tcg/aarch64: limit mul_vec size

[Qemu-devel] [PATCH] tcg/aarch64: limit mul_vec size

Alex Bennée posted 1 patch 5 years, 9 months ago

Download mbox

Patches applied successfully (tree, apply log)
git fetch https://github.com/patchew-project/qemu tags/patchew/20180719154248.29669-1-alex.bennee@linaro.org

Test checkpatch passed

Test docker-mingw@fedora passed

Test docker-quick@centos7 passed

tcg/aarch64/tcg-target.inc.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)

Expand all Fold all

[Qemu-devel] [PATCH] tcg/aarch64: limit mul_vec size

Posted by Alex Bennée 5 years, 9 months ago

In AdvSIMD we can only do 32x32 integer multiples although SVE is
capable of larger 64 bit multiples. As a result we can end up
generating invalid opcodes. Fix this by only reprting we can emit mul
vector ops if the size is small enough.

Fixes a crash on:

  sve-all-short-v8.3+sve@vq3/insn_mul_z_zi___INC.risu.bin

When running on AArch64 hardware.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
---
 tcg/aarch64/tcg-target.inc.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/tcg/aarch64/tcg-target.inc.c b/tcg/aarch64/tcg-target.inc.c
index 4562d36d1b..3d08bdd2e3 100644
--- a/tcg/aarch64/tcg-target.inc.c
+++ b/tcg/aarch64/tcg-target.inc.c
@@ -718,6 +718,7 @@ static void tcg_out_insn_3614(TCGContext *s, AArch64Insn insn, bool q,
 static void tcg_out_insn_3616(TCGContext *s, AArch64Insn insn, bool q,
                               unsigned size, TCGReg rd, TCGReg rn, TCGReg rm)
 {
+    tcg_debug_assert(!(insn == I3616_MUL && size == 3));
     tcg_out32(s, insn | q << 30 | (size << 22) | (rm & 0x1f) << 16
               | (rn & 0x1f) << 5 | (rd & 0x1f));
 }
@@ -2219,7 +2220,6 @@ int tcg_can_emit_vec_op(TCGOpcode opc, TCGType type, unsigned vece)
     switch (opc) {
     case INDEX_op_add_vec:
     case INDEX_op_sub_vec:
-    case INDEX_op_mul_vec:
     case INDEX_op_and_vec:
     case INDEX_op_or_vec:
     case INDEX_op_xor_vec:
@@ -2232,6 +2232,8 @@ int tcg_can_emit_vec_op(TCGOpcode opc, TCGType type, unsigned vece)
     case INDEX_op_shri_vec:
     case INDEX_op_sari_vec:
         return 1;
+    case INDEX_op_mul_vec:
+        return vece < MO_64 ? 1 : 0;
 
     default:
         return 0;
-- 
2.17.1

Re: [Qemu-devel] [PATCH] tcg/aarch64: limit mul_vec size

Posted by Richard Henderson 5 years, 9 months ago

On 07/19/2018 08:42 AM, Alex Bennée wrote:
> In AdvSIMD we can only do 32x32 integer multiples although SVE is
> capable of larger 64 bit multiples. As a result we can end up
> generating invalid opcodes. Fix this by only reprting we can emit mul
> vector ops if the size is small enough.
> 
> Fixes a crash on:
> 
>   sve-all-short-v8.3+sve@vq3/insn_mul_z_zi___INC.risu.bin
> 
> When running on AArch64 hardware.
> 
> Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
> ---
>  tcg/aarch64/tcg-target.inc.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)

Queued.  Thanks,


r~