From b2754227139512adecb6fda067632b587ff4a017 Mon Sep 17 00:00:00 2001
From: "Hu, Lin1" <lin1.hu@intel.com>
Date: Mon, 19 Aug 2024 10:09:01 +0800
Subject: AVX10.2 ymm rounding: Support vcvttps2{,u}{dq,qq} and
 vcvtu{dq,qq}2p{s,d,h} intrins

gcc/ChangeLog:

	* config/i386/avx10_2roundingintrin.h: New intrins.
	* config/i386/i386-builtin.def (BDESC): Add new builtins.
	* config/i386/sse.md
	(unspec_fix_truncv8sfv8si2<mask_name>): Extend rounding control.
	(<mask_codefor>fixuns_trunc<mode><sseintvecmodelower>2<mask_name>):
	Ditto.
	(<mask_codefor>floatuns<sseintvecmodelower><mode>2<mask_name><round_name>):
	Add condition check.
	(fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
	Remove round_saeonly_name.

gcc/testsuite/ChangeLog:

	* gcc.target/i386/avx-1.c: Add new builtin test.
	* gcc.target/i386/sse-13.c: Ditto.
	* gcc.target/i386/sse-14.c: Ditto.
	* gcc.target/i386/sse-22.c: Add new macro test.
	* gcc.target/i386/sse-23.c: Ditto.
	* gcc.target/i386/avx10_2-rounding-2.c: Add test.
---
 gcc/config/i386/avx10_2roundingintrin.h | 492 ++++++++++++++++++++++++++++++++
 gcc/config/i386/i386-builtin.def        |   9 +
 gcc/config/i386/sse.md                  |  27 +-
 3 files changed, 515 insertions(+), 13 deletions(-)

(limited to 'gcc/config')

diff --git a/gcc/config/i386/avx10_2roundingintrin.h b/gcc/config/i386/avx10_2roundingintrin.h
index 45a04e5..384facb 100644
--- a/gcc/config/i386/avx10_2roundingintrin.h
+++ b/gcc/config/i386/avx10_2roundingintrin.h
@@ -1451,6 +1451,312 @@ _mm256_maskz_cvtt_roundph_epi16 (__mmask16 __U, __m256h __A, const int __R)
 						      (__mmask16) __U,
 						      __R);
 }
+
+extern __inline __m256i
+__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_cvtt_roundps_epi32 (__m256 __A, const int __R)
+{
+  return
+    (__m256i) __builtin_ia32_cvttps2dq256_mask_round ((__v8sf) __A,
+						      (__v8si)
+						      _mm256_undefined_si256 (),
+						      (__mmask8) -1,
+						      __R);
+}
+
+extern __inline __m256i
+__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_mask_cvtt_roundps_epi32 (__m256i __W, __mmask8 __U, __m256 __A,
+				const int __R)
+{
+  return (__m256i) __builtin_ia32_cvttps2dq256_mask_round ((__v8sf) __A,
+							   (__v8si) __W,
+							   (__mmask8) __U,
+							   __R);
+}
+
+extern __inline __m256i
+__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_maskz_cvtt_roundps_epi32 (__mmask8 __U, __m256 __A, const int __R)
+{
+  return
+    (__m256i) __builtin_ia32_cvttps2dq256_mask_round ((__v8sf) __A,
+						      (__v8si)
+						      _mm256_setzero_si256 (),
+						      (__mmask8) __U,
+						      __R);
+}
+
+extern __inline __m256i
+__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_cvtt_roundps_epi64 (__m128 __A, const int __R)
+{
+  return
+    (__m256i) __builtin_ia32_cvttps2qq256_mask_round ((__v4sf) __A,
+						      (__v4di)
+						      _mm256_setzero_si256 (),
+						      (__mmask8) -1,
+						      __R);
+}
+
+extern __inline __m256i
+__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_mask_cvtt_roundps_epi64 (__m256i __W, __mmask8 __U, __m128 __A,
+				const int __R)
+{
+  return (__m256i) __builtin_ia32_cvttps2qq256_mask_round ((__v4sf) __A,
+							   (__v4di) __W,
+							   (__mmask8) __U,
+							   __R);
+}
+
+extern __inline __m256i
+__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_maskz_cvtt_roundps_epi64 (__mmask8 __U, __m128 __A, const int __R)
+{
+  return
+    (__m256i) __builtin_ia32_cvttps2qq256_mask_round ((__v4sf) __A,
+						      (__v4di)
+						      _mm256_setzero_si256 (),
+						      (__mmask8) __U,
+						      __R);
+}
+
+extern __inline __m256i
+__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_cvtt_roundps_epu32 (__m256 __A, const int __R)
+{
+  return
+    (__m256i) __builtin_ia32_cvttps2udq256_mask_round ((__v8sf) __A,
+						       (__v8si)
+						       _mm256_undefined_si256 (),
+						       (__mmask8) -1,
+						       __R);
+}
+
+extern __inline __m256i
+__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_mask_cvtt_roundps_epu32 (__m256i __W, __mmask8 __U, __m256 __A,
+				const int __R)
+{
+  return (__m256i) __builtin_ia32_cvttps2udq256_mask_round ((__v8sf) __A,
+							    (__v8si) __W,
+							    (__mmask8) __U,
+							    __R);
+}
+
+extern __inline __m256i
+__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_maskz_cvtt_roundps_epu32 (__mmask8 __U, __m256 __A, const int __R)
+{
+  return
+    (__m256i) __builtin_ia32_cvttps2udq256_mask_round ((__v8sf) __A,
+						       (__v8si)
+						       _mm256_setzero_si256 (),
+						       (__mmask8) __U,
+						       __R);
+}
+
+extern __inline __m256i
+__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_cvtt_roundps_epu64 (__m128 __A, const int __R)
+{
+  return
+    (__m256i) __builtin_ia32_cvttps2uqq256_mask_round ((__v4sf) __A,
+						       (__v4di)
+						       _mm256_setzero_si256 (),
+						       (__mmask8) -1,
+						       __R);
+}
+
+extern __inline __m256i
+__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_mask_cvtt_roundps_epu64 (__m256i __W, __mmask8 __U, __m128 __A,
+				const int __R)
+{
+  return (__m256i) __builtin_ia32_cvttps2uqq256_mask_round ((__v4sf) __A,
+							    (__v4di) __W,
+							    (__mmask8) __U,
+							    __R);
+}
+
+extern __inline __m256i
+__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_maskz_cvtt_roundps_epu64 (__mmask8 __U, __m128 __A, const int __R)
+{
+  return
+    (__m256i) __builtin_ia32_cvttps2uqq256_mask_round ((__v4sf) __A,
+						       (__v4di)
+						       _mm256_setzero_si256 (),
+						       (__mmask8) __U,
+						       __R);
+}
+
+extern __inline __m128h
+__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_cvt_roundepu32_ph (__m256i __A, const int __R)
+{
+  return (__m128h) __builtin_ia32_vcvtudq2ph256_mask_round ((__v8si) __A,
+							    (__v8hf)
+							    _mm_setzero_ph (),
+							    (__mmask8) -1,
+							    __R);
+}
+
+extern __inline __m128h
+__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_mask_cvt_roundepu32_ph (__m128h __W, __mmask8 __U, __m256i __A,
+			       const int __R)
+{
+  return (__m128h) __builtin_ia32_vcvtudq2ph256_mask_round ((__v8si) __A,
+							    (__v8hf) __W,
+							    (__mmask8) __U,
+							    __R);
+}
+
+extern __inline __m128h
+__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_maskz_cvt_roundepu32_ph (__mmask8 __U, __m256i __A, const int __R)
+{
+  return (__m128h) __builtin_ia32_vcvtudq2ph256_mask_round ((__v8si) __A,
+							    (__v8hf)
+							    _mm_setzero_ph (),
+							    (__mmask8) __U,
+							    __R);
+}
+
+extern __inline __m256
+__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_cvt_roundepu32_ps (__m256i __A, const int __R)
+{
+  return
+    (__m256) __builtin_ia32_cvtudq2ps256_mask_round ((__v8si) __A,
+						     (__v8sf)
+						     _mm256_undefined_ps (),
+						     (__mmask8) -1,
+						     __R);
+}
+
+extern __inline __m256
+__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_mask_cvt_roundepu32_ps (__m256 __W, __mmask8 __U, __m256i __A,
+			       const int __R)
+{
+  return (__m256) __builtin_ia32_cvtudq2ps256_mask_round ((__v8si) __A,
+							  (__v8sf) __W,
+							  (__mmask8) __U,
+							  __R);
+}
+
+extern __inline __m256
+__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_maskz_cvt_roundepu32_ps (__mmask8 __U, __m256i __A, const int __R)
+{
+  return (__m256) __builtin_ia32_cvtudq2ps256_mask_round ((__v8si) __A,
+							  (__v8sf)
+							  _mm256_setzero_ps (),
+							  (__mmask8) __U,
+							  __R);
+}
+
+extern __inline __m256d
+__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_cvt_roundepu64_pd (__m256i __A, const int __R)
+{
+  return (__m256d) __builtin_ia32_cvtuqq2pd256_mask_round ((__v4di) __A,
+							   (__v4df)
+							   _mm256_setzero_pd (),
+							   (__mmask8) -1,
+							   __R);
+}
+
+extern __inline __m256d
+__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_mask_cvt_roundepu64_pd (__m256d __W, __mmask8 __U, __m256i __A,
+			       const int __R)
+{
+  return (__m256d) __builtin_ia32_cvtuqq2pd256_mask_round ((__v4di) __A,
+							   (__v4df) __W,
+							   (__mmask8) __U,
+							   __R);
+}
+
+extern __inline __m256d
+__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_maskz_cvt_roundepu64_pd (__mmask8 __U, __m256i __A, const int __R)
+{
+  return (__m256d) __builtin_ia32_cvtuqq2pd256_mask_round ((__v4di) __A,
+							   (__v4df)
+							   _mm256_setzero_pd (),
+							   (__mmask8) __U,
+							   __R);
+}
+
+extern __inline __m128h
+__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_cvt_roundepu64_ph (__m256i __A, const int __R)
+{
+  return (__m128h) __builtin_ia32_vcvtuqq2ph256_mask_round ((__v4di) __A,
+							    (__v8hf)
+							    _mm_setzero_ph (),
+							    (__mmask8) -1,
+							    __R);
+}
+
+extern __inline __m128h
+__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_mask_cvt_roundepu64_ph (__m128h __W, __mmask8 __U, __m256i __A,
+			       const int __R)
+{
+  return (__m128h) __builtin_ia32_vcvtuqq2ph256_mask_round ((__v4di) __A,
+							    (__v8hf) __W,
+							    (__mmask8) __U,
+							    __R);
+}
+
+extern __inline __m128h
+__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_maskz_cvt_roundepu64_ph (__mmask8 __U, __m256i __A, const int __R)
+{
+  return (__m128h) __builtin_ia32_vcvtuqq2ph256_mask_round ((__v4di) __A,
+							    (__v8hf)
+							    _mm_setzero_ph (),
+							    (__mmask8) __U,
+							    __R);
+}
+
+extern __inline __m128
+__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_cvt_roundepu64_ps (__m256i __A, const int __R)
+{
+  return (__m128) __builtin_ia32_cvtuqq2ps256_mask_round ((__v4di) __A,
+							  (__v4sf)
+							  _mm_setzero_ps (),
+							  (__mmask8) -1,
+							  __R);
+}
+
+extern __inline __m128
+__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_mask_cvt_roundepu64_ps (__m128 __W, __mmask8 __U, __m256i __A,
+			       const int __R)
+{
+  return (__m128) __builtin_ia32_cvtuqq2ps256_mask_round ((__v4di) __A,
+							  (__v4sf) __W,
+							  (__mmask8) __U,
+							  __R);
+}
+
+extern __inline __m128
+__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_maskz_cvt_roundepu64_ps (__mmask8 __U, __m256i __A, const int __R)
+{
+  return (__m128) __builtin_ia32_cvtuqq2ps256_mask_round ((__v4di) __A,
+							  (__v4sf)
+							  _mm_setzero_ps (),
+							  (__mmask8) __U,
+							  __R);
+}
 #else
 #define _mm256_add_round_pd(A, B, R) \
   ((__m256d) __builtin_ia32_addpd256_mask_round ((__v4df) (A), \
@@ -2310,6 +2616,192 @@ _mm256_maskz_cvtt_roundph_epi16 (__mmask16 __U, __m256h __A, const int __R)
 					    (_mm256_setzero_si256 ()), \
 					    (__mmask16) (U), \
 					    (R)))
+
+#define _mm256_cvtt_roundps_epi32(A, R) \
+  ((__m256i) \
+   __builtin_ia32_cvttps2dq256_mask_round ((__v8sf) (A), \
+					   (__v8si) \
+					   (_mm256_undefined_si256 ()), \
+					   (__mmask8) (-1), \
+					   (R)))
+
+#define _mm256_mask_cvtt_roundps_epi32(W, U, A, R) \
+  ((__m256i) __builtin_ia32_cvttps2dq256_mask_round ((__v8sf) (A), \
+						     (__v8si) (W), \
+						     (__mmask8) (U), \
+						     (R)))
+
+#define _mm256_maskz_cvtt_roundps_epi32(U, A, R) \
+  ((__m256i) \
+   __builtin_ia32_cvttps2dq256_mask_round ((__v8sf) (A), \
+					   (__v8si) \
+					   (_mm256_setzero_si256 ()),  \
+					   (__mmask8) (U),  \
+					   (R)))
+
+#define _mm256_cvtt_roundps_epi64(A, R) \
+  ((__m256i) __builtin_ia32_cvttps2qq256_mask_round ((__v4sf) (A), \
+						     (__v4di) \
+						     (_mm256_setzero_si256 ()),\
+						     (__mmask8) (-1), \
+						     (R)))
+
+#define _mm256_mask_cvtt_roundps_epi64(W, U, A, R) \
+  ((__m256i) __builtin_ia32_cvttps2qq256_mask_round ((__v4sf) (A), \
+						     (__v4di) (W), \
+						     (__mmask8) (U), \
+						     (R)))
+
+#define _mm256_maskz_cvtt_roundps_epi64(U, A, R) \
+  ((__m256i) __builtin_ia32_cvttps2qq256_mask_round ((__v4sf) (A), \
+						     (__v4di) \
+						     (_mm256_setzero_si256 ()),\
+						     (__mmask8) (U), \
+						     (R)))
+
+#define _mm256_cvtt_roundps_epu32(A, R) \
+  ((__m256i) \
+   __builtin_ia32_cvttps2udq256_mask_round ((__v8sf) (A), \
+					    (__v8si) \
+					    (_mm256_undefined_si256 ()), \
+					    (__mmask8) (-1), \
+					    (R)))
+
+#define _mm256_mask_cvtt_roundps_epu32(W, U, A, R) \
+  ((__m256i) __builtin_ia32_cvttps2udq256_mask_round ((__v8sf) (A), \
+						      (__v8si) (W), \
+						      (__mmask8) (U), \
+						      (R)))
+
+#define _mm256_maskz_cvtt_roundps_epu32(U, A, R) \
+  ((__m256i) \
+   __builtin_ia32_cvttps2udq256_mask_round ((__v8sf) (A), \
+					    (__v8si) \
+					    (_mm256_setzero_si256 ()), \
+					    (__mmask8) (U), \
+					    (R)))
+
+#define _mm256_cvtt_roundps_epu64(A, R) \
+  ((__m256i) \
+   __builtin_ia32_cvttps2uqq256_mask_round ((__v4sf) (A), \
+					    (__v4di) \
+					    (_mm256_setzero_si256 ()), \
+					    (__mmask8) (-1), \
+					    (R)))
+
+#define _mm256_mask_cvtt_roundps_epu64(W, U, A, R) \
+  ((__m256i) __builtin_ia32_cvttps2uqq256_mask_round ((__v4sf) (A), \
+						      (__v4di) (W), \
+						      (__mmask8) (U), \
+						      (R)))
+
+#define _mm256_maskz_cvtt_roundps_epu64(U, A, R) \
+  ((__m256i) \
+   __builtin_ia32_cvttps2uqq256_mask_round ((__v4sf) (A), \
+					    (__v4di) \
+					    (_mm256_setzero_si256 ()), \
+					    (__mmask8) (U),  \
+					    (R)))
+
+#define _mm256_cvt_roundepu32_ph(A, R) \
+  ((__m128h) __builtin_ia32_vcvtudq2ph256_mask_round ((__v8si) (A), \
+						      (__v8hf) \
+						      (_mm_setzero_ph ()), \
+						      (__mmask8) (-1), \
+						      (R)))
+
+#define _mm256_mask_cvt_roundepu32_ph(W, U, A, R) \
+  ((__m128h) __builtin_ia32_vcvtudq2ph256_mask_round ((__v8si) (A), \
+						      (__v8hf) (W), \
+						      (__mmask8) (U), \
+						      (R)))
+
+#define _mm256_maskz_cvt_roundepu32_ph(U, A, R) \
+  ((__m128h) __builtin_ia32_vcvtudq2ph256_mask_round ((__v8si) (A), \
+						      (__v8hf) \
+						      (_mm_setzero_ph ()), \
+						      (__mmask8) (U), \
+						      (R)))
+
+#define _mm256_cvt_roundepu32_ps(A, R) \
+  ((__m256) __builtin_ia32_cvtudq2ps256_mask_round ((__v8si) (A), \
+						    (__v8sf) \
+						    (_mm256_undefined_ps ()), \
+						    (__mmask8) (-1), \
+						    (R)))
+
+#define _mm256_mask_cvt_roundepu32_ps(W, U, A, R) \
+  ((__m256) __builtin_ia32_cvtudq2ps256_mask_round ((__v8si) (A), \
+						    (__v8sf) (W), \
+						    (__mmask8) (U), \
+						    (R)))
+
+#define _mm256_maskz_cvt_roundepu32_ps(U, A, R) \
+  ((__m256) __builtin_ia32_cvtudq2ps256_mask_round ((__v8si) (A), \
+						    (__v8sf) \
+						    (_mm256_setzero_ps ()), \
+						    (__mmask8) (U), \
+						    (R)))
+
+#define _mm256_cvt_roundepu64_pd(A, R) \
+  ((__m256d) __builtin_ia32_cvtuqq2pd256_mask_round ((__v4di) (A), \
+						     (__v4df) \
+						     (_mm256_setzero_pd ()), \
+						     (__mmask8) (-1),  \
+						     (R)))
+
+#define _mm256_mask_cvt_roundepu64_pd(W, U, A, R) \
+  ((__m256d) __builtin_ia32_cvtuqq2pd256_mask_round ((__v4di) (A), \
+						     (__v4df) (W),  \
+						     (__mmask8) (U),  \
+						     (R)))
+
+#define _mm256_maskz_cvt_roundepu64_pd(U, A, R) \
+  ((__m256d) __builtin_ia32_cvtuqq2pd256_mask_round ((__v4di) (A), \
+						     (__v4df) \
+						     (_mm256_setzero_pd ()), \
+						     (__mmask8) (U), \
+						     (R)))
+
+#define _mm256_cvt_roundepu64_ph(A, R) \
+  ((__m128h) __builtin_ia32_vcvtuqq2ph256_mask_round ((__v4di) (A), \
+						      (__v8hf) \
+						      (_mm_setzero_ph ()), \
+						      (__mmask8) (-1), \
+						      (R)))
+
+#define _mm256_mask_cvt_roundepu64_ph(W, U, A, R) \
+  ((__m128h) __builtin_ia32_vcvtuqq2ph256_mask_round ((__v4di) (A), \
+						      (__v8hf) (W), \
+						      (__mmask8) (U), \
+						      (R)))
+
+#define _mm256_maskz_cvt_roundepu64_ph(U, A, R) \
+  ((__m128h) __builtin_ia32_vcvtuqq2ph256_mask_round ((__v4di) (A), \
+						      (__v8hf) \
+						      (_mm_setzero_ph ()), \
+						      (__mmask8) (U), \
+						      (R)))
+
+#define _mm256_cvt_roundepu64_ps(A, R) \
+  ((__m128) __builtin_ia32_cvtuqq2ps256_mask_round ((__v4di) (A), \
+						    (__v4sf) \
+						    (_mm_setzero_ps ()), \
+						    (__mmask8) (-1), \
+						    (R)))
+
+#define _mm256_mask_cvt_roundepu64_ps(W, U, A, R) \
+  ((__m128) __builtin_ia32_cvtuqq2ps256_mask_round ((__v4di) (A), \
+						    (__v4sf) (W), \
+						    (__mmask8) (U), \
+						    (R)))
+
+#define _mm256_maskz_cvt_roundepu64_ps(U, A, R) \
+  ((__m128) __builtin_ia32_cvtuqq2ps256_mask_round ((__v4di) (A), \
+						    (__v4sf) \
+						    (_mm_setzero_ps ()), \
+						    (__mmask8) (U), \
+						    (R)))
 #endif
 
 #ifdef __DISABLE_AVX10_2_256__
diff --git a/gcc/config/i386/i386-builtin.def b/gcc/config/i386/i386-builtin.def
index a883336..1290ae6 100644
--- a/gcc/config/i386/i386-builtin.def
+++ b/gcc/config/i386/i386-builtin.def
@@ -3361,6 +3361,15 @@ BDESC (0, OPTION_MASK_ISA2_AVX10_2_256, CODE_FOR_unspec_avx512fp16_fixuns_truncv
 BDESC (0, OPTION_MASK_ISA2_AVX10_2_256, CODE_FOR_unspec_avx512fp16_fixuns_truncv4di2_mask_round, "__builtin_ia32_vcvttph2uqq256_mask_round", IX86_BUILTIN_VCVTTPH2UQQ256_MASK_ROUND, UNKNOWN, (int) V4DI_FTYPE_V8HF_V4DI_UQI_INT)
 BDESC (0, OPTION_MASK_ISA2_AVX10_2_256, CODE_FOR_unspec_avx512fp16_fixuns_truncv16hi2_mask_round, "__builtin_ia32_vcvttph2uw256_mask_round", IX86_BUILTIN_VCVTTPH2UW256_MASK_ROUND, UNKNOWN, (int) V16HI_FTYPE_V16HF_V16HI_UHI_INT)
 BDESC (0, OPTION_MASK_ISA2_AVX10_2_256, CODE_FOR_unspec_avx512fp16_fix_truncv16hi2_mask_round, "__builtin_ia32_vcvttph2w256_mask_round", IX86_BUILTIN_VCVTTPH2W256_MASK_ROUND, UNKNOWN, (int) V16HI_FTYPE_V16HF_V16HI_UHI_INT)
+BDESC (0, OPTION_MASK_ISA2_AVX10_2_256, CODE_FOR_unspec_fix_truncv8sfv8si2_mask_round, "__builtin_ia32_cvttps2dq256_mask_round", IX86_BUILTIN_VCVTTPS2DQ256_MASK_ROUND, UNKNOWN, (int) V8SI_FTYPE_V8SF_V8SI_UQI_INT)
+BDESC (0, OPTION_MASK_ISA2_AVX10_2_256, CODE_FOR_unspec_fix_truncv4sfv4di2_mask_round, "__builtin_ia32_cvttps2qq256_mask_round", IX86_BUILTIN_VCVTTPS2QQ256_MASK_ROUND, UNKNOWN, (int) V4DI_FTYPE_V4SF_V4DI_UQI_INT)
+BDESC (0, OPTION_MASK_ISA2_AVX10_2_256, CODE_FOR_unspec_fixuns_truncv8sfv8si2_mask_round, "__builtin_ia32_cvttps2udq256_mask_round", IX86_BUILTIN_VCVTTPS2UDQ256_MASK_ROUND, UNKNOWN, (int) V8SI_FTYPE_V8SF_V8SI_UQI_INT)
+BDESC (0, OPTION_MASK_ISA2_AVX10_2_256, CODE_FOR_unspec_fixuns_truncv4sfv4di2_mask_round, "__builtin_ia32_cvttps2uqq256_mask_round", IX86_BUILTIN_VCVTTPS2UQQ256_MASK_ROUND, UNKNOWN, (int) V4DI_FTYPE_V4SF_V4DI_UQI_INT)
+BDESC (0, OPTION_MASK_ISA2_AVX10_2_256, CODE_FOR_avx512fp16_vcvtudq2ph_v8si_mask_round, "__builtin_ia32_vcvtudq2ph256_mask_round", IX86_BUILTIN_VCVTUDQ2PH256_MASK_ROUND, UNKNOWN, (int) V8HF_FTYPE_V8SI_V8HF_UQI_INT)
+BDESC (0, OPTION_MASK_ISA2_AVX10_2_256, CODE_FOR_floatunsv8siv8sf2_mask_round, "__builtin_ia32_cvtudq2ps256_mask_round", IX86_BUILTIN_VCVTUDQ2PS256_MASK_ROUND, UNKNOWN, (int) V8SF_FTYPE_V8SI_V8SF_UQI_INT)
+BDESC (0, OPTION_MASK_ISA2_AVX10_2_256, CODE_FOR_floatunsv4div4df2_mask_round, "__builtin_ia32_cvtuqq2pd256_mask_round", IX86_BUILTIN_VCVTUQQ2PD256_MASK_ROUND, UNKNOWN, (int) V4DF_FTYPE_V4DI_V4DF_UQI_INT)
+BDESC (0, OPTION_MASK_ISA2_AVX10_2_256, CODE_FOR_avx512fp16_vcvtuqq2ph_v4di_mask_round, "__builtin_ia32_vcvtuqq2ph256_mask_round", IX86_BUILTIN_VCVTUQQ2PH256_MASK_ROUND, UNKNOWN, (int) V8HF_FTYPE_V4DI_V8HF_UQI_INT)
+BDESC (0, OPTION_MASK_ISA2_AVX10_2_256, CODE_FOR_floatunsv4div4sf2_mask_round, "__builtin_ia32_cvtuqq2ps256_mask_round", IX86_BUILTIN_VCVTUQQ2PS256_MASK_ROUND, UNKNOWN, (int) V4SF_FTYPE_V4DI_V4SF_UQI_INT)
 
 BDESC_END (ROUND_ARGS, MULTI_ARG)
 
diff --git a/gcc/config/i386/sse.md b/gcc/config/i386/sse.md
index 53c15d1..25de88b 100644
--- a/gcc/config/i386/sse.md
+++ b/gcc/config/i386/sse.md
@@ -8401,7 +8401,7 @@
   [(set (match_operand:VF1_AVX512VL 0 "register_operand" "=v")
 	(unsigned_float:VF1_AVX512VL
 	  (match_operand:<sseintvecmode> 1 "nonimmediate_operand" "<round_constraint>")))]
-  "TARGET_AVX512F"
+  "TARGET_AVX512F && <round_mode_condition>"
   "vcvtudq2ps\t{<round_mask_op2>%1, %0<mask_operand2>|%0<mask_operand2>, %1<round_mask_op2>}"
   [(set_attr "type" "ssecvt")
    (set_attr "prefix" "evex")
@@ -8530,12 +8530,13 @@
    (set_attr "prefix" "evex")
    (set_attr "mode" "XI")])
 
-(define_insn "unspec_fix_truncv8sfv8si2<mask_name>"
+(define_insn "unspec_fix_truncv8sfv8si2<mask_name><round_saeonly_name>"
   [(set (match_operand:V8SI 0 "register_operand" "=v")
-	(unspec:V8SI [(match_operand:V8SF 1 "nonimmediate_operand" "vm")]
+	(unspec:V8SI [(match_operand:V8SF 1 "<round_saeonly_nimm_predicate>" "<round_saeonly_constraint>")]
 	 UNSPEC_VCVTT))]
-  "TARGET_AVX && <mask_avx512vl_condition>"
-  "vcvttps2dq\t{%1, %0<mask_operand2>|%0<mask_operand2>, %1}"
+  "TARGET_AVX && <mask_avx512vl_condition>
+  && (!<round_saeonly_applied> || TARGET_AVX10_2_256)"
+  "vcvttps2dq\t{<round_saeonly_mask_op2>%1, %0<mask_operand2>|%0<mask_operand2>, %1<round_saeonly_mask_op2>}"
   [(set_attr "type" "ssecvt")
    (set_attr "prefix" "<mask_prefix>")
    (set_attr "mode" "OI")])
@@ -9488,12 +9489,12 @@
    (set_attr "prefix" "evex")
    (set_attr "mode" "<sseinsnmode>")])
 
-(define_insn "fix<fixunssuffix>_trunc<ssePSmode2lower><mode>2<mask_name><round_saeonly_name>"
+(define_insn "fix<fixunssuffix>_trunc<ssePSmode2lower><mode>2<mask_name>"
   [(set (match_operand:VI8_256_512 0 "register_operand" "=v")
 	(any_fix:VI8_256_512
-	  (match_operand:<ssePSmode2> 1 "<round_saeonly_nimm_predicate>" "<round_saeonly_constraint>")))]
-  "TARGET_AVX512DQ && <round_saeonly_mode_condition>"
-  "vcvttps2<fixsuffix>qq\t{<round_saeonly_mask_op2>%1, %0<mask_operand2>|%0<mask_operand2>, %1<round_saeonly_mask_op2>}"
+	  (match_operand:<ssePSmode2> 1 "nonimmediate_operand" "vm")))]
+  "TARGET_AVX512DQ"
+  "vcvttps2<fixsuffix>qq\t{%1, %0<mask_operand2>|%0<mask_operand2>, %1}"
   [(set_attr "type" "ssecvt")
    (set_attr "prefix" "evex")
    (set_attr "mode" "<sseinsnmode>")])
@@ -9648,13 +9649,13 @@
   DONE;
 })
 
-(define_insn "<mask_codefor>unspec_fixuns_trunc<mode><sseintvecmodelower>2<mask_name>"
+(define_insn "<mask_codefor>unspec_fixuns_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>"
   [(set (match_operand:<sseintvecmode> 0 "register_operand" "=v")
 	(unspec:<sseintvecmode>
-	  [(match_operand:VF1_128_256 1 "nonimmediate_operand" "vm")]
+	  [(match_operand:VF1_128_256 1 "<round_saeonly_nimm_predicate>" "<round_saeonly_constraint>")]
 	  UNSPEC_VCVTTU))]
-  "TARGET_AVX512VL"
-  "vcvttps2udq\t{%1, %0<mask_operand2>|%0<mask_operand2>, %1}"
+  "TARGET_AVX512VL && <round_saeonly_mode_condition>"
+  "vcvttps2udq\t{<round_saeonly_mask_op2>%1, %0<mask_operand2>|%0<mask_operand2>, %1<round_saeonly_mask_op2>}"
   [(set_attr "type" "ssecvt")
    (set_attr "prefix" "evex")
    (set_attr "mode" "<sseintvecmode2>")])
-- 
cgit v1.1