed before calling this function, since it calls - * qemu_tcg_mttcg_enabled(). - * - * In user-mode we use a single region. Having multiple regions in user-mode - * is not supported, because the number of vCPU threads (recall that each thread - * spawned by the guest corresponds to a vCPU thread) is only bounded by the - * OS, and usually this number is huge (tens of thousands is not uncommon). - * Thus, given this large bound on the number of vCPU threads and the fact - * that code_gen_buffer is allocated at compile-time, we cannot guarantee - * that the availability of at least one region per vCPU thread. - * - * However, this user-mode limitation is unlikely to be a significant problem - * in practice. Multi-threaded guests share most if not all of their translated - * code, which makes parallel code generation less appealing than in softmmu. - */ -void tcg_region_init(void) -{ - void *buf = tcg_init_ctx.code_gen_buffer; - void *aligned; - size_t size = tcg_init_ctx.code_gen_buffer_size; - size_t page_size = qemu_real_host_page_size; - size_t region_size; - size_t n_regions; - size_t i; - - n_regions = tcg_n_regions(); - - /* The first region will be 'aligned - buf' bytes larger than the others */ - aligned = QEMU_ALIGN_PTR_UP(buf, page_size); - g_assert(aligned < tcg_init_ctx.code_gen_buffer + size); - /* - * Make region_size a multiple of page_size, using aligned as the start. - * As a result of this we might end up with a few extra pages at the end of - * the buffer; we will assign those to the last region. - */ - region_size = (size - (aligned - buf)) / n_regions; - region_size = QEMU_ALIGN_DOWN(region_size, page_size); - - /* A region must have at least 2 pages; one code, one guard */ - g_assert(region_size >= 2 * page_size); - - /* init the region struct */ - qemu_mutex_init(&region.lock); - region.n = n_regions; - region.size = region_size - page_size; - region.stride = region_size; - region.start = buf; - region.start_aligned = aligned; - /* page-align the end, since its last page will be a guard page */ - region.end = QEMU_ALIGN_PTR_DOWN(buf + size, page_size); - /* account for that last guard page */ - region.end -= page_size; - - /* - * Set guard pages in the rw buffer, as that's the one into which - * buffer overruns could occur. Do not set guard pages in the rx - * buffer -- let that one use hugepages throughout. - */ - for (i = 0; i < region.n; i++) { - void *start, *end; - - tcg_region_bounds(i, &start, &end); - - /* - * macOS 11.2 has a bug (Apple Feedback FB8994773) in which mprotect - * rejects a permission change from RWX -> NONE. Guard pages are - * nice for bug detection but are not essential; ignore any failure. - */ - (void)qemu_mprotect_none(end, page_size); - } - - tcg_region_trees_init(); - - /* - * Leave the initial context initialized to the first region. - * This will be the context into which we generate the prologue. - * It is also the only context for CONFIG_USER_ONLY. - */ - tcg_region_initial_alloc__locked(&tcg_init_ctx); -} - -static void tcg_region_prologue_set(TCGContext *s) -{ - /* Deduct the prologue from the first region. */ - g_assert(region.start == s->code_gen_buffer); - region.start = s->code_ptr; - - /* Recompute boundaries of the first region. */ - tcg_region_assign(s, 0); - - /* Register the balance of the buffer with gdb. */ - tcg_register_jit(tcg_splitwx_to_rx(region.start), - region.end - region.start); -} - #ifdef CONFIG_DEBUG_TCG const void *tcg_splitwx_to_rx(void *rw) { @@ -XXX,XX +XXX,XX @@ void tcg_register_thread(void) } #endif /* !CONFIG_USER_ONLY */ -/* - * Returns the size (in bytes) of all translated code (i.e. from all regions) - * currently in the cache. - * See also: tcg_code_capacity() - * Do not confuse with tcg_current_code_size(); that one applies to a single - * TCG context. - */ -size_t tcg_code_size(void) -{ - unsigned int n_ctxs = qatomic_read(&n_tcg_ctxs); - unsigned int i; - size_t total; - - qemu_mutex_lock(&region.lock); - total = region.agg_size_full; - for (i = 0; i < n_ctxs; i++) { - const TCGContext *s = qatomic_read(&tcg_ctxs[i]); - size_t size; - - size = qatomic_read(&s->code_gen_ptr) - s->code_gen_buffer; - g_assert(size <= s->code_gen_buffer_size); - total += size; - } - qemu_mutex_unlock(&region.lock); - return total; -} - -/* - * Returns the code capacity (in bytes) of the entire cache, i.e. including all - * regions. - * See also: tcg_code_size() - */ -size_t tcg_code_capacity(void) -{ - size_t guard_size, capacity; - - /* no need for synchronization; these variables are set at init time */ - guard_size = region.stride - region.size; - capacity = region.end + guard_size - region.start; - capacity -= region.n * (guard_size + TCG_HIGHWATER); - return capacity; -} - -size_t tcg_tb_phys_invalidate_count(void) -{ - unsigned int n_ctxs = qatomic_read(&n_tcg_ctxs); - unsigned int i; - size_t total = 0; - - for (i = 0; i < n_ctxs; i++) { - const TCGContext *s = qatomic_read(&tcg_ctxs[i]); - - total += qatomic_read(&s->tb_phys_invalidate_count); - } - return total; -} - /* pool based memory allocation */ void *tcg_malloc_internal(TCGContext *s, int size) { diff --git a/tcg/meson.build b/tcg/meson.build index XXXXXXX..XXXXXXX 100644 --- a/tcg/meson.build +++ b/tcg/meson.build @@ -XXX,XX +XXX,XX @@ tcg_ss = ss.source_set() tcg_ss.add(files( 'optimize.c', + 'region.c', 'tcg.c', 'tcg-common.c', 'tcg-op.c', -- 2.25.1 It consists of one function call and has only one caller. Reviewed-by: Luis Pires <luis.pires@eldorado.org.br> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- accel/tcg/translate-all.c | 7 +------ 1 file changed, 1 insertion(+), 6 deletions(-) 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 void page_table_config_init(void) assert(v_l2_levels >= 0); } -static void cpu_gen_init(void) -{ - tcg_context_init(&tcg_init_ctx); -} - /* Encode VAL as a signed leb128 sequence at P. Return P incremented past the encoded value. */ static uint8_t *encode_sleb128(uint8_t *p, target_long val) @@ -XXX,XX +XXX,XX @@ void tcg_exec_init(unsigned long tb_size, int splitwx) bool ok; tcg_allowed = true; - cpu_gen_init(); + tcg_context_init(&tcg_init_ctx); page_init(); tb_htable_init(); -- 2.25.1 Buffer management is integral to tcg. Do not leave the allocation to code outside of tcg/. This is code movement, with further cleanups to follow. Reviewed-by: Luis Pires <luis.pires@eldorado.org.br> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- include/tcg/tcg.h | 2 +- accel/tcg/translate-all.c | 414 +----------------------------------- tcg/region.c | 431 +++++++++++++++++++++++++++++++++++++- 3 files changed, 428 insertions(+), 419 deletions(-) diff --git a/include/tcg/tcg.h b/include/tcg/tcg.h index XXXXXXX..XXXXXXX 100644 --- a/include/tcg/tcg.h +++ b/include/tcg/tcg.h @@ -XXX,XX +XXX,XX @@ void *tcg_malloc_internal(TCGContext *s, int size); void tcg_pool_reset(TCGContext *s); TranslationBlock *tcg_tb_alloc(TCGContext *s); -void tcg_region_init(void); +void tcg_region_init(size_t tb_size, int splitwx); void tb_destroy(TranslationBlock *tb); void tcg_region_reset_all(void); 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 @@ */ #include "qemu/osdep.h" -#include "qemu/units.h" #include "qemu-common.h" #define NO_CPU_IO_DEFS @@ -XXX,XX +XXX,XX @@ #include "exec/cputlb.h" #include "exec/translate-all.h" #include "qemu/bitmap.h" -#include "qemu/error-report.h" #include "qemu/qemu-print.h" #include "qemu/timer.h" #include "qemu/main-loop.h" @@ -XXX,XX +XXX,XX @@ static void page_lock_pair(PageDesc **ret_p1, tb_page_addr_t phys1, } } -/* Minimum size of the code gen buffer. This number is randomly chosen, - but not so small that we can't have a fair number of TB's live. */ -#define MIN_CODE_GEN_BUFFER_SIZE (1 * MiB) - -/* Maximum size of the code gen buffer we'd like to use. Unless otherwise - indicated, this is constrained by the range of direct branches on the - host cpu, as used by the TCG implementation of goto_tb. */ -#if defined(__x86_64__) -# define MAX_CODE_GEN_BUFFER_SIZE (2 * GiB) -#elif defined(__sparc__) -# define MAX_CODE_GEN_BUFFER_SIZE (2 * GiB) -#elif defined(__powerpc64__) -# define MAX_CODE_GEN_BUFFER_SIZE (2 * GiB) -#elif defined(__powerpc__) -# define MAX_CODE_GEN_BUFFER_SIZE (32 * MiB) -#elif defined(__aarch64__) -# define MAX_CODE_GEN_BUFFER_SIZE (2 * GiB) -#elif defined(__s390x__) - /* We have a +- 4GB range on the branches; leave some slop. */ -# define MAX_CODE_GEN_BUFFER_SIZE (3 * GiB) -#elif defined(__mips__) - /* We have a 256MB branch region, but leave room to make sure the - main executable is also within that region. */ -# define MAX_CODE_GEN_BUFFER_SIZE (128 * MiB) -#else -# define MAX_CODE_GEN_BUFFER_SIZE ((size_t)-1) -#endif - -#if TCG_TARGET_REG_BITS == 32 -#define DEFAULT_CODE_GEN_BUFFER_SIZE_1 (32 * MiB) -#ifdef CONFIG_USER_ONLY -/* - * For user mode on smaller 32 bit systems we may run into trouble - * allocating big chunks of data in the right place. On these systems - * we utilise a static code generation buffer directly in the binary. - */ -#define USE_STATIC_CODE_GEN_BUFFER -#endif -#else /* TCG_TARGET_REG_BITS == 64 */ -#ifdef CONFIG_USER_ONLY -/* - * As user-mode emulation typically means running multiple instances - * of the translator don't go too nuts with our default code gen - * buffer lest we make things too hard for the OS. - */ -#define DEFAULT_CODE_GEN_BUFFER_SIZE_1 (128 * MiB) -#else -/* - * We expect most system emulation to run one or two guests per host. - * Users running large scale system emulation may want to tweak their - * runtime setup via the tb-size control on the command line. - */ -#define DEFAULT_CODE_GEN_BUFFER_SIZE_1 (1 * GiB) -#endif -#endif - -#define DEFAULT_CODE_GEN_BUFFER_SIZE \ - (DEFAULT_CODE_GEN_BUFFER_SIZE_1 < MAX_CODE_GEN_BUFFER_SIZE \ - ? DEFAULT_CODE_GEN_BUFFER_SIZE_1 : MAX_CODE_GEN_BUFFER_SIZE) - -static size_t size_code_gen_buffer(size_t tb_size) -{ - /* Size the buffer. */ - if (tb_size == 0) { - size_t phys_mem = qemu_get_host_physmem(); - if (phys_mem == 0) { - tb_size = DEFAULT_CODE_GEN_BUFFER_SIZE; - } else { - tb_size = MIN(DEFAULT_CODE_GEN_BUFFER_SIZE, phys_mem / 8); - } - } - if (tb_size < MIN_CODE_GEN_BUFFER_SIZE) { - tb_size = MIN_CODE_GEN_BUFFER_SIZE; - } - if (tb_size > MAX_CODE_GEN_BUFFER_SIZE) { - tb_size = MAX_CODE_GEN_BUFFER_SIZE; - } - return tb_size; -} - -#ifdef __mips__ -/* In order to use J and JAL within the code_gen_buffer, we require - that the buffer not cross a 256MB boundary. */ -static inline bool cross_256mb(void *addr, size_t size) -{ - return ((uintptr_t)addr ^ ((uintptr_t)addr + size)) & ~0x0ffffffful; -} - -/* We weren't able to allocate a buffer without crossing that boundary, - so make do with the larger portion of the buffer that doesn't cross. - Returns the new base of the buffer, and adjusts code_gen_buffer_size. */ -static inline void *split_cross_256mb(void *buf1, size_t size1) -{ - void *buf2 = (void *)(((uintptr_t)buf1 + size1) & ~0x0ffffffful); - size_t size2 = buf1 + size1 - buf2; - - size1 = buf2 - buf1; - if (size1 < size2) { - size1 = size2; - buf1 = buf2; - } - - tcg_ctx->code_gen_buffer_size = size1; - return buf1; -} -#endif - -#ifdef USE_STATIC_CODE_GEN_BUFFER -static uint8_t static_code_gen_buffer[DEFAULT_CODE_GEN_BUFFER_SIZE] - __attribute__((aligned(CODE_GEN_ALIGN))); - -static bool alloc_code_gen_buffer(size_t tb_size, int splitwx, Error **errp) -{ - void *buf, *end; - size_t size; - - if (splitwx > 0) { - error_setg(errp, "jit split-wx not supported"); - return false; - } - - /* page-align the beginning and end of the buffer */ - buf = static_code_gen_buffer; - end = static_code_gen_buffer + sizeof(static_code_gen_buffer); - buf = QEMU_ALIGN_PTR_UP(buf, qemu_real_host_page_size); - end = QEMU_ALIGN_PTR_DOWN(end, qemu_real_host_page_size); - - size = end - buf; - - /* Honor a command-line option limiting the size of the buffer. */ - if (size > tb_size) { - size = QEMU_ALIGN_DOWN(tb_size, qemu_real_host_page_size); - } - tcg_ctx->code_gen_buffer_size = size; - -#ifdef __mips__ - if (cross_256mb(buf, size)) { - buf = split_cross_256mb(buf, size); - size = tcg_ctx->code_gen_buffer_size; - } -#endif - - if (qemu_mprotect_rwx(buf, size)) { - error_setg_errno(errp, errno, "mprotect of jit buffer"); - return false; - } - qemu_madvise(buf, size, QEMU_MADV_HUGEPAGE); - - tcg_ctx->code_gen_buffer = buf; - return true; -} -#elif defined(_WIN32) -static bool alloc_code_gen_buffer(size_t size, int splitwx, Error **errp) -{ - void *buf; - - if (splitwx > 0) { - error_setg(errp, "jit split-wx not supported"); - return false; - } - - buf = VirtualAlloc(NULL, size, MEM_RESERVE | MEM_COMMIT, - PAGE_EXECUTE_READWRITE); - if (buf == NULL) { - error_setg_win32(errp, GetLastError(), - "allocate %zu bytes for jit buffer", size); - return false; - } - - tcg_ctx->code_gen_buffer = buf; - tcg_ctx->code_gen_buffer_size = size; - return true; -} -#else -static bool alloc_code_gen_buffer_anon(size_t size, int prot, - int flags, Error **errp) -{ - void *buf; - - buf = mmap(NULL, size, prot, flags, -1, 0); - if (buf == MAP_FAILED) { - error_setg_errno(errp, errno, - "allocate %zu bytes for jit buffer", size); - return false; - } - tcg_ctx->code_gen_buffer_size = size; - -#ifdef __mips__ - if (cross_256mb(buf, size)) { - /* - * Try again, with the original still mapped, to avoid re-acquiring - * the same 256mb crossing. - */ - size_t size2; - void *buf2 = mmap(NULL, size, prot, flags, -1, 0); - switch ((int)(buf2 != MAP_FAILED)) { - case 1: - if (!cross_256mb(buf2, size)) { - /* Success! Use the new buffer. */ - munmap(buf, size); - break; - } - /* Failure. Work with what we had. */ - munmap(buf2, size); - /* fallthru */ - default: - /* Split the original buffer. Free the smaller half. */ - buf2 = split_cross_256mb(buf, size); - size2 = tcg_ctx->code_gen_buffer_size; - if (buf == buf2) { - munmap(buf + size2, size - size2); - } else { - munmap(buf, size - size2); - } - size = size2; - break; - } - buf = buf2; - } -#endif - - /* Request large pages for the buffer. */ - qemu_madvise(buf, size, QEMU_MADV_HUGEPAGE); - - tcg_ctx->code_gen_buffer = buf; - return true; -} - -#ifndef CONFIG_TCG_INTERPRETER -#ifdef CONFIG_POSIX -#include "qemu/memfd.h" - -static bool alloc_code_gen_buffer_splitwx_memfd(size_t size, Error **errp) -{ - void *buf_rw = NULL, *buf_rx = MAP_FAILED; - int fd = -1; - -#ifdef __mips__ - /* Find space for the RX mapping, vs the 256MiB regions. */ - if (!alloc_code_gen_buffer_anon(size, PROT_NONE, - MAP_PRIVATE | MAP_ANONYMOUS | - MAP_NORESERVE, errp)) { - return false; - } - /* The size of the mapping may have been adjusted. */ - size = tcg_ctx->code_gen_buffer_size; - buf_rx = tcg_ctx->code_gen_buffer; -#endif - - buf_rw = qemu_memfd_alloc("tcg-jit", size, 0, &fd, errp); - if (buf_rw == NULL) { - goto fail; - } - -#ifdef __mips__ - void *tmp = mmap(buf_rx, size, PROT_READ | PROT_EXEC, - MAP_SHARED | MAP_FIXED, fd, 0); - if (tmp != buf_rx) { - goto fail_rx; - } -#else - buf_rx = mmap(NULL, size, PROT_READ | PROT_EXEC, MAP_SHARED, fd, 0); - if (buf_rx == MAP_FAILED) { - goto fail_rx; - } -#endif - - close(fd); - tcg_ctx->code_gen_buffer = buf_rw; - tcg_ctx->code_gen_buffer_size = size; - tcg_splitwx_diff = buf_rx - buf_rw; - - /* Request large pages for the buffer and the splitwx. */ - qemu_madvise(buf_rw, size, QEMU_MADV_HUGEPAGE); - qemu_madvise(buf_rx, size, QEMU_MADV_HUGEPAGE); - return true; - - fail_rx: - error_setg_errno(errp, errno, "failed to map shared memory for execute"); - fail: - if (buf_rx != MAP_FAILED) { - munmap(buf_rx, size); - } - if (buf_rw) { - munmap(buf_rw, size); - } - if (fd >= 0) { - close(fd); - } - return false; -} -#endif /* CONFIG_POSIX */ - -#ifdef CONFIG_DARWIN -#include <mach/mach.h> - -extern kern_return_t mach_vm_remap(vm_map_t target_task, - mach_vm_address_t *target_address, - mach_vm_size_t size, - mach_vm_offset_t mask, - int flags, - vm_map_t src_task, - mach_vm_address_t src_address, - boolean_t copy, - vm_prot_t *cur_protection, - vm_prot_t *max_protection, - vm_inherit_t inheritance); - -static bool alloc_code_gen_buffer_splitwx_vmremap(size_t size, Error **errp) -{ - kern_return_t ret; - mach_vm_address_t buf_rw, buf_rx; - vm_prot_t cur_prot, max_prot; - - /* Map the read-write portion via normal anon memory. */ - if (!alloc_code_gen_buffer_anon(size, PROT_READ | PROT_WRITE, - MAP_PRIVATE | MAP_ANONYMOUS, errp)) { - return false; - } - - buf_rw = (mach_vm_address_t)tcg_ctx->code_gen_buffer; - buf_rx = 0; - ret = mach_vm_remap(mach_task_self(), - &buf_rx, - size, - 0, - VM_FLAGS_ANYWHERE, - mach_task_self(), - buf_rw, - false, - &cur_prot, - &max_prot, - VM_INHERIT_NONE); - if (ret != KERN_SUCCESS) { - /* TODO: Convert "ret" to a human readable error message. */ - error_setg(errp, "vm_remap for jit splitwx failed"); - munmap((void *)buf_rw, size); - return false; - } - - if (mprotect((void *)buf_rx, size, PROT_READ | PROT_EXEC) != 0) { - error_setg_errno(errp, errno, "mprotect for jit splitwx"); - munmap((void *)buf_rx, size); - munmap((void *)buf_rw, size); - return false; - } - - tcg_splitwx_diff = buf_rx - buf_rw; - return true; -} -#endif /* CONFIG_DARWIN */ -#endif /* CONFIG_TCG_INTERPRETER */ - -static bool alloc_code_gen_buffer_splitwx(size_t size, Error **errp) -{ -#ifndef CONFIG_TCG_INTERPRETER -# ifdef CONFIG_DARWIN - return alloc_code_gen_buffer_splitwx_vmremap(size, errp); -# endif -# ifdef CONFIG_POSIX - return alloc_code_gen_buffer_splitwx_memfd(size, errp); -# endif -#endif - error_setg(errp, "jit split-wx not supported"); - return false; -} - -static bool alloc_code_gen_buffer(size_t size, int splitwx, Error **errp) -{ - ERRP_GUARD(); - int prot, flags; - - if (splitwx) { - if (alloc_code_gen_buffer_splitwx(size, errp)) { - return true; - } - /* - * If splitwx force-on (1), fail; - * if splitwx default-on (-1), fall through to splitwx off. - */ - if (splitwx > 0) { - return false; - } - error_free_or_abort(errp); - } - - prot = PROT_READ | PROT_WRITE | PROT_EXEC; - flags = MAP_PRIVATE | MAP_ANONYMOUS; -#ifdef CONFIG_TCG_INTERPRETER - /* The tcg interpreter does not need execute permission. */ - prot = PROT_READ | PROT_WRITE; -#elif defined(CONFIG_DARWIN) - /* Applicable to both iOS and macOS (Apple Silicon). */ - if (!splitwx) { - flags |= MAP_JIT; - } -#endif - - return alloc_code_gen_buffer_anon(size, prot, flags, errp); -} -#endif /* USE_STATIC_CODE_GEN_BUFFER, WIN32, POSIX */ - static bool tb_cmp(const void *ap, const void *bp) { const TranslationBlock *a = ap; @@ -XXX,XX +XXX,XX @@ static void tb_htable_init(void) size. */ void tcg_exec_init(unsigned long tb_size, int splitwx) { - bool ok; - tcg_allowed = true; tcg_context_init(&tcg_init_ctx); page_init(); tb_htable_init(); - - ok = alloc_code_gen_buffer(size_code_gen_buffer(tb_size), - splitwx, &error_fatal); - assert(ok); - - /* TODO: allocating regions is hand-in-glove with code_gen_buffer. */ - tcg_region_init(); + tcg_region_init(tb_size, splitwx); #if defined(CONFIG_SOFTMMU) /* There's no guest base to take into account, so go ahead and diff --git a/tcg/region.c b/tcg/region.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/region.c +++ b/tcg/region.c @@ -XXX,XX +XXX,XX @@ */ #include "qemu/osdep.h" +#include "qemu/units.h" +#include "qapi/error.h" #include "exec/exec-all.h" #include "tcg/tcg.h" #if !defined(CONFIG_USER_ONLY) @@ -XXX,XX +XXX,XX @@ static size_t tcg_n_regions(void) } #endif +/* + * Minimum size of the code gen buffer. This number is randomly chosen, + * but not so small that we can't have a fair number of TB's live. + */ +#define MIN_CODE_GEN_BUFFER_SIZE (1 * MiB) + +/* + * Maximum size of the code gen buffer we'd like to use. Unless otherwise + * indicated, this is constrained by the range of direct branches on the + * host cpu, as used by the TCG implementation of goto_tb. + */ +#if defined(__x86_64__) +# define MAX_CODE_GEN_BUFFER_SIZE (2 * GiB) +#elif defined(__sparc__) +# define MAX_CODE_GEN_BUFFER_SIZE (2 * GiB) +#elif defined(__powerpc64__) +# define MAX_CODE_GEN_BUFFER_SIZE (2 * GiB) +#elif defined(__powerpc__) +# define MAX_CODE_GEN_BUFFER_SIZE (32 * MiB) +#elif defined(__aarch64__) +# define MAX_CODE_GEN_BUFFER_SIZE (2 * GiB) +#elif defined(__s390x__) + /* We have a +- 4GB range on the branches; leave some slop. */ +# define MAX_CODE_GEN_BUFFER_SIZE (3 * GiB) +#elif defined(__mips__) + /* + * We have a 256MB branch region, but leave room to make sure the + * main executable is also within that region. + */ +# define MAX_CODE_GEN_BUFFER_SIZE (128 * MiB) +#else +# define MAX_CODE_GEN_BUFFER_SIZE ((size_t)-1) +#endif + +#if TCG_TARGET_REG_BITS == 32 +#define DEFAULT_CODE_GEN_BUFFER_SIZE_1 (32 * MiB) +#ifdef CONFIG_USER_ONLY +/* + * For user mode on smaller 32 bit systems we may run into trouble + * allocating big chunks of data in the right place. On these systems + * we utilise a static code generation buffer directly in the binary. + */ +#define USE_STATIC_CODE_GEN_BUFFER +#endif +#else /* TCG_TARGET_REG_BITS == 64 */ +#ifdef CONFIG_USER_ONLY +/* + * As user-mode emulation typically means running multiple instances + * of the translator don't go too nuts with our default code gen + * buffer lest we make things too hard for the OS. + */ +#define DEFAULT_CODE_GEN_BUFFER_SIZE_1 (128 * MiB) +#else +/* + * We expect most system emulation to run one or two guests per host. + * Users running large scale system emulation may want to tweak their + * runtime setup via the tb-size control on the command line. + */ +#define DEFAULT_CODE_GEN_BUFFER_SIZE_1 (1 * GiB) +#endif +#endif + +#define DEFAULT_CODE_GEN_BUFFER_SIZE \ + (DEFAULT_CODE_GEN_BUFFER_SIZE_1 < MAX_CODE_GEN_BUFFER_SIZE \ + ? DEFAULT_CODE_GEN_BUFFER_SIZE_1 : MAX_CODE_GEN_BUFFER_SIZE) + +static size_t size_code_gen_buffer(size_t tb_size) +{ + /* Size the buffer. */ + if (tb_size == 0) { + size_t phys_mem = qemu_get_host_physmem(); + if (phys_mem == 0) { + tb_size = DEFAULT_CODE_GEN_BUFFER_SIZE; + } else { + tb_size = MIN(DEFAULT_CODE_GEN_BUFFER_SIZE, phys_mem / 8); + } + } + if (tb_size < MIN_CODE_GEN_BUFFER_SIZE) { + tb_size = MIN_CODE_GEN_BUFFER_SIZE; + } + if (tb_size > MAX_CODE_GEN_BUFFER_SIZE) { + tb_size = MAX_CODE_GEN_BUFFER_SIZE; + } + return tb_size; +} + +#ifdef __mips__ +/* + * In order to use J and JAL within the code_gen_buffer, we require + * that the buffer not cross a 256MB boundary. + */ +static inline bool cross_256mb(void *addr, size_t size) +{ + return ((uintptr_t)addr ^ ((uintptr_t)addr + size)) & ~0x0ffffffful; +} + +/* + * We weren't able to allocate a buffer without crossing that boundary, + * so make do with the larger portion of the buffer that doesn't cross. + * Returns the new base of the buffer, and adjusts code_gen_buffer_size. + */ +static inline void *split_cross_256mb(void *buf1, size_t size1) +{ + void *buf2 = (void *)(((uintptr_t)buf1 + size1) & ~0x0ffffffful); + size_t size2 = buf1 + size1 - buf2; + + size1 = buf2 - buf1; + if (size1 < size2) { + size1 = size2; + buf1 = buf2; + } + + tcg_ctx->code_gen_buffer_size = size1; + return buf1; +} +#endif + +#ifdef USE_STATIC_CODE_GEN_BUFFER +static uint8_t static_code_gen_buffer[DEFAULT_CODE_GEN_BUFFER_SIZE] + __attribute__((aligned(CODE_GEN_ALIGN))); + +static bool alloc_code_gen_buffer(size_t tb_size, int splitwx, Error **errp) +{ + void *buf, *end; + size_t size; + + if (splitwx > 0) { + error_setg(errp, "jit split-wx not supported"); + return false; + } + + /* page-align the beginning and end of the buffer */ + buf = static_code_gen_buffer; + end = static_code_gen_buffer + sizeof(static_code_gen_buffer); + buf = QEMU_ALIGN_PTR_UP(buf, qemu_real_host_page_size); + end = QEMU_ALIGN_PTR_DOWN(end, qemu_real_host_page_size); + + size = end - buf; + + /* Honor a command-line option limiting the size of the buffer. */ + if (size > tb_size) { + size = QEMU_ALIGN_DOWN(tb_size, qemu_real_host_page_size); + } + tcg_ctx->code_gen_buffer_size = size; + +#ifdef __mips__ + if (cross_256mb(buf, size)) { + buf = split_cross_256mb(buf, size); + size = tcg_ctx->code_gen_buffer_size; + } +#endif + + if (qemu_mprotect_rwx(buf, size)) { + error_setg_errno(errp, errno, "mprotect of jit buffer"); + return false; + } + qemu_madvise(buf, size, QEMU_MADV_HUGEPAGE); + + tcg_ctx->code_gen_buffer = buf; + return true; +} +#elif defined(_WIN32) +static bool alloc_code_gen_buffer(size_t size, int splitwx, Error **errp) +{ + void *buf; + + if (splitwx > 0) { + error_setg(errp, "jit split-wx not supported"); + return false; + } + + buf = VirtualAlloc(NULL, size, MEM_RESERVE | MEM_COMMIT, + PAGE_EXECUTE_READWRITE); + if (buf == NULL) { + error_setg_win32(errp, GetLastError(), + "allocate %zu bytes for jit buffer", size); + return false; + } + + tcg_ctx->code_gen_buffer = buf; + tcg_ctx->code_gen_buffer_size = size; + return true; +} +#else +static bool alloc_code_gen_buffer_anon(size_t size, int prot, + int flags, Error **errp) +{ + void *buf; + + buf = mmap(NULL, size, prot, flags, -1, 0); + if (buf == MAP_FAILED) { + error_setg_errno(errp, errno, + "allocate %zu bytes for jit buffer", size); + return false; + } + tcg_ctx->code_gen_buffer_size = size; + +#ifdef __mips__ + if (cross_256mb(buf, size)) { + /* + * Try again, with the original still mapped, to avoid re-acquiring + * the same 256mb crossing. + */ + size_t size2; + void *buf2 = mmap(NULL, size, prot, flags, -1, 0); + switch ((int)(buf2 != MAP_FAILED)) { + case 1: + if (!cross_256mb(buf2, size)) { + /* Success! Use the new buffer. */ + munmap(buf, size); + break; + } + /* Failure. Work with what we had. */ + munmap(buf2, size); + /* fallthru */ + default: + /* Split the original buffer. Free the smaller half. */ + buf2 = split_cross_256mb(buf, size); + size2 = tcg_ctx->code_gen_buffer_size; + if (buf == buf2) { + munmap(buf + size2, size - size2); + } else { + munmap(buf, size - size2); + } + size = size2; + break; + } + buf = buf2; + } +#endif + + /* Request large pages for the buffer. */ + qemu_madvise(buf, size, QEMU_MADV_HUGEPAGE); + + tcg_ctx->code_gen_buffer = buf; + return true; +} + +#ifndef CONFIG_TCG_INTERPRETER +#ifdef CONFIG_POSIX +#include "qemu/memfd.h" + +static bool alloc_code_gen_buffer_splitwx_memfd(size_t size, Error **errp) +{ + void *buf_rw = NULL, *buf_rx = MAP_FAILED; + int fd = -1; + +#ifdef __mips__ + /* Find space for the RX mapping, vs the 256MiB regions. */ + if (!alloc_code_gen_buffer_anon(size, PROT_NONE, + MAP_PRIVATE | MAP_ANONYMOUS | + MAP_NORESERVE, errp)) { + return false; + } + /* The size of the mapping may have been adjusted. */ + size = tcg_ctx->code_gen_buffer_size; + buf_rx = tcg_ctx->code_gen_buffer; +#endif + + buf_rw = qemu_memfd_alloc("tcg-jit", size, 0, &fd, errp); + if (buf_rw == NULL) { + goto fail; + } + +#ifdef __mips__ + void *tmp = mmap(buf_rx, size, PROT_READ | PROT_EXEC, + MAP_SHARED | MAP_FIXED, fd, 0); + if (tmp != buf_rx) { + goto fail_rx; + } +#else + buf_rx = mmap(NULL, size, PROT_READ | PROT_EXEC, MAP_SHARED, fd, 0); + if (buf_rx == MAP_FAILED) { + goto fail_rx; + } +#endif + + close(fd); + tcg_ctx->code_gen_buffer = buf_rw; + tcg_ctx->code_gen_buffer_size = size; + tcg_splitwx_diff = buf_rx - buf_rw; + + /* Request large pages for the buffer and the splitwx. */ + qemu_madvise(buf_rw, size, QEMU_MADV_HUGEPAGE); + qemu_madvise(buf_rx, size, QEMU_MADV_HUGEPAGE); + return true; + + fail_rx: + error_setg_errno(errp, errno, "failed to map shared memory for execute"); + fail: + if (buf_rx != MAP_FAILED) { + munmap(buf_rx, size); + } + if (buf_rw) { + munmap(buf_rw, size); + } + if (fd >= 0) { + close(fd); + } + return false; +} +#endif /* CONFIG_POSIX */ + +#ifdef CONFIG_DARWIN +#include <mach/mach.h> + +extern kern_return_t mach_vm_remap(vm_map_t target_task, + mach_vm_address_t *target_address, + mach_vm_size_t size, + mach_vm_offset_t mask, + int flags, + vm_map_t src_task, + mach_vm_address_t src_address, + boolean_t copy, + vm_prot_t *cur_protection, + vm_prot_t *max_protection, + vm_inherit_t inheritance); + +static bool alloc_code_gen_buffer_splitwx_vmremap(size_t size, Error **errp) +{ + kern_return_t ret; + mach_vm_address_t buf_rw, buf_rx; + vm_prot_t cur_prot, max_prot; + + /* Map the read-write portion via normal anon memory. */ + if (!alloc_code_gen_buffer_anon(size, PROT_READ | PROT_WRITE, + MAP_PRIVATE | MAP_ANONYMOUS, errp)) { + return false; + } + + buf_rw = (mach_vm_address_t)tcg_ctx->code_gen_buffer; + buf_rx = 0; + ret = mach_vm_remap(mach_task_self(), + &buf_rx, + size, + 0, + VM_FLAGS_ANYWHERE, + mach_task_self(), + buf_rw, + false, + &cur_prot, + &max_prot, + VM_INHERIT_NONE); + if (ret != KERN_SUCCESS) { + /* TODO: Convert "ret" to a human readable error message. */ + error_setg(errp, "vm_remap for jit splitwx failed"); + munmap((void *)buf_rw, size); + return false; + } + + if (mprotect((void *)buf_rx, size, PROT_READ | PROT_EXEC) != 0) { + error_setg_errno(errp, errno, "mprotect for jit splitwx"); + munmap((void *)buf_rx, size); + munmap((void *)buf_rw, size); + return false; + } + + tcg_splitwx_diff = buf_rx - buf_rw; + return true; +} +#endif /* CONFIG_DARWIN */ +#endif /* CONFIG_TCG_INTERPRETER */ + +static bool alloc_code_gen_buffer_splitwx(size_t size, Error **errp) +{ +#ifndef CONFIG_TCG_INTERPRETER +# ifdef CONFIG_DARWIN + return alloc_code_gen_buffer_splitwx_vmremap(size, errp); +# endif +# ifdef CONFIG_POSIX + return alloc_code_gen_buffer_splitwx_memfd(size, errp); +# endif +#endif + error_setg(errp, "jit split-wx not supported"); + return false; +} + +static bool alloc_code_gen_buffer(size_t size, int splitwx, Error **errp) +{ + ERRP_GUARD(); + int prot, flags; + + if (splitwx) { + if (alloc_code_gen_buffer_splitwx(size, errp)) { + return true; + } + /* + * If splitwx force-on (1), fail; + * if splitwx default-on (-1), fall through to splitwx off. + */ + if (splitwx > 0) { + return false; + } + error_free_or_abort(errp); + } + + prot = PROT_READ | PROT_WRITE | PROT_EXEC; + flags = MAP_PRIVATE | MAP_ANONYMOUS; +#ifdef CONFIG_TCG_INTERPRETER + /* The tcg interpreter does not need execute permission. */ + prot = PROT_READ | PROT_WRITE; +#elif defined(CONFIG_DARWIN) + /* Applicable to both iOS and macOS (Apple Silicon). */ + if (!splitwx) { + flags |= MAP_JIT; + } +#endif + + return alloc_code_gen_buffer_anon(size, prot, flags, errp); +} +#endif /* USE_STATIC_CODE_GEN_BUFFER, WIN32, POSIX */ + /* * Initializes region partitioning. * @@ -XXX,XX +XXX,XX @@ static size_t tcg_n_regions(void) * in practice. Multi-threaded guests share most if not all of their translated * code, which makes parallel code generation less appealing than in softmmu. */ -void tcg_region_init(void) +void tcg_region_init(size_t tb_size, int splitwx) { - void *buf = tcg_init_ctx.code_gen_buffer; - void *aligned; - size_t size = tcg_init_ctx.code_gen_buffer_size; - size_t page_size = qemu_real_host_page_size; + void *buf, *aligned; + size_t size; + size_t page_size; size_t region_size; size_t n_regions; size_t i; + bool ok; + ok = alloc_code_gen_buffer(size_code_gen_buffer(tb_size), + splitwx, &error_fatal); + assert(ok); + + buf = tcg_init_ctx.code_gen_buffer; + size = tcg_init_ctx.code_gen_buffer_size; + page_size = qemu_real_host_page_size; n_regions = tcg_n_regions(); /* The first region will be 'aligned - buf' bytes larger than the others */ -- 2.25.1 We shortly want to use tcg_init for something else. Since the hook is called init_machine, match that. Reviewed-by: Luis Pires <luis.pires@eldorado.org.br> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- accel/tcg/tcg-all.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/accel/tcg/tcg-all.c b/accel/tcg/tcg-all.c index XXXXXXX..XXXXXXX 100644 --- a/accel/tcg/tcg-all.c +++ b/accel/tcg/tcg-all.c @@ -XXX,XX +XXX,XX @@ static void tcg_accel_instance_init(Object *obj) bool mttcg_enabled; -static int tcg_init(MachineState *ms) +static int tcg_init_machine(MachineState *ms) { TCGState *s = TCG_STATE(current_accel()); @@ -XXX,XX +XXX,XX @@ static void tcg_accel_class_init(ObjectClass *oc, void *data) { AccelClass *ac = ACCEL_CLASS(oc); ac->name = "tcg"; - ac->init_machine = tcg_init; + ac->init_machine = tcg_init_machine; ac->allowed = &tcg_allowed; object_class_property_add_str(oc, "thread", -- 2.25.1 Perform both tcg_context_init and tcg_region_init. Do not leave this split to the caller. Reviewed-by: Luis Pires <luis.pires@eldorado.org.br> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- include/tcg/tcg.h | 3 +-- tcg/tcg-internal.h | 1 + accel/tcg/translate-all.c | 3 +-- tcg/tcg.c | 9 ++++++++- 4 files changed, 11 insertions(+), 5 deletions(-) diff --git a/include/tcg/tcg.h b/include/tcg/tcg.h index XXXXXXX..XXXXXXX 100644 --- a/include/tcg/tcg.h +++ b/include/tcg/tcg.h @@ -XXX,XX +XXX,XX @@ void *tcg_malloc_internal(TCGContext *s, int size); void tcg_pool_reset(TCGContext *s); TranslationBlock *tcg_tb_alloc(TCGContext *s); -void tcg_region_init(size_t tb_size, int splitwx); void tb_destroy(TranslationBlock *tb); void tcg_region_reset_all(void); @@ -XXX,XX +XXX,XX @@ static inline void *tcg_malloc(int size) } } -void tcg_context_init(TCGContext *s); +void tcg_init(size_t tb_size, int splitwx); void tcg_register_thread(void); void tcg_prologue_init(TCGContext *s); void tcg_func_start(TCGContext *s); diff --git a/tcg/tcg-internal.h b/tcg/tcg-internal.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/tcg-internal.h +++ b/tcg/tcg-internal.h @@ -XXX,XX +XXX,XX @@ extern TCGContext **tcg_ctxs; extern unsigned int n_tcg_ctxs; +void tcg_region_init(size_t tb_size, int splitwx); bool tcg_region_alloc(TCGContext *s); void tcg_region_initial_alloc(TCGContext *s); void tcg_region_prologue_set(TCGContext *s); 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 void tb_htable_init(void) void tcg_exec_init(unsigned long tb_size, int splitwx) { tcg_allowed = true; - tcg_context_init(&tcg_init_ctx); page_init(); tb_htable_init(); - tcg_region_init(tb_size, splitwx); + tcg_init(tb_size, splitwx); #if defined(CONFIG_SOFTMMU) /* There's no guest base to take into account, so go ahead and diff --git a/tcg/tcg.c b/tcg/tcg.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/tcg.c +++ b/tcg/tcg.c @@ -XXX,XX +XXX,XX @@ static void process_op_defs(TCGContext *s); static TCGTemp *tcg_global_reg_new_internal(TCGContext *s, TCGType type, TCGReg reg, const char *name); -void tcg_context_init(TCGContext *s) +static void tcg_context_init(void) { + TCGContext *s = &tcg_init_ctx; int op, total_args, n, i; TCGOpDef *def; TCGArgConstraint *args_ct; @@ -XXX,XX +XXX,XX @@ void tcg_context_init(TCGContext *s) cpu_env = temp_tcgv_ptr(ts); } +void tcg_init(size_t tb_size, int splitwx) +{ + tcg_context_init(); + tcg_region_init(tb_size, splitwx); +} + /* * Allocate TBs right before their corresponding translated code, making * sure that TBs and code are on different cache lines. -- 2.25.1 There is only one caller, and shortly we will need access to the MachineState, which tcg_init_machine already has. Reviewed-by: Luis Pires <luis.pires@eldorado.org.br> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- accel/tcg/internal.h | 2 ++ include/sysemu/tcg.h | 2 -- accel/tcg/tcg-all.c | 16 +++++++++++++++- accel/tcg/translate-all.c | 21 ++------------------- bsd-user/main.c | 2 +- 5 files changed, 20 insertions(+), 23 deletions(-) diff --git a/accel/tcg/internal.h b/accel/tcg/internal.h index XXXXXXX..XXXXXXX 100644 --- a/accel/tcg/internal.h +++ b/accel/tcg/internal.h @@ -XXX,XX +XXX,XX @@ TranslationBlock *tb_gen_code(CPUState *cpu, target_ulong pc, int cflags); void QEMU_NORETURN cpu_io_recompile(CPUState *cpu, uintptr_t retaddr); +void page_init(void); +void tb_htable_init(void); #endif /* ACCEL_TCG_INTERNAL_H */ diff --git a/include/sysemu/tcg.h b/include/sysemu/tcg.h index XXXXXXX..XXXXXXX 100644 --- a/include/sysemu/tcg.h +++ b/include/sysemu/tcg.h @@ -XXX,XX +XXX,XX @@ #ifndef SYSEMU_TCG_H #define SYSEMU_TCG_H -void tcg_exec_init(unsigned long tb_size, int splitwx); - #ifdef CONFIG_TCG extern bool tcg_allowed; #define tcg_enabled() (tcg_allowed) diff --git a/accel/tcg/tcg-all.c b/accel/tcg/tcg-all.c index XXXXXXX..XXXXXXX 100644 --- a/accel/tcg/tcg-all.c +++ b/accel/tcg/tcg-all.c @@ -XXX,XX +XXX,XX @@ #include "qemu/error-report.h" #include "qemu/accel.h" #include "qapi/qapi-builtin-visit.h" +#include "internal.h" struct TCGState { AccelState parent_obj; @@ -XXX,XX +XXX,XX @@ static int tcg_init_machine(MachineState *ms) { TCGState *s = TCG_STATE(current_accel()); - tcg_exec_init(s->tb_size * 1024 * 1024, s->splitwx_enabled); + tcg_allowed = true; mttcg_enabled = s->mttcg_enabled; + + page_init(); + tb_htable_init(); + tcg_init(s->tb_size * 1024 * 1024, s->splitwx_enabled); + +#if defined(CONFIG_SOFTMMU) + /* + * There's no guest base to take into account, so go ahead and + * initialize the prologue now. + */ + tcg_prologue_init(tcg_ctx); +#endif + return 0; } 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 @@ bool cpu_restore_state(CPUState *cpu, uintptr_t host_pc, bool will_exit) return false; } -static void page_init(void) +void page_init(void) { page_size_init(); page_table_config_init(); @@ -XXX,XX +XXX,XX @@ static bool tb_cmp(const void *ap, const void *bp) a->page_addr[1] == b->page_addr[1]; } -static void tb_htable_init(void) +void tb_htable_init(void) { unsigned int mode = QHT_MODE_AUTO_RESIZE; qht_init(&tb_ctx.htable, tb_cmp, CODE_GEN_HTABLE_SIZE, mode); } -/* Must be called before using the QEMU cpus. 'tb_size' is the size - (in bytes) allocated to the translation buffer. Zero means default - size. */ -void tcg_exec_init(unsigned long tb_size, int splitwx) -{ - tcg_allowed = true; - page_init(); - tb_htable_init(); - tcg_init(tb_size, splitwx); - -#if defined(CONFIG_SOFTMMU) - /* There's no guest base to take into account, so go ahead and - initialize the prologue now. */ - tcg_prologue_init(tcg_ctx); -#endif -} - /* call with @p->lock held */ static inline void invalidate_page_bitmap(PageDesc *p) { diff --git a/bsd-user/main.c b/bsd-user/main.c index XXXXXXX..XXXXXXX 100644 --- a/bsd-user/main.c +++ b/bsd-user/main.c @@ -XXX,XX +XXX,XX @@ int main(int argc, char **argv) envlist_free(envlist); /* - * Now that page sizes are configured in tcg_exec_init() we can do + * Now that page sizes are configured we can do * proper page alignment for guest_base. */ guest_base = HOST_PAGE_ALIGN(guest_base); -- 2.25.1 Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Reviewed-by: Luis Pires <luis.pires@eldorado.org.br> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- accel/tcg/tcg-all.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/accel/tcg/tcg-all.c b/accel/tcg/tcg-all.c index XXXXXXX..XXXXXXX 100644 --- a/accel/tcg/tcg-all.c +++ b/accel/tcg/tcg-all.c @@ -XXX,XX +XXX,XX @@ #include "qemu/error-report.h" #include "qemu/accel.h" #include "qapi/qapi-builtin-visit.h" +#include "qemu/units.h" #include "internal.h" struct TCGState { @@ -XXX,XX +XXX,XX @@ static int tcg_init_machine(MachineState *ms) page_init(); tb_htable_init(); - tcg_init(s->tb_size * 1024 * 1024, s->splitwx_enabled); + tcg_init(s->tb_size * MiB, s->splitwx_enabled); #if defined(CONFIG_SOFTMMU) /* -- 2.25.1 Start removing the include of hw/boards.h from tcg/. Pass down the max_cpus value from tcg_init_machine, where we have the MachineState already. Reviewed-by: Luis Pires <luis.pires@eldorado.org.br> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- include/tcg/tcg.h | 2 +- tcg/tcg-internal.h | 2 +- accel/tcg/tcg-all.c | 10 +++++++++- tcg/region.c | 32 +++++++++++--------------------- tcg/tcg.c | 10 ++++------ 5 files changed, 26 insertions(+), 30 deletions(-) diff --git a/include/tcg/tcg.h b/include/tcg/tcg.h index XXXXXXX..XXXXXXX 100644 --- a/include/tcg/tcg.h +++ b/include/tcg/tcg.h @@ -XXX,XX +XXX,XX @@ static inline void *tcg_malloc(int size) } } -void tcg_init(size_t tb_size, int splitwx); +void tcg_init(size_t tb_size, int splitwx, unsigned max_cpus); void tcg_register_thread(void); void tcg_prologue_init(TCGContext *s); void tcg_func_start(TCGContext *s); diff --git a/tcg/tcg-internal.h b/tcg/tcg-internal.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/tcg-internal.h +++ b/tcg/tcg-internal.h @@ -XXX,XX +XXX,XX @@ extern TCGContext **tcg_ctxs; extern unsigned int n_tcg_ctxs; -void tcg_region_init(size_t tb_size, int splitwx); +void tcg_region_init(size_t tb_size, int splitwx, unsigned max_cpus); bool tcg_region_alloc(TCGContext *s); void tcg_region_initial_alloc(TCGContext *s); void tcg_region_prologue_set(TCGContext *s); diff --git a/accel/tcg/tcg-all.c b/accel/tcg/tcg-all.c index XXXXXXX..XXXXXXX 100644 --- a/accel/tcg/tcg-all.c +++ b/accel/tcg/tcg-all.c @@ -XXX,XX +XXX,XX @@ #include "qemu/accel.h" #include "qapi/qapi-builtin-visit.h" #include "qemu/units.h" +#if !defined(CONFIG_USER_ONLY) +#include "hw/boards.h" +#endif #include "internal.h" struct TCGState { @@ -XXX,XX +XXX,XX @@ bool mttcg_enabled; static int tcg_init_machine(MachineState *ms) { TCGState *s = TCG_STATE(current_accel()); +#ifdef CONFIG_USER_ONLY + unsigned max_cpus = 1; +#else + unsigned max_cpus = ms->smp.max_cpus; +#endif tcg_allowed = true; mttcg_enabled = s->mttcg_enabled; page_init(); tb_htable_init(); - tcg_init(s->tb_size * MiB, s->splitwx_enabled); + tcg_init(s->tb_size * MiB, s->splitwx_enabled, max_cpus); #if defined(CONFIG_SOFTMMU) /* diff --git a/tcg/region.c b/tcg/region.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/region.c +++ b/tcg/region.c @@ -XXX,XX +XXX,XX @@ #include "qapi/error.h" #include "exec/exec-all.h" #include "tcg/tcg.h" -#if !defined(CONFIG_USER_ONLY) -#include "hw/boards.h" -#endif #include "tcg-internal.h" @@ -XXX,XX +XXX,XX @@ void tcg_region_reset_all(void) tcg_region_tree_reset_all(); } +static size_t tcg_n_regions(unsigned max_cpus) +{ #ifdef CONFIG_USER_ONLY -static size_t tcg_n_regions(void) -{ return 1; -} #else -/* - * It is likely that some vCPUs will translate more code than others, so we - * first try to set more regions than max_cpus, with those regions being of - * reasonable size. If that's not possible we make do by evenly dividing - * the code_gen_buffer among the vCPUs. - */ -static size_t tcg_n_regions(void) -{ + /* + * It is likely that some vCPUs will translate more code than others, + * so we first try to set more regions than max_cpus, with those regions + * being of reasonable size. If that's not possible we make do by evenly + * dividing the code_gen_buffer among the vCPUs. + */ size_t i; /* Use a single region if all we have is one vCPU thread */ -#if !defined(CONFIG_USER_ONLY) - MachineState *ms = MACHINE(qdev_get_machine()); - unsigned int max_cpus = ms->smp.max_cpus; -#endif if (max_cpus == 1 || !qemu_tcg_mttcg_enabled()) { return 1; } @@ -XXX,XX +XXX,XX @@ static size_t tcg_n_regions(void) } /* If we can't, then just allocate one region per vCPU thread */ return max_cpus; -} #endif +} /* * Minimum size of the code gen buffer. This number is randomly chosen, @@ -XXX,XX +XXX,XX @@ static bool alloc_code_gen_buffer(size_t size, int splitwx, Error **errp) * in practice. Multi-threaded guests share most if not all of their translated * code, which makes parallel code generation less appealing than in softmmu. */ -void tcg_region_init(size_t tb_size, int splitwx) +void tcg_region_init(size_t tb_size, int splitwx, unsigned max_cpus) { void *buf, *aligned; size_t size; @@ -XXX,XX +XXX,XX @@ void tcg_region_init(size_t tb_size, int splitwx) buf = tcg_init_ctx.code_gen_buffer; size = tcg_init_ctx.code_gen_buffer_size; page_size = qemu_real_host_page_size; - n_regions = tcg_n_regions(); + n_regions = tcg_n_regions(max_cpus); /* The first region will be 'aligned - buf' bytes larger than the others */ aligned = QEMU_ALIGN_PTR_UP(buf, page_size); diff --git a/tcg/tcg.c b/tcg/tcg.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/tcg.c +++ b/tcg/tcg.c @@ -XXX,XX +XXX,XX @@ static void process_op_defs(TCGContext *s); static TCGTemp *tcg_global_reg_new_internal(TCGContext *s, TCGType type, TCGReg reg, const char *name); -static void tcg_context_init(void) +static void tcg_context_init(unsigned max_cpus) { TCGContext *s = &tcg_init_ctx; int op, total_args, n, i; @@ -XXX,XX +XXX,XX @@ static void tcg_context_init(void) tcg_ctxs = &tcg_ctx; n_tcg_ctxs = 1; #else - MachineState *ms = MACHINE(qdev_get_machine()); - unsigned int max_cpus = ms->smp.max_cpus; tcg_ctxs = g_new(TCGContext *, max_cpus); #endif @@ -XXX,XX +XXX,XX @@ static void tcg_context_init(void) cpu_env = temp_tcgv_ptr(ts); } -void tcg_init(size_t tb_size, int splitwx) +void tcg_init(size_t tb_size, int splitwx, unsigned max_cpus) { - tcg_context_init(); - tcg_region_init(tb_size, splitwx); + tcg_context_init(max_cpus); + tcg_region_init(tb_size, splitwx, max_cpus); } /* -- 2.25.1 Finish the divorce of tcg/ from hw/, and do not take the max cpu value from MachineState; just remember what we were passed in tcg_init. Reviewed-by: Luis Pires <luis.pires@eldorado.org.br> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/tcg-internal.h | 3 ++- tcg/region.c | 6 +++--- tcg/tcg.c | 23 ++++++++++------------- 3 files changed, 15 insertions(+), 17 deletions(-) diff --git a/tcg/tcg-internal.h b/tcg/tcg-internal.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/tcg-internal.h +++ b/tcg/tcg-internal.h @@ -XXX,XX +XXX,XX @@ #define TCG_HIGHWATER 1024 extern TCGContext **tcg_ctxs; -extern unsigned int n_tcg_ctxs; +extern unsigned int tcg_cur_ctxs; +extern unsigned int tcg_max_ctxs; void tcg_region_init(size_t tb_size, int splitwx, unsigned max_cpus); bool tcg_region_alloc(TCGContext *s); diff --git a/tcg/region.c b/tcg/region.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/region.c +++ b/tcg/region.c @@ -XXX,XX +XXX,XX @@ void tcg_region_initial_alloc(TCGContext *s) /* Call from a safe-work context */ void tcg_region_reset_all(void) { - unsigned int n_ctxs = qatomic_read(&n_tcg_ctxs); + unsigned int n_ctxs = qatomic_read(&tcg_cur_ctxs); unsigned int i; qemu_mutex_lock(&region.lock); @@ -XXX,XX +XXX,XX @@ void tcg_region_prologue_set(TCGContext *s) */ size_t tcg_code_size(void) { - unsigned int n_ctxs = qatomic_read(&n_tcg_ctxs); + unsigned int n_ctxs = qatomic_read(&tcg_cur_ctxs); unsigned int i; size_t total; @@ -XXX,XX +XXX,XX @@ size_t tcg_code_capacity(void) size_t tcg_tb_phys_invalidate_count(void) { - unsigned int n_ctxs = qatomic_read(&n_tcg_ctxs); + unsigned int n_ctxs = qatomic_read(&tcg_cur_ctxs); unsigned int i; size_t total = 0; diff --git a/tcg/tcg.c b/tcg/tcg.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/tcg.c +++ b/tcg/tcg.c @@ -XXX,XX +XXX,XX @@ #define NO_CPU_IO_DEFS #include "exec/exec-all.h" - -#if !defined(CONFIG_USER_ONLY) -#include "hw/boards.h" -#endif - #include "tcg/tcg-op.h" #if UINTPTR_MAX == UINT32_MAX @@ -XXX,XX +XXX,XX @@ static int tcg_out_ldst_finalize(TCGContext *s); #endif TCGContext **tcg_ctxs; -unsigned int n_tcg_ctxs; +unsigned int tcg_cur_ctxs; +unsigned int tcg_max_ctxs; TCGv_env cpu_env = 0; const void *tcg_code_gen_epilogue; uintptr_t tcg_splitwx_diff; @@ -XXX,XX +XXX,XX @@ void tcg_register_thread(void) #else void tcg_register_thread(void) { - MachineState *ms = MACHINE(qdev_get_machine()); TCGContext *s = g_malloc(sizeof(*s)); unsigned int i, n; @@ -XXX,XX +XXX,XX @@ void tcg_register_thread(void) } /* Claim an entry in tcg_ctxs */ - n = qatomic_fetch_inc(&n_tcg_ctxs); - g_assert(n < ms->smp.max_cpus); + n = qatomic_fetch_inc(&tcg_cur_ctxs); + g_assert(n < tcg_max_ctxs); qatomic_set(&tcg_ctxs[n], s); if (n > 0) { @@ -XXX,XX +XXX,XX @@ static void tcg_context_init(unsigned max_cpus) */ #ifdef CONFIG_USER_ONLY tcg_ctxs = &tcg_ctx; - n_tcg_ctxs = 1; + tcg_cur_ctxs = 1; + tcg_max_ctxs = 1; #else - tcg_ctxs = g_new(TCGContext *, max_cpus); + tcg_max_ctxs = max_cpus; + tcg_ctxs = g_new0(TCGContext *, max_cpus); #endif tcg_debug_assert(!tcg_regset_test_reg(s->reserved_regs, TCG_AREG0)); @@ -XXX,XX +XXX,XX @@ static void tcg_reg_alloc_call(TCGContext *s, TCGOp *op) static inline void tcg_profile_snapshot(TCGProfile *prof, bool counters, bool table) { - unsigned int n_ctxs = qatomic_read(&n_tcg_ctxs); + unsigned int n_ctxs = qatomic_read(&tcg_cur_ctxs); unsigned int i; for (i = 0; i < n_ctxs; i++) { @@ -XXX,XX +XXX,XX @@ void tcg_dump_op_count(void) int64_t tcg_cpu_exec_time(void) { - unsigned int n_ctxs = qatomic_read(&n_tcg_ctxs); + unsigned int n_ctxs = qatomic_read(&tcg_cur_ctxs); unsigned int i; int64_t ret = 0; -- 2.25.1 Remove the ifdef ladder and move each define into the appropriate header file. Reviewed-by: Luis Pires <luis.pires@eldorado.org.br> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/aarch64/tcg-target.h | 1 + tcg/arm/tcg-target.h | 1 + tcg/i386/tcg-target.h | 2 ++ tcg/mips/tcg-target.h | 6 ++++++ tcg/ppc/tcg-target.h | 2 ++ tcg/riscv/tcg-target.h | 1 + tcg/s390/tcg-target.h | 3 +++ tcg/sparc/tcg-target.h | 1 + tcg/tci/tcg-target.h | 1 + tcg/region.c | 33 +++++---------------------------- 10 files changed, 23 insertions(+), 28 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 @@ #define TCG_TARGET_INSN_UNIT_SIZE 4 #define TCG_TARGET_TLB_DISPLACEMENT_BITS 24 +#define MAX_CODE_GEN_BUFFER_SIZE (2 * GiB) #undef TCG_TARGET_STACK_GROWSUP typedef enum { 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; #undef TCG_TARGET_STACK_GROWSUP #define TCG_TARGET_INSN_UNIT_SIZE 4 #define TCG_TARGET_TLB_DISPLACEMENT_BITS 16 +#define MAX_CODE_GEN_BUFFER_SIZE UINT32_MAX typedef enum { TCG_REG_R0 = 0, diff --git a/tcg/i386/tcg-target.h b/tcg/i386/tcg-target.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/i386/tcg-target.h +++ b/tcg/i386/tcg-target.h @@ -XXX,XX +XXX,XX @@ #ifdef __x86_64__ # define TCG_TARGET_REG_BITS 64 # define TCG_TARGET_NB_REGS 32 +# define MAX_CODE_GEN_BUFFER_SIZE (2 * GiB) #else # define TCG_TARGET_REG_BITS 32 # define TCG_TARGET_NB_REGS 24 +# define MAX_CODE_GEN_BUFFER_SIZE UINT32_MAX #endif typedef enum { 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 @@ #define TCG_TARGET_TLB_DISPLACEMENT_BITS 16 #define TCG_TARGET_NB_REGS 32 +/* + * We have a 256MB branch region, but leave room to make sure the + * main executable is also within that region. + */ +#define MAX_CODE_GEN_BUFFER_SIZE (128 * MiB) + typedef enum { TCG_REG_ZERO = 0, TCG_REG_AT, 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 @@ #ifdef _ARCH_PPC64 # define TCG_TARGET_REG_BITS 64 +# define MAX_CODE_GEN_BUFFER_SIZE (2 * GiB) #else # define TCG_TARGET_REG_BITS 32 +# define MAX_CODE_GEN_BUFFER_SIZE (32 * MiB) #endif #define TCG_TARGET_NB_REGS 64 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 @@ #define TCG_TARGET_INSN_UNIT_SIZE 4 #define TCG_TARGET_TLB_DISPLACEMENT_BITS 20 #define TCG_TARGET_NB_REGS 32 +#define MAX_CODE_GEN_BUFFER_SIZE ((size_t)-1) typedef enum { TCG_REG_ZERO, diff --git a/tcg/s390/tcg-target.h b/tcg/s390/tcg-target.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/s390/tcg-target.h +++ b/tcg/s390/tcg-target.h @@ -XXX,XX +XXX,XX @@ #define TCG_TARGET_INSN_UNIT_SIZE 2 #define TCG_TARGET_TLB_DISPLACEMENT_BITS 19 +/* We have a +- 4GB range on the branches; leave some slop. */ +#define MAX_CODE_GEN_BUFFER_SIZE (3 * GiB) + typedef enum TCGReg { TCG_REG_R0 = 0, TCG_REG_R1, diff --git a/tcg/sparc/tcg-target.h b/tcg/sparc/tcg-target.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/sparc/tcg-target.h +++ b/tcg/sparc/tcg-target.h @@ -XXX,XX +XXX,XX @@ #define TCG_TARGET_INSN_UNIT_SIZE 4 #define TCG_TARGET_TLB_DISPLACEMENT_BITS 32 #define TCG_TARGET_NB_REGS 32 +#define MAX_CODE_GEN_BUFFER_SIZE (2 * GiB) typedef enum { TCG_REG_G0 = 0, diff --git a/tcg/tci/tcg-target.h b/tcg/tci/tcg-target.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/tci/tcg-target.h +++ b/tcg/tci/tcg-target.h @@ -XXX,XX +XXX,XX @@ #define TCG_TARGET_INTERPRETER 1 #define TCG_TARGET_INSN_UNIT_SIZE 1 #define TCG_TARGET_TLB_DISPLACEMENT_BITS 32 +#define MAX_CODE_GEN_BUFFER_SIZE ((size_t)-1) #if UINTPTR_MAX == UINT32_MAX # define TCG_TARGET_REG_BITS 32 diff --git a/tcg/region.c b/tcg/region.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/region.c +++ b/tcg/region.c @@ -XXX,XX +XXX,XX @@ static size_t tcg_n_regions(unsigned max_cpus) /* * Minimum size of the code gen buffer. This number is randomly chosen, * but not so small that we can't have a fair number of TB's live. + * + * Maximum size, MAX_CODE_GEN_BUFFER_SIZE, is defined in tcg-target.h. + * Unless otherwise indicated, this is constrained by the range of + * direct branches on the host cpu, as used by the TCG implementation + * of goto_tb. */ #define MIN_CODE_GEN_BUFFER_SIZE (1 * MiB) -/* - * Maximum size of the code gen buffer we'd like to use. Unless otherwise - * indicated, this is constrained by the range of direct branches on the - * host cpu, as used by the TCG implementation of goto_tb. - */ -#if defined(__x86_64__) -# define MAX_CODE_GEN_BUFFER_SIZE (2 * GiB) -#elif defined(__sparc__) -# define MAX_CODE_GEN_BUFFER_SIZE (2 * GiB) -#elif defined(__powerpc64__) -# define MAX_CODE_GEN_BUFFER_SIZE (2 * GiB) -#elif defined(__powerpc__) -# define MAX_CODE_GEN_BUFFER_SIZE (32 * MiB) -#elif defined(__aarch64__) -# define MAX_CODE_GEN_BUFFER_SIZE (2 * GiB) -#elif defined(__s390x__) - /* We have a +- 4GB range on the branches; leave some slop. */ -# define MAX_CODE_GEN_BUFFER_SIZE (3 * GiB) -#elif defined(__mips__) - /* - * We have a 256MB branch region, but leave room to make sure the - * main executable is also within that region. - */ -# define MAX_CODE_GEN_BUFFER_SIZE (128 * MiB) -#else -# define MAX_CODE_GEN_BUFFER_SIZE ((size_t)-1) -#endif - #if TCG_TARGET_REG_BITS == 32 #define DEFAULT_CODE_GEN_BUFFER_SIZE_1 (32 * MiB) #ifdef CONFIG_USER_ONLY -- 2.25.1 A size is easier to work with than an end point, particularly during initial buffer allocation. Reviewed-by: Luis Pires <luis.pires@eldorado.org.br> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/region.c | 30 ++++++++++++++++++------------ 1 file changed, 18 insertions(+), 12 deletions(-) diff --git a/tcg/region.c b/tcg/region.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/region.c +++ b/tcg/region.c @@ -XXX,XX +XXX,XX @@ struct tcg_region_state { /* fields set at init time */ void *start; void *start_aligned; - void *end; size_t n; size_t size; /* size of one region */ size_t stride; /* .size + guard size */ + size_t total_size; /* size of entire buffer, >= n * stride */ /* fields protected by the lock */ size_t current; /* current region index */ @@ -XXX,XX +XXX,XX @@ static void tcg_region_bounds(size_t curr_region, void **pstart, void **pend) if (curr_region == 0) { start = region.start; } + /* The final region may have a few extra pages due to earlier rounding. */ if (curr_region == region.n - 1) { - end = region.end; + end = region.start_aligned + region.total_size; } *pstart = start; @@ -XXX,XX +XXX,XX @@ static bool alloc_code_gen_buffer(size_t size, int splitwx, Error **errp) */ void tcg_region_init(size_t tb_size, int splitwx, unsigned max_cpus) { - void *buf, *aligned; - size_t size; + void *buf, *aligned, *end; + size_t total_size; size_t page_size; size_t region_size; size_t n_regions; @@ -XXX,XX +XXX,XX @@ void tcg_region_init(size_t tb_size, int splitwx, unsigned max_cpus) assert(ok); buf = tcg_init_ctx.code_gen_buffer; - size = tcg_init_ctx.code_gen_buffer_size; + total_size = tcg_init_ctx.code_gen_buffer_size; page_size = qemu_real_host_page_size; n_regions = tcg_n_regions(max_cpus); /* The first region will be 'aligned - buf' bytes larger than the others */ aligned = QEMU_ALIGN_PTR_UP(buf, page_size); - g_assert(aligned < tcg_init_ctx.code_gen_buffer + size); + g_assert(aligned < tcg_init_ctx.code_gen_buffer + total_size); + /* * Make region_size a multiple of page_size, using aligned as the start. * As a result of this we might end up with a few extra pages at the end of * the buffer; we will assign those to the last region. */ - region_size = (size - (aligned - buf)) / n_regions; + region_size = (total_size - (aligned - buf)) / n_regions; region_size = QEMU_ALIGN_DOWN(region_size, page_size); /* A region must have at least 2 pages; one code, one guard */ @@ -XXX,XX +XXX,XX @@ void tcg_region_init(size_t tb_size, int splitwx, unsigned max_cpus) region.start = buf; region.start_aligned = aligned; /* page-align the end, since its last page will be a guard page */ - region.end = QEMU_ALIGN_PTR_DOWN(buf + size, page_size); + end = QEMU_ALIGN_PTR_DOWN(buf + total_size, page_size); /* account for that last guard page */ - region.end -= page_size; + end -= page_size; + total_size = end - aligned; + region.total_size = total_size; /* * Set guard pages in the rw buffer, as that's the one into which @@ -XXX,XX +XXX,XX @@ void tcg_region_prologue_set(TCGContext *s) /* Register the balance of the buffer with gdb. */ tcg_register_jit(tcg_splitwx_to_rx(region.start), - region.end - region.start); + region.start_aligned + region.total_size - region.start); } /* @@ -XXX,XX +XXX,XX @@ size_t tcg_code_capacity(void) /* no need for synchronization; these variables are set at init time */ guard_size = region.stride - region.size; - capacity = region.end + guard_size - region.start; - capacity -= region.n * (guard_size + TCG_HIGHWATER); + capacity = region.total_size; + capacity -= (region.n - 1) * guard_size; + capacity -= region.n * TCG_HIGHWATER; + return capacity; } -- 2.25.1 Give the field a name reflecting its actual meaning. Reviewed-by: Luis Pires <luis.pires@eldorado.org.br> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/region.c | 15 ++++++++------- 1 file changed, 8 insertions(+), 7 deletions(-) diff --git a/tcg/region.c b/tcg/region.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/region.c +++ b/tcg/region.c @@ -XXX,XX +XXX,XX @@ struct tcg_region_state { QemuMutex lock; /* fields set at init time */ - void *start; void *start_aligned; + void *after_prologue; size_t n; size_t size; /* size of one region */ size_t stride; /* .size + guard size */ @@ -XXX,XX +XXX,XX @@ static void tcg_region_bounds(size_t curr_region, void **pstart, void **pend) end = start + region.size; if (curr_region == 0) { - start = region.start; + start = region.after_prologue; } /* The final region may have a few extra pages due to earlier rounding. */ if (curr_region == region.n - 1) { @@ -XXX,XX +XXX,XX @@ void tcg_region_init(size_t tb_size, int splitwx, unsigned max_cpus) region.n = n_regions; region.size = region_size - page_size; region.stride = region_size; - region.start = buf; + region.after_prologue = buf; region.start_aligned = aligned; /* page-align the end, since its last page will be a guard page */ end = QEMU_ALIGN_PTR_DOWN(buf + total_size, page_size); @@ -XXX,XX +XXX,XX @@ void tcg_region_init(size_t tb_size, int splitwx, unsigned max_cpus) void tcg_region_prologue_set(TCGContext *s) { /* Deduct the prologue from the first region. */ - g_assert(region.start == s->code_gen_buffer); - region.start = s->code_ptr; + g_assert(region.start_aligned == s->code_gen_buffer); + region.after_prologue = s->code_ptr; /* Recompute boundaries of the first region. */ tcg_region_assign(s, 0); /* Register the balance of the buffer with gdb. */ - tcg_register_jit(tcg_splitwx_to_rx(region.start), - region.start_aligned + region.total_size - region.start); + tcg_register_jit(tcg_splitwx_to_rx(region.after_prologue), + region.start_aligned + region.total_size - + region.after_prologue); } /* -- 2.25.1 Compute the value using straight division and bounds, rather than a loop. Pass in tb_size rather than reading from tcg_init_ctx.code_gen_buffer_size, Reviewed-by: Luis Pires <luis.pires@eldorado.org.br> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/region.c | 29 ++++++++++++----------------- 1 file changed, 12 insertions(+), 17 deletions(-) diff --git a/tcg/region.c b/tcg/region.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/region.c +++ b/tcg/region.c @@ -XXX,XX +XXX,XX @@ void tcg_region_reset_all(void) tcg_region_tree_reset_all(); } -static size_t tcg_n_regions(unsigned max_cpus) +static size_t tcg_n_regions(size_t tb_size, unsigned max_cpus) { #ifdef CONFIG_USER_ONLY return 1; #else + size_t n_regions; + /* * It is likely that some vCPUs will translate more code than others, * so we first try to set more regions than max_cpus, with those regions * being of reasonable size. If that's not possible we make do by evenly * dividing the code_gen_buffer among the vCPUs. */ - size_t i; - /* Use a single region if all we have is one vCPU thread */ if (max_cpus == 1 || !qemu_tcg_mttcg_enabled()) { return 1; } - /* Try to have more regions than max_cpus, with each region being >= 2 MB */ - for (i = 8; i > 0; i--) { - size_t regions_per_thread = i; - size_t region_size; - - region_size = tcg_init_ctx.code_gen_buffer_size; - region_size /= max_cpus * regions_per_thread; - - if (region_size >= 2 * 1024u * 1024) { - return max_cpus * regions_per_thread; - } + /* + * Try to have more regions than max_cpus, with each region being >= 2 MB. + * If we can't, then just allocate one region per vCPU thread. + */ + n_regions = tb_size / (2 * MiB); + if (n_regions <= max_cpus) { + return max_cpus; } - /* If we can't, then just allocate one region per vCPU thread */ - return max_cpus; + return MIN(n_regions, max_cpus * 8); #endif } @@ -XXX,XX +XXX,XX @@ void tcg_region_init(size_t tb_size, int splitwx, unsigned max_cpus) buf = tcg_init_ctx.code_gen_buffer; total_size = tcg_init_ctx.code_gen_buffer_size; page_size = qemu_real_host_page_size; - n_regions = tcg_n_regions(max_cpus); + n_regions = tcg_n_regions(total_size, max_cpus); /* The first region will be 'aligned - buf' bytes larger than the others */ aligned = QEMU_ALIGN_PTR_UP(buf, page_size); -- 2.25.1 Return output buffer and size via output pointer arguments, rather than returning size via tcg_ctx->code_gen_buffer_size. Reviewed-by: Luis Pires <luis.pires@eldorado.org.br> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/region.c | 19 +++++++++---------- 1 file changed, 9 insertions(+), 10 deletions(-) diff --git a/tcg/region.c b/tcg/region.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/region.c +++ b/tcg/region.c @@ -XXX,XX +XXX,XX @@ static inline bool cross_256mb(void *addr, size_t size) /* * We weren't able to allocate a buffer without crossing that boundary, * so make do with the larger portion of the buffer that doesn't cross. - * Returns the new base of the buffer, and adjusts code_gen_buffer_size. + * Returns the new base and size of the buffer in *obuf and *osize. */ -static inline void *split_cross_256mb(void *buf1, size_t size1) +static inline void split_cross_256mb(void **obuf, size_t *osize, + void *buf1, size_t size1) { void *buf2 = (void *)(((uintptr_t)buf1 + size1) & ~0x0ffffffful); size_t size2 = buf1 + size1 - buf2; @@ -XXX,XX +XXX,XX @@ static inline void *split_cross_256mb(void *buf1, size_t size1) buf1 = buf2; } - tcg_ctx->code_gen_buffer_size = size1; - return buf1; + *obuf = buf1; + *osize = size1; } #endif @@ -XXX,XX +XXX,XX @@ static bool alloc_code_gen_buffer(size_t tb_size, int splitwx, Error **errp) if (size > tb_size) { size = QEMU_ALIGN_DOWN(tb_size, qemu_real_host_page_size); } - tcg_ctx->code_gen_buffer_size = size; #ifdef __mips__ if (cross_256mb(buf, size)) { - buf = split_cross_256mb(buf, size); - size = tcg_ctx->code_gen_buffer_size; + split_cross_256mb(&buf, &size, buf, size); } #endif @@ -XXX,XX +XXX,XX @@ static bool alloc_code_gen_buffer(size_t tb_size, int splitwx, Error **errp) qemu_madvise(buf, size, QEMU_MADV_HUGEPAGE); tcg_ctx->code_gen_buffer = buf; + tcg_ctx->code_gen_buffer_size = size; return true; } #elif defined(_WIN32) @@ -XXX,XX +XXX,XX @@ static bool alloc_code_gen_buffer_anon(size_t size, int prot, "allocate %zu bytes for jit buffer", size); return false; } - tcg_ctx->code_gen_buffer_size = size; #ifdef __mips__ if (cross_256mb(buf, size)) { @@ -XXX,XX +XXX,XX @@ static bool alloc_code_gen_buffer_anon(size_t size, int prot, /* fallthru */ default: /* Split the original buffer. Free the smaller half. */ - buf2 = split_cross_256mb(buf, size); - size2 = tcg_ctx->code_gen_buffer_size; + split_cross_256mb(&buf2, &size2, buf, size); if (buf == buf2) { munmap(buf + size2, size - size2); } else { @@ -XXX,XX +XXX,XX @@ static bool alloc_code_gen_buffer_anon(size_t size, int prot, qemu_madvise(buf, size, QEMU_MADV_HUGEPAGE); tcg_ctx->code_gen_buffer = buf; + tcg_ctx->code_gen_buffer_size = size; return true; } -- 2.25.1 Shortly, the full code_gen_buffer will only be visible to region.c, so move in_code_gen_buffer out-of-line. Move the debugging versions of tcg_splitwx_to_{rx,rw} to region.c as well, so that the compiler gets to see the implementation of in_code_gen_buffer. This leaves exactly one use of in_code_gen_buffer outside of region.c, in cpu_restore_state. Which, being on the exception path, is not performance critical. Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Reviewed-by: Luis Pires <luis.pires@eldorado.org.br> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- include/tcg/tcg.h | 11 +---------- tcg/region.c | 34 ++++++++++++++++++++++++++++++++++ tcg/tcg.c | 23 ----------------------- 3 files changed, 35 insertions(+), 33 deletions(-) diff --git a/include/tcg/tcg.h b/include/tcg/tcg.h index XXXXXXX..XXXXXXX 100644 --- a/include/tcg/tcg.h +++ b/include/tcg/tcg.h @@ -XXX,XX +XXX,XX @@ extern const void *tcg_code_gen_epilogue; extern uintptr_t tcg_splitwx_diff; extern TCGv_env cpu_env; -static inline bool in_code_gen_buffer(const void *p) -{ - const TCGContext *s = &tcg_init_ctx; - /* - * Much like it is valid to have a pointer to the byte past the - * end of an array (so long as you don't dereference it), allow - * a pointer to the byte past the end of the code gen buffer. - */ - return (size_t)(p - s->code_gen_buffer) <= s->code_gen_buffer_size; -} +bool in_code_gen_buffer(const void *p); #ifdef CONFIG_DEBUG_TCG const void *tcg_splitwx_to_rx(void *rw); diff --git a/tcg/region.c b/tcg/region.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/region.c +++ b/tcg/region.c @@ -XXX,XX +XXX,XX @@ static struct tcg_region_state region; static void *region_trees; static size_t tree_size; +bool in_code_gen_buffer(const void *p) +{ + const TCGContext *s = &tcg_init_ctx; + /* + * Much like it is valid to have a pointer to the byte past the + * end of an array (so long as you don't dereference it), allow + * a pointer to the byte past the end of the code gen buffer. + */ + return (size_t)(p - s->code_gen_buffer) <= s->code_gen_buffer_size; +} + +#ifdef CONFIG_DEBUG_TCG +const void *tcg_splitwx_to_rx(void *rw) +{ + /* Pass NULL pointers unchanged. */ + if (rw) { + g_assert(in_code_gen_buffer(rw)); + rw += tcg_splitwx_diff; + } + return rw; +} + +void *tcg_splitwx_to_rw(const void *rx) +{ + /* Pass NULL pointers unchanged. */ + if (rx) { + rx -= tcg_splitwx_diff; + /* Assert that we end with a pointer in the rw region. */ + g_assert(in_code_gen_buffer(rx)); + } + return (void *)rx; +} +#endif /* CONFIG_DEBUG_TCG */ + /* compare a pointer @ptr and a tb_tc @s */ static int ptr_cmp_tb_tc(const void *ptr, const struct tb_tc *s) { diff --git a/tcg/tcg.c b/tcg/tcg.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/tcg.c +++ b/tcg/tcg.c @@ -XXX,XX +XXX,XX @@ static const TCGTargetOpDef constraint_sets[] = { #include "tcg-target.c.inc" -#ifdef CONFIG_DEBUG_TCG -const void *tcg_splitwx_to_rx(void *rw) -{ - /* Pass NULL pointers unchanged. */ - if (rw) { - g_assert(in_code_gen_buffer(rw)); - rw += tcg_splitwx_diff; - } - return rw; -} - -void *tcg_splitwx_to_rw(const void *rx) -{ - /* Pass NULL pointers unchanged. */ - if (rx) { - rx -= tcg_splitwx_diff; - /* Assert that we end with a pointer in the rw region. */ - g_assert(in_code_gen_buffer(rx)); - } - return (void *)rx; -} -#endif /* CONFIG_DEBUG_TCG */ - static void alloc_tcg_plugin_context(TCGContext *s) { #ifdef CONFIG_PLUGIN -- 2.25.1 Do not mess around with setting values within tcg_init_ctx. Put the values into 'region' directly, which is where they will live for the lifetime of the program. Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Reviewed-by: Luis Pires <luis.pires@eldorado.org.br> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/region.c | 64 ++++++++++++++++++++++------------------------------ 1 file changed, 27 insertions(+), 37 deletions(-) diff --git a/tcg/region.c b/tcg/region.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/region.c +++ b/tcg/region.c @@ -XXX,XX +XXX,XX @@ static size_t tree_size; bool in_code_gen_buffer(const void *p) { - const TCGContext *s = &tcg_init_ctx; /* * Much like it is valid to have a pointer to the byte past the * end of an array (so long as you don't dereference it), allow * a pointer to the byte past the end of the code gen buffer. */ - return (size_t)(p - s->code_gen_buffer) <= s->code_gen_buffer_size; + return (size_t)(p - region.start_aligned) <= region.total_size; } #ifdef CONFIG_DEBUG_TCG @@ -XXX,XX +XXX,XX @@ static bool alloc_code_gen_buffer(size_t tb_size, int splitwx, Error **errp) } qemu_madvise(buf, size, QEMU_MADV_HUGEPAGE); - tcg_ctx->code_gen_buffer = buf; - tcg_ctx->code_gen_buffer_size = size; + region.start_aligned = buf; + region.total_size = size; return true; } #elif defined(_WIN32) @@ -XXX,XX +XXX,XX @@ static bool alloc_code_gen_buffer(size_t size, int splitwx, Error **errp) return false; } - tcg_ctx->code_gen_buffer = buf; - tcg_ctx->code_gen_buffer_size = size; + region.start_aligned = buf; + region.total_size = size; return true; } #else @@ -XXX,XX +XXX,XX @@ static bool alloc_code_gen_buffer_anon(size_t size, int prot, /* Request large pages for the buffer. */ qemu_madvise(buf, size, QEMU_MADV_HUGEPAGE); - tcg_ctx->code_gen_buffer = buf; - tcg_ctx->code_gen_buffer_size = size; + region.start_aligned = buf; + region.total_size = size; return true; } @@ -XXX,XX +XXX,XX @@ static bool alloc_code_gen_buffer_splitwx_memfd(size_t size, Error **errp) return false; } /* The size of the mapping may have been adjusted. */ - size = tcg_ctx->code_gen_buffer_size; - buf_rx = tcg_ctx->code_gen_buffer; + buf_rx = region.start_aligned; + size = region.total_size; #endif buf_rw = qemu_memfd_alloc("tcg-jit", size, 0, &fd, errp); @@ -XXX,XX +XXX,XX @@ static bool alloc_code_gen_buffer_splitwx_memfd(size_t size, Error **errp) #endif close(fd); - tcg_ctx->code_gen_buffer = buf_rw; - tcg_ctx->code_gen_buffer_size = size; + region.start_aligned = buf_rw; + region.total_size = size; tcg_splitwx_diff = buf_rx - buf_rw; /* Request large pages for the buffer and the splitwx. */ @@ -XXX,XX +XXX,XX @@ static bool alloc_code_gen_buffer_splitwx_vmremap(size_t size, Error **errp) return false; } - buf_rw = (mach_vm_address_t)tcg_ctx->code_gen_buffer; + buf_rw = region.start_aligned; buf_rx = 0; ret = mach_vm_remap(mach_task_self(), &buf_rx, @@ -XXX,XX +XXX,XX @@ static bool alloc_code_gen_buffer(size_t size, int splitwx, Error **errp) */ void tcg_region_init(size_t tb_size, int splitwx, unsigned max_cpus) { - void *buf, *aligned, *end; - size_t total_size; size_t page_size; size_t region_size; - size_t n_regions; size_t i; bool ok; @@ -XXX,XX +XXX,XX @@ void tcg_region_init(size_t tb_size, int splitwx, unsigned max_cpus) splitwx, &error_fatal); assert(ok); - buf = tcg_init_ctx.code_gen_buffer; - total_size = tcg_init_ctx.code_gen_buffer_size; - page_size = qemu_real_host_page_size; - n_regions = tcg_n_regions(total_size, max_cpus); - - /* The first region will be 'aligned - buf' bytes larger than the others */ - aligned = QEMU_ALIGN_PTR_UP(buf, page_size); - g_assert(aligned < tcg_init_ctx.code_gen_buffer + total_size); - /* * Make region_size a multiple of page_size, using aligned as the start. * As a result of this we might end up with a few extra pages at the end of * the buffer; we will assign those to the last region. */ - region_size = (total_size - (aligned - buf)) / n_regions; + region.n = tcg_n_regions(region.total_size, max_cpus); + page_size = qemu_real_host_page_size; + region_size = region.total_size / region.n; region_size = QEMU_ALIGN_DOWN(region_size, page_size); /* A region must have at least 2 pages; one code, one guard */ g_assert(region_size >= 2 * page_size); + region.stride = region_size; + + /* Reserve space for guard pages. */ + region.size = region_size - page_size; + region.total_size -= page_size; + + /* + * The first region will be smaller than the others, via the prologue, + * which has yet to be allocated. For now, the first region begins at + * the page boundary. + */ + region.after_prologue = region.start_aligned; /* init the region struct */ qemu_mutex_init(&region.lock); - region.n = n_regions; - region.size = region_size - page_size; - region.stride = region_size; - region.after_prologue = buf; - region.start_aligned = aligned; - /* page-align the end, since its last page will be a guard page */ - end = QEMU_ALIGN_PTR_DOWN(buf + total_size, page_size); - /* account for that last guard page */ - end -= page_size; - total_size = end - aligned; - region.total_size = total_size; /* * Set guard pages in the rw buffer, as that's the one into which -- 2.25.1 Change the interface from a boolean error indication to a negative error vs a non-negative protection. For the moment this is only interface change, not making use of the new data. Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Reviewed-by: Luis Pires <luis.pires@eldorado.org.br> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/region.c | 63 +++++++++++++++++++++++++++------------------------- 1 file changed, 33 insertions(+), 30 deletions(-) diff --git a/tcg/region.c b/tcg/region.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/region.c +++ b/tcg/region.c @@ -XXX,XX +XXX,XX @@ static inline void split_cross_256mb(void **obuf, size_t *osize, static uint8_t static_code_gen_buffer[DEFAULT_CODE_GEN_BUFFER_SIZE] __attribute__((aligned(CODE_GEN_ALIGN))); -static bool alloc_code_gen_buffer(size_t tb_size, int splitwx, Error **errp) +static int alloc_code_gen_buffer(size_t tb_size, int splitwx, Error **errp) { void *buf, *end; size_t size; if (splitwx > 0) { error_setg(errp, "jit split-wx not supported"); - return false; + return -1; } /* page-align the beginning and end of the buffer */ @@ -XXX,XX +XXX,XX @@ static bool alloc_code_gen_buffer(size_t tb_size, int splitwx, Error **errp) region.start_aligned = buf; region.total_size = size; - return true; + + return PROT_READ | PROT_WRITE; } #elif defined(_WIN32) -static bool alloc_code_gen_buffer(size_t size, int splitwx, Error **errp) +static int alloc_code_gen_buffer(size_t size, int splitwx, Error **errp) { void *buf; if (splitwx > 0) { error_setg(errp, "jit split-wx not supported"); - return false; + return -1; } buf = VirtualAlloc(NULL, size, MEM_RESERVE | MEM_COMMIT, @@ -XXX,XX +XXX,XX @@ static bool alloc_code_gen_buffer(size_t size, int splitwx, Error **errp) region.start_aligned = buf; region.total_size = size; - return true; + + return PAGE_READ | PAGE_WRITE | PAGE_EXEC; } #else -static bool alloc_code_gen_buffer_anon(size_t size, int prot, - int flags, Error **errp) +static int alloc_code_gen_buffer_anon(size_t size, int prot, + int flags, Error **errp) { void *buf; @@ -XXX,XX +XXX,XX @@ static bool alloc_code_gen_buffer_anon(size_t size, int prot, if (buf == MAP_FAILED) { error_setg_errno(errp, errno, "allocate %zu bytes for jit buffer", size); - return false; + return -1; } #ifdef __mips__ @@ -XXX,XX +XXX,XX @@ static bool alloc_code_gen_buffer_anon(size_t size, int prot, region.start_aligned = buf; region.total_size = size; - return true; + return prot; } #ifndef CONFIG_TCG_INTERPRETER @@ -XXX,XX +XXX,XX @@ static bool alloc_code_gen_buffer_splitwx_memfd(size_t size, Error **errp) #ifdef __mips__ /* Find space for the RX mapping, vs the 256MiB regions. */ - if (!alloc_code_gen_buffer_anon(size, PROT_NONE, - MAP_PRIVATE | MAP_ANONYMOUS | - MAP_NORESERVE, errp)) { + if (alloc_code_gen_buffer_anon(size, PROT_NONE, + MAP_PRIVATE | MAP_ANONYMOUS | + MAP_NORESERVE, errp) < 0) { return false; } /* The size of the mapping may have been adjusted. */ @@ -XXX,XX +XXX,XX @@ static bool alloc_code_gen_buffer_splitwx_memfd(size_t size, Error **errp) /* Request large pages for the buffer and the splitwx. */ qemu_madvise(buf_rw, size, QEMU_MADV_HUGEPAGE); qemu_madvise(buf_rx, size, QEMU_MADV_HUGEPAGE); - return true; + return PROT_READ | PROT_WRITE; fail_rx: error_setg_errno(errp, errno, "failed to map shared memory for execute"); @@ -XXX,XX +XXX,XX @@ static bool alloc_code_gen_buffer_splitwx_memfd(size_t size, Error **errp) if (fd >= 0) { close(fd); } - return false; + return -1; } #endif /* CONFIG_POSIX */ @@ -XXX,XX +XXX,XX @@ extern kern_return_t mach_vm_remap(vm_map_t target_task, vm_prot_t *max_protection, vm_inherit_t inheritance); -static bool alloc_code_gen_buffer_splitwx_vmremap(size_t size, Error **errp) +static int alloc_code_gen_buffer_splitwx_vmremap(size_t size, Error **errp) { kern_return_t ret; mach_vm_address_t buf_rw, buf_rx; @@ -XXX,XX +XXX,XX @@ static bool alloc_code_gen_buffer_splitwx_vmremap(size_t size, Error **errp) /* Map the read-write portion via normal anon memory. */ if (!alloc_code_gen_buffer_anon(size, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, errp)) { - return false; + return -1; } buf_rw = region.start_aligned; @@ -XXX,XX +XXX,XX @@ static bool alloc_code_gen_buffer_splitwx_vmremap(size_t size, Error **errp) /* TODO: Convert "ret" to a human readable error message. */ error_setg(errp, "vm_remap for jit splitwx failed"); munmap((void *)buf_rw, size); - return false; + return -1; } if (mprotect((void *)buf_rx, size, PROT_READ | PROT_EXEC) != 0) { error_setg_errno(errp, errno, "mprotect for jit splitwx"); munmap((void *)buf_rx, size); munmap((void *)buf_rw, size); - return false; + return -1; } tcg_splitwx_diff = buf_rx - buf_rw; - return true; + return PROT_READ | PROT_WRITE; } #endif /* CONFIG_DARWIN */ #endif /* CONFIG_TCG_INTERPRETER */ -static bool alloc_code_gen_buffer_splitwx(size_t size, Error **errp) +static int alloc_code_gen_buffer_splitwx(size_t size, Error **errp) { #ifndef CONFIG_TCG_INTERPRETER # ifdef CONFIG_DARWIN @@ -XXX,XX +XXX,XX @@ static bool alloc_code_gen_buffer_splitwx(size_t size, Error **errp) # endif #endif error_setg(errp, "jit split-wx not supported"); - return false; + return -1; } -static bool alloc_code_gen_buffer(size_t size, int splitwx, Error **errp) +static int alloc_code_gen_buffer(size_t size, int splitwx, Error **errp) { ERRP_GUARD(); int prot, flags; if (splitwx) { - if (alloc_code_gen_buffer_splitwx(size, errp)) { - return true; + prot = alloc_code_gen_buffer_splitwx(size, errp); + if (prot >= 0) { + return prot; } /* * If splitwx force-on (1), fail; * if splitwx default-on (-1), fall through to splitwx off. */ if (splitwx > 0) { - return false; + return -1; } error_free_or_abort(errp); } @@ -XXX,XX +XXX,XX @@ void tcg_region_init(size_t tb_size, int splitwx, unsigned max_cpus) size_t page_size; size_t region_size; size_t i; - bool ok; + int have_prot; - ok = alloc_code_gen_buffer(size_code_gen_buffer(tb_size), - splitwx, &error_fatal); - assert(ok); + have_prot = alloc_code_gen_buffer(size_code_gen_buffer(tb_size), + splitwx, &error_fatal); + assert(have_prot >= 0); /* * Make region_size a multiple of page_size, using aligned as the start. -- 2.25.1 Move the call out of the N versions of alloc_code_gen_buffer and into tcg_region_init. Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Reviewed-by: Luis Pires <luis.pires@eldorado.org.br> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/region.c | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-) diff --git a/tcg/region.c b/tcg/region.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/region.c +++ b/tcg/region.c @@ -XXX,XX +XXX,XX @@ static int alloc_code_gen_buffer(size_t tb_size, int splitwx, Error **errp) error_setg_errno(errp, errno, "mprotect of jit buffer"); return false; } - qemu_madvise(buf, size, QEMU_MADV_HUGEPAGE); region.start_aligned = buf; region.total_size = size; @@ -XXX,XX +XXX,XX @@ static int alloc_code_gen_buffer_anon(size_t size, int prot, } #endif - /* Request large pages for the buffer. */ - qemu_madvise(buf, size, QEMU_MADV_HUGEPAGE); - region.start_aligned = buf; region.total_size = size; return prot; @@ -XXX,XX +XXX,XX @@ static bool alloc_code_gen_buffer_splitwx_memfd(size_t size, Error **errp) region.total_size = size; tcg_splitwx_diff = buf_rx - buf_rw; - /* Request large pages for the buffer and the splitwx. */ - qemu_madvise(buf_rw, size, QEMU_MADV_HUGEPAGE); - qemu_madvise(buf_rx, size, QEMU_MADV_HUGEPAGE); return PROT_READ | PROT_WRITE; fail_rx: @@ -XXX,XX +XXX,XX @@ void tcg_region_init(size_t tb_size, int splitwx, unsigned max_cpus) splitwx, &error_fatal); assert(have_prot >= 0); + /* Request large pages for the buffer and the splitwx. */ + qemu_madvise(region.start_aligned, region.total_size, QEMU_MADV_HUGEPAGE); + if (tcg_splitwx_diff) { + qemu_madvise(region.start_aligned + tcg_splitwx_diff, + region.total_size, QEMU_MADV_HUGEPAGE); + } + /* * Make region_size a multiple of page_size, using aligned as the start. * As a result of this we might end up with a few extra pages at the end of -- 2.25.1 For --enable-tcg-interpreter on Windows, we will need this. Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Reviewed-by: Luis Pires <luis.pires@eldorado.org.br> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- include/qemu/osdep.h | 1 + util/osdep.c | 9 +++++++++ 2 files changed, 10 insertions(+) diff --git a/include/qemu/osdep.h b/include/qemu/osdep.h index XXXXXXX..XXXXXXX 100644 --- a/include/qemu/osdep.h +++ b/include/qemu/osdep.h @@ -XXX,XX +XXX,XX @@ void sigaction_invoke(struct sigaction *action, #endif int qemu_madvise(void *addr, size_t len, int advice); +int qemu_mprotect_rw(void *addr, size_t size); int qemu_mprotect_rwx(void *addr, size_t size); int qemu_mprotect_none(void *addr, size_t size); diff --git a/util/osdep.c b/util/osdep.c index XXXXXXX..XXXXXXX 100644 --- a/util/osdep.c +++ b/util/osdep.c @@ -XXX,XX +XXX,XX @@ static int qemu_mprotect__osdep(void *addr, size_t size, int prot) #endif } +int qemu_mprotect_rw(void *addr, size_t size) +{ +#ifdef _WIN32 + return qemu_mprotect__osdep(addr, size, PAGE_READWRITE); +#else + return qemu_mprotect__osdep(addr, size, PROT_READ | PROT_WRITE); +#endif +} + int qemu_mprotect_rwx(void *addr, size_t size) { #ifdef _WIN32 -- 2.25.1 If qemu_get_host_physmem returns an odd number of pages, then physmem / 8 will not be a multiple of the page size. The following was observed on a gitlab runner: ERROR qtest-arm/boot-serial-test - Bail out! ERROR:../util/osdep.c:80:qemu_mprotect__osdep: \ assertion failed: (!(size & ~qemu_real_host_page_mask)) Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Reviewed-by: Luis Pires <luis.pires@eldorado.org.br> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/region.c | 47 +++++++++++++++++++++-------------------------- 1 file changed, 21 insertions(+), 26 deletions(-) diff --git a/tcg/region.c b/tcg/region.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/region.c +++ b/tcg/region.c @@ -XXX,XX +XXX,XX @@ static size_t tcg_n_regions(size_t tb_size, unsigned max_cpus) (DEFAULT_CODE_GEN_BUFFER_SIZE_1 < MAX_CODE_GEN_BUFFER_SIZE \ ? DEFAULT_CODE_GEN_BUFFER_SIZE_1 : MAX_CODE_GEN_BUFFER_SIZE) -static size_t size_code_gen_buffer(size_t tb_size) -{ - /* Size the buffer. */ - if (tb_size == 0) { - size_t phys_mem = qemu_get_host_physmem(); - if (phys_mem == 0) { - tb_size = DEFAULT_CODE_GEN_BUFFER_SIZE; - } else { - tb_size = MIN(DEFAULT_CODE_GEN_BUFFER_SIZE, phys_mem / 8); - } - } - if (tb_size < MIN_CODE_GEN_BUFFER_SIZE) { - tb_size = MIN_CODE_GEN_BUFFER_SIZE; - } - if (tb_size > MAX_CODE_GEN_BUFFER_SIZE) { - tb_size = MAX_CODE_GEN_BUFFER_SIZE; - } - return tb_size; -} - #ifdef __mips__ /* * In order to use J and JAL within the code_gen_buffer, we require @@ -XXX,XX +XXX,XX @@ static int alloc_code_gen_buffer(size_t size, int splitwx, Error **errp) */ void tcg_region_init(size_t tb_size, int splitwx, unsigned max_cpus) { - size_t page_size; + const size_t page_size = qemu_real_host_page_size; size_t region_size; size_t i; int have_prot; - have_prot = alloc_code_gen_buffer(size_code_gen_buffer(tb_size), - splitwx, &error_fatal); + /* Size the buffer. */ + if (tb_size == 0) { + size_t phys_mem = qemu_get_host_physmem(); + if (phys_mem == 0) { + tb_size = DEFAULT_CODE_GEN_BUFFER_SIZE; + } else { + tb_size = QEMU_ALIGN_DOWN(phys_mem / 8, page_size); + tb_size = MIN(DEFAULT_CODE_GEN_BUFFER_SIZE, tb_size); + } + } + if (tb_size < MIN_CODE_GEN_BUFFER_SIZE) { + tb_size = MIN_CODE_GEN_BUFFER_SIZE; + } + if (tb_size > MAX_CODE_GEN_BUFFER_SIZE) { + tb_size = MAX_CODE_GEN_BUFFER_SIZE; + } + + have_prot = alloc_code_gen_buffer(tb_size, splitwx, &error_fatal); assert(have_prot >= 0); /* Request large pages for the buffer and the splitwx. */ @@ -XXX,XX +XXX,XX @@ void tcg_region_init(size_t tb_size, int splitwx, unsigned max_cpus) * As a result of this we might end up with a few extra pages at the end of * the buffer; we will assign those to the last region. */ - region.n = tcg_n_regions(region.total_size, max_cpus); - page_size = qemu_real_host_page_size; - region_size = region.total_size / region.n; + region.n = tcg_n_regions(tb_size, max_cpus); + region_size = tb_size / region.n; region_size = QEMU_ALIGN_DOWN(region_size, page_size); /* A region must have at least 2 pages; one code, one guard */ -- 2.25.1 Do not handle protections on a case-by-case basis in the various alloc_code_gen_buffer instances; do it within a single loop in tcg_region_init. Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Reviewed-by: Luis Pires <luis.pires@eldorado.org.br> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/region.c | 45 +++++++++++++++++++++++++++++++-------------- 1 file changed, 31 insertions(+), 14 deletions(-) diff --git a/tcg/region.c b/tcg/region.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/region.c +++ b/tcg/region.c @@ -XXX,XX +XXX,XX @@ static int alloc_code_gen_buffer(size_t tb_size, int splitwx, Error **errp) } #endif - if (qemu_mprotect_rwx(buf, size)) { - error_setg_errno(errp, errno, "mprotect of jit buffer"); - return false; - } - region.start_aligned = buf; region.total_size = size; @@ -XXX,XX +XXX,XX @@ void tcg_region_init(size_t tb_size, int splitwx, unsigned max_cpus) { const size_t page_size = qemu_real_host_page_size; size_t region_size; - size_t i; - int have_prot; + int have_prot, need_prot; /* Size the buffer. */ if (tb_size == 0) { @@ -XXX,XX +XXX,XX @@ void tcg_region_init(size_t tb_size, int splitwx, unsigned max_cpus) * Set guard pages in the rw buffer, as that's the one into which * buffer overruns could occur. Do not set guard pages in the rx * buffer -- let that one use hugepages throughout. + * Work with the page protections set up with the initial mapping. */ - for (i = 0; i < region.n; i++) { + need_prot = PAGE_READ | PAGE_WRITE; +#ifndef CONFIG_TCG_INTERPRETER + if (tcg_splitwx_diff == 0) { + need_prot |= PAGE_EXEC; + } +#endif + for (size_t i = 0, n = region.n; i < n; i++) { void *start, *end; tcg_region_bounds(i, &start, &end); + if (have_prot != need_prot) { + int rc; - /* - * macOS 11.2 has a bug (Apple Feedback FB8994773) in which mprotect - * rejects a permission change from RWX -> NONE. Guard pages are - * nice for bug detection but are not essential; ignore any failure. - */ - (void)qemu_mprotect_none(end, page_size); + if (need_prot == (PAGE_READ | PAGE_WRITE | PAGE_EXEC)) { + rc = qemu_mprotect_rwx(start, end - start); + } else if (need_prot == (PAGE_READ | PAGE_WRITE)) { + rc = qemu_mprotect_rw(start, end - start); + } else { + g_assert_not_reached(); + } + if (rc) { + error_setg_errno(&error_fatal, errno, + "mprotect of jit buffer"); + } + } + if (have_prot != 0) { + /* + * macOS 11.2 has a bug (Apple Feedback FB8994773) in which mprotect + * rejects a permission change from RWX -> NONE. Guard pages are + * nice for bug detection but are not essential; ignore any failure. + */ + (void)qemu_mprotect_none(end, page_size); + } } tcg_region_trees_init(); -- 2.25.1 There's a change in mprotect() behaviour [1] in the latest macOS on M1 and it's not yet clear if it's going to be fixed by Apple. In this case, instead of changing permissions of N guard pages, we change permissions of N rwx regions. The same number of syscalls are required either way. [1] https://gist.github.com/hikalium/75ae822466ee4da13cbbe486498a191f Reviewed-by: Luis Pires <luis.pires@eldorado.org.br> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/region.c | 19 +++++++++---------- 1 file changed, 9 insertions(+), 10 deletions(-) diff --git a/tcg/region.c b/tcg/region.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/region.c +++ b/tcg/region.c @@ -XXX,XX +XXX,XX @@ static int alloc_code_gen_buffer(size_t size, int splitwx, Error **errp) error_free_or_abort(errp); } - prot = PROT_READ | PROT_WRITE | PROT_EXEC; + /* + * macOS 11.2 has a bug (Apple Feedback FB8994773) in which mprotect + * rejects a permission change from RWX -> NONE when reserving the + * guard pages later. We can go the other way with the same number + * of syscalls, so always begin with PROT_NONE. + */ + prot = PROT_NONE; flags = MAP_PRIVATE | MAP_ANONYMOUS; -#ifdef CONFIG_TCG_INTERPRETER - /* The tcg interpreter does not need execute permission. */ - prot = PROT_READ | PROT_WRITE; -#elif defined(CONFIG_DARWIN) +#ifdef CONFIG_DARWIN /* Applicable to both iOS and macOS (Apple Silicon). */ if (!splitwx) { flags |= MAP_JIT; @@ -XXX,XX +XXX,XX @@ void tcg_region_init(size_t tb_size, int splitwx, unsigned max_cpus) } } if (have_prot != 0) { - /* - * macOS 11.2 has a bug (Apple Feedback FB8994773) in which mprotect - * rejects a permission change from RWX -> NONE. Guard pages are - * nice for bug detection but are not essential; ignore any failure. - */ + /* Guard pages are nice for bug detection but are not essential. */ (void)qemu_mprotect_none(end, page_size); } } -- 2.25.1 These variables belong to the jit side, not the user side. Since tcg_init_ctx is no longer used outside of tcg/, move the declaration to tcg-internal.h. Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Reviewed-by: Luis Pires <luis.pires@eldorado.org.br> Suggested-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- include/tcg/tcg.h | 1 - tcg/tcg-internal.h | 1 + accel/tcg/translate-all.c | 3 --- tcg/tcg.c | 3 +++ 4 files changed, 4 insertions(+), 4 deletions(-) diff --git a/include/tcg/tcg.h b/include/tcg/tcg.h index XXXXXXX..XXXXXXX 100644 --- a/include/tcg/tcg.h +++ b/include/tcg/tcg.h @@ -XXX,XX +XXX,XX @@ static inline bool temp_readonly(TCGTemp *ts) return ts->kind >= TEMP_FIXED; } -extern TCGContext tcg_init_ctx; extern __thread TCGContext *tcg_ctx; extern const void *tcg_code_gen_epilogue; extern uintptr_t tcg_splitwx_diff; diff --git a/tcg/tcg-internal.h b/tcg/tcg-internal.h index XXXXXXX..XXXXXXX 100644 --- a/tcg/tcg-internal.h +++ b/tcg/tcg-internal.h @@ -XXX,XX +XXX,XX @@ #define TCG_HIGHWATER 1024 +extern TCGContext tcg_init_ctx; extern TCGContext **tcg_ctxs; extern unsigned int tcg_cur_ctxs; extern unsigned int tcg_max_ctxs; 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 v_l2_levels; static void *l1_map[V_L1_MAX_SIZE]; -/* code generation context */ -TCGContext tcg_init_ctx; -__thread TCGContext *tcg_ctx; TBContext tb_ctx; static void page_table_config_init(void) diff --git a/tcg/tcg.c b/tcg/tcg.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/tcg.c +++ b/tcg/tcg.c @@ -XXX,XX +XXX,XX @@ static bool tcg_target_const_match(int64_t val, TCGType type, int ct); static int tcg_out_ldst_finalize(TCGContext *s); #endif +TCGContext tcg_init_ctx; +__thread TCGContext *tcg_ctx; + TCGContext **tcg_ctxs; unsigned int tcg_cur_ctxs; unsigned int tcg_max_ctxs; -- 2.25.1 Introduce a function to remove everything emitted since a given point. Reviewed-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- include/tcg/tcg.h | 10 ++++++++++ tcg/tcg.c | 13 +++++++++++++ 2 files changed, 23 insertions(+) diff --git a/include/tcg/tcg.h b/include/tcg/tcg.h index XXXXXXX..XXXXXXX 100644 --- a/include/tcg/tcg.h +++ b/include/tcg/tcg.h @@ -XXX,XX +XXX,XX @@ void tcg_op_remove(TCGContext *s, TCGOp *op); TCGOp *tcg_op_insert_before(TCGContext *s, TCGOp *op, TCGOpcode opc); TCGOp *tcg_op_insert_after(TCGContext *s, TCGOp *op, TCGOpcode opc); +/** + * tcg_remove_ops_after: + * @op: target operation + * + * Discard any opcodes emitted since @op. Expected usage is to save + * a starting point with tcg_last_op(), speculatively emit opcodes, + * then decide whether or not to keep those opcodes after the fact. + */ +void tcg_remove_ops_after(TCGOp *op); + void tcg_optimize(TCGContext *s); /* Allocate a new temporary and initialize it with a constant. */ diff --git a/tcg/tcg.c b/tcg/tcg.c index XXXXXXX..XXXXXXX 100644 --- a/tcg/tcg.c +++ b/tcg/tcg.c @@ -XXX,XX +XXX,XX @@ void tcg_op_remove(TCGContext *s, TCGOp *op) #endif } +void tcg_remove_ops_after(TCGOp *op) +{ + TCGContext *s = tcg_ctx; + + while (true) { + TCGOp *last = tcg_last_op(); + if (last == op) { + return; + } + tcg_op_remove(s, last); + } +} + static TCGOp *tcg_op_alloc(TCGOpcode opc) { TCGContext *s = tcg_ctx; -- 2.25.1 At some point during the development of tcg_constant_*, I changed my mind about whether such temps should be able to be passed to tcg_temp_free_*. The final version committed allows this, but the commentary was not updated to match. Fixes: c0522136adf Reported-by: Peter Maydell <peter.maydell@linaro.org> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Reviewed-by: Luis Pires <luis.pires@eldorado.org.br> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- include/tcg/tcg.h | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/include/tcg/tcg.h b/include/tcg/tcg.h index XXXXXXX..XXXXXXX 100644 --- a/include/tcg/tcg.h +++ b/include/tcg/tcg.h @@ -XXX,XX +XXX,XX @@ TCGv_vec tcg_const_ones_vec_matching(TCGv_vec); /* * Locate or create a read-only temporary that is a constant. - * This kind of temporary need not and should not be freed. + * This kind of temporary need not be freed, but for convenience + * will be silently ignored by tcg_temp_free_*. */ TCGTemp *tcg_constant_internal(TCGType type, int64_t val); -- 2.25.1 From: "Jose R. Ziviani" <jziviani@suse.de> Commit 5e8892db93 fixed several function signatures but tcg_out_op for arm is missing. This patch fixes it as well. Signed-off-by: Jose R. Ziviani <jziviani@suse.de> Message-Id: <20210610224450.23425-1-jziviani@suse.de> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- tcg/arm/tcg-target.c.inc | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) 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_qemu_st(TCGContext *s, const TCGArg *args, bool is64) static void tcg_out_epilogue(TCGContext *s); static inline void tcg_out_op(TCGContext *s, TCGOpcode opc, - const TCGArg *args, const int *const_args) + const TCGArg args[TCG_MAX_OP_ARGS], + const int const_args[TCG_MAX_OP_ARGS]) { TCGArg a0, a1, a2, a3, a4, a5; int c; -- 2.25.1 Typo in the conversion to FloatParts64. Fixes: 572c4d862ff2 Fixes: Coverity CID 1457457 Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Message-Id: <20210607223812.110596-1-richard.henderson@linaro.org> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- fpu/softfloat.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) 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 float32_exp2(float32 a, float_status *status) float_raise(float_flag_inexact, status); - float64_unpack_canonical(&xnp, float64_ln2, status); + float64_unpack_canonical(&tp, float64_ln2, status); xp = *parts_mul(&xp, &tp, status); xnp = xp; -- 2.25.1 From: Luis Pires <luis.pires@eldorado.org.br> Signed-off-by: Luis Pires <luis.pires@eldorado.org.br> Message-Id: <20210601125143.191165-1-luis.pires@eldorado.org.br> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- docs/devel/tcg.rst | 101 ++++++++++++++++++++++++++++++++++++++++----- 1 file changed, 90 insertions(+), 11 deletions(-) diff --git a/docs/devel/tcg.rst b/docs/devel/tcg.rst index XXXXXXX..XXXXXXX 100644 --- a/docs/devel/tcg.rst +++ b/docs/devel/tcg.rst @@ -XXX,XX +XXX,XX @@ performances. QEMU's dynamic translation backend is called TCG, for "Tiny Code Generator". For more information, please take a look at ``tcg/README``. -Some notable features of QEMU's dynamic translator are: +The following sections outline some notable features and implementation +details of QEMU's dynamic translator. CPU state optimisations ----------------------- -The target CPUs have many internal states which change the way it -evaluates instructions. In order to achieve a good speed, the +The target CPUs have many internal states which change the way they +evaluate instructions. In order to achieve a good speed, the translation phase considers that some state information of the virtual CPU cannot change in it. The state is recorded in the Translation Block (TB). If the state changes (e.g. privilege level), a new TB will @@ -XXX,XX +XXX,XX @@ Direct block chaining --------------------- After each translated basic block is executed, QEMU uses the simulated -Program Counter (PC) and other cpu state information (such as the CS +Program Counter (PC) and other CPU state information (such as the CS segment base value) to find the next basic block. -In order to accelerate the most common cases where the new simulated PC -is known, QEMU can patch a basic block so that it jumps directly to the -next one. +In its simplest, less optimized form, this is done by exiting from the +current TB, going through the TB epilogue, and then back to the +main loop. That’s where QEMU looks for the next TB to execute, +translating it from the guest architecture if it isn’t already available +in memory. Then QEMU proceeds to execute this next TB, starting at the +prologue and then moving on to the translated instructions. -The most portable code uses an indirect jump. An indirect jump makes -it easier to make the jump target modification atomic. On some host -architectures (such as x86 or PowerPC), the ``JUMP`` opcode is -directly patched so that the block chaining has no overhead. +Exiting from the TB this way will cause the ``cpu_exec_interrupt()`` +callback to be re-evaluated before executing additional instructions. +It is mandatory to exit this way after any CPU state changes that may +unmask interrupts. + +In order to accelerate the cases where the TB for the new +simulated PC is already available, QEMU has mechanisms that allow +multiple TBs to be chained directly, without having to go back to the +main loop as described above. These mechanisms are: + +``lookup_and_goto_ptr`` +^^^^^^^^^^^^^^^^^^^^^^^ + +Calling ``tcg_gen_lookup_and_goto_ptr()`` will emit a call to +``helper_lookup_tb_ptr``. This helper will look for an existing TB that +matches the current CPU state. If the destination TB is available its +code address is returned, otherwise the address of the JIT epilogue is +returned. The call to the helper is always followed by the tcg ``goto_ptr`` +opcode, which branches to the returned address. In this way, we either +branch to the next TB or return to the main loop. + +``goto_tb + exit_tb`` +^^^^^^^^^^^^^^^^^^^^^ + +The translation code usually implements branching by performing the +following steps: + +1. Call ``tcg_gen_goto_tb()`` passing a jump slot index (either 0 or 1) + as a parameter. + +2. Emit TCG instructions to update the CPU state with any information + that has been assumed constant and is required by the main loop to + correctly locate and execute the next TB. For most guests, this is + just the PC of the branch destination, but others may store additional + data. The information updated in this step must be inferable from both + ``cpu_get_tb_cpu_state()`` and ``cpu_restore_state()``. + +3. Call ``tcg_gen_exit_tb()`` passing the address of the current TB and + the jump slot index again. + +Step 1, ``tcg_gen_goto_tb()``, will emit a ``goto_tb`` TCG +instruction that later on gets translated to a jump to an address +associated with the specified jump slot. Initially, this is the address +of step 2's instructions, which update the CPU state information. Step 3, +``tcg_gen_exit_tb()``, exits from the current TB returning a tagged +pointer composed of the last executed TB’s address and the jump slot +index. + +The first time this whole sequence is executed, step 1 simply jumps +to step 2. Then the CPU state information gets updated and we exit from +the current TB. As a result, the behavior is very similar to the less +optimized form described earlier in this section. + +Next, the main loop looks for the next TB to execute using the +current CPU state information (creating the TB if it wasn’t already +available) and, before starting to execute the new TB’s instructions, +patches the previously executed TB by associating one of its jump +slots (the one specified in the call to ``tcg_gen_exit_tb()``) with the +address of the new TB. + +The next time this previous TB is executed and we get to that same +``goto_tb`` step, it will already be patched (assuming the destination TB +is still in memory) and will jump directly to the first instruction of +the destination TB, without going back to the main loop. + +For the ``goto_tb + exit_tb`` mechanism to be used, the following +conditions need to be satisfied: + +* The change in CPU state must be constant, e.g., a direct branch and + not an indirect branch. + +* The direct branch cannot cross a page boundary. Memory mappings + may change, causing the code at the destination address to change. + +Note that, on step 3 (``tcg_gen_exit_tb()``), in addition to the +jump slot index, the address of the TB just executed is also returned. +This address corresponds to the TB that will be patched; it may be +different than the one that was directly executed from the main loop +if the latter had already been chained to other TBs. Self-modifying code and translated code invalidation ---------------------------------------------------- -- 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