* THE SOFTWARE.
*/
-/* define it to suppress various consistency checks (faster) */
-#define NDEBUG
-
/* define it to use liveness analysis (better code) */
#define USE_LIVENESS_ANALYSIS
-#include <assert.h>
+#include "config.h"
+
+#ifndef DEBUG_TCG
+/* define it to suppress various consistency checks (faster) */
+#define NDEBUG
+#endif
+
#include <stdarg.h>
#include <stdlib.h>
#include <stdio.h>
#ifdef _WIN32
#include <malloc.h>
#endif
+#ifdef _AIX
+#include <alloca.h>
+#endif
-#include "config.h"
-#include "osdep.h"
+#include "qemu-common.h"
+#include "cache-utils.h"
/* Note: the long term plan is to reduce the dependancies on the QEMU
CPU definitions. Currently they are used for qemu_ld/st
static void patch_reloc(uint8_t *code_ptr, int type,
- tcg_target_long value);
+ tcg_target_long value, tcg_target_long addend);
-TCGOpDef tcg_op_defs[] = {
+static TCGOpDef tcg_op_defs[] = {
#define DEF(s, n, copy_size) { #s, 0, 0, n, n, 0, copy_size },
#define DEF2(s, iargs, oargs, cargs, flags) { #s, iargs, oargs, cargs, iargs + oargs + cargs, flags, 0 },
#include "tcg-opc.h"
#undef DEF2
};
-TCGRegSet tcg_target_available_regs[2];
-TCGRegSet tcg_target_call_clobber_regs;
+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;
/* FIXME: This may break relocations on RISC targets that
modify instruction fields in place. The caller may not have
written the initial value. */
- patch_reloc(code_ptr, type, l->u.value + addend);
+ patch_reloc(code_ptr, type, l->u.value, addend);
} else {
/* add a new relocation entry */
r = tcg_malloc(sizeof(TCGRelocation));
tcg_abort();
r = l->u.first_reloc;
while (r != NULL) {
- patch_reloc(r->ptr, r->type, value + r->addend);
+ patch_reloc(r->ptr, r->type, value, r->addend);
r = r->next;
}
l->has_value = 1;
#include "tcg-target.c"
-/* XXX: factorize */
-static void pstrcpy(char *buf, int buf_size, const char *str)
-{
- int c;
- char *q = buf;
-
- if (buf_size <= 0)
- return;
-
- for(;;) {
- c = *str++;
- if (c == 0 || q >= buf + buf_size - 1)
- break;
- *q++ = c;
- }
- *q = '\0';
-}
-
-#if TCG_TARGET_REG_BITS == 32
-/* strcat and truncate. */
-static char *pstrcat(char *buf, int buf_size, const char *s)
-{
- int len;
- len = strlen(buf);
- if (len < buf_size)
- pstrcpy(buf + len, buf_size - len, s);
- return buf;
-}
-#endif
-
/* pool based memory allocation */
void *tcg_malloc_internal(TCGContext *s, int size)
{
s->pool_current = NULL;
}
-/* free all the pool */
-void tcg_pool_free(TCGContext *s)
-{
- TCGPool *p, *p1;
-
- for(p = s->pool_first; p != NULL; p = p1) {
- p1 = p->next;
- qemu_free(p);
- }
- s->pool_first = NULL;
- s->pool_cur = s->pool_end = NULL;
-}
-
void tcg_context_init(TCGContext *s)
{
int op, total_args, n;
}
tcg_target_init(s);
+
+ /* init global prologue and epilogue */
+ s->code_buf = code_gen_prologue;
+ s->code_ptr = s->code_buf;
+ tcg_target_qemu_prologue(s);
+ flush_icache_range((unsigned long)s->code_buf,
+ (unsigned long)s->code_ptr);
}
void tcg_set_frame(TCGContext *s, int reg,
s->frame_reg = reg;
}
-void tcg_set_macro_func(TCGContext *s, TCGMacroFunc *func)
-{
- s->macro_func = func;
-}
-
void tcg_func_start(TCGContext *s)
{
+ int i;
tcg_pool_reset(s);
s->nb_temps = s->nb_globals;
+ for(i = 0; i < (TCG_TYPE_COUNT * 2); i++)
+ s->first_free_temp[i] = -1;
s->labels = tcg_malloc(sizeof(TCGLabel) * TCG_MAX_LABELS);
s->nb_labels = 0;
s->current_frame_offset = s->frame_start;
tcg_abort();
}
-TCGv tcg_global_reg_new(TCGType type, int reg, const char *name)
+static inline int tcg_global_reg_new_internal(TCGType type, int reg,
+ const char *name)
{
TCGContext *s = &tcg_ctx;
TCGTemp *ts;
ts->type = type;
ts->fixed_reg = 1;
ts->reg = reg;
- ts->val_type = TEMP_VAL_REG;
ts->name = name;
s->nb_globals++;
tcg_regset_set_reg(s->reserved_regs, reg);
- return MAKE_TCGV(idx);
+ return idx;
+}
+
+TCGv_i32 tcg_global_reg_new_i32(int reg, const char *name)
+{
+ int idx;
+
+ idx = tcg_global_reg_new_internal(TCG_TYPE_I32, reg, name);
+ return MAKE_TCGV_I32(idx);
+}
+
+TCGv_i64 tcg_global_reg_new_i64(int reg, const char *name)
+{
+ int idx;
+
+ idx = tcg_global_reg_new_internal(TCG_TYPE_I64, reg, name);
+ return MAKE_TCGV_I64(idx);
}
-TCGv tcg_global_mem_new(TCGType type, int reg, tcg_target_long offset,
- const char *name)
+static inline int tcg_global_mem_new_internal(TCGType type, int reg,
+ tcg_target_long offset,
+ const char *name)
{
TCGContext *s = &tcg_ctx;
TCGTemp *ts;
#if TCG_TARGET_REG_BITS == 32
if (type == TCG_TYPE_I64) {
char buf[64];
- tcg_temp_alloc(s, s->nb_globals + 1);
+ tcg_temp_alloc(s, s->nb_globals + 2);
ts = &s->temps[s->nb_globals];
ts->base_type = type;
ts->type = TCG_TYPE_I32;
#else
ts->mem_offset = offset;
#endif
- ts->val_type = TEMP_VAL_MEM;
pstrcpy(buf, sizeof(buf), name);
pstrcat(buf, sizeof(buf), "_0");
ts->name = strdup(buf);
#else
ts->mem_offset = offset + 4;
#endif
- ts->val_type = TEMP_VAL_MEM;
pstrcpy(buf, sizeof(buf), name);
pstrcat(buf, sizeof(buf), "_1");
ts->name = strdup(buf);
ts->mem_allocated = 1;
ts->mem_reg = reg;
ts->mem_offset = offset;
- ts->val_type = TEMP_VAL_MEM;
ts->name = name;
s->nb_globals++;
}
- return MAKE_TCGV(idx);
+ return idx;
}
-TCGv tcg_temp_new(TCGType type)
+TCGv_i32 tcg_global_mem_new_i32(int reg, tcg_target_long offset,
+ const char *name)
{
- TCGContext *s = &tcg_ctx;
- TCGTemp *ts;
int idx;
- idx = s->nb_temps;
+ 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)
+{
+ int idx;
+
+ idx = tcg_global_mem_new_internal(TCG_TYPE_I64, reg, offset, name);
+ return MAKE_TCGV_I64(idx);
+}
+
+static inline int tcg_temp_new_internal(TCGType type, int temp_local)
+{
+ TCGContext *s = &tcg_ctx;
+ TCGTemp *ts;
+ int idx, k;
+
+ k = type;
+ if (temp_local)
+ k += TCG_TYPE_COUNT;
+ idx = s->first_free_temp[k];
+ if (idx != -1) {
+ /* There is already an available temp with the
+ right type */
+ ts = &s->temps[idx];
+ s->first_free_temp[k] = ts->next_free_temp;
+ ts->temp_allocated = 1;
+ assert(ts->temp_local == temp_local);
+ } else {
+ idx = s->nb_temps;
#if TCG_TARGET_REG_BITS == 32
- if (type == TCG_TYPE_I64) {
- tcg_temp_alloc(s, s->nb_temps + 1);
- ts = &s->temps[s->nb_temps];
- ts->base_type = type;
- ts->type = TCG_TYPE_I32;
- ts->fixed_reg = 0;
- ts->val_type = TEMP_VAL_DEAD;
- ts->mem_allocated = 0;
- ts->name = NULL;
- ts++;
- ts->base_type = TCG_TYPE_I32;
- ts->type = TCG_TYPE_I32;
- ts->val_type = TEMP_VAL_DEAD;
- ts->fixed_reg = 0;
- ts->mem_allocated = 0;
- ts->name = NULL;
- s->nb_temps += 2;
- } else
+ if (type == TCG_TYPE_I64) {
+ tcg_temp_alloc(s, s->nb_temps + 2);
+ ts = &s->temps[s->nb_temps];
+ ts->base_type = type;
+ ts->type = TCG_TYPE_I32;
+ ts->temp_allocated = 1;
+ ts->temp_local = temp_local;
+ ts->name = NULL;
+ ts++;
+ ts->base_type = TCG_TYPE_I32;
+ ts->type = TCG_TYPE_I32;
+ ts->temp_allocated = 1;
+ ts->temp_local = temp_local;
+ ts->name = NULL;
+ s->nb_temps += 2;
+ } else
#endif
- {
- tcg_temp_alloc(s, s->nb_temps + 1);
- ts = &s->temps[s->nb_temps];
- ts->base_type = type;
- ts->type = type;
- ts->fixed_reg = 0;
- ts->val_type = TEMP_VAL_DEAD;
- ts->mem_allocated = 0;
- ts->name = NULL;
- s->nb_temps++;
+ {
+ tcg_temp_alloc(s, s->nb_temps + 1);
+ ts = &s->temps[s->nb_temps];
+ ts->base_type = type;
+ ts->type = type;
+ ts->temp_allocated = 1;
+ ts->temp_local = temp_local;
+ ts->name = NULL;
+ s->nb_temps++;
+ }
}
- return MAKE_TCGV(idx);
+ return idx;
}
-TCGv tcg_const_i32(int32_t val)
+TCGv_i32 tcg_temp_new_internal_i32(int temp_local)
{
- TCGContext *s = &tcg_ctx;
- TCGTemp *ts;
int idx;
- idx = s->nb_temps;
- tcg_temp_alloc(s, idx + 1);
- ts = &s->temps[idx];
- ts->base_type = ts->type = TCG_TYPE_I32;
- ts->val_type = TEMP_VAL_CONST;
- ts->name = NULL;
- ts->val = val;
- s->nb_temps++;
- return MAKE_TCGV(idx);
+ idx = tcg_temp_new_internal(TCG_TYPE_I32, temp_local);
+ return MAKE_TCGV_I32(idx);
+}
+
+TCGv_i64 tcg_temp_new_internal_i64(int temp_local)
+{
+ int idx;
+
+ idx = tcg_temp_new_internal(TCG_TYPE_I64, temp_local);
+ return MAKE_TCGV_I64(idx);
}
-TCGv tcg_const_i64(int64_t val)
+static inline void tcg_temp_free_internal(int idx)
{
TCGContext *s = &tcg_ctx;
TCGTemp *ts;
- int idx;
+ int k;
- idx = s->nb_temps;
-#if TCG_TARGET_REG_BITS == 32
- tcg_temp_alloc(s, idx + 2);
- ts = &s->temps[idx];
- ts->base_type = TCG_TYPE_I64;
- ts->type = TCG_TYPE_I32;
- ts->val_type = TEMP_VAL_CONST;
- ts->name = NULL;
- ts->val = val;
- ts++;
- ts->base_type = TCG_TYPE_I32;
- ts->type = TCG_TYPE_I32;
- ts->val_type = TEMP_VAL_CONST;
- ts->name = NULL;
- ts->val = val >> 32;
- s->nb_temps += 2;
-#else
- tcg_temp_alloc(s, idx + 1);
+ assert(idx >= s->nb_globals && idx < s->nb_temps);
ts = &s->temps[idx];
- ts->base_type = ts->type = TCG_TYPE_I64;
- ts->val_type = TEMP_VAL_CONST;
- ts->name = NULL;
- ts->val = val;
- s->nb_temps++;
-#endif
- return MAKE_TCGV(idx);
+ assert(ts->temp_allocated != 0);
+ ts->temp_allocated = 0;
+ k = ts->base_type;
+ if (ts->temp_local)
+ k += TCG_TYPE_COUNT;
+ ts->next_free_temp = s->first_free_temp[k];
+ s->first_free_temp[k] = idx;
+}
+
+void tcg_temp_free_i32(TCGv_i32 arg)
+{
+ tcg_temp_free_internal(GET_TCGV_I32(arg));
+}
+
+void tcg_temp_free_i64(TCGv_i64 arg)
+{
+ tcg_temp_free_internal(GET_TCGV_I64(arg));
+}
+
+TCGv_i32 tcg_const_i32(int32_t val)
+{
+ TCGv_i32 t0;
+ t0 = tcg_temp_new_i32();
+ tcg_gen_movi_i32(t0, val);
+ return t0;
+}
+
+TCGv_i64 tcg_const_i64(int64_t val)
+{
+ TCGv_i64 t0;
+ t0 = tcg_temp_new_i64();
+ tcg_gen_movi_i64(t0, val);
+ return t0;
+}
+
+TCGv_i32 tcg_const_local_i32(int32_t val)
+{
+ TCGv_i32 t0;
+ t0 = tcg_temp_local_new_i32();
+ tcg_gen_movi_i32(t0, val);
+ return t0;
+}
+
+TCGv_i64 tcg_const_local_i64(int64_t val)
+{
+ TCGv_i64 t0;
+ t0 = tcg_temp_local_new_i64();
+ tcg_gen_movi_i64(t0, val);
+ return t0;
}
void tcg_register_helper(void *func, const char *name)
s->helpers = realloc(s->helpers, n * sizeof(TCGHelperInfo));
s->allocated_helpers = n;
}
- s->helpers[s->nb_helpers].func = func;
+ s->helpers[s->nb_helpers].func = (tcg_target_ulong)func;
s->helpers[s->nb_helpers].name = name;
s->nb_helpers++;
}
-const char *tcg_helper_get_name(TCGContext *s, void *func)
-{
- int i;
-
- for(i = 0; i < s->nb_helpers; i++) {
- if (s->helpers[i].func == func)
- return s->helpers[i].name;
- }
- return NULL;
-}
-
-static inline TCGType tcg_get_base_type(TCGContext *s, TCGv arg)
-{
- return s->temps[GET_TCGV(arg)].base_type;
-}
-
-static void tcg_gen_call_internal(TCGContext *s, TCGv func,
- unsigned int flags,
- unsigned int nb_rets, const TCGv *rets,
- unsigned int nb_params, const TCGv *params)
+/* Note: we convert the 64 bit args to 32 bit and do some alignment
+ and endian swap. Maybe it would be better to do the alignment
+ and endian swap in tcg_reg_alloc_call(). */
+void tcg_gen_callN(TCGContext *s, TCGv_ptr func, unsigned int flags,
+ int sizemask, TCGArg ret, int nargs, TCGArg *args)
{
+ int call_type;
int i;
+ int real_args;
+ int nb_rets;
+ TCGArg *nparam;
*gen_opc_ptr++ = INDEX_op_call;
- *gen_opparam_ptr++ = (nb_rets << 16) | (nb_params + 1);
- for(i = 0; i < nb_rets; i++) {
- *gen_opparam_ptr++ = GET_TCGV(rets[i]);
- }
- for(i = 0; i < nb_params; i++) {
- *gen_opparam_ptr++ = GET_TCGV(params[i]);
- }
- *gen_opparam_ptr++ = GET_TCGV(func);
-
- *gen_opparam_ptr++ = flags;
- /* total parameters, needed to go backward in the instruction stream */
- *gen_opparam_ptr++ = 1 + nb_rets + nb_params + 3;
-}
-
-
+ nparam = gen_opparam_ptr++;
+ call_type = (flags & TCG_CALL_TYPE_MASK);
+ if (ret != TCG_CALL_DUMMY_ARG) {
#if TCG_TARGET_REG_BITS < 64
-/* Note: we convert the 64 bit args to 32 bit */
-void tcg_gen_call(TCGContext *s, TCGv func, unsigned int flags,
- unsigned int nb_rets, const TCGv *rets,
- unsigned int nb_params, const TCGv *args1)
-{
- TCGv ret, *args2, rets_2[2], arg;
- int j, i, call_type;
-
- if (nb_rets == 1) {
- ret = rets[0];
- if (tcg_get_base_type(s, ret) == TCG_TYPE_I64) {
+ if (sizemask & 1) {
+#ifdef TCG_TARGET_WORDS_BIGENDIAN
+ *gen_opparam_ptr++ = ret + 1;
+ *gen_opparam_ptr++ = ret;
+#else
+ *gen_opparam_ptr++ = ret;
+ *gen_opparam_ptr++ = ret + 1;
+#endif
nb_rets = 2;
- rets_2[0] = ret;
- rets_2[1] = TCGV_HIGH(ret);
- rets = rets_2;
+ } else
+#endif
+ {
+ *gen_opparam_ptr++ = ret;
+ nb_rets = 1;
}
+ } else {
+ nb_rets = 0;
}
- args2 = alloca((nb_params * 2) * sizeof(TCGv));
- j = 0;
- call_type = (flags & TCG_CALL_TYPE_MASK);
- for(i = 0; i < nb_params; i++) {
- arg = args1[i];
- if (tcg_get_base_type(s, arg) == TCG_TYPE_I64) {
+ real_args = 0;
+ for (i = 0; i < nargs; i++) {
+#if TCG_TARGET_REG_BITS < 64
+ if (sizemask & (2 << i)) {
#ifdef TCG_TARGET_I386
/* REGPARM case: if the third parameter is 64 bit, it is
allocated on the stack */
- if (j == 2 && call_type == TCG_CALL_TYPE_REGPARM) {
+ if (i == 2 && call_type == TCG_CALL_TYPE_REGPARM) {
call_type = TCG_CALL_TYPE_REGPARM_2;
flags = (flags & ~TCG_CALL_TYPE_MASK) | call_type;
}
- args2[j++] = arg;
- args2[j++] = TCGV_HIGH(arg);
-#else
+#endif
+#ifdef TCG_TARGET_CALL_ALIGN_ARGS
+ /* some targets want aligned 64 bit args */
+ if (real_args & 1) {
+ *gen_opparam_ptr++ = TCG_CALL_DUMMY_ARG;
+ real_args++;
+ }
+#endif
#ifdef TCG_TARGET_WORDS_BIGENDIAN
- args2[j++] = TCGV_HOGH(arg);
- args2[j++] = arg;
+ *gen_opparam_ptr++ = args[i] + 1;
+ *gen_opparam_ptr++ = args[i];
#else
- args2[j++] = arg;
- args2[j++] = TCGV_HIGH(arg);
+ *gen_opparam_ptr++ = args[i];
+ *gen_opparam_ptr++ = args[i] + 1;
#endif
+ real_args += 2;
+ } else
#endif
- } else {
- args2[j++] = arg;
+ {
+ *gen_opparam_ptr++ = args[i];
+ real_args++;
}
}
- tcg_gen_call_internal(s, func, flags,
- nb_rets, rets, j, args2);
-}
-#else
-void tcg_gen_call(TCGContext *s, TCGv func, unsigned int flags,
- unsigned int nb_rets, const TCGv *rets,
- unsigned int nb_params, const TCGv *args1)
-{
- tcg_gen_call_internal(s, func, flags,
- nb_rets, rets, nb_params, args1);
+ *gen_opparam_ptr++ = GET_TCGV_PTR(func);
+
+ *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;
}
-#endif
#if TCG_TARGET_REG_BITS == 32
-void tcg_gen_shifti_i64(TCGv ret, TCGv arg1,
+void tcg_gen_shifti_i64(TCGv_i64 ret, TCGv_i64 arg1,
int c, int right, int arith)
{
- if (c == 0)
- return;
- if (c >= 32) {
+ if (c == 0) {
+ tcg_gen_mov_i32(TCGV_LOW(ret), TCGV_LOW(arg1));
+ tcg_gen_mov_i32(TCGV_HIGH(ret), TCGV_HIGH(arg1));
+ } else if (c >= 32) {
c -= 32;
if (right) {
if (arith) {
- tcg_gen_sari_i32(ret, TCGV_HIGH(arg1), c);
+ tcg_gen_sari_i32(TCGV_LOW(ret), TCGV_HIGH(arg1), c);
tcg_gen_sari_i32(TCGV_HIGH(ret), TCGV_HIGH(arg1), 31);
} else {
- tcg_gen_shri_i32(ret, TCGV_HIGH(arg1), c);
+ tcg_gen_shri_i32(TCGV_LOW(ret), TCGV_HIGH(arg1), c);
tcg_gen_movi_i32(TCGV_HIGH(ret), 0);
}
} else {
- tcg_gen_shli_i32(TCGV_HIGH(ret), arg1, c);
- tcg_gen_movi_i32(ret, 0);
+ tcg_gen_shli_i32(TCGV_HIGH(ret), TCGV_LOW(arg1), c);
+ tcg_gen_movi_i32(TCGV_LOW(ret), 0);
}
} else {
- TCGv t0, t1;
+ TCGv_i32 t0, t1;
- t0 = tcg_temp_new(TCG_TYPE_I32);
- t1 = tcg_temp_new(TCG_TYPE_I32);
+ t0 = tcg_temp_new_i32();
+ t1 = tcg_temp_new_i32();
if (right) {
tcg_gen_shli_i32(t0, TCGV_HIGH(arg1), 32 - c);
if (arith)
tcg_gen_sari_i32(t1, TCGV_HIGH(arg1), c);
- else
+ else
tcg_gen_shri_i32(t1, TCGV_HIGH(arg1), c);
- tcg_gen_shri_i32(ret, arg1, c);
- tcg_gen_or_i32(ret, ret, t0);
+ tcg_gen_shri_i32(TCGV_LOW(ret), TCGV_LOW(arg1), c);
+ tcg_gen_or_i32(TCGV_LOW(ret), TCGV_LOW(ret), t0);
tcg_gen_mov_i32(TCGV_HIGH(ret), t1);
} else {
- tcg_gen_shri_i32(t0, arg1, 32 - c);
+ tcg_gen_shri_i32(t0, TCGV_LOW(arg1), 32 - c);
/* Note: ret can be the same as arg1, so we use t1 */
- tcg_gen_shli_i32(t1, arg1, c);
+ tcg_gen_shli_i32(t1, TCGV_LOW(arg1), c);
tcg_gen_shli_i32(TCGV_HIGH(ret), TCGV_HIGH(arg1), c);
tcg_gen_or_i32(TCGV_HIGH(ret), TCGV_HIGH(ret), t0);
- tcg_gen_mov_i32(ret, t1);
+ tcg_gen_mov_i32(TCGV_LOW(ret), t1);
}
+ tcg_temp_free_i32(t0);
+ tcg_temp_free_i32(t1);
}
}
#endif
-void tcg_reg_alloc_start(TCGContext *s)
+static void tcg_reg_alloc_start(TCGContext *s)
{
int i;
TCGTemp *ts;
ts->val_type = TEMP_VAL_MEM;
}
}
+ for(i = s->nb_globals; i < s->nb_temps; i++) {
+ ts = &s->temps[i];
+ ts->val_type = TEMP_VAL_DEAD;
+ ts->mem_allocated = 0;
+ ts->fixed_reg = 0;
+ }
for(i = 0; i < TCG_TARGET_NB_REGS; i++) {
s->reg_to_temp[i] = -1;
}
if (idx < s->nb_globals) {
pstrcpy(buf, buf_size, ts->name);
} else {
- if (ts->val_type == TEMP_VAL_CONST) {
- snprintf(buf, buf_size, "$0x%" TCG_PRIlx , ts->val);
- } else {
+ if (ts->temp_local)
+ snprintf(buf, buf_size, "loc%d", idx - s->nb_globals);
+ else
snprintf(buf, buf_size, "tmp%d", idx - s->nb_globals);
- }
}
return buf;
}
-char *tcg_get_arg_str(TCGContext *s, char *buf, int buf_size, TCGv arg)
+char *tcg_get_arg_str_i32(TCGContext *s, char *buf, int buf_size, TCGv_i32 arg)
+{
+ return tcg_get_arg_str_idx(s, buf, buf_size, GET_TCGV_I32(arg));
+}
+
+char *tcg_get_arg_str_i64(TCGContext *s, char *buf, int buf_size, TCGv_i64 arg)
{
- return tcg_get_arg_str_idx(s, buf, buf_size, GET_TCGV(arg));
+ return tcg_get_arg_str_idx(s, buf, buf_size, GET_TCGV_I64(arg));
}
+static int helper_cmp(const void *p1, const void *p2)
+{
+ const TCGHelperInfo *th1 = p1;
+ const TCGHelperInfo *th2 = p2;
+ if (th1->func < th2->func)
+ return -1;
+ else if (th1->func == th2->func)
+ return 0;
+ else
+ return 1;
+}
+
+/* find helper definition (Note: A hash table would be better) */
+static TCGHelperInfo *tcg_find_helper(TCGContext *s, tcg_target_ulong val)
+{
+ int m, m_min, m_max;
+ TCGHelperInfo *th;
+ tcg_target_ulong v;
+
+ if (unlikely(!s->helpers_sorted)) {
+ qsort(s->helpers, s->nb_helpers, sizeof(TCGHelperInfo),
+ helper_cmp);
+ s->helpers_sorted = 1;
+ }
+
+ /* binary search */
+ m_min = 0;
+ m_max = s->nb_helpers - 1;
+ while (m_min <= m_max) {
+ m = (m_min + m_max) >> 1;
+ th = &s->helpers[m];
+ v = th->func;
+ if (v == val)
+ return th;
+ else if (val < v) {
+ m_max = m - 1;
+ } else {
+ m_min = m + 1;
+ }
+ }
+ return NULL;
+}
+
+static const char * const cond_name[] =
+{
+ [TCG_COND_EQ] = "eq",
+ [TCG_COND_NE] = "ne",
+ [TCG_COND_LT] = "lt",
+ [TCG_COND_GE] = "ge",
+ [TCG_COND_LE] = "le",
+ [TCG_COND_GT] = "gt",
+ [TCG_COND_LTU] = "ltu",
+ [TCG_COND_GEU] = "geu",
+ [TCG_COND_LEU] = "leu",
+ [TCG_COND_GTU] = "gtu"
+};
+
void tcg_dump_ops(TCGContext *s, FILE *outfile)
{
const uint16_t *opc_ptr;
const TCGArg *args;
TCGArg arg;
- int c, i, k, nb_oargs, nb_iargs, nb_cargs;
+ int c, i, k, nb_oargs, nb_iargs, nb_cargs, first_insn;
const TCGOpDef *def;
char buf[128];
+ first_insn = 1;
opc_ptr = gen_opc_buf;
args = gen_opparam_buf;
while (opc_ptr < gen_opc_ptr) {
c = *opc_ptr++;
def = &tcg_op_defs[c];
- fprintf(outfile, " %s ", def->name);
- if (c == INDEX_op_call) {
+ if (c == INDEX_op_debug_insn_start) {
+ uint64_t pc;
+#if TARGET_LONG_BITS > TCG_TARGET_REG_BITS
+ pc = ((uint64_t)args[1] << 32) | args[0];
+#else
+ pc = args[0];
+#endif
+ if (!first_insn)
+ fprintf(outfile, "\n");
+ fprintf(outfile, " ---- 0x%" PRIx64, pc);
+ first_insn = 0;
+ nb_oargs = def->nb_oargs;
+ nb_iargs = def->nb_iargs;
+ nb_cargs = def->nb_cargs;
+ } else if (c == INDEX_op_call) {
TCGArg arg;
+
/* variable number of arguments */
arg = *args++;
nb_oargs = arg >> 16;
nb_iargs = arg & 0xffff;
nb_cargs = def->nb_cargs;
- } else if (c == INDEX_op_nopn) {
- /* variable number of arguments */
- nb_cargs = *args;
- nb_oargs = 0;
- nb_iargs = 0;
- } else {
- nb_oargs = def->nb_oargs;
- nb_iargs = def->nb_iargs;
- nb_cargs = def->nb_cargs;
- }
- k = 0;
- for(i = 0; i < nb_oargs; i++) {
- if (k != 0)
- fprintf(outfile, ",");
+ fprintf(outfile, " %s ", def->name);
+
+ /* function name */
fprintf(outfile, "%s",
- tcg_get_arg_str_idx(s, buf, sizeof(buf), args[k++]));
- }
- for(i = 0; i < nb_iargs; i++) {
- if (k != 0)
+ tcg_get_arg_str_idx(s, buf, sizeof(buf), args[nb_oargs + nb_iargs - 1]));
+ /* flags */
+ fprintf(outfile, ",$0x%" TCG_PRIlx,
+ args[nb_oargs + nb_iargs]);
+ /* nb out args */
+ fprintf(outfile, ",$%d", nb_oargs);
+ for(i = 0; i < nb_oargs; i++) {
fprintf(outfile, ",");
- /* XXX: dump helper name for call */
- fprintf(outfile, "%s",
- tcg_get_arg_str_idx(s, buf, sizeof(buf), args[k++]));
- }
- for(i = 0; i < nb_cargs; i++) {
- if (k != 0)
+ fprintf(outfile, "%s",
+ tcg_get_arg_str_idx(s, buf, sizeof(buf), args[i]));
+ }
+ for(i = 0; i < (nb_iargs - 1); i++) {
fprintf(outfile, ",");
- arg = args[k++];
- fprintf(outfile, "$0x%" TCG_PRIlx, arg);
+ if (args[nb_oargs + i] == TCG_CALL_DUMMY_ARG) {
+ fprintf(outfile, "<dummy>");
+ } else {
+ fprintf(outfile, "%s",
+ tcg_get_arg_str_idx(s, buf, sizeof(buf), args[nb_oargs + i]));
+ }
+ }
+ } else if (c == INDEX_op_movi_i32
+#if TCG_TARGET_REG_BITS == 64
+ || c == INDEX_op_movi_i64
+#endif
+ ) {
+ tcg_target_ulong val;
+ TCGHelperInfo *th;
+
+ nb_oargs = def->nb_oargs;
+ nb_iargs = def->nb_iargs;
+ nb_cargs = def->nb_cargs;
+ fprintf(outfile, " %s %s,$", def->name,
+ tcg_get_arg_str_idx(s, buf, sizeof(buf), args[0]));
+ val = args[1];
+ th = tcg_find_helper(s, val);
+ if (th) {
+ fprintf(outfile, "%s", th->name);
+ } else {
+ if (c == INDEX_op_movi_i32)
+ fprintf(outfile, "0x%x", (uint32_t)val);
+ else
+ fprintf(outfile, "0x%" PRIx64 , (uint64_t)val);
+ }
+ } else {
+ fprintf(outfile, " %s ", def->name);
+ if (c == INDEX_op_nopn) {
+ /* variable number of arguments */
+ nb_cargs = *args;
+ nb_oargs = 0;
+ nb_iargs = 0;
+ } else {
+ nb_oargs = def->nb_oargs;
+ nb_iargs = def->nb_iargs;
+ nb_cargs = def->nb_cargs;
+ }
+
+ k = 0;
+ for(i = 0; i < nb_oargs; i++) {
+ if (k != 0)
+ fprintf(outfile, ",");
+ fprintf(outfile, "%s",
+ tcg_get_arg_str_idx(s, buf, sizeof(buf), args[k++]));
+ }
+ for(i = 0; i < nb_iargs; i++) {
+ if (k != 0)
+ fprintf(outfile, ",");
+ fprintf(outfile, "%s",
+ tcg_get_arg_str_idx(s, buf, sizeof(buf), args[k++]));
+ }
+ if (c == INDEX_op_brcond_i32
+#if TCG_TARGET_REG_BITS == 32
+ || c == INDEX_op_brcond2_i32
+#elif TCG_TARGET_REG_BITS == 64
+ || c == INDEX_op_brcond_i64
+#endif
+ ) {
+ if (args[k] < ARRAY_SIZE(cond_name) && cond_name[args[k]])
+ fprintf(outfile, ",%s", cond_name[args[k++]]);
+ else
+ fprintf(outfile, ",$0x%" TCG_PRIlx, args[k++]);
+ i = 1;
+ }
+ else
+ i = 0;
+ for(; i < nb_cargs; i++) {
+ if (k != 0)
+ fprintf(outfile, ",");
+ arg = args[k++];
+ fprintf(outfile, "$0x%" TCG_PRIlx, arg);
+ }
}
fprintf(outfile, "\n");
args += nb_iargs + nb_oargs + nb_cargs;
}
}
-/* liveness analysis: end of basic block: globals are live, temps are dead */
-static inline void tcg_la_bb_end(TCGContext *s, uint8_t *dead_temps)
+/* 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)
{
memset(dead_temps, 0, s->nb_globals);
memset(dead_temps + s->nb_globals, 1, 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)
+{
+ int i;
+ TCGTemp *ts;
+
+ memset(dead_temps, 0, s->nb_globals);
+ ts = &s->temps[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++;
+ }
+}
+
/* Liveness analysis : update the opc_dead_iargs array to tell if a
given input arguments is dead. Instructions updating dead
temporaries are removed. */
-void tcg_liveness_analysis(TCGContext *s)
+static void tcg_liveness_analysis(TCGContext *s)
{
int i, op_index, op, nb_args, nb_iargs, nb_oargs, arg, nb_ops;
TCGArg *args;
def = &tcg_op_defs[op];
switch(op) {
case INDEX_op_call:
- nb_args = args[-1];
- args -= nb_args;
- nb_iargs = args[0] & 0xffff;
- nb_oargs = args[0] >> 16;
- args++;
+ {
+ int call_flags;
- /* output args are dead */
- for(i = 0; i < nb_oargs; i++) {
- arg = args[i];
- dead_temps[arg] = 1;
- }
-
- /* globals are live (they may be used by the call) */
- memset(dead_temps, 0, s->nb_globals);
+ nb_args = args[-1];
+ args -= nb_args;
+ nb_iargs = args[0] & 0xffff;
+ nb_oargs = args[0] >> 16;
+ args++;
+ call_flags = args[nb_oargs + nb_iargs];
+
+ /* pure functions can be removed if their result is not
+ used */
+ if (call_flags & TCG_CALL_PURE) {
+ for(i = 0; i < nb_oargs; i++) {
+ arg = args[i];
+ if (!dead_temps[arg])
+ goto do_not_remove_call;
+ }
+ tcg_set_nop(s, gen_opc_buf + op_index,
+ args - 1, nb_args);
+ } else {
+ do_not_remove_call:
- /* input args are live */
- dead_iargs = 0;
- for(i = 0; i < nb_iargs; i++) {
- arg = args[i + nb_oargs];
- if (dead_temps[arg]) {
- dead_iargs |= (1 << i);
+ /* output args are dead */
+ for(i = 0; i < nb_oargs; i++) {
+ arg = args[i];
+ dead_temps[arg] = 1;
+ }
+
+ if (!(call_flags & TCG_CALL_CONST)) {
+ /* globals are live (they may be used by the call) */
+ memset(dead_temps, 0, s->nb_globals);
+ }
+
+ /* input args are live */
+ dead_iargs = 0;
+ for(i = 0; i < nb_iargs; i++) {
+ arg = args[i + nb_oargs];
+ if (arg != TCG_CALL_DUMMY_ARG) {
+ if (dead_temps[arg]) {
+ dead_iargs |= (1 << i);
+ }
+ dead_temps[arg] = 0;
+ }
+ }
+ s->op_dead_iargs[op_index] = dead_iargs;
}
- dead_temps[arg] = 0;
+ args--;
}
- s->op_dead_iargs[op_index] = dead_iargs;
- args--;
break;
case INDEX_op_set_label:
args--;
/* mark end of basic block */
tcg_la_bb_end(s, dead_temps);
break;
+ case INDEX_op_debug_insn_start:
+ args -= def->nb_args;
+ break;
case INDEX_op_nopn:
nb_args = args[-1];
args -= nb_args;
/* mark the temporary as dead */
dead_temps[args[0]] = 1;
break;
- case INDEX_op_macro_2:
- {
- int dead_args[2], macro_id;
- int saved_op_index, saved_arg_index;
- int macro_op_index, macro_arg_index;
- int macro_end_op_index, macro_end_arg_index;
- int last_nb_temps;
-
- nb_args = 3;
- args -= nb_args;
- dead_args[0] = dead_temps[args[0]];
- dead_args[1] = dead_temps[args[1]];
- macro_id = args[2];
-
- /* call the macro function which generate code
- depending on the live outputs */
- saved_op_index = op_index;
- saved_arg_index = args - gen_opparam_buf;
-
- /* add a macro start instruction */
- *gen_opc_ptr++ = INDEX_op_macro_start;
- *gen_opparam_ptr++ = saved_op_index;
- *gen_opparam_ptr++ = saved_arg_index;
-
- macro_op_index = gen_opc_ptr - gen_opc_buf;
- macro_arg_index = gen_opparam_ptr - gen_opparam_buf;
-
- last_nb_temps = s->nb_temps;
-
- s->macro_func(s, macro_id, dead_args);
-
- /* realloc temp info (XXX: make it faster) */
- if (s->nb_temps > last_nb_temps) {
- uint8_t *new_dead_temps;
-
- new_dead_temps = tcg_malloc(s->nb_temps);
- memcpy(new_dead_temps, dead_temps, last_nb_temps);
- memset(new_dead_temps + last_nb_temps, 1,
- s->nb_temps - last_nb_temps);
- dead_temps = new_dead_temps;
- }
-
- macro_end_op_index = gen_opc_ptr - gen_opc_buf;
- macro_end_arg_index = gen_opparam_ptr - gen_opparam_buf;
-
- /* end of macro: add a goto to the next instruction */
- *gen_opc_ptr++ = INDEX_op_macro_end;
- *gen_opparam_ptr++ = op_index + 1;
- *gen_opparam_ptr++ = saved_arg_index + nb_args;
-
- /* modify the macro operation to be a macro_goto */
- gen_opc_buf[op_index] = INDEX_op_macro_goto;
- args[0] = macro_op_index;
- args[1] = macro_arg_index;
- args[2] = 0; /* dummy third arg to match the
- macro parameters */
-
- /* set the next instruction to the end of the macro */
- op_index = macro_end_op_index;
- args = macro_end_arg_index + gen_opparam_buf;
- }
- break;
- case INDEX_op_macro_start:
- args -= 2;
- op_index = args[0];
- args = gen_opparam_buf + args[1];
- break;
- case INDEX_op_macro_goto:
- case INDEX_op_macro_end:
- tcg_abort(); /* should never happen in liveness analysis */
case INDEX_op_end:
break;
/* XXX: optimize by hardcoding common cases (e.g. triadic ops) */
default:
- if (op > INDEX_op_end) {
- args -= def->nb_args;
- nb_iargs = def->nb_iargs;
- nb_oargs = def->nb_oargs;
+ args -= def->nb_args;
+ nb_iargs = def->nb_iargs;
+ nb_oargs = def->nb_oargs;
- /* Test if the operation can be removed because all
- its outputs are dead. We assume that nb_oargs == 0
- implies side effects */
- if (!(def->flags & TCG_OPF_SIDE_EFFECTS) && nb_oargs != 0) {
- for(i = 0; i < nb_oargs; i++) {
- arg = args[i];
- if (!dead_temps[arg])
- goto do_not_remove;
- }
- tcg_set_nop(s, gen_opc_buf + op_index, args, def->nb_args);
+ /* Test if the operation can be removed because all
+ its outputs are dead. We assume that nb_oargs == 0
+ implies side effects */
+ if (!(def->flags & TCG_OPF_SIDE_EFFECTS) && nb_oargs != 0) {
+ for(i = 0; i < nb_oargs; i++) {
+ arg = args[i];
+ if (!dead_temps[arg])
+ goto do_not_remove;
+ }
+ tcg_set_nop(s, gen_opc_buf + op_index, args, def->nb_args);
#ifdef CONFIG_PROFILER
- {
- extern int64_t dyngen_tcg_del_op_count;
- dyngen_tcg_del_op_count++;
- }
+ s->del_op_count++;
#endif
- } else {
- do_not_remove:
+ } else {
+ do_not_remove:
- /* output args are dead */
- for(i = 0; i < nb_oargs; i++) {
- arg = args[i];
- dead_temps[arg] = 1;
- }
-
- /* if end of basic block, update */
- if (def->flags & TCG_OPF_BB_END) {
- tcg_la_bb_end(s, dead_temps);
- }
-
- /* input args are live */
- dead_iargs = 0;
- for(i = 0; i < nb_iargs; i++) {
- arg = args[i + nb_oargs];
- if (dead_temps[arg]) {
- dead_iargs |= (1 << i);
- }
- dead_temps[arg] = 0;
+ /* output args are dead */
+ for(i = 0; i < nb_oargs; i++) {
+ arg = args[i];
+ dead_temps[arg] = 1;
+ }
+
+ /* 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);
+ }
+
+ /* input args are live */
+ dead_iargs = 0;
+ for(i = 0; i < nb_iargs; i++) {
+ arg = args[i + nb_oargs];
+ if (dead_temps[arg]) {
+ dead_iargs |= (1 << i);
}
- s->op_dead_iargs[op_index] = dead_iargs;
+ dead_temps[arg] = 0;
}
- } else {
- /* legacy dyngen operations */
- args -= def->nb_args;
- /* mark end of basic block */
- tcg_la_bb_end(s, dead_temps);
+ s->op_dead_iargs[op_index] = dead_iargs;
}
break;
}
ts->reg != reg) {
printf("Inconsistency for register %s:\n",
tcg_target_reg_names[reg]);
- printf("reg state:\n");
- dump_regs(s);
- tcg_abort();
+ goto fail;
}
}
}
s->reg_to_temp[ts->reg] != k) {
printf("Inconsistency for temp %s:\n",
tcg_get_arg_str_idx(s, buf, sizeof(buf), k));
+ fail:
printf("reg state:\n");
dump_regs(s);
tcg_abort();
if (!ts->mem_coherent) {
if (!ts->mem_allocated)
temp_allocate_frame(s, temp);
- tcg_out_st(s, reg, ts->mem_reg, ts->mem_offset);
+ tcg_out_st(s, ts->type, reg, ts->mem_reg, ts->mem_offset);
}
ts->val_type = TEMP_VAL_MEM;
s->reg_to_temp[reg] = -1;
tcg_regset_andnot(reg_ct, reg1, reg2);
/* first try free registers */
- for(i = 0; i < TCG_TARGET_NB_REGS; i++) {
+ for(i = 0; i < ARRAY_SIZE(tcg_target_reg_alloc_order); i++) {
reg = tcg_target_reg_alloc_order[i];
if (tcg_regset_test_reg(reg_ct, reg) && s->reg_to_temp[reg] == -1)
return reg;
}
/* XXX: do better spill choice */
- for(i = 0; i < TCG_TARGET_NB_REGS; i++) {
+ for(i = 0; i < ARRAY_SIZE(tcg_target_reg_alloc_order); i++) {
reg = tcg_target_reg_alloc_order[i];
if (tcg_regset_test_reg(reg_ct, reg)) {
tcg_reg_free(s, reg);
tcg_abort();
}
-/* at the end of a basic block, we assume all temporaries are dead and
- all globals are stored at their canonical location */
-/* XXX: optimize by handling constants in another array ? */
-void tcg_reg_alloc_bb_end(TCGContext *s)
+/* save 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)
{
TCGTemp *ts;
+ int reg;
+
+ ts = &s->temps[temp];
+ if (!ts->fixed_reg) {
+ switch(ts->val_type) {
+ case TEMP_VAL_REG:
+ tcg_reg_free(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:
+ tcg_abort();
+ }
+ }
+}
+
+/* save globals to their cannonical 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. */
+static void save_globals(TCGContext *s, TCGRegSet allocated_regs)
+{
int i;
for(i = 0; i < s->nb_globals; i++) {
- ts = &s->temps[i];
- if (!ts->fixed_reg) {
- if (ts->val_type == TEMP_VAL_REG) {
- tcg_reg_free(s, ts->reg);
- }
- }
+ temp_save(s, i, allocated_regs);
}
+}
+
+/* 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)
+{
+ TCGTemp *ts;
+ int i;
for(i = s->nb_globals; i < s->nb_temps; i++) {
ts = &s->temps[i];
- if (ts->val_type != TEMP_VAL_CONST) {
+ 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;
}
}
+
+ save_globals(s, allocated_regs);
}
#define IS_DEAD_IARG(n) ((dead_iargs >> (n)) & 1)
+static void tcg_reg_alloc_movi(TCGContext *s, const TCGArg *args)
+{
+ TCGTemp *ots;
+ tcg_target_ulong val;
+
+ ots = &s->temps[args[0]];
+ val = args[1];
+
+ if (ots->fixed_reg) {
+ /* for fixed registers, we do not do any constant
+ propagation */
+ tcg_out_movi(s, ots->type, ots->reg, val);
+ } else {
+ /* The movi is not explicitly generated here */
+ if (ots->val_type == TEMP_VAL_REG)
+ s->reg_to_temp[ots->reg] = -1;
+ ots->val_type = TEMP_VAL_CONST;
+ ots->val = val;
+ }
+}
+
static void tcg_reg_alloc_mov(TCGContext *s, const TCGOpDef *def,
const TCGArg *args,
unsigned int dead_iargs)
ts = &s->temps[args[1]];
arg_ct = &def->args_ct[0];
+ /* XXX: always mark arg dead if IS_DEAD_IARG(0) */
if (ts->val_type == TEMP_VAL_REG) {
if (IS_DEAD_IARG(0) && !ts->fixed_reg && !ots->fixed_reg) {
/* the mov can be suppressed */
} else {
reg = tcg_reg_alloc(s, arg_ct->u.regs, s->reserved_regs);
}
- tcg_out_ld(s, reg, ts->mem_reg, ts->mem_offset);
+ tcg_out_ld(s, ts->type, reg, ts->mem_reg, ts->mem_offset);
} else if (ts->val_type == TEMP_VAL_CONST) {
- if (ots->val_type == TEMP_VAL_REG) {
+ if (ots->fixed_reg) {
reg = ots->reg;
+ tcg_out_movi(s, ots->type, reg, ts->val);
} else {
- reg = tcg_reg_alloc(s, arg_ct->u.regs, s->reserved_regs);
+ /* 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;
}
- tcg_out_movi(s, ots->type, reg, ts->val);
} else {
tcg_abort();
}
ts = &s->temps[arg];
if (ts->val_type == TEMP_VAL_MEM) {
reg = tcg_reg_alloc(s, arg_ct->u.regs, allocated_regs);
- tcg_out_ld(s, reg, ts->mem_reg, ts->mem_offset);
+ tcg_out_ld(s, ts->type, reg, ts->mem_reg, ts->mem_offset);
ts->val_type = TEMP_VAL_REG;
ts->reg = reg;
ts->mem_coherent = 1;
new_args[i] = ts->val;
goto iarg_end;
} else {
- /* need to move to a register*/
+ /* need to move to a register */
reg = tcg_reg_alloc(s, arg_ct->u.regs, allocated_regs);
tcg_out_movi(s, ts->type, reg, ts->val);
- goto iarg_end1;
+ ts->val_type = TEMP_VAL_REG;
+ ts->reg = reg;
+ ts->mem_coherent = 0;
+ s->reg_to_temp[reg] = arg;
}
}
assert(ts->val_type == TEMP_VAL_REG);
reg = tcg_reg_alloc(s, arg_ct->u.regs, allocated_regs);
tcg_out_mov(s, reg, ts->reg);
}
- iarg_end1:
new_args[i] = reg;
const_args[i] = 0;
tcg_regset_set_reg(allocated_regs, reg);
iarg_end: ;
}
- /* mark dead temporaries and free the associated registers */
- for(i = 0; i < nb_iargs; i++) {
- arg = args[nb_oargs + i];
- if (IS_DEAD_IARG(i)) {
- ts = &s->temps[arg];
- if (ts->val_type != TEMP_VAL_CONST && !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_BB_END) {
+ tcg_reg_alloc_bb_end(s, allocated_regs);
+ } else {
+ /* mark dead temporaries and free the associated registers */
+ for(i = 0; i < nb_iargs; i++) {
+ arg = args[nb_oargs + i];
+ if (IS_DEAD_IARG(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;
+ }
}
}
- }
-
- /* XXX: permit generic clobber register list ? */
- if (def->flags & TCG_OPF_CALL_CLOBBER) {
- for(reg = 0; reg < TCG_TARGET_NB_REGS; reg++) {
- if (tcg_regset_test_reg(tcg_target_call_clobber_regs, reg)) {
- tcg_reg_free(s, reg);
+
+ if (def->flags & TCG_OPF_CALL_CLOBBER) {
+ /* XXX: permit generic clobber register list ? */
+ for(reg = 0; reg < TCG_TARGET_NB_REGS; reg++) {
+ if (tcg_regset_test_reg(tcg_target_call_clobber_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);
}
- }
-
- /* satisfy the output constraints */
- tcg_regset_set(allocated_regs, s->reserved_regs);
- for(k = 0; k < nb_oargs; k++) {
- i = def->sorted_args[k];
- arg = args[i];
- arg_ct = &def->args_ct[i];
- ts = &s->temps[arg];
- if (arg_ct->ct & TCG_CT_ALIAS) {
- reg = new_args[arg_ct->alias_index];
- } else {
- /* if fixed register, we try to use it */
- reg = ts->reg;
- if (ts->fixed_reg &&
- tcg_regset_test_reg(arg_ct->u.regs, reg)) {
- goto oarg_end;
+
+ /* satisfy the output constraints */
+ tcg_regset_set(allocated_regs, s->reserved_regs);
+ for(k = 0; k < nb_oargs; k++) {
+ i = def->sorted_args[k];
+ arg = args[i];
+ arg_ct = &def->args_ct[i];
+ ts = &s->temps[arg];
+ if (arg_ct->ct & TCG_CT_ALIAS) {
+ reg = new_args[arg_ct->alias_index];
+ } else {
+ /* if fixed register, we try to use it */
+ reg = ts->reg;
+ if (ts->fixed_reg &&
+ tcg_regset_test_reg(arg_ct->u.regs, reg)) {
+ goto oarg_end;
+ }
+ reg = tcg_reg_alloc(s, arg_ct->u.regs, allocated_regs);
}
- reg = tcg_reg_alloc(s, arg_ct->u.regs, allocated_regs);
- }
- 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)
- s->reg_to_temp[ts->reg] = -1;
- 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;
+ 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)
+ s->reg_to_temp[ts->reg] = -1;
+ 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;
}
- oarg_end:
- new_args[i] = reg;
}
- if (def->flags & TCG_OPF_BB_END)
- tcg_reg_alloc_bb_end(s);
-
/* emit instruction */
tcg_out_op(s, opc, new_args, const_args);
}
}
+#ifdef TCG_TARGET_STACK_GROWSUP
+#define STACK_DIR(x) (-(x))
+#else
+#define STACK_DIR(x) (x)
+#endif
+
static int tcg_reg_alloc_call(TCGContext *s, const TCGOpDef *def,
int opc, const TCGArg *args,
unsigned int dead_iargs)
int nb_iargs, nb_oargs, flags, nb_regs, i, reg, nb_params;
TCGArg arg, func_arg;
TCGTemp *ts;
- tcg_target_long stack_offset, call_stack_size;
- int const_func_arg;
+ tcg_target_long stack_offset, call_stack_size, func_addr;
+ int const_func_arg, allocate_args;
TCGRegSet allocated_regs;
const TCGArgConstraint *arg_ct;
call_stack_size = (nb_params - nb_regs) * sizeof(tcg_target_long);
call_stack_size = (call_stack_size + TCG_TARGET_STACK_ALIGN - 1) &
~(TCG_TARGET_STACK_ALIGN - 1);
- tcg_out_addi(s, TCG_REG_CALL_STACK, -call_stack_size);
+ allocate_args = (call_stack_size > TCG_STATIC_CALL_ARGS_SIZE);
+ if (allocate_args) {
+ tcg_out_addi(s, TCG_REG_CALL_STACK, -STACK_DIR(call_stack_size));
+ }
- stack_offset = 0;
+ stack_offset = TCG_TARGET_CALL_STACK_OFFSET;
for(i = nb_regs; i < nb_params; i++) {
arg = args[nb_oargs + i];
- ts = &s->temps[arg];
- if (ts->val_type == TEMP_VAL_REG) {
- tcg_out_st(s, ts->reg, TCG_REG_CALL_STACK, stack_offset);
- } else if (ts->val_type == TEMP_VAL_MEM) {
- reg = tcg_reg_alloc(s, tcg_target_available_regs[ts->type],
- s->reserved_regs);
- /* XXX: not correct if reading values from the stack */
- tcg_out_ld(s, reg, ts->mem_reg, ts->mem_offset);
- tcg_out_st(s, reg, TCG_REG_CALL_STACK, stack_offset);
- } else if (ts->val_type == TEMP_VAL_CONST) {
- reg = tcg_reg_alloc(s, tcg_target_available_regs[ts->type],
- s->reserved_regs);
- /* XXX: sign extend may be needed on some targets */
- tcg_out_movi(s, ts->type, reg, ts->val);
- tcg_out_st(s, reg, TCG_REG_CALL_STACK, stack_offset);
- } else {
- tcg_abort();
+#ifdef TCG_TARGET_STACK_GROWSUP
+ stack_offset -= sizeof(tcg_target_long);
+#endif
+ if (arg != TCG_CALL_DUMMY_ARG) {
+ ts = &s->temps[arg];
+ if (ts->val_type == TEMP_VAL_REG) {
+ tcg_out_st(s, ts->type, ts->reg, TCG_REG_CALL_STACK, stack_offset);
+ } else if (ts->val_type == TEMP_VAL_MEM) {
+ reg = tcg_reg_alloc(s, tcg_target_available_regs[ts->type],
+ s->reserved_regs);
+ /* XXX: not correct if reading values from the stack */
+ tcg_out_ld(s, ts->type, reg, ts->mem_reg, ts->mem_offset);
+ tcg_out_st(s, ts->type, reg, TCG_REG_CALL_STACK, stack_offset);
+ } else if (ts->val_type == TEMP_VAL_CONST) {
+ reg = tcg_reg_alloc(s, tcg_target_available_regs[ts->type],
+ s->reserved_regs);
+ /* XXX: sign extend may be needed on some targets */
+ tcg_out_movi(s, ts->type, reg, ts->val);
+ tcg_out_st(s, ts->type, reg, TCG_REG_CALL_STACK, stack_offset);
+ } else {
+ tcg_abort();
+ }
}
+#ifndef TCG_TARGET_STACK_GROWSUP
stack_offset += sizeof(tcg_target_long);
+#endif
}
/* assign input registers */
tcg_regset_set(allocated_regs, s->reserved_regs);
for(i = 0; i < nb_regs; i++) {
arg = args[nb_oargs + i];
- ts = &s->temps[arg];
- reg = tcg_target_call_iarg_regs[i];
- tcg_reg_free(s, reg);
- if (ts->val_type == TEMP_VAL_REG) {
- if (ts->reg != reg) {
- tcg_out_mov(s, reg, ts->reg);
+ if (arg != TCG_CALL_DUMMY_ARG) {
+ ts = &s->temps[arg];
+ reg = tcg_target_call_iarg_regs[i];
+ tcg_reg_free(s, reg);
+ if (ts->val_type == TEMP_VAL_REG) {
+ if (ts->reg != reg) {
+ tcg_out_mov(s, reg, ts->reg);
+ }
+ } else if (ts->val_type == TEMP_VAL_MEM) {
+ tcg_out_ld(s, ts->type, reg, ts->mem_reg, ts->mem_offset);
+ } else if (ts->val_type == TEMP_VAL_CONST) {
+ /* XXX: sign extend ? */
+ tcg_out_movi(s, ts->type, reg, ts->val);
+ } else {
+ tcg_abort();
}
- } else if (ts->val_type == TEMP_VAL_MEM) {
- tcg_out_ld(s, reg, ts->mem_reg, ts->mem_offset);
- } else if (ts->val_type == TEMP_VAL_CONST) {
- /* XXX: sign extend ? */
- tcg_out_movi(s, ts->type, reg, ts->val);
- } else {
- tcg_abort();
+ tcg_regset_set_reg(allocated_regs, reg);
}
- tcg_regset_set_reg(allocated_regs, reg);
}
/* assign function address */
func_arg = args[nb_oargs + nb_iargs - 1];
arg_ct = &def->args_ct[0];
ts = &s->temps[func_arg];
+ func_addr = ts->val;
const_func_arg = 0;
if (ts->val_type == TEMP_VAL_MEM) {
reg = tcg_reg_alloc(s, arg_ct->u.regs, allocated_regs);
- tcg_out_ld(s, reg, ts->mem_reg, ts->mem_offset);
+ tcg_out_ld(s, ts->type, reg, ts->mem_reg, ts->mem_offset);
func_arg = reg;
+ tcg_regset_set_reg(allocated_regs, reg);
} else if (ts->val_type == TEMP_VAL_REG) {
reg = ts->reg;
if (!tcg_regset_test_reg(arg_ct->u.regs, reg)) {
tcg_out_mov(s, reg, ts->reg);
}
func_arg = reg;
+ tcg_regset_set_reg(allocated_regs, reg);
} else if (ts->val_type == TEMP_VAL_CONST) {
- if (tcg_target_const_match(ts->val, arg_ct)) {
+ if (tcg_target_const_match(func_addr, arg_ct)) {
const_func_arg = 1;
- func_arg = ts->val;
+ func_arg = func_addr;
} else {
reg = tcg_reg_alloc(s, arg_ct->u.regs, allocated_regs);
- tcg_out_movi(s, ts->type, reg, ts->val);
+ tcg_out_movi(s, ts->type, reg, func_addr);
func_arg = reg;
+ tcg_regset_set_reg(allocated_regs, reg);
}
} else {
tcg_abort();
}
+
/* mark dead temporaries and free the associated registers */
- for(i = 0; i < nb_params; i++) {
+ for(i = 0; i < nb_iargs; i++) {
arg = args[nb_oargs + i];
if (IS_DEAD_IARG(i)) {
ts = &s->temps[arg];
- if (ts->val_type != TEMP_VAL_CONST && !ts->fixed_reg) {
+ if (!ts->fixed_reg) {
if (ts->val_type == TEMP_VAL_REG)
s->reg_to_temp[ts->reg] = -1;
ts->val_type = TEMP_VAL_DEAD;
/* store globals and free associated registers (we assume the call
can modify any global. */
- for(i = 0; i < s->nb_globals; i++) {
- ts = &s->temps[i];
- if (!ts->fixed_reg) {
- if (ts->val_type == TEMP_VAL_REG) {
- tcg_reg_free(s, ts->reg);
- }
- }
+ if (!(flags & TCG_CALL_CONST)) {
+ save_globals(s, allocated_regs);
}
tcg_out_op(s, opc, &func_arg, &const_func_arg);
- tcg_out_addi(s, TCG_REG_CALL_STACK, call_stack_size);
+ if (allocate_args) {
+ tcg_out_addi(s, TCG_REG_CALL_STACK, STACK_DIR(call_stack_size));
+ }
/* assign output registers and emit moves if needed */
for(i = 0; i < nb_oargs; i++) {
arg = args[i];
ts = &s->temps[arg];
reg = tcg_target_call_oarg_regs[i];
- tcg_reg_free(s, reg);
+ assert(s->reg_to_temp[reg] == -1);
if (ts->fixed_reg) {
if (ts->reg != reg) {
tcg_out_mov(s, ts->reg, reg);
#ifdef CONFIG_PROFILER
-static int64_t dyngen_table_op_count[NB_OPS];
+static int64_t tcg_table_op_count[NB_OPS];
void dump_op_count(void)
{
int i;
FILE *f;
- f = fopen("/tmp/op1.log", "w");
- for(i = 0; i < INDEX_op_end; i++) {
- fprintf(f, "%s %" PRId64 "\n", tcg_op_defs[i].name, dyngen_table_op_count[i]);
- }
- fclose(f);
- f = fopen("/tmp/op2.log", "w");
+ f = fopen("/tmp/op.log", "w");
for(i = INDEX_op_end; i < NB_OPS; i++) {
- fprintf(f, "%s %" PRId64 "\n", tcg_op_defs[i].name, dyngen_table_op_count[i]);
+ fprintf(f, "%s %" PRId64 "\n", tcg_op_defs[i].name, tcg_table_op_count[i]);
}
fclose(f);
}
static inline int tcg_gen_code_common(TCGContext *s, uint8_t *gen_code_buf,
long search_pc)
{
- int opc, op_index, macro_op_index;
+ int opc, op_index;
const TCGOpDef *def;
unsigned int dead_iargs;
const TCGArg *args;
#ifdef DEBUG_DISAS
- if (unlikely(loglevel & CPU_LOG_TB_OP)) {
- fprintf(logfile, "OP:\n");
+ if (unlikely(qemu_loglevel_mask(CPU_LOG_TB_OP))) {
+ qemu_log("OP:\n");
tcg_dump_ops(s, logfile);
- fprintf(logfile, "\n");
+ qemu_log("\n");
}
#endif
+#ifdef CONFIG_PROFILER
+ s->la_time -= profile_getclock();
+#endif
tcg_liveness_analysis(s);
+#ifdef CONFIG_PROFILER
+ s->la_time += profile_getclock();
+#endif
#ifdef DEBUG_DISAS
- if (unlikely(loglevel & CPU_LOG_TB_OP_OPT)) {
- fprintf(logfile, "OP after la:\n");
+ if (unlikely(qemu_loglevel_mask(CPU_LOG_TB_OP_OPT))) {
+ qemu_log("OP after la:\n");
tcg_dump_ops(s, logfile);
- fprintf(logfile, "\n");
+ qemu_log("\n");
}
#endif
s->code_buf = gen_code_buf;
s->code_ptr = gen_code_buf;
- macro_op_index = -1;
args = gen_opparam_buf;
op_index = 0;
+
for(;;) {
opc = gen_opc_buf[op_index];
#ifdef CONFIG_PROFILER
- dyngen_table_op_count[opc]++;
+ tcg_table_op_count[opc]++;
#endif
def = &tcg_op_defs[opc];
#if 0
dead_iargs = s->op_dead_iargs[op_index];
tcg_reg_alloc_mov(s, def, args, dead_iargs);
break;
+ case INDEX_op_movi_i32:
+#if TCG_TARGET_REG_BITS == 64
+ case INDEX_op_movi_i64:
+#endif
+ tcg_reg_alloc_movi(s, args);
+ break;
+ case INDEX_op_debug_insn_start:
+ /* debug instruction */
+ break;
case INDEX_op_nop:
case INDEX_op_nop1:
case INDEX_op_nop2:
TCGTemp *ts;
ts = &s->temps[args[0]];
/* mark the temporary as dead */
- if (ts->val_type != TEMP_VAL_CONST && !ts->fixed_reg) {
+ if (!ts->fixed_reg) {
if (ts->val_type == TEMP_VAL_REG)
s->reg_to_temp[ts->reg] = -1;
ts->val_type = TEMP_VAL_DEAD;
}
}
break;
- case INDEX_op_macro_goto:
- macro_op_index = op_index; /* only used for exceptions */
- op_index = args[0] - 1;
- args = gen_opparam_buf + args[1];
- goto next;
- case INDEX_op_macro_end:
- macro_op_index = -1; /* only used for exceptions */
- op_index = args[0] - 1;
- args = gen_opparam_buf + args[1];
- goto next;
- case INDEX_op_macro_start:
- /* must never happen here */
- tcg_abort();
case INDEX_op_set_label:
- tcg_reg_alloc_bb_end(s);
+ tcg_reg_alloc_bb_end(s, s->reserved_regs);
tcg_out_label(s, args[0], (long)s->code_ptr);
break;
case INDEX_op_call:
goto next;
case INDEX_op_end:
goto the_end;
- case 0 ... INDEX_op_end - 1:
- /* legacy dyngen ops */
-#ifdef CONFIG_PROFILER
- {
- extern int64_t dyngen_old_op_count;
- dyngen_old_op_count++;
- }
-#endif
- tcg_reg_alloc_bb_end(s);
- if (search_pc >= 0) {
- s->code_ptr += def->copy_size;
- args += def->nb_args;
- } else {
- args = dyngen_op(s, opc, args);
- }
- goto next;
default:
/* Note: in order to speed up the code, it would be much
faster to have specialized register allocator functions for
break;
}
args += def->nb_args;
- next: ;
+ next:
if (search_pc >= 0 && search_pc < s->code_ptr - gen_code_buf) {
- if (macro_op_index >= 0)
- return macro_op_index;
- else
- return op_index;
+ return op_index;
}
op_index++;
#ifndef NDEBUG
return -1;
}
-int dyngen_code(TCGContext *s, uint8_t *gen_code_buf)
+int tcg_gen_code(TCGContext *s, uint8_t *gen_code_buf)
{
#ifdef CONFIG_PROFILER
{
- extern int64_t dyngen_op_count;
- extern int dyngen_op_count_max;
int n;
n = (gen_opc_ptr - gen_opc_buf);
- dyngen_op_count += n;
- if (n > dyngen_op_count_max)
- dyngen_op_count_max = n;
+ s->op_count += n;
+ if (n > s->op_count_max)
+ s->op_count_max = n;
+
+ s->temp_count += s->nb_temps;
+ if (s->nb_temps > s->temp_count_max)
+ s->temp_count_max = s->nb_temps;
}
#endif
offset bytes from the start of the TB. The contents of gen_code_buf must
not be changed, though writing the same values is ok.
Return -1 if not found. */
-int dyngen_code_search_pc(TCGContext *s, uint8_t *gen_code_buf, long offset)
+int tcg_gen_code_search_pc(TCGContext *s, uint8_t *gen_code_buf, long offset)
{
return tcg_gen_code_common(s, gen_code_buf, offset);
}
+
+#ifdef CONFIG_PROFILER
+void tcg_dump_info(FILE *f,
+ int (*cpu_fprintf)(FILE *f, const char *fmt, ...))
+{
+ TCGContext *s = &tcg_ctx;
+ int64_t tot;
+
+ tot = s->interm_time + s->code_time;
+ cpu_fprintf(f, "JIT cycles %" PRId64 " (%0.3f s at 2.4 GHz)\n",
+ tot, tot / 2.4e9);
+ cpu_fprintf(f, "translated TBs %" PRId64 " (aborted=%" PRId64 " %0.1f%%)\n",
+ s->tb_count,
+ s->tb_count1 - s->tb_count,
+ s->tb_count1 ? (double)(s->tb_count1 - s->tb_count) / s->tb_count1 * 100.0 : 0);
+ cpu_fprintf(f, "avg ops/TB %0.1f max=%d\n",
+ s->tb_count ? (double)s->op_count / s->tb_count : 0, s->op_count_max);
+ cpu_fprintf(f, "deleted ops/TB %0.2f\n",
+ s->tb_count ?
+ (double)s->del_op_count / s->tb_count : 0);
+ cpu_fprintf(f, "avg temps/TB %0.2f max=%d\n",
+ s->tb_count ?
+ (double)s->temp_count / s->tb_count : 0,
+ s->temp_count_max);
+
+ cpu_fprintf(f, "cycles/op %0.1f\n",
+ s->op_count ? (double)tot / s->op_count : 0);
+ cpu_fprintf(f, "cycles/in byte %0.1f\n",
+ s->code_in_len ? (double)tot / s->code_in_len : 0);
+ cpu_fprintf(f, "cycles/out byte %0.1f\n",
+ s->code_out_len ? (double)tot / s->code_out_len : 0);
+ if (tot == 0)
+ tot = 1;
+ cpu_fprintf(f, " gen_interm time %0.1f%%\n",
+ (double)s->interm_time / tot * 100.0);
+ cpu_fprintf(f, " gen_code time %0.1f%%\n",
+ (double)s->code_time / tot * 100.0);
+ cpu_fprintf(f, "liveness/code time %0.1f%%\n",
+ (double)s->la_time / (s->code_time ? s->code_time : 1) * 100.0);
+ cpu_fprintf(f, "cpu_restore count %" PRId64 "\n",
+ s->restore_count);
+ cpu_fprintf(f, " avg cycles %0.1f\n",
+ s->restore_count ? (double)s->restore_time / s->restore_count : 0);
+ {
+ extern void dump_op_count(void);
+ dump_op_count();
+ }
+}
+#else
+void tcg_dump_info(FILE *f,
+ int (*cpu_fprintf)(FILE *f, const char *fmt, ...))
+{
+ cpu_fprintf(f, "[TCG profiler not compiled]\n");
+}
+#endif
projects / qemu / blobdiff