* [PATCH i386 AVX512] [28/n] Add si/di masked interleave.
@ 2014-08-28 7:07 Kirill Yukhin
2014-08-28 7:51 ` Uros Bizjak
0 siblings, 1 reply; 2+ messages in thread
From: Kirill Yukhin @ 2014-08-28 7:07 UTC (permalink / raw)
To: Uros Bizjak; +Cc: Jakub Jelinek, Richard Henderson, GCC Patches, kirill.yukhin
Hello,
This patch adds support for SI/DO masked interleaves.
Bootstrapped.
AVX-512* tests on top of patch-set all pass
under simulator.
Is it ok for trunk?
gcc/
* config/i386/sse.md
(define_insn "avx2_interleave_highv4di<mask_name>"): Add masking.
(define_insn "vec_interleave_highv2di<mask_name>"): Ditto.
(define_insn "avx2_interleave_lowv4di<mask_name>"): Ditto.
(define_insn "vec_interleave_lowv2di<mask_name>"): Ditto.
--
Thanks, K
diff --git a/gcc/config/i386/sse.md b/gcc/config/i386/sse.md
index c602eeb..c9931b4 100644
--- a/gcc/config/i386/sse.md
+++ b/gcc/config/i386/sse.md
@@ -7515,18 +7515,18 @@
})
;; punpcklqdq and punpckhqdq are shorter than shufpd.
-(define_insn "avx2_interleave_highv4di"
- [(set (match_operand:V4DI 0 "register_operand" "=x")
+(define_insn "avx2_interleave_highv4di<mask_name>"
+ [(set (match_operand:V4DI 0 "register_operand" "=v")
(vec_select:V4DI
(vec_concat:V8DI
- (match_operand:V4DI 1 "register_operand" "x")
- (match_operand:V4DI 2 "nonimmediate_operand" "xm"))
+ (match_operand:V4DI 1 "register_operand" "v")
+ (match_operand:V4DI 2 "nonimmediate_operand" "vm"))
(parallel [(const_int 1)
(const_int 5)
(const_int 3)
(const_int 7)])))]
- "TARGET_AVX2"
- "vpunpckhqdq\t{%2, %1, %0|%0, %1, %2}"
+ "TARGET_AVX2 && <mask_avx512vl_condition>"
+ "vpunpckhqdq\t{%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %2}"
[(set_attr "type" "sselog")
(set_attr "prefix" "vex")
(set_attr "mode" "OI")])
@@ -7547,36 +7547,36 @@
(set_attr "prefix" "evex")
(set_attr "mode" "XI")])
-(define_insn "vec_interleave_highv2di"
- [(set (match_operand:V2DI 0 "register_operand" "=x,x")
+(define_insn "vec_interleave_highv2di<mask_name>"
+ [(set (match_operand:V2DI 0 "register_operand" "=x,v")
(vec_select:V2DI
(vec_concat:V4DI
- (match_operand:V2DI 1 "register_operand" "0,x")
- (match_operand:V2DI 2 "nonimmediate_operand" "xm,xm"))
+ (match_operand:V2DI 1 "register_operand" "0,v")
+ (match_operand:V2DI 2 "nonimmediate_operand" "xm,vm"))
(parallel [(const_int 1)
(const_int 3)])))]
- "TARGET_SSE2"
+ "TARGET_SSE2 && <mask_avx512vl_condition>"
"@
punpckhqdq\t{%2, %0|%0, %2}
- vpunpckhqdq\t{%2, %1, %0|%0, %1, %2}"
+ vpunpckhqdq\t{%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %2}"
[(set_attr "isa" "noavx,avx")
(set_attr "type" "sselog")
(set_attr "prefix_data16" "1,*")
- (set_attr "prefix" "orig,vex")
+ (set_attr "prefix" "orig,<mask_prefix>")
(set_attr "mode" "TI")])
-(define_insn "avx2_interleave_lowv4di"
- [(set (match_operand:V4DI 0 "register_operand" "=x")
+(define_insn "avx2_interleave_lowv4di<mask_name>"
+ [(set (match_operand:V4DI 0 "register_operand" "=v")
(vec_select:V4DI
(vec_concat:V8DI
- (match_operand:V4DI 1 "register_operand" "x")
- (match_operand:V4DI 2 "nonimmediate_operand" "xm"))
+ (match_operand:V4DI 1 "register_operand" "v")
+ (match_operand:V4DI 2 "nonimmediate_operand" "vm"))
(parallel [(const_int 0)
(const_int 4)
(const_int 2)
(const_int 6)])))]
- "TARGET_AVX2"
- "vpunpcklqdq\t{%2, %1, %0|%0, %1, %2}"
+ "TARGET_AVX2 && <mask_avx512vl_condition>"
+ "vpunpcklqdq\t{%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %2}"
[(set_attr "type" "sselog")
(set_attr "prefix" "vex")
(set_attr "mode" "OI")])
@@ -7597,18 +7597,18 @@
(set_attr "prefix" "evex")
(set_attr "mode" "XI")])
-(define_insn "vec_interleave_lowv2di"
- [(set (match_operand:V2DI 0 "register_operand" "=x,x")
+(define_insn "vec_interleave_lowv2di<mask_name>"
+ [(set (match_operand:V2DI 0 "register_operand" "=x,v")
(vec_select:V2DI
(vec_concat:V4DI
- (match_operand:V2DI 1 "register_operand" "0,x")
- (match_operand:V2DI 2 "nonimmediate_operand" "xm,xm"))
+ (match_operand:V2DI 1 "register_operand" "0,v")
+ (match_operand:V2DI 2 "nonimmediate_operand" "xm,vm"))
(parallel [(const_int 0)
(const_int 2)])))]
- "TARGET_SSE2"
+ "TARGET_SSE2 && <mask_avx512vl_condition>"
"@
punpcklqdq\t{%2, %0|%0, %2}
- vpunpcklqdq\t{%2, %1, %0|%0, %1, %2}"
+ vpunpcklqdq\t{%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %2}"
[(set_attr "isa" "noavx,avx")
(set_attr "type" "sselog")
(set_attr "prefix_data16" "1,*")
^ permalink raw reply [flat|nested] 2+ messages in thread
* Re: [PATCH i386 AVX512] [28/n] Add si/di masked interleave.
2014-08-28 7:07 [PATCH i386 AVX512] [28/n] Add si/di masked interleave Kirill Yukhin
@ 2014-08-28 7:51 ` Uros Bizjak
0 siblings, 0 replies; 2+ messages in thread
From: Uros Bizjak @ 2014-08-28 7:51 UTC (permalink / raw)
To: Kirill Yukhin; +Cc: Jakub Jelinek, Richard Henderson, GCC Patches
On Thu, Aug 28, 2014 at 9:05 AM, Kirill Yukhin <kirill.yukhin@gmail.com> wrote:
> Hello,
> This patch adds support for SI/DO masked interleaves.
>
> Bootstrapped.
> AVX-512* tests on top of patch-set all pass
> under simulator.
>
> Is it ok for trunk?
>
> gcc/
> * config/i386/sse.md
> (define_insn "avx2_interleave_highv4di<mask_name>"): Add masking.
> (define_insn "vec_interleave_highv2di<mask_name>"): Ditto.
> (define_insn "avx2_interleave_lowv4di<mask_name>"): Ditto.
> (define_insn "vec_interleave_lowv2di<mask_name>"): Ditto.
It looks that latest patches all extend patterns in the same way. I
have looked briefly through the patch and rubberstamped it as OK.
OK.
Thanks,
Uros.
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2014-08-28 7:51 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2014-08-28 7:07 [PATCH i386 AVX512] [28/n] Add si/di masked interleave Kirill Yukhin
2014-08-28 7:51 ` Uros Bizjak
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).