aboutsummaryrefslogtreecommitdiff
path: root/llvm/test/Analysis/BasicAA/scalable-dse-aa.ll
diff options
context:
space:
mode:
Diffstat (limited to 'llvm/test/Analysis/BasicAA/scalable-dse-aa.ll')
-rw-r--r--llvm/test/Analysis/BasicAA/scalable-dse-aa.ll149
1 files changed, 123 insertions, 26 deletions
diff --git a/llvm/test/Analysis/BasicAA/scalable-dse-aa.ll b/llvm/test/Analysis/BasicAA/scalable-dse-aa.ll
index 7e980c9..ffd8259 100644
--- a/llvm/test/Analysis/BasicAA/scalable-dse-aa.ll
+++ b/llvm/test/Analysis/BasicAA/scalable-dse-aa.ll
@@ -1,10 +1,23 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 6
; RUN: opt < %s -aa-pipeline=basic-aa -passes=dse -S | FileCheck %s
define <vscale x 4 x float> @dead_scalable_store(ptr %0) {
; CHECK-LABEL: define <vscale x 4 x float> @dead_scalable_store(
-; CHECK: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> %load.0.16, ptr nonnull %gep.arr.16, i32 1, <vscale x 4 x i1> %mask)
-; CHECK-NOT: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> %load.0.32, ptr nonnull %gep.arr.32, i32 1, <vscale x 4 x i1> %mask)
-; CHECK: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> %load.0.48, ptr nonnull %gep.arr.48, i32 1, <vscale x 4 x i1> %mask)
+; CHECK-SAME: ptr [[TMP0:%.*]]) {
+; CHECK-NEXT: [[ARR:%.*]] = alloca [64 x i32], align 4
+; CHECK-NEXT: [[MASK:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i32(i32 0, i32 4)
+; CHECK-NEXT: [[GEP_0_16:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP0]], i64 16
+; CHECK-NEXT: [[GEP_0_48:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP0]], i64 48
+; CHECK-NEXT: [[GEP_ARR_16:%.*]] = getelementptr inbounds nuw i8, ptr [[ARR]], i64 16
+; CHECK-NEXT: [[GEP_ARR_48:%.*]] = getelementptr inbounds nuw i8, ptr [[ARR]], i64 48
+; CHECK-NEXT: [[LOAD_0_16:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[GEP_0_16]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer)
+; CHECK-NEXT: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> [[LOAD_0_16]], ptr nonnull [[GEP_ARR_16]], i32 1, <vscale x 4 x i1> [[MASK]])
+; CHECK-NEXT: [[LOAD_0_48:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[GEP_0_48]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer)
+; CHECK-NEXT: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> [[LOAD_0_48]], ptr nonnull [[GEP_ARR_48]], i32 1, <vscale x 4 x i1> [[MASK]])
+; CHECK-NEXT: [[FADDOP0:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[GEP_ARR_16]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer)
+; CHECK-NEXT: [[FADDOP1:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[GEP_ARR_48]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer)
+; CHECK-NEXT: [[FADD:%.*]] = fadd <vscale x 4 x float> [[FADDOP0]], [[FADDOP1]]
+; CHECK-NEXT: ret <vscale x 4 x float> [[FADD]]
;
%arr = alloca [64 x i32], align 4
%mask = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i32(i32 0, i32 4)
@@ -34,9 +47,21 @@ define <vscale x 4 x float> @dead_scalable_store(ptr %0) {
define <4 x float> @dead_scalable_store_fixed(ptr %0) {
; CHECK-LABEL: define <4 x float> @dead_scalable_store_fixed(
-; CHECK: call void @llvm.masked.store.v4f32.p0(<4 x float> %load.0.16, ptr nonnull %gep.arr.16, i32 1, <4 x i1> %mask)
-; CHECK-NOT: call void @llvm.masked.store.v4f32.p0(<4 x float> %load.0.32, ptr nonnull %gep.arr.36, i32 1, <4 x i1> %mask2)
-; CHECK: call void @llvm.masked.store.v4f32.p0(<4 x float> %load.0.48, ptr nonnull %gep.arr.48, i32 1, <4 x i1> %mask)
+; CHECK-SAME: ptr [[TMP0:%.*]]) {
+; CHECK-NEXT: [[ARR:%.*]] = alloca [64 x i32], align 4
+; CHECK-NEXT: [[MASK:%.*]] = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i32(i32 0, i32 4)
+; CHECK-NEXT: [[GEP_0_16:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP0]], i64 16
+; CHECK-NEXT: [[GEP_0_48:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP0]], i64 48
+; CHECK-NEXT: [[GEP_ARR_16:%.*]] = getelementptr inbounds nuw i8, ptr [[ARR]], i64 16
+; CHECK-NEXT: [[GEP_ARR_48:%.*]] = getelementptr inbounds nuw i8, ptr [[ARR]], i64 48
+; CHECK-NEXT: [[LOAD_0_16:%.*]] = call <4 x float> @llvm.masked.load.v4f32.p0(ptr nonnull [[GEP_0_16]], i32 1, <4 x i1> [[MASK]], <4 x float> zeroinitializer)
+; CHECK-NEXT: call void @llvm.masked.store.v4f32.p0(<4 x float> [[LOAD_0_16]], ptr nonnull [[GEP_ARR_16]], i32 1, <4 x i1> [[MASK]])
+; CHECK-NEXT: [[LOAD_0_48:%.*]] = call <4 x float> @llvm.masked.load.v4f32.p0(ptr nonnull [[GEP_0_48]], i32 1, <4 x i1> [[MASK]], <4 x float> zeroinitializer)
+; CHECK-NEXT: call void @llvm.masked.store.v4f32.p0(<4 x float> [[LOAD_0_48]], ptr nonnull [[GEP_ARR_48]], i32 1, <4 x i1> [[MASK]])
+; CHECK-NEXT: [[FADDOP0:%.*]] = call <4 x float> @llvm.masked.load.v4f32.p0(ptr nonnull [[GEP_ARR_16]], i32 1, <4 x i1> [[MASK]], <4 x float> zeroinitializer)
+; CHECK-NEXT: [[FADDOP1:%.*]] = call <4 x float> @llvm.masked.load.v4f32.p0(ptr nonnull [[GEP_ARR_48]], i32 1, <4 x i1> [[MASK]], <4 x float> zeroinitializer)
+; CHECK-NEXT: [[FADD:%.*]] = fadd <4 x float> [[FADDOP0]], [[FADDOP1]]
+; CHECK-NEXT: ret <4 x float> [[FADD]]
;
%arr = alloca [64 x i32], align 4
%mask = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i32(i32 0, i32 4)
@@ -67,9 +92,25 @@ define <4 x float> @dead_scalable_store_fixed(ptr %0) {
define <vscale x 4 x float> @scalable_store_partial_overwrite(ptr %0) {
; CHECK-LABEL: define <vscale x 4 x float> @scalable_store_partial_overwrite(
-; CHECK: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> %load.0.16, ptr nonnull %gep.arr.16, i32 1, <vscale x 4 x i1> %mask)
-; CHECK: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> %load.0.30, ptr nonnull %gep.arr.30, i32 1, <vscale x 4 x i1> %mask)
-; CHECK: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> %load.0.48, ptr nonnull %gep.arr.48, i32 1, <vscale x 4 x i1> %mask)
+; CHECK-SAME: ptr [[TMP0:%.*]]) {
+; CHECK-NEXT: [[ARR:%.*]] = alloca [64 x i32], align 4
+; CHECK-NEXT: [[MASK:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i32(i32 0, i32 4)
+; CHECK-NEXT: [[GEP_0_16:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP0]], i64 16
+; CHECK-NEXT: [[GEP_0_30:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP0]], i64 30
+; CHECK-NEXT: [[GEP_0_48:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP0]], i64 48
+; CHECK-NEXT: [[GEP_ARR_16:%.*]] = getelementptr inbounds nuw i8, ptr [[ARR]], i64 16
+; CHECK-NEXT: [[GEP_ARR_30:%.*]] = getelementptr inbounds nuw i8, ptr [[ARR]], i64 30
+; CHECK-NEXT: [[GEP_ARR_48:%.*]] = getelementptr inbounds nuw i8, ptr [[ARR]], i64 48
+; CHECK-NEXT: [[LOAD_0_16:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[GEP_0_16]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer)
+; CHECK-NEXT: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> [[LOAD_0_16]], ptr nonnull [[GEP_ARR_16]], i32 1, <vscale x 4 x i1> [[MASK]])
+; CHECK-NEXT: [[LOAD_0_30:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[GEP_0_30]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer)
+; CHECK-NEXT: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> [[LOAD_0_30]], ptr nonnull [[GEP_ARR_30]], i32 1, <vscale x 4 x i1> [[MASK]])
+; CHECK-NEXT: [[LOAD_0_48:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[GEP_0_48]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer)
+; CHECK-NEXT: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> [[LOAD_0_48]], ptr nonnull [[GEP_ARR_48]], i32 1, <vscale x 4 x i1> [[MASK]])
+; CHECK-NEXT: [[FADDOP0:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[GEP_ARR_16]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer)
+; CHECK-NEXT: [[FADDOP1:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[GEP_ARR_48]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer)
+; CHECK-NEXT: [[FADD:%.*]] = fadd <vscale x 4 x float> [[FADDOP0]], [[FADDOP1]]
+; CHECK-NEXT: ret <vscale x 4 x float> [[FADD]]
;
%arr = alloca [64 x i32], align 4
%mask = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i32(i32 0, i32 4)
@@ -99,9 +140,23 @@ define <vscale x 4 x float> @scalable_store_partial_overwrite(ptr %0) {
define <vscale x 4 x float> @dead_scalable_store_small_mask(ptr %0) {
; CHECK-LABEL: define <vscale x 4 x float> @dead_scalable_store_small_mask(
-; CHECK: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> %load.0.16, ptr nonnull %gep.arr.16, i32 1, <vscale x 4 x i1> %mask)
-; CHECK-NOT: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> %load.0.30, ptr nonnull %gep.arr.30, i32 1, <vscale x 4 x i1> %mask)
-; CHECK: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> %load.0.46, ptr nonnull %gep.arr.46, i32 1, <vscale x 4 x i1> %mask)
+; CHECK-SAME: ptr [[TMP0:%.*]]) {
+; CHECK-NEXT: [[ARR:%.*]] = alloca [64 x i32], align 4
+; CHECK-NEXT: [[MASK:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i32(i32 0, i32 4)
+; CHECK-NEXT: [[GEP_0_16:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP0]], i64 16
+; CHECK-NEXT: [[GEP_0_46:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP0]], i64 46
+; CHECK-NEXT: [[GEP_ARR_16:%.*]] = getelementptr inbounds nuw i8, ptr [[ARR]], i64 16
+; CHECK-NEXT: [[GEP_ARR_46:%.*]] = getelementptr inbounds nuw i8, ptr [[ARR]], i64 46
+; CHECK-NEXT: [[LOAD_0_16:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[GEP_0_16]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer)
+; CHECK-NEXT: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> [[LOAD_0_16]], ptr nonnull [[GEP_ARR_16]], i32 1, <vscale x 4 x i1> [[MASK]])
+; CHECK-NEXT: [[LOAD_0_46:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[GEP_0_46]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer)
+; CHECK-NEXT: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> [[LOAD_0_46]], ptr nonnull [[GEP_ARR_46]], i32 1, <vscale x 4 x i1> [[MASK]])
+; CHECK-NEXT: [[SMALLMASK:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i32(i32 0, i32 2)
+; CHECK-NEXT: [[FADDOP0:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[GEP_ARR_16]], i32 1, <vscale x 4 x i1> [[SMALLMASK]], <vscale x 4 x float> zeroinitializer)
+; CHECK-NEXT: [[FADDOP1:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[GEP_ARR_46]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer)
+; CHECK-NEXT: [[FADD:%.*]] = fadd <vscale x 4 x float> [[FADDOP0]], [[FADDOP1]]
+; CHECK-NEXT: ret <vscale x 4 x float> [[FADD]]
+;
%arr = alloca [64 x i32], align 4
%mask = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i32(i32 0, i32 4)
@@ -131,7 +186,12 @@ define <vscale x 4 x float> @dead_scalable_store_small_mask(ptr %0) {
define <vscale x 4 x float> @dead_scalar_store(ptr noalias %0, ptr %1) {
; CHECK-LABEL: define <vscale x 4 x float> @dead_scalar_store(
-; CHECK-NOT: store i32 20, ptr %gep.1.12
+; CHECK-SAME: ptr noalias [[TMP0:%.*]], ptr [[TMP1:%.*]]) {
+; CHECK-NEXT: [[MASK:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i128(i128 0, i128 4)
+; CHECK-NEXT: [[LOAD_0:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[TMP0]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer)
+; CHECK-NEXT: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> [[LOAD_0]], ptr nonnull [[TMP1]], i32 1, <vscale x 4 x i1> [[MASK]])
+; CHECK-NEXT: [[RETVAL:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[TMP1]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer)
+; CHECK-NEXT: ret <vscale x 4 x float> [[RETVAL]]
;
%mask = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i128(i128 0, i128 4)
%gep.1.12 = getelementptr inbounds nuw i8, ptr %1, i64 12
@@ -144,10 +204,17 @@ define <vscale x 4 x float> @dead_scalar_store(ptr noalias %0, ptr %1) {
}
-; CHECK-LABEL: define <4 x float> @dead_scalable_store_fixed_large_mask(
-; CHECK-NOT: store i32 20, ptr %1
-; CHECK: store i32 50, ptr %gep.5
define <4 x float> @dead_scalable_store_fixed_large_mask(ptr noalias %0, ptr %1) {
+; CHECK-LABEL: define <4 x float> @dead_scalable_store_fixed_large_mask(
+; CHECK-SAME: ptr noalias [[TMP0:%.*]], ptr [[TMP1:%.*]]) {
+; CHECK-NEXT: [[MASK:%.*]] = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i32(i32 0, i32 7)
+; CHECK-NEXT: [[GEP_5:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP1]], i64 5
+; CHECK-NEXT: store i32 50, ptr [[GEP_5]], align 4
+; CHECK-NEXT: [[LOAD_0:%.*]] = call <4 x float> @llvm.masked.load.v4f32.p0(ptr nonnull [[TMP0]], i32 1, <4 x i1> [[MASK]], <4 x float> zeroinitializer)
+; CHECK-NEXT: call void @llvm.masked.store.v4f32.p0(<4 x float> [[LOAD_0]], ptr nonnull [[TMP1]], i32 1, <4 x i1> [[MASK]])
+; CHECK-NEXT: [[RETVAL:%.*]] = call <4 x float> @llvm.masked.load.v4f32.p0(ptr nonnull [[TMP1]], i32 1, <4 x i1> [[MASK]], <4 x float> zeroinitializer)
+; CHECK-NEXT: ret <4 x float> [[RETVAL]]
+;
%mask = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i32(i32 0, i32 7)
store i32 20, ptr %1
@@ -164,8 +231,16 @@ define <4 x float> @dead_scalable_store_fixed_large_mask(ptr noalias %0, ptr %1)
; This get active lane mask may cover 4 or 8 integers
define <vscale x 4 x float> @mask_gt_minimum_num_elts(ptr noalias %0, ptr %1) {
; CHECK-LABEL: define <vscale x 4 x float> @mask_gt_minimum_num_elts(
-; CHECK: store i32 10, ptr %gep.1.12
-; CHECK: store i32 20, ptr %gep.1.28
+; CHECK-SAME: ptr noalias [[TMP0:%.*]], ptr [[TMP1:%.*]]) {
+; CHECK-NEXT: [[MASK:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i32(i32 0, i32 8)
+; CHECK-NEXT: [[GEP_1_12:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP1]], i64 12
+; CHECK-NEXT: store i32 10, ptr [[GEP_1_12]], align 4
+; CHECK-NEXT: [[GEP_1_28:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP1]], i64 28
+; CHECK-NEXT: store i32 20, ptr [[GEP_1_28]], align 4
+; CHECK-NEXT: [[LOAD_0:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[TMP0]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer)
+; CHECK-NEXT: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> [[LOAD_0]], ptr nonnull [[TMP1]], i32 1, <vscale x 4 x i1> [[MASK]])
+; CHECK-NEXT: [[RETVAL:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[TMP1]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer)
+; CHECK-NEXT: ret <vscale x 4 x float> [[RETVAL]]
;
%mask = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i32(i32 0, i32 8)
%gep.1.12 = getelementptr inbounds nuw i8, ptr %1, i64 12
@@ -182,7 +257,13 @@ define <vscale x 4 x float> @mask_gt_minimum_num_elts(ptr noalias %0, ptr %1) {
; Don't do anything if the mask's Op1 < Op0
define <vscale x 4 x float> @active_lane_mask_lt(ptr noalias %0, ptr %1) {
; CHECK-LABEL: define <vscale x 4 x float> @active_lane_mask_lt(
-; CHECK: store i32 20, ptr %1
+; CHECK-SAME: ptr noalias [[TMP0:%.*]], ptr [[TMP1:%.*]]) {
+; CHECK-NEXT: [[MASK:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i32(i32 4, i32 2)
+; CHECK-NEXT: store i32 20, ptr [[TMP1]], align 4
+; CHECK-NEXT: [[LOAD_0:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[TMP0]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer)
+; CHECK-NEXT: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> [[LOAD_0]], ptr nonnull [[TMP1]], i32 1, <vscale x 4 x i1> [[MASK]])
+; CHECK-NEXT: [[RETVAL:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[TMP1]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer)
+; CHECK-NEXT: ret <vscale x 4 x float> [[RETVAL]]
;
%mask = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i32(i32 4, i32 2)
store i32 20, ptr %1
@@ -196,7 +277,13 @@ define <vscale x 4 x float> @active_lane_mask_lt(ptr noalias %0, ptr %1) {
; Don't do anything if the mask's Op1 == Op0
define <vscale x 4 x float> @active_lane_mask_eq(ptr noalias %0, ptr %1) {
; CHECK-LABEL: define <vscale x 4 x float> @active_lane_mask_eq(
-; CHECK: store i32 20, ptr %1
+; CHECK-SAME: ptr noalias [[TMP0:%.*]], ptr [[TMP1:%.*]]) {
+; CHECK-NEXT: [[MASK:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i32(i32 2, i32 2)
+; CHECK-NEXT: store i32 20, ptr [[TMP1]], align 4
+; CHECK-NEXT: [[LOAD_0:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[TMP0]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer)
+; CHECK-NEXT: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> [[LOAD_0]], ptr nonnull [[TMP1]], i32 1, <vscale x 4 x i1> [[MASK]])
+; CHECK-NEXT: [[RETVAL:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[TMP1]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer)
+; CHECK-NEXT: ret <vscale x 4 x float> [[RETVAL]]
;
%mask = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i32(i32 2, i32 2)
store i32 20, ptr %1
@@ -209,8 +296,14 @@ define <vscale x 4 x float> @active_lane_mask_eq(ptr noalias %0, ptr %1) {
define <vscale x 16 x i8> @scalar_stores_small_mask(ptr noalias %0, ptr %1) {
; CHECK-LABEL: define <vscale x 16 x i8> @scalar_stores_small_mask(
-; CHECK-NOT: store i8 60, ptr %gep.1.6
-; CHECK: store i8 120, ptr %gep.1.8
+; CHECK-SAME: ptr noalias [[TMP0:%.*]], ptr [[TMP1:%.*]]) {
+; CHECK-NEXT: [[MASK:%.*]] = call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i8(i8 0, i8 7)
+; CHECK-NEXT: [[GEP_1_8:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP1]], i64 8
+; CHECK-NEXT: store i8 120, ptr [[GEP_1_8]], align 1
+; CHECK-NEXT: [[LOAD_0:%.*]] = call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0(ptr nonnull [[TMP0]], i32 1, <vscale x 16 x i1> [[MASK]], <vscale x 16 x i8> zeroinitializer)
+; CHECK-NEXT: call void @llvm.masked.store.nxv16i8.p0(<vscale x 16 x i8> [[LOAD_0]], ptr [[TMP1]], i32 1, <vscale x 16 x i1> [[MASK]])
+; CHECK-NEXT: [[RETVAL:%.*]] = call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0(ptr [[TMP1]], i32 1, <vscale x 16 x i1> [[MASK]], <vscale x 16 x i8> zeroinitializer)
+; CHECK-NEXT: ret <vscale x 16 x i8> [[RETVAL]]
;
%mask = call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i8.i8(i8 0, i8 7)
%gep.1.6 = getelementptr inbounds nuw i8, ptr %1, i64 6
@@ -226,10 +319,14 @@ define <vscale x 16 x i8> @scalar_stores_small_mask(ptr noalias %0, ptr %1) {
define <vscale x 4 x float> @dead_scalar_store_offset(ptr noalias %0, ptr %1) {
; CHECK-LABEL: define <vscale x 4 x float> @dead_scalar_store_offset(
-; CHECK-NOT: store i32 10, ptr %gep.1.0
-; CHECK-NOT: store i32 20, ptr %gep.1.4
-; CHECK-NOT: store i32 30, ptr %gep.1.8
-; CHECK: store i32 40, ptr %gep.1.12
+; CHECK-SAME: ptr noalias [[TMP0:%.*]], ptr [[TMP1:%.*]]) {
+; CHECK-NEXT: [[MASK:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i32(i32 1, i32 4)
+; CHECK-NEXT: [[GEP_1_12:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP1]], i64 12
+; CHECK-NEXT: store i32 40, ptr [[GEP_1_12]], align 4
+; CHECK-NEXT: [[LOAD_0:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[TMP0]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer)
+; CHECK-NEXT: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> [[LOAD_0]], ptr nonnull [[TMP1]], i32 1, <vscale x 4 x i1> [[MASK]])
+; CHECK-NEXT: [[RETVAL:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[TMP1]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer)
+; CHECK-NEXT: ret <vscale x 4 x float> [[RETVAL]]
;
%mask = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i32(i32 1, i32 4)
%gep.1.0 = getelementptr inbounds nuw i8, ptr %1, i64 0
generated by cgit v1.1 at 2025-10-16 19:21:45 +0000