OpenE2K/qemu-e2k
13
4
Fork 0
Code Issues 4 Pull Requests Projects 2 Releases Activity

tcg: Don't free vector results

Browse Source 
Avoid reusing vector temporaries so that we may re-use them
when propagating stores to loads.

Reviewed-by: Song Gao <gaosong@loongson.cn>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
This commit is contained in:
Richard Henderson 2023-08-23 20:35:05 -07:00
parent b701f195d3
commit 9628d008bd
1 changed files with 38 additions and 74 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

112
tcg/tcg-op-gvec.c
View File

@ -561,7 +561,6 @@ static void do_dup(unsigned vece, uint32_t dofs, uint32_t oprsz,
tcg_gen_dupi_vec(vece, t_vec, in_c);
}
do_dup_store(type, dofs, oprsz, maxsz, t_vec);
tcg_temp_free_vec(t_vec);
return;
}
@ -1024,11 +1023,10 @@ static void expand_2_vec(unsigned vece, uint32_t dofs, uint32_t aofs,
bool load_dest,
void (*fni)(unsigned, TCGv_vec, TCGv_vec))
{
TCGv_vec t0 = tcg_temp_new_vec(type);
TCGv_vec t1 = tcg_temp_new_vec(type);
uint32_t i;
for (uint32_t i = 0; i < oprsz; i += tysz) {
TCGv_vec t0 = tcg_temp_new_vec(type);
TCGv_vec t1 = tcg_temp_new_vec(type);
for (i = 0; i < oprsz; i += tysz) {
tcg_gen_ld_vec(t0, tcg_env, aofs + i);
if (load_dest) {
tcg_gen_ld_vec(t1, tcg_env, dofs + i);
@ -1036,8 +1034,6 @@ static void expand_2_vec(unsigned vece, uint32_t dofs, uint32_t aofs,
fni(vece, t1, t0);
tcg_gen_st_vec(t1, tcg_env, dofs + i);
}
tcg_temp_free_vec(t0);
tcg_temp_free_vec(t1);
}
/* Expand OPSZ bytes worth of two-vector operands and an immediate operand
@ -1047,11 +1043,10 @@ static void expand_2i_vec(unsigned vece, uint32_t dofs, uint32_t aofs,
int64_t c, bool load_dest,
void (*fni)(unsigned, TCGv_vec, TCGv_vec, int64_t))
{
TCGv_vec t0 = tcg_temp_new_vec(type);
TCGv_vec t1 = tcg_temp_new_vec(type);
uint32_t i;
for (uint32_t i = 0; i < oprsz; i += tysz) {
TCGv_vec t0 = tcg_temp_new_vec(type);
TCGv_vec t1 = tcg_temp_new_vec(type);
for (i = 0; i < oprsz; i += tysz) {
tcg_gen_ld_vec(t0, tcg_env, aofs + i);
if (load_dest) {
tcg_gen_ld_vec(t1, tcg_env, dofs + i);
@ -1059,8 +1054,6 @@ static void expand_2i_vec(unsigned vece, uint32_t dofs, uint32_t aofs,
fni(vece, t1, t0, c);
tcg_gen_st_vec(t1, tcg_env, dofs + i);
}
tcg_temp_free_vec(t0);
tcg_temp_free_vec(t1);
}
static void expand_2s_vec(unsigned vece, uint32_t dofs, uint32_t aofs,
@ -1068,11 +1061,10 @@ static void expand_2s_vec(unsigned vece, uint32_t dofs, uint32_t aofs,
TCGv_vec c, bool scalar_first,
void (*fni)(unsigned, TCGv_vec, TCGv_vec, TCGv_vec))
{
TCGv_vec t0 = tcg_temp_new_vec(type);
TCGv_vec t1 = tcg_temp_new_vec(type);
uint32_t i;
for (uint32_t i = 0; i < oprsz; i += tysz) {
TCGv_vec t0 = tcg_temp_new_vec(type);
TCGv_vec t1 = tcg_temp_new_vec(type);
for (i = 0; i < oprsz; i += tysz) {
tcg_gen_ld_vec(t0, tcg_env, aofs + i);
if (scalar_first) {
fni(vece, t1, c, t0);
@ -1081,8 +1073,6 @@ static void expand_2s_vec(unsigned vece, uint32_t dofs, uint32_t aofs,
}
tcg_gen_st_vec(t1, tcg_env, dofs + i);
}
tcg_temp_free_vec(t0);
tcg_temp_free_vec(t1);
}
/* Expand OPSZ bytes worth of three-operand operations using host vectors. */
@ -1091,12 +1081,11 @@ static void expand_3_vec(unsigned vece, uint32_t dofs, uint32_t aofs,
uint32_t tysz, TCGType type, bool load_dest,
void (*fni)(unsigned, TCGv_vec, TCGv_vec, TCGv_vec))
{
TCGv_vec t0 = tcg_temp_new_vec(type);
TCGv_vec t1 = tcg_temp_new_vec(type);
TCGv_vec t2 = tcg_temp_new_vec(type);
uint32_t i;
for (uint32_t i = 0; i < oprsz; i += tysz) {
TCGv_vec t0 = tcg_temp_new_vec(type);
TCGv_vec t1 = tcg_temp_new_vec(type);
TCGv_vec t2 = tcg_temp_new_vec(type);
for (i = 0; i < oprsz; i += tysz) {
tcg_gen_ld_vec(t0, tcg_env, aofs + i);
tcg_gen_ld_vec(t1, tcg_env, bofs + i);
if (load_dest) {
@ -1105,9 +1094,6 @@ static void expand_3_vec(unsigned vece, uint32_t dofs, uint32_t aofs,
fni(vece, t2, t0, t1);
tcg_gen_st_vec(t2, tcg_env, dofs + i);
}
tcg_temp_free_vec(t2);
tcg_temp_free_vec(t1);
tcg_temp_free_vec(t0);
}
/*
@ -1120,12 +1106,11 @@ static void expand_3i_vec(unsigned vece, uint32_t dofs, uint32_t aofs,
void (*fni)(unsigned, TCGv_vec, TCGv_vec, TCGv_vec,
int64_t))
{
TCGv_vec t0 = tcg_temp_new_vec(type);
TCGv_vec t1 = tcg_temp_new_vec(type);
TCGv_vec t2 = tcg_temp_new_vec(type);
uint32_t i;
for (uint32_t i = 0; i < oprsz; i += tysz) {
TCGv_vec t0 = tcg_temp_new_vec(type);
TCGv_vec t1 = tcg_temp_new_vec(type);
TCGv_vec t2 = tcg_temp_new_vec(type);
for (i = 0; i < oprsz; i += tysz) {
tcg_gen_ld_vec(t0, tcg_env, aofs + i);
tcg_gen_ld_vec(t1, tcg_env, bofs + i);
if (load_dest) {
@ -1134,9 +1119,6 @@ static void expand_3i_vec(unsigned vece, uint32_t dofs, uint32_t aofs,
fni(vece, t2, t0, t1, c);
tcg_gen_st_vec(t2, tcg_env, dofs + i);
}
tcg_temp_free_vec(t0);
tcg_temp_free_vec(t1);
tcg_temp_free_vec(t2);
}
/* Expand OPSZ bytes worth of four-operand operations using host vectors. */
@ -1146,13 +1128,12 @@ static void expand_4_vec(unsigned vece, uint32_t dofs, uint32_t aofs,
void (*fni)(unsigned, TCGv_vec, TCGv_vec,
TCGv_vec, TCGv_vec))
{
TCGv_vec t0 = tcg_temp_new_vec(type);
TCGv_vec t1 = tcg_temp_new_vec(type);
TCGv_vec t2 = tcg_temp_new_vec(type);
TCGv_vec t3 = tcg_temp_new_vec(type);
uint32_t i;
for (uint32_t i = 0; i < oprsz; i += tysz) {
TCGv_vec t0 = tcg_temp_new_vec(type);
TCGv_vec t1 = tcg_temp_new_vec(type);
TCGv_vec t2 = tcg_temp_new_vec(type);
TCGv_vec t3 = tcg_temp_new_vec(type);
for (i = 0; i < oprsz; i += tysz) {
tcg_gen_ld_vec(t1, tcg_env, aofs + i);
tcg_gen_ld_vec(t2, tcg_env, bofs + i);
tcg_gen_ld_vec(t3, tcg_env, cofs + i);
@ -1162,10 +1143,6 @@ static void expand_4_vec(unsigned vece, uint32_t dofs, uint32_t aofs,
tcg_gen_st_vec(t1, tcg_env, aofs + i);
}
}
tcg_temp_free_vec(t3);
tcg_temp_free_vec(t2);
tcg_temp_free_vec(t1);
tcg_temp_free_vec(t0);
}
/*
@ -1178,23 +1155,18 @@ static void expand_4i_vec(unsigned vece, uint32_t dofs, uint32_t aofs,
void (*fni)(unsigned, TCGv_vec, TCGv_vec,
TCGv_vec, TCGv_vec, int64_t))
{
TCGv_vec t0 = tcg_temp_new_vec(type);
TCGv_vec t1 = tcg_temp_new_vec(type);
TCGv_vec t2 = tcg_temp_new_vec(type);
TCGv_vec t3 = tcg_temp_new_vec(type);
uint32_t i;
for (uint32_t i = 0; i < oprsz; i += tysz) {
TCGv_vec t0 = tcg_temp_new_vec(type);
TCGv_vec t1 = tcg_temp_new_vec(type);
TCGv_vec t2 = tcg_temp_new_vec(type);
TCGv_vec t3 = tcg_temp_new_vec(type);
for (i = 0; i < oprsz; i += tysz) {
tcg_gen_ld_vec(t1, tcg_env, aofs + i);
tcg_gen_ld_vec(t2, tcg_env, bofs + i);
tcg_gen_ld_vec(t3, tcg_env, cofs + i);
fni(vece, t0, t1, t2, t3, c);
tcg_gen_st_vec(t0, tcg_env, dofs + i);
}
tcg_temp_free_vec(t3);
tcg_temp_free_vec(t2);
tcg_temp_free_vec(t1);
tcg_temp_free_vec(t0);
}
/* Expand a vector two-operand operation. */
@ -1732,7 +1704,6 @@ void tcg_gen_gvec_dup_mem(unsigned vece, uint32_t dofs, uint32_t aofs,
TCGv_vec t_vec = tcg_temp_new_vec(type);
tcg_gen_dup_mem_vec(vece, t_vec, tcg_env, aofs);
do_dup_store(type, dofs, oprsz, maxsz, t_vec);
tcg_temp_free_vec(t_vec);
} else if (vece <= MO_32) {
TCGv_i32 in = tcg_temp_ebb_new_i32();
switch (vece) {
@ -1766,7 +1737,6 @@ void tcg_gen_gvec_dup_mem(unsigned vece, uint32_t dofs, uint32_t aofs,
for (i = (aofs == dofs) * 16; i < oprsz; i += 16) {
tcg_gen_st_vec(in, tcg_env, dofs + i);
}
tcg_temp_free_vec(in);
} else {
TCGv_i64 in0 = tcg_temp_ebb_new_i64();
TCGv_i64 in1 = tcg_temp_ebb_new_i64();
@ -1796,7 +1766,6 @@ void tcg_gen_gvec_dup_mem(unsigned vece, uint32_t dofs, uint32_t aofs,
for (i = (aofs == dofs) * 32; i < oprsz; i += 32) {
tcg_gen_st_vec(in, tcg_env, dofs + i);
}
tcg_temp_free_vec(in);
} else if (TCG_TARGET_HAS_v128) {
TCGv_vec in0 = tcg_temp_new_vec(TCG_TYPE_V128);
TCGv_vec in1 = tcg_temp_new_vec(TCG_TYPE_V128);
@ -1807,8 +1776,6 @@ void tcg_gen_gvec_dup_mem(unsigned vece, uint32_t dofs, uint32_t aofs,
tcg_gen_st_vec(in0, tcg_env, dofs + i);
tcg_gen_st_vec(in1, tcg_env, dofs + i + 16);
}
tcg_temp_free_vec(in0);
tcg_temp_free_vec(in1);
} else {
TCGv_i64 in[4];
int j;
@ -3136,15 +3103,14 @@ static void expand_2sh_vec(unsigned vece, uint32_t dofs, uint32_t aofs,
TCGv_i32 shift,
void (*fni)(unsigned, TCGv_vec, TCGv_vec, TCGv_i32))
{
TCGv_vec t0 = tcg_temp_new_vec(type);
uint32_t i;
for (uint32_t i = 0; i < oprsz; i += tysz) {
TCGv_vec t0 = tcg_temp_new_vec(type);
TCGv_vec t1 = tcg_temp_new_vec(type);
for (i = 0; i < oprsz; i += tysz) {
tcg_gen_ld_vec(t0, tcg_env, aofs + i);
fni(vece, t0, t0, shift);
tcg_gen_st_vec(t0, tcg_env, dofs + i);
fni(vece, t1, t0, shift);
tcg_gen_st_vec(t1, tcg_env, dofs + i);
}
tcg_temp_free_vec(t0);
}
static void
@ -3720,18 +3686,16 @@ static void expand_cmp_vec(unsigned vece, uint32_t dofs, uint32_t aofs,
uint32_t bofs, uint32_t oprsz, uint32_t tysz,
TCGType type, TCGCond cond)
{
TCGv_vec t0 = tcg_temp_new_vec(type);
TCGv_vec t1 = tcg_temp_new_vec(type);
uint32_t i;
for (uint32_t i = 0; i < oprsz; i += tysz) {
TCGv_vec t0 = tcg_temp_new_vec(type);
TCGv_vec t1 = tcg_temp_new_vec(type);
TCGv_vec t2 = tcg_temp_new_vec(type);
for (i = 0; i < oprsz; i += tysz) {
tcg_gen_ld_vec(t0, tcg_env, aofs + i);
tcg_gen_ld_vec(t1, tcg_env, bofs + i);
tcg_gen_cmp_vec(cond, vece, t0, t0, t1);
tcg_gen_st_vec(t0, tcg_env, dofs + i);
tcg_gen_cmp_vec(cond, vece, t2, t0, t1);
tcg_gen_st_vec(t2, tcg_env, dofs + i);
}
tcg_temp_free_vec(t1);
tcg_temp_free_vec(t0);
}
void tcg_gen_gvec_cmp(TCGCond cond, unsigned vece, uint32_t dofs,