Skip to content

Commit f3c408d

Browse files
[SLP]Add debug counter support
Fixes #110725 Reviewers: aeubanks Reviewed By: aeubanks Pull Request: #110734
1 parent 60af154 commit f3c408d

File tree

2 files changed

+242
-0
lines changed

2 files changed

+242
-0
lines changed

llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp

Lines changed: 7 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -79,6 +79,7 @@
7979
#include "llvm/Support/Compiler.h"
8080
#include "llvm/Support/DOTGraphTraits.h"
8181
#include "llvm/Support/Debug.h"
82+
#include "llvm/Support/DebugCounter.h"
8283
#include "llvm/Support/ErrorHandling.h"
8384
#include "llvm/Support/GraphWriter.h"
8485
#include "llvm/Support/InstructionCost.h"
@@ -109,6 +110,9 @@ using namespace slpvectorizer;
109110

110111
STATISTIC(NumVectorInstructions, "Number of vector instructions generated");
111112

113+
DEBUG_COUNTER(VectorizedGraphs, "slp-vectorized",
114+
"Controls which SLP graphs should be vectorized.");
115+
112116
static cl::opt<bool>
113117
RunSLPVectorization("vectorize-slp", cl::init(true), cl::Hidden,
114118
cl::desc("Run the SLP vectorization passes"));
@@ -11175,6 +11179,9 @@ bool BoUpSLP::isLoadCombineCandidate(ArrayRef<Value *> Stores) const {
1117511179
}
1117611180

1117711181
bool BoUpSLP::isTreeTinyAndNotFullyVectorizable(bool ForReduction) const {
11182+
if (!DebugCounter::shouldExecute(VectorizedGraphs))
11183+
return true;
11184+
1117811185
// No need to vectorize inserts of gathered values.
1117911186
if (VectorizableTree.size() == 2 &&
1118011187
isa<InsertElementInst>(VectorizableTree[0]->Scalars[0]) &&
Lines changed: 235 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,235 @@
1+
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 2
2+
; RUN: opt -S -passes=slp-vectorizer -debug-counter=slp-vectorized=0 -slp-threshold=-99999 < %s | FileCheck %s --check-prefix=COUNT0
3+
; RUN: opt -S -passes=slp-vectorizer -debug-counter=slp-vectorized=1 -slp-threshold=-99999 < %s | FileCheck %s --check-prefix=COUNT1
4+
; RUN: opt -S -passes=slp-vectorizer -debug-counter=slp-vectorized=2 -slp-threshold=-99999 < %s | FileCheck %s --check-prefix=COUNT2
5+
; RUN: opt -S -passes=slp-vectorizer -debug-counter=slp-vectorized=0-1 -slp-threshold=-99999 < %s | FileCheck %s --check-prefix=COUNT-1
6+
7+
define void @blam(ptr %arg, double %load2, i1 %fcmp3) {
8+
; CHECK-LABEL: define void @blam
9+
; CHECK-SAME: (ptr [[ARG:%.*]], double [[LOAD2:%.*]], i1 [[FCMP3:%.*]]) {
10+
; CHECK-NEXT: bb:
11+
; CHECK-NEXT: [[GETELEMENTPTR13:%.*]] = getelementptr double, ptr [[ARG]], i64 3
12+
; CHECK-NEXT: [[TMP0:%.*]] = load <2 x double>, ptr [[ARG]], align 8
13+
; CHECK-NEXT: [[TMP1:%.*]] = insertelement <2 x i1> poison, i1 [[FCMP3]], i32 0
14+
; CHECK-NEXT: [[TMP2:%.*]] = shufflevector <2 x i1> [[TMP1]], <2 x i1> poison, <2 x i32> zeroinitializer
15+
; CHECK-NEXT: [[TMP3:%.*]] = select <2 x i1> [[TMP2]], <2 x double> zeroinitializer, <2 x double> [[TMP0]]
16+
; CHECK-NEXT: [[TMP4:%.*]] = insertelement <2 x double> [[TMP0]], double [[LOAD2]], i32 0
17+
; CHECK-NEXT: [[TMP5:%.*]] = fcmp olt <2 x double> [[TMP4]], zeroinitializer
18+
; CHECK-NEXT: [[TMP6:%.*]] = select <2 x i1> [[TMP5]], <2 x double> zeroinitializer, <2 x double> [[TMP0]]
19+
; CHECK-NEXT: [[TMP7:%.*]] = fcmp olt <2 x double> [[TMP3]], zeroinitializer
20+
; CHECK-NEXT: [[TMP8:%.*]] = select <2 x i1> [[TMP7]], <2 x double> <double 0.000000e+00, double 1.000000e+00>, <2 x double> <double 1.000000e+00, double 0.000000e+00>
21+
; CHECK-NEXT: [[TMP9:%.*]] = shufflevector <2 x double> [[TMP8]], <2 x double> poison, <2 x i32> <i32 1, i32 0>
22+
; CHECK-NEXT: [[TMP10:%.*]] = fcmp olt <2 x double> [[TMP9]], [[TMP6]]
23+
; CHECK-NEXT: [[TMP11:%.*]] = shufflevector <2 x double> [[TMP4]], <2 x double> <double poison, double 0.000000e+00>, <2 x i32> <i32 0, i32 3>
24+
; CHECK-NEXT: [[TMP12:%.*]] = shufflevector <2 x double> [[TMP4]], <2 x double> <double 0.000000e+00, double poison>, <2 x i32> <i32 2, i32 0>
25+
; CHECK-NEXT: [[TMP13:%.*]] = select <2 x i1> [[TMP10]], <2 x double> [[TMP11]], <2 x double> [[TMP12]]
26+
; CHECK-NEXT: [[TMP14:%.*]] = fcmp olt <2 x double> [[TMP13]], zeroinitializer
27+
; CHECK-NEXT: [[TMP15:%.*]] = select <2 x i1> [[TMP14]], <2 x double> zeroinitializer, <2 x double> <double 1.000000e+00, double 1.000000e+00>
28+
; CHECK-NEXT: [[TMP16:%.*]] = fcmp ogt <2 x double> [[TMP15]], zeroinitializer
29+
; CHECK-NEXT: [[TMP17:%.*]] = shufflevector <2 x double> [[TMP4]], <2 x double> poison, <2 x i32> zeroinitializer
30+
; CHECK-NEXT: [[TMP18:%.*]] = select <2 x i1> [[TMP16]], <2 x double> zeroinitializer, <2 x double> [[TMP17]]
31+
; CHECK-NEXT: [[TMP19:%.*]] = fcmp olt <2 x double> [[TMP18]], zeroinitializer
32+
; CHECK-NEXT: [[TMP20:%.*]] = select <2 x i1> [[TMP19]], <2 x double> <double 1.000000e+00, double 1.000000e+00>, <2 x double> zeroinitializer
33+
; CHECK-NEXT: store <2 x double> [[TMP20]], ptr [[GETELEMENTPTR13]], align 8
34+
; CHECK-NEXT: ret void
35+
;
36+
; COUNT0-LABEL: define void @blam
37+
; COUNT0-SAME: (ptr [[ARG:%.*]], double [[LOAD2:%.*]], i1 [[FCMP3:%.*]]) {
38+
; COUNT0-NEXT: bb:
39+
; COUNT0-NEXT: [[GETELEMENTPTR:%.*]] = getelementptr double, ptr [[ARG]], i64 1
40+
; COUNT0-NEXT: [[LOAD:%.*]] = load double, ptr [[GETELEMENTPTR]], align 8
41+
; COUNT0-NEXT: [[FCMP:%.*]] = fcmp olt double [[LOAD]], 0.000000e+00
42+
; COUNT0-NEXT: [[SELECT3:%.*]] = select i1 [[FCMP]], double 0.000000e+00, double [[LOAD]]
43+
; COUNT0-NEXT: [[SELECT4:%.*]] = select i1 [[FCMP3]], double 0.000000e+00, double [[LOAD]]
44+
; COUNT0-NEXT: [[LOAD7:%.*]] = load double, ptr [[ARG]], align 8
45+
; COUNT0-NEXT: [[SELECT10:%.*]] = select i1 [[FCMP3]], double 0.000000e+00, double [[LOAD7]]
46+
; COUNT0-NEXT: [[GETELEMENTPTR13:%.*]] = getelementptr double, ptr [[ARG]], i64 3
47+
; COUNT0-NEXT: [[GETELEMENTPTR21:%.*]] = getelementptr double, ptr [[ARG]], i64 4
48+
; COUNT0-NEXT: [[FCMP23:%.*]] = fcmp olt double [[SELECT10]], 0.000000e+00
49+
; COUNT0-NEXT: [[SELECT24:%.*]] = select i1 [[FCMP23]], double 0.000000e+00, double 1.000000e+00
50+
; COUNT0-NEXT: [[TMP0:%.*]] = insertelement <2 x double> poison, double [[LOAD2]], i32 1
51+
; COUNT0-NEXT: [[TMP1:%.*]] = insertelement <2 x double> [[TMP0]], double [[SELECT4]], i32 0
52+
; COUNT0-NEXT: [[TMP2:%.*]] = fcmp olt <2 x double> [[TMP1]], zeroinitializer
53+
; COUNT0-NEXT: [[TMP3:%.*]] = insertelement <2 x double> <double 0.000000e+00, double poison>, double [[LOAD7]], i32 1
54+
; COUNT0-NEXT: [[TMP4:%.*]] = select <2 x i1> [[TMP2]], <2 x double> <double 1.000000e+00, double 0.000000e+00>, <2 x double> [[TMP3]]
55+
; COUNT0-NEXT: [[FCMP33:%.*]] = fcmp olt double [[SELECT24]], [[SELECT3]]
56+
; COUNT0-NEXT: [[SELECT34:%.*]] = select i1 [[FCMP33]], double 0.000000e+00, double [[LOAD2]]
57+
; COUNT0-NEXT: [[TMP5:%.*]] = extractelement <2 x double> [[TMP4]], i32 0
58+
; COUNT0-NEXT: [[TMP6:%.*]] = extractelement <2 x double> [[TMP4]], i32 1
59+
; COUNT0-NEXT: [[FCMP39:%.*]] = fcmp olt double [[TMP5]], [[TMP6]]
60+
; COUNT0-NEXT: [[SELECT40:%.*]] = select i1 [[FCMP39]], double [[LOAD2]], double 0.000000e+00
61+
; COUNT0-NEXT: [[FCMP62:%.*]] = fcmp olt double [[SELECT34]], 0.000000e+00
62+
; COUNT0-NEXT: [[SELECT639:%.*]] = select i1 [[FCMP62]], double 0.000000e+00, double 1.000000e+00
63+
; COUNT0-NEXT: [[FCMP76:%.*]] = fcmp olt double [[SELECT40]], 0.000000e+00
64+
; COUNT0-NEXT: [[SELECT77:%.*]] = select i1 [[FCMP76]], double 0.000000e+00, double 1.000000e+00
65+
; COUNT0-NEXT: [[FCMP90:%.*]] = fcmp ogt double [[SELECT639]], 0.000000e+00
66+
; COUNT0-NEXT: [[SELECT91:%.*]] = select i1 [[FCMP90]], double 0.000000e+00, double [[LOAD2]]
67+
; COUNT0-NEXT: [[FCMP92:%.*]] = fcmp ogt double [[SELECT77]], 0.000000e+00
68+
; COUNT0-NEXT: [[SELECT93:%.*]] = select i1 [[FCMP92]], double 0.000000e+00, double [[LOAD2]]
69+
; COUNT0-NEXT: [[FCMP108:%.*]] = fcmp olt double [[SELECT93]], 0.000000e+00
70+
; COUNT0-NEXT: [[SELECT109:%.*]] = select i1 [[FCMP108]], double 1.000000e+00, double 0.000000e+00
71+
; COUNT0-NEXT: [[FCMP110:%.*]] = fcmp olt double [[SELECT91]], 0.000000e+00
72+
; COUNT0-NEXT: [[SELECT111:%.*]] = select i1 [[FCMP110]], double 1.000000e+00, double 0.000000e+00
73+
; COUNT0-NEXT: store double [[SELECT111]], ptr [[GETELEMENTPTR21]], align 8
74+
; COUNT0-NEXT: store double [[SELECT109]], ptr [[GETELEMENTPTR13]], align 8
75+
; COUNT0-NEXT: ret void
76+
;
77+
; COUNT1-LABEL: define void @blam
78+
; COUNT1-SAME: (ptr [[ARG:%.*]], double [[LOAD2:%.*]], i1 [[FCMP3:%.*]]) {
79+
; COUNT1-NEXT: bb:
80+
; COUNT1-NEXT: [[GETELEMENTPTR:%.*]] = getelementptr double, ptr [[ARG]], i64 1
81+
; COUNT1-NEXT: [[LOAD:%.*]] = load double, ptr [[GETELEMENTPTR]], align 8
82+
; COUNT1-NEXT: [[SELECT4:%.*]] = select i1 [[FCMP3]], double 0.000000e+00, double [[LOAD]]
83+
; COUNT1-NEXT: [[LOAD7:%.*]] = load double, ptr [[ARG]], align 8
84+
; COUNT1-NEXT: [[SELECT10:%.*]] = select i1 [[FCMP3]], double 0.000000e+00, double [[LOAD7]]
85+
; COUNT1-NEXT: [[FCMP11:%.*]] = fcmp olt double [[LOAD2]], 0.000000e+00
86+
; COUNT1-NEXT: [[SELECT128:%.*]] = select i1 [[FCMP11]], double 0.000000e+00, double [[LOAD7]]
87+
; COUNT1-NEXT: [[GETELEMENTPTR13:%.*]] = getelementptr double, ptr [[ARG]], i64 3
88+
; COUNT1-NEXT: [[GETELEMENTPTR21:%.*]] = getelementptr double, ptr [[ARG]], i64 4
89+
; COUNT1-NEXT: [[TMP0:%.*]] = insertelement <2 x double> poison, double [[SELECT10]], i32 0
90+
; COUNT1-NEXT: [[TMP1:%.*]] = insertelement <2 x double> [[TMP0]], double [[LOAD]], i32 1
91+
; COUNT1-NEXT: [[TMP2:%.*]] = fcmp olt <2 x double> [[TMP1]], zeroinitializer
92+
; COUNT1-NEXT: [[TMP3:%.*]] = shufflevector <2 x double> [[TMP1]], <2 x double> <double 1.000000e+00, double poison>, <2 x i32> <i32 2, i32 1>
93+
; COUNT1-NEXT: [[TMP4:%.*]] = select <2 x i1> [[TMP2]], <2 x double> zeroinitializer, <2 x double> [[TMP3]]
94+
; COUNT1-NEXT: [[FCMP29:%.*]] = fcmp olt double [[SELECT4]], 0.000000e+00
95+
; COUNT1-NEXT: [[SELECT30:%.*]] = select i1 [[FCMP29]], double 1.000000e+00, double 0.000000e+00
96+
; COUNT1-NEXT: [[TMP5:%.*]] = extractelement <2 x double> [[TMP4]], i32 0
97+
; COUNT1-NEXT: [[TMP6:%.*]] = extractelement <2 x double> [[TMP4]], i32 1
98+
; COUNT1-NEXT: [[FCMP33:%.*]] = fcmp olt double [[TMP5]], [[TMP6]]
99+
; COUNT1-NEXT: [[SELECT34:%.*]] = select i1 [[FCMP33]], double 0.000000e+00, double [[LOAD2]]
100+
; COUNT1-NEXT: [[FCMP39:%.*]] = fcmp olt double [[SELECT30]], [[SELECT128]]
101+
; COUNT1-NEXT: [[SELECT40:%.*]] = select i1 [[FCMP39]], double [[LOAD2]], double 0.000000e+00
102+
; COUNT1-NEXT: [[FCMP62:%.*]] = fcmp olt double [[SELECT34]], 0.000000e+00
103+
; COUNT1-NEXT: [[SELECT639:%.*]] = select i1 [[FCMP62]], double 0.000000e+00, double 1.000000e+00
104+
; COUNT1-NEXT: [[FCMP76:%.*]] = fcmp olt double [[SELECT40]], 0.000000e+00
105+
; COUNT1-NEXT: [[SELECT77:%.*]] = select i1 [[FCMP76]], double 0.000000e+00, double 1.000000e+00
106+
; COUNT1-NEXT: [[FCMP90:%.*]] = fcmp ogt double [[SELECT639]], 0.000000e+00
107+
; COUNT1-NEXT: [[SELECT91:%.*]] = select i1 [[FCMP90]], double 0.000000e+00, double [[LOAD2]]
108+
; COUNT1-NEXT: [[FCMP92:%.*]] = fcmp ogt double [[SELECT77]], 0.000000e+00
109+
; COUNT1-NEXT: [[SELECT93:%.*]] = select i1 [[FCMP92]], double 0.000000e+00, double [[LOAD2]]
110+
; COUNT1-NEXT: [[FCMP108:%.*]] = fcmp olt double [[SELECT93]], 0.000000e+00
111+
; COUNT1-NEXT: [[SELECT109:%.*]] = select i1 [[FCMP108]], double 1.000000e+00, double 0.000000e+00
112+
; COUNT1-NEXT: [[FCMP110:%.*]] = fcmp olt double [[SELECT91]], 0.000000e+00
113+
; COUNT1-NEXT: [[SELECT111:%.*]] = select i1 [[FCMP110]], double 1.000000e+00, double 0.000000e+00
114+
; COUNT1-NEXT: store double [[SELECT111]], ptr [[GETELEMENTPTR21]], align 8
115+
; COUNT1-NEXT: store double [[SELECT109]], ptr [[GETELEMENTPTR13]], align 8
116+
; COUNT1-NEXT: ret void
117+
;
118+
; COUNT2-LABEL: define void @blam
119+
; COUNT2-SAME: (ptr [[ARG:%.*]], double [[LOAD2:%.*]], i1 [[FCMP3:%.*]]) {
120+
; COUNT2-NEXT: bb:
121+
; COUNT2-NEXT: [[GETELEMENTPTR:%.*]] = getelementptr double, ptr [[ARG]], i64 1
122+
; COUNT2-NEXT: [[LOAD:%.*]] = load double, ptr [[GETELEMENTPTR]], align 8
123+
; COUNT2-NEXT: [[FCMP:%.*]] = fcmp olt double [[LOAD]], 0.000000e+00
124+
; COUNT2-NEXT: [[SELECT3:%.*]] = select i1 [[FCMP]], double 0.000000e+00, double [[LOAD]]
125+
; COUNT2-NEXT: [[SELECT4:%.*]] = select i1 [[FCMP3]], double 0.000000e+00, double [[LOAD]]
126+
; COUNT2-NEXT: [[LOAD7:%.*]] = load double, ptr [[ARG]], align 8
127+
; COUNT2-NEXT: [[SELECT10:%.*]] = select i1 [[FCMP3]], double 0.000000e+00, double [[LOAD7]]
128+
; COUNT2-NEXT: [[FCMP11:%.*]] = fcmp olt double [[LOAD2]], 0.000000e+00
129+
; COUNT2-NEXT: [[SELECT128:%.*]] = select i1 [[FCMP11]], double 0.000000e+00, double [[LOAD7]]
130+
; COUNT2-NEXT: [[GETELEMENTPTR13:%.*]] = getelementptr double, ptr [[ARG]], i64 3
131+
; COUNT2-NEXT: [[GETELEMENTPTR21:%.*]] = getelementptr double, ptr [[ARG]], i64 4
132+
; COUNT2-NEXT: [[FCMP23:%.*]] = fcmp olt double [[SELECT10]], 0.000000e+00
133+
; COUNT2-NEXT: [[SELECT24:%.*]] = select i1 [[FCMP23]], double 0.000000e+00, double 1.000000e+00
134+
; COUNT2-NEXT: [[FCMP29:%.*]] = fcmp olt double [[SELECT4]], 0.000000e+00
135+
; COUNT2-NEXT: [[SELECT30:%.*]] = select i1 [[FCMP29]], double 1.000000e+00, double 0.000000e+00
136+
; COUNT2-NEXT: [[FCMP33:%.*]] = fcmp olt double [[SELECT24]], [[SELECT3]]
137+
; COUNT2-NEXT: [[SELECT34:%.*]] = select i1 [[FCMP33]], double 0.000000e+00, double [[LOAD2]]
138+
; COUNT2-NEXT: [[FCMP39:%.*]] = fcmp olt double [[SELECT30]], [[SELECT128]]
139+
; COUNT2-NEXT: [[SELECT40:%.*]] = select i1 [[FCMP39]], double [[LOAD2]], double 0.000000e+00
140+
; COUNT2-NEXT: [[FCMP62:%.*]] = fcmp olt double [[SELECT34]], 0.000000e+00
141+
; COUNT2-NEXT: [[SELECT639:%.*]] = select i1 [[FCMP62]], double 0.000000e+00, double 1.000000e+00
142+
; COUNT2-NEXT: [[FCMP76:%.*]] = fcmp olt double [[SELECT40]], 0.000000e+00
143+
; COUNT2-NEXT: [[SELECT77:%.*]] = select i1 [[FCMP76]], double 0.000000e+00, double 1.000000e+00
144+
; COUNT2-NEXT: [[FCMP90:%.*]] = fcmp ogt double [[SELECT639]], 0.000000e+00
145+
; COUNT2-NEXT: [[SELECT91:%.*]] = select i1 [[FCMP90]], double 0.000000e+00, double [[LOAD2]]
146+
; COUNT2-NEXT: [[FCMP92:%.*]] = fcmp ogt double [[SELECT77]], 0.000000e+00
147+
; COUNT2-NEXT: [[SELECT93:%.*]] = select i1 [[FCMP92]], double 0.000000e+00, double [[LOAD2]]
148+
; COUNT2-NEXT: [[FCMP108:%.*]] = fcmp olt double [[SELECT93]], 0.000000e+00
149+
; COUNT2-NEXT: [[SELECT109:%.*]] = select i1 [[FCMP108]], double 1.000000e+00, double 0.000000e+00
150+
; COUNT2-NEXT: [[FCMP110:%.*]] = fcmp olt double [[SELECT91]], 0.000000e+00
151+
; COUNT2-NEXT: [[SELECT111:%.*]] = select i1 [[FCMP110]], double 1.000000e+00, double 0.000000e+00
152+
; COUNT2-NEXT: store double [[SELECT111]], ptr [[GETELEMENTPTR21]], align 8
153+
; COUNT2-NEXT: store double [[SELECT109]], ptr [[GETELEMENTPTR13]], align 8
154+
; COUNT2-NEXT: ret void
155+
;
156+
; COUNT-1-LABEL: define void @blam
157+
; COUNT-1-SAME: (ptr [[ARG:%.*]], double [[LOAD2:%.*]], i1 [[FCMP3:%.*]]) {
158+
; COUNT-1-NEXT: bb:
159+
; COUNT-1-NEXT: [[GETELEMENTPTR:%.*]] = getelementptr double, ptr [[ARG]], i64 1
160+
; COUNT-1-NEXT: [[LOAD:%.*]] = load double, ptr [[GETELEMENTPTR]], align 8
161+
; COUNT-1-NEXT: [[SELECT4:%.*]] = select i1 [[FCMP3]], double 0.000000e+00, double [[LOAD]]
162+
; COUNT-1-NEXT: [[LOAD7:%.*]] = load double, ptr [[ARG]], align 8
163+
; COUNT-1-NEXT: [[SELECT10:%.*]] = select i1 [[FCMP3]], double 0.000000e+00, double [[LOAD7]]
164+
; COUNT-1-NEXT: [[GETELEMENTPTR13:%.*]] = getelementptr double, ptr [[ARG]], i64 3
165+
; COUNT-1-NEXT: [[GETELEMENTPTR21:%.*]] = getelementptr double, ptr [[ARG]], i64 4
166+
; COUNT-1-NEXT: [[TMP0:%.*]] = insertelement <2 x double> poison, double [[SELECT10]], i32 0
167+
; COUNT-1-NEXT: [[TMP1:%.*]] = insertelement <2 x double> [[TMP0]], double [[LOAD]], i32 1
168+
; COUNT-1-NEXT: [[TMP2:%.*]] = fcmp olt <2 x double> [[TMP1]], zeroinitializer
169+
; COUNT-1-NEXT: [[TMP3:%.*]] = shufflevector <2 x double> [[TMP1]], <2 x double> <double 1.000000e+00, double poison>, <2 x i32> <i32 2, i32 1>
170+
; COUNT-1-NEXT: [[TMP4:%.*]] = select <2 x i1> [[TMP2]], <2 x double> zeroinitializer, <2 x double> [[TMP3]]
171+
; COUNT-1-NEXT: [[TMP5:%.*]] = insertelement <2 x double> poison, double [[LOAD2]], i32 1
172+
; COUNT-1-NEXT: [[TMP6:%.*]] = insertelement <2 x double> [[TMP5]], double [[SELECT4]], i32 0
173+
; COUNT-1-NEXT: [[TMP7:%.*]] = fcmp olt <2 x double> [[TMP6]], zeroinitializer
174+
; COUNT-1-NEXT: [[TMP8:%.*]] = insertelement <2 x double> <double 0.000000e+00, double poison>, double [[LOAD7]], i32 1
175+
; COUNT-1-NEXT: [[TMP9:%.*]] = select <2 x i1> [[TMP7]], <2 x double> <double 1.000000e+00, double 0.000000e+00>, <2 x double> [[TMP8]]
176+
; COUNT-1-NEXT: [[TMP10:%.*]] = extractelement <2 x double> [[TMP4]], i32 0
177+
; COUNT-1-NEXT: [[TMP11:%.*]] = extractelement <2 x double> [[TMP4]], i32 1
178+
; COUNT-1-NEXT: [[FCMP33:%.*]] = fcmp olt double [[TMP10]], [[TMP11]]
179+
; COUNT-1-NEXT: [[SELECT34:%.*]] = select i1 [[FCMP33]], double 0.000000e+00, double [[LOAD2]]
180+
; COUNT-1-NEXT: [[TMP12:%.*]] = extractelement <2 x double> [[TMP9]], i32 0
181+
; COUNT-1-NEXT: [[TMP13:%.*]] = extractelement <2 x double> [[TMP9]], i32 1
182+
; COUNT-1-NEXT: [[FCMP39:%.*]] = fcmp olt double [[TMP12]], [[TMP13]]
183+
; COUNT-1-NEXT: [[SELECT40:%.*]] = select i1 [[FCMP39]], double [[LOAD2]], double 0.000000e+00
184+
; COUNT-1-NEXT: [[FCMP62:%.*]] = fcmp olt double [[SELECT34]], 0.000000e+00
185+
; COUNT-1-NEXT: [[SELECT639:%.*]] = select i1 [[FCMP62]], double 0.000000e+00, double 1.000000e+00
186+
; COUNT-1-NEXT: [[FCMP76:%.*]] = fcmp olt double [[SELECT40]], 0.000000e+00
187+
; COUNT-1-NEXT: [[SELECT77:%.*]] = select i1 [[FCMP76]], double 0.000000e+00, double 1.000000e+00
188+
; COUNT-1-NEXT: [[FCMP90:%.*]] = fcmp ogt double [[SELECT639]], 0.000000e+00
189+
; COUNT-1-NEXT: [[SELECT91:%.*]] = select i1 [[FCMP90]], double 0.000000e+00, double [[LOAD2]]
190+
; COUNT-1-NEXT: [[FCMP92:%.*]] = fcmp ogt double [[SELECT77]], 0.000000e+00
191+
; COUNT-1-NEXT: [[SELECT93:%.*]] = select i1 [[FCMP92]], double 0.000000e+00, double [[LOAD2]]
192+
; COUNT-1-NEXT: [[FCMP108:%.*]] = fcmp olt double [[SELECT93]], 0.000000e+00
193+
; COUNT-1-NEXT: [[SELECT109:%.*]] = select i1 [[FCMP108]], double 1.000000e+00, double 0.000000e+00
194+
; COUNT-1-NEXT: [[FCMP110:%.*]] = fcmp olt double [[SELECT91]], 0.000000e+00
195+
; COUNT-1-NEXT: [[SELECT111:%.*]] = select i1 [[FCMP110]], double 1.000000e+00, double 0.000000e+00
196+
; COUNT-1-NEXT: store double [[SELECT111]], ptr [[GETELEMENTPTR21]], align 8
197+
; COUNT-1-NEXT: store double [[SELECT109]], ptr [[GETELEMENTPTR13]], align 8
198+
; COUNT-1-NEXT: ret void
199+
;
200+
bb:
201+
%getelementptr = getelementptr double, ptr %arg, i64 1
202+
%load = load double, ptr %getelementptr, align 8
203+
%fcmp = fcmp olt double %load, 0.000000e+00
204+
%select3 = select i1 %fcmp, double 0.000000e+00, double %load
205+
%select4 = select i1 %fcmp3, double 0.000000e+00, double %load
206+
%load7 = load double, ptr %arg, align 8
207+
%select10 = select i1 %fcmp3, double 0.000000e+00, double %load7
208+
%fcmp11 = fcmp olt double %load2, 0.000000e+00
209+
%select128 = select i1 %fcmp11, double 0.000000e+00, double %load7
210+
%getelementptr13 = getelementptr double, ptr %arg, i64 3
211+
%getelementptr21 = getelementptr double, ptr %arg, i64 4
212+
%fcmp23 = fcmp olt double %select10, 0.000000e+00
213+
%select24 = select i1 %fcmp23, double 0.000000e+00, double 1.000000e+00
214+
%fcmp29 = fcmp olt double %select4, 0.000000e+00
215+
%select30 = select i1 %fcmp29, double 1.000000e+00, double 0.000000e+00
216+
%fcmp33 = fcmp olt double %select24, %select3
217+
%select34 = select i1 %fcmp33, double 0.000000e+00, double %load2
218+
%fcmp39 = fcmp olt double %select30, %select128
219+
%select40 = select i1 %fcmp39, double %load2, double 0.000000e+00
220+
%fcmp62 = fcmp olt double %select34, 0.000000e+00
221+
%select639 = select i1 %fcmp62, double 0.000000e+00, double 1.000000e+00
222+
%fcmp76 = fcmp olt double %select40, 0.000000e+00
223+
%select77 = select i1 %fcmp76, double 0.000000e+00, double 1.000000e+00
224+
%fcmp90 = fcmp ogt double %select639, 0.000000e+00
225+
%select91 = select i1 %fcmp90, double 0.000000e+00, double %load2
226+
%fcmp92 = fcmp ogt double %select77, 0.000000e+00
227+
%select93 = select i1 %fcmp92, double 0.000000e+00, double %load2
228+
%fcmp108 = fcmp olt double %select93, 0.000000e+00
229+
%select109 = select i1 %fcmp108, double 1.000000e+00, double 0.000000e+00
230+
%fcmp110 = fcmp olt double %select91, 0.000000e+00
231+
%select111 = select i1 %fcmp110, double 1.000000e+00, double 0.000000e+00
232+
store double %select111, ptr %getelementptr21, align 8
233+
store double %select109, ptr %getelementptr13, align 8
234+
ret void
235+
}

0 commit comments

Comments
 (0)