diff --git a/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp b/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp
index 4e509816716698..359fc55035fa21 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp
@@ -984,65 +984,13 @@ Value *InstCombiner::simplifyAMDGCNMemoryIntrinsicDemanded(IntrinsicInst *II,
   if (VWidth == 1)
     return nullptr;
 
-  IRBuilderBase::InsertPointGuard Guard(Builder);
-  Builder.SetInsertPoint(II);
-
-  // Assume the arguments are unchanged and later override them, if needed.
-  SmallVector<Value *, 16> Args(II->arg_begin(), II->arg_end());
+  ConstantInt *NewDMask = nullptr;
 
   if (DMaskIdx < 0) {
-    // Buffer case.
-
-    const unsigned ActiveBits = DemandedElts.getActiveBits();
-    const unsigned UnusedComponentsAtFront = DemandedElts.countTrailingZeros();
-
-    // Start assuming the prefix of elements is demanded, but possibly clear some other bits if
-    // there are trailing zeros (unused components at front) and update offset.
-    DemandedElts = (1 << ActiveBits) - 1;
-
-    if (UnusedComponentsAtFront > 0) {
-      static const unsigned InvalidOffsetIdx = 0xf;
-
-      unsigned OffsetIdx;
-      switch (II->getIntrinsicID()) {
-      case Intrinsic::amdgcn_raw_buffer_load:
-      case Intrinsic::amdgcn_raw_buffer_load_format:
-        OffsetIdx = 1;
-        break;
-      case Intrinsic::amdgcn_s_buffer_load:
-        // If resulting type is vec3, there is no point in trimming the
-        // load with updated offset, as the vec3 would most likely be widened to
-        // vec4 anyway during lowering.
-        if (ActiveBits == 4 && UnusedComponentsAtFront == 1)
-          OffsetIdx = InvalidOffsetIdx;
-        else
-          OffsetIdx = 1;
-        break;
-      case Intrinsic::amdgcn_buffer_load:
-      case Intrinsic::amdgcn_buffer_load_format:
-      case Intrinsic::amdgcn_struct_buffer_load:
-      case Intrinsic::amdgcn_struct_buffer_load_format:
-        OffsetIdx = 2;
-        break;
-      default:
-        // TODO: handle *tbuffer* intrinsics.
-        OffsetIdx = InvalidOffsetIdx;
-        break;
-      }
-
-      if (OffsetIdx != InvalidOffsetIdx) {
-        // Clear demanded bits and update the offset.
-        DemandedElts &= ~((1 << UnusedComponentsAtFront) - 1);
-        auto Offset = II->getArgOperand(OffsetIdx);
-        unsigned SingleComponentSizeInBits = getDataLayout().getTypeSizeInBits(II->getType()->getScalarType());
-        unsigned OffsetAdd = UnusedComponentsAtFront * SingleComponentSizeInBits / 8;
-        auto OffsetAddVal = ConstantInt::get(Offset->getType(), OffsetAdd);
-        Args[OffsetIdx] = Builder.CreateAdd(Offset, OffsetAddVal);
-      }
-    }
+    // Pretend that a prefix of elements is demanded to simplify the code
+    // below.
+    DemandedElts = (1 << DemandedElts.getActiveBits()) - 1;
   } else {
-    // Image case.
-
     ConstantInt *DMask = cast<ConstantInt>(II->getArgOperand(DMaskIdx));
     unsigned DMaskVal = DMask->getZExtValue() & 0xf;
 
@@ -1061,7 +1009,7 @@ Value *InstCombiner::simplifyAMDGCNMemoryIntrinsicDemanded(IntrinsicInst *II,
     }
 
     if (DMaskVal != NewDMaskVal)
-      Args[DMaskIdx] = ConstantInt::get(DMask->getType(), NewDMaskVal);
+      NewDMask = ConstantInt::get(DMask->getType(), NewDMaskVal);
   }
 
   unsigned NewNumElts = DemandedElts.countPopulation();
@@ -1069,8 +1017,8 @@ Value *InstCombiner::simplifyAMDGCNMemoryIntrinsicDemanded(IntrinsicInst *II,
     return UndefValue::get(II->getType());
 
   if (NewNumElts >= VWidth && DemandedElts.isMask()) {
-    if (DMaskIdx >= 0)
-      II->setArgOperand(DMaskIdx, Args[DMaskIdx]);
+    if (NewDMask)
+      II->setArgOperand(DMaskIdx, NewDMask);
     return nullptr;
   }
 
@@ -1093,6 +1041,16 @@ Value *InstCombiner::simplifyAMDGCNMemoryIntrinsicDemanded(IntrinsicInst *II,
   OverloadTys[0] = NewTy;
   Function *NewIntrin = Intrinsic::getDeclaration(M, IID, OverloadTys);
 
+  SmallVector<Value *, 16> Args;
+  for (unsigned I = 0, E = II->getNumArgOperands(); I != E; ++I)
+    Args.push_back(II->getArgOperand(I));
+
+  if (NewDMask)
+    Args[DMaskIdx] = NewDMask;
+
+  IRBuilderBase::InsertPointGuard Guard(Builder);
+  Builder.SetInsertPoint(II);
+
   CallInst *NewCall = Builder.CreateCall(NewIntrin, Args);
   NewCall->takeName(II);
   NewCall->copyMetadata(*II);
@@ -1761,7 +1719,6 @@ Value *InstCombiner::SimplifyDemandedVectorElts(Value *V, APInt DemandedElts,
     case Intrinsic::amdgcn_raw_buffer_load:
     case Intrinsic::amdgcn_raw_buffer_load_format:
     case Intrinsic::amdgcn_raw_tbuffer_load:
-    case Intrinsic::amdgcn_s_buffer_load:
     case Intrinsic::amdgcn_struct_buffer_load:
     case Intrinsic::amdgcn_struct_buffer_load_format:
     case Intrinsic::amdgcn_struct_tbuffer_load:
diff --git a/llvm/test/Transforms/InstCombine/AMDGPU/amdgcn-demanded-vector-elts.ll b/llvm/test/Transforms/InstCombine/AMDGPU/amdgcn-demanded-vector-elts.ll
index 40797ec88ab0cb..389ca6010049c2 100644
--- a/llvm/test/Transforms/InstCombine/AMDGPU/amdgcn-demanded-vector-elts.ll
+++ b/llvm/test/Transforms/InstCombine/AMDGPU/amdgcn-demanded-vector-elts.ll
@@ -46,9 +46,9 @@ define amdgpu_ps float @extract_elt0_buffer_load_v2f32(<4 x i32> inreg %rsrc, i3
 }
 
 ; CHECK-LABEL: @extract_elt1_buffer_load_v2f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 4
-; CHECK-NEXT: %data = call float @llvm.amdgcn.buffer.load.f32(<4 x i32> %rsrc, i32 %idx, i32 %1, i1 false, i1 false)
-; CHECK-NEXT: ret float %data
+; CHECK-NEXT: %data = call <2 x float> @llvm.amdgcn.buffer.load.v2f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i1 false, i1 false)
+; CHECK-NEXT: %elt1 = extractelement <2 x float> %data, i32 1
+; CHECK-NEXT: ret float %elt1
 define amdgpu_ps float @extract_elt1_buffer_load_v2f32(<4 x i32> inreg %rsrc, i32 %idx, i32 %ofs) #0 {
   %data = call <2 x float> @llvm.amdgcn.buffer.load.v2f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i1 false, i1 false)
   %elt1 = extractelement <2 x float> %data, i32 1
@@ -65,9 +65,9 @@ define amdgpu_ps float @extract_elt0_buffer_load_v4f32(<4 x i32> inreg %rsrc, i3
 }
 
 ; CHECK-LABEL: @extract_elt1_buffer_load_v4f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 4
-; CHECK-NEXT: %data = call float @llvm.amdgcn.buffer.load.f32(<4 x i32> %rsrc, i32 %idx, i32 %1, i1 false, i1 false)
-; CHECK-NEXT: ret float %data
+; CHECK-NEXT: %data = call <2 x float> @llvm.amdgcn.buffer.load.v2f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i1 false, i1 false)
+; CHECK-NEXT: %elt1 = extractelement <2 x float> %data, i32 1
+; CHECK-NEXT: ret float %elt1
 define amdgpu_ps float @extract_elt1_buffer_load_v4f32(<4 x i32> inreg %rsrc, i32 %idx, i32 %ofs) #0 {
   %data = call <4 x float> @llvm.amdgcn.buffer.load.v4f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i1 false, i1 false)
   %elt1 = extractelement <4 x float> %data, i32 1
@@ -75,9 +75,9 @@ define amdgpu_ps float @extract_elt1_buffer_load_v4f32(<4 x i32> inreg %rsrc, i3
 }
 
 ; CHECK-LABEL: @extract_elt2_buffer_load_v4f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 8
-; CHECK-NEXT: %data = call float @llvm.amdgcn.buffer.load.f32(<4 x i32> %rsrc, i32 %idx, i32 %1, i1 false, i1 false)
-; CHECK-NEXT: ret float %data
+; CHECK-NEXT: %data = call <3 x float> @llvm.amdgcn.buffer.load.v3f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i1 false, i1 false)
+; CHECK-NEXT: %elt1 = extractelement <3 x float> %data, i32 2
+; CHECK-NEXT: ret float %elt1
 define amdgpu_ps float @extract_elt2_buffer_load_v4f32(<4 x i32> inreg %rsrc, i32 %idx, i32 %ofs) #0 {
   %data = call <4 x float> @llvm.amdgcn.buffer.load.v4f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i1 false, i1 false)
   %elt1 = extractelement <4 x float> %data, i32 2
@@ -85,9 +85,9 @@ define amdgpu_ps float @extract_elt2_buffer_load_v4f32(<4 x i32> inreg %rsrc, i3
 }
 
 ; CHECK-LABEL: @extract_elt3_buffer_load_v4f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 12
-; CHECK-NEXT: %data = call float @llvm.amdgcn.buffer.load.f32(<4 x i32> %rsrc, i32 %idx, i32 %1, i1 false, i1 false)
-; CHECK-NEXT: ret float %data
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.buffer.load.v4f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i1 false, i1 false)
+; CHECK-NEXT: %elt1 = extractelement <4 x float> %data, i32 3
+; CHECK-NEXT: ret float %elt1
 define amdgpu_ps float @extract_elt3_buffer_load_v4f32(<4 x i32> inreg %rsrc, i32 %idx, i32 %ofs) #0 {
   %data = call <4 x float> @llvm.amdgcn.buffer.load.v4f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i1 false, i1 false)
   %elt1 = extractelement <4 x float> %data, i32 3
@@ -104,9 +104,9 @@ define amdgpu_ps <2 x float> @extract_elt0_elt1_buffer_load_v4f32(<4 x i32> inre
 }
 
 ; CHECK-LABEL: @extract_elt1_elt2_buffer_load_v4f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 4
-; CHECK-NEXT: %data = call <2 x float> @llvm.amdgcn.buffer.load.v2f32(<4 x i32> %rsrc, i32 %idx, i32 %1, i1 false, i1 false)
-; CHECK-NEXT: ret <2 x float> %data
+; CHECK-NEXT: %data = call <3 x float> @llvm.amdgcn.buffer.load.v3f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i1 false, i1 false)
+; CHECK-NEXT: %shuf = shufflevector <3 x float> %data, <3 x float> undef, <2 x i32> <i32 1, i32 2>
+; CHECK-NEXT: ret <2 x float> %shuf
 define amdgpu_ps <2 x float> @extract_elt1_elt2_buffer_load_v4f32(<4 x i32> inreg %rsrc, i32 %idx, i32 %ofs) #0 {
   %data = call <4 x float> @llvm.amdgcn.buffer.load.v4f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i1 false, i1 false)
   %shuf = shufflevector <4 x float> %data, <4 x float> undef, <2 x i32> <i32 1, i32 2>
@@ -114,9 +114,9 @@ define amdgpu_ps <2 x float> @extract_elt1_elt2_buffer_load_v4f32(<4 x i32> inre
 }
 
 ; CHECK-LABEL: @extract_elt2_elt3_buffer_load_v4f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 8
-; CHECK-NEXT: %data = call <2 x float> @llvm.amdgcn.buffer.load.v2f32(<4 x i32> %rsrc, i32 %idx, i32 %1, i1 false, i1 false)
-; CHECK-NEXT: ret <2 x float> %data
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.buffer.load.v4f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i1 false, i1 false)
+; CHECK-NEXT: %shuf = shufflevector <4 x float> %data, <4 x float> undef, <2 x i32> <i32 2, i32 3>
+; CHECK-NEXT: ret <2 x float> %shuf
 define amdgpu_ps <2 x float> @extract_elt2_elt3_buffer_load_v4f32(<4 x i32> inreg %rsrc, i32 %idx, i32 %ofs) #0 {
   %data = call <4 x float> @llvm.amdgcn.buffer.load.v4f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i1 false, i1 false)
   %shuf = shufflevector <4 x float> %data, <4 x float> undef, <2 x i32> <i32 2, i32 3>
@@ -133,9 +133,9 @@ define amdgpu_ps <3 x float> @extract_elt0_elt1_elt2_buffer_load_v4f32(<4 x i32>
 }
 
 ; CHECK-LABEL: @extract_elt1_elt2_elt3_buffer_load_v4f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 4
-; CHECK-NEXT: %data = call <3 x float> @llvm.amdgcn.buffer.load.v3f32(<4 x i32> %rsrc, i32 %idx, i32 %1, i1 false, i1 false)
-; CHECK-NEXT: ret <3 x float> %data
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.buffer.load.v4f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i1 false, i1 false)
+; CHECK-NEXT: %shuf = shufflevector <4 x float> %data, <4 x float> undef, <3 x i32> <i32 1, i32 2, i32 3>
+; CHECK-NEXT: ret <3 x float> %shuf
 define amdgpu_ps <3 x float> @extract_elt1_elt2_elt3_buffer_load_v4f32(<4 x i32> inreg %rsrc, i32 %idx, i32 %ofs) #0 {
   %data = call <4 x float> @llvm.amdgcn.buffer.load.v4f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i1 false, i1 false)
   %shuf = shufflevector <4 x float> %data, <4 x float> undef, <3 x i32> <i32 1, i32 2, i32 3>
@@ -246,9 +246,9 @@ define amdgpu_ps float @extract_elt0_buffer_load_v3f32(<4 x i32> inreg %rsrc, i3
 }
 
 ; CHECK-LABEL: @extract_elt1_buffer_load_v3f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 4
-; CHECK-NEXT: %data = call float @llvm.amdgcn.buffer.load.f32(<4 x i32> %rsrc, i32 %idx, i32 %1, i1 false, i1 false)
-; CHECK-NEXT: ret float %data
+; CHECK-NEXT: %data = call <2 x float> @llvm.amdgcn.buffer.load.v2f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i1 false, i1 false)
+; CHECK-NEXT: %elt1 = extractelement <2 x float> %data, i32 1
+; CHECK-NEXT: ret float %elt1
 define amdgpu_ps float @extract_elt1_buffer_load_v3f32(<4 x i32> inreg %rsrc, i32 %idx, i32 %ofs) #0 {
   %data = call <3 x float> @llvm.amdgcn.buffer.load.v3f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i1 false, i1 false)
   %elt1 = extractelement <3 x float> %data, i32 1
@@ -256,9 +256,9 @@ define amdgpu_ps float @extract_elt1_buffer_load_v3f32(<4 x i32> inreg %rsrc, i3
 }
 
 ; CHECK-LABEL: @extract_elt2_buffer_load_v3f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 8
-; CHECK-NEXT: %data = call float @llvm.amdgcn.buffer.load.f32(<4 x i32> %rsrc, i32 %idx, i32 %1, i1 false, i1 false)
-; CHECK-NEXT: ret float %data
+; CHECK-NEXT: %data = call <3 x float> @llvm.amdgcn.buffer.load.v3f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i1 false, i1 false)
+; CHECK-NEXT: %elt1 = extractelement <3 x float> %data, i32 2
+; CHECK-NEXT: ret float %elt1
 define amdgpu_ps float @extract_elt2_buffer_load_v3f32(<4 x i32> inreg %rsrc, i32 %idx, i32 %ofs) #0 {
   %data = call <3 x float> @llvm.amdgcn.buffer.load.v3f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i1 false, i1 false)
   %elt1 = extractelement <3 x float> %data, i32 2
@@ -275,9 +275,9 @@ define amdgpu_ps <2 x float> @extract_elt0_elt1_buffer_load_v3f32(<4 x i32> inre
 }
 
 ; CHECK-LABEL: @extract_elt1_elt2_buffer_load_v3f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 4
-; CHECK-NEXT: %data = call <2 x float> @llvm.amdgcn.buffer.load.v2f32(<4 x i32> %rsrc, i32 %idx, i32 %1, i1 false, i1 false)
-; CHECK-NEXT: ret <2 x float> %data
+; CHECK-NEXT: %data = call <3 x float> @llvm.amdgcn.buffer.load.v3f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i1 false, i1 false)
+; CHECK-NEXT: %shuf = shufflevector <3 x float> %data, <3 x float> undef, <2 x i32> <i32 1, i32 2>
+; CHECK-NEXT: ret <2 x float> %shuf
 define amdgpu_ps <2 x float> @extract_elt1_elt2_buffer_load_v3f32(<4 x i32> inreg %rsrc, i32 %idx, i32 %ofs) #0 {
   %data = call <3 x float> @llvm.amdgcn.buffer.load.v3f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i1 false, i1 false)
   %shuf = shufflevector <3 x float> %data, <3 x float> undef, <2 x i32> <i32 1, i32 2>
@@ -427,9 +427,9 @@ define amdgpu_ps float @extract_elt0_raw_buffer_load_v2f32(<4 x i32> inreg %rsrc
 }
 
 ; CHECK-LABEL: @extract_elt1_raw_buffer_load_v2f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 4
-; CHECK-NEXT: %data = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret float %data
+; CHECK-NEXT: %data = call <2 x float> @llvm.amdgcn.raw.buffer.load.v2f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %elt1 = extractelement <2 x float> %data, i32 1
+; CHECK-NEXT: ret float %elt1
 define amdgpu_ps float @extract_elt1_raw_buffer_load_v2f32(<4 x i32> inreg %rsrc, i32 %ofs, i32 %sofs) #0 {
   %data = call <2 x float> @llvm.amdgcn.raw.buffer.load.v2f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
   %elt1 = extractelement <2 x float> %data, i32 1
@@ -446,9 +446,9 @@ define amdgpu_ps float @extract_elt0_raw_buffer_load_v4f32(<4 x i32> inreg %rsrc
 }
 
 ; CHECK-LABEL: @extract_elt1_raw_buffer_load_v4f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 4
-; CHECK-NEXT: %data = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret float %data
+; CHECK-NEXT: %data = call <2 x float> @llvm.amdgcn.raw.buffer.load.v2f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %elt1 = extractelement <2 x float> %data, i32 1
+; CHECK-NEXT: ret float %elt1
 define amdgpu_ps float @extract_elt1_raw_buffer_load_v4f32(<4 x i32> inreg %rsrc, i32 %ofs, i32 %sofs) #0 {
   %data = call <4 x float> @llvm.amdgcn.raw.buffer.load.v4f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
   %elt1 = extractelement <4 x float> %data, i32 1
@@ -456,9 +456,9 @@ define amdgpu_ps float @extract_elt1_raw_buffer_load_v4f32(<4 x i32> inreg %rsrc
 }
 
 ; CHECK-LABEL: @extract_elt2_raw_buffer_load_v4f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 8
-; CHECK-NEXT: %data = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret float %data
+; CHECK-NEXT: %data = call <3 x float> @llvm.amdgcn.raw.buffer.load.v3f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %elt1 = extractelement <3 x float> %data, i32 2
+; CHECK-NEXT: ret float %elt1
 define amdgpu_ps float @extract_elt2_raw_buffer_load_v4f32(<4 x i32> inreg %rsrc, i32 %ofs, i32 %sofs) #0 {
   %data = call <4 x float> @llvm.amdgcn.raw.buffer.load.v4f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
   %elt1 = extractelement <4 x float> %data, i32 2
@@ -466,9 +466,9 @@ define amdgpu_ps float @extract_elt2_raw_buffer_load_v4f32(<4 x i32> inreg %rsrc
 }
 
 ; CHECK-LABEL: @extract_elt3_raw_buffer_load_v4f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 12
-; CHECK-NEXT: %data = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret float %data
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.raw.buffer.load.v4f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %elt1 = extractelement <4 x float> %data, i32 3
+; CHECK-NEXT: ret float %elt1
 define amdgpu_ps float @extract_elt3_raw_buffer_load_v4f32(<4 x i32> inreg %rsrc, i32 %ofs, i32 %sofs) #0 {
   %data = call <4 x float> @llvm.amdgcn.raw.buffer.load.v4f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
   %elt1 = extractelement <4 x float> %data, i32 3
@@ -485,9 +485,9 @@ define amdgpu_ps <2 x float> @extract_elt0_elt1_raw_buffer_load_v4f32(<4 x i32>
 }
 
 ; CHECK-LABEL: @extract_elt1_elt2_raw_buffer_load_v4f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 4
-; CHECK-NEXT: %data = call <2 x float> @llvm.amdgcn.raw.buffer.load.v2f32(<4 x i32> %rsrc, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret <2 x float> %data
+; CHECK-NEXT: %data = call <3 x float> @llvm.amdgcn.raw.buffer.load.v3f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %shuf = shufflevector <3 x float> %data, <3 x float> undef, <2 x i32> <i32 1, i32 2>
+; CHECK-NEXT: ret <2 x float> %shuf
 define amdgpu_ps <2 x float> @extract_elt1_elt2_raw_buffer_load_v4f32(<4 x i32> inreg %rsrc, i32 %ofs, i32 %sofs) #0 {
   %data = call <4 x float> @llvm.amdgcn.raw.buffer.load.v4f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
   %shuf = shufflevector <4 x float> %data, <4 x float> undef, <2 x i32> <i32 1, i32 2>
@@ -495,9 +495,9 @@ define amdgpu_ps <2 x float> @extract_elt1_elt2_raw_buffer_load_v4f32(<4 x i32>
 }
 
 ; CHECK-LABEL: @extract_elt2_elt3_raw_buffer_load_v4f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 8
-; CHECK-NEXT: %data = call <2 x float> @llvm.amdgcn.raw.buffer.load.v2f32(<4 x i32> %rsrc, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret <2 x float> %data
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.raw.buffer.load.v4f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %shuf = shufflevector <4 x float> %data, <4 x float> undef, <2 x i32> <i32 2, i32 3>
+; CHECK-NEXT: ret <2 x float> %shuf
 define amdgpu_ps <2 x float> @extract_elt2_elt3_raw_buffer_load_v4f32(<4 x i32> inreg %rsrc, i32 %ofs, i32 %sofs) #0 {
   %data = call <4 x float> @llvm.amdgcn.raw.buffer.load.v4f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
   %shuf = shufflevector <4 x float> %data, <4 x float> undef, <2 x i32> <i32 2, i32 3>
@@ -514,9 +514,9 @@ define amdgpu_ps <3 x float> @extract_elt0_elt1_elt2_raw_buffer_load_v4f32(<4 x
 }
 
 ; CHECK-LABEL: @extract_elt1_elt2_elt3_raw_buffer_load_v4f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 4
-; CHECK-NEXT: %data = call <3 x float> @llvm.amdgcn.raw.buffer.load.v3f32(<4 x i32> %rsrc, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret <3 x float> %data
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.raw.buffer.load.v4f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %shuf = shufflevector <4 x float> %data, <4 x float> undef, <3 x i32> <i32 1, i32 2, i32 3>
+; CHECK-NEXT: ret <3 x float> %shuf
 define amdgpu_ps <3 x float> @extract_elt1_elt2_elt3_raw_buffer_load_v4f32(<4 x i32> inreg %rsrc, i32 %ofs, i32 %sofs) #0 {
   %data = call <4 x float> @llvm.amdgcn.raw.buffer.load.v4f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
   %shuf = shufflevector <4 x float> %data, <4 x float> undef, <3 x i32> <i32 1, i32 2, i32 3>
@@ -543,9 +543,9 @@ define amdgpu_ps float @extract_elt0_raw_buffer_load_v3f32(<4 x i32> inreg %rsrc
 }
 
 ; CHECK-LABEL: @extract_elt1_raw_buffer_load_v3f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 4
-; CHECK-NEXT: %data = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret float %data
+; CHECK-NEXT: %data = call <2 x float> @llvm.amdgcn.raw.buffer.load.v2f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %elt1 = extractelement <2 x float> %data, i32 1
+; CHECK-NEXT: ret float %elt1
 define amdgpu_ps float @extract_elt1_raw_buffer_load_v3f32(<4 x i32> inreg %rsrc, i32 %ofs, i32 %sofs) #0 {
   %data = call <3 x float> @llvm.amdgcn.raw.buffer.load.v3f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
   %elt1 = extractelement <3 x float> %data, i32 1
@@ -553,9 +553,9 @@ define amdgpu_ps float @extract_elt1_raw_buffer_load_v3f32(<4 x i32> inreg %rsrc
 }
 
 ; CHECK-LABEL: @extract_elt2_raw_buffer_load_v3f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 8
-; CHECK-NEXT: %data = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret float %data
+; CHECK-NEXT: %data = call <3 x float> @llvm.amdgcn.raw.buffer.load.v3f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %elt1 = extractelement <3 x float> %data, i32 2
+; CHECK-NEXT: ret float %elt1
 define amdgpu_ps float @extract_elt2_raw_buffer_load_v3f32(<4 x i32> inreg %rsrc, i32 %ofs, i32 %sofs) #0 {
   %data = call <3 x float> @llvm.amdgcn.raw.buffer.load.v3f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
   %elt1 = extractelement <3 x float> %data, i32 2
@@ -572,9 +572,9 @@ define amdgpu_ps <2 x float> @extract_elt0_elt1_raw_buffer_load_v3f32(<4 x i32>
 }
 
 ; CHECK-LABEL: @extract_elt1_elt2_raw_buffer_load_v3f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 4
-; CHECK-NEXT: %data = call <2 x float> @llvm.amdgcn.raw.buffer.load.v2f32(<4 x i32> %rsrc, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret <2 x float> %data
+; CHECK-NEXT: %data = call <3 x float> @llvm.amdgcn.raw.buffer.load.v3f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %shuf = shufflevector <3 x float> %data, <3 x float> undef, <2 x i32> <i32 1, i32 2>
+; CHECK-NEXT: ret <2 x float> %shuf
 define amdgpu_ps <2 x float> @extract_elt1_elt2_raw_buffer_load_v3f32(<4 x i32> inreg %rsrc, i32 %ofs, i32 %sofs) #0 {
   %data = call <3 x float> @llvm.amdgcn.raw.buffer.load.v3f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
   %shuf = shufflevector <3 x float> %data, <3 x float> undef, <2 x i32> <i32 1, i32 2>
@@ -666,9 +666,9 @@ define amdgpu_ps float @extract_elt0_raw_buffer_load_format_v2f32(<4 x i32> inre
 }
 
 ; CHECK-LABEL: @extract_elt1_raw_buffer_load_format_v2f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 4
-; CHECK-NEXT: %data = call float @llvm.amdgcn.raw.buffer.load.format.f32(<4 x i32> %rsrc, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret float %data
+; CHECK-NEXT: %data = call <2 x float> @llvm.amdgcn.raw.buffer.load.format.v2f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %elt1 = extractelement <2 x float> %data, i32 1
+; CHECK-NEXT: ret float %elt1
 define amdgpu_ps float @extract_elt1_raw_buffer_load_format_v2f32(<4 x i32> inreg %rsrc, i32 %ofs, i32 %sofs) #0 {
   %data = call <2 x float> @llvm.amdgcn.raw.buffer.load.format.v2f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
   %elt1 = extractelement <2 x float> %data, i32 1
@@ -685,9 +685,9 @@ define amdgpu_ps float @extract_elt0_raw_buffer_load_format_v4f32(<4 x i32> inre
 }
 
 ; CHECK-LABEL: @extract_elt1_raw_buffer_load_format_v4f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 4
-; CHECK-NEXT: %data = call float @llvm.amdgcn.raw.buffer.load.format.f32(<4 x i32> %rsrc, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret float %data
+; CHECK-NEXT: %data = call <2 x float> @llvm.amdgcn.raw.buffer.load.format.v2f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %elt1 = extractelement <2 x float> %data, i32 1
+; CHECK-NEXT: ret float %elt1
 define amdgpu_ps float @extract_elt1_raw_buffer_load_format_v4f32(<4 x i32> inreg %rsrc, i32 %ofs, i32 %sofs) #0 {
   %data = call <4 x float> @llvm.amdgcn.raw.buffer.load.format.v4f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
   %elt1 = extractelement <4 x float> %data, i32 1
@@ -695,9 +695,9 @@ define amdgpu_ps float @extract_elt1_raw_buffer_load_format_v4f32(<4 x i32> inre
 }
 
 ; CHECK-LABEL: @extract_elt2_raw_buffer_load_format_v4f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 8
-; CHECK-NEXT: %data = call float @llvm.amdgcn.raw.buffer.load.format.f32(<4 x i32> %rsrc, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret float %data
+; CHECK-NEXT: %data = call <3 x float> @llvm.amdgcn.raw.buffer.load.format.v3f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %elt1 = extractelement <3 x float> %data, i32 2
+; CHECK-NEXT: ret float %elt1
 define amdgpu_ps float @extract_elt2_raw_buffer_load_format_v4f32(<4 x i32> inreg %rsrc, i32 %ofs, i32 %sofs) #0 {
   %data = call <4 x float> @llvm.amdgcn.raw.buffer.load.format.v4f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
   %elt1 = extractelement <4 x float> %data, i32 2
@@ -705,9 +705,9 @@ define amdgpu_ps float @extract_elt2_raw_buffer_load_format_v4f32(<4 x i32> inre
 }
 
 ; CHECK-LABEL: @extract_elt3_raw_buffer_load_format_v4f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 12
-; CHECK-NEXT: %data = call float @llvm.amdgcn.raw.buffer.load.format.f32(<4 x i32> %rsrc, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret float %data
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.raw.buffer.load.format.v4f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %elt1 = extractelement <4 x float> %data, i32 3
+; CHECK-NEXT: ret float %elt1
 define amdgpu_ps float @extract_elt3_raw_buffer_load_format_v4f32(<4 x i32> inreg %rsrc, i32 %ofs, i32 %sofs) #0 {
   %data = call <4 x float> @llvm.amdgcn.raw.buffer.load.format.v4f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
   %elt1 = extractelement <4 x float> %data, i32 3
@@ -724,9 +724,9 @@ define amdgpu_ps <2 x float> @extract_elt0_elt1_raw_buffer_load_format_v4f32(<4
 }
 
 ; CHECK-LABEL: @extract_elt1_elt2_raw_buffer_load_format_v4f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 4
-; CHECK-NEXT: %data = call <2 x float> @llvm.amdgcn.raw.buffer.load.format.v2f32(<4 x i32> %rsrc, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret <2 x float> %data
+; CHECK-NEXT: %data = call <3 x float> @llvm.amdgcn.raw.buffer.load.format.v3f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %shuf = shufflevector <3 x float> %data, <3 x float> undef, <2 x i32> <i32 1, i32 2>
+; CHECK-NEXT: ret <2 x float> %shuf
 define amdgpu_ps <2 x float> @extract_elt1_elt2_raw_buffer_load_format_v4f32(<4 x i32> inreg %rsrc, i32 %ofs, i32 %sofs) #0 {
   %data = call <4 x float> @llvm.amdgcn.raw.buffer.load.format.v4f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
   %shuf = shufflevector <4 x float> %data, <4 x float> undef, <2 x i32> <i32 1, i32 2>
@@ -734,9 +734,9 @@ define amdgpu_ps <2 x float> @extract_elt1_elt2_raw_buffer_load_format_v4f32(<4
 }
 
 ; CHECK-LABEL: @extract_elt2_elt3_raw_buffer_load_format_v4f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 8
-; CHECK-NEXT: %data = call <2 x float> @llvm.amdgcn.raw.buffer.load.format.v2f32(<4 x i32> %rsrc, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret <2 x float> %data
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.raw.buffer.load.format.v4f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %shuf = shufflevector <4 x float> %data, <4 x float> undef, <2 x i32> <i32 2, i32 3>
+; CHECK-NEXT: ret <2 x float> %shuf
 define amdgpu_ps <2 x float> @extract_elt2_elt3_raw_buffer_load_format_v4f32(<4 x i32> inreg %rsrc, i32 %ofs, i32 %sofs) #0 {
   %data = call <4 x float> @llvm.amdgcn.raw.buffer.load.format.v4f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
   %shuf = shufflevector <4 x float> %data, <4 x float> undef, <2 x i32> <i32 2, i32 3>
@@ -753,9 +753,9 @@ define amdgpu_ps <3 x float> @extract_elt0_elt1_elt2_raw_buffer_load_format_v4f3
 }
 
 ; CHECK-LABEL: @extract_elt1_elt2_elt3_raw_buffer_load_format_v4f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 4
-; CHECK-NEXT: %data = call <3 x float> @llvm.amdgcn.raw.buffer.load.format.v3f32(<4 x i32> %rsrc, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret <3 x float> %data
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.raw.buffer.load.format.v4f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %shuf = shufflevector <4 x float> %data, <4 x float> undef, <3 x i32> <i32 1, i32 2, i32 3>
+; CHECK-NEXT: ret <3 x float> %shuf
 define amdgpu_ps <3 x float> @extract_elt1_elt2_elt3_raw_buffer_load_format_v4f32(<4 x i32> inreg %rsrc, i32 %ofs, i32 %sofs) #0 {
   %data = call <4 x float> @llvm.amdgcn.raw.buffer.load.format.v4f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
   %shuf = shufflevector <4 x float> %data, <4 x float> undef, <3 x i32> <i32 1, i32 2, i32 3>
@@ -782,9 +782,9 @@ define amdgpu_ps float @extract_elt0_raw_buffer_load_format_v3f32(<4 x i32> inre
 }
 
 ; CHECK-LABEL: @extract_elt1_raw_buffer_load_format_v3f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 4
-; CHECK-NEXT: %data = call float @llvm.amdgcn.raw.buffer.load.format.f32(<4 x i32> %rsrc, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret float %data
+; CHECK-NEXT: %data = call <2 x float> @llvm.amdgcn.raw.buffer.load.format.v2f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %elt1 = extractelement <2 x float> %data, i32 1
+; CHECK-NEXT: ret float %elt1
 define amdgpu_ps float @extract_elt1_raw_buffer_load_format_v3f32(<4 x i32> inreg %rsrc, i32 %ofs, i32 %sofs) #0 {
   %data = call <3 x float> @llvm.amdgcn.raw.buffer.load.format.v3f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
   %elt1 = extractelement <3 x float> %data, i32 1
@@ -792,9 +792,9 @@ define amdgpu_ps float @extract_elt1_raw_buffer_load_format_v3f32(<4 x i32> inre
 }
 
 ; CHECK-LABEL: @extract_elt2_raw_buffer_load_format_v3f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 8
-; CHECK-NEXT: %data = call float @llvm.amdgcn.raw.buffer.load.format.f32(<4 x i32> %rsrc, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret float %data
+; CHECK-NEXT: %data = call <3 x float> @llvm.amdgcn.raw.buffer.load.format.v3f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %elt1 = extractelement <3 x float> %data, i32 2
+; CHECK-NEXT: ret float %elt1
 define amdgpu_ps float @extract_elt2_raw_buffer_load_format_v3f32(<4 x i32> inreg %rsrc, i32 %ofs, i32 %sofs) #0 {
   %data = call <3 x float> @llvm.amdgcn.raw.buffer.load.format.v3f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
   %elt1 = extractelement <3 x float> %data, i32 2
@@ -811,9 +811,9 @@ define amdgpu_ps <2 x float> @extract_elt0_elt1_raw_buffer_load_format_v3f32(<4
 }
 
 ; CHECK-LABEL: @extract_elt1_elt2_raw_buffer_load_format_v3f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 4
-; CHECK-NEXT: %data = call <2 x float> @llvm.amdgcn.raw.buffer.load.format.v2f32(<4 x i32> %rsrc, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret <2 x float> %data
+; CHECK-NEXT: %data = call <3 x float> @llvm.amdgcn.raw.buffer.load.format.v3f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %shuf = shufflevector <3 x float> %data, <3 x float> undef, <2 x i32> <i32 1, i32 2>
+; CHECK-NEXT: ret <2 x float> %shuf
 define amdgpu_ps <2 x float> @extract_elt1_elt2_raw_buffer_load_format_v3f32(<4 x i32> inreg %rsrc, i32 %ofs, i32 %sofs) #0 {
   %data = call <3 x float> @llvm.amdgcn.raw.buffer.load.format.v3f32(<4 x i32> %rsrc, i32 %ofs, i32 %sofs, i32 0)
   %shuf = shufflevector <3 x float> %data, <3 x float> undef, <2 x i32> <i32 1, i32 2>
@@ -901,9 +901,9 @@ define amdgpu_ps float @extract_elt0_struct_buffer_load_v2f32(<4 x i32> inreg %r
 }
 
 ; CHECK-LABEL: @extract_elt1_struct_buffer_load_v2f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 4
-; CHECK-NEXT: %data = call float @llvm.amdgcn.struct.buffer.load.f32(<4 x i32> %rsrc, i32 %idx, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret float %data
+; CHECK-NEXT: %data = call <2 x float> @llvm.amdgcn.struct.buffer.load.v2f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %elt1 = extractelement <2 x float> %data, i32 1
+; CHECK-NEXT: ret float %elt1
 define amdgpu_ps float @extract_elt1_struct_buffer_load_v2f32(<4 x i32> inreg %rsrc, i32 %idx, i32 %ofs, i32 %sofs) #0 {
   %data = call <2 x float> @llvm.amdgcn.struct.buffer.load.v2f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
   %elt1 = extractelement <2 x float> %data, i32 1
@@ -920,9 +920,9 @@ define amdgpu_ps float @extract_elt0_struct_buffer_load_v4f32(<4 x i32> inreg %r
 }
 
 ; CHECK-LABEL: @extract_elt1_struct_buffer_load_v4f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 4
-; CHECK-NEXT: %data = call float @llvm.amdgcn.struct.buffer.load.f32(<4 x i32> %rsrc, i32 %idx, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret float %data
+; CHECK-NEXT: %data = call <2 x float> @llvm.amdgcn.struct.buffer.load.v2f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %elt1 = extractelement <2 x float> %data, i32 1
+; CHECK-NEXT: ret float %elt1
 define amdgpu_ps float @extract_elt1_struct_buffer_load_v4f32(<4 x i32> inreg %rsrc, i32 %idx, i32 %ofs, i32 %sofs) #0 {
   %data = call <4 x float> @llvm.amdgcn.struct.buffer.load.v4f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
   %elt1 = extractelement <4 x float> %data, i32 1
@@ -930,9 +930,9 @@ define amdgpu_ps float @extract_elt1_struct_buffer_load_v4f32(<4 x i32> inreg %r
 }
 
 ; CHECK-LABEL: @extract_elt2_struct_buffer_load_v4f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 8
-; CHECK-NEXT: %data = call float @llvm.amdgcn.struct.buffer.load.f32(<4 x i32> %rsrc, i32 %idx, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret float %data
+; CHECK-NEXT: %data = call <3 x float> @llvm.amdgcn.struct.buffer.load.v3f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %elt1 = extractelement <3 x float> %data, i32 2
+; CHECK-NEXT: ret float %elt1
 define amdgpu_ps float @extract_elt2_struct_buffer_load_v4f32(<4 x i32> inreg %rsrc, i32 %idx, i32 %ofs, i32 %sofs) #0 {
   %data = call <4 x float> @llvm.amdgcn.struct.buffer.load.v4f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
   %elt1 = extractelement <4 x float> %data, i32 2
@@ -940,9 +940,9 @@ define amdgpu_ps float @extract_elt2_struct_buffer_load_v4f32(<4 x i32> inreg %r
 }
 
 ; CHECK-LABEL: @extract_elt3_struct_buffer_load_v4f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 12
-; CHECK-NEXT: %data = call float @llvm.amdgcn.struct.buffer.load.f32(<4 x i32> %rsrc, i32 %idx, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret float %data
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.struct.buffer.load.v4f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %elt1 = extractelement <4 x float> %data, i32 3
+; CHECK-NEXT: ret float %elt1
 define amdgpu_ps float @extract_elt3_struct_buffer_load_v4f32(<4 x i32> inreg %rsrc, i32 %idx, i32 %ofs, i32 %sofs) #0 {
   %data = call <4 x float> @llvm.amdgcn.struct.buffer.load.v4f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
   %elt1 = extractelement <4 x float> %data, i32 3
@@ -959,9 +959,9 @@ define amdgpu_ps <2 x float> @extract_elt0_elt1_struct_buffer_load_v4f32(<4 x i3
 }
 
 ; CHECK-LABEL: @extract_elt1_elt2_struct_buffer_load_v4f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 4
-; CHECK-NEXT: %data = call <2 x float> @llvm.amdgcn.struct.buffer.load.v2f32(<4 x i32> %rsrc, i32 %idx, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret <2 x float> %data
+; CHECK-NEXT: %data = call <3 x float> @llvm.amdgcn.struct.buffer.load.v3f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %shuf = shufflevector <3 x float> %data, <3 x float> undef, <2 x i32> <i32 1, i32 2>
+; CHECK-NEXT: ret <2 x float> %shuf
 define amdgpu_ps <2 x float> @extract_elt1_elt2_struct_buffer_load_v4f32(<4 x i32> inreg %rsrc, i32 %idx, i32 %ofs, i32 %sofs) #0 {
   %data = call <4 x float> @llvm.amdgcn.struct.buffer.load.v4f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
   %shuf = shufflevector <4 x float> %data, <4 x float> undef, <2 x i32> <i32 1, i32 2>
@@ -969,9 +969,9 @@ define amdgpu_ps <2 x float> @extract_elt1_elt2_struct_buffer_load_v4f32(<4 x i3
 }
 
 ; CHECK-LABEL: @extract_elt2_elt3_struct_buffer_load_v4f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 8
-; CHECK-NEXT: %data = call <2 x float> @llvm.amdgcn.struct.buffer.load.v2f32(<4 x i32> %rsrc, i32 %idx, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret <2 x float> %data
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.struct.buffer.load.v4f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %shuf = shufflevector <4 x float> %data, <4 x float> undef, <2 x i32> <i32 2, i32 3>
+; CHECK-NEXT: ret <2 x float> %shuf
 define amdgpu_ps <2 x float> @extract_elt2_elt3_struct_buffer_load_v4f32(<4 x i32> inreg %rsrc, i32 %idx, i32 %ofs, i32 %sofs) #0 {
   %data = call <4 x float> @llvm.amdgcn.struct.buffer.load.v4f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
   %shuf = shufflevector <4 x float> %data, <4 x float> undef, <2 x i32> <i32 2, i32 3>
@@ -988,9 +988,9 @@ define amdgpu_ps <3 x float> @extract_elt0_elt1_elt2_struct_buffer_load_v4f32(<4
 }
 
 ; CHECK-LABEL: @extract_elt1_elt2_elt3_struct_buffer_load_v4f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 4
-; CHECK-NEXT: %data = call <3 x float> @llvm.amdgcn.struct.buffer.load.v3f32(<4 x i32> %rsrc, i32 %idx, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret <3 x float> %data
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.struct.buffer.load.v4f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %shuf = shufflevector <4 x float> %data, <4 x float> undef, <3 x i32> <i32 1, i32 2, i32 3>
+; CHECK-NEXT: ret <3 x float> %shuf
 define amdgpu_ps <3 x float> @extract_elt1_elt2_elt3_struct_buffer_load_v4f32(<4 x i32> inreg %rsrc, i32 %idx, i32 %ofs, i32 %sofs) #0 {
   %data = call <4 x float> @llvm.amdgcn.struct.buffer.load.v4f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
   %shuf = shufflevector <4 x float> %data, <4 x float> undef, <3 x i32> <i32 1, i32 2, i32 3>
@@ -1017,9 +1017,9 @@ define amdgpu_ps float @extract_elt0_struct_buffer_load_v3f32(<4 x i32> inreg %r
 }
 
 ; CHECK-LABEL: @extract_elt1_struct_buffer_load_v3f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 4
-; CHECK-NEXT: %data = call float @llvm.amdgcn.struct.buffer.load.f32(<4 x i32> %rsrc, i32 %idx, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret float %data
+; CHECK-NEXT: %data = call <2 x float> @llvm.amdgcn.struct.buffer.load.v2f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %elt1 = extractelement <2 x float> %data, i32 1
+; CHECK-NEXT: ret float %elt1
 define amdgpu_ps float @extract_elt1_struct_buffer_load_v3f32(<4 x i32> inreg %rsrc, i32 %idx, i32 %ofs, i32 %sofs) #0 {
   %data = call <3 x float> @llvm.amdgcn.struct.buffer.load.v3f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
   %elt1 = extractelement <3 x float> %data, i32 1
@@ -1027,9 +1027,9 @@ define amdgpu_ps float @extract_elt1_struct_buffer_load_v3f32(<4 x i32> inreg %r
 }
 
 ; CHECK-LABEL: @extract_elt2_struct_buffer_load_v3f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 8
-; CHECK-NEXT: %data = call float @llvm.amdgcn.struct.buffer.load.f32(<4 x i32> %rsrc, i32 %idx, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret float %data
+; CHECK-NEXT: %data = call <3 x float> @llvm.amdgcn.struct.buffer.load.v3f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %elt1 = extractelement <3 x float> %data, i32 2
+; CHECK-NEXT: ret float %elt1
 define amdgpu_ps float @extract_elt2_struct_buffer_load_v3f32(<4 x i32> inreg %rsrc, i32 %idx, i32 %ofs, i32 %sofs) #0 {
   %data = call <3 x float> @llvm.amdgcn.struct.buffer.load.v3f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
   %elt1 = extractelement <3 x float> %data, i32 2
@@ -1046,9 +1046,9 @@ define amdgpu_ps <2 x float> @extract_elt0_elt1_struct_buffer_load_v3f32(<4 x i3
 }
 
 ; CHECK-LABEL: @extract_elt1_elt2_struct_buffer_load_v3f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 4
-; CHECK-NEXT: %data = call <2 x float> @llvm.amdgcn.struct.buffer.load.v2f32(<4 x i32> %rsrc, i32 %idx, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret <2 x float> %data
+; CHECK-NEXT: %data = call <3 x float> @llvm.amdgcn.struct.buffer.load.v3f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %shuf = shufflevector <3 x float> %data, <3 x float> undef, <2 x i32> <i32 1, i32 2>
+; CHECK-NEXT: ret <2 x float> %shuf
 define amdgpu_ps <2 x float> @extract_elt1_elt2_struct_buffer_load_v3f32(<4 x i32> inreg %rsrc, i32 %idx, i32 %ofs, i32 %sofs) #0 {
   %data = call <3 x float> @llvm.amdgcn.struct.buffer.load.v3f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
   %shuf = shufflevector <3 x float> %data, <3 x float> undef, <2 x i32> <i32 1, i32 2>
@@ -1140,9 +1140,9 @@ define amdgpu_ps float @extract_elt0_struct_buffer_load_format_v2f32(<4 x i32> i
 }
 
 ; CHECK-LABEL: @extract_elt1_struct_buffer_load_format_v2f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 4
-; CHECK-NEXT: %data = call float @llvm.amdgcn.struct.buffer.load.format.f32(<4 x i32> %rsrc, i32 %idx, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret float %data
+; CHECK-NEXT: %data = call <2 x float> @llvm.amdgcn.struct.buffer.load.format.v2f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %elt1 = extractelement <2 x float> %data, i32 1
+; CHECK-NEXT: ret float %elt1
 define amdgpu_ps float @extract_elt1_struct_buffer_load_format_v2f32(<4 x i32> inreg %rsrc, i32 %idx, i32 %ofs, i32 %sofs) #0 {
   %data = call <2 x float> @llvm.amdgcn.struct.buffer.load.format.v2f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
   %elt1 = extractelement <2 x float> %data, i32 1
@@ -1159,9 +1159,9 @@ define amdgpu_ps float @extract_elt0_struct_buffer_load_format_v4f32(<4 x i32> i
 }
 
 ; CHECK-LABEL: @extract_elt1_struct_buffer_load_format_v4f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 4
-; CHECK-NEXT: %data = call float @llvm.amdgcn.struct.buffer.load.format.f32(<4 x i32> %rsrc, i32 %idx, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret float %data
+; CHECK-NEXT: %data = call <2 x float> @llvm.amdgcn.struct.buffer.load.format.v2f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %elt1 = extractelement <2 x float> %data, i32 1
+; CHECK-NEXT: ret float %elt1
 define amdgpu_ps float @extract_elt1_struct_buffer_load_format_v4f32(<4 x i32> inreg %rsrc, i32 %idx, i32 %ofs, i32 %sofs) #0 {
   %data = call <4 x float> @llvm.amdgcn.struct.buffer.load.format.v4f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
   %elt1 = extractelement <4 x float> %data, i32 1
@@ -1169,9 +1169,9 @@ define amdgpu_ps float @extract_elt1_struct_buffer_load_format_v4f32(<4 x i32> i
 }
 
 ; CHECK-LABEL: @extract_elt2_struct_buffer_load_format_v4f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 8
-; CHECK-NEXT: %data = call float @llvm.amdgcn.struct.buffer.load.format.f32(<4 x i32> %rsrc, i32 %idx, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret float %data
+; CHECK-NEXT: %data = call <3 x float> @llvm.amdgcn.struct.buffer.load.format.v3f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %elt1 = extractelement <3 x float> %data, i32 2
+; CHECK-NEXT: ret float %elt1
 define amdgpu_ps float @extract_elt2_struct_buffer_load_format_v4f32(<4 x i32> inreg %rsrc, i32 %idx, i32 %ofs, i32 %sofs) #0 {
   %data = call <4 x float> @llvm.amdgcn.struct.buffer.load.format.v4f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
   %elt1 = extractelement <4 x float> %data, i32 2
@@ -1179,9 +1179,9 @@ define amdgpu_ps float @extract_elt2_struct_buffer_load_format_v4f32(<4 x i32> i
 }
 
 ; CHECK-LABEL: @extract_elt3_struct_buffer_load_format_v4f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 12
-; CHECK-NEXT: %data = call float @llvm.amdgcn.struct.buffer.load.format.f32(<4 x i32> %rsrc, i32 %idx, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret float %data
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.struct.buffer.load.format.v4f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %elt1 = extractelement <4 x float> %data, i32 3
+; CHECK-NEXT: ret float %elt1
 define amdgpu_ps float @extract_elt3_struct_buffer_load_format_v4f32(<4 x i32> inreg %rsrc, i32 %idx, i32 %ofs, i32 %sofs) #0 {
   %data = call <4 x float> @llvm.amdgcn.struct.buffer.load.format.v4f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
   %elt1 = extractelement <4 x float> %data, i32 3
@@ -1198,9 +1198,9 @@ define amdgpu_ps <2 x float> @extract_elt0_elt1_struct_buffer_load_format_v4f32(
 }
 
 ; CHECK-LABEL: @extract_elt1_elt2_struct_buffer_load_format_v4f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 4
-; CHECK-NEXT: %data = call <2 x float> @llvm.amdgcn.struct.buffer.load.format.v2f32(<4 x i32> %rsrc, i32 %idx, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret <2 x float> %data
+; CHECK-NEXT: %data = call <3 x float> @llvm.amdgcn.struct.buffer.load.format.v3f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %shuf = shufflevector <3 x float> %data, <3 x float> undef, <2 x i32> <i32 1, i32 2>
+; CHECK-NEXT: ret <2 x float> %shuf
 define amdgpu_ps <2 x float> @extract_elt1_elt2_struct_buffer_load_format_v4f32(<4 x i32> inreg %rsrc, i32 %idx, i32 %ofs, i32 %sofs) #0 {
   %data = call <4 x float> @llvm.amdgcn.struct.buffer.load.format.v4f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
   %shuf = shufflevector <4 x float> %data, <4 x float> undef, <2 x i32> <i32 1, i32 2>
@@ -1208,9 +1208,9 @@ define amdgpu_ps <2 x float> @extract_elt1_elt2_struct_buffer_load_format_v4f32(
 }
 
 ; CHECK-LABEL: @extract_elt2_elt3_struct_buffer_load_format_v4f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 8
-; CHECK-NEXT: %data = call <2 x float> @llvm.amdgcn.struct.buffer.load.format.v2f32(<4 x i32> %rsrc, i32 %idx, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret <2 x float> %data
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.struct.buffer.load.format.v4f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %shuf = shufflevector <4 x float> %data, <4 x float> undef, <2 x i32> <i32 2, i32 3>
+; CHECK-NEXT: ret <2 x float> %shuf
 define amdgpu_ps <2 x float> @extract_elt2_elt3_struct_buffer_load_format_v4f32(<4 x i32> inreg %rsrc, i32 %idx, i32 %ofs, i32 %sofs) #0 {
   %data = call <4 x float> @llvm.amdgcn.struct.buffer.load.format.v4f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
   %shuf = shufflevector <4 x float> %data, <4 x float> undef, <2 x i32> <i32 2, i32 3>
@@ -1227,9 +1227,9 @@ define amdgpu_ps <3 x float> @extract_elt0_elt1_elt2_struct_buffer_load_format_v
 }
 
 ; CHECK-LABEL: @extract_elt1_elt2_elt3_struct_buffer_load_format_v4f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 4
-; CHECK-NEXT: %data = call <3 x float> @llvm.amdgcn.struct.buffer.load.format.v3f32(<4 x i32> %rsrc, i32 %idx, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret <3 x float> %data
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.struct.buffer.load.format.v4f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %shuf = shufflevector <4 x float> %data, <4 x float> undef, <3 x i32> <i32 1, i32 2, i32 3>
+; CHECK-NEXT: ret <3 x float> %shuf
 define amdgpu_ps <3 x float> @extract_elt1_elt2_elt3_struct_buffer_load_format_v4f32(<4 x i32> inreg %rsrc, i32 %idx, i32 %ofs, i32 %sofs) #0 {
   %data = call <4 x float> @llvm.amdgcn.struct.buffer.load.format.v4f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
   %shuf = shufflevector <4 x float> %data, <4 x float> undef, <3 x i32> <i32 1, i32 2, i32 3>
@@ -1256,9 +1256,9 @@ define amdgpu_ps float @extract_elt0_struct_buffer_load_format_v3f32(<4 x i32> i
 }
 
 ; CHECK-LABEL: @extract_elt1_struct_buffer_load_format_v3f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 4
-; CHECK-NEXT: %data = call float @llvm.amdgcn.struct.buffer.load.format.f32(<4 x i32> %rsrc, i32 %idx, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret float %data
+; CHECK-NEXT: %data = call <2 x float> @llvm.amdgcn.struct.buffer.load.format.v2f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %elt1 = extractelement <2 x float> %data, i32 1
+; CHECK-NEXT: ret float %elt1
 define amdgpu_ps float @extract_elt1_struct_buffer_load_format_v3f32(<4 x i32> inreg %rsrc, i32 %idx, i32 %ofs, i32 %sofs) #0 {
   %data = call <3 x float> @llvm.amdgcn.struct.buffer.load.format.v3f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
   %elt1 = extractelement <3 x float> %data, i32 1
@@ -1266,9 +1266,9 @@ define amdgpu_ps float @extract_elt1_struct_buffer_load_format_v3f32(<4 x i32> i
 }
 
 ; CHECK-LABEL: @extract_elt2_struct_buffer_load_format_v3f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 8
-; CHECK-NEXT: %data = call float @llvm.amdgcn.struct.buffer.load.format.f32(<4 x i32> %rsrc, i32 %idx, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret float %data
+; CHECK-NEXT: %data = call <3 x float> @llvm.amdgcn.struct.buffer.load.format.v3f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %elt1 = extractelement <3 x float> %data, i32 2
+; CHECK-NEXT: ret float %elt1
 define amdgpu_ps float @extract_elt2_struct_buffer_load_format_v3f32(<4 x i32> inreg %rsrc, i32 %idx, i32 %ofs, i32 %sofs) #0 {
   %data = call <3 x float> @llvm.amdgcn.struct.buffer.load.format.v3f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
   %elt1 = extractelement <3 x float> %data, i32 2
@@ -1285,9 +1285,9 @@ define amdgpu_ps <2 x float> @extract_elt0_elt1_struct_buffer_load_format_v3f32(
 }
 
 ; CHECK-LABEL: @extract_elt1_elt2_struct_buffer_load_format_v3f32(
-; CHECK-NEXT: %1 = add i32 %ofs, 4
-; CHECK-NEXT: %data = call <2 x float> @llvm.amdgcn.struct.buffer.load.format.v2f32(<4 x i32> %rsrc, i32 %idx, i32 %1, i32 %sofs, i32 0)
-; CHECK-NEXT: ret <2 x float> %data
+; CHECK-NEXT: %data = call <3 x float> @llvm.amdgcn.struct.buffer.load.format.v3f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
+; CHECK-NEXT: %shuf = shufflevector <3 x float> %data, <3 x float> undef, <2 x i32> <i32 1, i32 2>
+; CHECK-NEXT: ret <2 x float> %shuf
 define amdgpu_ps <2 x float> @extract_elt1_elt2_struct_buffer_load_format_v3f32(<4 x i32> inreg %rsrc, i32 %idx, i32 %ofs, i32 %sofs) #0 {
   %data = call <3 x float> @llvm.amdgcn.struct.buffer.load.format.v3f32(<4 x i32> %rsrc, i32 %idx, i32 %ofs, i32 %sofs, i32 0)
   %shuf = shufflevector <3 x float> %data, <3 x float> undef, <2 x i32> <i32 1, i32 2>