yuzu/unicorn
1
0
Fork 0
mirror of https://github.com/yuzu-emu/unicorn.git synced 2024-12-23 19:15:34 +00:00
Code Issues Packages Projects Releases Wiki Activity

target/arm: Remove VLDR/VSTR/VLDM/VSTM use of cpu_F0s and cpu_F0d

Browse source 
Expand out the sequences in the new decoder VLDR/VSTR/VLDM/VSTM trans
functions which perform the memory accesses by going via the TCG
globals cpu_F0s and cpu_F0d, to use local TCG temps instead.

Backports commit 3993d0407dff7233e42f2251db971e126a0497e9 from qemu
This commit is contained in:
Peter Maydell 2019-06-13 17:31:25 -04:00 committed by Lioncash
parent ff7042567e
commit 93fe4cbe9e
No known key found for this signature in database
GPG key ID: 4E3C3CC1031BA9C7
2 changed files with 28 additions and 36 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

46
qemu/target/arm/translate-vfp.inc.c
View file

@ -871,7 +871,7 @@ static bool trans_VLDR_VSTR_sp(DisasContext *s, arg_VLDR_VSTR_sp *a)
{ {
TCGContext *tcg_ctx = s->uc->tcg_ctx; TCGContext *tcg_ctx = s->uc->tcg_ctx;
uint32_t offset; uint32_t offset;
TCGv_i32 addr; TCGv_i32 addr, tmp;
if (!vfp_access_check(s)) { if (!vfp_access_check(s)) {
return true; return true;
@ -890,13 +890,15 @@ static bool trans_VLDR_VSTR_sp(DisasContext *s, arg_VLDR_VSTR_sp *a)
addr = load_reg(s, a->rn); addr = load_reg(s, a->rn);
} }
tcg_gen_addi_i32(tcg_ctx, addr, addr, offset); tcg_gen_addi_i32(tcg_ctx, addr, addr, offset);
tmp = tcg_temp_new_i32(tcg_ctx);
if (a->l) { if (a->l) {
gen_vfp_ld(s, false, addr); gen_aa32_ld32u(s, tmp, addr, get_mem_index(s));
gen_mov_vreg_F0(s, false, a->vd); neon_store_reg32(s, tmp, a->vd);
} else { } else {
gen_mov_F0_vreg(s, false, a->vd); neon_load_reg32(s, tmp, a->vd);
gen_vfp_st(s, false, addr); gen_aa32_st32(s, tmp, addr, get_mem_index(s));
} }
tcg_temp_free_i32(tcg_ctx, tmp);
tcg_temp_free_i32(tcg_ctx, addr); tcg_temp_free_i32(tcg_ctx, addr);
return true; return true;
@ -907,6 +909,7 @@ static bool trans_VLDR_VSTR_dp(DisasContext *s, arg_VLDR_VSTR_sp *a)
TCGContext *tcg_ctx = s->uc->tcg_ctx; TCGContext *tcg_ctx = s->uc->tcg_ctx;
uint32_t offset; uint32_t offset;
TCGv_i32 addr; TCGv_i32 addr;
TCGv_i64 tmp;
/* UNDEF accesses to D16-D31 if they don't exist */ /* UNDEF accesses to D16-D31 if they don't exist */
if (!dc_isar_feature(aa32_fp_d32, s) && (a->vd & 0x10)) { if (!dc_isar_feature(aa32_fp_d32, s) && (a->vd & 0x10)) {
@ -930,13 +933,15 @@ static bool trans_VLDR_VSTR_dp(DisasContext *s, arg_VLDR_VSTR_sp *a)
addr = load_reg(s, a->rn); addr = load_reg(s, a->rn);
} }
tcg_gen_addi_i32(tcg_ctx, addr, addr, offset); tcg_gen_addi_i32(tcg_ctx, addr, addr, offset);
tmp = tcg_temp_new_i64(tcg_ctx);
if (a->l) { if (a->l) {
gen_vfp_ld(s, true, addr); gen_aa32_ld64(s, tmp, addr, get_mem_index(s));
gen_mov_vreg_F0(s, true, a->vd); neon_store_reg64(s, tmp, a->vd);
} else { } else {
gen_mov_F0_vreg(s, true, a->vd); neon_load_reg64(s, tmp, a->vd);
gen_vfp_st(s, true, addr); gen_aa32_st64(s, tmp, addr, get_mem_index(s));
} }
tcg_temp_free_i64(tcg_ctx, tmp);
tcg_temp_free_i32(tcg_ctx, addr); tcg_temp_free_i32(tcg_ctx, addr);
return true; return true;
@ -946,7 +951,7 @@ static bool trans_VLDM_VSTM_sp(DisasContext *s, arg_VLDM_VSTM_sp *a)
{ {
TCGContext *tcg_ctx = s->uc->tcg_ctx; TCGContext *tcg_ctx = s->uc->tcg_ctx;
uint32_t offset; uint32_t offset;
TCGv_i32 addr; TCGv_i32 addr, tmp;
int i, n; int i, n;
n = a->imm; n = a->imm;
@ -992,18 +997,20 @@ static bool trans_VLDM_VSTM_sp(DisasContext *s, arg_VLDM_VSTM_sp *a)
} }
offset = 4; offset = 4;
tmp = tcg_temp_new_i32(tcg_ctx);
for (i = 0; i < n; i++) { for (i = 0; i < n; i++) {
if (a->l) { if (a->l) {
/* load */ /* load */
gen_vfp_ld(s, false, addr); gen_aa32_ld32u(s, tmp, addr, get_mem_index(s));
gen_mov_vreg_F0(s, false, a->vd + i); neon_store_reg32(s, tmp, a->vd + i);
} else { } else {
/* store */ /* store */
gen_mov_F0_vreg(s, false, a->vd + i); neon_load_reg32(s, tmp, a->vd + i);
gen_vfp_st(s, false, addr); gen_aa32_st32(s, tmp, addr, get_mem_index(s));
} }
tcg_gen_addi_i32(tcg_ctx, addr, addr, offset); tcg_gen_addi_i32(tcg_ctx, addr, addr, offset);
} }
tcg_temp_free_i32(tcg_ctx, tmp);
if (a->w) { if (a->w) {
/* writeback */ /* writeback */
if (a->p) { if (a->p) {
@ -1023,6 +1030,7 @@ static bool trans_VLDM_VSTM_dp(DisasContext *s, arg_VLDM_VSTM_dp *a)
TCGContext *tcg_ctx = s->uc->tcg_ctx; TCGContext *tcg_ctx = s->uc->tcg_ctx;
uint32_t offset; uint32_t offset;
TCGv_i32 addr; TCGv_i32 addr;
TCGv_i64 tmp;
int i, n; int i, n;
n = a->imm >> 1; n = a->imm >> 1;
@ -1073,18 +1081,20 @@ static bool trans_VLDM_VSTM_dp(DisasContext *s, arg_VLDM_VSTM_dp *a)
} }
offset = 8; offset = 8;
tmp = tcg_temp_new_i64(tcg_ctx);
for (i = 0; i < n; i++) { for (i = 0; i < n; i++) {
if (a->l) { if (a->l) {
/* load */ /* load */
gen_vfp_ld(s, true, addr); gen_aa32_ld64(s, tmp, addr, get_mem_index(s));
gen_mov_vreg_F0(s, true, a->vd + i); neon_store_reg64(s, tmp, a->vd + i);
} else { } else {
/* store */ /* store */
gen_mov_F0_vreg(s, true, a->vd + i); neon_load_reg64(s, tmp, a->vd + i);
gen_vfp_st(s, true, addr); gen_aa32_st64(s, tmp, addr, get_mem_index(s));
} }
tcg_gen_addi_i32(tcg_ctx, addr, addr, offset); tcg_gen_addi_i32(tcg_ctx, addr, addr, offset);
} }
tcg_temp_free_i64(tcg_ctx, tmp);
if (a->w) { if (a->w) {
/* writeback */ /* writeback */
if (a->p) { if (a->p) {

18
qemu/target/arm/translate.c
View file

@ -1589,24 +1589,6 @@ VFP_GEN_FIX(uhto, )
VFP_GEN_FIX(ulto, ) VFP_GEN_FIX(ulto, )
#undef VFP_GEN_FIX #undef VFP_GEN_FIX
static inline void gen_vfp_ld(DisasContext *s, int dp, TCGv_i32 addr)
{
if (dp) {
gen_aa32_ld64(s, s->F0d, addr, get_mem_index(s));
} else {
gen_aa32_ld32u(s, s->F0s, addr, get_mem_index(s));
}
}
static inline void gen_vfp_st(DisasContext *s, int dp, TCGv_i32 addr)
{
if (dp) {
gen_aa32_st64(s, s->F0d, addr, get_mem_index(s));
} else {
gen_aa32_st32(s, s->F0s, addr, get_mem_index(s));
}
}
static inline long vfp_reg_offset(bool dp, unsigned reg) static inline long vfp_reg_offset(bool dp, unsigned reg)
{ {
if (dp) { if (dp) {