#endif
#include "qemu-common.h"
-#include "cache-utils.h"
-#include "host-utils.h"
-#include "qemu-timer.h"
+#include "qemu/cache-utils.h"
+#include "qemu/host-utils.h"
+#include "qemu/timer.h"
/* Note: the long term plan is to reduce the dependancies on the QEMU
CPU definitions. Currently they are used for qemu_ld/st
#include "elf.h"
-#if defined(CONFIG_USE_GUEST_BASE) && !defined(TCG_TARGET_HAS_GUEST_BASE)
-#error GUEST_BASE not supported on this host.
-#endif
-
/* Forward declarations for functions declared in tcg-target.c and used here. */
static void tcg_target_init(TCGContext *s);
static void tcg_target_qemu_prologue(TCGContext *s);
static void patch_reloc(uint8_t *code_ptr, int type,
tcg_target_long value, tcg_target_long addend);
+/* The CIE and FDE header definitions will be common to all hosts. */
+typedef struct {
+ uint32_t len __attribute__((aligned((sizeof(void *)))));
+ uint32_t id;
+ uint8_t version;
+ char augmentation[1];
+ uint8_t code_align;
+ uint8_t data_align;
+ uint8_t return_column;
+} DebugFrameCIE;
+
+typedef struct QEMU_PACKED {
+ uint32_t len __attribute__((aligned((sizeof(void *)))));
+ uint32_t cie_offset;
+ tcg_target_long func_start;
+ tcg_target_long func_len;
+} DebugFrameFDEHeader;
+
static void tcg_register_jit_int(void *buf, size_t size,
void *debug_frame, size_t debug_frame_size)
__attribute__((unused));
static TCGRegSet tcg_target_available_regs[2];
static TCGRegSet tcg_target_call_clobber_regs;
-/* XXX: move that inside the context */
-uint16_t *gen_opc_ptr;
-TCGArg *gen_opparam_ptr;
-
static inline void tcg_out8(TCGContext *s, uint8_t v)
{
*s->code_ptr++ = v;
static inline void tcg_out16(TCGContext *s, uint16_t v)
{
- *(uint16_t *)s->code_ptr = v;
- s->code_ptr += 2;
+ uint8_t *p = s->code_ptr;
+ *(uint16_t *)p = v;
+ s->code_ptr = p + 2;
}
static inline void tcg_out32(TCGContext *s, uint32_t v)
{
- *(uint32_t *)s->code_ptr = v;
- s->code_ptr += 4;
+ uint8_t *p = s->code_ptr;
+ *(uint32_t *)p = v;
+ s->code_ptr = p + 4;
+}
+
+static inline void tcg_out64(TCGContext *s, uint64_t v)
+{
+ uint8_t *p = s->code_ptr;
+ *(uint64_t *)p = v;
+ s->code_ptr = p + 8;
}
/* label relocation processing */
int *sorted_args;
memset(s, 0, sizeof(*s));
- s->temps = s->static_temps;
s->nb_globals = 0;
/* Count total number of arguments and allocate the corresponding
void tcg_prologue_init(TCGContext *s)
{
/* init global prologue and epilogue */
- s->code_buf = code_gen_prologue;
+ s->code_buf = s->code_gen_prologue;
s->code_ptr = s->code_buf;
tcg_target_qemu_prologue(s);
- flush_icache_range((tcg_target_ulong)s->code_buf,
- (tcg_target_ulong)s->code_ptr);
+ flush_icache_range((uintptr_t)s->code_buf, (uintptr_t)s->code_ptr);
+
+#ifdef DEBUG_DISAS
+ if (qemu_loglevel_mask(CPU_LOG_TB_OUT_ASM)) {
+ size_t size = s->code_ptr - s->code_buf;
+ qemu_log("PROLOGUE: [size=%zu]\n", size);
+ log_disas(s->code_buf, size);
+ qemu_log("\n");
+ qemu_log_flush();
+ }
+#endif
}
-void tcg_set_frame(TCGContext *s, int reg,
- tcg_target_long start, tcg_target_long size)
+void tcg_set_frame(TCGContext *s, int reg, intptr_t start, intptr_t size)
{
s->frame_start = start;
s->frame_end = start + size;
s->goto_tb_issue_mask = 0;
#endif
- gen_opc_ptr = gen_opc_buf;
- gen_opparam_ptr = gen_opparam_buf;
+ s->gen_opc_ptr = s->gen_opc_buf;
+ s->gen_opparam_ptr = s->gen_opparam_buf;
+
+#if defined(CONFIG_QEMU_LDST_OPTIMIZATION) && defined(CONFIG_SOFTMMU)
+ /* Initialize qemu_ld/st labels to assist code generation at the end of TB
+ for TLB miss cases at the end of TB */
+ s->qemu_ldst_labels = tcg_malloc(sizeof(TCGLabelQemuLdst) *
+ TCG_MAX_QEMU_LDST);
+ s->nb_qemu_ldst_labels = 0;
+#endif
}
static inline void tcg_temp_alloc(TCGContext *s, int n)
}
static inline int tcg_global_mem_new_internal(TCGType type, int reg,
- tcg_target_long offset,
+ intptr_t offset,
const char *name)
{
TCGContext *s = &tcg_ctx;
return idx;
}
-TCGv_i32 tcg_global_mem_new_i32(int reg, tcg_target_long offset,
- const char *name)
+TCGv_i32 tcg_global_mem_new_i32(int reg, intptr_t offset, const char *name)
{
- int idx;
-
- idx = tcg_global_mem_new_internal(TCG_TYPE_I32, reg, offset, name);
+ int idx = tcg_global_mem_new_internal(TCG_TYPE_I32, reg, offset, name);
return MAKE_TCGV_I32(idx);
}
-TCGv_i64 tcg_global_mem_new_i64(int reg, tcg_target_long offset,
- const char *name)
+TCGv_i64 tcg_global_mem_new_i64(int reg, intptr_t offset, const char *name)
{
- int idx;
-
- idx = tcg_global_mem_new_internal(TCG_TYPE_I64, reg, offset, name);
+ int idx = tcg_global_mem_new_internal(TCG_TYPE_I64, reg, offset, name);
return MAKE_TCGV_I64(idx);
}
}
#endif /* TCG_TARGET_EXTEND_ARGS */
- *gen_opc_ptr++ = INDEX_op_call;
- nparam = gen_opparam_ptr++;
+ *s->gen_opc_ptr++ = INDEX_op_call;
+ nparam = s->gen_opparam_ptr++;
if (ret != TCG_CALL_DUMMY_ARG) {
#if TCG_TARGET_REG_BITS < 64
if (sizemask & 1) {
#ifdef TCG_TARGET_WORDS_BIGENDIAN
- *gen_opparam_ptr++ = ret + 1;
- *gen_opparam_ptr++ = ret;
+ *s->gen_opparam_ptr++ = ret + 1;
+ *s->gen_opparam_ptr++ = ret;
#else
- *gen_opparam_ptr++ = ret;
- *gen_opparam_ptr++ = ret + 1;
+ *s->gen_opparam_ptr++ = ret;
+ *s->gen_opparam_ptr++ = ret + 1;
#endif
nb_rets = 2;
} else
#endif
{
- *gen_opparam_ptr++ = ret;
+ *s->gen_opparam_ptr++ = ret;
nb_rets = 1;
}
} else {
#ifdef TCG_TARGET_CALL_ALIGN_ARGS
/* some targets want aligned 64 bit args */
if (real_args & 1) {
- *gen_opparam_ptr++ = TCG_CALL_DUMMY_ARG;
+ *s->gen_opparam_ptr++ = TCG_CALL_DUMMY_ARG;
real_args++;
}
#endif
have to get more complicated to differentiate between
stack arguments and register arguments. */
#if defined(TCG_TARGET_WORDS_BIGENDIAN) != defined(TCG_TARGET_STACK_GROWSUP)
- *gen_opparam_ptr++ = args[i] + 1;
- *gen_opparam_ptr++ = args[i];
+ *s->gen_opparam_ptr++ = args[i] + 1;
+ *s->gen_opparam_ptr++ = args[i];
#else
- *gen_opparam_ptr++ = args[i];
- *gen_opparam_ptr++ = args[i] + 1;
+ *s->gen_opparam_ptr++ = args[i];
+ *s->gen_opparam_ptr++ = args[i] + 1;
#endif
real_args += 2;
continue;
}
#endif /* TCG_TARGET_REG_BITS < 64 */
- *gen_opparam_ptr++ = args[i];
+ *s->gen_opparam_ptr++ = args[i];
real_args++;
}
- *gen_opparam_ptr++ = GET_TCGV_PTR(func);
+ *s->gen_opparam_ptr++ = GET_TCGV_PTR(func);
- *gen_opparam_ptr++ = flags;
+ *s->gen_opparam_ptr++ = flags;
*nparam = (nb_rets << 16) | (real_args + 1);
/* total parameters, needed to go backward in the instruction stream */
- *gen_opparam_ptr++ = 1 + nb_rets + real_args + 3;
+ *s->gen_opparam_ptr++ = 1 + nb_rets + real_args + 3;
#if defined(TCG_TARGET_EXTEND_ARGS) && TCG_TARGET_REG_BITS == 64
for (i = 0; i < nargs; ++i) {
}
for(i = s->nb_globals; i < s->nb_temps; i++) {
ts = &s->temps[i];
- ts->val_type = TEMP_VAL_DEAD;
+ if (ts->temp_local) {
+ ts->val_type = TEMP_VAL_MEM;
+ } else {
+ ts->val_type = TEMP_VAL_DEAD;
+ }
ts->mem_allocated = 0;
ts->fixed_reg = 0;
}
assert(idx >= 0 && idx < s->nb_temps);
ts = &s->temps[idx];
- assert(ts);
if (idx < s->nb_globals) {
pstrcpy(buf, buf_size, ts->name);
} else {
static const char * const cond_name[] =
{
+ [TCG_COND_NEVER] = "never",
+ [TCG_COND_ALWAYS] = "always",
[TCG_COND_EQ] = "eq",
[TCG_COND_NE] = "ne",
[TCG_COND_LT] = "lt",
char buf[128];
first_insn = 1;
- opc_ptr = gen_opc_buf;
- args = gen_opparam_buf;
- while (opc_ptr < gen_opc_ptr) {
+ opc_ptr = s->gen_opc_buf;
+ args = s->gen_opparam_buf;
+ while (opc_ptr < s->gen_opc_ptr) {
c = *opc_ptr++;
def = &tcg_op_defs[c];
if (c == INDEX_op_debug_insn_start) {
i = 0;
for (op = 0; op < ARRAY_SIZE(tcg_op_defs); op++) {
const TCGOpDef *def = &tcg_op_defs[op];
- if (op < INDEX_op_call
- || op == INDEX_op_debug_insn_start
- || (def->flags & TCG_OPF_NOT_PRESENT)) {
+ if (def->flags & TCG_OPF_NOT_PRESENT) {
/* Wrong entry in op definitions? */
if (def->used) {
fprintf(stderr, "Invalid op definition for %s\n", def->name);
}
}
-/* liveness analysis: end of function: globals are live, temps are
- dead. */
-/* XXX: at this stage, not used as there would be little gains because
- most TBs end with a conditional jump. */
-static inline void tcg_la_func_end(TCGContext *s, uint8_t *dead_temps)
+/* liveness analysis: end of function: all temps are dead, and globals
+ should be in memory. */
+static inline void tcg_la_func_end(TCGContext *s, uint8_t *dead_temps,
+ uint8_t *mem_temps)
{
- memset(dead_temps, 0, s->nb_globals);
- memset(dead_temps + s->nb_globals, 1, s->nb_temps - s->nb_globals);
+ memset(dead_temps, 1, s->nb_temps);
+ memset(mem_temps, 1, s->nb_globals);
+ memset(mem_temps + s->nb_globals, 0, s->nb_temps - s->nb_globals);
}
-/* liveness analysis: end of basic block: globals are live, temps are
- dead, local temps are live. */
-static inline void tcg_la_bb_end(TCGContext *s, uint8_t *dead_temps)
+/* liveness analysis: end of basic block: all temps are dead, globals
+ and local temps should be in memory. */
+static inline void tcg_la_bb_end(TCGContext *s, uint8_t *dead_temps,
+ uint8_t *mem_temps)
{
int i;
- TCGTemp *ts;
- memset(dead_temps, 0, s->nb_globals);
- ts = &s->temps[s->nb_globals];
+ memset(dead_temps, 1, s->nb_temps);
+ memset(mem_temps, 1, s->nb_globals);
for(i = s->nb_globals; i < s->nb_temps; i++) {
- if (ts->temp_local)
- dead_temps[i] = 0;
- else
- dead_temps[i] = 1;
- ts++;
+ mem_temps[i] = s->temps[i].temp_local;
}
}
static void tcg_liveness_analysis(TCGContext *s)
{
int i, op_index, nb_args, nb_iargs, nb_oargs, arg, nb_ops;
- TCGOpcode op;
+ TCGOpcode op, op_new, op_new2;
TCGArg *args;
const TCGOpDef *def;
- uint8_t *dead_temps;
- unsigned int dead_args;
+ uint8_t *dead_temps, *mem_temps;
+ uint16_t dead_args;
+ uint8_t sync_args;
+ bool have_op_new2;
- gen_opc_ptr++; /* skip end */
+ s->gen_opc_ptr++; /* skip end */
- nb_ops = gen_opc_ptr - gen_opc_buf;
+ nb_ops = s->gen_opc_ptr - s->gen_opc_buf;
s->op_dead_args = tcg_malloc(nb_ops * sizeof(uint16_t));
+ s->op_sync_args = tcg_malloc(nb_ops * sizeof(uint8_t));
dead_temps = tcg_malloc(s->nb_temps);
- memset(dead_temps, 1, s->nb_temps);
+ mem_temps = tcg_malloc(s->nb_temps);
+ tcg_la_func_end(s, dead_temps, mem_temps);
- args = gen_opparam_ptr;
+ args = s->gen_opparam_ptr;
op_index = nb_ops - 1;
while (op_index >= 0) {
- op = gen_opc_buf[op_index];
+ op = s->gen_opc_buf[op_index];
def = &tcg_op_defs[op];
switch(op) {
case INDEX_op_call:
/* pure functions can be removed if their result is not
used */
- if (call_flags & TCG_CALL_PURE) {
+ if (call_flags & TCG_CALL_NO_SIDE_EFFECTS) {
for(i = 0; i < nb_oargs; i++) {
arg = args[i];
- if (!dead_temps[arg])
+ if (!dead_temps[arg] || mem_temps[arg]) {
goto do_not_remove_call;
+ }
}
- tcg_set_nop(s, gen_opc_buf + op_index,
+ tcg_set_nop(s, s->gen_opc_buf + op_index,
args - 1, nb_args);
} else {
do_not_remove_call:
/* output args are dead */
dead_args = 0;
+ sync_args = 0;
for(i = 0; i < nb_oargs; i++) {
arg = args[i];
if (dead_temps[arg]) {
dead_args |= (1 << i);
}
+ if (mem_temps[arg]) {
+ sync_args |= (1 << i);
+ }
dead_temps[arg] = 1;
+ mem_temps[arg] = 0;
}
-
- if (!(call_flags & TCG_CALL_CONST)) {
- /* globals are live (they may be used by the call) */
- memset(dead_temps, 0, s->nb_globals);
+
+ if (!(call_flags & TCG_CALL_NO_READ_GLOBALS)) {
+ /* globals should be synced to memory */
+ memset(mem_temps, 1, s->nb_globals);
+ }
+ if (!(call_flags & (TCG_CALL_NO_WRITE_GLOBALS |
+ TCG_CALL_NO_READ_GLOBALS))) {
+ /* globals should go back to memory */
+ memset(dead_temps, 1, s->nb_globals);
}
/* input args are live */
}
}
s->op_dead_args[op_index] = dead_args;
+ s->op_sync_args[op_index] = sync_args;
}
args--;
}
args--;
/* mark the temporary as dead */
dead_temps[args[0]] = 1;
+ mem_temps[args[0]] = 0;
break;
case INDEX_op_end:
break;
- /* XXX: optimize by hardcoding common cases (e.g. triadic ops) */
+
+ case INDEX_op_add2_i32:
+ op_new = INDEX_op_add_i32;
+ goto do_addsub2;
+ case INDEX_op_sub2_i32:
+ op_new = INDEX_op_sub_i32;
+ goto do_addsub2;
+ case INDEX_op_add2_i64:
+ op_new = INDEX_op_add_i64;
+ goto do_addsub2;
+ case INDEX_op_sub2_i64:
+ op_new = INDEX_op_sub_i64;
+ do_addsub2:
+ args -= 6;
+ nb_iargs = 4;
+ nb_oargs = 2;
+ /* Test if the high part of the operation is dead, but not
+ the low part. The result can be optimized to a simple
+ add or sub. This happens often for x86_64 guest when the
+ cpu mode is set to 32 bit. */
+ if (dead_temps[args[1]] && !mem_temps[args[1]]) {
+ if (dead_temps[args[0]] && !mem_temps[args[0]]) {
+ goto do_remove;
+ }
+ /* Create the single operation plus nop. */
+ s->gen_opc_buf[op_index] = op = op_new;
+ args[1] = args[2];
+ args[2] = args[4];
+ assert(s->gen_opc_buf[op_index + 1] == INDEX_op_nop);
+ tcg_set_nop(s, s->gen_opc_buf + op_index + 1, args + 3, 3);
+ /* Fall through and mark the single-word operation live. */
+ nb_iargs = 2;
+ nb_oargs = 1;
+ }
+ goto do_not_remove;
+
+ case INDEX_op_mulu2_i32:
+ op_new = INDEX_op_mul_i32;
+ op_new2 = INDEX_op_muluh_i32;
+ have_op_new2 = TCG_TARGET_HAS_muluh_i32;
+ goto do_mul2;
+ case INDEX_op_muls2_i32:
+ op_new = INDEX_op_mul_i32;
+ op_new2 = INDEX_op_mulsh_i32;
+ have_op_new2 = TCG_TARGET_HAS_mulsh_i32;
+ goto do_mul2;
+ case INDEX_op_mulu2_i64:
+ op_new = INDEX_op_mul_i64;
+ op_new2 = INDEX_op_muluh_i64;
+ have_op_new2 = TCG_TARGET_HAS_muluh_i64;
+ goto do_mul2;
+ case INDEX_op_muls2_i64:
+ op_new = INDEX_op_mul_i64;
+ op_new2 = INDEX_op_mulsh_i64;
+ have_op_new2 = TCG_TARGET_HAS_mulsh_i64;
+ goto do_mul2;
+ do_mul2:
+ args -= 4;
+ nb_iargs = 2;
+ nb_oargs = 2;
+ if (dead_temps[args[1]] && !mem_temps[args[1]]) {
+ if (dead_temps[args[0]] && !mem_temps[args[0]]) {
+ /* Both parts of the operation are dead. */
+ goto do_remove;
+ }
+ /* The high part of the operation is dead; generate the low. */
+ s->gen_opc_buf[op_index] = op = op_new;
+ args[1] = args[2];
+ args[2] = args[3];
+ } else if (have_op_new2 && dead_temps[args[0]]
+ && !mem_temps[args[0]]) {
+ /* The low part of the operation is dead; generate the high. */
+ s->gen_opc_buf[op_index] = op = op_new2;
+ args[0] = args[1];
+ args[1] = args[2];
+ args[2] = args[3];
+ } else {
+ goto do_not_remove;
+ }
+ assert(s->gen_opc_buf[op_index + 1] == INDEX_op_nop);
+ tcg_set_nop(s, s->gen_opc_buf + op_index + 1, args + 3, 1);
+ /* Mark the single-word operation live. */
+ nb_oargs = 1;
+ goto do_not_remove;
+
default:
+ /* XXX: optimize by hardcoding common cases (e.g. triadic ops) */
args -= def->nb_args;
nb_iargs = def->nb_iargs;
nb_oargs = def->nb_oargs;
if (!(def->flags & TCG_OPF_SIDE_EFFECTS) && nb_oargs != 0) {
for(i = 0; i < nb_oargs; i++) {
arg = args[i];
- if (!dead_temps[arg])
+ if (!dead_temps[arg] || mem_temps[arg]) {
goto do_not_remove;
+ }
}
- tcg_set_nop(s, gen_opc_buf + op_index, args, def->nb_args);
+ do_remove:
+ tcg_set_nop(s, s->gen_opc_buf + op_index, args, def->nb_args);
#ifdef CONFIG_PROFILER
s->del_op_count++;
#endif
/* output args are dead */
dead_args = 0;
+ sync_args = 0;
for(i = 0; i < nb_oargs; i++) {
arg = args[i];
if (dead_temps[arg]) {
dead_args |= (1 << i);
}
+ if (mem_temps[arg]) {
+ sync_args |= (1 << i);
+ }
dead_temps[arg] = 1;
+ mem_temps[arg] = 0;
}
/* if end of basic block, update */
if (def->flags & TCG_OPF_BB_END) {
- tcg_la_bb_end(s, dead_temps);
- } else if (def->flags & TCG_OPF_CALL_CLOBBER) {
- /* globals are live */
- memset(dead_temps, 0, s->nb_globals);
+ tcg_la_bb_end(s, dead_temps, mem_temps);
+ } else if (def->flags & TCG_OPF_SIDE_EFFECTS) {
+ /* globals should be synced to memory */
+ memset(mem_temps, 1, s->nb_globals);
}
/* input args are live */
dead_temps[arg] = 0;
}
s->op_dead_args[op_index] = dead_args;
+ s->op_sync_args[op_index] = sync_args;
}
break;
}
op_index--;
}
- if (args != gen_opparam_buf)
+ if (args != s->gen_opparam_buf) {
tcg_abort();
+ }
}
#else
/* dummy liveness analysis */
static void tcg_liveness_analysis(TCGContext *s)
{
int nb_ops;
- nb_ops = gen_opc_ptr - gen_opc_buf;
+ nb_ops = s->gen_opc_ptr - s->gen_opc_buf;
s->op_dead_args = tcg_malloc(nb_ops * sizeof(uint16_t));
memset(s->op_dead_args, 0, nb_ops * sizeof(uint16_t));
+ s->op_sync_args = tcg_malloc(nb_ops * sizeof(uint8_t));
+ memset(s->op_sync_args, 0, nb_ops * sizeof(uint8_t));
}
#endif
ts->mem_offset = s->current_frame_offset;
ts->mem_reg = s->frame_reg;
ts->mem_allocated = 1;
- s->current_frame_offset += (tcg_target_long)sizeof(tcg_target_long);
+ s->current_frame_offset += sizeof(tcg_target_long);
+}
+
+/* sync register 'reg' by saving it to the corresponding temporary */
+static inline void tcg_reg_sync(TCGContext *s, int reg)
+{
+ TCGTemp *ts;
+ int temp;
+
+ temp = s->reg_to_temp[reg];
+ ts = &s->temps[temp];
+ assert(ts->val_type == TEMP_VAL_REG);
+ if (!ts->mem_coherent && !ts->fixed_reg) {
+ if (!ts->mem_allocated) {
+ temp_allocate_frame(s, temp);
+ }
+ tcg_out_st(s, ts->type, reg, ts->mem_reg, ts->mem_offset);
+ }
+ ts->mem_coherent = 1;
}
/* free register 'reg' by spilling the corresponding temporary if necessary */
static void tcg_reg_free(TCGContext *s, int reg)
{
- TCGTemp *ts;
int temp;
temp = s->reg_to_temp[reg];
if (temp != -1) {
- ts = &s->temps[temp];
- assert(ts->val_type == TEMP_VAL_REG);
- if (!ts->mem_coherent) {
- if (!ts->mem_allocated)
- temp_allocate_frame(s, temp);
- tcg_out_st(s, ts->type, reg, ts->mem_reg, ts->mem_offset);
- }
- ts->val_type = TEMP_VAL_MEM;
+ tcg_reg_sync(s, reg);
+ s->temps[temp].val_type = TEMP_VAL_MEM;
s->reg_to_temp[reg] = -1;
}
}
tcg_abort();
}
-/* save a temporary to memory. 'allocated_regs' is used in case a
+/* mark a temporary as dead. */
+static inline void temp_dead(TCGContext *s, int temp)
+{
+ TCGTemp *ts;
+
+ ts = &s->temps[temp];
+ if (!ts->fixed_reg) {
+ if (ts->val_type == TEMP_VAL_REG) {
+ s->reg_to_temp[ts->reg] = -1;
+ }
+ if (temp < s->nb_globals || ts->temp_local) {
+ ts->val_type = TEMP_VAL_MEM;
+ } else {
+ ts->val_type = TEMP_VAL_DEAD;
+ }
+ }
+}
+
+/* sync a temporary to memory. 'allocated_regs' is used in case a
temporary registers needs to be allocated to store a constant. */
-static void temp_save(TCGContext *s, int temp, TCGRegSet allocated_regs)
+static inline void temp_sync(TCGContext *s, int temp, TCGRegSet allocated_regs)
{
TCGTemp *ts;
- int reg;
ts = &s->temps[temp];
if (!ts->fixed_reg) {
switch(ts->val_type) {
+ case TEMP_VAL_CONST:
+ ts->reg = tcg_reg_alloc(s, tcg_target_available_regs[ts->type],
+ allocated_regs);
+ ts->val_type = TEMP_VAL_REG;
+ s->reg_to_temp[ts->reg] = temp;
+ ts->mem_coherent = 0;
+ tcg_out_movi(s, ts->type, ts->reg, ts->val);
+ /* fallthrough*/
case TEMP_VAL_REG:
- tcg_reg_free(s, ts->reg);
+ tcg_reg_sync(s, ts->reg);
break;
case TEMP_VAL_DEAD:
- ts->val_type = TEMP_VAL_MEM;
- break;
- case TEMP_VAL_CONST:
- reg = tcg_reg_alloc(s, tcg_target_available_regs[ts->type],
- allocated_regs);
- if (!ts->mem_allocated)
- temp_allocate_frame(s, temp);
- tcg_out_movi(s, ts->type, reg, ts->val);
- tcg_out_st(s, ts->type, reg, ts->mem_reg, ts->mem_offset);
- ts->val_type = TEMP_VAL_MEM;
- break;
case TEMP_VAL_MEM:
break;
default:
}
}
+/* save a temporary to memory. 'allocated_regs' is used in case a
+ temporary registers needs to be allocated to store a constant. */
+static inline void temp_save(TCGContext *s, int temp, TCGRegSet allocated_regs)
+{
+#ifdef USE_LIVENESS_ANALYSIS
+ /* The liveness analysis already ensures that globals are back
+ in memory. Keep an assert for safety. */
+ assert(s->temps[temp].val_type == TEMP_VAL_MEM || s->temps[temp].fixed_reg);
+#else
+ temp_sync(s, temp, allocated_regs);
+ temp_dead(s, temp);
+#endif
+}
+
/* save globals to their canonical location and assume they can be
modified be the following code. 'allocated_regs' is used in case a
temporary registers needs to be allocated to store a constant. */
}
}
+/* sync globals to their canonical location and assume they can be
+ read by the following code. 'allocated_regs' is used in case a
+ temporary registers needs to be allocated to store a constant. */
+static void sync_globals(TCGContext *s, TCGRegSet allocated_regs)
+{
+ int i;
+
+ for (i = 0; i < s->nb_globals; i++) {
+#ifdef USE_LIVENESS_ANALYSIS
+ assert(s->temps[i].val_type != TEMP_VAL_REG || s->temps[i].fixed_reg ||
+ s->temps[i].mem_coherent);
+#else
+ temp_sync(s, i, allocated_regs);
+#endif
+ }
+}
+
/* at the end of a basic block, we assume all temporaries are dead and
all globals are stored at their canonical location. */
static void tcg_reg_alloc_bb_end(TCGContext *s, TCGRegSet allocated_regs)
if (ts->temp_local) {
temp_save(s, i, allocated_regs);
} else {
- if (ts->val_type == TEMP_VAL_REG) {
- s->reg_to_temp[ts->reg] = -1;
- }
- ts->val_type = TEMP_VAL_DEAD;
+#ifdef USE_LIVENESS_ANALYSIS
+ /* The liveness analysis already ensures that temps are dead.
+ Keep an assert for safety. */
+ assert(ts->val_type == TEMP_VAL_DEAD);
+#else
+ temp_dead(s, i);
+#endif
}
}
}
#define IS_DEAD_ARG(n) ((dead_args >> (n)) & 1)
+#define NEED_SYNC_ARG(n) ((sync_args >> (n)) & 1)
-static void tcg_reg_alloc_movi(TCGContext *s, const TCGArg *args)
+static void tcg_reg_alloc_movi(TCGContext *s, const TCGArg *args,
+ uint16_t dead_args, uint8_t sync_args)
{
TCGTemp *ots;
tcg_target_ulong val;
ots->val_type = TEMP_VAL_CONST;
ots->val = val;
}
+ if (NEED_SYNC_ARG(0)) {
+ temp_sync(s, args[0], s->reserved_regs);
+ }
+ if (IS_DEAD_ARG(0)) {
+ temp_dead(s, args[0]);
+ }
}
static void tcg_reg_alloc_mov(TCGContext *s, const TCGOpDef *def,
- const TCGArg *args,
- unsigned int dead_args)
+ const TCGArg *args, uint16_t dead_args,
+ uint8_t sync_args)
{
+ TCGRegSet allocated_regs;
TCGTemp *ts, *ots;
- int reg;
- const TCGArgConstraint *arg_ct;
+ const TCGArgConstraint *arg_ct, *oarg_ct;
+ tcg_regset_set(allocated_regs, s->reserved_regs);
ots = &s->temps[args[0]];
ts = &s->temps[args[1]];
- arg_ct = &def->args_ct[0];
+ oarg_ct = &def->args_ct[0];
+ arg_ct = &def->args_ct[1];
+
+ /* If the source value is not in a register, and we're going to be
+ forced to have it in a register in order to perform the copy,
+ then copy the SOURCE value into its own register first. That way
+ we don't have to reload SOURCE the next time it is used. */
+ if (((NEED_SYNC_ARG(0) || ots->fixed_reg) && ts->val_type != TEMP_VAL_REG)
+ || ts->val_type == TEMP_VAL_MEM) {
+ ts->reg = tcg_reg_alloc(s, arg_ct->u.regs, allocated_regs);
+ if (ts->val_type == TEMP_VAL_MEM) {
+ tcg_out_ld(s, ts->type, ts->reg, ts->mem_reg, ts->mem_offset);
+ ts->mem_coherent = 1;
+ } else if (ts->val_type == TEMP_VAL_CONST) {
+ tcg_out_movi(s, ts->type, ts->reg, ts->val);
+ }
+ s->reg_to_temp[ts->reg] = args[1];
+ ts->val_type = TEMP_VAL_REG;
+ }
- /* XXX: always mark arg dead if IS_DEAD_ARG(1) */
- if (ts->val_type == TEMP_VAL_REG) {
+ if (IS_DEAD_ARG(0) && !ots->fixed_reg) {
+ /* mov to a non-saved dead register makes no sense (even with
+ liveness analysis disabled). */
+ assert(NEED_SYNC_ARG(0));
+ /* The code above should have moved the temp to a register. */
+ assert(ts->val_type == TEMP_VAL_REG);
+ if (!ots->mem_allocated) {
+ temp_allocate_frame(s, args[0]);
+ }
+ tcg_out_st(s, ots->type, ts->reg, ots->mem_reg, ots->mem_offset);
+ if (IS_DEAD_ARG(1)) {
+ temp_dead(s, args[1]);
+ }
+ temp_dead(s, args[0]);
+ } else if (ts->val_type == TEMP_VAL_CONST) {
+ /* propagate constant */
+ if (ots->val_type == TEMP_VAL_REG) {
+ s->reg_to_temp[ots->reg] = -1;
+ }
+ ots->val_type = TEMP_VAL_CONST;
+ ots->val = ts->val;
+ } else {
+ /* The code in the first if block should have moved the
+ temp to a register. */
+ assert(ts->val_type == TEMP_VAL_REG);
if (IS_DEAD_ARG(1) && !ts->fixed_reg && !ots->fixed_reg) {
/* the mov can be suppressed */
- if (ots->val_type == TEMP_VAL_REG)
- s->reg_to_temp[ots->reg] = -1;
- reg = ts->reg;
- s->reg_to_temp[reg] = -1;
- ts->val_type = TEMP_VAL_DEAD;
- } else {
if (ots->val_type == TEMP_VAL_REG) {
- reg = ots->reg;
- } else {
- reg = tcg_reg_alloc(s, arg_ct->u.regs, s->reserved_regs);
- }
- if (ts->reg != reg) {
- tcg_out_mov(s, ots->type, reg, ts->reg);
+ s->reg_to_temp[ots->reg] = -1;
}
- }
- } else if (ts->val_type == TEMP_VAL_MEM) {
- if (ots->val_type == TEMP_VAL_REG) {
- reg = ots->reg;
+ ots->reg = ts->reg;
+ temp_dead(s, args[1]);
} else {
- reg = tcg_reg_alloc(s, arg_ct->u.regs, s->reserved_regs);
+ if (ots->val_type != TEMP_VAL_REG) {
+ /* When allocating a new register, make sure to not spill the
+ input one. */
+ tcg_regset_set_reg(allocated_regs, ts->reg);
+ ots->reg = tcg_reg_alloc(s, oarg_ct->u.regs, allocated_regs);
+ }
+ tcg_out_mov(s, ots->type, ots->reg, ts->reg);
}
- tcg_out_ld(s, ts->type, reg, ts->mem_reg, ts->mem_offset);
- } else if (ts->val_type == TEMP_VAL_CONST) {
- if (ots->fixed_reg) {
- reg = ots->reg;
- tcg_out_movi(s, ots->type, reg, ts->val);
- } else {
- /* propagate constant */
- if (ots->val_type == TEMP_VAL_REG)
- s->reg_to_temp[ots->reg] = -1;
- ots->val_type = TEMP_VAL_CONST;
- ots->val = ts->val;
- return;
+ ots->val_type = TEMP_VAL_REG;
+ ots->mem_coherent = 0;
+ s->reg_to_temp[ots->reg] = args[0];
+ if (NEED_SYNC_ARG(0)) {
+ tcg_reg_sync(s, ots->reg);
}
- } else {
- tcg_abort();
}
- s->reg_to_temp[reg] = args[0];
- ots->reg = reg;
- ots->val_type = TEMP_VAL_REG;
- ots->mem_coherent = 0;
}
static void tcg_reg_alloc_op(TCGContext *s,
const TCGOpDef *def, TCGOpcode opc,
- const TCGArg *args,
- unsigned int dead_args)
+ const TCGArg *args, uint16_t dead_args,
+ uint8_t sync_args)
{
TCGRegSet allocated_regs;
int i, k, nb_iargs, nb_oargs, reg;
iarg_end: ;
}
+ /* mark dead temporaries and free the associated registers */
+ for (i = nb_oargs; i < nb_oargs + nb_iargs; i++) {
+ if (IS_DEAD_ARG(i)) {
+ temp_dead(s, args[i]);
+ }
+ }
+
if (def->flags & TCG_OPF_BB_END) {
tcg_reg_alloc_bb_end(s, allocated_regs);
} else {
- /* mark dead temporaries and free the associated registers */
- for(i = nb_oargs; i < nb_oargs + nb_iargs; i++) {
- arg = args[i];
- if (IS_DEAD_ARG(i)) {
- ts = &s->temps[arg];
- if (!ts->fixed_reg) {
- if (ts->val_type == TEMP_VAL_REG)
- s->reg_to_temp[ts->reg] = -1;
- ts->val_type = TEMP_VAL_DEAD;
- }
- }
- }
-
if (def->flags & TCG_OPF_CALL_CLOBBER) {
/* XXX: permit generic clobber register list ? */
for(reg = 0; reg < TCG_TARGET_NB_REGS; reg++) {
tcg_reg_free(s, reg);
}
}
- /* XXX: for load/store we could do that only for the slow path
- (i.e. when a memory callback is called) */
-
- /* store globals and free associated registers (we assume the insn
- can modify any global. */
- save_globals(s, allocated_regs);
+ }
+ if (def->flags & TCG_OPF_SIDE_EFFECTS) {
+ /* sync globals if the op has side effects and might trigger
+ an exception. */
+ sync_globals(s, allocated_regs);
}
/* satisfy the output constraints */
tcg_regset_set_reg(allocated_regs, reg);
/* if a fixed register is used, then a move will be done afterwards */
if (!ts->fixed_reg) {
- if (ts->val_type == TEMP_VAL_REG)
+ if (ts->val_type == TEMP_VAL_REG) {
s->reg_to_temp[ts->reg] = -1;
- if (IS_DEAD_ARG(i)) {
- ts->val_type = TEMP_VAL_DEAD;
- } else {
- ts->val_type = TEMP_VAL_REG;
- ts->reg = reg;
- /* temp value is modified, so the value kept in memory is
- potentially not the same */
- ts->mem_coherent = 0;
- s->reg_to_temp[reg] = arg;
- }
+ }
+ ts->val_type = TEMP_VAL_REG;
+ ts->reg = reg;
+ /* temp value is modified, so the value kept in memory is
+ potentially not the same */
+ ts->mem_coherent = 0;
+ s->reg_to_temp[reg] = arg;
}
oarg_end:
new_args[i] = reg;
if (ts->fixed_reg && ts->reg != reg) {
tcg_out_mov(s, ts->type, ts->reg, reg);
}
+ if (NEED_SYNC_ARG(i)) {
+ tcg_reg_sync(s, reg);
+ }
+ if (IS_DEAD_ARG(i)) {
+ temp_dead(s, args[i]);
+ }
}
}
static int tcg_reg_alloc_call(TCGContext *s, const TCGOpDef *def,
TCGOpcode opc, const TCGArg *args,
- unsigned int dead_args)
+ uint16_t dead_args, uint8_t sync_args)
{
int nb_iargs, nb_oargs, flags, nb_regs, i, reg, nb_params;
TCGArg arg, func_arg;
/* mark dead temporaries and free the associated registers */
for(i = nb_oargs; i < nb_iargs + nb_oargs; i++) {
- arg = args[i];
if (IS_DEAD_ARG(i)) {
- ts = &s->temps[arg];
- if (!ts->fixed_reg) {
- if (ts->val_type == TEMP_VAL_REG)
- s->reg_to_temp[ts->reg] = -1;
- ts->val_type = TEMP_VAL_DEAD;
- }
+ temp_dead(s, args[i]);
}
}
tcg_reg_free(s, reg);
}
}
-
- /* store globals and free associated registers (we assume the call
- can modify any global. */
- if (!(flags & TCG_CALL_CONST)) {
+
+ /* Save globals if they might be written by the helper, sync them if
+ they might be read. */
+ if (flags & TCG_CALL_NO_READ_GLOBALS) {
+ /* Nothing to do */
+ } else if (flags & TCG_CALL_NO_WRITE_GLOBALS) {
+ sync_globals(s, allocated_regs);
+ } else {
save_globals(s, allocated_regs);
}
tcg_out_mov(s, ts->type, ts->reg, reg);
}
} else {
- if (ts->val_type == TEMP_VAL_REG)
+ if (ts->val_type == TEMP_VAL_REG) {
s->reg_to_temp[ts->reg] = -1;
+ }
+ ts->val_type = TEMP_VAL_REG;
+ ts->reg = reg;
+ ts->mem_coherent = 0;
+ s->reg_to_temp[reg] = arg;
+ if (NEED_SYNC_ARG(i)) {
+ tcg_reg_sync(s, reg);
+ }
if (IS_DEAD_ARG(i)) {
- ts->val_type = TEMP_VAL_DEAD;
- } else {
- ts->val_type = TEMP_VAL_REG;
- ts->reg = reg;
- ts->mem_coherent = 0;
- s->reg_to_temp[reg] = arg;
+ temp_dead(s, args[i]);
}
}
}
TCGOpcode opc;
int op_index;
const TCGOpDef *def;
- unsigned int dead_args;
const TCGArg *args;
#ifdef DEBUG_DISAS
#endif
#ifdef USE_TCG_OPTIMIZATIONS
- gen_opparam_ptr =
- tcg_optimize(s, gen_opc_ptr, gen_opparam_buf, tcg_op_defs);
+ s->gen_opparam_ptr =
+ tcg_optimize(s, s->gen_opc_ptr, s->gen_opparam_buf, tcg_op_defs);
#endif
#ifdef CONFIG_PROFILER
s->code_buf = gen_code_buf;
s->code_ptr = gen_code_buf;
- args = gen_opparam_buf;
+ args = s->gen_opparam_buf;
op_index = 0;
for(;;) {
- opc = gen_opc_buf[op_index];
+ opc = s->gen_opc_buf[op_index];
#ifdef CONFIG_PROFILER
tcg_table_op_count[opc]++;
#endif
switch(opc) {
case INDEX_op_mov_i32:
case INDEX_op_mov_i64:
- dead_args = s->op_dead_args[op_index];
- tcg_reg_alloc_mov(s, def, args, dead_args);
+ tcg_reg_alloc_mov(s, def, args, s->op_dead_args[op_index],
+ s->op_sync_args[op_index]);
break;
case INDEX_op_movi_i32:
case INDEX_op_movi_i64:
- tcg_reg_alloc_movi(s, args);
+ tcg_reg_alloc_movi(s, args, s->op_dead_args[op_index],
+ s->op_sync_args[op_index]);
break;
case INDEX_op_debug_insn_start:
/* debug instruction */
args += args[0];
goto next;
case INDEX_op_discard:
- {
- TCGTemp *ts;
- ts = &s->temps[args[0]];
- /* mark the temporary as dead */
- if (!ts->fixed_reg) {
- if (ts->val_type == TEMP_VAL_REG)
- s->reg_to_temp[ts->reg] = -1;
- ts->val_type = TEMP_VAL_DEAD;
- }
- }
+ temp_dead(s, args[0]);
break;
case INDEX_op_set_label:
tcg_reg_alloc_bb_end(s, s->reserved_regs);
tcg_out_label(s, args[0], s->code_ptr);
break;
case INDEX_op_call:
- dead_args = s->op_dead_args[op_index];
- args += tcg_reg_alloc_call(s, def, opc, args, dead_args);
+ args += tcg_reg_alloc_call(s, def, opc, args,
+ s->op_dead_args[op_index],
+ s->op_sync_args[op_index]);
goto next;
case INDEX_op_end:
goto the_end;
/* Note: in order to speed up the code, it would be much
faster to have specialized register allocator functions for
some common argument patterns */
- dead_args = s->op_dead_args[op_index];
- tcg_reg_alloc_op(s, def, opc, args, dead_args);
+ tcg_reg_alloc_op(s, def, opc, args, s->op_dead_args[op_index],
+ s->op_sync_args[op_index]);
break;
}
args += def->nb_args;
#endif
}
the_end:
+#if defined(CONFIG_QEMU_LDST_OPTIMIZATION) && defined(CONFIG_SOFTMMU)
+ /* Generate TB finalization at the end of block */
+ tcg_out_tb_finalize(s);
+#endif
return -1;
}
#ifdef CONFIG_PROFILER
{
int n;
- n = (gen_opc_ptr - gen_opc_buf);
+ n = (s->gen_opc_ptr - s->gen_opc_buf);
s->op_count += n;
if (n > s->op_count_max)
s->op_count_max = n;
tcg_gen_code_common(s, gen_code_buf, -1);
/* flush instruction cache */
- flush_icache_range((tcg_target_ulong)gen_code_buf,
- (tcg_target_ulong)s->code_ptr);
+ flush_icache_range((uintptr_t)gen_code_buf, (uintptr_t)s->code_ptr);
return s->code_ptr - gen_code_buf;
}
img->sym[1].st_size = buf_size;
img->di.cu_low_pc = buf;
- img->di.cu_high_pc = buf_size;
+ img->di.cu_high_pc = buf + buf_size;
img->di.fn_low_pc = buf;
- img->di.fn_high_pc = buf_size;
+ img->di.fn_high_pc = buf + buf_size;
#ifdef DEBUG_JIT
/* Enable this block to be able to debug the ELF image file creation.
projects / qemu.git / blobdiff