[clang] [llvm] [X86][AVX10.2] Support AVX10.2-MINMAX new instructions. (PR #101598)

[Freddy Ye via llvm-commits]

================
@@ -0,0 +1,188 @@
+/*===--------------- avx10_2minmaxintrin.h - AVX10_2MINMAX intrinsics
+ *-----------------===
+ *
+ * Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+ * See https://llvm.org/LICENSE.txt for license information.
+ * SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+ *
+ *===-----------------------------------------------------------------------===
+ */
+#ifndef __IMMINTRIN_H
+#error                                                                         \
+    "Never use <avx10_2minmaxintrin.h> directly; include <immintrin.h> instead."
+#endif // __IMMINTRIN_H
+
+#ifndef __AVX10_2MINMAXINTRIN_H
+#define __AVX10_2MINMAXINTRIN_H
+
+#define _mm_minmaxne_pbh(A, B, C)                                              \
+  ((__m128bh)__builtin_ia32_vminmaxnepbf16128(                                 \
+      (__m128bh)(__v8bf)(A), (__m128bh)(__v8bf)(B), (int)(C)))
+
+#define _mm_mask_minmaxne_pbh(W, U, A, B, C)                                   \
+  ((__m128bh)__builtin_ia32_selectpbf_128(                                     \
+      (__mmask8)(U),                                                           \
+      (__v8bf)__builtin_ia32_vminmaxnepbf16128(                                \
+          (__m128bh)(__v8bf)(A), (__m128bh)(__v8bf)(B), (int)(C)),             \
+      (__v8bf)(W)))
+
+#define _mm_maskz_minmaxne_pbh(U, A, B, C)                                     \
+  ((__m128bh)__builtin_ia32_selectpbf_128(                                     \
+      (__mmask8)(U),                                                           \
+      (__v8bf)__builtin_ia32_vminmaxnepbf16128(                                \
+          (__m128bh)(__v8bf)(A), (__m128bh)(__v8bf)(B), (int)(C)),             \
+      (__v8bf) __builtin_bit_cast(__m128bh, _mm_setzero_ps())))
+
+#define _mm256_minmaxne_pbh(A, B, C)                                           \
+  ((__m256bh)__builtin_ia32_vminmaxnepbf16256(                                 \
+      (__m256bh)(__v16bf)(A), (__m256bh)(__v16bf)(B), (int)(C)))
+
+#define _mm256_mask_minmaxne_pbh(W, U, A, B, C)                                \
+  ((__m256bh)__builtin_ia32_selectpbf_256(                                     \
+      (__mmask16)(U),                                                          \
+      (__v16bf)__builtin_ia32_vminmaxnepbf16256(                               \
+          (__m256bh)(__v16bf)(A), (__m256bh)(__v16bf)(B), (int)(C)),           \
+      (__v16bf)(W)))
+
+#define _mm256_maskz_minmaxne_pbh(U, A, B, C)                                  \
+  ((__m256bh)__builtin_ia32_selectpbf_256(                                     \
+      (__mmask16)(U),                                                          \
+      (__v16bf)__builtin_ia32_vminmaxnepbf16256(                               \
+          (__m256bh)(__v16bf)(A), (__m256bh)(__v16bf)(B), (int)(C)),           \
+      (__v16bf) __builtin_bit_cast(__m256bh, _mm256_setzero_ps())))
+
+#define _mm_minmax_pd(A, B, C)                                                 \
+  ((__m128d)__builtin_ia32_vminmaxpd128_mask(                                  \
+      (__v2df)(__m128d)(A), (__v2df)(__m128d)(B), (int)(C),                    \
+      (__v2df)_mm_setzero_pd(), (__mmask8)(-1)))
+
+#define _mm_mask_minmax_pd(W, U, A, B, C)                                      \
+  ((__m128d)__builtin_ia32_vminmaxpd128_mask(                                  \
+      (__v2df)(__m128d)(A), (__v2df)(__m128d)(B), (int)(C),                    \
+      (__v2df)(__m128d)(W), (__mmask8)(U)))
+
+#define _mm_maskz_minmax_pd(U, A, B, C)                                        \
+  ((__m128d)__builtin_ia32_vminmaxpd128_mask(                                  \
+      (__v2df)(__m128d)(A), (__v2df)(__m128d)(B), (int)(C),                    \
+      (__v2df)_mm_setzero_pd(), (__mmask8)(U)))
+
+#define _mm256_minmax_pd(A, B, C)                                              \
+  ((__m256d)__builtin_ia32_vminmaxpd256_round_mask(                            \
+      (__v4df)(__m256d)(A), (__v4df)(__m256d)(B), (int)(C),                    \
+      (__v4df)_mm256_setzero_pd(), (__mmask8)(-1), _MM_FROUND_NO_EXC))
----------------
FreddyLeaf wrote:

[adfe6cd](https://github.com/llvm/llvm-project/pull/101598/commits/adfe6cd518f7f75e8dbbdfcdc6ddadae1a27fc4a)

https://github.com/llvm/llvm-project/pull/101598