[ARM] MVE ICmp costing tests. NFC
GitOrigin-RevId: 1fbb3287fcda77d69a405dea17d7d11e1d3ac152
diff --git a/test/Analysis/CostModel/ARM/mve-cmp.ll b/test/Analysis/CostModel/ARM/mve-cmp.ll
new file mode 100644
index 0000000..0ca8bc0
--- /dev/null
+++ b/test/Analysis/CostModel/ARM/mve-cmp.ll
@@ -0,0 +1,78 @@
+; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py
+; RUN: opt -cost-model -analyze -mtriple=thumbv8.1m.main-none-eabi -mattr=+mve < %s | FileCheck %s --check-prefixes=CHECK
+; RUN: opt -cost-model -analyze -mtriple=thumbv8.1m.main-none-eabi -mattr=+mve.fp < %s | FileCheck %s --check-prefixes=CHECK
+
+define void @icmp() {
+; CHECK-LABEL: 'icmp'
+; CHECK-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %v2i8 = icmp slt <2 x i8> undef, undef
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v4i8 = icmp slt <4 x i8> undef, undef
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v8i8 = icmp slt <8 x i8> undef, undef
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v16i8 = icmp slt <16 x i8> undef, undef
+; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %v32i8 = icmp slt <32 x i8> undef, undef
+; CHECK-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %v2i16 = icmp slt <2 x i16> undef, undef
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v4i16 = icmp slt <4 x i16> undef, undef
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v8i16 = icmp slt <8 x i16> undef, undef
+; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %v16i16 = icmp slt <16 x i16> undef, undef
+; CHECK-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %v2i32 = icmp slt <2 x i32> undef, undef
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v4i32 = icmp slt <4 x i32> undef, undef
+; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %v8i32 = icmp slt <8 x i32> undef, undef
+; CHECK-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %v16i32 = icmp slt <16 x i32> undef, undef
+; CHECK-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %v2i64 = icmp slt <2 x i64> undef, undef
+; CHECK-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %v4i64 = icmp slt <4 x i64> undef, undef
+; CHECK-NEXT: Cost Model: Found an estimated cost of 144 for instruction: %v8i64 = icmp slt <8 x i64> undef, undef
+; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+ %v2i8 = icmp slt <2 x i8> undef, undef
+ %v4i8 = icmp slt <4 x i8> undef, undef
+ %v8i8 = icmp slt <8 x i8> undef, undef
+ %v16i8 = icmp slt <16 x i8> undef, undef
+ %v32i8 = icmp slt <32 x i8> undef, undef
+
+ %v2i16 = icmp slt <2 x i16> undef, undef
+ %v4i16 = icmp slt <4 x i16> undef, undef
+ %v8i16 = icmp slt <8 x i16> undef, undef
+ %v16i16 = icmp slt <16 x i16> undef, undef
+
+ %v2i32 = icmp slt <2 x i32> undef, undef
+ %v4i32 = icmp slt <4 x i32> undef, undef
+ %v8i32 = icmp slt <8 x i32> undef, undef
+ %v16i32 = icmp slt <16 x i32> undef, undef
+
+ %v2i64 = icmp slt <2 x i64> undef, undef
+ %v4i64 = icmp slt <4 x i64> undef, undef
+ %v8i64 = icmp slt <8 x i64> undef, undef
+
+ ret void
+}
+
+define void @fcmp() {
+; CHECK-LABEL: 'fcmp'
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v2f16 = fcmp olt <2 x half> undef, undef
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v4f16 = fcmp olt <4 x half> undef, undef
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v8f16 = fcmp olt <8 x half> undef, undef
+; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %v16f16 = fcmp olt <16 x half> undef, undef
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v2f32 = fcmp olt <2 x float> undef, undef
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v4f32 = fcmp olt <4 x float> undef, undef
+; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %v8f32 = fcmp olt <8 x float> undef, undef
+; CHECK-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %v16f32 = fcmp olt <16 x float> undef, undef
+; CHECK-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %v2f64 = fcmp olt <2 x double> undef, undef
+; CHECK-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %v4f64 = fcmp olt <4 x double> undef, undef
+; CHECK-NEXT: Cost Model: Found an estimated cost of 144 for instruction: %v8f64 = fcmp olt <8 x double> undef, undef
+; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+ %v2f16 = fcmp olt <2 x half> undef, undef
+ %v4f16 = fcmp olt <4 x half> undef, undef
+ %v8f16 = fcmp olt <8 x half> undef, undef
+ %v16f16 = fcmp olt <16 x half> undef, undef
+
+ %v2f32 = fcmp olt <2 x float> undef, undef
+ %v4f32 = fcmp olt <4 x float> undef, undef
+ %v8f32 = fcmp olt <8 x float> undef, undef
+ %v16f32 = fcmp olt <16 x float> undef, undef
+
+ %v2f64 = fcmp olt <2 x double> undef, undef
+ %v4f64 = fcmp olt <4 x double> undef, undef
+ %v8f64 = fcmp olt <8 x double> undef, undef
+
+ ret void
+}
diff --git a/test/CodeGen/ARM/vselect_imax.ll b/test/CodeGen/ARM/vselect_imax.ll
index 03e212c..31f19e0 100644
--- a/test/CodeGen/ARM/vselect_imax.ll
+++ b/test/CodeGen/ARM/vselect_imax.ll
@@ -21,6 +21,7 @@
; CHECK: vmin.s16
; CHECK: vmin.s16
; COST: func_blend10
+; COST: cost of 2 {{.*}} icmp
; COST: cost of 2 {{.*}} select
%r = select %T1_10 %c, %T0_10 %v0, %T0_10 %v1
store %T0_10 %r, %T0_10* %storeaddr
@@ -37,6 +38,7 @@
; CHECK: vmin.s32
; CHECK: vmin.s32
; COST: func_blend14
+; COST: cost of 2 {{.*}} icmp
; COST: cost of 2 {{.*}} select
%r = select %T1_14 %c, %T0_14 %v0, %T0_14 %v1
store %T0_14 %r, %T0_14* %storeaddr
@@ -53,6 +55,7 @@
%v1 = load %T0_15, %T0_15* %loadaddr2
%c = icmp slt %T0_15 %v0, %v1
; COST: func_blend15
+; COST: cost of 4 {{.*}} icmp
; COST: cost of 4 {{.*}} select
%r = select %T1_15 %c, %T0_15 %v0, %T0_15 %v1
store %T0_15 %r, %T0_15* %storeaddr
@@ -127,6 +130,7 @@
%v1 = load %T0_18, %T0_18* %loadaddr2
%c = icmp slt %T0_18 %v0, %v1
; COST: func_blend18
+; COST: cost of 2 {{.*}} icmp
; COST: cost of 19 {{.*}} select
%r = select %T1_18 %c, %T0_18 %v0, %T0_18 %v1
store %T0_18 %r, %T0_18* %storeaddr
@@ -256,6 +260,7 @@
%v1 = load %T0_19, %T0_19* %loadaddr2
%c = icmp slt %T0_19 %v0, %v1
; COST: func_blend19
+; COST: cost of 4 {{.*}} icmp
; COST: cost of 50 {{.*}} select
%r = select %T1_19 %c, %T0_19 %v0, %T0_19 %v1
store %T0_19 %r, %T0_19* %storeaddr
@@ -511,6 +516,7 @@
%v1 = load %T0_20, %T0_20* %loadaddr2
%c = icmp slt %T0_20 %v0, %v1
; COST: func_blend20
+; COST: cost of 8 {{.*}} icmp
; COST: cost of 100 {{.*}} select
%r = select %T1_20 %c, %T0_20 %v0, %T0_20 %v1
store %T0_20 %r, %T0_20* %storeaddr
diff --git a/test/Transforms/LoopVectorize/ARM/mve-icmpcost.ll b/test/Transforms/LoopVectorize/ARM/mve-icmpcost.ll
new file mode 100644
index 0000000..45730ab
--- /dev/null
+++ b/test/Transforms/LoopVectorize/ARM/mve-icmpcost.ll
@@ -0,0 +1,240 @@
+; RUN: opt -loop-vectorize -debug-only=loop-vectorize -disable-output < %s 2>&1 | FileCheck %s
+; REQUIRES: asserts
+
+target datalayout = "e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64"
+target triple = "thumbv8.1m.main-arm-none-eabi"
+
+; CHECK: LV: Found an estimated cost of 0 for VF 1 For instruction: %i.016 = phi i32 [ 0, %for.body.lr.ph ], [ %inc, %for.inc ]
+; CHECK: LV: Found an estimated cost of 0 for VF 1 For instruction: %arrayidx = getelementptr inbounds i16, i16* %s, i32 %i.016
+; CHECK: LV: Found an estimated cost of 1 for VF 1 For instruction: %1 = load i16, i16* %arrayidx, align 2
+; CHECK: LV: Found an estimated cost of 0 for VF 1 For instruction: %conv = sext i16 %1 to i32
+; CHECK: LV: Found an estimated cost of 1 for VF 1 For instruction: %cmp2 = icmp sgt i32 %conv, %conv1
+; CHECK: LV: Found an estimated cost of 0 for VF 1 For instruction: br i1 %cmp2, label %if.then, label %for.inc
+; CHECK: LV: Found an estimated cost of 1 for VF 1 For instruction: %conv6 = add i16 %1, %0
+; CHECK: LV: Found an estimated cost of 0 for VF 1 For instruction: %arrayidx7 = getelementptr inbounds i16, i16* %d, i32 %i.016
+; CHECK: LV: Found an estimated cost of 1 for VF 1 For instruction: store i16 %conv6, i16* %arrayidx7, align 2
+; CHECK: LV: Found an estimated cost of 0 for VF 1 For instruction: br label %for.inc
+; CHECK: LV: Found an estimated cost of 1 for VF 1 For instruction: %inc = add nuw nsw i32 %i.016, 1
+; CHECK: LV: Found an estimated cost of 1 for VF 1 For instruction: %exitcond.not = icmp eq i32 %inc, %n
+; CHECK: LV: Found an estimated cost of 0 for VF 1 For instruction: br i1 %exitcond.not, label %for.cond.cleanup.loopexit, label %for.body
+; CHECK: LV: Scalar loop costs: 5.
+; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction: %i.016 = phi i32 [ 0, %for.body.lr.ph ], [ %inc, %for.inc ]
+; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction: %arrayidx = getelementptr inbounds i16, i16* %s, i32 %i.016
+; CHECK: LV: Found an estimated cost of 10 for VF 2 For instruction: %1 = load i16, i16* %arrayidx, align 2
+; CHECK: LV: Found an estimated cost of 4 for VF 2 For instruction: %conv = sext i16 %1 to i32
+; CHECK: LV: Found an estimated cost of 12 for VF 2 For instruction: %cmp2 = icmp sgt i32 %conv, %conv1
+; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction: br i1 %cmp2, label %if.then, label %for.inc
+; CHECK: LV: Found an estimated cost of 14 for VF 2 For instruction: %conv6 = add i16 %1, %0
+; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction: %arrayidx7 = getelementptr inbounds i16, i16* %d, i32 %i.016
+; CHECK: LV: Found an estimated cost of 16 for VF 2 For instruction: store i16 %conv6, i16* %arrayidx7, align 2
+; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction: br label %for.inc
+; CHECK: LV: Found an estimated cost of 1 for VF 2 For instruction: %inc = add nuw nsw i32 %i.016, 1
+; CHECK: LV: Found an estimated cost of 1 for VF 2 For instruction: %exitcond.not = icmp eq i32 %inc, %n
+; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction: br i1 %exitcond.not, label %for.cond.cleanup.loopexit, label %for.body
+; CHECK: LV: Vector loop of width 2 costs: 29.
+; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction: %i.016 = phi i32 [ 0, %for.body.lr.ph ], [ %inc, %for.inc ]
+; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction: %arrayidx = getelementptr inbounds i16, i16* %s, i32 %i.016
+; CHECK: LV: Found an estimated cost of 2 for VF 4 For instruction: %1 = load i16, i16* %arrayidx, align 2
+; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction: %conv = sext i16 %1 to i32
+; CHECK: LV: Found an estimated cost of 2 for VF 4 For instruction: %cmp2 = icmp sgt i32 %conv, %conv1
+; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction: br i1 %cmp2, label %if.then, label %for.inc
+; CHECK: LV: Found an estimated cost of 2 for VF 4 For instruction: %conv6 = add i16 %1, %0
+; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction: %arrayidx7 = getelementptr inbounds i16, i16* %d, i32 %i.016
+; CHECK: LV: Found an estimated cost of 2 for VF 4 For instruction: store i16 %conv6, i16* %arrayidx7, align 2
+; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction: br label %for.inc
+; CHECK: LV: Found an estimated cost of 1 for VF 4 For instruction: %inc = add nuw nsw i32 %i.016, 1
+; CHECK: LV: Found an estimated cost of 1 for VF 4 For instruction: %exitcond.not = icmp eq i32 %inc, %n
+; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction: br i1 %exitcond.not, label %for.cond.cleanup.loopexit, label %for.body
+; CHECK: LV: Vector loop of width 4 costs: 2.
+; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction: %i.016 = phi i32 [ 0, %for.body.lr.ph ], [ %inc, %for.inc ]
+; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction: %arrayidx = getelementptr inbounds i16, i16* %s, i32 %i.016
+; CHECK: LV: Found an estimated cost of 2 for VF 8 For instruction: %1 = load i16, i16* %arrayidx, align 2
+; CHECK: LV: Found an estimated cost of 2 for VF 8 For instruction: %conv = sext i16 %1 to i32
+; CHECK: LV: Found an estimated cost of 4 for VF 8 For instruction: %cmp2 = icmp sgt i32 %conv, %conv1
+; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction: br i1 %cmp2, label %if.then, label %for.inc
+; CHECK: LV: Found an estimated cost of 2 for VF 8 For instruction: %conv6 = add i16 %1, %0
+; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction: %arrayidx7 = getelementptr inbounds i16, i16* %d, i32 %i.016
+; CHECK: LV: Found an estimated cost of 2 for VF 8 For instruction: store i16 %conv6, i16* %arrayidx7, align 2
+; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction: br label %for.inc
+; CHECK: LV: Found an estimated cost of 1 for VF 8 For instruction: %inc = add nuw nsw i32 %i.016, 1
+; CHECK: LV: Found an estimated cost of 1 for VF 8 For instruction: %exitcond.not = icmp eq i32 %inc, %n
+; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction: br i1 %exitcond.not, label %for.cond.cleanup.loopexit, label %for.body
+; CHECK: LV: Vector loop of width 8 costs: 1.
+; CHECK: LV: Selecting VF: 8.
+define void @expensive_icmp(i16* noalias nocapture %d, i16* nocapture readonly %s, i32 %n, i16 zeroext %m) #0 {
+entry:
+ %cmp15 = icmp sgt i32 %n, 0
+ br i1 %cmp15, label %for.body.lr.ph, label %for.cond.cleanup
+
+for.body.lr.ph: ; preds = %entry
+ %conv1 = zext i16 %m to i32
+ %0 = trunc i32 %n to i16
+ br label %for.body
+
+for.cond.cleanup: ; preds = %for.inc, %entry
+ ret void
+
+for.body: ; preds = %for.body.lr.ph, %for.inc
+ %i.016 = phi i32 [ 0, %for.body.lr.ph ], [ %inc, %for.inc ]
+ %arrayidx = getelementptr inbounds i16, i16* %s, i32 %i.016
+ %1 = load i16, i16* %arrayidx, align 2
+ %conv = sext i16 %1 to i32
+ %cmp2 = icmp sgt i32 %conv, %conv1
+ br i1 %cmp2, label %if.then, label %for.inc
+
+if.then: ; preds = %for.body
+ %conv6 = add i16 %1, %0
+ %arrayidx7 = getelementptr inbounds i16, i16* %d, i32 %i.016
+ store i16 %conv6, i16* %arrayidx7, align 2
+ br label %for.inc
+
+for.inc: ; preds = %for.body, %if.then
+ %inc = add nuw nsw i32 %i.016, 1
+ %exitcond.not = icmp eq i32 %inc, %n
+ br i1 %exitcond.not, label %for.cond.cleanup, label %for.body
+}
+
+; CHECK: LV: Found an estimated cost of 0 for VF 1 For instruction: %blkCnt.012 = phi i32 [ %dec, %while.body ], [ %blockSize, %while.body.preheader ]
+; CHECK: LV: Found an estimated cost of 0 for VF 1 For instruction: %pSrcA.addr.011 = phi i8* [ %incdec.ptr, %while.body ], [ %pSrcA, %while.body.preheader ]
+; CHECK: LV: Found an estimated cost of 0 for VF 1 For instruction: %pDst.addr.010 = phi i8* [ %incdec.ptr5, %while.body ], [ %pDst, %while.body.preheader ]
+; CHECK: LV: Found an estimated cost of 0 for VF 1 For instruction: %pSrcB.addr.09 = phi i8* [ %incdec.ptr2, %while.body ], [ %pSrcB, %while.body.preheader ]
+; CHECK: LV: Found an estimated cost of 0 for VF 1 For instruction: %incdec.ptr = getelementptr inbounds i8, i8* %pSrcA.addr.011, i32 1
+; CHECK: LV: Found an estimated cost of 1 for VF 1 For instruction: %0 = load i8, i8* %pSrcA.addr.011, align 1
+; CHECK: LV: Found an estimated cost of 0 for VF 1 For instruction: %conv1 = sext i8 %0 to i32
+; CHECK: LV: Found an estimated cost of 0 for VF 1 For instruction: %incdec.ptr2 = getelementptr inbounds i8, i8* %pSrcB.addr.09, i32 1
+; CHECK: LV: Found an estimated cost of 1 for VF 1 For instruction: %1 = load i8, i8* %pSrcB.addr.09, align 1
+; CHECK: LV: Found an estimated cost of 0 for VF 1 For instruction: %conv3 = sext i8 %1 to i32
+; CHECK: LV: Found an estimated cost of 1 for VF 1 For instruction: %mul = mul nsw i32 %conv3, %conv1
+; CHECK: LV: Found an estimated cost of 1 for VF 1 For instruction: %shr = ashr i32 %mul, 7
+; CHECK: LV: Found an estimated cost of 1 for VF 1 For instruction: %2 = icmp slt i32 %shr, 127
+; CHECK: LV: Found an estimated cost of 1 for VF 1 For instruction: %spec.select.i = select i1 %2, i32 %shr, i32 127
+; CHECK: LV: Found an estimated cost of 0 for VF 1 For instruction: %conv4 = trunc i32 %spec.select.i to i8
+; CHECK: LV: Found an estimated cost of 0 for VF 1 For instruction: %incdec.ptr5 = getelementptr inbounds i8, i8* %pDst.addr.010, i32 1
+; CHECK: LV: Found an estimated cost of 1 for VF 1 For instruction: store i8 %conv4, i8* %pDst.addr.010, align 1
+; CHECK: LV: Found an estimated cost of 1 for VF 1 For instruction: %dec = add i32 %blkCnt.012, -1
+; CHECK: LV: Found an estimated cost of 1 for VF 1 For instruction: %cmp.not = icmp eq i32 %dec, 0
+; CHECK: LV: Found an estimated cost of 0 for VF 1 For instruction: br i1 %cmp.not, label %while.end.loopexit, label %while.body
+; CHECK: LV: Scalar loop costs: 9.
+; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction: %blkCnt.012 = phi i32 [ %dec, %while.body ], [ %blockSize, %while.body.preheader ]
+; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction: %pSrcA.addr.011 = phi i8* [ %incdec.ptr, %while.body ], [ %pSrcA, %while.body.preheader ]
+; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction: %pDst.addr.010 = phi i8* [ %incdec.ptr5, %while.body ], [ %pDst, %while.body.preheader ]
+; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction: %pSrcB.addr.09 = phi i8* [ %incdec.ptr2, %while.body ], [ %pSrcB, %while.body.preheader ]
+; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction: %incdec.ptr = getelementptr inbounds i8, i8* %pSrcA.addr.011, i32 1
+; CHECK: LV: Found an estimated cost of 10 for VF 2 For instruction: %0 = load i8, i8* %pSrcA.addr.011, align 1
+; CHECK: LV: Found an estimated cost of 4 for VF 2 For instruction: %conv1 = sext i8 %0 to i32
+; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction: %incdec.ptr2 = getelementptr inbounds i8, i8* %pSrcB.addr.09, i32 1
+; CHECK: LV: Found an estimated cost of 10 for VF 2 For instruction: %1 = load i8, i8* %pSrcB.addr.09, align 1
+; CHECK: LV: Found an estimated cost of 4 for VF 2 For instruction: %conv3 = sext i8 %1 to i32
+; CHECK: LV: Found an estimated cost of 14 for VF 2 For instruction: %mul = mul nsw i32 %conv3, %conv1
+; CHECK: LV: Found an estimated cost of 10 for VF 2 For instruction: %shr = ashr i32 %mul, 7
+; CHECK: LV: Found an estimated cost of 12 for VF 2 For instruction: %2 = icmp slt i32 %shr, 127
+; CHECK: LV: Found an estimated cost of 12 for VF 2 For instruction: %spec.select.i = select i1 %2, i32 %shr, i32 127
+; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction: %conv4 = trunc i32 %spec.select.i to i8
+; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction: %incdec.ptr5 = getelementptr inbounds i8, i8* %pDst.addr.010, i32 1
+; CHECK: LV: Found an estimated cost of 10 for VF 2 For instruction: store i8 %conv4, i8* %pDst.addr.010, align 1
+; CHECK: LV: Found an estimated cost of 1 for VF 2 For instruction: %dec = add i32 %blkCnt.012, -1
+; CHECK: LV: Found an estimated cost of 1 for VF 2 For instruction: %cmp.not = icmp eq i32 %dec, 0
+; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction: br i1 %cmp.not, label %while.end.loopexit, label %while.body
+; CHECK: LV: Vector loop of width 2 costs: 44.
+; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction: %blkCnt.012 = phi i32 [ %dec, %while.body ], [ %blockSize, %while.body.preheader ]
+; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction: %pSrcA.addr.011 = phi i8* [ %incdec.ptr, %while.body ], [ %pSrcA, %while.body.preheader ]
+; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction: %pDst.addr.010 = phi i8* [ %incdec.ptr5, %while.body ], [ %pDst, %while.body.preheader ]
+; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction: %pSrcB.addr.09 = phi i8* [ %incdec.ptr2, %while.body ], [ %pSrcB, %while.body.preheader ]
+; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction: %incdec.ptr = getelementptr inbounds i8, i8* %pSrcA.addr.011, i32 1
+; CHECK: LV: Found an estimated cost of 2 for VF 4 For instruction: %0 = load i8, i8* %pSrcA.addr.011, align 1
+; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction: %conv1 = sext i8 %0 to i32
+; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction: %incdec.ptr2 = getelementptr inbounds i8, i8* %pSrcB.addr.09, i32 1
+; CHECK: LV: Found an estimated cost of 2 for VF 4 For instruction: %1 = load i8, i8* %pSrcB.addr.09, align 1
+; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction: %conv3 = sext i8 %1 to i32
+; CHECK: LV: Found an estimated cost of 2 for VF 4 For instruction: %mul = mul nsw i32 %conv3, %conv1
+; CHECK: LV: Found an estimated cost of 2 for VF 4 For instruction: %shr = ashr i32 %mul, 7
+; CHECK: LV: Found an estimated cost of 2 for VF 4 For instruction: %2 = icmp slt i32 %shr, 127
+; CHECK: LV: Found an estimated cost of 2 for VF 4 For instruction: %spec.select.i = select i1 %2, i32 %shr, i32 127
+; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction: %conv4 = trunc i32 %spec.select.i to i8
+; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction: %incdec.ptr5 = getelementptr inbounds i8, i8* %pDst.addr.010, i32 1
+; CHECK: LV: Found an estimated cost of 2 for VF 4 For instruction: store i8 %conv4, i8* %pDst.addr.010, align 1
+; CHECK: LV: Found an estimated cost of 1 for VF 4 For instruction: %dec = add i32 %blkCnt.012, -1
+; CHECK: LV: Found an estimated cost of 1 for VF 4 For instruction: %cmp.not = icmp eq i32 %dec, 0
+; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction: br i1 %cmp.not, label %while.end.loopexit, label %while.body
+; CHECK: LV: Vector loop of width 4 costs: 4.
+; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction: %blkCnt.012 = phi i32 [ %dec, %while.body ], [ %blockSize, %while.body.preheader ]
+; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction: %pSrcA.addr.011 = phi i8* [ %incdec.ptr, %while.body ], [ %pSrcA, %while.body.preheader ]
+; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction: %pDst.addr.010 = phi i8* [ %incdec.ptr5, %while.body ], [ %pDst, %while.body.preheader ]
+; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction: %pSrcB.addr.09 = phi i8* [ %incdec.ptr2, %while.body ], [ %pSrcB, %while.body.preheader ]
+; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction: %incdec.ptr = getelementptr inbounds i8, i8* %pSrcA.addr.011, i32 1
+; CHECK: LV: Found an estimated cost of 2 for VF 8 For instruction: %0 = load i8, i8* %pSrcA.addr.011, align 1
+; CHECK: LV: Found an estimated cost of 2 for VF 8 For instruction: %conv1 = sext i8 %0 to i32
+; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction: %incdec.ptr2 = getelementptr inbounds i8, i8* %pSrcB.addr.09, i32 1
+; CHECK: LV: Found an estimated cost of 2 for VF 8 For instruction: %1 = load i8, i8* %pSrcB.addr.09, align 1
+; CHECK: LV: Found an estimated cost of 2 for VF 8 For instruction: %conv3 = sext i8 %1 to i32
+; CHECK: LV: Found an estimated cost of 4 for VF 8 For instruction: %mul = mul nsw i32 %conv3, %conv1
+; CHECK: LV: Found an estimated cost of 4 for VF 8 For instruction: %shr = ashr i32 %mul, 7
+; CHECK: LV: Found an estimated cost of 4 for VF 8 For instruction: %2 = icmp slt i32 %shr, 127
+; CHECK: LV: Found an estimated cost of 4 for VF 8 For instruction: %spec.select.i = select i1 %2, i32 %shr, i32 127
+; CHECK: LV: Found an estimated cost of 2 for VF 8 For instruction: %conv4 = trunc i32 %spec.select.i to i8
+; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction: %incdec.ptr5 = getelementptr inbounds i8, i8* %pDst.addr.010, i32 1
+; CHECK: LV: Found an estimated cost of 2 for VF 8 For instruction: store i8 %conv4, i8* %pDst.addr.010, align 1
+; CHECK: LV: Found an estimated cost of 1 for VF 8 For instruction: %dec = add i32 %blkCnt.012, -1
+; CHECK: LV: Found an estimated cost of 1 for VF 8 For instruction: %cmp.not = icmp eq i32 %dec, 0
+; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction: br i1 %cmp.not, label %while.end.loopexit, label %while.body
+; CHECK: LV: Vector loop of width 8 costs: 3.
+; CHECK: LV: Found an estimated cost of 0 for VF 16 For instruction: %blkCnt.012 = phi i32 [ %dec, %while.body ], [ %blockSize, %while.body.preheader ]
+; CHECK: LV: Found an estimated cost of 0 for VF 16 For instruction: %pSrcA.addr.011 = phi i8* [ %incdec.ptr, %while.body ], [ %pSrcA, %while.body.preheader ]
+; CHECK: LV: Found an estimated cost of 0 for VF 16 For instruction: %pDst.addr.010 = phi i8* [ %incdec.ptr5, %while.body ], [ %pDst, %while.body.preheader ]
+; CHECK: LV: Found an estimated cost of 0 for VF 16 For instruction: %pSrcB.addr.09 = phi i8* [ %incdec.ptr2, %while.body ], [ %pSrcB, %while.body.preheader ]
+; CHECK: LV: Found an estimated cost of 0 for VF 16 For instruction: %incdec.ptr = getelementptr inbounds i8, i8* %pSrcA.addr.011, i32 1
+; CHECK: LV: Found an estimated cost of 2 for VF 16 For instruction: %0 = load i8, i8* %pSrcA.addr.011, align 1
+; CHECK: LV: Found an estimated cost of 6 for VF 16 For instruction: %conv1 = sext i8 %0 to i32
+; CHECK: LV: Found an estimated cost of 0 for VF 16 For instruction: %incdec.ptr2 = getelementptr inbounds i8, i8* %pSrcB.addr.09, i32 1
+; CHECK: LV: Found an estimated cost of 2 for VF 16 For instruction: %1 = load i8, i8* %pSrcB.addr.09, align 1
+; CHECK: LV: Found an estimated cost of 6 for VF 16 For instruction: %conv3 = sext i8 %1 to i32
+; CHECK: LV: Found an estimated cost of 8 for VF 16 For instruction: %mul = mul nsw i32 %conv3, %conv1
+; CHECK: LV: Found an estimated cost of 8 for VF 16 For instruction: %shr = ashr i32 %mul, 7
+; CHECK: LV: Found an estimated cost of 8 for VF 16 For instruction: %2 = icmp slt i32 %shr, 127
+; CHECK: LV: Found an estimated cost of 8 for VF 16 For instruction: %spec.select.i = select i1 %2, i32 %shr, i32 127
+; CHECK: LV: Found an estimated cost of 6 for VF 16 For instruction: %conv4 = trunc i32 %spec.select.i to i8
+; CHECK: LV: Found an estimated cost of 0 for VF 16 For instruction: %incdec.ptr5 = getelementptr inbounds i8, i8* %pDst.addr.010, i32 1
+; CHECK: LV: Found an estimated cost of 2 for VF 16 For instruction: store i8 %conv4, i8* %pDst.addr.010, align 1
+; CHECK: LV: Found an estimated cost of 1 for VF 16 For instruction: %dec = add i32 %blkCnt.012, -1
+; CHECK: LV: Found an estimated cost of 1 for VF 16 For instruction: %cmp.not = icmp eq i32 %dec, 0
+; CHECK: LV: Found an estimated cost of 0 for VF 16 For instruction: br i1 %cmp.not, label %while.end.loopexit, label %while.body
+; CHECK: LV: Vector loop of width 16 costs: 3.
+; CHECK: LV: Selecting VF: 16.
+define void @cheap_icmp(i8* nocapture readonly %pSrcA, i8* nocapture readonly %pSrcB, i8* nocapture %pDst, i32 %blockSize) #0 {
+entry:
+ %cmp.not8 = icmp eq i32 %blockSize, 0
+ br i1 %cmp.not8, label %while.end, label %while.body.preheader
+
+while.body.preheader: ; preds = %entry
+ br label %while.body
+
+while.body: ; preds = %while.body.preheader, %while.body
+ %blkCnt.012 = phi i32 [ %dec, %while.body ], [ %blockSize, %while.body.preheader ]
+ %pSrcA.addr.011 = phi i8* [ %incdec.ptr, %while.body ], [ %pSrcA, %while.body.preheader ]
+ %pDst.addr.010 = phi i8* [ %incdec.ptr5, %while.body ], [ %pDst, %while.body.preheader ]
+ %pSrcB.addr.09 = phi i8* [ %incdec.ptr2, %while.body ], [ %pSrcB, %while.body.preheader ]
+ %incdec.ptr = getelementptr inbounds i8, i8* %pSrcA.addr.011, i32 1
+ %0 = load i8, i8* %pSrcA.addr.011, align 1
+ %conv1 = sext i8 %0 to i32
+ %incdec.ptr2 = getelementptr inbounds i8, i8* %pSrcB.addr.09, i32 1
+ %1 = load i8, i8* %pSrcB.addr.09, align 1
+ %conv3 = sext i8 %1 to i32
+ %mul = mul nsw i32 %conv3, %conv1
+ %shr = ashr i32 %mul, 7
+ %2 = icmp slt i32 %shr, 127
+ %spec.select.i = select i1 %2, i32 %shr, i32 127
+ %conv4 = trunc i32 %spec.select.i to i8
+ %incdec.ptr5 = getelementptr inbounds i8, i8* %pDst.addr.010, i32 1
+ store i8 %conv4, i8* %pDst.addr.010, align 1
+ %dec = add i32 %blkCnt.012, -1
+ %cmp.not = icmp eq i32 %dec, 0
+ br i1 %cmp.not, label %while.end.loopexit, label %while.body
+
+while.end.loopexit: ; preds = %while.body
+ br label %while.end
+
+while.end: ; preds = %while.end.loopexit, %entry
+ ret void
+}
+
+attributes #0 = { "target-features"="+mve.fp" }
