motorhead/qemu
1
0
Fork 0
mirror of https://github.com/Motorhead1991/qemu.git synced 2025-08-02 15:23:53 -06:00
Code Issues Projects Releases Packages Wiki Activity Actions

tcg: Split out tcg_gen_gvec_3_var

Browse source 
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
This commit is contained in:
Richard Henderson 2024-03-06 13:24:31 -10:00
parent 7a74c13468
commit 465b21ffbe
2 changed files with 95 additions and 40 deletions
Download patch file Download diff file Expand all files Collapse all files

33
include/tcg/tcg-op-gvec-common.h
View file

@ -241,13 +241,42 @@ void tcg_gen_gvec_2i(uint32_t dofs, uint32_t aofs, uint32_t oprsz,
void tcg_gen_gvec_2s(uint32_t dofs, uint32_t aofs, uint32_t oprsz,
uint32_t maxsz, TCGv_i64 c, const GVecGen2s *op);
/*
* Expand (dbase+dofs) = op(abase+aofs, bbase+bofs),
* length @oprsz, clearing to @maxsz.
*/
void tcg_gen_gvec_3_var(TCGv_ptr dbase, uint32_t dofs,
TCGv_ptr abase, uint32_t aofs,
TCGv_ptr bbase, uint32_t bofs,
uint32_t oprsz, uint32_t maxsz, const GVecGen3 *op);
/* Similarly, expand (env+dofs) = op(env+aofs, env+bofs). */
void tcg_gen_gvec_3(uint32_t dofs, uint32_t aofs, uint32_t bofs,
uint32_t oprsz, uint32_t maxsz, const GVecGen3 *);
uint32_t oprsz, uint32_t maxsz, const GVecGen3 *op);
/*
* Depending on op->load_dest and op->write_aofs, expand
* (env+dofs) = op(env+aofs, env+bofs, c)
* or
* (env+dofs) = op(env+dofs, env+aofs, env+bofs, c)
* or
* (env+dofs), (env+aofs) = op(env+aofs, env+bofs, c)
* or
* (env+dofs), (env+aofs) = op(env+dofs, env+aofs, env+bofs, c)
*/
void tcg_gen_gvec_3i(uint32_t dofs, uint32_t aofs, uint32_t bofs,
uint32_t oprsz, uint32_t maxsz, int64_t c,
const GVecGen3i *);
const GVecGen3i *op);
/*
* Depending on op->write_aofs, expand
* (env+dofs) = op(env+aofs, env+bofs, env+cofs)
* or
* (env+dofs), (env+aofs) = op(env+aofs, env+bofs, env+cofs)
*/
void tcg_gen_gvec_4(uint32_t dofs, uint32_t aofs, uint32_t bofs, uint32_t cofs,
uint32_t oprsz, uint32_t maxsz, const GVecGen4 *);
/* Expand (env+dofs) = op(env+aofs, env+bofs, env+cofs, c). */
void tcg_gen_gvec_4i(uint32_t dofs, uint32_t aofs, uint32_t bofs, uint32_t cofs,
uint32_t oprsz, uint32_t maxsz, int64_t c,
const GVecGen4i *);

102
tcg/tcg-op-gvec.c
View file

@ -182,9 +182,11 @@ void tcg_gen_gvec_2i_ool(uint32_t dofs, uint32_t aofs, TCGv_i64 c,
}
/* Generate a call to a gvec-style helper with three vector operands. */
void tcg_gen_gvec_3_ool(uint32_t dofs, uint32_t aofs, uint32_t bofs,
uint32_t oprsz, uint32_t maxsz, int32_t data,
gen_helper_gvec_3 *fn)
static void expand_3_ool(TCGv_ptr dbase, uint32_t dofs,
TCGv_ptr abase, uint32_t aofs,
TCGv_ptr bbase, uint32_t bofs,
uint32_t oprsz, uint32_t maxsz,
int32_t data, gen_helper_gvec_3 *fn)
{
TCGv_ptr a0, a1, a2;
TCGv_i32 desc = tcg_constant_i32(simd_desc(oprsz, maxsz, data));
@ -193,9 +195,9 @@ void tcg_gen_gvec_3_ool(uint32_t dofs, uint32_t aofs, uint32_t bofs,
a1 = tcg_temp_ebb_new_ptr();
a2 = tcg_temp_ebb_new_ptr();
tcg_gen_addi_ptr(a0, tcg_env, dofs);
tcg_gen_addi_ptr(a1, tcg_env, aofs);
tcg_gen_addi_ptr(a2, tcg_env, bofs);
tcg_gen_addi_ptr(a0, dbase, dofs);
tcg_gen_addi_ptr(a1, abase, aofs);
tcg_gen_addi_ptr(a2, bbase, bofs);
fn(a0, a1, a2, desc);
@ -204,6 +206,14 @@ void tcg_gen_gvec_3_ool(uint32_t dofs, uint32_t aofs, uint32_t bofs,
tcg_temp_free_ptr(a2);
}
void tcg_gen_gvec_3_ool(uint32_t dofs, uint32_t aofs, uint32_t bofs,
uint32_t oprsz, uint32_t maxsz, int32_t data,
gen_helper_gvec_3 *fn)
{
expand_3_ool(tcg_env, dofs, tcg_env, aofs, tcg_env, bofs,
oprsz, maxsz, data, fn);
}
/* Generate a call to a gvec-style helper with four vector operands. */
void tcg_gen_gvec_4_ool(uint32_t dofs, uint32_t aofs, uint32_t bofs,
uint32_t cofs, uint32_t oprsz, uint32_t maxsz,
@ -794,8 +804,10 @@ static void expand_2s_i32(uint32_t dofs, uint32_t aofs, uint32_t oprsz,
}
/* Expand OPSZ bytes worth of three-operand operations using i32 elements. */
static void expand_3_i32(uint32_t dofs, uint32_t aofs,
uint32_t bofs, uint32_t oprsz, bool load_dest,
static void expand_3_i32(TCGv_ptr dbase, uint32_t dofs,
TCGv_ptr abase, uint32_t aofs,
TCGv_ptr bbase, uint32_t bofs,
uint32_t oprsz, bool load_dest,
void (*fni)(TCGv_i32, TCGv_i32, TCGv_i32))
{
TCGv_i32 t0 = tcg_temp_new_i32();
@ -804,13 +816,13 @@ static void expand_3_i32(uint32_t dofs, uint32_t aofs,
uint32_t i;
for (i = 0; i < oprsz; i += 4) {
tcg_gen_ld_i32(t0, tcg_env, aofs + i);
tcg_gen_ld_i32(t1, tcg_env, bofs + i);
tcg_gen_ld_i32(t0, abase, aofs + i);
tcg_gen_ld_i32(t1, bbase, bofs + i);
if (load_dest) {
tcg_gen_ld_i32(t2, tcg_env, dofs + i);
tcg_gen_ld_i32(t2, dbase, dofs + i);
}
fni(t2, t0, t1);
tcg_gen_st_i32(t2, tcg_env, dofs + i);
tcg_gen_st_i32(t2, dbase, dofs + i);
}
tcg_temp_free_i32(t2);
tcg_temp_free_i32(t1);
@ -958,8 +970,10 @@ static void expand_2s_i64(uint32_t dofs, uint32_t aofs, uint32_t oprsz,
}
/* Expand OPSZ bytes worth of three-operand operations using i64 elements. */
static void expand_3_i64(uint32_t dofs, uint32_t aofs,
uint32_t bofs, uint32_t oprsz, bool load_dest,
static void expand_3_i64(TCGv_ptr dbase, uint32_t dofs,
TCGv_ptr abase, uint32_t aofs,
TCGv_ptr bbase, uint32_t bofs,
uint32_t oprsz, bool load_dest,
void (*fni)(TCGv_i64, TCGv_i64, TCGv_i64))
{
TCGv_i64 t0 = tcg_temp_new_i64();
@ -968,13 +982,13 @@ static void expand_3_i64(uint32_t dofs, uint32_t aofs,
uint32_t i;
for (i = 0; i < oprsz; i += 8) {
tcg_gen_ld_i64(t0, tcg_env, aofs + i);
tcg_gen_ld_i64(t1, tcg_env, bofs + i);
tcg_gen_ld_i64(t0, abase, aofs + i);
tcg_gen_ld_i64(t1, bbase, bofs + i);
if (load_dest) {
tcg_gen_ld_i64(t2, tcg_env, dofs + i);
tcg_gen_ld_i64(t2, dbase, dofs + i);
}
fni(t2, t0, t1);
tcg_gen_st_i64(t2, tcg_env, dofs + i);
tcg_gen_st_i64(t2, dbase, dofs + i);
}
tcg_temp_free_i64(t2);
tcg_temp_free_i64(t1);
@ -1119,8 +1133,9 @@ static void expand_2s_vec(unsigned vece, uint32_t dofs, uint32_t aofs,
}
/* Expand OPSZ bytes worth of three-operand operations using host vectors. */
static void expand_3_vec(unsigned vece, uint32_t dofs, uint32_t aofs,
uint32_t bofs, uint32_t oprsz,
static void expand_3_vec(unsigned vece, TCGv_ptr dbase, uint32_t dofs,
TCGv_ptr abase, uint32_t aofs,
TCGv_ptr bbase, uint32_t bofs, uint32_t oprsz,
uint32_t tysz, TCGType type, bool load_dest,
void (*fni)(unsigned, TCGv_vec, TCGv_vec, TCGv_vec))
{
@ -1129,13 +1144,13 @@ static void expand_3_vec(unsigned vece, uint32_t dofs, uint32_t aofs,
TCGv_vec t1 = tcg_temp_new_vec(type);
TCGv_vec t2 = tcg_temp_new_vec(type);
tcg_gen_ld_vec(t0, tcg_env, aofs + i);
tcg_gen_ld_vec(t1, tcg_env, bofs + i);
tcg_gen_ld_vec(t0, abase, aofs + i);
tcg_gen_ld_vec(t1, bbase, bofs + i);
if (load_dest) {
tcg_gen_ld_vec(t2, tcg_env, dofs + i);
tcg_gen_ld_vec(t2, dbase, dofs + i);
}
fni(vece, t2, t0, t1);
tcg_gen_st_vec(t2, tcg_env, dofs + i);
tcg_gen_st_vec(t2, dbase, dofs + i);
}
}
@ -1437,8 +1452,10 @@ void tcg_gen_gvec_2s(uint32_t dofs, uint32_t aofs, uint32_t oprsz,
}
/* Expand a vector three-operand operation. */
void tcg_gen_gvec_3(uint32_t dofs, uint32_t aofs, uint32_t bofs,
uint32_t oprsz, uint32_t maxsz, const GVecGen3 *g)
void tcg_gen_gvec_3_var(TCGv_ptr dbase, uint32_t dofs,
TCGv_ptr abase, uint32_t aofs,
TCGv_ptr bbase, uint32_t bofs,
uint32_t oprsz, uint32_t maxsz, const GVecGen3 *g)
{
const TCGOpcode *this_list = g->opt_opc ? : vecop_list_empty;
const TCGOpcode *hold_list = tcg_swap_vecop_list(this_list);
@ -1446,7 +1463,7 @@ void tcg_gen_gvec_3(uint32_t dofs, uint32_t aofs, uint32_t bofs,
uint32_t some;
check_size_align(oprsz, maxsz, dofs | aofs | bofs);
check_overlap_3(tcg_env, dofs, tcg_env, aofs, tcg_env, bofs, maxsz);
check_overlap_3(dbase, dofs, abase, aofs, bbase, bofs, maxsz);
type = 0;
if (g->fniv) {
@ -1459,8 +1476,8 @@ void tcg_gen_gvec_3(uint32_t dofs, uint32_t aofs, uint32_t bofs,
* that e.g. size == 80 would be expanded with 2x32 + 1x16.
*/
some = QEMU_ALIGN_DOWN(oprsz, 32);
expand_3_vec(g->vece, dofs, aofs, bofs, some, 32, TCG_TYPE_V256,
g->load_dest, g->fniv);
expand_3_vec(g->vece, dbase, dofs, abase, aofs, bbase, bofs,
some, 32, TCG_TYPE_V256, g->load_dest, g->fniv);
if (some == oprsz) {
break;
}
@ -1471,23 +1488,25 @@ void tcg_gen_gvec_3(uint32_t dofs, uint32_t aofs, uint32_t bofs,
maxsz -= some;
/* fallthru */
case TCG_TYPE_V128:
expand_3_vec(g->vece, dofs, aofs, bofs, oprsz, 16, TCG_TYPE_V128,
g->load_dest, g->fniv);
expand_3_vec(g->vece, dbase, dofs, abase, aofs, bbase, bofs,
oprsz, 16, TCG_TYPE_V128, g->load_dest, g->fniv);
break;
case TCG_TYPE_V64:
expand_3_vec(g->vece, dofs, aofs, bofs, oprsz, 8, TCG_TYPE_V64,
g->load_dest, g->fniv);
expand_3_vec(g->vece, dbase, dofs, abase, aofs, bbase, bofs,
oprsz, 8, TCG_TYPE_V64, g->load_dest, g->fniv);
break;
case 0:
if (g->fni8 && check_size_impl(oprsz, 8)) {
expand_3_i64(dofs, aofs, bofs, oprsz, g->load_dest, g->fni8);
expand_3_i64(dbase, dofs, abase, aofs, bbase, bofs,
oprsz, g->load_dest, g->fni8);
} else if (g->fni4 && check_size_impl(oprsz, 4)) {
expand_3_i32(dofs, aofs, bofs, oprsz, g->load_dest, g->fni4);
expand_3_i32(dbase, dofs, abase, aofs, bbase, bofs,
oprsz, g->load_dest, g->fni4);
} else {
assert(g->fno != NULL);
tcg_gen_gvec_3_ool(dofs, aofs, bofs, oprsz,
maxsz, g->data, g->fno);
expand_3_ool(dbase, dofs, abase, aofs, bbase, bofs,
oprsz, maxsz, g->data, g->fno);
oprsz = maxsz;
}
break;
@ -1498,10 +1517,17 @@ void tcg_gen_gvec_3(uint32_t dofs, uint32_t aofs, uint32_t bofs,
tcg_swap_vecop_list(hold_list);
if (oprsz < maxsz) {
expand_clr(tcg_env, dofs + oprsz, maxsz - oprsz);
expand_clr(dbase, dofs + oprsz, maxsz - oprsz);
}
}
void tcg_gen_gvec_3(uint32_t dofs, uint32_t aofs, uint32_t bofs,
uint32_t oprsz, uint32_t maxsz, const GVecGen3 *g)
{
tcg_gen_gvec_3_var(tcg_env, dofs, tcg_env, aofs, tcg_env, bofs,
oprsz, maxsz, g);
}
/* Expand a vector operation with three vectors and an immediate. */
void tcg_gen_gvec_3i(uint32_t dofs, uint32_t aofs, uint32_t bofs,
uint32_t oprsz, uint32_t maxsz, int64_t c,