On Tue, Jan 28, 2020 at 10:04 AM Uros Bizjak <ubiz...@gmail.com> wrote:
>
> On Tue, Jan 28, 2020 at 6:51 PM H.J. Lu <hjl.to...@gmail.com> wrote:
> >
> > On Tue, Jan 28, 2020 at 9:12 AM Uros Bizjak <ubiz...@gmail.com> wrote:
> > >
> > > On Tue, Jan 28, 2020 at 4:34 PM H.J. Lu <hjl.to...@gmail.com> wrote:
> > >
> > > > > You could move
> > > > >
> > > > > (match_test "TARGET_AVX")
> > > > > (const_string "TI")
> > > > >
> > > > > up to bypass the cases below.
> > > > >
> > > >
> > > > I don't think we can do that. There are 2 cases where we prefer
> > > > movaps/movups:
> > > >
> > > > /* Use packed single precision instructions where posisble. I.e.
> > > > movups instead of movupd. */
> > > > DEF_TUNE (X86_TUNE_SSE_PACKED_SINGLE_INSN_OPTIMAL,
> > > > "sse_packed_single_insn_optimal",
> > > > m_BDVER | m_ZNVER)
> > > >
> > > > /* X86_TUNE_SSE_TYPELESS_STORES: Always movaps/movups for 128bit
> > > > stores. */
> > > > DEF_TUNE (X86_TUNE_SSE_TYPELESS_STORES, "sse_typeless_stores",
> > > > m_AMD_MULTIPLE | m_CORE_ALL | m_GENERIC)
> > > >
> > > > We should always use movaps/movups for
> > > > TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
> > > > It is wrong to bypass TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with
> > > > TARGET_AVX
> > > > as m_BDVER | m_ZNVER support AVX.
> > >
> > > The reason for TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL on AMD target is
> > > only insn size, as advised in e.g. Software Optimization Guide for the
> > > AMD Family 15h Processors [1], section 7.1.2, where it is said:
> > >
> > > --quote--
> > > 7.1.2 Reduce Instruction SizeOptimization
> > >
> > > Reduce the size of instructions when possible.
> > >
> > > Rationale
> > >
> > > Using smaller instruction sizes improves instruction fetch throughput.
> > > Specific examples include the following:
> > >
> > > *In SIMD code, use the single-precision (PS) form of instructions
> > > instead of the double-precision (PD) form. For example, for register
> > > to register moves, MOVAPS achieves the same result as MOVAPD, but uses
> > > one less byte to encode the instruction and has no prefix byte. Other
> > > examples in which single-precision forms can be substituted for
> > > double-precision forms include MOVUPS, MOVNTPS, XORPS, ORPS, ANDPS,
> > > and SHUFPS.
> > > ...
> > > --/quote--
> > >
> > > Please note that this optimization applies only to non-AVX forms, as
> > > demonstrated by:
> > >
> > > 0: 0f 28 c8 movaps %xmm0,%xmm1
> > > 3: 66 0f 28 c8 movapd %xmm0,%xmm1
> > > 7: c5 f8 28 d1 vmovaps %xmm1,%xmm2
> > > b: c5 f9 28 d1 vmovapd %xmm1,%xmm2
> > >
> > > Also note that MOVDQA is missing in the above optimization. It is
> > > harmful to substitute MOVDQA with MOVAPS, as it can (and does)
> > > introduce +1 cycle forwarding penalty between FLT (FPA/FPM) and INT
> > > (VALU) FP clusters.
> > >
> > > Following the above optimization, it is obvious that
> > > TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling was cargo-culted from
> > > one pattern to another. Its use should be reviewed and fixed where not
> > > appropriate.
> > >
> > > [1] https://www.amd.com/system/files/TechDocs/47414_15h_sw_opt_guide.pdf
> > >
> > > Uros.
> >
> > Here is the updated patch which moves TARGET_AVX before
> > TARGET_SSE_TYPELESS_STORES. OK for master if there is
> > no regression?
> >
> > Thanks.
>
>
> + (match_test "TARGET_AVX")
> + (const_string "<sseinsnmode>")
> (and (match_test "<MODE_SIZE> == 16")
>
> Only MODE_SIZE == 16 cases will be left here, since TARGET_AVX is
> necessary for MODE_SIZE > 16. This test can be removed.
>
> OK with the above change.
>
This is the patch I am going to check in.
Thanks.
--
H.J.
From 66c534dedc7a9a632aa38c32e3f7c251b8f2c778 Mon Sep 17 00:00:00 2001
From: "H.J. Lu" <hjl.to...@gmail.com>
Date: Mon, 27 Jan 2020 09:35:11 -0800
Subject: [PATCH] i386: Prefer TARGET_AVX over TARGET_SSE_TYPELESS_STORES
movaps/movups is one byte shorter than movdaq/movdqu. But it isn't the
case for AVX nor AVX512. This patch prefers TARGET_AVX over
TARGET_SSE_TYPELESS_STORES and adjust vmovups checks in assembly ouputs.
gcc/
PR target/91461
* config/i386/i386.md (*movoi_internal_avx): Remove
TARGET_SSE_TYPELESS_STORES check.
(*movti_internal): Prefer TARGET_AVX over
TARGET_SSE_TYPELESS_STORES.
(*movtf_internal): Likewise.
* config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
from TARGET_SSE_TYPELESS_STORES.
gcc/testsuite/
PR target/91461
* gcc.target/i386/avx256-unaligned-store-2.c: Don't check
vmovups.
* gcc.target/i386/avx256-unaligned-store-3.c: Likewise.
* gcc.target/i386/pieces-memcpy-4.c: Likewise.
* gcc.target/i386/pieces-memcpy-5.c: Likewise.
* gcc.target/i386/pieces-memcpy-6.c: Likewise.
* gcc.target/i386/pieces-strcpy-2.c: Likewise.
* gcc.target/i386/pr90980-1.c: Likewise.
* gcc.target/i386/pr87317-4.c: Check "\tvmovd\t" instead of
"vmovd" to avoid matching "vmovdqu".
* gcc.target/i386/pr87317-5.c: Likewise.
* gcc.target/i386/pr87317-7.c: Likewise.
* gcc.target/i386/pr91461-1.c: New test.
* gcc.target/i386/pr91461-2.c: Likewise.
* gcc.target/i386/pr91461-3.c: Likewise.
* gcc.target/i386/pr91461-4.c: Likewise.
* gcc.target/i386/pr91461-5.c: Likewise.
---
gcc/config/i386/i386.md | 12 ++-
gcc/config/i386/sse.md | 9 +--
.../i386/avx256-unaligned-store-2.c | 4 +-
.../i386/avx256-unaligned-store-3.c | 4 +-
.../gcc.target/i386/pieces-memcpy-4.c | 3 +-
.../gcc.target/i386/pieces-memcpy-5.c | 3 +-
.../gcc.target/i386/pieces-memcpy-6.c | 3 +-
.../gcc.target/i386/pieces-strcpy-2.c | 2 +-
gcc/testsuite/gcc.target/i386/pr87317-4.c | 2 +-
gcc/testsuite/gcc.target/i386/pr87317-5.c | 2 +-
gcc/testsuite/gcc.target/i386/pr87317-7.c | 2 +-
gcc/testsuite/gcc.target/i386/pr90980-1.c | 2 +-
gcc/testsuite/gcc.target/i386/pr91461-1.c | 66 ++++++++++++++++
gcc/testsuite/gcc.target/i386/pr91461-2.c | 19 +++++
gcc/testsuite/gcc.target/i386/pr91461-3.c | 76 +++++++++++++++++++
gcc/testsuite/gcc.target/i386/pr91461-4.c | 21 +++++
gcc/testsuite/gcc.target/i386/pr91461-5.c | 17 +++++
17 files changed, 220 insertions(+), 27 deletions(-)
create mode 100644 gcc/testsuite/gcc.target/i386/pr91461-1.c
create mode 100644 gcc/testsuite/gcc.target/i386/pr91461-2.c
create mode 100644 gcc/testsuite/gcc.target/i386/pr91461-3.c
create mode 100644 gcc/testsuite/gcc.target/i386/pr91461-4.c
create mode 100644 gcc/testsuite/gcc.target/i386/pr91461-5.c
diff --git a/gcc/config/i386/i386.md b/gcc/config/i386/i386.md
index a125ab350bb..9f0077d59a9 100644
--- a/gcc/config/i386/i386.md
+++ b/gcc/config/i386/i386.md
@@ -1980,9 +1980,7 @@
(and (eq_attr "alternative" "1")
(match_test "TARGET_AVX512VL"))
(const_string "XI")
- (ior (match_test "TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL")
- (and (eq_attr "alternative" "3")
- (match_test "TARGET_SSE_TYPELESS_STORES")))
+ (match_test "TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL")
(const_string "V8SF")
]
(const_string "OI")))])
@@ -2059,13 +2057,13 @@
(and (eq_attr "alternative" "3")
(match_test "TARGET_AVX512VL"))
(const_string "XI")
+ (match_test "TARGET_AVX")
+ (const_string "TI")
(ior (not (match_test "TARGET_SSE2"))
(ior (match_test "TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL")
(and (eq_attr "alternative" "5")
(match_test "TARGET_SSE_TYPELESS_STORES"))))
(const_string "V4SF")
- (match_test "TARGET_AVX")
- (const_string "TI")
(match_test "optimize_function_for_size_p (cfun)")
(const_string "V4SF")
]
@@ -3324,13 +3322,13 @@
(set (attr "mode")
(cond [(eq_attr "alternative" "3,4")
(const_string "DI")
+ (match_test "TARGET_AVX")
+ (const_string "TI")
(match_test "TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL")
(const_string "V4SF")
(and (eq_attr "alternative" "2")
(match_test "TARGET_SSE_TYPELESS_STORES"))
(const_string "V4SF")
- (match_test "TARGET_AVX")
- (const_string "TI")
(ior (not (match_test "TARGET_SSE2"))
(match_test "optimize_function_for_size_p (cfun)"))
(const_string "V4SF")
diff --git a/gcc/config/i386/sse.md b/gcc/config/i386/sse.md
index 04a8c5e56b9..abbd879aab3 100644
--- a/gcc/config/i386/sse.md
+++ b/gcc/config/i386/sse.md
@@ -1116,13 +1116,12 @@
(cond [(and (eq_attr "alternative" "1")
(match_test "TARGET_AVX512VL"))
(const_string "<sseinsnmode>")
- (and (match_test "<MODE_SIZE> == 16")
- (ior (match_test "TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL")
- (and (eq_attr "alternative" "3")
- (match_test "TARGET_SSE_TYPELESS_STORES"))))
- (const_string "<ssePSmode>")
(match_test "TARGET_AVX")
(const_string "<sseinsnmode>")
+ (ior (match_test "TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL")
+ (and (eq_attr "alternative" "3")
+ (match_test "TARGET_SSE_TYPELESS_STORES")))
+ (const_string "<ssePSmode>")
(ior (not (match_test "TARGET_SSE2"))
(match_test "optimize_function_for_size_p (cfun)"))
(const_string "V4SF")
diff --git a/gcc/testsuite/gcc.target/i386/avx256-unaligned-store-2.c b/gcc/testsuite/gcc.target/i386/avx256-unaligned-store-2.c
index 1e7969bb47d..be12529e8d5 100644
--- a/gcc/testsuite/gcc.target/i386/avx256-unaligned-store-2.c
+++ b/gcc/testsuite/gcc.target/i386/avx256-unaligned-store-2.c
@@ -23,6 +23,6 @@ avx_test (void)
}
}
-/* { dg-final { scan-assembler-not "vmovups.*movv32qi_internal/3" } } */
-/* { dg-final { scan-assembler "vmovups.*movv16qi_internal/3" } } */
+/* { dg-final { scan-assembler-not "vmovdqu.*movv32qi_internal/3" } } */
+/* { dg-final { scan-assembler "vmovdqu.*movv16qi_internal/3" } } */
/* { dg-final { scan-assembler "vextract.128" } } */
diff --git a/gcc/testsuite/gcc.target/i386/avx256-unaligned-store-3.c b/gcc/testsuite/gcc.target/i386/avx256-unaligned-store-3.c
index a439a66ff34..918028df9ed 100644
--- a/gcc/testsuite/gcc.target/i386/avx256-unaligned-store-3.c
+++ b/gcc/testsuite/gcc.target/i386/avx256-unaligned-store-3.c
@@ -17,6 +17,6 @@ avx_test (void)
d[i] = c[i] * 20.0;
}
-/* { dg-final { scan-assembler-not "vmovups.*movv4df_internal/3" } } */
-/* { dg-final { scan-assembler "vmovups.*movv2df_internal/3" } } */
+/* { dg-final { scan-assembler-not "vmovupd.*movv4df_internal/3" } } */
+/* { dg-final { scan-assembler "vmovupd.*movv2df_internal/3" } } */
/* { dg-final { scan-assembler "vextractf128" } } */
diff --git a/gcc/testsuite/gcc.target/i386/pieces-memcpy-4.c b/gcc/testsuite/gcc.target/i386/pieces-memcpy-4.c
index 64e8921abe2..6f20203a146 100644
--- a/gcc/testsuite/gcc.target/i386/pieces-memcpy-4.c
+++ b/gcc/testsuite/gcc.target/i386/pieces-memcpy-4.c
@@ -9,5 +9,4 @@ foo (void)
__builtin_memcpy (dst, src, 18);
}
-/* { dg-final { scan-assembler-times "vmovdqu\[ \\t\]+\[^\n\]*%xmm" 1 } } */
-/* { dg-final { scan-assembler-times "vmovups\[ \\t\]+\[^\n\]*%xmm" 1 } } */
+/* { dg-final { scan-assembler-times "vmovdqu\[ \\t\]+\[^\n\]*%xmm" 2 } } */
diff --git a/gcc/testsuite/gcc.target/i386/pieces-memcpy-5.c b/gcc/testsuite/gcc.target/i386/pieces-memcpy-5.c
index 3c464c32f8e..5a1c7b3d512 100644
--- a/gcc/testsuite/gcc.target/i386/pieces-memcpy-5.c
+++ b/gcc/testsuite/gcc.target/i386/pieces-memcpy-5.c
@@ -9,5 +9,4 @@ foo (void)
__builtin_memcpy (dst, src, 19);
}
-/* { dg-final { scan-assembler-times "vmovdqu\[ \\t\]+\[^\n\]*%xmm" 1 } } */
-/* { dg-final { scan-assembler-times "vmovups\[ \\t\]+\[^\n\]*%xmm" 1 } } */
+/* { dg-final { scan-assembler-times "vmovdqu\[ \\t\]+\[^\n\]*%xmm" 2 } } */
diff --git a/gcc/testsuite/gcc.target/i386/pieces-memcpy-6.c b/gcc/testsuite/gcc.target/i386/pieces-memcpy-6.c
index cdb00e05bc1..5f99cc98c47 100644
--- a/gcc/testsuite/gcc.target/i386/pieces-memcpy-6.c
+++ b/gcc/testsuite/gcc.target/i386/pieces-memcpy-6.c
@@ -9,5 +9,4 @@ foo (void)
__builtin_memcpy (dst, src, 33);
}
-/* { dg-final { scan-assembler-times "vmovdqu\[ \\t\]+\[^\n\]*%xmm" 2 } } */
-/* { dg-final { scan-assembler-times "vmovups\[ \\t\]+\[^\n\]*%xmm" 2 } } */
+/* { dg-final { scan-assembler-times "vmovdqu\[ \\t\]+\[^\n\]*%xmm" 4 } } */
diff --git a/gcc/testsuite/gcc.target/i386/pieces-strcpy-2.c b/gcc/testsuite/gcc.target/i386/pieces-strcpy-2.c
index 74212559508..90446edb4f3 100644
--- a/gcc/testsuite/gcc.target/i386/pieces-strcpy-2.c
+++ b/gcc/testsuite/gcc.target/i386/pieces-strcpy-2.c
@@ -12,4 +12,4 @@ foo (char *s)
}
/* { dg-final { scan-assembler-times "vmovdqa\[ \\t\]+\[^\n\]*%xmm" 4 } } */
-/* { dg-final { scan-assembler-times "vmovups\[ \\t\]+\[^\n\]*%xmm" 4 } } */
+/* { dg-final { scan-assembler-times "vmovdqu\[ \\t\]+\[^\n\]*%xmm" 4 } } */
diff --git a/gcc/testsuite/gcc.target/i386/pr87317-4.c b/gcc/testsuite/gcc.target/i386/pr87317-4.c
index 2d4f24a89e9..d802575f4c5 100644
--- a/gcc/testsuite/gcc.target/i386/pr87317-4.c
+++ b/gcc/testsuite/gcc.target/i386/pr87317-4.c
@@ -1,7 +1,7 @@
/* { dg-do compile } */
/* { dg-options "-O2 -march=haswell" } */
/* { dg-final { scan-assembler-times "vpmovzxbd" 1 } } */
-/* { dg-final { scan-assembler-not "vmovd" } } */
+/* { dg-final { scan-assembler-not "\tvmovd\t" } } */
#include <immintrin.h>
diff --git a/gcc/testsuite/gcc.target/i386/pr87317-5.c b/gcc/testsuite/gcc.target/i386/pr87317-5.c
index 96f82847e5d..42cf7dc0ffe 100644
--- a/gcc/testsuite/gcc.target/i386/pr87317-5.c
+++ b/gcc/testsuite/gcc.target/i386/pr87317-5.c
@@ -1,7 +1,7 @@
/* { dg-do compile } */
/* { dg-options "-O2 -march=haswell" } */
/* { dg-final { scan-assembler-times "vpmovzxwq" 1 } } */
-/* { dg-final { scan-assembler-not "vmovd" } } */
+/* { dg-final { scan-assembler-not "\tvmovd\t" } } */
#include <immintrin.h>
diff --git a/gcc/testsuite/gcc.target/i386/pr87317-7.c b/gcc/testsuite/gcc.target/i386/pr87317-7.c
index 2c043d9eb26..c76af7efd5f 100644
--- a/gcc/testsuite/gcc.target/i386/pr87317-7.c
+++ b/gcc/testsuite/gcc.target/i386/pr87317-7.c
@@ -1,7 +1,7 @@
/* { dg-do compile { target { ! ia32 } } } */
/* { dg-options "-O2 -march=haswell" } */
/* { dg-final { scan-assembler-times "vpmovzxbd" 1 } } */
-/* { dg-final { scan-assembler-not "vmovd" } } */
+/* { dg-final { scan-assembler-not "\tvmovd\t" } } */
#include <immintrin.h>
diff --git a/gcc/testsuite/gcc.target/i386/pr90980-1.c b/gcc/testsuite/gcc.target/i386/pr90980-1.c
index 72a30dc8da2..885518984c5 100644
--- a/gcc/testsuite/gcc.target/i386/pr90980-1.c
+++ b/gcc/testsuite/gcc.target/i386/pr90980-1.c
@@ -1,6 +1,6 @@
/* { dg-do compile } */
/* { dg-options "-march=skylake-avx512 -O2" } */
-/* { dg-final { scan-assembler-times "(?:vmovups|vmovdqu)\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\]*\\)(?:\n|\[ \\t\]+#)" 2 } } */
+/* { dg-final { scan-assembler-times "vmovdqu\[2346\]*\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\]*\\)(?:\n|\[ \\t\]+#)" 2 } } */
#include <immintrin.h>
diff --git a/gcc/testsuite/gcc.target/i386/pr91461-1.c b/gcc/testsuite/gcc.target/i386/pr91461-1.c
new file mode 100644
index 00000000000..0c94b8e2b76
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr91461-1.c
@@ -0,0 +1,66 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -mavx" } */
+/* { dg-final { scan-assembler "\tvmovdqa\t" } } */
+/* { dg-final { scan-assembler "\tvmovdqu\t" } } */
+/* { dg-final { scan-assembler "\tvmovapd\t" } } */
+/* { dg-final { scan-assembler "\tvmovupd\t" } } */
+/* { dg-final { scan-assembler-not "\tvmovaps\t" } } */
+/* { dg-final { scan-assembler-not "\tvmovups\t" } } */
+
+#include <immintrin.h>
+
+void
+foo1 (__m128i *p, __m128i x)
+{
+ *p = x;
+}
+
+void
+foo2 (__m128d *p, __m128d x)
+{
+ *p = x;
+}
+
+void
+foo3 (__float128 *p, __float128 x)
+{
+ *p = x;
+}
+
+void
+foo4 (__m128i_u *p, __m128i x)
+{
+ *p = x;
+}
+
+void
+foo5 (__m128d_u *p, __m128d x)
+{
+ *p = x;
+}
+
+typedef __float128 __float128_u __attribute__ ((__aligned__ (1)));
+
+void
+foo6 (__float128_u *p, __float128 x)
+{
+ *p = x;
+}
+
+#ifdef __x86_64__
+typedef __int128 __int128_u __attribute__ ((__aligned__ (1)));
+
+extern __int128 int128;
+
+void
+foo7 (__int128 *p)
+{
+ *p = int128;
+}
+
+void
+foo8 (__int128_u *p)
+{
+ *p = int128;
+}
+#endif
diff --git a/gcc/testsuite/gcc.target/i386/pr91461-2.c b/gcc/testsuite/gcc.target/i386/pr91461-2.c
new file mode 100644
index 00000000000..921cfaf9780
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr91461-2.c
@@ -0,0 +1,19 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -mavx" } */
+/* { dg-final { scan-assembler "\tvmovdqa\t" } } */
+/* { dg-final { scan-assembler "\tvmovapd\t" } } */
+/* { dg-final { scan-assembler-not "\tvmovaps\t" } } */
+
+#include <immintrin.h>
+
+void
+foo1 (__m256i *p, __m256i x)
+{
+ *p = x;
+}
+
+void
+foo2 (__m256d *p, __m256d x)
+{
+ *p = x;
+}
diff --git a/gcc/testsuite/gcc.target/i386/pr91461-3.c b/gcc/testsuite/gcc.target/i386/pr91461-3.c
new file mode 100644
index 00000000000..c67a48063bf
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr91461-3.c
@@ -0,0 +1,76 @@
+/* { dg-do compile { target { ! ia32 } } } */
+/* { dg-options "-O2 -mavx512f -mavx512vl" } */
+/* { dg-final { scan-assembler-not "\tvmovaps\t" } } */
+/* { dg-final { scan-assembler-not "\tvmovups\t" } } */
+
+#include <immintrin.h>
+
+void
+foo1 (__m128i *p, __m128i a)
+{
+ register __m128i x __asm ("xmm16") = a;
+ asm volatile ("" : "+v" (x));
+ *p = x;
+}
+
+void
+foo2 (__m128d *p, __m128d a)
+{
+ register __m128d x __asm ("xmm16") = a;
+ asm volatile ("" : "+v" (x));
+ *p = x;
+}
+
+void
+foo3 (__float128 *p, __float128 a)
+{
+ register __float128 x __asm ("xmm16") = a;
+ asm volatile ("" : "+v" (x));
+ *p = x;
+}
+
+void
+foo4 (__m128i_u *p, __m128i a)
+{
+ register __m128i x __asm ("xmm16") = a;
+ asm volatile ("" : "+v" (x));
+ *p = x;
+}
+
+void
+foo5 (__m128d_u *p, __m128d a)
+{
+ register __m128d x __asm ("xmm16") = a;
+ asm volatile ("" : "+v" (x));
+ *p = x;
+}
+
+typedef __float128 __float128_u __attribute__ ((__aligned__ (1)));
+
+void
+foo6 (__float128_u *p, __float128 a)
+{
+ register __float128 x __asm ("xmm16") = a;
+ asm volatile ("" : "+v" (x));
+ *p = x;
+}
+
+typedef __int128 __int128_u __attribute__ ((__aligned__ (1)));
+
+extern __int128 int128;
+
+void
+foo7 (__int128 *p)
+{
+ register __int128 x __asm ("xmm16") = int128;
+ asm volatile ("" : "+v" (x));
+ *p = x;
+}
+
+void
+foo8 (__int128_u *p)
+{
+ register __int128 x __asm ("xmm16") = int128;
+ asm volatile ("" : "+v" (x));
+ *p = x;
+}
diff --git a/gcc/testsuite/gcc.target/i386/pr91461-4.c b/gcc/testsuite/gcc.target/i386/pr91461-4.c
new file mode 100644
index 00000000000..69df590de3a
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr91461-4.c
@@ -0,0 +1,21 @@
+/* { dg-do compile { target { ! ia32 } } } */
+/* { dg-options "-O2 -mavx512f -mavx512vl" } */
+/* { dg-final { scan-assembler-not "\tvmovaps\t" } } */
+
+#include <immintrin.h>
+
+void
+foo1 (__m256i *p, __m256i a)
+{
+ register __m256i x __asm ("xmm16") = a;
+ asm volatile ("" : "+v" (x));
+ *p = x;
+}
+
+void
+foo2 (__m256d *p, __m256d a)
+{
+ register __m256d x __asm ("xmm16") = a;
+ asm volatile ("" : "+v" (x));
+ *p = x;
+}
diff --git a/gcc/testsuite/gcc.target/i386/pr91461-5.c b/gcc/testsuite/gcc.target/i386/pr91461-5.c
new file mode 100644
index 00000000000..974263042f3
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr91461-5.c
@@ -0,0 +1,17 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -mavx512f" } */
+/* { dg-final { scan-assembler-not "\tvmovaps\t" } } */
+
+#include <immintrin.h>
+
+void
+foo1 (__m512i *p, __m512i x)
+{
+ *p = x;
+}
+
+void
+foo2 (__m512d *p, __m512d x)
+{
+ *p = x;
+}
--
2.24.1
