[ARM] Don't pretend we know how to generate MVE VLDn

[dmgreen]

Summary

We don't yet know how to generate these instructions for MVE. And in the case of VLD3, we don't have the instruction.

[dmgreen]

I need a test for this..

[dmgreen]

Now with test.

[ostannard]

LGTM with a minor style nit.

llvm/lib/Target/ARM/ARMISelLowering.cpp

16161

Opening brace should be on the line above.

llvm/lib/Target/ARM/ARMISelLowering.h

public:
 
     void emitAtomicCmpXchgNoStoreLLBalance(IRBuilder<> &Builder) const override;
 
     Instruction *emitLeadingFence(IRBuilder<> &Builder, Instruction *Inst,
                                   AtomicOrdering Ord) const override;
     Instruction *emitTrailingFence(IRBuilder<> &Builder, Instruction *Inst,
                                    AtomicOrdering Ord) const override;
 
-    unsigned getMaxSupportedInterleaveFactor() const override { return 4; }
+    unsigned getMaxSupportedInterleaveFactor() const override;
 
     bool lowerInterleavedLoad(LoadInst *LI,
                               ArrayRef<ShuffleVectorInst *> Shuffles,
                               ArrayRef<unsigned> Indices,
                               unsigned Factor) const override;
     bool lowerInterleavedStore(StoreInst *SI, ShuffleVectorInst *SVI,
                                unsigned Factor) const override;
 

llvm/lib/Target/ARM/ARMISelLowering.cpp

bool ARMTargetLowering::isLegalInterleavedAccessType(
   if (ElSize != 8 && ElSize != 16 && ElSize != 32)
     return false;
 
   // Ensure the total vector size is 64 or a multiple of 128. Types larger than
   // 128 will be split into multiple interleaved accesses.
   return VecSize == 64 || VecSize % 128 == 0;
 }
 
+unsigned ARMTargetLowering::getMaxSupportedInterleaveFactor() const {
+  if (Subtarget->hasNEON())
+    return 4;
+  return TargetLoweringBase::getMaxSupportedInterleaveFactor();
+}
+
 /// Lower an interleaved load into a vldN intrinsic.
 ///
 /// E.g. Lower an interleaved load (Factor = 2):
 ///        %wide.vec = load <8 x i32>, <8 x i32>* %ptr, align 4
 ///        %v0 = shuffle %wide.vec, undef, <0, 2, 4, 6>  ; Extract even elements
 ///        %v1 = shuffle %wide.vec, undef, <1, 3, 5, 7>  ; Extract odd elements
 ///
 ///      Into:

llvm/test/Transforms/LoopVectorize/ARM/mve-interleaved-cost.ll

+; RUN: opt -loop-vectorize -force-vector-width=2 -debug-only=loop-vectorize -disable-output < %s 2>&1 | FileCheck %s --check-prefix=VF_2
+; RUN: opt -loop-vectorize -force-vector-width=4 -debug-only=loop-vectorize -disable-output < %s 2>&1 | FileCheck %s --check-prefix=VF_4
+; RUN: opt -loop-vectorize -force-vector-width=8 -debug-only=loop-vectorize -disable-output < %s 2>&1 | FileCheck %s --check-prefix=VF_8
+; RUN: opt -loop-vectorize -force-vector-width=16 -debug-only=loop-vectorize -disable-output < %s 2>&1 | FileCheck %s --check-prefix=VF_16
+; REQUIRES: asserts
+
+target datalayout = "e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64"
+target triple = "thumbv8.1-m.main-none-eabi"
+
+%i8.2 = type {i8, i8}
+define void @i8_factor_2(%i8.2* %data, i64 %n) #0 {
+entry:
+  br label %for.body
+
+; VF_8-LABEL:  Checking a loop in "i8_factor_2"
+; VF_8:          Found an estimated cost of 2 for VF 8 For instruction: %tmp2 = load i8, i8* %tmp0, align 1
+; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp3 = load i8, i8* %tmp1, align 1
+; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store i8 0, i8* %tmp0, align 1
+; VF_8-NEXT:     Found an estimated cost of 2 for VF 8 For instruction: store i8 0, i8* %tmp1, align 1
+; VF_16-LABEL: Checking a loop in "i8_factor_2"
+; VF_16:         Found an estimated cost of 2 for VF 16 For instruction: %tmp2 = load i8, i8* %tmp0, align 1
+; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp3 = load i8, i8* %tmp1, align 1
+; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store i8 0, i8* %tmp0, align 1
+; VF_16-NEXT:    Found an estimated cost of 2 for VF 16 For instruction: store i8 0, i8* %tmp1, align 1
+for.body:
+  %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]
+  %tmp0 = getelementptr inbounds %i8.2, %i8.2* %data, i64 %i, i32 0
+  %tmp1 = getelementptr inbounds %i8.2, %i8.2* %data, i64 %i, i32 1
+  %tmp2 = load i8, i8* %tmp0, align 1
+  %tmp3 = load i8, i8* %tmp1, align 1
+  store i8 0, i8* %tmp0, align 1
+  store i8 0, i8* %tmp1, align 1
+  %i.next = add nuw nsw i64 %i, 1
+  %cond = icmp slt i64 %i.next, %n
+  br i1 %cond, label %for.body, label %for.end
+
+for.end:
+  ret void
+}
+
+%i16.2 = type {i16, i16}
+define void @i16_factor_2(%i16.2* %data, i64 %n) #0 {
+entry:
+  br label %for.body
+
+; VF_4-LABEL:  Checking a loop in "i16_factor_2"
+; VF_4:          Found an estimated cost of 2 for VF 4 For instruction: %tmp2 = load i16, i16* %tmp0, align 2
+; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: %tmp3 = load i16, i16* %tmp1, align 2
+; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: store i16 0, i16* %tmp0, align 2
+; VF_4-NEXT:     Found an estimated cost of 2 for VF 4 For instruction: store i16 0, i16* %tmp1, align 2
+; VF_8-LABEL:  Checking a loop in "i16_factor_2"
+; VF_8:          Found an estimated cost of 2 for VF 8 For instruction: %tmp2 = load i16, i16* %tmp0, align 2
+; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp3 = load i16, i16* %tmp1, align 2
+; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store i16 0, i16* %tmp0, align 2
+; VF_8-NEXT:     Found an estimated cost of 2 for VF 8 For instruction: store i16 0, i16* %tmp1, align 2
+; VF_16-LABEL: Checking a loop in "i16_factor_2"
+; VF_16:         Found an estimated cost of 4 for VF 16 For instruction: %tmp2 = load i16, i16* %tmp0, align 2
+; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp3 = load i16, i16* %tmp1, align 2
+; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store i16 0, i16* %tmp0, align 2
+; VF_16-NEXT:    Found an estimated cost of 4 for VF 16 For instruction: store i16 0, i16* %tmp1, align 2
+for.body:
+  %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]
+  %tmp0 = getelementptr inbounds %i16.2, %i16.2* %data, i64 %i, i32 0
+  %tmp1 = getelementptr inbounds %i16.2, %i16.2* %data, i64 %i, i32 1
+  %tmp2 = load i16, i16* %tmp0, align 2
+  %tmp3 = load i16, i16* %tmp1, align 2
+  store i16 0, i16* %tmp0, align 2
+  store i16 0, i16* %tmp1, align 2
+  %i.next = add nuw nsw i64 %i, 1
+  %cond = icmp slt i64 %i.next, %n
+  br i1 %cond, label %for.body, label %for.end
+
+for.end:
+  ret void
+}
+
+%i32.2 = type {i32, i32}
+define void @i32_factor_2(%i32.2* %data, i64 %n) #0 {
+entry:
+  br label %for.body
+
+; VF_2-LABEL:  Checking a loop in "i32_factor_2"
+; VF_2:          Found an estimated cost of 2 for VF 2 For instruction: %tmp2 = load i32, i32* %tmp0, align 4
+; VF_2-NEXT:     Found an estimated cost of 0 for VF 2 For instruction: %tmp3 = load i32, i32* %tmp1, align 4
+; VF_2-NEXT:     Found an estimated cost of 0 for VF 2 For instruction: store i32 0, i32* %tmp0, align 4
+; VF_2-NEXT:     Found an estimated cost of 2 for VF 2 For instruction: store i32 0, i32* %tmp1, align 4
+; VF_4-LABEL:  Checking a loop in "i32_factor_2"
+; VF_4:          Found an estimated cost of 2 for VF 4 For instruction: %tmp2 = load i32, i32* %tmp0, align 4
+; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: %tmp3 = load i32, i32* %tmp1, align 4
+; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: store i32 0, i32* %tmp0, align 4
+; VF_4-NEXT:     Found an estimated cost of 2 for VF 4 For instruction: store i32 0, i32* %tmp1, align 4
+; VF_8-LABEL:  Checking a loop in "i32_factor_2"
+; VF_8:          Found an estimated cost of 4 for VF 8 For instruction: %tmp2 = load i32, i32* %tmp0, align 4
+; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp3 = load i32, i32* %tmp1, align 4
+; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store i32 0, i32* %tmp0, align 4
+; VF_8-NEXT:     Found an estimated cost of 4 for VF 8 For instruction: store i32 0, i32* %tmp1, align 4
+; VF_16-LABEL: Checking a loop in "i32_factor_2"
+; VF_16:         Found an estimated cost of 8 for VF 16 For instruction: %tmp2 = load i32, i32* %tmp0, align 4
+; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp3 = load i32, i32* %tmp1, align 4
+; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store i32 0, i32* %tmp0, align 4
+; VF_16-NEXT:    Found an estimated cost of 8 for VF 16 For instruction: store i32 0, i32* %tmp1, align 4
+for.body:
+  %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]
+  %tmp0 = getelementptr inbounds %i32.2, %i32.2* %data, i64 %i, i32 0
+  %tmp1 = getelementptr inbounds %i32.2, %i32.2* %data, i64 %i, i32 1
+  %tmp2 = load i32, i32* %tmp0, align 4
+  %tmp3 = load i32, i32* %tmp1, align 4
+  store i32 0, i32* %tmp0, align 4
+  store i32 0, i32* %tmp1, align 4
+  %i.next = add nuw nsw i64 %i, 1
+  %cond = icmp slt i64 %i.next, %n
+  br i1 %cond, label %for.body, label %for.end
+
+for.end:
+  ret void
+}
+
+%i8.3 = type {i8, i8, i8}
+define void @i8_factor_3(%i8.3* %data, i64 %n) #0 {
+entry:
+  br label %for.body
+
+; VF_8-LABEL:  Checking a loop in "i8_factor_3"
+; VF_8:          Found an estimated cost of 408 for VF 8 For instruction: %tmp3 = load i8, i8* %tmp0, align 1
+; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp4 = load i8, i8* %tmp1, align 1
+; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp5 = load i8, i8* %tmp2, align 1
+; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store i8 0, i8* %tmp0, align 1
+; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store i8 0, i8* %tmp1, align 1
+; VF_8-NEXT:     Found an estimated cost of 216 for VF 8 For instruction: store i8 0, i8* %tmp2, align 1
+; VF_16-LABEL: Checking a loop in "i8_factor_3"
+; VF_16:         Found an estimated cost of 1584 for VF 16 For instruction: %tmp3 = load i8, i8* %tmp0, align 1
+; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp4 = load i8, i8* %tmp1, align 1
+; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp5 = load i8, i8* %tmp2, align 1
+; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store i8 0, i8* %tmp0, align 1
+; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store i8 0, i8* %tmp1, align 1
+; VF_16-NEXT:    Found an estimated cost of 816 for VF 16 For instruction: store i8 0, i8* %tmp2, align 1
+for.body:
+  %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]
+  %tmp0 = getelementptr inbounds %i8.3, %i8.3* %data, i64 %i, i32 0
+  %tmp1 = getelementptr inbounds %i8.3, %i8.3* %data, i64 %i, i32 1
+  %tmp2 = getelementptr inbounds %i8.3, %i8.3* %data, i64 %i, i32 2
+  %tmp3 = load i8, i8* %tmp0, align 1
+  %tmp4 = load i8, i8* %tmp1, align 1
+  %tmp5 = load i8, i8* %tmp2, align 1
+  store i8 0, i8* %tmp0, align 1
+  store i8 0, i8* %tmp1, align 1
+  store i8 0, i8* %tmp2, align 1
+  %i.next = add nuw nsw i64 %i, 1
+  %cond = icmp slt i64 %i.next, %n
+  br i1 %cond, label %for.body, label %for.end
+
+for.end:
+  ret void
+}
+
+%i16.3 = type {i16, i16, i16}
+define void @i16_factor_3(%i16.3* %data, i64 %n) #0 {
+entry:
+  br label %for.body
+
+; VF_4-LABEL:  Checking a loop in "i16_factor_3"
+; VF_4:          Found an estimated cost of 108 for VF 4 For instruction: %tmp3 = load i16, i16* %tmp0, align 2
+; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: %tmp4 = load i16, i16* %tmp1, align 2
+; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: %tmp5 = load i16, i16* %tmp2, align 2
+; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: store i16 0, i16* %tmp0, align 2
+; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: store i16 0, i16* %tmp1, align 2
+; VF_4-NEXT:     Found an estimated cost of 60 for VF 4 For instruction: store i16 0, i16* %tmp2, align 2
+; VF_8-LABEL:  Checking a loop in "i16_factor_3"
+; VF_8:          Found an estimated cost of 408 for VF 8 For instruction: %tmp3 = load i16, i16* %tmp0, align 2
+; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp4 = load i16, i16* %tmp1, align 2
+; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp5 = load i16, i16* %tmp2, align 2
+; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store i16 0, i16* %tmp0, align 2
+; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store i16 0, i16* %tmp1, align 2
+; VF_8-NEXT:     Found an estimated cost of 216 for VF 8 For instruction: store i16 0, i16* %tmp2, align 2
+; VF_16-LABEL: Checking a loop in "i16_factor_3"
+; VF_16:         Found an estimated cost of 1584 for VF 16 For instruction: %tmp3 = load i16, i16* %tmp0, align 2
+; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp4 = load i16, i16* %tmp1, align 2
+; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp5 = load i16, i16* %tmp2, align 2
+; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store i16 0, i16* %tmp0, align 2
+; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store i16 0, i16* %tmp1, align 2
+; VF_16-NEXT:    Found an estimated cost of 816 for VF 16 For instruction: store i16 0, i16* %tmp2, align 2
+for.body:
+  %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]
+  %tmp0 = getelementptr inbounds %i16.3, %i16.3* %data, i64 %i, i32 0
+  %tmp1 = getelementptr inbounds %i16.3, %i16.3* %data, i64 %i, i32 1
+  %tmp2 = getelementptr inbounds %i16.3, %i16.3* %data, i64 %i, i32 2
+  %tmp3 = load i16, i16* %tmp0, align 2
+  %tmp4 = load i16, i16* %tmp1, align 2
+  %tmp5 = load i16, i16* %tmp2, align 2
+  store i16 0, i16* %tmp0, align 2
+  store i16 0, i16* %tmp1, align 2
+  store i16 0, i16* %tmp2, align 2
+  %i.next = add nuw nsw i64 %i, 1
+  %cond = icmp slt i64 %i.next, %n
+  br i1 %cond, label %for.body, label %for.end
+
+for.end:
+  ret void
+}
+
+%i32.3 = type {i32, i32, i32}
+define void @i32_factor_3(%i32.3* %data, i64 %n) #0 {
+entry:
+  br label %for.body
+
+; VF_2-LABEL:  Checking a loop in "i32_factor_3"
+; VF_2:          Found an estimated cost of 30 for VF 2 For instruction: %tmp3 = load i32, i32* %tmp0, align 4
+; VF_2-NEXT:     Found an estimated cost of 0 for VF 2 For instruction: %tmp4 = load i32, i32* %tmp1, align 4
+; VF_2-NEXT:     Found an estimated cost of 0 for VF 2 For instruction: %tmp5 = load i32, i32* %tmp2, align 4
+; VF_2-NEXT:     Found an estimated cost of 0 for VF 2 For instruction: store i32 0, i32* %tmp0, align 4
+; VF_2-NEXT:     Found an estimated cost of 0 for VF 2 For instruction: store i32 0, i32* %tmp1, align 4
+; VF_2-NEXT:     Found an estimated cost of 18 for VF 2 For instruction: store i32 0, i32* %tmp2, align 4
+; VF_4-LABEL:  Checking a loop in "i32_factor_3"
+; VF_4:          Found an estimated cost of 108 for VF 4 For instruction: %tmp3 = load i32, i32* %tmp0, align 4
+; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: %tmp4 = load i32, i32* %tmp1, align 4
+; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: %tmp5 = load i32, i32* %tmp2, align 4
+; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: store i32 0, i32* %tmp0, align 4
+; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: store i32 0, i32* %tmp1, align 4
+; VF_4-NEXT:     Found an estimated cost of 60 for VF 4 For instruction: store i32 0, i32* %tmp2, align 4
+; VF_8-LABEL:  Checking a loop in "i32_factor_3"
+; VF_8:          Found an estimated cost of 408 for VF 8 For instruction: %tmp3 = load i32, i32* %tmp0, align 4
+; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp4 = load i32, i32* %tmp1, align 4
+; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp5 = load i32, i32* %tmp2, align 4
+; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store i32 0, i32* %tmp0, align 4
+; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store i32 0, i32* %tmp1, align 4
+; VF_8-NEXT:     Found an estimated cost of 216 for VF 8 For instruction: store i32 0, i32* %tmp2, align 4
+; VF_16-LABEL: Checking a loop in "i32_factor_3"
+; VF_16:         Found an estimated cost of 1584 for VF 16 For instruction: %tmp3 = load i32, i32* %tmp0, align 4
+; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp4 = load i32, i32* %tmp1, align 4
+; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp5 = load i32, i32* %tmp2, align 4
+; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store i32 0, i32* %tmp0, align 4
+; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store i32 0, i32* %tmp1, align 4
+; VF_16-NEXT:    Found an estimated cost of 816 for VF 16 For instruction: store i32 0, i32* %tmp2, align 4
+for.body:
+  %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]
+  %tmp0 = getelementptr inbounds %i32.3, %i32.3* %data, i64 %i, i32 0
+  %tmp1 = getelementptr inbounds %i32.3, %i32.3* %data, i64 %i, i32 1
+  %tmp2 = getelementptr inbounds %i32.3, %i32.3* %data, i64 %i, i32 2
+  %tmp3 = load i32, i32* %tmp0, align 4
+  %tmp4 = load i32, i32* %tmp1, align 4
+  %tmp5 = load i32, i32* %tmp2, align 4
+  store i32 0, i32* %tmp0, align 4
+  store i32 0, i32* %tmp1, align 4
+  store i32 0, i32* %tmp2, align 4
+  %i.next = add nuw nsw i64 %i, 1
+  %cond = icmp slt i64 %i.next, %n
+  br i1 %cond, label %for.body, label %for.end
+
+for.end:
+  ret void
+}
+
+
+%i8.4 = type {i8, i8, i8, i8}
+define void @i8_factor_4(%i8.4* %data, i64 %n) #0 {
+entry:
+  br label %for.body
+
+; VF_8-LABEL:  Checking a loop in "i8_factor_4"
+; VF_8:          Found an estimated cost of 544 for VF 8 For instruction: %tmp4 = load i8, i8* %tmp0, align 1
+; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp5 = load i8, i8* %tmp1, align 1
+; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp6 = load i8, i8* %tmp2, align 1
+; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp7 = load i8, i8* %tmp3, align 1
+; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store i8 0, i8* %tmp0, align 1
+; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store i8 0, i8* %tmp1, align 1
+; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store i8 0, i8* %tmp2, align 1
+; VF_8-NEXT:     Found an estimated cost of 288 for VF 8 For instruction: store i8 0, i8* %tmp3, align 1
+; VF_16-LABEL: Checking a loop in "i8_factor_4"
+; VF_16:         Found an estimated cost of 2112 for VF 16 For instruction: %tmp4 = load i8, i8* %tmp0, align 1
+; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp5 = load i8, i8* %tmp1, align 1
+; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp6 = load i8, i8* %tmp2, align 1
+; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp7 = load i8, i8* %tmp3, align 1
+; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store i8 0, i8* %tmp0, align 1
+; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store i8 0, i8* %tmp1, align 1
+; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store i8 0, i8* %tmp2, align 1
+; VF_16-NEXT:    Found an estimated cost of 1088 for VF 16 For instruction: store i8 0, i8* %tmp3, align 1
+for.body:
+  %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]
+  %tmp0 = getelementptr inbounds %i8.4, %i8.4* %data, i64 %i, i32 0
+  %tmp1 = getelementptr inbounds %i8.4, %i8.4* %data, i64 %i, i32 1
+  %tmp2 = getelementptr inbounds %i8.4, %i8.4* %data, i64 %i, i32 2
+  %tmp3 = getelementptr inbounds %i8.4, %i8.4* %data, i64 %i, i32 3
+  %tmp4 = load i8, i8* %tmp0, align 1
+  %tmp5 = load i8, i8* %tmp1, align 1
+  %tmp6 = load i8, i8* %tmp2, align 1
+  %tmp7 = load i8, i8* %tmp3, align 1
+  store i8 0, i8* %tmp0, align 1
+  store i8 0, i8* %tmp1, align 1
+  store i8 0, i8* %tmp2, align 1
+  store i8 0, i8* %tmp3, align 1
+  %i.next = add nuw nsw i64 %i, 1
+  %cond = icmp slt i64 %i.next, %n
+  br i1 %cond, label %for.body, label %for.end
+
+for.end:
+  ret void
+}
+
+%i16.4 = type {i16, i16, i16, i16}
+define void @i16_factor_4(%i16.4* %data, i64 %n) #0 {
+entry:
+  br label %for.body
+
+; VF_4-LABEL:  Checking a loop in "i16_factor_4"
+; VF_4:          Found an estimated cost of 144 for VF 4 For instruction: %tmp4 = load i16, i16* %tmp0, align 2
+; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: %tmp5 = load i16, i16* %tmp1, align 2
+; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: %tmp6 = load i16, i16* %tmp2, align 2
+; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: %tmp7 = load i16, i16* %tmp3, align 2
+; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: store i16 0, i16* %tmp0, align 2
+; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: store i16 0, i16* %tmp1, align 2
+; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: store i16 0, i16* %tmp2, align 2
+; VF_4-NEXT:     Found an estimated cost of 80 for VF 4 For instruction: store i16 0, i16* %tmp3, align 2
+; VF_8-LABEL:  Checking a loop in "i16_factor_4"
+; VF_8:          Found an estimated cost of 544 for VF 8 For instruction: %tmp4 = load i16, i16* %tmp0, align 2
+; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp5 = load i16, i16* %tmp1, align 2
+; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp6 = load i16, i16* %tmp2, align 2
+; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp7 = load i16, i16* %tmp3, align 2
+; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store i16 0, i16* %tmp0, align 2
+; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store i16 0, i16* %tmp1, align 2
+; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store i16 0, i16* %tmp2, align 2
+; VF_8-NEXT:     Found an estimated cost of 288 for VF 8 For instruction: store i16 0, i16* %tmp3, align 2
+; VF_16-LABEL: Checking a loop in "i16_factor_4"
+; VF_16:         Found an estimated cost of 2112 for VF 16 For instruction: %tmp4 = load i16, i16* %tmp0, align 2
+; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp5 = load i16, i16* %tmp1, align 2
+; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp6 = load i16, i16* %tmp2, align 2
+; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp7 = load i16, i16* %tmp3, align 2
+; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store i16 0, i16* %tmp0, align 2
+; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store i16 0, i16* %tmp1, align 2
+; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store i16 0, i16* %tmp2, align 2
+; VF_16-NEXT:    Found an estimated cost of 1088 for VF 16 For instruction: store i16 0, i16* %tmp3, align 2
+for.body:
+  %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]
+  %tmp0 = getelementptr inbounds %i16.4, %i16.4* %data, i64 %i, i32 0
+  %tmp1 = getelementptr inbounds %i16.4, %i16.4* %data, i64 %i, i32 1
+  %tmp2 = getelementptr inbounds %i16.4, %i16.4* %data, i64 %i, i32 2
+  %tmp3 = getelementptr inbounds %i16.4, %i16.4* %data, i64 %i, i32 3
+  %tmp4 = load i16, i16* %tmp0, align 2
+  %tmp5 = load i16, i16* %tmp1, align 2
+  %tmp6 = load i16, i16* %tmp2, align 2
+  %tmp7 = load i16, i16* %tmp3, align 2
+  store i16 0, i16* %tmp0, align 2
+  store i16 0, i16* %tmp1, align 2
+  store i16 0, i16* %tmp2, align 2
+  store i16 0, i16* %tmp3, align 2
+  %i.next = add nuw nsw i64 %i, 1
+  %cond = icmp slt i64 %i.next, %n
+  br i1 %cond, label %for.body, label %for.end
+
+for.end:
+  ret void
+}
+
+%i32.4 = type {i32, i32, i32, i32}
+define void @i32_factor_4(%i32.4* %data, i64 %n) #0 {
+entry:
+  br label %for.body
+
+; VF_2-LABEL:  Checking a loop in "i32_factor_4"
+; VF_2:          Found an estimated cost of 40 for VF 2 For instruction: %tmp4 = load i32, i32* %tmp0, align 4
+; VF_2-NEXT:     Found an estimated cost of 0 for VF 2 For instruction: %tmp5 = load i32, i32* %tmp1, align 4
+; VF_2-NEXT:     Found an estimated cost of 0 for VF 2 For instruction: %tmp6 = load i32, i32* %tmp2, align 4
+; VF_2-NEXT:     Found an estimated cost of 0 for VF 2 For instruction: %tmp7 = load i32, i32* %tmp3, align 4
+; VF_2-NEXT:     Found an estimated cost of 0 for VF 2 For instruction: store i32 0, i32* %tmp0, align 4
+; VF_2-NEXT:     Found an estimated cost of 0 for VF 2 For instruction: store i32 0, i32* %tmp1, align 4
+; VF_2-NEXT:     Found an estimated cost of 0 for VF 2 For instruction: store i32 0, i32* %tmp2, align 4
+; VF_2-NEXT:     Found an estimated cost of 24 for VF 2 For instruction: store i32 0, i32* %tmp3, align 4
+; VF_4-LABEL:  Checking a loop in "i32_factor_4"
+; VF_4:          Found an estimated cost of 144 for VF 4 For instruction: %tmp4 = load i32, i32* %tmp0, align 4
+; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: %tmp5 = load i32, i32* %tmp1, align 4
+; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: %tmp6 = load i32, i32* %tmp2, align 4
+; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: %tmp7 = load i32, i32* %tmp3, align 4
+; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: store i32 0, i32* %tmp0, align 4
+; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: store i32 0, i32* %tmp1, align 4
+; VF_4-NEXT:     Found an estimated cost of 0 for VF 4 For instruction: store i32 0, i32* %tmp2, align 4
+; VF_4-NEXT:     Found an estimated cost of 80 for VF 4 For instruction: store i32 0, i32* %tmp3, align 4
+; VF_8-LABEL:  Checking a loop in "i32_factor_4"
+; VF_8:          Found an estimated cost of 544 for VF 8 For instruction: %tmp4 = load i32, i32* %tmp0, align 4
+; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp5 = load i32, i32* %tmp1, align 4
+; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp6 = load i32, i32* %tmp2, align 4
+; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: %tmp7 = load i32, i32* %tmp3, align 4
+; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store i32 0, i32* %tmp0, align 4
+; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store i32 0, i32* %tmp1, align 4
+; VF_8-NEXT:     Found an estimated cost of 0 for VF 8 For instruction: store i32 0, i32* %tmp2, align 4
+; VF_8-NEXT:     Found an estimated cost of 288 for VF 8 For instruction: store i32 0, i32* %tmp3, align 4
+; VF_16-LABEL: Checking a loop in "i32_factor_4"
+; VF_16:         Found an estimated cost of 2112 for VF 16 For instruction: %tmp4 = load i32, i32* %tmp0, align 4
+; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp5 = load i32, i32* %tmp1, align 4
+; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp6 = load i32, i32* %tmp2, align 4
+; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: %tmp7 = load i32, i32* %tmp3, align 4
+; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store i32 0, i32* %tmp0, align 4
+; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store i32 0, i32* %tmp1, align 4
+; VF_16-NEXT:    Found an estimated cost of 0 for VF 16 For instruction: store i32 0, i32* %tmp2, align 4
+; VF_16-NEXT:    Found an estimated cost of 1088 for VF 16 For instruction: store i32 0, i32* %tmp3, align 4
+for.body:
+  %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]
+  %tmp0 = getelementptr inbounds %i32.4, %i32.4* %data, i64 %i, i32 0
+  %tmp1 = getelementptr inbounds %i32.4, %i32.4* %data, i64 %i, i32 1
+  %tmp2 = getelementptr inbounds %i32.4, %i32.4* %data, i64 %i, i32 2
+  %tmp3 = getelementptr inbounds %i32.4, %i32.4* %data, i64 %i, i32 3
+  %tmp4 = load i32, i32* %tmp0, align 4
+  %tmp5 = load i32, i32* %tmp1, align 4
+  %tmp6 = load i32, i32* %tmp2, align 4
+  %tmp7 = load i32, i32* %tmp3, align 4
+  store i32 0, i32* %tmp0, align 4
+  store i32 0, i32* %tmp1, align 4
+  store i32 0, i32* %tmp2, align 4
+  store i32 0, i32* %tmp3, align 4
+  %i.next = add nuw nsw i64 %i, 1
+  %cond = icmp slt i64 %i.next, %n
+  br i1 %cond, label %for.body, label %for.end
+
+for.end:
+  ret void
+}
+
+attributes #0 = { "target-features"="+mve.fp" }