Signed-off-by: Richard Henderson <richard.hender...@linaro.org>
---
include/tcg/tcg-op-gvec-common.h | 2 ++
tcg/tcg-op-gvec.c | 30 ++++++++++++++++++++++--------
2 files changed, 24 insertions(+), 8 deletions(-)
diff --git a/include/tcg/tcg-op-gvec-common.h b/include/tcg/tcg-op-gvec-common.h
index 4db8a58c14..65553f5f97 100644
--- a/include/tcg/tcg-op-gvec-common.h
+++ b/include/tcg/tcg-op-gvec-common.h
@@ -183,6 +183,8 @@ typedef struct {
bool prefer_i64;
/* Load dest as a 3rd source operand. */
bool load_dest;
+ /* Write aofs as a 2nd dest operand. */
+ bool write_aofs;
} GVecGen3i;
typedef struct {
diff --git a/tcg/tcg-op-gvec.c b/tcg/tcg-op-gvec.c
index bb88943f79..0308732d9b 100644
--- a/tcg/tcg-op-gvec.c
+++ b/tcg/tcg-op-gvec.c
@@ -785,7 +785,8 @@ static void expand_3_i32(uint32_t dofs, uint32_t aofs,
}
static void expand_3i_i32(uint32_t dofs, uint32_t aofs, uint32_t bofs,
- uint32_t oprsz, int32_t c, bool load_dest,
+ uint32_t oprsz, int32_t c,
+ bool load_dest, bool write_aofs,
void (*fni)(TCGv_i32, TCGv_i32, TCGv_i32, int32_t))
{
TCGv_i32 t0 = tcg_temp_new_i32();
@@ -801,6 +802,9 @@ static void expand_3i_i32(uint32_t dofs, uint32_t aofs,
uint32_t bofs,
}
fni(t2, t0, t1, c);
tcg_gen_st_i32(t2, tcg_env, dofs + i);
+ if (write_aofs) {
+ tcg_gen_st_i32(t0, tcg_env, aofs + i);
+ }
}
tcg_temp_free_i32(t0);
tcg_temp_free_i32(t1);
@@ -944,7 +948,8 @@ static void expand_3_i64(uint32_t dofs, uint32_t aofs,
}
static void expand_3i_i64(uint32_t dofs, uint32_t aofs, uint32_t bofs,
- uint32_t oprsz, int64_t c, bool load_dest,
+ uint32_t oprsz, int64_t c,
+ bool load_dest, bool write_aofs,
void (*fni)(TCGv_i64, TCGv_i64, TCGv_i64, int64_t))
{
TCGv_i64 t0 = tcg_temp_new_i64();
@@ -960,6 +965,9 @@ static void expand_3i_i64(uint32_t dofs, uint32_t aofs,
uint32_t bofs,
}
fni(t2, t0, t1, c);
tcg_gen_st_i64(t2, tcg_env, dofs + i);
+ if (write_aofs) {
+ tcg_gen_st_i64(t0, tcg_env, aofs + i);
+ }
}
tcg_temp_free_i64(t0);
tcg_temp_free_i64(t1);
@@ -1102,7 +1110,8 @@ static void expand_3_vec(unsigned vece, uint32_t dofs,
uint32_t aofs,
*/
static void expand_3i_vec(unsigned vece, uint32_t dofs, uint32_t aofs,
uint32_t bofs, uint32_t oprsz, uint32_t tysz,
- TCGType type, int64_t c, bool load_dest,
+ TCGType type, int64_t c,
+ bool load_dest, bool write_aofs,
void (*fni)(unsigned, TCGv_vec, TCGv_vec, TCGv_vec,
int64_t))
{
@@ -1118,6 +1127,9 @@ 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);
+ if (write_aofs) {
+ tcg_gen_st_vec(t0, tcg_env, aofs + i);
+ }
}
}
@@ -1471,7 +1483,7 @@ void tcg_gen_gvec_3i(uint32_t dofs, uint32_t aofs,
uint32_t bofs,
*/
some = QEMU_ALIGN_DOWN(oprsz, 32);
expand_3i_vec(g->vece, dofs, aofs, bofs, some, 32, TCG_TYPE_V256,
- c, g->load_dest, g->fniv);
+ c, g->load_dest, g->write_aofs, g->fniv);
if (some == oprsz) {
break;
}
@@ -1483,18 +1495,20 @@ void tcg_gen_gvec_3i(uint32_t dofs, uint32_t aofs,
uint32_t bofs,
/* fallthru */
case TCG_TYPE_V128:
expand_3i_vec(g->vece, dofs, aofs, bofs, oprsz, 16, TCG_TYPE_V128,
- c, g->load_dest, g->fniv);
+ c, g->load_dest, g->write_aofs, g->fniv);
break;
case TCG_TYPE_V64:
expand_3i_vec(g->vece, dofs, aofs, bofs, oprsz, 8, TCG_TYPE_V64,
- c, g->load_dest, g->fniv);
+ c, g->load_dest, g->write_aofs, g->fniv);
break;
case 0:
if (g->fni8 && check_size_impl(oprsz, 8)) {
- expand_3i_i64(dofs, aofs, bofs, oprsz, c, g->load_dest, g->fni8);
+ expand_3i_i64(dofs, aofs, bofs, oprsz, c,
+ g->load_dest, g->write_aofs, g->fni8);
} else if (g->fni4 && check_size_impl(oprsz, 4)) {
- expand_3i_i32(dofs, aofs, bofs, oprsz, c, g->load_dest, g->fni4);
+ expand_3i_i32(dofs, aofs, bofs, oprsz, c,
+ g->load_dest, g->write_aofs, g->fni4);
} else {
assert(g->fno != NULL);
tcg_gen_gvec_3_ool(dofs, aofs, bofs, oprsz, maxsz, c, g->fno);
--
2.34.1
