expand-masked-expandload.ll | Explore in Territory

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt -S %s -passes=scalarize-masked-mem-intrin -mtriple=x86_64-linux-gnu | FileCheck %s

define <2 x i64> @scalarize_v2i64(ptr %p, <2 x i1> %mask, <2 x i64> %passthru) {
; CHECK-LABEL: @scalarize_v2i64(
; CHECK-NEXT:    [[SCALAR_MASK:%.*]] = bitcast <2 x i1> [[MASK:%.*]] to i2
; CHECK-NEXT:    [[TMP1:%.*]] = and i2 [[SCALAR_MASK]], 1
; CHECK-NEXT:    [[TMP2:%.*]] = icmp ne i2 [[TMP1]], 0
; CHECK-NEXT:    br i1 [[TMP2]], label [[COND_LOAD:%.*]], label [[ELSE:%.*]]
; CHECK:       cond.load:
; CHECK-NEXT:    [[TMP3:%.*]] = load i64, ptr [[P:%.*]], align 1
; CHECK-NEXT:    [[TMP4:%.*]] = insertelement <2 x i64> [[PASSTHRU:%.*]], i64 [[TMP3]], i64 0
; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[P]], i32 1
; CHECK-NEXT:    br label [[ELSE]]
; CHECK:       else:
; CHECK-NEXT:    [[RES_PHI_ELSE:%.*]] = phi <2 x i64> [ [[TMP4]], [[COND_LOAD]] ], [ [[PASSTHRU]], [[TMP0:%.*]] ]
; CHECK-NEXT:    [[PTR_PHI_ELSE:%.*]] = phi ptr [ [[TMP5]], [[COND_LOAD]] ], [ [[P]], [[TMP0]] ]
; CHECK-NEXT:    [[TMP6:%.*]] = and i2 [[SCALAR_MASK]], -2
; CHECK-NEXT:    [[TMP7:%.*]] = icmp ne i2 [[TMP6]], 0
; CHECK-NEXT:    br i1 [[TMP7]], label [[COND_LOAD1:%.*]], label [[ELSE2:%.*]]
; CHECK:       cond.load1:
; CHECK-NEXT:    [[TMP8:%.*]] = load i64, ptr [[PTR_PHI_ELSE]], align 1
; CHECK-NEXT:    [[TMP9:%.*]] = insertelement <2 x i64> [[RES_PHI_ELSE]], i64 [[TMP8]], i64 1
; CHECK-NEXT:    br label [[ELSE2]]
; CHECK:       else2:
; CHECK-NEXT:    [[RES_PHI_ELSE3:%.*]] = phi <2 x i64> [ [[TMP9]], [[COND_LOAD1]] ], [ [[RES_PHI_ELSE]], [[ELSE]] ]
; CHECK-NEXT:    ret <2 x i64> [[RES_PHI_ELSE3]]
;
  %ret = call <2 x i64> @llvm.masked.expandload.v2i64.p0(ptr %p, <2 x i1> %mask, <2 x i64> %passthru)
  ret <2 x i64> %ret
}

define <2 x i64> @scalarize_v2i64_ones_mask(ptr %p, <2 x i64> %passthru) {
; CHECK-LABEL: @scalarize_v2i64_ones_mask(
; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[P:%.*]], i32 0
; CHECK-NEXT:    [[LOAD0:%.*]] = load i64, ptr [[TMP1]], align 1
; CHECK-NEXT:    [[RES0:%.*]] = insertelement <2 x i64> poison, i64 [[LOAD0]], i64 0
; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[P]], i32 1
; CHECK-NEXT:    [[LOAD1:%.*]] = load i64, ptr [[TMP2]], align 1
; CHECK-NEXT:    [[RES1:%.*]] = insertelement <2 x i64> [[RES0]], i64 [[LOAD1]], i64 1
; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <2 x i64> [[RES1]], <2 x i64> [[PASSTHRU:%.*]], <2 x i32> <i32 0, i32 1>
; CHECK-NEXT:    ret <2 x i64> [[TMP3]]
;
  %ret = call <2 x i64> @llvm.masked.expandload.v2i64.p0(ptr %p, <2 x i1> <i1 true, i1 true>, <2 x i64> %passthru)
  ret <2 x i64> %ret
}

define <2 x i64> @scalarize_v2i64_zero_mask(ptr %p, <2 x i64> %passthru) {
; CHECK-LABEL: @scalarize_v2i64_zero_mask(
; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <2 x i64> poison, <2 x i64> [[PASSTHRU:%.*]], <2 x i32> <i32 2, i32 3>
; CHECK-NEXT:    ret <2 x i64> [[TMP1]]
;
  %ret = call <2 x i64> @llvm.masked.expandload.v2i64.p0(ptr %p, <2 x i1> <i1 false, i1 false>, <2 x i64> %passthru)
  ret <2 x i64> %ret
}

define <2 x i64> @scalarize_v2i64_const_mask(ptr %p, <2 x i64> %passthru) {
; CHECK-LABEL: @scalarize_v2i64_const_mask(
; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[P:%.*]], i32 0
; CHECK-NEXT:    [[LOAD1:%.*]] = load i64, ptr [[TMP1]], align 1
; CHECK-NEXT:    [[RES1:%.*]] = insertelement <2 x i64> poison, i64 [[LOAD1]], i64 1
; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <2 x i64> [[RES1]], <2 x i64> [[PASSTHRU:%.*]], <2 x i32> <i32 2, i32 1>
; CHECK-NEXT:    ret <2 x i64> [[TMP2]]
;
  %ret = call <2 x i64> @llvm.masked.expandload.v2i64.p0(ptr %p, <2 x i1> <i1 false, i1 true>, <2 x i64> %passthru)
  ret <2 x i64> %ret
}

declare <2 x i64> @llvm.masked.expandload.v2i64.p0(ptr,  <2 x i1>, <2 x i64>)
llvm/llvm/test/Transforms/ScalarizeMaskedMemIntrin/X86/expand-masked-expandload.ll
