================
@@ -126,6 +197,43 @@
#define _tile_cvtrowps2bf16l(tsrc, row)
\
__builtin_ia32_tcvtrowps2bf16l(tsrc, row)
+/// Moves a row from a tile register to a zmm destination register, converting
+/// the fp32 source elements to bf16. It places the resulting bf16 elements
+/// in the low 16 bits within each dword. The row of the tile is selected
+/// by a 8b immediate value.
+///
+/// \headerfile <x86intrin.h>
+///
+/// \code
+/// __m512i _tile_cvtrowps2bf16li(__tile tsrc, const unsigned int imm8);
+/// \endcode
+///
+/// \code{.operation}
+/// VL := 512
+/// VL_bytes := VL >> 3
+/// row_index := imm8 & 0x3ff
----------------
fzou1 wrote:
Same as above.
https://github.com/llvm/llvm-project/pull/173545
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
