public inbox for gcc-cvs@sourceware.org
help / color / mirror / Atom feed
* [gcc r12-3542] Output vextract{i, f}{32x4, 64x2} for (vec_select:(reg:Vmode) idx) when byte_offset of idx % 16 == 0.
@ 2021-09-15 8:38 hongtao Liu
0 siblings, 0 replies; only message in thread
From: hongtao Liu @ 2021-09-15 8:38 UTC (permalink / raw)
To: gcc-cvs
https://gcc.gnu.org/g:243e0a5b1942879bc005bf150a744e69a4fcdc87
commit r12-3542-g243e0a5b1942879bc005bf150a744e69a4fcdc87
Author: liuhongt <hongtao.liu@intel.com>
Date: Mon Sep 13 10:27:51 2021 +0800
Output vextract{i,f}{32x4,64x2} for (vec_select:(reg:Vmode) idx) when byte_offset of idx % 16 == 0.
2020-09-13 Hongtao Liu <hongtao.liu@intel.com>
Peter Cordes <peter@cordes.ca>
gcc/ChangeLog:
PR target/91103
* config/i386/sse.md (extract_suf): Add V8SF/V8SI/V4DF/V4DI.
(*vec_extract<mode><ssescalarmodelower>_valign): Output
vextract{i,f}{32x4,64x2} instruction when byte_offset % 16 ==
0.
gcc/testsuite/ChangeLog:
PR target/91103
* gcc.target/i386/pr91103-1.c: Add extract tests.
* gcc.target/i386/pr91103-2.c: Ditto.
Diff:
---
gcc/config/i386/sse.md | 21 +++++++++++++++++----
gcc/testsuite/gcc.target/i386/pr91103-1.c | 7 ++++++-
gcc/testsuite/gcc.target/i386/pr91103-2.c | 4 ++++
3 files changed, 27 insertions(+), 5 deletions(-)
diff --git a/gcc/config/i386/sse.md b/gcc/config/i386/sse.md
index b67ec1507d9..15645395d83 100644
--- a/gcc/config/i386/sse.md
+++ b/gcc/config/i386/sse.md
@@ -9089,7 +9089,8 @@
[(V16SF "avx512f") (V16SI "avx512f") (V8DF "avx512dq") (V8DI "avx512dq")])
(define_mode_attr extract_suf
- [(V16SF "32x4") (V16SI "32x4") (V8DF "64x2") (V8DI "64x2")])
+ [(V16SF "32x4") (V16SI "32x4") (V8DF "64x2") (V8DI "64x2")
+ (V8SF "32x4") (V8SI "32x4") (V4DF "64x2") (V4DI "64x2")])
(define_mode_iterator AVX512_VEC
[(V8DF "TARGET_AVX512DQ") (V8DI "TARGET_AVX512DQ") V16SF V16SI])
@@ -10661,9 +10662,21 @@
(match_operand:V48_256_512_AVX512VL 1 "register_operand" "v")
(parallel [(match_operand 2 "<vec_extract_imm_predicate>")])))]
"TARGET_AVX512F
- && INTVAL(operands[2]) >= 16 / GET_MODE_SIZE (<ssescalarmode>mode)"
- "valign<ternlogsuffix>\t{%2, %1, %1, %<xtg_mode>0|%<xtg_mode>0, %1, %1, %2}";
- [(set_attr "prefix" "evex")
+ && INTVAL(operands[2]) * GET_MODE_SIZE (<ssescalarmode>mode) >= 16"
+{
+ int byte_offset = INTVAL (operands[2]) * GET_MODE_SIZE (<ssescalarmode>mode);
+ if (byte_offset % 16 == 0)
+ {
+ operands[2] = GEN_INT (byte_offset / 16);
+ if (byte_offset / 16 == 1)
+ return "vextract<shuffletype><extract_suf>\t{%2, %t1, %x0|%x0, %t1, %2}";
+ else
+ return "vextract<shuffletype><extract_suf>\t{%2, %1, %x0|%x0, %1, %2}";
+ }
+ else
+ return "valign<ternlogsuffix>\t{%2, %1, %1, %<xtg_mode>0|%<xtg_mode>0, %1, %1, %2}";
+}
+ [(set_attr "prefix" "maybe_evex")
(set_attr "mode" "<sseintvecinsnmode>")])
(define_expand "avx512f_shufps512_mask"
diff --git a/gcc/testsuite/gcc.target/i386/pr91103-1.c b/gcc/testsuite/gcc.target/i386/pr91103-1.c
index 11caaa8bd1b..2d78a6da0f3 100644
--- a/gcc/testsuite/gcc.target/i386/pr91103-1.c
+++ b/gcc/testsuite/gcc.target/i386/pr91103-1.c
@@ -1,6 +1,7 @@
/* { dg-do compile } */
/* { dg-options "-mavx512vl -O2" } */
-/* { dg-final { scan-assembler-times "valign\[dq\]" 16 } } */
+/* { dg-final { scan-assembler-times "valign\[dq\]" 8 } } */
+/* { dg-final { scan-assembler-times "vextract" 12 } } */
typedef float v8sf __attribute__((vector_size(32)));
typedef float v16sf __attribute__((vector_size(64)));
@@ -23,9 +24,13 @@ EXTRACT (v8sf, float, 4);
EXTRACT (v8sf, float, 7);
EXTRACT (v8si, int, 4);
EXTRACT (v8si, int, 7);
+EXTRACT (v16sf, float, 4);
EXTRACT (v16sf, float, 8);
+EXTRACT (v16sf, float, 12);
EXTRACT (v16sf, float, 15);
+EXTRACT (v16si, int, 4);
EXTRACT (v16si, int, 8);
+EXTRACT (v16si, int, 12);
EXTRACT (v16si, int, 15);
EXTRACT (v4df, double, 2);
EXTRACT (v4df, double, 3);
diff --git a/gcc/testsuite/gcc.target/i386/pr91103-2.c b/gcc/testsuite/gcc.target/i386/pr91103-2.c
index 010e4775723..a928d87f99a 100644
--- a/gcc/testsuite/gcc.target/i386/pr91103-2.c
+++ b/gcc/testsuite/gcc.target/i386/pr91103-2.c
@@ -61,9 +61,13 @@ RUNCHECK (f2, v8sf, float, 4);
RUNCHECK (f2, v8sf, float, 7);
RUNCHECK (di2, v8si, int, 4);
RUNCHECK (di2, v8si, int, 7);
+RUNCHECK (f1, v16sf, float, 4);
RUNCHECK (f1, v16sf, float, 8);
+RUNCHECK (f1, v16sf, float, 12);
RUNCHECK (f1, v16sf, float, 15);
+RUNCHECK (di1, v16si, int, 4);
RUNCHECK (di1, v16si, int, 8);
+RUNCHECK (di1, v16si, int, 12);
RUNCHECK (di1, v16si, int, 15);
RUNCHECK (d2, v4df, double, 2);
RUNCHECK (d2, v4df, double, 3);
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2021-09-15 8:38 UTC | newest]
Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-09-15 8:38 [gcc r12-3542] Output vextract{i, f}{32x4, 64x2} for (vec_select:(reg:Vmode) idx) when byte_offset of idx % 16 == 0 hongtao Liu
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).