* [PATCH] Fix a vbmi2 ICE (PR target/83604)
@ 2017-12-28 9:10 Jakub Jelinek
2018-01-05 16:06 ` Kirill Yukhin
0 siblings, 1 reply; 2+ messages in thread
From: Jakub Jelinek @ 2017-12-28 9:10 UTC (permalink / raw)
To: Uros Bizjak, Kirill Yukhin; +Cc: gcc-patches, Julia Koval, Sebastian Peryt
Hi!
These insns don't really need AVX512BW in any way themselves, only their
masked variants might need it for reloading of the mask register, but that
should be covered in builtins.def, doesn't need duplication in sse.md.
For non-masked it causes ICEs, because the builtins properly aren't guarded
with AVX512BW, but the insns incorrectly require that.
Fixed thusly, bootstrapped/regtested on x86_64-linux and i686-linux, ok for
trunk?
2017-12-28 Jakub Jelinek <jakub@redhat.com>
PR target/83604
* config/i386/sse.md (VI248_VLBW): Rename to ...
(VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
(vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
mode iterator instead of VI248_VLBW.
* gcc.target/i386/pr83604.c: New test.
--- gcc/config/i386/sse.md.jj 2017-12-27 19:19:58.081660733 +0100
+++ gcc/config/i386/sse.md 2017-12-27 18:43:32.200347561 +0100
@@ -448,8 +448,8 @@ (define_mode_iterator VI124_AVX2
(define_mode_iterator VI2_AVX2_AVX512BW
[(V32HI "TARGET_AVX512BW") (V16HI "TARGET_AVX2") V8HI])
-(define_mode_iterator VI248_VLBW
- [(V32HI "TARGET_AVX512BW") V16SI V8DI
+(define_mode_iterator VI248_AVX512VL
+ [V32HI V16SI V8DI
(V16HI "TARGET_AVX512VL") (V8SI "TARGET_AVX512VL")
(V4DI "TARGET_AVX512VL") (V8HI "TARGET_AVX512VL")
(V4SI "TARGET_AVX512VL") (V2DI "TARGET_AVX512VL")])
@@ -20134,10 +20134,10 @@ (define_insn "vgf2p8mulb_<mode><mask_nam
(set_attr "mode" "<sseinsnmode>")])
(define_insn "vpshrd_<mode><mask_name>"
- [(set (match_operand:VI248_VLBW 0 "register_operand" "=v")
- (unspec:VI248_VLBW
- [(match_operand:VI248_VLBW 1 "register_operand" "v")
- (match_operand:VI248_VLBW 2 "nonimmediate_operand" "vm")
+ [(set (match_operand:VI248_AVX512VL 0 "register_operand" "=v")
+ (unspec:VI248_AVX512VL
+ [(match_operand:VI248_AVX512VL 1 "register_operand" "v")
+ (match_operand:VI248_AVX512VL 2 "nonimmediate_operand" "vm")
(match_operand:SI 3 "const_0_to_255_operand" "n")]
UNSPEC_VPSHRD))]
"TARGET_AVX512VBMI2"
@@ -20145,10 +20145,10 @@ (define_insn "vpshrd_<mode><mask_name>"
[(set_attr ("prefix") ("evex"))])
(define_insn "vpshld_<mode><mask_name>"
- [(set (match_operand:VI248_VLBW 0 "register_operand" "=v")
- (unspec:VI248_VLBW
- [(match_operand:VI248_VLBW 1 "register_operand" "v")
- (match_operand:VI248_VLBW 2 "nonimmediate_operand" "vm")
+ [(set (match_operand:VI248_AVX512VL 0 "register_operand" "=v")
+ (unspec:VI248_AVX512VL
+ [(match_operand:VI248_AVX512VL 1 "register_operand" "v")
+ (match_operand:VI248_AVX512VL 2 "nonimmediate_operand" "vm")
(match_operand:SI 3 "const_0_to_255_operand" "n")]
UNSPEC_VPSHLD))]
"TARGET_AVX512VBMI2"
@@ -20156,11 +20156,11 @@ (define_insn "vpshld_<mode><mask_name>"
[(set_attr ("prefix") ("evex"))])
(define_insn "vpshrdv_<mode>"
- [(set (match_operand:VI248_VLBW 0 "register_operand" "=v")
- (unspec:VI248_VLBW
- [(match_operand:VI248_VLBW 1 "register_operand" "0")
- (match_operand:VI248_VLBW 2 "register_operand" "v")
- (match_operand:VI248_VLBW 3 "nonimmediate_operand" "vm")]
+ [(set (match_operand:VI248_AVX512VL 0 "register_operand" "=v")
+ (unspec:VI248_AVX512VL
+ [(match_operand:VI248_AVX512VL 1 "register_operand" "0")
+ (match_operand:VI248_AVX512VL 2 "register_operand" "v")
+ (match_operand:VI248_AVX512VL 3 "nonimmediate_operand" "vm")]
UNSPEC_VPSHRDV))]
"TARGET_AVX512VBMI2"
"vpshrdv<ssemodesuffix>\t{%3, %2, %0|%0, %2, %3 }"
@@ -20168,12 +20168,12 @@ (define_insn "vpshrdv_<mode>"
(set_attr "mode" "<sseinsnmode>")])
(define_insn "vpshrdv_<mode>_mask"
- [(set (match_operand:VI248_VLBW 0 "register_operand" "=v")
- (vec_merge:VI248_VLBW
- (unspec:VI248_VLBW
- [(match_operand:VI248_VLBW 1 "register_operand" "0")
- (match_operand:VI248_VLBW 2 "register_operand" "v")
- (match_operand:VI248_VLBW 3 "nonimmediate_operand" "vm")]
+ [(set (match_operand:VI248_AVX512VL 0 "register_operand" "=v")
+ (vec_merge:VI248_AVX512VL
+ (unspec:VI248_AVX512VL
+ [(match_operand:VI248_AVX512VL 1 "register_operand" "0")
+ (match_operand:VI248_AVX512VL 2 "register_operand" "v")
+ (match_operand:VI248_AVX512VL 3 "nonimmediate_operand" "vm")]
UNSPEC_VPSHRDV)
(match_dup 1)
(match_operand:<avx512fmaskmode> 4 "register_operand" "Yk")))]
@@ -20183,10 +20183,10 @@ (define_insn "vpshrdv_<mode>_mask"
(set_attr "mode" "<sseinsnmode>")])
(define_expand "vpshrdv_<mode>_maskz"
- [(match_operand:VI248_VLBW 0 "register_operand")
- (match_operand:VI248_VLBW 1 "register_operand")
- (match_operand:VI248_VLBW 2 "register_operand")
- (match_operand:VI248_VLBW 3 "nonimmediate_operand")
+ [(match_operand:VI248_AVX512VL 0 "register_operand")
+ (match_operand:VI248_AVX512VL 1 "register_operand")
+ (match_operand:VI248_AVX512VL 2 "register_operand")
+ (match_operand:VI248_AVX512VL 3 "nonimmediate_operand")
(match_operand:<avx512fmaskmode> 4 "register_operand")]
"TARGET_AVX512VBMI2"
{
@@ -20198,14 +20198,14 @@ (define_expand "vpshrdv_<mode>_maskz"
})
(define_insn "vpshrdv_<mode>_maskz_1"
- [(set (match_operand:VI248_VLBW 0 "register_operand" "=v")
- (vec_merge:VI248_VLBW
- (unspec:VI248_VLBW
- [(match_operand:VI248_VLBW 1 "register_operand" "0")
- (match_operand:VI248_VLBW 2 "register_operand" "v")
- (match_operand:VI248_VLBW 3 "nonimmediate_operand" "vm")]
+ [(set (match_operand:VI248_AVX512VL 0 "register_operand" "=v")
+ (vec_merge:VI248_AVX512VL
+ (unspec:VI248_AVX512VL
+ [(match_operand:VI248_AVX512VL 1 "register_operand" "0")
+ (match_operand:VI248_AVX512VL 2 "register_operand" "v")
+ (match_operand:VI248_AVX512VL 3 "nonimmediate_operand" "vm")]
UNSPEC_VPSHRDV)
- (match_operand:VI248_VLBW 4 "const0_operand" "C")
+ (match_operand:VI248_AVX512VL 4 "const0_operand" "C")
(match_operand:<avx512fmaskmode> 5 "register_operand" "Yk")))]
"TARGET_AVX512VBMI2"
"vpshrdv<ssemodesuffix>\t{%3, %2, %0%{%5%}%{z%}|%0%{%5%}%{z%}, %2, %3 }"
@@ -20213,11 +20213,11 @@ (define_insn "vpshrdv_<mode>_maskz_1"
(set_attr "mode" "<sseinsnmode>")])
(define_insn "vpshldv_<mode>"
- [(set (match_operand:VI248_VLBW 0 "register_operand" "=v")
- (unspec:VI248_VLBW
- [(match_operand:VI248_VLBW 1 "register_operand" "0")
- (match_operand:VI248_VLBW 2 "register_operand" "v")
- (match_operand:VI248_VLBW 3 "nonimmediate_operand" "vm")]
+ [(set (match_operand:VI248_AVX512VL 0 "register_operand" "=v")
+ (unspec:VI248_AVX512VL
+ [(match_operand:VI248_AVX512VL 1 "register_operand" "0")
+ (match_operand:VI248_AVX512VL 2 "register_operand" "v")
+ (match_operand:VI248_AVX512VL 3 "nonimmediate_operand" "vm")]
UNSPEC_VPSHLDV))]
"TARGET_AVX512VBMI2"
"vpshldv<ssemodesuffix>\t{%3, %2, %0|%0, %2, %3 }"
@@ -20225,12 +20225,12 @@ (define_insn "vpshldv_<mode>"
(set_attr "mode" "<sseinsnmode>")])
(define_insn "vpshldv_<mode>_mask"
- [(set (match_operand:VI248_VLBW 0 "register_operand" "=v")
- (vec_merge:VI248_VLBW
- (unspec:VI248_VLBW
- [(match_operand:VI248_VLBW 1 "register_operand" "0")
- (match_operand:VI248_VLBW 2 "register_operand" "v")
- (match_operand:VI248_VLBW 3 "nonimmediate_operand" "vm")]
+ [(set (match_operand:VI248_AVX512VL 0 "register_operand" "=v")
+ (vec_merge:VI248_AVX512VL
+ (unspec:VI248_AVX512VL
+ [(match_operand:VI248_AVX512VL 1 "register_operand" "0")
+ (match_operand:VI248_AVX512VL 2 "register_operand" "v")
+ (match_operand:VI248_AVX512VL 3 "nonimmediate_operand" "vm")]
UNSPEC_VPSHLDV)
(match_dup 1)
(match_operand:<avx512fmaskmode> 4 "register_operand" "Yk")))]
@@ -20240,10 +20240,10 @@ (define_insn "vpshldv_<mode>_mask"
(set_attr "mode" "<sseinsnmode>")])
(define_expand "vpshldv_<mode>_maskz"
- [(match_operand:VI248_VLBW 0 "register_operand")
- (match_operand:VI248_VLBW 1 "register_operand")
- (match_operand:VI248_VLBW 2 "register_operand")
- (match_operand:VI248_VLBW 3 "nonimmediate_operand")
+ [(match_operand:VI248_AVX512VL 0 "register_operand")
+ (match_operand:VI248_AVX512VL 1 "register_operand")
+ (match_operand:VI248_AVX512VL 2 "register_operand")
+ (match_operand:VI248_AVX512VL 3 "nonimmediate_operand")
(match_operand:<avx512fmaskmode> 4 "register_operand")]
"TARGET_AVX512VBMI2"
{
@@ -20255,14 +20255,14 @@ (define_expand "vpshldv_<mode>_maskz"
})
(define_insn "vpshldv_<mode>_maskz_1"
- [(set (match_operand:VI248_VLBW 0 "register_operand" "=v")
- (vec_merge:VI248_VLBW
- (unspec:VI248_VLBW
- [(match_operand:VI248_VLBW 1 "register_operand" "0")
- (match_operand:VI248_VLBW 2 "register_operand" "v")
- (match_operand:VI248_VLBW 3 "nonimmediate_operand" "vm")]
+ [(set (match_operand:VI248_AVX512VL 0 "register_operand" "=v")
+ (vec_merge:VI248_AVX512VL
+ (unspec:VI248_AVX512VL
+ [(match_operand:VI248_AVX512VL 1 "register_operand" "0")
+ (match_operand:VI248_AVX512VL 2 "register_operand" "v")
+ (match_operand:VI248_AVX512VL 3 "nonimmediate_operand" "vm")]
UNSPEC_VPSHLDV)
- (match_operand:VI248_VLBW 4 "const0_operand" "C")
+ (match_operand:VI248_AVX512VL 4 "const0_operand" "C")
(match_operand:<avx512fmaskmode> 5 "register_operand" "Yk")))]
"TARGET_AVX512VBMI2"
"vpshldv<ssemodesuffix>\t{%3, %2, %0%{%5%}%{z%}|%0%{%5%}%{z%}, %2, %3 }"
--- gcc/testsuite/gcc.target/i386/pr83604.c.jj 2017-12-27 19:33:19.835077783 +0100
+++ gcc/testsuite/gcc.target/i386/pr83604.c 2017-12-27 19:32:37.661052000 +0100
@@ -0,0 +1,11 @@
+/* PR target/83604 */
+/* { dg-do compile } */
+/* { dg-options "-O2 -mno-avx" } */
+
+typedef short V __attribute__((__vector_size__(64)));
+
+__attribute__((target ("avx512vbmi2"))) V
+foo (V x, V y, V z)
+{
+ return __builtin_ia32_vpshrdv_v32hi (x, y, z);
+}
Jakub
^ permalink raw reply [flat|nested] 2+ messages in thread
* Re: [PATCH] Fix a vbmi2 ICE (PR target/83604)
2017-12-28 9:10 [PATCH] Fix a vbmi2 ICE (PR target/83604) Jakub Jelinek
@ 2018-01-05 16:06 ` Kirill Yukhin
0 siblings, 0 replies; 2+ messages in thread
From: Kirill Yukhin @ 2018-01-05 16:06 UTC (permalink / raw)
To: Jakub Jelinek; +Cc: Uros Bizjak, gcc-patches, Julia Koval, Sebastian Peryt
Hello Jakub!
On 28 Dec 10:10, Jakub Jelinek wrote:
> Hi!
>
> These insns don't really need AVX512BW in any way themselves, only their
> masked variants might need it for reloading of the mask register, but that
> should be covered in builtins.def, doesn't need duplication in sse.md.
> For non-masked it causes ICEs, because the builtins properly aren't guarded
> with AVX512BW, but the insns incorrectly require that.
>
> Fixed thusly, bootstrapped/regtested on x86_64-linux and i686-linux, ok for
> trunk?
Your patch is OK.
>
> 2017-12-28 Jakub Jelinek <jakub@redhat.com>
>
> PR target/83604
> * config/i386/sse.md (VI248_VLBW): Rename to ...
> (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
> (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
> vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
> vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
> vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
> mode iterator instead of VI248_VLBW.
>
> * gcc.target/i386/pr83604.c: New test.
--
Thanks, K
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2018-01-05 16:06 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2017-12-28 9:10 [PATCH] Fix a vbmi2 ICE (PR target/83604) Jakub Jelinek
2018-01-05 16:06 ` Kirill Yukhin
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).