yuzu/unicorn
1
0
Fork 0
mirror of https://github.com/yuzu-emu/unicorn.git synced 2024-12-23 18:35:27 +00:00
Code Issues Packages Projects Releases Wiki Activity

target-sparc: Allow 4-byte alignment on fp mem ops

Browse source 
The cpu is allowed to require stricter alignment on these 8- and 16-byte
operations, and the OS is required to fix up the accesses as necessary,
so the previous code was not wrong.

However, we can easily handle this misalignment for all direct 8-byte
operations and for direct 16-byte loads.

We must retain 16-byte alignment for 16-byte stores, so that we don't have
to probe for writability of a second page before performing the first of
two 8-byte stores. We also retain 8-byte alignment for no-fault loads,
since they are rare and it's not worth extending the helpers for this.

Backports commit cb21b4da6cca1bb4e3f5fefb698fb9e4d00c8f66 from qemu
This commit is contained in:
Richard Henderson 2018-03-01 08:29:07 -05:00 committed by Lioncash
parent eec264526e
commit 00fc847229
No known key found for this signature in database
GPG key ID: 4E3C3CC1031BA9C7
1 changed files with 44 additions and 45 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

75
qemu/target-sparc/translate.c
View file

@ -2617,6 +2617,7 @@ static void gen_ldf_asi(DisasContext *dc, TCGv addr,
TCGContext *tcg_ctx = dc->uc->tcg_ctx;
DisasASI da = get_asi(dc, insn, (size == 4 ? MO_TEUL : MO_TEQ));
TCGv_i32 d32;
TCGv_i64 d64;
switch (da.type) {
case GET_ASI_EXCP:
@ -2668,12 +2669,17 @@ static void gen_ldf_asi(DisasContext *dc, TCGv addr,
gen_store_fpr_F(dc, rd, d32);
break;
case 8:
tcg_gen_qemu_ld_i64(dc->uc, tcg_ctx->cpu_fpr[rd / 2], addr, da.mem_idx, da.memop);
tcg_gen_qemu_ld_i64(dc->uc, tcg_ctx->cpu_fpr[rd / 2], addr, da.mem_idx,
da.memop | MO_ALIGN_4);
break;
case 16:
tcg_gen_qemu_ld_i64(dc->uc, tcg_ctx->cpu_fpr[rd / 2], addr, da.mem_idx, da.memop);
d64 = tcg_temp_new_i64(tcg_ctx);
tcg_gen_qemu_ld_i64(dc->uc, d64, addr, da.mem_idx, da.memop | MO_ALIGN_4);
tcg_gen_addi_tl(tcg_ctx, addr, addr, 8);
tcg_gen_qemu_ld_i64(dc->uc, tcg_ctx->cpu_fpr[rd/2+1], addr, da.mem_idx, da.memop);
tcg_gen_qemu_ld_i64(dc->uc, tcg_ctx->cpu_fpr[rd/2+1], addr, da.mem_idx,
da.memop | MO_ALIGN_4);
tcg_gen_mov_i64(tcg_ctx, tcg_ctx->cpu_fpr[rd / 2], d64);
tcg_temp_free_i64(tcg_ctx, d64);
break;
default:
g_assert_not_reached();
@ -2692,25 +2698,23 @@ static void gen_ldf_asi(DisasContext *dc, TCGv addr,
but we can just use the integer asi helper for them. */
switch (size) {
case 4:
{
TCGv d64 = tcg_temp_new_i64(tcg_ctx);
d64 = tcg_temp_new_i64(tcg_ctx);
gen_helper_ld_asi(tcg_ctx, d64, tcg_ctx->cpu_env, addr, r_asi, r_mop);
d32 = gen_dest_fpr_F(dc);
tcg_gen_extrl_i64_i32(tcg_ctx, d32, d64);
tcg_temp_free_i64(tcg_ctx, d64);
gen_store_fpr_F(dc, rd, d32);
}
break;
case 8:
gen_helper_ld_asi(tcg_ctx, tcg_ctx->cpu_fpr[rd / 2],
tcg_ctx->cpu_env, addr, r_asi, r_mop);
break;
case 16:
gen_helper_ld_asi(tcg_ctx, tcg_ctx->cpu_fpr[rd / 2],
tcg_ctx->cpu_env, addr, r_asi, r_mop);
d64 = tcg_temp_new_i64(tcg_ctx);
gen_helper_ld_asi(tcg_ctx, d64, tcg_ctx->cpu_env, addr, r_asi, r_mop);
tcg_gen_addi_tl(tcg_ctx, addr, addr, 8);
gen_helper_ld_asi(tcg_ctx, tcg_ctx->cpu_fpr[rd/2+1],
tcg_ctx->cpu_env, addr, r_asi, r_mop);
tcg_gen_mov_i64(tcg_ctx, tcg_ctx->cpu_fpr[rd / 2], d64);
tcg_temp_free_i64(tcg_ctx, d64);
break;
default:
g_assert_not_reached();
@ -2778,15 +2782,15 @@ static void gen_stf_asi(DisasContext *dc, TCGv addr,
tcg_gen_qemu_st_i32(dc->uc, d32, addr, da.mem_idx, da.memop);
break;
case 8:
/* ??? Only 4-byte alignment required. However, it is legal
for the cpu to signal the alignment fault, and the OS trap
handler is required to fix it up. */
tcg_gen_qemu_st_i64(dc->uc, tcg_ctx->cpu_fpr[rd / 2], addr, da.mem_idx, da.memop);
tcg_gen_qemu_st_i64(dc->uc, tcg_ctx->cpu_fpr[rd / 2], addr, da.mem_idx,
da.memop | MO_ALIGN_4);
break;
case 16:
/* Only 4-byte alignment required. See above. Requiring
16-byte alignment here avoids having to probe the second
page before performing the first write. */
/* Only 4-byte alignment required. However, it is legal for the
cpu to signal the alignment fault, and the OS trap handler is
required to fix it up. Requiring 16-byte alignment here avoids
having to probe the second page before performing the first
write. */
tcg_gen_qemu_st_i64(dc->uc, tcg_ctx->cpu_fpr[rd / 2], addr, da.mem_idx,
da.memop | MO_ALIGN_16);
tcg_gen_addi_tl(tcg_ctx, addr, addr, 8);
@ -5604,8 +5608,6 @@ static void disas_sparc_insn(DisasContext * dc, unsigned int insn, bool hook_ins
skip_move: ;
#endif
} else if (xop >= 0x20 && xop < 0x24) {
TCGv t0;
if (gen_trap_ifnofpu(dc)) {
goto jmp_insn;
}
@ -5613,10 +5615,9 @@ static void disas_sparc_insn(DisasContext * dc, unsigned int insn, bool hook_ins
switch (xop) {
case 0x20: /* ldf, load fpreg */
gen_address_mask(dc, cpu_addr);
t0 = get_temp_tl(dc);
tcg_gen_qemu_ld32u(dc->uc, t0, cpu_addr, dc->mem_idx);
cpu_dst_32 = gen_dest_fpr_F(dc);
tcg_gen_trunc_tl_i32(tcg_ctx, cpu_dst_32, t0);
tcg_gen_qemu_ld_i32(dc->uc, cpu_dst_32, cpu_addr,
dc->mem_idx, MO_TEUL);
gen_store_fpr_F(dc, rd, cpu_dst_32);
break;
case 0x21: /* ldfsr, V9 ldxfsr */
@ -5624,26 +5625,28 @@ static void disas_sparc_insn(DisasContext * dc, unsigned int insn, bool hook_ins
gen_address_mask(dc, cpu_addr);
if (rd == 1) {
TCGv_i64 t64 = tcg_temp_new_i64(tcg_ctx);
tcg_gen_qemu_ld64(dc->uc, t64, cpu_addr, dc->mem_idx);
tcg_gen_qemu_ld_i64(dc->uc, t64, cpu_addr,
dc->mem_idx, MO_TEQ);
gen_helper_ldxfsr(tcg_ctx, tcg_ctx->cpu_fsr, tcg_ctx->cpu_env, tcg_ctx->cpu_fsr, t64);
tcg_temp_free_i64(tcg_ctx, t64);
break;
}
#endif
cpu_dst_32 = get_temp_i32(dc);
t0 = get_temp_tl(dc);
tcg_gen_qemu_ld32u(dc->uc, t0, cpu_addr, dc->mem_idx);
tcg_gen_trunc_tl_i32(tcg_ctx, cpu_dst_32, t0);
tcg_gen_qemu_ld_i32(dc->uc, cpu_dst_32, cpu_addr,
dc->mem_idx, MO_TEUL);
gen_helper_ldfsr(tcg_ctx, tcg_ctx->cpu_fsr, tcg_ctx->cpu_env, tcg_ctx->cpu_fsr, cpu_dst_32);
break;
case 0x22: /* ldqf, load quad fpreg */
CHECK_FPU_FEATURE(dc, FLOAT128);
gen_address_mask(dc, cpu_addr);
cpu_src1_64 = tcg_temp_new_i64(tcg_ctx);
tcg_gen_qemu_ld64(dc->uc, cpu_src1_64, cpu_addr, dc->mem_idx);
tcg_gen_qemu_ld_i64(dc->uc, cpu_src1_64, cpu_addr, dc->mem_idx,
MO_TEQ | MO_ALIGN_4);
tcg_gen_addi_tl(tcg_ctx, cpu_addr, cpu_addr, 8);
cpu_src2_64 = tcg_temp_new_i64(tcg_ctx);
tcg_gen_qemu_ld64(dc->uc, cpu_src2_64, cpu_addr, dc->mem_idx);
tcg_gen_qemu_ld_i64(dc->uc, cpu_src2_64, cpu_addr, dc->mem_idx,
MO_TEQ | MO_ALIGN_4);
gen_store_fpr_Q(dc, rd, cpu_src1_64, cpu_src2_64);
tcg_temp_free_i64(tcg_ctx, cpu_src1_64);
tcg_temp_free_i64(tcg_ctx, cpu_src2_64);
@ -5651,7 +5654,8 @@ static void disas_sparc_insn(DisasContext * dc, unsigned int insn, bool hook_ins
case 0x23: /* lddf, load double fpreg */
gen_address_mask(dc, cpu_addr);
cpu_dst_64 = gen_dest_fpr_D(dc, rd);
tcg_gen_qemu_ld64(dc->uc, cpu_dst_64, cpu_addr, dc->mem_idx);
tcg_gen_qemu_ld_i64(dc->uc, cpu_dst_64, cpu_addr, dc->mem_idx,
MO_TEQ | MO_ALIGN_4);
gen_store_fpr_D(dc, rd, cpu_dst_64);
break;
default:
@ -5725,13 +5729,10 @@ static void disas_sparc_insn(DisasContext * dc, unsigned int insn, bool hook_ins
save_state(dc);
switch (xop) {
case 0x24: /* stf, store fpreg */
{
TCGv t = get_temp_tl(dc);
gen_address_mask(dc, cpu_addr);
cpu_src1_32 = gen_load_fpr_F(dc, rd);
tcg_gen_ext_i32_tl(tcg_ctx, t, cpu_src1_32);
tcg_gen_qemu_st32(dc->uc, t, cpu_addr, dc->mem_idx);
}
tcg_gen_qemu_st_i32(dc->uc, cpu_src1_32, cpu_addr,
dc->mem_idx, MO_TEUL);
break;
case 0x25: /* stfsr, V9 stxfsr */
{
@ -5777,12 +5778,10 @@ static void disas_sparc_insn(DisasContext * dc, unsigned int insn, bool hook_ins
#endif
#endif
case 0x27: /* stdf, store double fpreg */
/* ??? Only 4-byte alignment required. However, it is
legal for the cpu to signal the alignment fault, and
the OS trap handler is required to fix it up. */
gen_address_mask(dc, cpu_addr);
cpu_src1_64 = gen_load_fpr_D(dc, rd);
tcg_gen_qemu_st64(dc->uc, cpu_src1_64, cpu_addr, dc->mem_idx);
tcg_gen_qemu_st_i64(dc->uc, cpu_src1_64, cpu_addr, dc->mem_idx,
MO_TEQ | MO_ALIGN_4);
break;
default:
goto illegal_insn;