[clang] [llvm] [RISCV][P-ext] Add packed multiply high accumulate with byte/halfword index intrinsics (PR #224261)

Craig Topper via cfe-commits cfe-commits at lists.llvm.org
Thu Sep 17 10:05:07 PDT 2026


================
@@ -911,6 +905,22 @@ __packed_ternary_builtin_mixed(pmhaccsu_i32x2, int32x2_t, int32x2_t, uint32x2_t,
 __packed_ternary_builtin_mixed(pmhraccsu_i16x4, int16x4_t, int16x4_t, uint16x4_t, __builtin_riscv_pmhraccsu_i16x4)
 __packed_ternary_builtin_mixed(pmhraccsu_i32x2, int32x2_t, int32x2_t, uint32x2_t, __builtin_riscv_pmhraccsu_i32x2)
 
+/* Packed Multiply High Accumulate with Byte Index (32-bit) */
----------------
topperc wrote:

Where did the word "Index" come from. That word is not used in relation to these instruction in the specs.

https://github.com/llvm/llvm-project/pull/224261


More information about the cfe-commits mailing list