test/Transforms/LowerMatrixIntrinsics/transpose-opts-lifting-constant-folds.ll

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
; RUN: opt -p lower-matrix-intrinsics -S %s | FileCheck %s

target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"

define <8 x float> @transpose_constant_fold_fadd_AT_BT() {
; CHECK-LABEL: define <8 x float> @transpose_constant_fold_fadd_AT_BT() {
; CHECK-NEXT:  [[ENTRY:.*:]]
; CHECK-NEXT:    ret <8 x float> splat (float 2.000000e+00)
;
entry:
  %t = tail call <8 x float> @llvm.matrix.transpose.v8f32(<8 x float> splat (float 1.0), i32 8, i32 1)
  %f = fadd <8 x float> %t, %t
  ret <8 x float> %f
}

define <8 x float> @transpose_constant_fold_fmul_A_k() {
; CHECK-LABEL: define <8 x float> @transpose_constant_fold_fmul_A_k() {
; CHECK-NEXT:  [[ENTRY:.*:]]
; CHECK-NEXT:    [[SPLAT:%.*]] = shufflevector <8 x float> splat (float 3.000000e+00), <8 x float> poison, <8 x i32> zeroinitializer
; CHECK-NEXT:    [[SPLIT:%.*]] = shufflevector <8 x float> [[SPLAT]], <8 x float> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
; CHECK-NEXT:    [[SPLIT1:%.*]] = shufflevector <8 x float> [[SPLAT]], <8 x float> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
; CHECK-NEXT:    [[TMP0:%.*]] = fmul <4 x float> splat (float 1.000000e+00), [[SPLIT]]
; CHECK-NEXT:    [[TMP1:%.*]] = fmul <4 x float> splat (float 1.000000e+00), [[SPLIT1]]
; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <4 x float> [[TMP0]], <4 x float> [[TMP1]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
; CHECK-NEXT:    ret <8 x float> [[TMP2]]
;
entry:
  %t.1 = tail call <8 x float> @llvm.matrix.transpose.v8f32(<8 x float> splat (float 1.0), i32 4, i32 2)
  %splat = shufflevector <8 x float> splat (float 3.0), <8 x float> poison, <8 x i32> zeroinitializer
  %m = fmul <8 x float> %t.1, %splat
  %t.2 = tail call <8 x float> @llvm.matrix.transpose.v8f32(<8 x float> %m, i32 2, i32 4)
  ret <8 x float> %t.2
}

; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
declare <8 x float> @llvm.matrix.transpose.v8f32(<8 x float>, i32 immarg, i32 immarg) #0

attributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }