On Thu, Jul 23, 2026 at 11:06 AM Hu, Lin1 <[email protected]> wrote:
>
> Change v1 -> v2.
> Provide _si{32,64} aliases for the non-round scalar intrinsics.
>
> -------------------------------------------
>
> The previous _ep[i|u]{32,64} suffix for scalar SAT CVT intrinsics does
> not match the intrinsic naming convention: scalar convert intrinsics
> have used _[i|u]{32,64} (with _si{32,64} aliases for signed variants)
> since AVX-512F, so the AVX10.2 scalar SAT CVT names need to follow the
> same convention.
>
> This patch renames the scalar SAT CVT intrinsics to use _[i|u]{32,64}
> suffix and provides _si{32,64} aliases for the signed intrinsics.
Ok.
>
> gcc/ChangeLog:
>
> * config/i386/avx10_2satcvtintrin.h
> (_mm_cvtts_sd_i32): Renamed from _mm_cvtts_sd_epi32.
> (_mm_cvtts_sd_si32): New alias of _mm_cvtts_sd_i32.
> (_mm_cvtts_sd_u32): Renamed from _mm_cvtts_sd_epu32.
> (_mm_cvtts_ss_i32): Renamed from _mm_cvtts_ss_epi32.
> (_mm_cvtts_ss_si32): New alias of _mm_cvtts_ss_i32.
> (_mm_cvtts_ss_u32): Renamed from _mm_cvtts_ss_epu32.
> (_mm_cvtts_roundsd_i32): Renamed from _mm_cvtts_roundsd_epi32.
> (_mm_cvtts_roundsd_si32): New alias of _mm_cvtts_roundsd_i32.
> (_mm_cvtts_roundsd_u32): Renamed from _mm_cvtts_roundsd_epu32.
> (_mm_cvtts_roundss_i32): Renamed from _mm_cvtts_roundss_epi32.
> (_mm_cvtts_roundss_si32): New alias of _mm_cvtts_roundss_i32.
> (_mm_cvtts_roundss_u32): Renamed from _mm_cvtts_roundss_epu32.
> (_mm_cvtts_sd_i64): Renamed from _mm_cvtts_sd_epi64.
> (_mm_cvtts_sd_si64): New alias of _mm_cvtts_sd_i64.
> (_mm_cvtts_sd_u64): Renamed from _mm_cvtts_sd_epu64.
> (_mm_cvtts_ss_i64): Renamed from _mm_cvtts_ss_epi64.
> (_mm_cvtts_ss_si64): New alias of _mm_cvtts_ss_i64.
> (_mm_cvtts_ss_u64): Renamed from _mm_cvtts_ss_epu64.
> (_mm_cvtts_roundsd_i64): Renamed from _mm_cvtts_roundsd_epi64.
> (_mm_cvtts_roundsd_si64): New alias of _mm_cvtts_roundsd_i64.
> (_mm_cvtts_roundsd_u64): Renamed from _mm_cvtts_roundsd_epu64.
> (_mm_cvtts_roundss_i64): Renamed from _mm_cvtts_roundss_epi64.
> (_mm_cvtts_roundss_si64): New alias of _mm_cvtts_roundss_i64.
> (_mm_cvtts_roundss_u64): Renamed from _mm_cvtts_roundss_epu64.
>
> gcc/testsuite/ChangeLog:
>
> * gcc.target/i386/avx10_2-satcvt-1.c: Update intrin names and
> add _si alias tests.
> * gcc.target/i386/sse-14.c: Ditto.
> * gcc.target/i386/sse-22.c: Ditto.
> * gcc.target/i386/avx10_2-vcvttsd2sis-2.c: Update intrin names.
> * gcc.target/i386/avx10_2-vcvttsd2usis-2.c: Ditto.
> * gcc.target/i386/avx10_2-vcvttss2sis-2.c: Ditto.
> * gcc.target/i386/avx10_2-vcvttss2usis-2.c: Ditto.
> * gcc.target/i386/zext-avx10-1.c: Ditto.
> ---
> gcc/config/i386/avx10_2satcvtintrin.h | 127 ++++++++++++++----
> .../gcc.target/i386/avx10_2-satcvt-1.c | 56 ++++----
> .../gcc.target/i386/avx10_2-vcvttsd2sis-2.c | 8 +-
> .../gcc.target/i386/avx10_2-vcvttsd2usis-2.c | 8 +-
> .../gcc.target/i386/avx10_2-vcvttss2sis-2.c | 8 +-
> .../gcc.target/i386/avx10_2-vcvttss2usis-2.c | 8 +-
> gcc/testsuite/gcc.target/i386/sse-14.c | 20 +--
> gcc/testsuite/gcc.target/i386/sse-22.c | 20 +--
> gcc/testsuite/gcc.target/i386/zext-avx10-1.c | 16 +--
> 9 files changed, 183 insertions(+), 88 deletions(-)
>
> diff --git a/gcc/config/i386/avx10_2satcvtintrin.h
> b/gcc/config/i386/avx10_2satcvtintrin.h
> index 3b47d5491ab..f202ec862c9 100644
> --- a/gcc/config/i386/avx10_2satcvtintrin.h
> +++ b/gcc/config/i386/avx10_2satcvtintrin.h
> @@ -2769,7 +2769,15 @@ _mm512_maskz_cvtts_roundps_epu64 (__mmask8 __U, __m256
> __A, const int __R)
>
> extern __inline int
> __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
> -_mm_cvtts_sd_epi32 (__m128d __A)
> +_mm_cvtts_sd_i32 (__m128d __A)
> +{
> + return (int) __builtin_ia32_cvttsd2sis32_round ((__v2df) __A,
> + _MM_FROUND_CUR_DIRECTION);
> +}
> +
> +extern __inline int
> +__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
> +_mm_cvtts_sd_si32 (__m128d __A)
> {
> return (int) __builtin_ia32_cvttsd2sis32_round ((__v2df) __A,
> _MM_FROUND_CUR_DIRECTION);
> @@ -2777,7 +2785,7 @@ _mm_cvtts_sd_epi32 (__m128d __A)
>
> extern __inline unsigned int
> __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
> -_mm_cvtts_sd_epu32 (__m128d __A)
> +_mm_cvtts_sd_u32 (__m128d __A)
> {
> return (unsigned int) __builtin_ia32_cvttsd2usis32_round ((__v2df) __A,
>
> _MM_FROUND_CUR_DIRECTION);
> @@ -2785,7 +2793,15 @@ _mm_cvtts_sd_epu32 (__m128d __A)
>
> extern __inline int
> __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
> -_mm_cvtts_ss_epi32 (__m128 __A)
> +_mm_cvtts_ss_i32 (__m128 __A)
> +{
> + return (int) __builtin_ia32_cvttss2sis32_round ((__v4sf) __A,
> + _MM_FROUND_CUR_DIRECTION);
> +}
> +
> +extern __inline int
> +__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
> +_mm_cvtts_ss_si32 (__m128 __A)
> {
> return (int) __builtin_ia32_cvttss2sis32_round ((__v4sf) __A,
> _MM_FROUND_CUR_DIRECTION);
> @@ -2793,7 +2809,7 @@ _mm_cvtts_ss_epi32 (__m128 __A)
>
> extern __inline unsigned int
> __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
> -_mm_cvtts_ss_epu32 (__m128 __A)
> +_mm_cvtts_ss_u32 (__m128 __A)
> {
> return (unsigned int) __builtin_ia32_cvttss2usis32_round ((__v4sf) __A,
>
> _MM_FROUND_CUR_DIRECTION);
> @@ -2802,7 +2818,15 @@ _mm_cvtts_ss_epu32 (__m128 __A)
> #ifdef __OPTIMIZE__
> extern __inline int
> __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
> -_mm_cvtts_roundsd_epi32 (__m128d __A, const int __R)
> +_mm_cvtts_roundsd_i32 (__m128d __A, const int __R)
> +{
> + return (int) __builtin_ia32_cvttsd2sis32_round ((__v2df) __A,
> + __R);
> +}
> +
> +extern __inline int
> +__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
> +_mm_cvtts_roundsd_si32 (__m128d __A, const int __R)
> {
> return (int) __builtin_ia32_cvttsd2sis32_round ((__v2df) __A,
> __R);
> @@ -2810,7 +2834,7 @@ _mm_cvtts_roundsd_epi32 (__m128d __A, const int __R)
>
> extern __inline unsigned int
> __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
> -_mm_cvtts_roundsd_epu32 (__m128d __A, const int __R)
> +_mm_cvtts_roundsd_u32 (__m128d __A, const int __R)
> {
> return (unsigned int) __builtin_ia32_cvttsd2usis32_round ((__v2df) __A,
> __R);
> @@ -2818,7 +2842,15 @@ _mm_cvtts_roundsd_epu32 (__m128d __A, const int __R)
>
> extern __inline int
> __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
> -_mm_cvtts_roundss_epi32 (__m128 __A, const int __R)
> +_mm_cvtts_roundss_i32 (__m128 __A, const int __R)
> +{
> + return (int) __builtin_ia32_cvttss2sis32_round ((__v4sf) __A,
> + __R);
> +}
> +
> +extern __inline int
> +__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
> +_mm_cvtts_roundss_si32 (__m128 __A, const int __R)
> {
> return (int) __builtin_ia32_cvttss2sis32_round ((__v4sf) __A,
> __R);
> @@ -2826,25 +2858,33 @@ _mm_cvtts_roundss_epi32 (__m128 __A, const int __R)
>
> extern __inline unsigned int
> __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
> -_mm_cvtts_roundss_epu32 (__m128 __A, const int __R)
> +_mm_cvtts_roundss_u32 (__m128 __A, const int __R)
> {
> return (unsigned int) __builtin_ia32_cvttss2usis32_round ((__v4sf) __A,
> __R);
> }
> #else
> -#define _mm_cvtts_roundsd_epi32(A, R) \
> +#define _mm_cvtts_roundsd_i32(A, R) \
> + ((int) __builtin_ia32_cvttsd2sis32_round ((__v2df) (A), \
> + (R)))
> +
> +#define _mm_cvtts_roundsd_si32(A, R) \
> ((int) __builtin_ia32_cvttsd2sis32_round ((__v2df) (A), \
> (R)))
>
> -#define _mm_cvtts_roundsd_epu32(A, R) \
> +#define _mm_cvtts_roundsd_u32(A, R) \
> ((unsigned int) __builtin_ia32_cvttsd2usis32_round ((__v2df) (A), \
> (R)))
>
> -#define _mm_cvtts_roundss_epi32(A, R) \
> +#define _mm_cvtts_roundss_i32(A, R) \
> + ((int) __builtin_ia32_cvttss2sis32_round ((__v4sf) (A), \
> + (R)))
> +
> +#define _mm_cvtts_roundss_si32(A, R) \
> ((int) __builtin_ia32_cvttss2sis32_round ((__v4sf) (A), \
> (R)))
>
> -#define _mm_cvtts_roundss_epu32(A, R) \
> +#define _mm_cvtts_roundss_u32(A, R) \
> ((unsigned int) __builtin_ia32_cvttss2usis32_round ((__v4sf) (A), \
> (R)))
> #endif
> @@ -2852,7 +2892,15 @@ _mm_cvtts_roundss_epu32 (__m128 __A, const int __R)
> #ifdef __x86_64__
> extern __inline long long
> __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
> -_mm_cvtts_sd_epi64 (__m128d __A)
> +_mm_cvtts_sd_i64 (__m128d __A)
> +{
> + return (long long) __builtin_ia32_cvttsd2sis64_round ((__v2df) __A,
> +
> _MM_FROUND_CUR_DIRECTION);
> +}
> +
> +extern __inline long long
> +__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
> +_mm_cvtts_sd_si64 (__m128d __A)
> {
> return (long long) __builtin_ia32_cvttsd2sis64_round ((__v2df) __A,
>
> _MM_FROUND_CUR_DIRECTION);
> @@ -2860,7 +2908,7 @@ _mm_cvtts_sd_epi64 (__m128d __A)
>
> extern __inline unsigned long long
> __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
> -_mm_cvtts_sd_epu64 (__m128d __A)
> +_mm_cvtts_sd_u64 (__m128d __A)
> {
> return (unsigned long long) __builtin_ia32_cvttsd2usis64_round ((__v2df)
> __A,
>
> _MM_FROUND_CUR_DIRECTION);
> @@ -2868,16 +2916,23 @@ _mm_cvtts_sd_epu64 (__m128d __A)
>
> extern __inline long long
> __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
> -_mm_cvtts_ss_epi64 (__m128 __A)
> +_mm_cvtts_ss_i64 (__m128 __A)
> {
> return (long long) __builtin_ia32_cvttss2sis64_round ((__v4sf) __A,
>
> _MM_FROUND_CUR_DIRECTION);
> }
>
> +extern __inline long long
> +__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
> +_mm_cvtts_ss_si64 (__m128 __A)
> +{
> + return (long long) __builtin_ia32_cvttss2sis64_round ((__v4sf) __A,
> +
> _MM_FROUND_CUR_DIRECTION);
> +}
>
> extern __inline unsigned long long
> __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
> -_mm_cvtts_ss_epu64 (__m128 __A)
> +_mm_cvtts_ss_u64 (__m128 __A)
> {
> return (unsigned long long) __builtin_ia32_cvttss2usis64_round ((__v4sf)
> __A,
>
> _MM_FROUND_CUR_DIRECTION);
> @@ -2886,7 +2941,15 @@ _mm_cvtts_ss_epu64 (__m128 __A)
> #ifdef __OPTIMIZE__
> extern __inline long long
> __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
> -_mm_cvtts_roundsd_epi64 (__m128d __A, const int __R)
> +_mm_cvtts_roundsd_i64 (__m128d __A, const int __R)
> +{
> + return (long long) __builtin_ia32_cvttsd2sis64_round ((__v2df) __A,
> + __R);
> +}
> +
> +extern __inline long long
> +__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
> +_mm_cvtts_roundsd_si64 (__m128d __A, const int __R)
> {
> return (long long) __builtin_ia32_cvttsd2sis64_round ((__v2df) __A,
> __R);
> @@ -2894,7 +2957,7 @@ _mm_cvtts_roundsd_epi64 (__m128d __A, const int __R)
>
> extern __inline unsigned long long
> __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
> -_mm_cvtts_roundsd_epu64 (__m128d __A, const int __R)
> +_mm_cvtts_roundsd_u64 (__m128d __A, const int __R)
> {
> return (unsigned long long) __builtin_ia32_cvttsd2usis64_round ((__v2df)
> __A,
> __R);
> @@ -2902,7 +2965,15 @@ _mm_cvtts_roundsd_epu64 (__m128d __A, const int __R)
>
> extern __inline long long
> __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
> -_mm_cvtts_roundss_epi64 (__m128 __A, const int __R)
> +_mm_cvtts_roundss_i64 (__m128 __A, const int __R)
> +{
> + return (long long) __builtin_ia32_cvttss2sis64_round ((__v4sf) __A,
> + __R);
> +}
> +
> +extern __inline long long
> +__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
> +_mm_cvtts_roundss_si64 (__m128 __A, const int __R)
> {
> return (long long) __builtin_ia32_cvttss2sis64_round ((__v4sf) __A,
> __R);
> @@ -2910,25 +2981,33 @@ _mm_cvtts_roundss_epi64 (__m128 __A, const int __R)
>
> extern __inline unsigned long long
> __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
> -_mm_cvtts_roundss_epu64 (__m128 __A, const int __R)
> +_mm_cvtts_roundss_u64 (__m128 __A, const int __R)
> {
> return (unsigned long long) __builtin_ia32_cvttss2usis64_round ((__v4sf)
> __A,
> __R);
> }
> #else
> -#define _mm_cvtts_roundsd_epi64(A, R) \
> +#define _mm_cvtts_roundsd_i64(A, R) \
> + ((long long) __builtin_ia32_cvttsd2sis64_round ((__v2df) (A), \
> + (R)))
> +
> +#define _mm_cvtts_roundsd_si64(A, R) \
> ((long long) __builtin_ia32_cvttsd2sis64_round ((__v2df) (A), \
> (R)))
>
> -#define _mm_cvtts_roundsd_epu64(A, R) \
> +#define _mm_cvtts_roundsd_u64(A, R) \
> ((unsigned long long) __builtin_ia32_cvttsd2usis64_round ((__v2df) (A), \
> (R)))
>
> -#define _mm_cvtts_roundss_epi64(A, R) \
> +#define _mm_cvtts_roundss_i64(A, R) \
> + ((long long) __builtin_ia32_cvttss2sis64_round ((__v4sf) (A), \
> + (R)))
> +
> +#define _mm_cvtts_roundss_si64(A, R) \
> ((long long) __builtin_ia32_cvttss2sis64_round ((__v4sf) (A), \
> (R)))
>
> -#define _mm_cvtts_roundss_epu64(A, R) \
> +#define _mm_cvtts_roundss_u64(A, R) \
> ((unsigned long long) __builtin_ia32_cvttss2usis64_round ((__v4sf) (A), \
> (R)))
> #endif
> diff --git a/gcc/testsuite/gcc.target/i386/avx10_2-satcvt-1.c
> b/gcc/testsuite/gcc.target/i386/avx10_2-satcvt-1.c
> index 9f3acb066d7..717ccd2b5ae 100644
> --- a/gcc/testsuite/gcc.target/i386/avx10_2-satcvt-1.c
> +++ b/gcc/testsuite/gcc.target/i386/avx10_2-satcvt-1.c
> @@ -220,21 +220,21 @@
> /* { dg-final { scan-assembler-times "vcvttps2uqqs\[
> \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */
> /* { dg-final { scan-assembler-times "vcvttps2uqqs\[
> \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\{%k\[0-9\]\}(?:\n|\[
> \\t\]+#)" 1 } } */
> /* { dg-final { scan-assembler-times "vcvttps2uqqs\[
> \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\{%k\[0-9\]\}\{z\}(?:\n|\[
> \\t\]+#)" 1 } } */
> -/* { dg-final { scan-assembler-times "vcvttsd2sis\[
> \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%e.x+(?:\n|\[ \\t\]+#)" 1 } } */
> +/* { dg-final { scan-assembler-times "vcvttsd2sis\[
> \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%e.x+(?:\n|\[ \\t\]+#)" 2 } } */
> /* { dg-final { scan-assembler-times "vcvttsd2usis\[
> \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%e.x+(?:\n|\[ \\t\]+#)" 1 } } */
> -/* { dg-final { scan-assembler-times "vcvttss2sis\[
> \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%e.x+(?:\n|\[ \\t\]+#)" 1 } } */
> +/* { dg-final { scan-assembler-times "vcvttss2sis\[
> \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%e.x+(?:\n|\[ \\t\]+#)" 2 } } */
> /* { dg-final { scan-assembler-times "vcvttss2usis\[
> \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%e.x+(?:\n|\[ \\t\]+#)" 1 } } */
> -/* { dg-final { scan-assembler-times "vcvttsd2sis\[
> \\t\]+\{sae\}\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%e.x+(?:\n|\[ \\t\]+#)" 1 } } */
> +/* { dg-final { scan-assembler-times "vcvttsd2sis\[
> \\t\]+\{sae\}\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%e.x+(?:\n|\[ \\t\]+#)" 2 } } */
> /* { dg-final { scan-assembler-times "vcvttsd2usis\[
> \\t\]+\{sae\}\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%e.x+(?:\n|\[ \\t\]+#)" 1 } } */
> -/* { dg-final { scan-assembler-times "vcvttss2sis\[
> \\t\]+\{sae\}\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%e.x+(?:\n|\[ \\t\]+#)" 1 } } */
> +/* { dg-final { scan-assembler-times "vcvttss2sis\[
> \\t\]+\{sae\}\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%e.x+(?:\n|\[ \\t\]+#)" 2 } } */
> /* { dg-final { scan-assembler-times "vcvttss2usis\[
> \\t\]+\{sae\}\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%e.x+(?:\n|\[ \\t\]+#)" 1 } } */
> -/* { dg-final { scan-assembler-times "vcvttsd2sis\[
> \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%r.x+(?:\n|\[ \\t\]+#)" 1 { target { !
> ia32 } } } } */
> +/* { dg-final { scan-assembler-times "vcvttsd2sis\[
> \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%r.x+(?:\n|\[ \\t\]+#)" 2 { target { !
> ia32 } } } } */
> /* { dg-final { scan-assembler-times "vcvttsd2usis\[
> \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%r.x+(?:\n|\[ \\t\]+#)" 1 { target { !
> ia32 } } } } */
> -/* { dg-final { scan-assembler-times "vcvttss2sis\[
> \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%r.x+(?:\n|\[ \\t\]+#)" 1 { target { !
> ia32 } } } } */
> +/* { dg-final { scan-assembler-times "vcvttss2sis\[
> \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%r.x+(?:\n|\[ \\t\]+#)" 2 { target { !
> ia32 } } } } */
> /* { dg-final { scan-assembler-times "vcvttss2usis\[
> \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%r.x+(?:\n|\[ \\t\]+#)" 1 { target { !
> ia32 } } } } */
> -/* { dg-final { scan-assembler-times "vcvttsd2sis\[
> \\t\]+\{sae\}\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%r.x+(?:\n|\[ \\t\]+#)" 1 {
> target { ! ia32 } } } } */
> +/* { dg-final { scan-assembler-times "vcvttsd2sis\[
> \\t\]+\{sae\}\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%r.x+(?:\n|\[ \\t\]+#)" 2 {
> target { ! ia32 } } } } */
> /* { dg-final { scan-assembler-times "vcvttsd2usis\[
> \\t\]+\{sae\}\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%r.x+(?:\n|\[ \\t\]+#)" 1 {
> target { ! ia32 } } } } */
> -/* { dg-final { scan-assembler-times "vcvttss2sis\[
> \\t\]+\{sae\}\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%r.x+(?:\n|\[ \\t\]+#)" 1 {
> target { ! ia32 } } } } */
> +/* { dg-final { scan-assembler-times "vcvttss2sis\[
> \\t\]+\{sae\}\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%r.x+(?:\n|\[ \\t\]+#)" 2 {
> target { ! ia32 } } } } */
> /* { dg-final { scan-assembler-times "vcvttss2usis\[
> \\t\]+\{sae\}\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%r.x+(?:\n|\[ \\t\]+#)" 1 {
> target { ! ia32 } } } } */
>
> #include <immintrin.h>
> @@ -553,23 +553,31 @@ avx10_2_test (void)
> hxi = _mm_mask_cvtts_ps_epu64 (hxi, m8, hx);
> hxi = _mm_maskz_cvtts_ps_epu64 (m8, hx);
>
> - i = _mm_cvtts_sd_epi32 (hxd);
> - ui = _mm_cvtts_sd_epu32 (hxd);
> - i = _mm_cvtts_ss_epi32 (hx);
> - ui = _mm_cvtts_ss_epu32 (hx);
> - i = _mm_cvtts_roundsd_epi32 (hxd, 8);
> - ui = _mm_cvtts_roundsd_epu32 (hxd, 8);
> - i = _mm_cvtts_roundss_epi32 (hx, 8);
> - ui = _mm_cvtts_roundss_epu32 (hx, 8);
> + i = _mm_cvtts_sd_i32 (hxd);
> + i = _mm_cvtts_sd_si32 (hxd);
> + ui = _mm_cvtts_sd_u32 (hxd);
> + i = _mm_cvtts_ss_i32 (hx);
> + i = _mm_cvtts_ss_si32 (hx);
> + ui = _mm_cvtts_ss_u32 (hx);
> + i = _mm_cvtts_roundsd_i32 (hxd, 8);
> + i = _mm_cvtts_roundsd_si32 (hxd, 8);
> + ui = _mm_cvtts_roundsd_u32 (hxd, 8);
> + i = _mm_cvtts_roundss_i32 (hx, 8);
> + i = _mm_cvtts_roundss_si32 (hx, 8);
> + ui = _mm_cvtts_roundss_u32 (hx, 8);
>
> #ifdef __x86_64__
> - ll = _mm_cvtts_sd_epi64 (hxd);
> - ull = _mm_cvtts_sd_epu64 (hxd);
> - ll = _mm_cvtts_ss_epi64 (hx);
> - ull = _mm_cvtts_ss_epu64 (hx);
> - ll = _mm_cvtts_roundsd_epi64 (hxd, 8);
> - ull = _mm_cvtts_roundsd_epu64 (hxd, 8);
> - ll = _mm_cvtts_roundss_epi64 (hx, 8);
> - ull = _mm_cvtts_roundss_epu64 (hx, 8);
> + ll = _mm_cvtts_sd_i64 (hxd);
> + ll = _mm_cvtts_sd_si64 (hxd);
> + ull = _mm_cvtts_sd_u64 (hxd);
> + ll = _mm_cvtts_ss_i64 (hx);
> + ll = _mm_cvtts_ss_si64 (hx);
> + ull = _mm_cvtts_ss_u64 (hx);
> + ll = _mm_cvtts_roundsd_i64 (hxd, 8);
> + ll = _mm_cvtts_roundsd_si64 (hxd, 8);
> + ull = _mm_cvtts_roundsd_u64 (hxd, 8);
> + ll = _mm_cvtts_roundss_i64 (hx, 8);
> + ll = _mm_cvtts_roundss_si64 (hx, 8);
> + ull = _mm_cvtts_roundss_u64 (hx, 8);
> #endif
> }
> diff --git a/gcc/testsuite/gcc.target/i386/avx10_2-vcvttsd2sis-2.c
> b/gcc/testsuite/gcc.target/i386/avx10_2-vcvttsd2sis-2.c
> index 8069497d1e6..11c7c89437d 100644
> --- a/gcc/testsuite/gcc.target/i386/avx10_2-vcvttsd2sis-2.c
> +++ b/gcc/testsuite/gcc.target/i386/avx10_2-vcvttsd2sis-2.c
> @@ -19,7 +19,7 @@ TEST (void)
>
> s.a[0] = 2.46;
>
> - res1 = _mm_cvtts_roundsd_epi32 (s.x, 8);
> + res1 = _mm_cvtts_roundsd_i32 (s.x, 8);
>
> if (s.a[0] > INT_MAX)
> res1_ref = INT_MAX;
> @@ -31,7 +31,7 @@ TEST (void)
> if (res1 != res1_ref)
> abort();
>
> - res1 = _mm_cvtts_sd_epi32 (s.x);
> + res1 = _mm_cvtts_sd_i32 (s.x);
>
> if (s.a[0] > INT_MAX)
> res1_ref = INT_MAX;
> @@ -44,7 +44,7 @@ TEST (void)
> abort();
>
> #ifdef __x86_64__
> - res2 = _mm_cvtts_roundsd_epi64 (s.x, 8);
> + res2 = _mm_cvtts_roundsd_i64 (s.x, 8);
>
> if (s.a[0] > LLONG_MAX)
> res2_ref = LLONG_MAX;
> @@ -56,7 +56,7 @@ TEST (void)
> if (res2 != res2_ref)
> abort();
>
> - res2 = _mm_cvtts_sd_epi64 (s.x);
> + res2 = _mm_cvtts_sd_i64 (s.x);
>
> if (s.a[0] > LLONG_MAX)
> res2_ref = LLONG_MAX;
> diff --git a/gcc/testsuite/gcc.target/i386/avx10_2-vcvttsd2usis-2.c
> b/gcc/testsuite/gcc.target/i386/avx10_2-vcvttsd2usis-2.c
> index e2b18d6bb00..615428860eb 100644
> --- a/gcc/testsuite/gcc.target/i386/avx10_2-vcvttsd2usis-2.c
> +++ b/gcc/testsuite/gcc.target/i386/avx10_2-vcvttsd2usis-2.c
> @@ -19,7 +19,7 @@ TEST (void)
>
> s.a[0] = 2.46;
>
> - res1 = _mm_cvtts_roundsd_epu32 (s.x, 8);
> + res1 = _mm_cvtts_roundsd_u32 (s.x, 8);
>
> if (s.a[0] > UINT_MAX)
> res1_ref = UINT_MAX;
> @@ -31,7 +31,7 @@ TEST (void)
> if (res1 != res1_ref)
> abort();
>
> - res1 = _mm_cvtts_sd_epu32 (s.x);
> + res1 = _mm_cvtts_sd_u32 (s.x);
>
> if (s.a[0] > UINT_MAX)
> res1_ref = UINT_MAX;
> @@ -44,7 +44,7 @@ TEST (void)
> abort();
>
> #ifdef __x86_64__
> - res2 = _mm_cvtts_roundsd_epu64 (s.x, 8);
> + res2 = _mm_cvtts_roundsd_u64 (s.x, 8);
>
> if (s.a[0] > ULONG_MAX)
> res2_ref = ULONG_MAX;
> @@ -56,7 +56,7 @@ TEST (void)
> if (res2 != res2_ref)
> abort();
>
> - res2 = _mm_cvtts_sd_epu64 (s.x);
> + res2 = _mm_cvtts_sd_u64 (s.x);
>
> if (s.a[0] > ULONG_MAX)
> res2_ref = ULONG_MAX;
> diff --git a/gcc/testsuite/gcc.target/i386/avx10_2-vcvttss2sis-2.c
> b/gcc/testsuite/gcc.target/i386/avx10_2-vcvttss2sis-2.c
> index 82db8efc1f6..e3290000c90 100644
> --- a/gcc/testsuite/gcc.target/i386/avx10_2-vcvttss2sis-2.c
> +++ b/gcc/testsuite/gcc.target/i386/avx10_2-vcvttss2sis-2.c
> @@ -19,7 +19,7 @@ TEST (void)
>
> s.a[0] = 2.46;
>
> - res1 = _mm_cvtts_roundss_epi32 (s.x, 8);
> + res1 = _mm_cvtts_roundss_i32 (s.x, 8);
>
> if (s.a[0] > INT_MAX)
> res1_ref = INT_MAX;
> @@ -31,7 +31,7 @@ TEST (void)
> if (res1 != res1_ref)
> abort();
>
> - res1 = _mm_cvtts_ss_epi32 (s.x);
> + res1 = _mm_cvtts_ss_i32 (s.x);
>
> if (s.a[0] > INT_MAX)
> res1_ref = INT_MAX;
> @@ -44,7 +44,7 @@ TEST (void)
> abort();
>
> #ifdef __x86_64__
> - res2 = _mm_cvtts_roundss_epi64 (s.x, 8);
> + res2 = _mm_cvtts_roundss_i64 (s.x, 8);
>
> if (s.a[0] > LLONG_MAX)
> res2_ref = LLONG_MAX;
> @@ -56,7 +56,7 @@ TEST (void)
> if (res2 != res2_ref)
> abort();
>
> - res2 = _mm_cvtts_ss_epi64 (s.x);
> + res2 = _mm_cvtts_ss_i64 (s.x);
>
> if (s.a[0] > LLONG_MAX)
> res2_ref = LLONG_MAX;
> diff --git a/gcc/testsuite/gcc.target/i386/avx10_2-vcvttss2usis-2.c
> b/gcc/testsuite/gcc.target/i386/avx10_2-vcvttss2usis-2.c
> index c3d16ffec89..f428d059b86 100644
> --- a/gcc/testsuite/gcc.target/i386/avx10_2-vcvttss2usis-2.c
> +++ b/gcc/testsuite/gcc.target/i386/avx10_2-vcvttss2usis-2.c
> @@ -18,7 +18,7 @@ TEST (void)
>
> s.a[0] = 2.46;
>
> - res1 = _mm_cvtts_roundss_epu32 (s.x, 8);
> + res1 = _mm_cvtts_roundss_u32 (s.x, 8);
>
> if (s.a[0] > UINT_MAX)
> res1_ref = UINT_MAX;
> @@ -30,7 +30,7 @@ TEST (void)
> if (res1 != res1_ref)
> abort();
>
> - res1 = _mm_cvtts_ss_epu32 (s.x);
> + res1 = _mm_cvtts_ss_u32 (s.x);
>
> if (s.a[0] > UINT_MAX)
> res1_ref = UINT_MAX;
> @@ -43,7 +43,7 @@ TEST (void)
> abort();
>
> #ifdef __x86_64__
> - res2 = _mm_cvtts_roundss_epu64 (s.x, 8);
> + res2 = _mm_cvtts_roundss_u64 (s.x, 8);
>
> if (s.a[0] > ULONG_MAX)
> res2_ref = ULONG_MAX;
> @@ -55,7 +55,7 @@ TEST (void)
> if (res2 != res2_ref)
> abort();
>
> - res2 = _mm_cvtts_ss_epu64 (s.x);
> + res2 = _mm_cvtts_ss_u64 (s.x);
>
> if (s.a[0] > ULONG_MAX)
> res2_ref = ULONG_MAX;
> diff --git a/gcc/testsuite/gcc.target/i386/sse-14.c
> b/gcc/testsuite/gcc.target/i386/sse-14.c
> index c1053c892c0..84be5f939a9 100644
> --- a/gcc/testsuite/gcc.target/i386/sse-14.c
> +++ b/gcc/testsuite/gcc.target/i386/sse-14.c
> @@ -1122,15 +1122,19 @@ test_3 (_mm512_mask_cvtts_roundps_epu32, __m512i,
> __m512i, __mmask16, __m512, 8)
> test_1 (_mm512_cvtts_roundps_epu64, __m512i, __m256, 8)
> test_2 (_mm512_maskz_cvtts_roundps_epu64, __m512i, __mmask8, __m256, 8)
> test_3 (_mm512_mask_cvtts_roundps_epu64, __m512i, __m512i, __mmask8, __m256,
> 8)
> -test_1 (_mm_cvtts_roundsd_epi32, int, __m128d, 8)
> -test_1 (_mm_cvtts_roundsd_epu32, unsigned int, __m128d, 8)
> -test_1 (_mm_cvtts_roundss_epi32, int, __m128, 8)
> -test_1 (_mm_cvtts_roundss_epu32, unsigned int, __m128, 8)
> +test_1 (_mm_cvtts_roundsd_i32, int, __m128d, 8)
> +test_1 (_mm_cvtts_roundsd_si32, int, __m128d, 8)
> +test_1 (_mm_cvtts_roundsd_u32, unsigned int, __m128d, 8)
> +test_1 (_mm_cvtts_roundss_i32, int, __m128, 8)
> +test_1 (_mm_cvtts_roundss_si32, int, __m128, 8)
> +test_1 (_mm_cvtts_roundss_u32, unsigned int, __m128, 8)
> #ifdef __x86_64__
> -test_1 (_mm_cvtts_roundsd_epi64, long long, __m128d, 8)
> -test_1 (_mm_cvtts_roundsd_epu64, unsigned long long, __m128d, 8)
> -test_1 (_mm_cvtts_roundss_epi64, long long, __m128, 8)
> -test_1 (_mm_cvtts_roundss_epu64, unsigned long long, __m128, 8)
> +test_1 (_mm_cvtts_roundsd_i64, long long, __m128d, 8)
> +test_1 (_mm_cvtts_roundsd_si64, long long, __m128d, 8)
> +test_1 (_mm_cvtts_roundsd_u64, unsigned long long, __m128d, 8)
> +test_1 (_mm_cvtts_roundss_i64, long long, __m128, 8)
> +test_1 (_mm_cvtts_roundss_si64, long long, __m128, 8)
> +test_1 (_mm_cvtts_roundss_u64, unsigned long long, __m128, 8)
> #endif
>
> /* avx10_2minmaxintrin.h */
> diff --git a/gcc/testsuite/gcc.target/i386/sse-22.c
> b/gcc/testsuite/gcc.target/i386/sse-22.c
> index 6b78f6f950f..dd725c0871a 100644
> --- a/gcc/testsuite/gcc.target/i386/sse-22.c
> +++ b/gcc/testsuite/gcc.target/i386/sse-22.c
> @@ -1163,15 +1163,19 @@ test_3 (_mm512_mask_cvtts_roundps_epu32, __m512i,
> __m512i, __mmask16, __m512, 8)
> test_1 (_mm512_cvtts_roundps_epu64, __m512i, __m256, 8)
> test_2 (_mm512_maskz_cvtts_roundps_epu64, __m512i, __mmask8, __m256, 8)
> test_3 (_mm512_mask_cvtts_roundps_epu64, __m512i, __m512i, __mmask8, __m256,
> 8)
> -test_1 (_mm_cvtts_roundsd_epi32, int, __m128d, 8)
> -test_1 (_mm_cvtts_roundsd_epu32, unsigned int, __m128d, 8)
> -test_1 (_mm_cvtts_roundss_epi32, int, __m128, 8)
> -test_1 (_mm_cvtts_roundss_epu32, unsigned int, __m128, 8)
> +test_1 (_mm_cvtts_roundsd_i32, int, __m128d, 8)
> +test_1 (_mm_cvtts_roundsd_si32, int, __m128d, 8)
> +test_1 (_mm_cvtts_roundsd_u32, unsigned int, __m128d, 8)
> +test_1 (_mm_cvtts_roundss_i32, int, __m128, 8)
> +test_1 (_mm_cvtts_roundss_si32, int, __m128, 8)
> +test_1 (_mm_cvtts_roundss_u32, unsigned int, __m128, 8)
> #ifdef __x86_64__
> -test_1 (_mm_cvtts_roundsd_epi64, long long, __m128d, 8)
> -test_1 (_mm_cvtts_roundsd_epu64, unsigned long long, __m128d, 8)
> -test_1 (_mm_cvtts_roundss_epi64, long long, __m128, 8)
> -test_1 (_mm_cvtts_roundss_epu64, unsigned long long, __m128, 8)
> +test_1 (_mm_cvtts_roundsd_i64, long long, __m128d, 8)
> +test_1 (_mm_cvtts_roundsd_si64, long long, __m128d, 8)
> +test_1 (_mm_cvtts_roundsd_u64, unsigned long long, __m128d, 8)
> +test_1 (_mm_cvtts_roundss_i64, long long, __m128, 8)
> +test_1 (_mm_cvtts_roundss_si64, long long, __m128, 8)
> +test_1 (_mm_cvtts_roundss_u64, unsigned long long, __m128, 8)
> #endif
>
> /* avx10_2minmaxintrin.h */
> diff --git a/gcc/testsuite/gcc.target/i386/zext-avx10-1.c
> b/gcc/testsuite/gcc.target/i386/zext-avx10-1.c
> index 66f28125fff..087e1ca8709 100644
> --- a/gcc/testsuite/gcc.target/i386/zext-avx10-1.c
> +++ b/gcc/testsuite/gcc.target/i386/zext-avx10-1.c
> @@ -19,7 +19,7 @@
> unsigned long long
> func1 (__m128 x)
> {
> - return _mm_cvtts_ss_epu32 (x);
> + return _mm_cvtts_ss_u32 (x);
> }
>
> /*
> @@ -35,7 +35,7 @@ func1 (__m128 x)
> unsigned long long
> func2 (__m128 x)
> {
> - return _mm_cvtts_roundss_epu32
> + return _mm_cvtts_roundss_u32
> (x, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
> }
>
> @@ -52,7 +52,7 @@ func2 (__m128 x)
> unsigned long long
> func3 (__m128 x)
> {
> - return (unsigned int) _mm_cvtts_ss_epi32 (x);
> + return (unsigned int) _mm_cvtts_ss_i32 (x);
> }
>
> /*
> @@ -68,7 +68,7 @@ func3 (__m128 x)
> unsigned long long
> func4 (__m128 x)
> {
> - return (unsigned int) _mm_cvtts_roundss_epi32
> + return (unsigned int) _mm_cvtts_roundss_i32
> (x, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
> }
>
> @@ -85,7 +85,7 @@ func4 (__m128 x)
> unsigned long long
> func5 (__m128d x)
> {
> - return _mm_cvtts_sd_epu32 (x);
> + return _mm_cvtts_sd_u32 (x);
> }
>
> /*
> @@ -101,7 +101,7 @@ func5 (__m128d x)
> unsigned long long
> func6 (__m128d x)
> {
> - return _mm_cvtts_roundsd_epu32
> + return _mm_cvtts_roundsd_u32
> (x, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
> }
>
> @@ -118,7 +118,7 @@ func6 (__m128d x)
> unsigned long long
> func7 (__m128d x)
> {
> - return (unsigned int) _mm_cvtts_sd_epi32 (x);
> + return (unsigned int) _mm_cvtts_sd_i32 (x);
> }
>
> /*
> @@ -134,6 +134,6 @@ func7 (__m128d x)
> unsigned long long
> func8 (__m128d x)
> {
> - return (unsigned int) _mm_cvtts_roundsd_epi32
> + return (unsigned int) _mm_cvtts_roundsd_i32
> (x, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
> }
> --
> 2.31.1
>
--
BR,
Hongtao
