[llvm] 6403009 - SLP: honor requested max vector size merging PHIs

Stanislav Mekhanoshin via llvm-commits llvm-commits at lists.llvm.org
Wed Jul 8 08:19:47 PDT 2020


Author: Stanislav Mekhanoshin
Date: 2020-07-08T08:06:15-07:00
New Revision: 64030099c378062131fa1b29742a783f2ca14c17

URL: https://github.com/llvm/llvm-project/commit/64030099c378062131fa1b29742a783f2ca14c17
DIFF: https://github.com/llvm/llvm-project/commit/64030099c378062131fa1b29742a783f2ca14c17.diff

LOG: SLP: honor requested max vector size merging PHIs

At the moment this place does not check maximum size set
by TTI and just creates a maximum possible vectors.

Differential Revision: https://reviews.llvm.org/D82227

Added: 
    

Modified: 
    llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
    llvm/test/Transforms/SLPVectorizer/X86/remark_unsupported.ll
    llvm/test/Transforms/SLPVectorizer/slp-max-phi-size.ll

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index 5bb05c6ac3d1..d4b16fac985d 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -7361,6 +7361,7 @@ bool SLPVectorizerPass::vectorizeChainsInBlock(BasicBlock *BB, BoUpSLP &R) {
   bool Changed = false;
   SmallVector<Value *, 4> Incoming;
   SmallPtrSet<Value *, 16> VisitedInstrs;
+  unsigned MaxVecRegSize = R.getMaxVecRegSize();
 
   bool HaveVectorizedPhiNodes = true;
   while (HaveVectorizedPhiNodes) {
@@ -7387,8 +7388,18 @@ bool SLPVectorizerPass::vectorizeChainsInBlock(BasicBlock *BB, BoUpSLP &R) {
 
       // Look for the next elements with the same type.
       SmallVector<Value *, 4>::iterator SameTypeIt = IncIt;
+      Type *EltTy = (*IncIt)->getType();
+      unsigned EltSize = EltTy->isSized() ? DL->getTypeSizeInBits(EltTy)
+                                          : MaxVecRegSize;
+      unsigned MaxNumElts = MaxVecRegSize / EltSize;
+      if (MaxNumElts < 2) {
+        ++IncIt;
+        continue;
+      }
+
       while (SameTypeIt != E &&
-             (*SameTypeIt)->getType() == (*IncIt)->getType()) {
+             (*SameTypeIt)->getType() == EltTy &&
+             (SameTypeIt - IncIt) < MaxNumElts) {
         VisitedInstrs.insert(*SameTypeIt);
         ++SameTypeIt;
       }

diff  --git a/llvm/test/Transforms/SLPVectorizer/X86/remark_unsupported.ll b/llvm/test/Transforms/SLPVectorizer/X86/remark_unsupported.ll
index a134aec00bbb..afc08aa4fb27 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/remark_unsupported.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/remark_unsupported.ll
@@ -1,5 +1,5 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
-; RUN: opt -S -mtriple=x86_64-pc-linux-gnu -mcpu=generic -slp-vectorizer -pass-remarks-output=%t < %s | FileCheck %s
+; RUN: opt -S -mtriple=x86_64-pc-linux-gnu -mcpu=generic -slp-vectorizer --slp-max-reg-size=256 -pass-remarks-output=%t < %s | FileCheck %s
 ; RUN: FileCheck --input-file=%t --check-prefix=YAML %s
 
 ; This type is not supported by SLP

diff  --git a/llvm/test/Transforms/SLPVectorizer/slp-max-phi-size.ll b/llvm/test/Transforms/SLPVectorizer/slp-max-phi-size.ll
index e349016fa4f2..466e83d0260a 100644
--- a/llvm/test/Transforms/SLPVectorizer/slp-max-phi-size.ll
+++ b/llvm/test/Transforms/SLPVectorizer/slp-max-phi-size.ll
@@ -8,94 +8,74 @@ define void @phi_float32(half %hval, float %fval) {
 ; MAX32-NEXT:  bb:
 ; MAX32-NEXT:    br label [[BB1:%.*]]
 ; MAX32:       bb1:
-; MAX32-NEXT:    [[TMP0:%.*]] = insertelement <4 x half> undef, half [[HVAL:%.*]], i32 0
-; MAX32-NEXT:    [[TMP1:%.*]] = insertelement <4 x half> [[TMP0]], half [[HVAL]], i32 1
-; MAX32-NEXT:    [[TMP2:%.*]] = insertelement <4 x half> [[TMP1]], half [[HVAL]], i32 2
-; MAX32-NEXT:    [[TMP3:%.*]] = insertelement <4 x half> [[TMP2]], half [[HVAL]], i32 3
-; MAX32-NEXT:    [[TMP4:%.*]] = fpext <4 x half> [[TMP3]] to <4 x float>
-; MAX32-NEXT:    [[SHUFFLE:%.*]] = shufflevector <4 x float> [[TMP4]], <4 x float> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 3, i32 2, i32 1, i32 0, i32 3, i32 2, i32 1, i32 0, i32 3, i32 2, i32 1, i32 0, i32 3, i32 2, i32 1, i32 0, i32 3, i32 2, i32 1, i32 0, i32 3, i32 2, i32 1, i32 0>
-; MAX32-NEXT:    [[TMP5:%.*]] = insertelement <32 x float> undef, float [[FVAL:%.*]], i32 0
-; MAX32-NEXT:    [[TMP6:%.*]] = insertelement <32 x float> [[TMP5]], float [[FVAL]], i32 1
-; MAX32-NEXT:    [[TMP7:%.*]] = insertelement <32 x float> [[TMP6]], float [[FVAL]], i32 2
-; MAX32-NEXT:    [[TMP8:%.*]] = insertelement <32 x float> [[TMP7]], float [[FVAL]], i32 3
-; MAX32-NEXT:    [[TMP9:%.*]] = insertelement <32 x float> [[TMP8]], float [[FVAL]], i32 4
-; MAX32-NEXT:    [[TMP10:%.*]] = insertelement <32 x float> [[TMP9]], float [[FVAL]], i32 5
-; MAX32-NEXT:    [[TMP11:%.*]] = insertelement <32 x float> [[TMP10]], float [[FVAL]], i32 6
-; MAX32-NEXT:    [[TMP12:%.*]] = insertelement <32 x float> [[TMP11]], float [[FVAL]], i32 7
-; MAX32-NEXT:    [[TMP13:%.*]] = insertelement <32 x float> [[TMP12]], float [[FVAL]], i32 8
-; MAX32-NEXT:    [[TMP14:%.*]] = insertelement <32 x float> [[TMP13]], float [[FVAL]], i32 9
-; MAX32-NEXT:    [[TMP15:%.*]] = insertelement <32 x float> [[TMP14]], float [[FVAL]], i32 10
-; MAX32-NEXT:    [[TMP16:%.*]] = insertelement <32 x float> [[TMP15]], float [[FVAL]], i32 11
-; MAX32-NEXT:    [[TMP17:%.*]] = insertelement <32 x float> [[TMP16]], float [[FVAL]], i32 12
-; MAX32-NEXT:    [[TMP18:%.*]] = insertelement <32 x float> [[TMP17]], float [[FVAL]], i32 13
-; MAX32-NEXT:    [[TMP19:%.*]] = insertelement <32 x float> [[TMP18]], float [[FVAL]], i32 14
-; MAX32-NEXT:    [[TMP20:%.*]] = insertelement <32 x float> [[TMP19]], float [[FVAL]], i32 15
-; MAX32-NEXT:    [[TMP21:%.*]] = insertelement <32 x float> [[TMP20]], float [[FVAL]], i32 16
-; MAX32-NEXT:    [[TMP22:%.*]] = insertelement <32 x float> [[TMP21]], float [[FVAL]], i32 17
-; MAX32-NEXT:    [[TMP23:%.*]] = insertelement <32 x float> [[TMP22]], float [[FVAL]], i32 18
-; MAX32-NEXT:    [[TMP24:%.*]] = insertelement <32 x float> [[TMP23]], float [[FVAL]], i32 19
-; MAX32-NEXT:    [[TMP25:%.*]] = insertelement <32 x float> [[TMP24]], float [[FVAL]], i32 20
-; MAX32-NEXT:    [[TMP26:%.*]] = insertelement <32 x float> [[TMP25]], float [[FVAL]], i32 21
-; MAX32-NEXT:    [[TMP27:%.*]] = insertelement <32 x float> [[TMP26]], float [[FVAL]], i32 22
-; MAX32-NEXT:    [[TMP28:%.*]] = insertelement <32 x float> [[TMP27]], float [[FVAL]], i32 23
-; MAX32-NEXT:    [[TMP29:%.*]] = insertelement <32 x float> [[TMP28]], float [[FVAL]], i32 24
-; MAX32-NEXT:    [[TMP30:%.*]] = insertelement <32 x float> [[TMP29]], float [[FVAL]], i32 25
-; MAX32-NEXT:    [[TMP31:%.*]] = insertelement <32 x float> [[TMP30]], float [[FVAL]], i32 26
-; MAX32-NEXT:    [[TMP32:%.*]] = insertelement <32 x float> [[TMP31]], float [[FVAL]], i32 27
-; MAX32-NEXT:    [[TMP33:%.*]] = insertelement <32 x float> [[TMP32]], float [[FVAL]], i32 28
-; MAX32-NEXT:    [[TMP34:%.*]] = insertelement <32 x float> [[TMP33]], float [[FVAL]], i32 29
-; MAX32-NEXT:    [[TMP35:%.*]] = insertelement <32 x float> [[TMP34]], float [[FVAL]], i32 30
-; MAX32-NEXT:    [[TMP36:%.*]] = insertelement <32 x float> [[TMP35]], float [[FVAL]], i32 31
-; MAX32-NEXT:    [[TMP37:%.*]] = fmul <32 x float> [[SHUFFLE]], [[TMP36]]
-; MAX32-NEXT:    [[TMP38:%.*]] = fadd <32 x float> zeroinitializer, [[TMP37]]
-; MAX32-NEXT:    [[TMP39:%.*]] = extractelement <32 x float> [[TMP38]], i32 0
-; MAX32-NEXT:    [[TMP40:%.*]] = insertelement <32 x float> undef, float [[TMP39]], i32 0
-; MAX32-NEXT:    [[TMP41:%.*]] = extractelement <32 x float> [[TMP38]], i32 1
-; MAX32-NEXT:    [[TMP42:%.*]] = insertelement <32 x float> [[TMP40]], float [[TMP41]], i32 1
-; MAX32-NEXT:    [[TMP43:%.*]] = insertelement <32 x float> [[TMP42]], float [[FVAL]], i32 2
-; MAX32-NEXT:    [[TMP44:%.*]] = insertelement <32 x float> [[TMP43]], float [[FVAL]], i32 3
-; MAX32-NEXT:    [[TMP45:%.*]] = extractelement <32 x float> [[TMP38]], i32 4
-; MAX32-NEXT:    [[TMP46:%.*]] = insertelement <32 x float> [[TMP44]], float [[TMP45]], i32 4
-; MAX32-NEXT:    [[TMP47:%.*]] = extractelement <32 x float> [[TMP38]], i32 5
-; MAX32-NEXT:    [[TMP48:%.*]] = insertelement <32 x float> [[TMP46]], float [[TMP47]], i32 5
-; MAX32-NEXT:    [[TMP49:%.*]] = insertelement <32 x float> [[TMP48]], float [[FVAL]], i32 6
-; MAX32-NEXT:    [[TMP50:%.*]] = insertelement <32 x float> [[TMP49]], float [[FVAL]], i32 7
-; MAX32-NEXT:    [[TMP51:%.*]] = insertelement <32 x float> [[TMP50]], float [[FVAL]], i32 8
-; MAX32-NEXT:    [[TMP52:%.*]] = insertelement <32 x float> [[TMP51]], float [[FVAL]], i32 9
-; MAX32-NEXT:    [[TMP53:%.*]] = extractelement <32 x float> [[TMP38]], i32 10
-; MAX32-NEXT:    [[TMP54:%.*]] = insertelement <32 x float> [[TMP52]], float [[TMP53]], i32 10
-; MAX32-NEXT:    [[TMP55:%.*]] = extractelement <32 x float> [[TMP38]], i32 11
-; MAX32-NEXT:    [[TMP56:%.*]] = insertelement <32 x float> [[TMP54]], float [[TMP55]], i32 11
-; MAX32-NEXT:    [[TMP57:%.*]] = insertelement <32 x float> [[TMP56]], float [[FVAL]], i32 12
-; MAX32-NEXT:    [[TMP58:%.*]] = insertelement <32 x float> [[TMP57]], float [[FVAL]], i32 13
-; MAX32-NEXT:    [[TMP59:%.*]] = extractelement <32 x float> [[TMP38]], i32 14
-; MAX32-NEXT:    [[TMP60:%.*]] = insertelement <32 x float> [[TMP58]], float [[TMP59]], i32 14
-; MAX32-NEXT:    [[TMP61:%.*]] = extractelement <32 x float> [[TMP38]], i32 15
-; MAX32-NEXT:    [[TMP62:%.*]] = insertelement <32 x float> [[TMP60]], float [[TMP61]], i32 15
-; MAX32-NEXT:    [[TMP63:%.*]] = insertelement <32 x float> [[TMP62]], float [[FVAL]], i32 16
-; MAX32-NEXT:    [[TMP64:%.*]] = insertelement <32 x float> [[TMP63]], float [[FVAL]], i32 17
-; MAX32-NEXT:    [[TMP65:%.*]] = extractelement <32 x float> [[TMP38]], i32 18
-; MAX32-NEXT:    [[TMP66:%.*]] = insertelement <32 x float> [[TMP64]], float [[TMP65]], i32 18
-; MAX32-NEXT:    [[TMP67:%.*]] = extractelement <32 x float> [[TMP38]], i32 19
-; MAX32-NEXT:    [[TMP68:%.*]] = insertelement <32 x float> [[TMP66]], float [[TMP67]], i32 19
-; MAX32-NEXT:    [[TMP69:%.*]] = insertelement <32 x float> [[TMP68]], float [[FVAL]], i32 20
-; MAX32-NEXT:    [[TMP70:%.*]] = insertelement <32 x float> [[TMP69]], float [[FVAL]], i32 21
-; MAX32-NEXT:    [[TMP71:%.*]] = extractelement <32 x float> [[TMP38]], i32 22
-; MAX32-NEXT:    [[TMP72:%.*]] = insertelement <32 x float> [[TMP70]], float [[TMP71]], i32 22
-; MAX32-NEXT:    [[TMP73:%.*]] = extractelement <32 x float> [[TMP38]], i32 23
-; MAX32-NEXT:    [[TMP74:%.*]] = insertelement <32 x float> [[TMP72]], float [[TMP73]], i32 23
-; MAX32-NEXT:    [[TMP75:%.*]] = insertelement <32 x float> [[TMP74]], float [[FVAL]], i32 24
-; MAX32-NEXT:    [[TMP76:%.*]] = insertelement <32 x float> [[TMP75]], float [[FVAL]], i32 25
-; MAX32-NEXT:    [[TMP77:%.*]] = extractelement <32 x float> [[TMP38]], i32 26
-; MAX32-NEXT:    [[TMP78:%.*]] = insertelement <32 x float> [[TMP76]], float [[TMP77]], i32 26
-; MAX32-NEXT:    [[TMP79:%.*]] = extractelement <32 x float> [[TMP38]], i32 27
-; MAX32-NEXT:    [[TMP80:%.*]] = insertelement <32 x float> [[TMP78]], float [[TMP79]], i32 27
-; MAX32-NEXT:    [[TMP81:%.*]] = insertelement <32 x float> [[TMP80]], float [[FVAL]], i32 28
-; MAX32-NEXT:    [[TMP82:%.*]] = insertelement <32 x float> [[TMP81]], float [[FVAL]], i32 29
-; MAX32-NEXT:    [[TMP83:%.*]] = extractelement <32 x float> [[TMP38]], i32 30
-; MAX32-NEXT:    [[TMP84:%.*]] = insertelement <32 x float> [[TMP82]], float [[TMP83]], i32 30
-; MAX32-NEXT:    [[TMP85:%.*]] = extractelement <32 x float> [[TMP38]], i32 31
-; MAX32-NEXT:    [[TMP86:%.*]] = insertelement <32 x float> [[TMP84]], float [[TMP85]], i32 31
+; MAX32-NEXT:    [[I:%.*]] = fpext half [[HVAL:%.*]] to float
+; MAX32-NEXT:    [[I1:%.*]] = fmul float [[I]], [[FVAL:%.*]]
+; MAX32-NEXT:    [[I2:%.*]] = fadd float 0.000000e+00, [[I1]]
+; MAX32-NEXT:    [[I3:%.*]] = fpext half [[HVAL]] to float
+; MAX32-NEXT:    [[I4:%.*]] = fmul float [[I3]], [[FVAL]]
+; MAX32-NEXT:    [[I5:%.*]] = fadd float 0.000000e+00, [[I4]]
+; MAX32-NEXT:    [[I6:%.*]] = fpext half [[HVAL]] to float
+; MAX32-NEXT:    [[I7:%.*]] = fmul float [[I6]], [[FVAL]]
+; MAX32-NEXT:    [[I8:%.*]] = fadd float 0.000000e+00, [[I7]]
+; MAX32-NEXT:    [[I9:%.*]] = fpext half [[HVAL]] to float
+; MAX32-NEXT:    [[I10:%.*]] = fmul float [[I9]], [[FVAL]]
+; MAX32-NEXT:    [[I11:%.*]] = fadd float 0.000000e+00, [[I10]]
+; MAX32-NEXT:    [[I12:%.*]] = fmul float [[I]], [[FVAL]]
+; MAX32-NEXT:    [[I13:%.*]] = fadd float 0.000000e+00, [[I12]]
+; MAX32-NEXT:    [[I14:%.*]] = fmul float [[I3]], [[FVAL]]
+; MAX32-NEXT:    [[I15:%.*]] = fadd float 0.000000e+00, [[I14]]
+; MAX32-NEXT:    [[I16:%.*]] = fmul float [[I6]], [[FVAL]]
+; MAX32-NEXT:    [[I17:%.*]] = fadd float 0.000000e+00, [[I16]]
+; MAX32-NEXT:    [[I18:%.*]] = fmul float [[I9]], [[FVAL]]
+; MAX32-NEXT:    [[I19:%.*]] = fadd float 0.000000e+00, [[I18]]
+; MAX32-NEXT:    [[I20:%.*]] = fmul float [[I]], [[FVAL]]
+; MAX32-NEXT:    [[I21:%.*]] = fadd float 0.000000e+00, [[I20]]
+; MAX32-NEXT:    [[I22:%.*]] = fmul float [[I3]], [[FVAL]]
+; MAX32-NEXT:    [[I23:%.*]] = fadd float 0.000000e+00, [[I22]]
+; MAX32-NEXT:    [[I24:%.*]] = fmul float [[I6]], [[FVAL]]
+; MAX32-NEXT:    [[I25:%.*]] = fadd float 0.000000e+00, [[I24]]
+; MAX32-NEXT:    [[I26:%.*]] = fmul float [[I9]], [[FVAL]]
+; MAX32-NEXT:    [[I27:%.*]] = fadd float 0.000000e+00, [[I26]]
+; MAX32-NEXT:    [[I28:%.*]] = fmul float [[I]], [[FVAL]]
+; MAX32-NEXT:    [[I29:%.*]] = fadd float 0.000000e+00, [[I28]]
+; MAX32-NEXT:    [[I30:%.*]] = fmul float [[I3]], [[FVAL]]
+; MAX32-NEXT:    [[I31:%.*]] = fadd float 0.000000e+00, [[I30]]
+; MAX32-NEXT:    [[I32:%.*]] = fmul float [[I6]], [[FVAL]]
+; MAX32-NEXT:    [[I33:%.*]] = fadd float 0.000000e+00, [[I32]]
+; MAX32-NEXT:    [[I34:%.*]] = fmul float [[I9]], [[FVAL]]
+; MAX32-NEXT:    [[I35:%.*]] = fadd float 0.000000e+00, [[I34]]
+; MAX32-NEXT:    [[I36:%.*]] = fmul float [[I]], [[FVAL]]
+; MAX32-NEXT:    [[I37:%.*]] = fadd float 0.000000e+00, [[I36]]
+; MAX32-NEXT:    [[I38:%.*]] = fmul float [[I3]], [[FVAL]]
+; MAX32-NEXT:    [[I39:%.*]] = fadd float 0.000000e+00, [[I38]]
+; MAX32-NEXT:    [[I40:%.*]] = fmul float [[I6]], [[FVAL]]
+; MAX32-NEXT:    [[I41:%.*]] = fadd float 0.000000e+00, [[I40]]
+; MAX32-NEXT:    [[I42:%.*]] = fmul float [[I9]], [[FVAL]]
+; MAX32-NEXT:    [[I43:%.*]] = fadd float 0.000000e+00, [[I42]]
+; MAX32-NEXT:    [[I44:%.*]] = fmul float [[I]], [[FVAL]]
+; MAX32-NEXT:    [[I45:%.*]] = fadd float 0.000000e+00, [[I44]]
+; MAX32-NEXT:    [[I46:%.*]] = fmul float [[I3]], [[FVAL]]
+; MAX32-NEXT:    [[I47:%.*]] = fadd float 0.000000e+00, [[I46]]
+; MAX32-NEXT:    [[I48:%.*]] = fmul float [[I6]], [[FVAL]]
+; MAX32-NEXT:    [[I49:%.*]] = fadd float 0.000000e+00, [[I48]]
+; MAX32-NEXT:    [[I50:%.*]] = fmul float [[I9]], [[FVAL]]
+; MAX32-NEXT:    [[I51:%.*]] = fadd float 0.000000e+00, [[I50]]
+; MAX32-NEXT:    [[I52:%.*]] = fmul float [[I]], [[FVAL]]
+; MAX32-NEXT:    [[I53:%.*]] = fadd float 0.000000e+00, [[I52]]
+; MAX32-NEXT:    [[I54:%.*]] = fmul float [[I3]], [[FVAL]]
+; MAX32-NEXT:    [[I55:%.*]] = fadd float 0.000000e+00, [[I54]]
+; MAX32-NEXT:    [[I56:%.*]] = fmul float [[I6]], [[FVAL]]
+; MAX32-NEXT:    [[I57:%.*]] = fadd float 0.000000e+00, [[I56]]
+; MAX32-NEXT:    [[I58:%.*]] = fmul float [[I9]], [[FVAL]]
+; MAX32-NEXT:    [[I59:%.*]] = fadd float 0.000000e+00, [[I58]]
+; MAX32-NEXT:    [[I60:%.*]] = fmul float [[I]], [[FVAL]]
+; MAX32-NEXT:    [[I61:%.*]] = fadd float 0.000000e+00, [[I60]]
+; MAX32-NEXT:    [[I62:%.*]] = fmul float [[I3]], [[FVAL]]
+; MAX32-NEXT:    [[I63:%.*]] = fadd float 0.000000e+00, [[I62]]
+; MAX32-NEXT:    [[I64:%.*]] = fmul float [[I6]], [[FVAL]]
+; MAX32-NEXT:    [[I65:%.*]] = fadd float 0.000000e+00, [[I64]]
+; MAX32-NEXT:    [[I66:%.*]] = fmul float [[I9]], [[FVAL]]
+; MAX32-NEXT:    [[I67:%.*]] = fadd float 0.000000e+00, [[I66]]
 ; MAX32-NEXT:    switch i32 undef, label [[BB5:%.*]] [
 ; MAX32-NEXT:    i32 0, label [[BB2:%.*]]
 ; MAX32-NEXT:    i32 1, label [[BB3:%.*]]
@@ -104,89 +84,42 @@ define void @phi_float32(half %hval, float %fval) {
 ; MAX32:       bb3:
 ; MAX32-NEXT:    br label [[BB2]]
 ; MAX32:       bb4:
-; MAX32-NEXT:    [[TMP87:%.*]] = insertelement <32 x float> [[TMP40]], float [[FVAL]], i32 1
-; MAX32-NEXT:    [[TMP88:%.*]] = insertelement <32 x float> [[TMP87]], float [[FVAL]], i32 2
-; MAX32-NEXT:    [[TMP89:%.*]] = extractelement <32 x float> [[TMP38]], i32 3
-; MAX32-NEXT:    [[TMP90:%.*]] = insertelement <32 x float> [[TMP88]], float [[TMP89]], i32 3
-; MAX32-NEXT:    [[TMP91:%.*]] = insertelement <32 x float> [[TMP90]], float [[TMP45]], i32 4
-; MAX32-NEXT:    [[TMP92:%.*]] = insertelement <32 x float> [[TMP91]], float [[FVAL]], i32 5
-; MAX32-NEXT:    [[TMP93:%.*]] = insertelement <32 x float> [[TMP92]], float [[FVAL]], i32 6
-; MAX32-NEXT:    [[TMP94:%.*]] = extractelement <32 x float> [[TMP38]], i32 7
-; MAX32-NEXT:    [[TMP95:%.*]] = insertelement <32 x float> [[TMP93]], float [[TMP94]], i32 7
-; MAX32-NEXT:    [[TMP96:%.*]] = extractelement <32 x float> [[TMP38]], i32 8
-; MAX32-NEXT:    [[TMP97:%.*]] = insertelement <32 x float> [[TMP95]], float [[TMP96]], i32 8
-; MAX32-NEXT:    [[TMP98:%.*]] = insertelement <32 x float> [[TMP97]], float [[FVAL]], i32 9
-; MAX32-NEXT:    [[TMP99:%.*]] = insertelement <32 x float> [[TMP98]], float [[FVAL]], i32 10
-; MAX32-NEXT:    [[TMP100:%.*]] = insertelement <32 x float> [[TMP99]], float [[TMP55]], i32 11
-; MAX32-NEXT:    [[TMP101:%.*]] = extractelement <32 x float> [[TMP38]], i32 12
-; MAX32-NEXT:    [[TMP102:%.*]] = insertelement <32 x float> [[TMP100]], float [[TMP101]], i32 12
-; MAX32-NEXT:    [[TMP103:%.*]] = insertelement <32 x float> [[TMP102]], float [[FVAL]], i32 13
-; MAX32-NEXT:    [[TMP104:%.*]] = insertelement <32 x float> [[TMP103]], float [[FVAL]], i32 14
-; MAX32-NEXT:    [[TMP105:%.*]] = insertelement <32 x float> [[TMP104]], float [[TMP61]], i32 15
-; MAX32-NEXT:    [[TMP106:%.*]] = extractelement <32 x float> [[TMP38]], i32 16
-; MAX32-NEXT:    [[TMP107:%.*]] = insertelement <32 x float> [[TMP105]], float [[TMP106]], i32 16
-; MAX32-NEXT:    [[TMP108:%.*]] = insertelement <32 x float> [[TMP107]], float [[FVAL]], i32 17
-; MAX32-NEXT:    [[TMP109:%.*]] = insertelement <32 x float> [[TMP108]], float [[FVAL]], i32 18
-; MAX32-NEXT:    [[TMP110:%.*]] = insertelement <32 x float> [[TMP109]], float [[TMP67]], i32 19
-; MAX32-NEXT:    [[TMP111:%.*]] = extractelement <32 x float> [[TMP38]], i32 20
-; MAX32-NEXT:    [[TMP112:%.*]] = insertelement <32 x float> [[TMP110]], float [[TMP111]], i32 20
-; MAX32-NEXT:    [[TMP113:%.*]] = insertelement <32 x float> [[TMP112]], float [[FVAL]], i32 21
-; MAX32-NEXT:    [[TMP114:%.*]] = insertelement <32 x float> [[TMP113]], float [[FVAL]], i32 22
-; MAX32-NEXT:    [[TMP115:%.*]] = insertelement <32 x float> [[TMP114]], float [[TMP73]], i32 23
-; MAX32-NEXT:    [[TMP116:%.*]] = extractelement <32 x float> [[TMP38]], i32 24
-; MAX32-NEXT:    [[TMP117:%.*]] = insertelement <32 x float> [[TMP115]], float [[TMP116]], i32 24
-; MAX32-NEXT:    [[TMP118:%.*]] = insertelement <32 x float> [[TMP117]], float [[FVAL]], i32 25
-; MAX32-NEXT:    [[TMP119:%.*]] = insertelement <32 x float> [[TMP118]], float [[FVAL]], i32 26
-; MAX32-NEXT:    [[TMP120:%.*]] = insertelement <32 x float> [[TMP119]], float [[TMP79]], i32 27
-; MAX32-NEXT:    [[TMP121:%.*]] = extractelement <32 x float> [[TMP38]], i32 28
-; MAX32-NEXT:    [[TMP122:%.*]] = insertelement <32 x float> [[TMP120]], float [[TMP121]], i32 28
-; MAX32-NEXT:    [[TMP123:%.*]] = insertelement <32 x float> [[TMP122]], float [[FVAL]], i32 29
-; MAX32-NEXT:    [[TMP124:%.*]] = insertelement <32 x float> [[TMP123]], float [[FVAL]], i32 30
-; MAX32-NEXT:    [[TMP125:%.*]] = insertelement <32 x float> [[TMP124]], float [[TMP85]], i32 31
 ; MAX32-NEXT:    br label [[BB2]]
 ; MAX32:       bb5:
-; MAX32-NEXT:    [[TMP126:%.*]] = insertelement <32 x float> [[TMP5]], float [[TMP41]], i32 1
-; MAX32-NEXT:    [[TMP127:%.*]] = insertelement <32 x float> [[TMP126]], float [[FVAL]], i32 2
-; MAX32-NEXT:    [[TMP128:%.*]] = extractelement <32 x float> [[TMP38]], i32 3
-; MAX32-NEXT:    [[TMP129:%.*]] = insertelement <32 x float> [[TMP127]], float [[TMP128]], i32 3
-; MAX32-NEXT:    [[TMP130:%.*]] = insertelement <32 x float> [[TMP129]], float [[FVAL]], i32 4
-; MAX32-NEXT:    [[TMP131:%.*]] = insertelement <32 x float> [[TMP130]], float [[TMP47]], i32 5
-; MAX32-NEXT:    [[TMP132:%.*]] = insertelement <32 x float> [[TMP131]], float [[FVAL]], i32 6
-; MAX32-NEXT:    [[TMP133:%.*]] = extractelement <32 x float> [[TMP38]], i32 7
-; MAX32-NEXT:    [[TMP134:%.*]] = insertelement <32 x float> [[TMP132]], float [[TMP133]], i32 7
-; MAX32-NEXT:    [[TMP135:%.*]] = extractelement <32 x float> [[TMP38]], i32 8
-; MAX32-NEXT:    [[TMP136:%.*]] = insertelement <32 x float> [[TMP134]], float [[TMP135]], i32 8
-; MAX32-NEXT:    [[TMP137:%.*]] = insertelement <32 x float> [[TMP136]], float [[FVAL]], i32 9
-; MAX32-NEXT:    [[TMP138:%.*]] = insertelement <32 x float> [[TMP137]], float [[TMP53]], i32 10
-; MAX32-NEXT:    [[TMP139:%.*]] = insertelement <32 x float> [[TMP138]], float [[FVAL]], i32 11
-; MAX32-NEXT:    [[TMP140:%.*]] = extractelement <32 x float> [[TMP38]], i32 12
-; MAX32-NEXT:    [[TMP141:%.*]] = insertelement <32 x float> [[TMP139]], float [[TMP140]], i32 12
-; MAX32-NEXT:    [[TMP142:%.*]] = insertelement <32 x float> [[TMP141]], float [[FVAL]], i32 13
-; MAX32-NEXT:    [[TMP143:%.*]] = insertelement <32 x float> [[TMP142]], float [[TMP59]], i32 14
-; MAX32-NEXT:    [[TMP144:%.*]] = insertelement <32 x float> [[TMP143]], float [[FVAL]], i32 15
-; MAX32-NEXT:    [[TMP145:%.*]] = extractelement <32 x float> [[TMP38]], i32 16
-; MAX32-NEXT:    [[TMP146:%.*]] = insertelement <32 x float> [[TMP144]], float [[TMP145]], i32 16
-; MAX32-NEXT:    [[TMP147:%.*]] = insertelement <32 x float> [[TMP146]], float [[FVAL]], i32 17
-; MAX32-NEXT:    [[TMP148:%.*]] = insertelement <32 x float> [[TMP147]], float [[TMP65]], i32 18
-; MAX32-NEXT:    [[TMP149:%.*]] = insertelement <32 x float> [[TMP148]], float [[FVAL]], i32 19
-; MAX32-NEXT:    [[TMP150:%.*]] = extractelement <32 x float> [[TMP38]], i32 20
-; MAX32-NEXT:    [[TMP151:%.*]] = insertelement <32 x float> [[TMP149]], float [[TMP150]], i32 20
-; MAX32-NEXT:    [[TMP152:%.*]] = insertelement <32 x float> [[TMP151]], float [[FVAL]], i32 21
-; MAX32-NEXT:    [[TMP153:%.*]] = insertelement <32 x float> [[TMP152]], float [[TMP71]], i32 22
-; MAX32-NEXT:    [[TMP154:%.*]] = insertelement <32 x float> [[TMP153]], float [[FVAL]], i32 23
-; MAX32-NEXT:    [[TMP155:%.*]] = extractelement <32 x float> [[TMP38]], i32 24
-; MAX32-NEXT:    [[TMP156:%.*]] = insertelement <32 x float> [[TMP154]], float [[TMP155]], i32 24
-; MAX32-NEXT:    [[TMP157:%.*]] = insertelement <32 x float> [[TMP156]], float [[FVAL]], i32 25
-; MAX32-NEXT:    [[TMP158:%.*]] = insertelement <32 x float> [[TMP157]], float [[TMP77]], i32 26
-; MAX32-NEXT:    [[TMP159:%.*]] = insertelement <32 x float> [[TMP158]], float [[FVAL]], i32 27
-; MAX32-NEXT:    [[TMP160:%.*]] = extractelement <32 x float> [[TMP38]], i32 28
-; MAX32-NEXT:    [[TMP161:%.*]] = insertelement <32 x float> [[TMP159]], float [[TMP160]], i32 28
-; MAX32-NEXT:    [[TMP162:%.*]] = insertelement <32 x float> [[TMP161]], float [[FVAL]], i32 29
-; MAX32-NEXT:    [[TMP163:%.*]] = insertelement <32 x float> [[TMP162]], float [[TMP83]], i32 30
-; MAX32-NEXT:    [[TMP164:%.*]] = insertelement <32 x float> [[TMP163]], float [[FVAL]], i32 31
 ; MAX32-NEXT:    br label [[BB2]]
 ; MAX32:       bb2:
-; MAX32-NEXT:    [[TMP165:%.*]] = phi <32 x float> [ [[TMP38]], [[BB3]] ], [ [[TMP125]], [[BB4]] ], [ [[TMP164]], [[BB5]] ], [ [[TMP86]], [[BB1]] ]
+; MAX32-NEXT:    [[PHI1:%.*]] = phi float [ [[I19]], [[BB3]] ], [ [[I19]], [[BB4]] ], [ [[FVAL]], [[BB5]] ], [ [[I19]], [[BB1]] ]
+; MAX32-NEXT:    [[PHI2:%.*]] = phi float [ [[I17]], [[BB3]] ], [ [[FVAL]], [[BB4]] ], [ [[I17]], [[BB5]] ], [ [[I17]], [[BB1]] ]
+; MAX32-NEXT:    [[PHI3:%.*]] = phi float [ [[I15]], [[BB3]] ], [ [[FVAL]], [[BB4]] ], [ [[FVAL]], [[BB5]] ], [ [[FVAL]], [[BB1]] ]
+; MAX32-NEXT:    [[PHI4:%.*]] = phi float [ [[I13]], [[BB3]] ], [ [[I13]], [[BB4]] ], [ [[I13]], [[BB5]] ], [ [[FVAL]], [[BB1]] ]
+; MAX32-NEXT:    [[PHI5:%.*]] = phi float [ [[I11]], [[BB3]] ], [ [[I11]], [[BB4]] ], [ [[FVAL]], [[BB5]] ], [ [[I11]], [[BB1]] ]
+; MAX32-NEXT:    [[PHI6:%.*]] = phi float [ [[I8]], [[BB3]] ], [ [[FVAL]], [[BB4]] ], [ [[I8]], [[BB5]] ], [ [[I8]], [[BB1]] ]
+; MAX32-NEXT:    [[PHI7:%.*]] = phi float [ [[I5]], [[BB3]] ], [ [[FVAL]], [[BB4]] ], [ [[FVAL]], [[BB5]] ], [ [[FVAL]], [[BB1]] ]
+; MAX32-NEXT:    [[PHI8:%.*]] = phi float [ [[I2]], [[BB3]] ], [ [[I2]], [[BB4]] ], [ [[I2]], [[BB5]] ], [ [[FVAL]], [[BB1]] ]
+; MAX32-NEXT:    [[PHI9:%.*]] = phi float [ [[I21]], [[BB3]] ], [ [[I21]], [[BB4]] ], [ [[I21]], [[BB5]] ], [ [[FVAL]], [[BB1]] ]
+; MAX32-NEXT:    [[PHI10:%.*]] = phi float [ [[I23]], [[BB3]] ], [ [[FVAL]], [[BB4]] ], [ [[FVAL]], [[BB5]] ], [ [[FVAL]], [[BB1]] ]
+; MAX32-NEXT:    [[PHI11:%.*]] = phi float [ [[I25]], [[BB3]] ], [ [[FVAL]], [[BB4]] ], [ [[I25]], [[BB5]] ], [ [[I25]], [[BB1]] ]
+; MAX32-NEXT:    [[PHI12:%.*]] = phi float [ [[I27]], [[BB3]] ], [ [[I27]], [[BB4]] ], [ [[FVAL]], [[BB5]] ], [ [[I27]], [[BB1]] ]
+; MAX32-NEXT:    [[PHI13:%.*]] = phi float [ [[I29]], [[BB3]] ], [ [[I29]], [[BB4]] ], [ [[I29]], [[BB5]] ], [ [[FVAL]], [[BB1]] ]
+; MAX32-NEXT:    [[PHI14:%.*]] = phi float [ [[I31]], [[BB3]] ], [ [[FVAL]], [[BB4]] ], [ [[FVAL]], [[BB5]] ], [ [[FVAL]], [[BB1]] ]
+; MAX32-NEXT:    [[PHI15:%.*]] = phi float [ [[I33]], [[BB3]] ], [ [[FVAL]], [[BB4]] ], [ [[I33]], [[BB5]] ], [ [[I33]], [[BB1]] ]
+; MAX32-NEXT:    [[PHI16:%.*]] = phi float [ [[I35]], [[BB3]] ], [ [[I35]], [[BB4]] ], [ [[FVAL]], [[BB5]] ], [ [[I35]], [[BB1]] ]
+; MAX32-NEXT:    [[PHI17:%.*]] = phi float [ [[I37]], [[BB3]] ], [ [[I37]], [[BB4]] ], [ [[I37]], [[BB5]] ], [ [[FVAL]], [[BB1]] ]
+; MAX32-NEXT:    [[PHI18:%.*]] = phi float [ [[I39]], [[BB3]] ], [ [[FVAL]], [[BB4]] ], [ [[FVAL]], [[BB5]] ], [ [[FVAL]], [[BB1]] ]
+; MAX32-NEXT:    [[PHI19:%.*]] = phi float [ [[I41]], [[BB3]] ], [ [[FVAL]], [[BB4]] ], [ [[I41]], [[BB5]] ], [ [[I41]], [[BB1]] ]
+; MAX32-NEXT:    [[PHI20:%.*]] = phi float [ [[I43]], [[BB3]] ], [ [[I43]], [[BB4]] ], [ [[FVAL]], [[BB5]] ], [ [[I43]], [[BB1]] ]
+; MAX32-NEXT:    [[PHI21:%.*]] = phi float [ [[I45]], [[BB3]] ], [ [[I45]], [[BB4]] ], [ [[I45]], [[BB5]] ], [ [[FVAL]], [[BB1]] ]
+; MAX32-NEXT:    [[PHI22:%.*]] = phi float [ [[I47]], [[BB3]] ], [ [[FVAL]], [[BB4]] ], [ [[FVAL]], [[BB5]] ], [ [[FVAL]], [[BB1]] ]
+; MAX32-NEXT:    [[PHI23:%.*]] = phi float [ [[I49]], [[BB3]] ], [ [[FVAL]], [[BB4]] ], [ [[I49]], [[BB5]] ], [ [[I49]], [[BB1]] ]
+; MAX32-NEXT:    [[PHI24:%.*]] = phi float [ [[I51]], [[BB3]] ], [ [[I51]], [[BB4]] ], [ [[FVAL]], [[BB5]] ], [ [[I51]], [[BB1]] ]
+; MAX32-NEXT:    [[PHI25:%.*]] = phi float [ [[I53]], [[BB3]] ], [ [[I53]], [[BB4]] ], [ [[I53]], [[BB5]] ], [ [[FVAL]], [[BB1]] ]
+; MAX32-NEXT:    [[PHI26:%.*]] = phi float [ [[I55]], [[BB3]] ], [ [[FVAL]], [[BB4]] ], [ [[FVAL]], [[BB5]] ], [ [[FVAL]], [[BB1]] ]
+; MAX32-NEXT:    [[PHI27:%.*]] = phi float [ [[I57]], [[BB3]] ], [ [[FVAL]], [[BB4]] ], [ [[I57]], [[BB5]] ], [ [[I57]], [[BB1]] ]
+; MAX32-NEXT:    [[PHI28:%.*]] = phi float [ [[I59]], [[BB3]] ], [ [[I59]], [[BB4]] ], [ [[FVAL]], [[BB5]] ], [ [[I59]], [[BB1]] ]
+; MAX32-NEXT:    [[PHI29:%.*]] = phi float [ [[I61]], [[BB3]] ], [ [[I61]], [[BB4]] ], [ [[I61]], [[BB5]] ], [ [[FVAL]], [[BB1]] ]
+; MAX32-NEXT:    [[PHI30:%.*]] = phi float [ [[I63]], [[BB3]] ], [ [[FVAL]], [[BB4]] ], [ [[FVAL]], [[BB5]] ], [ [[FVAL]], [[BB1]] ]
+; MAX32-NEXT:    [[PHI31:%.*]] = phi float [ [[I65]], [[BB3]] ], [ [[FVAL]], [[BB4]] ], [ [[I65]], [[BB5]] ], [ [[I65]], [[BB1]] ]
+; MAX32-NEXT:    [[PHI32:%.*]] = phi float [ [[I67]], [[BB3]] ], [ [[I67]], [[BB4]] ], [ [[FVAL]], [[BB5]] ], [ [[I67]], [[BB1]] ]
 ; MAX32-NEXT:    ret void
 ;
 ; MAX256-LABEL: @phi_float32(
@@ -198,89 +131,77 @@ define void @phi_float32(half %hval, float %fval) {
 ; MAX256-NEXT:    [[TMP2:%.*]] = insertelement <4 x half> [[TMP1]], half [[HVAL]], i32 2
 ; MAX256-NEXT:    [[TMP3:%.*]] = insertelement <4 x half> [[TMP2]], half [[HVAL]], i32 3
 ; MAX256-NEXT:    [[TMP4:%.*]] = fpext <4 x half> [[TMP3]] to <4 x float>
-; MAX256-NEXT:    [[SHUFFLE:%.*]] = shufflevector <4 x float> [[TMP4]], <4 x float> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 3, i32 2, i32 1, i32 0, i32 3, i32 2, i32 1, i32 0, i32 3, i32 2, i32 1, i32 0, i32 3, i32 2, i32 1, i32 0, i32 3, i32 2, i32 1, i32 0, i32 3, i32 2, i32 1, i32 0>
-; MAX256-NEXT:    [[TMP5:%.*]] = insertelement <32 x float> undef, float [[FVAL:%.*]], i32 0
-; MAX256-NEXT:    [[TMP6:%.*]] = insertelement <32 x float> [[TMP5]], float [[FVAL]], i32 1
-; MAX256-NEXT:    [[TMP7:%.*]] = insertelement <32 x float> [[TMP6]], float [[FVAL]], i32 2
-; MAX256-NEXT:    [[TMP8:%.*]] = insertelement <32 x float> [[TMP7]], float [[FVAL]], i32 3
-; MAX256-NEXT:    [[TMP9:%.*]] = insertelement <32 x float> [[TMP8]], float [[FVAL]], i32 4
-; MAX256-NEXT:    [[TMP10:%.*]] = insertelement <32 x float> [[TMP9]], float [[FVAL]], i32 5
-; MAX256-NEXT:    [[TMP11:%.*]] = insertelement <32 x float> [[TMP10]], float [[FVAL]], i32 6
-; MAX256-NEXT:    [[TMP12:%.*]] = insertelement <32 x float> [[TMP11]], float [[FVAL]], i32 7
-; MAX256-NEXT:    [[TMP13:%.*]] = insertelement <32 x float> [[TMP12]], float [[FVAL]], i32 8
-; MAX256-NEXT:    [[TMP14:%.*]] = insertelement <32 x float> [[TMP13]], float [[FVAL]], i32 9
-; MAX256-NEXT:    [[TMP15:%.*]] = insertelement <32 x float> [[TMP14]], float [[FVAL]], i32 10
-; MAX256-NEXT:    [[TMP16:%.*]] = insertelement <32 x float> [[TMP15]], float [[FVAL]], i32 11
-; MAX256-NEXT:    [[TMP17:%.*]] = insertelement <32 x float> [[TMP16]], float [[FVAL]], i32 12
-; MAX256-NEXT:    [[TMP18:%.*]] = insertelement <32 x float> [[TMP17]], float [[FVAL]], i32 13
-; MAX256-NEXT:    [[TMP19:%.*]] = insertelement <32 x float> [[TMP18]], float [[FVAL]], i32 14
-; MAX256-NEXT:    [[TMP20:%.*]] = insertelement <32 x float> [[TMP19]], float [[FVAL]], i32 15
-; MAX256-NEXT:    [[TMP21:%.*]] = insertelement <32 x float> [[TMP20]], float [[FVAL]], i32 16
-; MAX256-NEXT:    [[TMP22:%.*]] = insertelement <32 x float> [[TMP21]], float [[FVAL]], i32 17
-; MAX256-NEXT:    [[TMP23:%.*]] = insertelement <32 x float> [[TMP22]], float [[FVAL]], i32 18
-; MAX256-NEXT:    [[TMP24:%.*]] = insertelement <32 x float> [[TMP23]], float [[FVAL]], i32 19
-; MAX256-NEXT:    [[TMP25:%.*]] = insertelement <32 x float> [[TMP24]], float [[FVAL]], i32 20
-; MAX256-NEXT:    [[TMP26:%.*]] = insertelement <32 x float> [[TMP25]], float [[FVAL]], i32 21
-; MAX256-NEXT:    [[TMP27:%.*]] = insertelement <32 x float> [[TMP26]], float [[FVAL]], i32 22
-; MAX256-NEXT:    [[TMP28:%.*]] = insertelement <32 x float> [[TMP27]], float [[FVAL]], i32 23
-; MAX256-NEXT:    [[TMP29:%.*]] = insertelement <32 x float> [[TMP28]], float [[FVAL]], i32 24
-; MAX256-NEXT:    [[TMP30:%.*]] = insertelement <32 x float> [[TMP29]], float [[FVAL]], i32 25
-; MAX256-NEXT:    [[TMP31:%.*]] = insertelement <32 x float> [[TMP30]], float [[FVAL]], i32 26
-; MAX256-NEXT:    [[TMP32:%.*]] = insertelement <32 x float> [[TMP31]], float [[FVAL]], i32 27
-; MAX256-NEXT:    [[TMP33:%.*]] = insertelement <32 x float> [[TMP32]], float [[FVAL]], i32 28
-; MAX256-NEXT:    [[TMP34:%.*]] = insertelement <32 x float> [[TMP33]], float [[FVAL]], i32 29
-; MAX256-NEXT:    [[TMP35:%.*]] = insertelement <32 x float> [[TMP34]], float [[FVAL]], i32 30
-; MAX256-NEXT:    [[TMP36:%.*]] = insertelement <32 x float> [[TMP35]], float [[FVAL]], i32 31
-; MAX256-NEXT:    [[TMP37:%.*]] = fmul <32 x float> [[SHUFFLE]], [[TMP36]]
-; MAX256-NEXT:    [[TMP38:%.*]] = fadd <32 x float> zeroinitializer, [[TMP37]]
-; MAX256-NEXT:    [[TMP39:%.*]] = extractelement <32 x float> [[TMP38]], i32 0
-; MAX256-NEXT:    [[TMP40:%.*]] = insertelement <32 x float> undef, float [[TMP39]], i32 0
-; MAX256-NEXT:    [[TMP41:%.*]] = extractelement <32 x float> [[TMP38]], i32 1
-; MAX256-NEXT:    [[TMP42:%.*]] = insertelement <32 x float> [[TMP40]], float [[TMP41]], i32 1
-; MAX256-NEXT:    [[TMP43:%.*]] = insertelement <32 x float> [[TMP42]], float [[FVAL]], i32 2
-; MAX256-NEXT:    [[TMP44:%.*]] = insertelement <32 x float> [[TMP43]], float [[FVAL]], i32 3
-; MAX256-NEXT:    [[TMP45:%.*]] = extractelement <32 x float> [[TMP38]], i32 4
-; MAX256-NEXT:    [[TMP46:%.*]] = insertelement <32 x float> [[TMP44]], float [[TMP45]], i32 4
-; MAX256-NEXT:    [[TMP47:%.*]] = extractelement <32 x float> [[TMP38]], i32 5
-; MAX256-NEXT:    [[TMP48:%.*]] = insertelement <32 x float> [[TMP46]], float [[TMP47]], i32 5
-; MAX256-NEXT:    [[TMP49:%.*]] = insertelement <32 x float> [[TMP48]], float [[FVAL]], i32 6
-; MAX256-NEXT:    [[TMP50:%.*]] = insertelement <32 x float> [[TMP49]], float [[FVAL]], i32 7
-; MAX256-NEXT:    [[TMP51:%.*]] = insertelement <32 x float> [[TMP50]], float [[FVAL]], i32 8
-; MAX256-NEXT:    [[TMP52:%.*]] = insertelement <32 x float> [[TMP51]], float [[FVAL]], i32 9
-; MAX256-NEXT:    [[TMP53:%.*]] = extractelement <32 x float> [[TMP38]], i32 10
-; MAX256-NEXT:    [[TMP54:%.*]] = insertelement <32 x float> [[TMP52]], float [[TMP53]], i32 10
-; MAX256-NEXT:    [[TMP55:%.*]] = extractelement <32 x float> [[TMP38]], i32 11
-; MAX256-NEXT:    [[TMP56:%.*]] = insertelement <32 x float> [[TMP54]], float [[TMP55]], i32 11
-; MAX256-NEXT:    [[TMP57:%.*]] = insertelement <32 x float> [[TMP56]], float [[FVAL]], i32 12
-; MAX256-NEXT:    [[TMP58:%.*]] = insertelement <32 x float> [[TMP57]], float [[FVAL]], i32 13
-; MAX256-NEXT:    [[TMP59:%.*]] = extractelement <32 x float> [[TMP38]], i32 14
-; MAX256-NEXT:    [[TMP60:%.*]] = insertelement <32 x float> [[TMP58]], float [[TMP59]], i32 14
-; MAX256-NEXT:    [[TMP61:%.*]] = extractelement <32 x float> [[TMP38]], i32 15
-; MAX256-NEXT:    [[TMP62:%.*]] = insertelement <32 x float> [[TMP60]], float [[TMP61]], i32 15
-; MAX256-NEXT:    [[TMP63:%.*]] = insertelement <32 x float> [[TMP62]], float [[FVAL]], i32 16
-; MAX256-NEXT:    [[TMP64:%.*]] = insertelement <32 x float> [[TMP63]], float [[FVAL]], i32 17
-; MAX256-NEXT:    [[TMP65:%.*]] = extractelement <32 x float> [[TMP38]], i32 18
-; MAX256-NEXT:    [[TMP66:%.*]] = insertelement <32 x float> [[TMP64]], float [[TMP65]], i32 18
-; MAX256-NEXT:    [[TMP67:%.*]] = extractelement <32 x float> [[TMP38]], i32 19
-; MAX256-NEXT:    [[TMP68:%.*]] = insertelement <32 x float> [[TMP66]], float [[TMP67]], i32 19
-; MAX256-NEXT:    [[TMP69:%.*]] = insertelement <32 x float> [[TMP68]], float [[FVAL]], i32 20
-; MAX256-NEXT:    [[TMP70:%.*]] = insertelement <32 x float> [[TMP69]], float [[FVAL]], i32 21
-; MAX256-NEXT:    [[TMP71:%.*]] = extractelement <32 x float> [[TMP38]], i32 22
-; MAX256-NEXT:    [[TMP72:%.*]] = insertelement <32 x float> [[TMP70]], float [[TMP71]], i32 22
-; MAX256-NEXT:    [[TMP73:%.*]] = extractelement <32 x float> [[TMP38]], i32 23
-; MAX256-NEXT:    [[TMP74:%.*]] = insertelement <32 x float> [[TMP72]], float [[TMP73]], i32 23
-; MAX256-NEXT:    [[TMP75:%.*]] = insertelement <32 x float> [[TMP74]], float [[FVAL]], i32 24
-; MAX256-NEXT:    [[TMP76:%.*]] = insertelement <32 x float> [[TMP75]], float [[FVAL]], i32 25
-; MAX256-NEXT:    [[TMP77:%.*]] = extractelement <32 x float> [[TMP38]], i32 26
-; MAX256-NEXT:    [[TMP78:%.*]] = insertelement <32 x float> [[TMP76]], float [[TMP77]], i32 26
-; MAX256-NEXT:    [[TMP79:%.*]] = extractelement <32 x float> [[TMP38]], i32 27
-; MAX256-NEXT:    [[TMP80:%.*]] = insertelement <32 x float> [[TMP78]], float [[TMP79]], i32 27
-; MAX256-NEXT:    [[TMP81:%.*]] = insertelement <32 x float> [[TMP80]], float [[FVAL]], i32 28
-; MAX256-NEXT:    [[TMP82:%.*]] = insertelement <32 x float> [[TMP81]], float [[FVAL]], i32 29
-; MAX256-NEXT:    [[TMP83:%.*]] = extractelement <32 x float> [[TMP38]], i32 30
-; MAX256-NEXT:    [[TMP84:%.*]] = insertelement <32 x float> [[TMP82]], float [[TMP83]], i32 30
-; MAX256-NEXT:    [[TMP85:%.*]] = extractelement <32 x float> [[TMP38]], i32 31
-; MAX256-NEXT:    [[TMP86:%.*]] = insertelement <32 x float> [[TMP84]], float [[TMP85]], i32 31
+; MAX256-NEXT:    [[SHUFFLE:%.*]] = shufflevector <4 x float> [[TMP4]], <4 x float> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3>
+; MAX256-NEXT:    [[TMP5:%.*]] = insertelement <8 x float> undef, float [[FVAL:%.*]], i32 0
+; MAX256-NEXT:    [[TMP6:%.*]] = insertelement <8 x float> [[TMP5]], float [[FVAL]], i32 1
+; MAX256-NEXT:    [[TMP7:%.*]] = insertelement <8 x float> [[TMP6]], float [[FVAL]], i32 2
+; MAX256-NEXT:    [[TMP8:%.*]] = insertelement <8 x float> [[TMP7]], float [[FVAL]], i32 3
+; MAX256-NEXT:    [[TMP9:%.*]] = insertelement <8 x float> [[TMP8]], float [[FVAL]], i32 4
+; MAX256-NEXT:    [[TMP10:%.*]] = insertelement <8 x float> [[TMP9]], float [[FVAL]], i32 5
+; MAX256-NEXT:    [[TMP11:%.*]] = insertelement <8 x float> [[TMP10]], float [[FVAL]], i32 6
+; MAX256-NEXT:    [[TMP12:%.*]] = insertelement <8 x float> [[TMP11]], float [[FVAL]], i32 7
+; MAX256-NEXT:    [[TMP13:%.*]] = fmul <8 x float> [[SHUFFLE]], [[TMP12]]
+; MAX256-NEXT:    [[TMP14:%.*]] = fadd <8 x float> zeroinitializer, [[TMP13]]
+; MAX256-NEXT:    [[TMP15:%.*]] = extractelement <8 x float> [[SHUFFLE]], i32 3
+; MAX256-NEXT:    [[TMP16:%.*]] = extractelement <8 x float> [[SHUFFLE]], i32 2
+; MAX256-NEXT:    [[TMP17:%.*]] = extractelement <8 x float> [[SHUFFLE]], i32 1
+; MAX256-NEXT:    [[TMP18:%.*]] = extractelement <8 x float> [[SHUFFLE]], i32 0
+; MAX256-NEXT:    [[TMP19:%.*]] = insertelement <8 x float> undef, float [[TMP15]], i32 0
+; MAX256-NEXT:    [[TMP20:%.*]] = insertelement <8 x float> [[TMP19]], float [[TMP16]], i32 1
+; MAX256-NEXT:    [[TMP21:%.*]] = insertelement <8 x float> [[TMP20]], float [[TMP17]], i32 2
+; MAX256-NEXT:    [[TMP22:%.*]] = insertelement <8 x float> [[TMP21]], float [[TMP18]], i32 3
+; MAX256-NEXT:    [[TMP23:%.*]] = insertelement <8 x float> [[TMP22]], float [[TMP15]], i32 4
+; MAX256-NEXT:    [[TMP24:%.*]] = insertelement <8 x float> [[TMP23]], float [[TMP16]], i32 5
+; MAX256-NEXT:    [[TMP25:%.*]] = insertelement <8 x float> [[TMP24]], float [[TMP17]], i32 6
+; MAX256-NEXT:    [[TMP26:%.*]] = insertelement <8 x float> [[TMP25]], float [[TMP18]], i32 7
+; MAX256-NEXT:    [[TMP27:%.*]] = fmul <8 x float> [[TMP26]], [[TMP12]]
+; MAX256-NEXT:    [[TMP28:%.*]] = fadd <8 x float> zeroinitializer, [[TMP27]]
+; MAX256-NEXT:    [[TMP29:%.*]] = fmul <8 x float> [[TMP26]], [[TMP12]]
+; MAX256-NEXT:    [[TMP30:%.*]] = fadd <8 x float> zeroinitializer, [[TMP29]]
+; MAX256-NEXT:    [[TMP31:%.*]] = fmul <8 x float> [[TMP26]], [[TMP12]]
+; MAX256-NEXT:    [[TMP32:%.*]] = fadd <8 x float> zeroinitializer, [[TMP31]]
+; MAX256-NEXT:    [[TMP33:%.*]] = extractelement <8 x float> [[TMP14]], i32 0
+; MAX256-NEXT:    [[TMP34:%.*]] = insertelement <8 x float> undef, float [[TMP33]], i32 0
+; MAX256-NEXT:    [[TMP35:%.*]] = extractelement <8 x float> [[TMP14]], i32 1
+; MAX256-NEXT:    [[TMP36:%.*]] = insertelement <8 x float> [[TMP34]], float [[TMP35]], i32 1
+; MAX256-NEXT:    [[TMP37:%.*]] = insertelement <8 x float> [[TMP36]], float [[FVAL]], i32 2
+; MAX256-NEXT:    [[TMP38:%.*]] = insertelement <8 x float> [[TMP37]], float [[FVAL]], i32 3
+; MAX256-NEXT:    [[TMP39:%.*]] = extractelement <8 x float> [[TMP14]], i32 4
+; MAX256-NEXT:    [[TMP40:%.*]] = insertelement <8 x float> [[TMP38]], float [[TMP39]], i32 4
+; MAX256-NEXT:    [[TMP41:%.*]] = extractelement <8 x float> [[TMP14]], i32 5
+; MAX256-NEXT:    [[TMP42:%.*]] = insertelement <8 x float> [[TMP40]], float [[TMP41]], i32 5
+; MAX256-NEXT:    [[TMP43:%.*]] = insertelement <8 x float> [[TMP42]], float [[FVAL]], i32 6
+; MAX256-NEXT:    [[TMP44:%.*]] = insertelement <8 x float> [[TMP43]], float [[FVAL]], i32 7
+; MAX256-NEXT:    [[TMP45:%.*]] = extractelement <8 x float> [[TMP28]], i32 2
+; MAX256-NEXT:    [[TMP46:%.*]] = insertelement <8 x float> [[TMP6]], float [[TMP45]], i32 2
+; MAX256-NEXT:    [[TMP47:%.*]] = extractelement <8 x float> [[TMP28]], i32 3
+; MAX256-NEXT:    [[TMP48:%.*]] = insertelement <8 x float> [[TMP46]], float [[TMP47]], i32 3
+; MAX256-NEXT:    [[TMP49:%.*]] = insertelement <8 x float> [[TMP48]], float [[FVAL]], i32 4
+; MAX256-NEXT:    [[TMP50:%.*]] = insertelement <8 x float> [[TMP49]], float [[FVAL]], i32 5
+; MAX256-NEXT:    [[TMP51:%.*]] = extractelement <8 x float> [[TMP28]], i32 6
+; MAX256-NEXT:    [[TMP52:%.*]] = insertelement <8 x float> [[TMP50]], float [[TMP51]], i32 6
+; MAX256-NEXT:    [[TMP53:%.*]] = extractelement <8 x float> [[TMP28]], i32 7
+; MAX256-NEXT:    [[TMP54:%.*]] = insertelement <8 x float> [[TMP52]], float [[TMP53]], i32 7
+; MAX256-NEXT:    [[TMP55:%.*]] = extractelement <8 x float> [[TMP30]], i32 2
+; MAX256-NEXT:    [[TMP56:%.*]] = insertelement <8 x float> [[TMP6]], float [[TMP55]], i32 2
+; MAX256-NEXT:    [[TMP57:%.*]] = extractelement <8 x float> [[TMP30]], i32 3
+; MAX256-NEXT:    [[TMP58:%.*]] = insertelement <8 x float> [[TMP56]], float [[TMP57]], i32 3
+; MAX256-NEXT:    [[TMP59:%.*]] = insertelement <8 x float> [[TMP58]], float [[FVAL]], i32 4
+; MAX256-NEXT:    [[TMP60:%.*]] = insertelement <8 x float> [[TMP59]], float [[FVAL]], i32 5
+; MAX256-NEXT:    [[TMP61:%.*]] = extractelement <8 x float> [[TMP30]], i32 6
+; MAX256-NEXT:    [[TMP62:%.*]] = insertelement <8 x float> [[TMP60]], float [[TMP61]], i32 6
+; MAX256-NEXT:    [[TMP63:%.*]] = extractelement <8 x float> [[TMP30]], i32 7
+; MAX256-NEXT:    [[TMP64:%.*]] = insertelement <8 x float> [[TMP62]], float [[TMP63]], i32 7
+; MAX256-NEXT:    [[TMP65:%.*]] = extractelement <8 x float> [[TMP32]], i32 2
+; MAX256-NEXT:    [[TMP66:%.*]] = insertelement <8 x float> [[TMP6]], float [[TMP65]], i32 2
+; MAX256-NEXT:    [[TMP67:%.*]] = extractelement <8 x float> [[TMP32]], i32 3
+; MAX256-NEXT:    [[TMP68:%.*]] = insertelement <8 x float> [[TMP66]], float [[TMP67]], i32 3
+; MAX256-NEXT:    [[TMP69:%.*]] = insertelement <8 x float> [[TMP68]], float [[FVAL]], i32 4
+; MAX256-NEXT:    [[TMP70:%.*]] = insertelement <8 x float> [[TMP69]], float [[FVAL]], i32 5
+; MAX256-NEXT:    [[TMP71:%.*]] = extractelement <8 x float> [[TMP32]], i32 6
+; MAX256-NEXT:    [[TMP72:%.*]] = insertelement <8 x float> [[TMP70]], float [[TMP71]], i32 6
+; MAX256-NEXT:    [[TMP73:%.*]] = extractelement <8 x float> [[TMP32]], i32 7
+; MAX256-NEXT:    [[TMP74:%.*]] = insertelement <8 x float> [[TMP72]], float [[TMP73]], i32 7
 ; MAX256-NEXT:    switch i32 undef, label [[BB5:%.*]] [
 ; MAX256-NEXT:    i32 0, label [[BB2:%.*]]
 ; MAX256-NEXT:    i32 1, label [[BB3:%.*]]
@@ -289,89 +210,92 @@ define void @phi_float32(half %hval, float %fval) {
 ; MAX256:       bb3:
 ; MAX256-NEXT:    br label [[BB2]]
 ; MAX256:       bb4:
-; MAX256-NEXT:    [[TMP87:%.*]] = insertelement <32 x float> [[TMP40]], float [[FVAL]], i32 1
-; MAX256-NEXT:    [[TMP88:%.*]] = insertelement <32 x float> [[TMP87]], float [[FVAL]], i32 2
-; MAX256-NEXT:    [[TMP89:%.*]] = extractelement <32 x float> [[TMP38]], i32 3
-; MAX256-NEXT:    [[TMP90:%.*]] = insertelement <32 x float> [[TMP88]], float [[TMP89]], i32 3
-; MAX256-NEXT:    [[TMP91:%.*]] = insertelement <32 x float> [[TMP90]], float [[TMP45]], i32 4
-; MAX256-NEXT:    [[TMP92:%.*]] = insertelement <32 x float> [[TMP91]], float [[FVAL]], i32 5
-; MAX256-NEXT:    [[TMP93:%.*]] = insertelement <32 x float> [[TMP92]], float [[FVAL]], i32 6
-; MAX256-NEXT:    [[TMP94:%.*]] = extractelement <32 x float> [[TMP38]], i32 7
-; MAX256-NEXT:    [[TMP95:%.*]] = insertelement <32 x float> [[TMP93]], float [[TMP94]], i32 7
-; MAX256-NEXT:    [[TMP96:%.*]] = extractelement <32 x float> [[TMP38]], i32 8
-; MAX256-NEXT:    [[TMP97:%.*]] = insertelement <32 x float> [[TMP95]], float [[TMP96]], i32 8
-; MAX256-NEXT:    [[TMP98:%.*]] = insertelement <32 x float> [[TMP97]], float [[FVAL]], i32 9
-; MAX256-NEXT:    [[TMP99:%.*]] = insertelement <32 x float> [[TMP98]], float [[FVAL]], i32 10
-; MAX256-NEXT:    [[TMP100:%.*]] = insertelement <32 x float> [[TMP99]], float [[TMP55]], i32 11
-; MAX256-NEXT:    [[TMP101:%.*]] = extractelement <32 x float> [[TMP38]], i32 12
-; MAX256-NEXT:    [[TMP102:%.*]] = insertelement <32 x float> [[TMP100]], float [[TMP101]], i32 12
-; MAX256-NEXT:    [[TMP103:%.*]] = insertelement <32 x float> [[TMP102]], float [[FVAL]], i32 13
-; MAX256-NEXT:    [[TMP104:%.*]] = insertelement <32 x float> [[TMP103]], float [[FVAL]], i32 14
-; MAX256-NEXT:    [[TMP105:%.*]] = insertelement <32 x float> [[TMP104]], float [[TMP61]], i32 15
-; MAX256-NEXT:    [[TMP106:%.*]] = extractelement <32 x float> [[TMP38]], i32 16
-; MAX256-NEXT:    [[TMP107:%.*]] = insertelement <32 x float> [[TMP105]], float [[TMP106]], i32 16
-; MAX256-NEXT:    [[TMP108:%.*]] = insertelement <32 x float> [[TMP107]], float [[FVAL]], i32 17
-; MAX256-NEXT:    [[TMP109:%.*]] = insertelement <32 x float> [[TMP108]], float [[FVAL]], i32 18
-; MAX256-NEXT:    [[TMP110:%.*]] = insertelement <32 x float> [[TMP109]], float [[TMP67]], i32 19
-; MAX256-NEXT:    [[TMP111:%.*]] = extractelement <32 x float> [[TMP38]], i32 20
-; MAX256-NEXT:    [[TMP112:%.*]] = insertelement <32 x float> [[TMP110]], float [[TMP111]], i32 20
-; MAX256-NEXT:    [[TMP113:%.*]] = insertelement <32 x float> [[TMP112]], float [[FVAL]], i32 21
-; MAX256-NEXT:    [[TMP114:%.*]] = insertelement <32 x float> [[TMP113]], float [[FVAL]], i32 22
-; MAX256-NEXT:    [[TMP115:%.*]] = insertelement <32 x float> [[TMP114]], float [[TMP73]], i32 23
-; MAX256-NEXT:    [[TMP116:%.*]] = extractelement <32 x float> [[TMP38]], i32 24
-; MAX256-NEXT:    [[TMP117:%.*]] = insertelement <32 x float> [[TMP115]], float [[TMP116]], i32 24
-; MAX256-NEXT:    [[TMP118:%.*]] = insertelement <32 x float> [[TMP117]], float [[FVAL]], i32 25
-; MAX256-NEXT:    [[TMP119:%.*]] = insertelement <32 x float> [[TMP118]], float [[FVAL]], i32 26
-; MAX256-NEXT:    [[TMP120:%.*]] = insertelement <32 x float> [[TMP119]], float [[TMP79]], i32 27
-; MAX256-NEXT:    [[TMP121:%.*]] = extractelement <32 x float> [[TMP38]], i32 28
-; MAX256-NEXT:    [[TMP122:%.*]] = insertelement <32 x float> [[TMP120]], float [[TMP121]], i32 28
-; MAX256-NEXT:    [[TMP123:%.*]] = insertelement <32 x float> [[TMP122]], float [[FVAL]], i32 29
-; MAX256-NEXT:    [[TMP124:%.*]] = insertelement <32 x float> [[TMP123]], float [[FVAL]], i32 30
-; MAX256-NEXT:    [[TMP125:%.*]] = insertelement <32 x float> [[TMP124]], float [[TMP85]], i32 31
+; MAX256-NEXT:    [[TMP75:%.*]] = insertelement <8 x float> [[TMP34]], float [[FVAL]], i32 1
+; MAX256-NEXT:    [[TMP76:%.*]] = insertelement <8 x float> [[TMP75]], float [[FVAL]], i32 2
+; MAX256-NEXT:    [[TMP77:%.*]] = extractelement <8 x float> [[TMP14]], i32 3
+; MAX256-NEXT:    [[TMP78:%.*]] = insertelement <8 x float> [[TMP76]], float [[TMP77]], i32 3
+; MAX256-NEXT:    [[TMP79:%.*]] = insertelement <8 x float> [[TMP78]], float [[TMP39]], i32 4
+; MAX256-NEXT:    [[TMP80:%.*]] = insertelement <8 x float> [[TMP79]], float [[FVAL]], i32 5
+; MAX256-NEXT:    [[TMP81:%.*]] = insertelement <8 x float> [[TMP80]], float [[FVAL]], i32 6
+; MAX256-NEXT:    [[TMP82:%.*]] = extractelement <8 x float> [[TMP14]], i32 7
+; MAX256-NEXT:    [[TMP83:%.*]] = insertelement <8 x float> [[TMP81]], float [[TMP82]], i32 7
+; MAX256-NEXT:    [[TMP84:%.*]] = extractelement <8 x float> [[TMP28]], i32 0
+; MAX256-NEXT:    [[TMP85:%.*]] = insertelement <8 x float> undef, float [[TMP84]], i32 0
+; MAX256-NEXT:    [[TMP86:%.*]] = insertelement <8 x float> [[TMP85]], float [[FVAL]], i32 1
+; MAX256-NEXT:    [[TMP87:%.*]] = insertelement <8 x float> [[TMP86]], float [[FVAL]], i32 2
+; MAX256-NEXT:    [[TMP88:%.*]] = insertelement <8 x float> [[TMP87]], float [[TMP47]], i32 3
+; MAX256-NEXT:    [[TMP89:%.*]] = extractelement <8 x float> [[TMP28]], i32 4
+; MAX256-NEXT:    [[TMP90:%.*]] = insertelement <8 x float> [[TMP88]], float [[TMP89]], i32 4
+; MAX256-NEXT:    [[TMP91:%.*]] = insertelement <8 x float> [[TMP90]], float [[FVAL]], i32 5
+; MAX256-NEXT:    [[TMP92:%.*]] = insertelement <8 x float> [[TMP91]], float [[FVAL]], i32 6
+; MAX256-NEXT:    [[TMP93:%.*]] = insertelement <8 x float> [[TMP92]], float [[TMP53]], i32 7
+; MAX256-NEXT:    [[TMP94:%.*]] = extractelement <8 x float> [[TMP30]], i32 0
+; MAX256-NEXT:    [[TMP95:%.*]] = insertelement <8 x float> undef, float [[TMP94]], i32 0
+; MAX256-NEXT:    [[TMP96:%.*]] = insertelement <8 x float> [[TMP95]], float [[FVAL]], i32 1
+; MAX256-NEXT:    [[TMP97:%.*]] = insertelement <8 x float> [[TMP96]], float [[FVAL]], i32 2
+; MAX256-NEXT:    [[TMP98:%.*]] = insertelement <8 x float> [[TMP97]], float [[TMP57]], i32 3
+; MAX256-NEXT:    [[TMP99:%.*]] = extractelement <8 x float> [[TMP30]], i32 4
+; MAX256-NEXT:    [[TMP100:%.*]] = insertelement <8 x float> [[TMP98]], float [[TMP99]], i32 4
+; MAX256-NEXT:    [[TMP101:%.*]] = insertelement <8 x float> [[TMP100]], float [[FVAL]], i32 5
+; MAX256-NEXT:    [[TMP102:%.*]] = insertelement <8 x float> [[TMP101]], float [[FVAL]], i32 6
+; MAX256-NEXT:    [[TMP103:%.*]] = insertelement <8 x float> [[TMP102]], float [[TMP63]], i32 7
+; MAX256-NEXT:    [[TMP104:%.*]] = extractelement <8 x float> [[TMP32]], i32 0
+; MAX256-NEXT:    [[TMP105:%.*]] = insertelement <8 x float> undef, float [[TMP104]], i32 0
+; MAX256-NEXT:    [[TMP106:%.*]] = insertelement <8 x float> [[TMP105]], float [[FVAL]], i32 1
+; MAX256-NEXT:    [[TMP107:%.*]] = insertelement <8 x float> [[TMP106]], float [[FVAL]], i32 2
+; MAX256-NEXT:    [[TMP108:%.*]] = insertelement <8 x float> [[TMP107]], float [[TMP67]], i32 3
+; MAX256-NEXT:    [[TMP109:%.*]] = extractelement <8 x float> [[TMP32]], i32 4
+; MAX256-NEXT:    [[TMP110:%.*]] = insertelement <8 x float> [[TMP108]], float [[TMP109]], i32 4
+; MAX256-NEXT:    [[TMP111:%.*]] = insertelement <8 x float> [[TMP110]], float [[FVAL]], i32 5
+; MAX256-NEXT:    [[TMP112:%.*]] = insertelement <8 x float> [[TMP111]], float [[FVAL]], i32 6
+; MAX256-NEXT:    [[TMP113:%.*]] = insertelement <8 x float> [[TMP112]], float [[TMP73]], i32 7
 ; MAX256-NEXT:    br label [[BB2]]
 ; MAX256:       bb5:
-; MAX256-NEXT:    [[TMP126:%.*]] = insertelement <32 x float> [[TMP5]], float [[TMP41]], i32 1
-; MAX256-NEXT:    [[TMP127:%.*]] = insertelement <32 x float> [[TMP126]], float [[FVAL]], i32 2
-; MAX256-NEXT:    [[TMP128:%.*]] = extractelement <32 x float> [[TMP38]], i32 3
-; MAX256-NEXT:    [[TMP129:%.*]] = insertelement <32 x float> [[TMP127]], float [[TMP128]], i32 3
-; MAX256-NEXT:    [[TMP130:%.*]] = insertelement <32 x float> [[TMP129]], float [[FVAL]], i32 4
-; MAX256-NEXT:    [[TMP131:%.*]] = insertelement <32 x float> [[TMP130]], float [[TMP47]], i32 5
-; MAX256-NEXT:    [[TMP132:%.*]] = insertelement <32 x float> [[TMP131]], float [[FVAL]], i32 6
-; MAX256-NEXT:    [[TMP133:%.*]] = extractelement <32 x float> [[TMP38]], i32 7
-; MAX256-NEXT:    [[TMP134:%.*]] = insertelement <32 x float> [[TMP132]], float [[TMP133]], i32 7
-; MAX256-NEXT:    [[TMP135:%.*]] = extractelement <32 x float> [[TMP38]], i32 8
-; MAX256-NEXT:    [[TMP136:%.*]] = insertelement <32 x float> [[TMP134]], float [[TMP135]], i32 8
-; MAX256-NEXT:    [[TMP137:%.*]] = insertelement <32 x float> [[TMP136]], float [[FVAL]], i32 9
-; MAX256-NEXT:    [[TMP138:%.*]] = insertelement <32 x float> [[TMP137]], float [[TMP53]], i32 10
-; MAX256-NEXT:    [[TMP139:%.*]] = insertelement <32 x float> [[TMP138]], float [[FVAL]], i32 11
-; MAX256-NEXT:    [[TMP140:%.*]] = extractelement <32 x float> [[TMP38]], i32 12
-; MAX256-NEXT:    [[TMP141:%.*]] = insertelement <32 x float> [[TMP139]], float [[TMP140]], i32 12
-; MAX256-NEXT:    [[TMP142:%.*]] = insertelement <32 x float> [[TMP141]], float [[FVAL]], i32 13
-; MAX256-NEXT:    [[TMP143:%.*]] = insertelement <32 x float> [[TMP142]], float [[TMP59]], i32 14
-; MAX256-NEXT:    [[TMP144:%.*]] = insertelement <32 x float> [[TMP143]], float [[FVAL]], i32 15
-; MAX256-NEXT:    [[TMP145:%.*]] = extractelement <32 x float> [[TMP38]], i32 16
-; MAX256-NEXT:    [[TMP146:%.*]] = insertelement <32 x float> [[TMP144]], float [[TMP145]], i32 16
-; MAX256-NEXT:    [[TMP147:%.*]] = insertelement <32 x float> [[TMP146]], float [[FVAL]], i32 17
-; MAX256-NEXT:    [[TMP148:%.*]] = insertelement <32 x float> [[TMP147]], float [[TMP65]], i32 18
-; MAX256-NEXT:    [[TMP149:%.*]] = insertelement <32 x float> [[TMP148]], float [[FVAL]], i32 19
-; MAX256-NEXT:    [[TMP150:%.*]] = extractelement <32 x float> [[TMP38]], i32 20
-; MAX256-NEXT:    [[TMP151:%.*]] = insertelement <32 x float> [[TMP149]], float [[TMP150]], i32 20
-; MAX256-NEXT:    [[TMP152:%.*]] = insertelement <32 x float> [[TMP151]], float [[FVAL]], i32 21
-; MAX256-NEXT:    [[TMP153:%.*]] = insertelement <32 x float> [[TMP152]], float [[TMP71]], i32 22
-; MAX256-NEXT:    [[TMP154:%.*]] = insertelement <32 x float> [[TMP153]], float [[FVAL]], i32 23
-; MAX256-NEXT:    [[TMP155:%.*]] = extractelement <32 x float> [[TMP38]], i32 24
-; MAX256-NEXT:    [[TMP156:%.*]] = insertelement <32 x float> [[TMP154]], float [[TMP155]], i32 24
-; MAX256-NEXT:    [[TMP157:%.*]] = insertelement <32 x float> [[TMP156]], float [[FVAL]], i32 25
-; MAX256-NEXT:    [[TMP158:%.*]] = insertelement <32 x float> [[TMP157]], float [[TMP77]], i32 26
-; MAX256-NEXT:    [[TMP159:%.*]] = insertelement <32 x float> [[TMP158]], float [[FVAL]], i32 27
-; MAX256-NEXT:    [[TMP160:%.*]] = extractelement <32 x float> [[TMP38]], i32 28
-; MAX256-NEXT:    [[TMP161:%.*]] = insertelement <32 x float> [[TMP159]], float [[TMP160]], i32 28
-; MAX256-NEXT:    [[TMP162:%.*]] = insertelement <32 x float> [[TMP161]], float [[FVAL]], i32 29
-; MAX256-NEXT:    [[TMP163:%.*]] = insertelement <32 x float> [[TMP162]], float [[TMP83]], i32 30
-; MAX256-NEXT:    [[TMP164:%.*]] = insertelement <32 x float> [[TMP163]], float [[FVAL]], i32 31
+; MAX256-NEXT:    [[TMP114:%.*]] = insertelement <8 x float> [[TMP5]], float [[TMP35]], i32 1
+; MAX256-NEXT:    [[TMP115:%.*]] = insertelement <8 x float> [[TMP114]], float [[FVAL]], i32 2
+; MAX256-NEXT:    [[TMP116:%.*]] = extractelement <8 x float> [[TMP14]], i32 3
+; MAX256-NEXT:    [[TMP117:%.*]] = insertelement <8 x float> [[TMP115]], float [[TMP116]], i32 3
+; MAX256-NEXT:    [[TMP118:%.*]] = insertelement <8 x float> [[TMP117]], float [[FVAL]], i32 4
+; MAX256-NEXT:    [[TMP119:%.*]] = insertelement <8 x float> [[TMP118]], float [[TMP41]], i32 5
+; MAX256-NEXT:    [[TMP120:%.*]] = insertelement <8 x float> [[TMP119]], float [[FVAL]], i32 6
+; MAX256-NEXT:    [[TMP121:%.*]] = extractelement <8 x float> [[TMP14]], i32 7
+; MAX256-NEXT:    [[TMP122:%.*]] = insertelement <8 x float> [[TMP120]], float [[TMP121]], i32 7
+; MAX256-NEXT:    [[TMP123:%.*]] = extractelement <8 x float> [[TMP28]], i32 0
+; MAX256-NEXT:    [[TMP124:%.*]] = insertelement <8 x float> undef, float [[TMP123]], i32 0
+; MAX256-NEXT:    [[TMP125:%.*]] = insertelement <8 x float> [[TMP124]], float [[FVAL]], i32 1
+; MAX256-NEXT:    [[TMP126:%.*]] = insertelement <8 x float> [[TMP125]], float [[TMP45]], i32 2
+; MAX256-NEXT:    [[TMP127:%.*]] = insertelement <8 x float> [[TMP126]], float [[FVAL]], i32 3
+; MAX256-NEXT:    [[TMP128:%.*]] = extractelement <8 x float> [[TMP28]], i32 4
+; MAX256-NEXT:    [[TMP129:%.*]] = insertelement <8 x float> [[TMP127]], float [[TMP128]], i32 4
+; MAX256-NEXT:    [[TMP130:%.*]] = insertelement <8 x float> [[TMP129]], float [[FVAL]], i32 5
+; MAX256-NEXT:    [[TMP131:%.*]] = insertelement <8 x float> [[TMP130]], float [[TMP51]], i32 6
+; MAX256-NEXT:    [[TMP132:%.*]] = insertelement <8 x float> [[TMP131]], float [[FVAL]], i32 7
+; MAX256-NEXT:    [[TMP133:%.*]] = extractelement <8 x float> [[TMP30]], i32 0
+; MAX256-NEXT:    [[TMP134:%.*]] = insertelement <8 x float> undef, float [[TMP133]], i32 0
+; MAX256-NEXT:    [[TMP135:%.*]] = insertelement <8 x float> [[TMP134]], float [[FVAL]], i32 1
+; MAX256-NEXT:    [[TMP136:%.*]] = insertelement <8 x float> [[TMP135]], float [[TMP55]], i32 2
+; MAX256-NEXT:    [[TMP137:%.*]] = insertelement <8 x float> [[TMP136]], float [[FVAL]], i32 3
+; MAX256-NEXT:    [[TMP138:%.*]] = extractelement <8 x float> [[TMP30]], i32 4
+; MAX256-NEXT:    [[TMP139:%.*]] = insertelement <8 x float> [[TMP137]], float [[TMP138]], i32 4
+; MAX256-NEXT:    [[TMP140:%.*]] = insertelement <8 x float> [[TMP139]], float [[FVAL]], i32 5
+; MAX256-NEXT:    [[TMP141:%.*]] = insertelement <8 x float> [[TMP140]], float [[TMP61]], i32 6
+; MAX256-NEXT:    [[TMP142:%.*]] = insertelement <8 x float> [[TMP141]], float [[FVAL]], i32 7
+; MAX256-NEXT:    [[TMP143:%.*]] = extractelement <8 x float> [[TMP32]], i32 0
+; MAX256-NEXT:    [[TMP144:%.*]] = insertelement <8 x float> undef, float [[TMP143]], i32 0
+; MAX256-NEXT:    [[TMP145:%.*]] = insertelement <8 x float> [[TMP144]], float [[FVAL]], i32 1
+; MAX256-NEXT:    [[TMP146:%.*]] = insertelement <8 x float> [[TMP145]], float [[TMP65]], i32 2
+; MAX256-NEXT:    [[TMP147:%.*]] = insertelement <8 x float> [[TMP146]], float [[FVAL]], i32 3
+; MAX256-NEXT:    [[TMP148:%.*]] = extractelement <8 x float> [[TMP32]], i32 4
+; MAX256-NEXT:    [[TMP149:%.*]] = insertelement <8 x float> [[TMP147]], float [[TMP148]], i32 4
+; MAX256-NEXT:    [[TMP150:%.*]] = insertelement <8 x float> [[TMP149]], float [[FVAL]], i32 5
+; MAX256-NEXT:    [[TMP151:%.*]] = insertelement <8 x float> [[TMP150]], float [[TMP71]], i32 6
+; MAX256-NEXT:    [[TMP152:%.*]] = insertelement <8 x float> [[TMP151]], float [[FVAL]], i32 7
 ; MAX256-NEXT:    br label [[BB2]]
 ; MAX256:       bb2:
-; MAX256-NEXT:    [[TMP165:%.*]] = phi <32 x float> [ [[TMP38]], [[BB3]] ], [ [[TMP125]], [[BB4]] ], [ [[TMP164]], [[BB5]] ], [ [[TMP86]], [[BB1]] ]
+; MAX256-NEXT:    [[TMP153:%.*]] = phi <8 x float> [ [[TMP14]], [[BB3]] ], [ [[TMP83]], [[BB4]] ], [ [[TMP122]], [[BB5]] ], [ [[TMP44]], [[BB1]] ]
+; MAX256-NEXT:    [[TMP154:%.*]] = phi <8 x float> [ [[TMP28]], [[BB3]] ], [ [[TMP93]], [[BB4]] ], [ [[TMP132]], [[BB5]] ], [ [[TMP54]], [[BB1]] ]
+; MAX256-NEXT:    [[TMP155:%.*]] = phi <8 x float> [ [[TMP30]], [[BB3]] ], [ [[TMP103]], [[BB4]] ], [ [[TMP142]], [[BB5]] ], [ [[TMP64]], [[BB1]] ]
+; MAX256-NEXT:    [[TMP156:%.*]] = phi <8 x float> [ [[TMP32]], [[BB3]] ], [ [[TMP113]], [[BB4]] ], [ [[TMP152]], [[BB5]] ], [ [[TMP74]], [[BB1]] ]
 ; MAX256-NEXT:    ret void
 ;
 ; MAX1024-LABEL: @phi_float32(


        


More information about the llvm-commits mailing list