[clang] [llvm] [X86][AMX] Support AMX-TRANSPOSE (PR #113532)

[Phoebe Wang via cfe-commits]

================
@@ -0,0 +1,248 @@
+/* ===--- amxtransposeintrin.h - AMX_TRANSPOSE intrinsics -*- C++ -*---------===
+ *
+ * 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 <amxtransposeintrin.h> directly; use <immintrin.h> instead."
+#endif /* __IMMINTRIN_H */
+
+#ifndef __AMX_TRANSPOSEINTRIN_H
+#define __AMX_TRANSPOSEINTRIN_H
+#ifdef __x86_64__
+
+#define __DEFAULT_FN_ATTRS_TRANSPOSE                                           \
+  __attribute__((__always_inline__, __nodebug__, __target__("amx-transpose")))
+
+#define _tile_2rpntlvwz0(tdst, base, stride)                                   \
+  __builtin_ia32_t2rpntlvwz0(tdst, base, stride)
+#define _tile_2rpntlvwz0t1(tdst, base, stride)                                 \
+  __builtin_ia32_t2rpntlvwz0t1(tdst, base, stride)
+#define _tile_2rpntlvwz1(tdst, base, stride)                                   \
+  __builtin_ia32_t2rpntlvwz1(tdst, base, stride)
+#define _tile_2rpntlvwz1t1(tdst, base, stride)                                 \
+  __builtin_ia32_t2rpntlvwz1t1(tdst, base, stride)
+
+/// Transpose 32-bit elements from \a src and write the result to \a dst.
+///
+/// \headerfile <immintrin.h>
+///
+/// \code
+/// void __tile_transposed(__tile dst, __tile src);
----------------
phoebewang wrote:

Done.

https://github.com/llvm/llvm-project/pull/113532