* [PATCH] i386: Autovectorize 4-byte vectors
@ 2021-05-26 18:49 Uros Bizjak
0 siblings, 0 replies; only message in thread
From: Uros Bizjak @ 2021-05-26 18:49 UTC (permalink / raw)
To: gcc-patches
[-- Attachment #1: Type: text/plain, Size: 888 bytes --]
2021-05-26 Uroš Bizjak <ubizjak@gmail.com>
gcc/
* config/i386/i386.c (ix86_autovectorize_vector_modes):
Add V4QImode and V16QImode for TARGET_SSE2.
* doc/sourcebuild.texi (Vector-specific attributes):
Add vect64 and vect32 description.
gcc/testsuite/
* lib/target-supports.exp (check_effective_target_vect32): New.
(available_vector_sizes): Append 32 for x86 targets.
* gcc.dg/vect/pr71264.c (dg-final): Xfail scan dump for vect32 targets.
* gcc.dg/vect/slp-28.c (dg-final): Adjust dumps for vect32 targets.
* gcc.dg/vect/slp-3.c (dg-final): Ditto.
* gcc.target/i386/pr100637-3b.c: New test.
* gcc.target/i386/pr100637-3w.c: Ditto.
* gcc.target/i386/pr100637-4b.c: Ditto.
* gcc.target/i386/pr100637-4w.c: Ditto.
Bootstrapped and regression tested on x86_64-linux-gnu {,-m32}.
Pushed to master.
Uros.
[-- Attachment #2: p.diff.txt --]
[-- Type: text/plain, Size: 8516 bytes --]
diff --git a/gcc/config/i386/i386.c b/gcc/config/i386/i386.c
index 28e6113a609..04649b42122 100644
--- a/gcc/config/i386/i386.c
+++ b/gcc/config/i386/i386.c
@@ -22190,12 +22190,15 @@ ix86_autovectorize_vector_modes (vector_modes *modes, bool all)
modes->safe_push (V16QImode);
modes->safe_push (V32QImode);
}
- else if (TARGET_MMX_WITH_SSE)
+ else if (TARGET_SSE2)
modes->safe_push (V16QImode);
if (TARGET_MMX_WITH_SSE)
modes->safe_push (V8QImode);
+ if (TARGET_SSE2)
+ modes->safe_push (V4QImode);
+
return 0;
}
diff --git a/gcc/doc/sourcebuild.texi b/gcc/doc/sourcebuild.texi
index cf3098749c0..16c6a3b8e99 100644
--- a/gcc/doc/sourcebuild.texi
+++ b/gcc/doc/sourcebuild.texi
@@ -1740,6 +1740,12 @@ circumstances.
@item vect_variable_length
Target has variable-length vectors.
+@item vect64
+Target supports vectors of 64 bits.
+
+@item vect32
+Target supports vectors of 32 bits.
+
@item vect_widen_sum_hi_to_si
Target supports a vector widening summation of @code{short} operands
into @code{int} results, or can promote (unpack) from @code{short}
diff --git a/gcc/testsuite/gcc.dg/vect/pr71264.c b/gcc/testsuite/gcc.dg/vect/pr71264.c
index dc849bf2797..1381e0ed132 100644
--- a/gcc/testsuite/gcc.dg/vect/pr71264.c
+++ b/gcc/testsuite/gcc.dg/vect/pr71264.c
@@ -19,5 +19,4 @@ void test(uint8_t *ptr, uint8_t *mask)
}
}
-/* { dg-final { scan-tree-dump "vectorized 1 loops in function" "vect" { xfail s390*-*-* sparc*-*-* } } } */
-
+/* { dg-final { scan-tree-dump "vectorized 1 loops in function" "vect" { xfail { { s390*-*-* sparc*-*-* } || vect32 } } } } */
diff --git a/gcc/testsuite/gcc.dg/vect/slp-28.c b/gcc/testsuite/gcc.dg/vect/slp-28.c
index 7778bad4465..0bb5f0eb0e4 100644
--- a/gcc/testsuite/gcc.dg/vect/slp-28.c
+++ b/gcc/testsuite/gcc.dg/vect/slp-28.c
@@ -88,6 +88,7 @@ int main (void)
return 0;
}
-/* { dg-final { scan-tree-dump-times "vectorized 1 loops" 1 "vect" } } */
-/* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 1 "vect" } } */
+/* { dg-final { scan-tree-dump-times "vectorized 1 loops" 1 "vect" { target { ! vect32 } } } } */
+/* { dg-final { scan-tree-dump-times "vectorized 2 loops" 1 "vect" { target vect32 } } } */
+/* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 1 "vect" { target { ! vect32 } } } } */
diff --git a/gcc/testsuite/gcc.dg/vect/slp-3.c b/gcc/testsuite/gcc.dg/vect/slp-3.c
index 46ab584419a..80ded1840ad 100644
--- a/gcc/testsuite/gcc.dg/vect/slp-3.c
+++ b/gcc/testsuite/gcc.dg/vect/slp-3.c
@@ -141,8 +141,8 @@ int main (void)
return 0;
}
-/* { dg-final { scan-tree-dump-times "vectorized 3 loops" 1 "vect" { target { ! vect_partial_vectors } } } } */
-/* { dg-final { scan-tree-dump-times "vectorized 4 loops" 1 "vect" { target vect_partial_vectors } } } */
-/* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 3 "vect" { target { ! vect_partial_vectors } } } }*/
-/* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 4 "vect" { target vect_partial_vectors } } } */
+/* { dg-final { scan-tree-dump-times "vectorized 3 loops" 1 "vect" { target { ! { vect_partial_vectors || vect32 } } } } } */
+/* { dg-final { scan-tree-dump-times "vectorized 4 loops" 1 "vect" { target { vect_partial_vectors || vect32 } } } } */
+/* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 3 "vect" { target { ! { vect_partial_vectors || vect32 } } } } }*/
+/* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 4 "vect" { target { vect_partial_vectors || vect32 } } } } */
diff --git a/gcc/testsuite/gcc.target/i386/pr100637-3b.c b/gcc/testsuite/gcc.target/i386/pr100637-3b.c
new file mode 100644
index 00000000000..16df70059a9
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr100637-3b.c
@@ -0,0 +1,56 @@
+/* PR target/100637 */
+/* { dg-do compile } */
+/* { dg-options "-O2 -ftree-vectorize -msse4" } */
+
+char r[4], a[4], b[4];
+unsigned char ur[4], ua[4], ub[4];
+
+void maxs (void)
+{
+ int i;
+
+ for (i = 0; i < 4; i++)
+ r[i] = a[i] > b[i] ? a[i] : b[i];
+}
+
+/* { dg-final { scan-assembler "pmaxsb" } } */
+
+void maxu (void)
+{
+ int i;
+
+ for (i = 0; i < 4; i++)
+ ur[i] = ua[i] > ub[i] ? ua[i] : ub[i];
+}
+
+/* { dg-final { scan-assembler "pmaxub" } } */
+
+void mins (void)
+{
+ int i;
+
+ for (i = 0; i < 4; i++)
+ r[i] = a[i] < b[i] ? a[i] : b[i];
+}
+
+/* { dg-final { scan-assembler "pminsb" } } */
+
+void minu (void)
+{
+ int i;
+
+ for (i = 0; i < 4; i++)
+ ur[i] = ua[i] < ub[i] ? ua[i] : ub[i];
+}
+
+/* { dg-final { scan-assembler "pminub" } } */
+
+void _abs (void)
+{
+ int i;
+
+ for (i = 0; i < 4; i++)
+ r[i] = a[i] < 0 ? -a[i] : a[i];
+}
+
+/* { dg-final { scan-assembler "pabsb" } } */
diff --git a/gcc/testsuite/gcc.target/i386/pr100637-3w.c b/gcc/testsuite/gcc.target/i386/pr100637-3w.c
new file mode 100644
index 00000000000..7f1882e7a56
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr100637-3w.c
@@ -0,0 +1,86 @@
+/* PR target/100637 */
+/* { dg-do compile } */
+/* { dg-options "-O2 -ftree-vectorize -msse4" } */
+
+short r[2], a[2], b[2];
+unsigned short ur[2], ua[2], ub[2];
+
+void mulh (void)
+{
+ int i;
+
+ for (i = 0; i < 2; i++)
+ r[i] = ((int) a[i] * b[i]) >> 16;
+}
+
+/* { dg-final { scan-assembler "pmulhw" { xfail *-*-* } } } */
+
+void mulhu (void)
+{
+ int i;
+
+ for (i = 0; i < 2; i++)
+ ur[i] = ((unsigned int) ua[i] * ub[i]) >> 16;
+}
+
+/* { dg-final { scan-assembler "pmulhuw" { xfail *-*-* } } } */
+
+void mulhrs (void)
+{
+ int i;
+
+ for (i = 0; i < 2; i++)
+ r[i] = ((((int) a[i] * b[i]) >> 14) + 1) >> 1;
+}
+
+/* { dg-final { scan-assembler "pmulhrsw" } } */
+
+void maxs (void)
+{
+ int i;
+
+ for (i = 0; i < 2; i++)
+ r[i] = a[i] > b[i] ? a[i] : b[i];
+}
+
+/* { dg-final { scan-assembler "pmaxsw" } } */
+
+void maxu (void)
+{
+ int i;
+
+ for (i = 0; i < 2; i++)
+ ur[i] = ua[i] > ub[i] ? ua[i] : ub[i];
+}
+
+/* { dg-final { scan-assembler "pmaxuw" } } */
+
+void mins (void)
+{
+ int i;
+
+ for (i = 0; i < 2; i++)
+ r[i] = a[i] < b[i] ? a[i] : b[i];
+}
+
+/* { dg-final { scan-assembler "pminsw" } } */
+
+void minu (void)
+{
+ int i;
+
+ for (i = 0; i < 2; i++)
+ ur[i] = ua[i] < ub[i] ? ua[i] : ub[i];
+}
+
+/* { dg-final { scan-assembler "pminuw" } } */
+
+void _abs (void)
+{
+ int i;
+
+ for (i = 0; i < 2; i++)
+ r[i] = a[i] < 0 ? -a[i] : a[i];
+}
+
+/* { dg-final { scan-assembler "pabsw" } } */
diff --git a/gcc/testsuite/gcc.target/i386/pr100637-4b.c b/gcc/testsuite/gcc.target/i386/pr100637-4b.c
new file mode 100644
index 00000000000..198e3dd3352
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr100637-4b.c
@@ -0,0 +1,19 @@
+/* PR target/100637 */
+/* { dg-do compile } */
+/* { dg-options "-O2 -ftree-vectorize -msse2" } */
+
+typedef char T;
+
+#define M 4
+
+extern T a[M], b[M], s1[M], s2[M], r[M];
+
+void foo (void)
+{
+ int j;
+
+ for (j = 0; j < M; j++)
+ r[j] = (a[j] < b[j]) ? s1[j] : s2[j];
+}
+
+/* { dg-final { scan-assembler "pcmpgtb" } } */
diff --git a/gcc/testsuite/gcc.target/i386/pr100637-4w.c b/gcc/testsuite/gcc.target/i386/pr100637-4w.c
new file mode 100644
index 00000000000..0f5dacce906
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr100637-4w.c
@@ -0,0 +1,19 @@
+/* PR target/100637 */
+/* { dg-do compile } */
+/* { dg-options "-O2 -ftree-vectorize -msse2" } */
+
+typedef short T;
+
+#define M 2
+
+extern T a[M], b[M], s1[M], s2[M], r[M];
+
+void foo (void)
+{
+ int j;
+
+ for (j = 0; j < M; j++)
+ r[j] = (a[j] < b[j]) ? s1[j] : s2[j];
+}
+
+/* { dg-final { scan-assembler "pcmpgtw" { xfail *-*-* } } } */
diff --git a/gcc/testsuite/lib/target-supports.exp b/gcc/testsuite/lib/target-supports.exp
index 849f1bbeda5..7f78c5593ac 100644
--- a/gcc/testsuite/lib/target-supports.exp
+++ b/gcc/testsuite/lib/target-supports.exp
@@ -7626,6 +7626,7 @@ proc available_vector_sizes { } {
if { ![is-effective-target ia32] } {
lappend result 64
}
+ lappend result 32
} elseif { [istarget sparc*-*-*] } {
lappend result 64
} elseif { [istarget amdgcn*-*-*] } {
@@ -7655,6 +7656,12 @@ proc check_effective_target_vect64 { } {
return [expr { [lsearch -exact [available_vector_sizes] 64] >= 0 }]
}
+# Return 1 if the target supports vectors of 32 bits.
+
+proc check_effective_target_vect32 { } {
+ return [expr { [lsearch -exact [available_vector_sizes] 32] >= 0 }]
+}
+
# Return 1 if the target supports vector copysignf calls.
proc check_effective_target_vect_call_copysignf { } {
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2021-05-26 18:49 UTC | newest]
Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-05-26 18:49 [PATCH] i386: Autovectorize 4-byte vectors 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).