[llvm-branch-commits] [llvm] [SLP] Price gathered lanes of vectorized entries as a shuffle (PR #240136)

Dmitry Sidorov via llvm-branch-commits llvm-branch-commits at lists.llvm.org
Sun Nov 8 06:59:07 PST 2026


================
@@ -6,14 +6,15 @@
 define void @swapped_operand_f32(ptr %a, ptr %out) {
 ; CHECK-LABEL: define void @swapped_operand_f32(
 ; CHECK-SAME: ptr [[A:%.*]], ptr [[OUT:%.*]]) {
-; CHECK-NEXT:    [[TMP1:%.*]] = load <2 x float>, ptr [[A]], align 4
-; CHECK-NEXT:    [[TMP2:%.*]] = fmul <2 x float> [[TMP1]], <float 2.000000e+00, float 3.000000e+00>
-; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <2 x float> [[TMP2]], <2 x float> poison, <2 x i32> <i32 1, i32 0>
-; CHECK-NEXT:    [[TMP4:%.*]] = fsub <2 x float> [[TMP2]], [[TMP3]]
+; CHECK-NEXT:    [[P1:%.*]] = getelementptr float, ptr [[A]], i64 1
----------------
MrSidims wrote:

I would doubt that, as there is no tripple hence cost model is inherited from default TTI.

https://github.com/llvm/llvm-project/pull/240136


More information about the llvm-branch-commits mailing list