AMDGPU: SimplifyDemandedElts for image intrinsics

Causes some VGPR usage improvements in shaderdb, but
introduces some SGPR spilling regressions due to random
scheduling changes later.

llvm-svn: 300453

diff --git a/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp b/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp
index abc7d07d4512bcb904240e959b80bcc4ff5afceb..3ce5a5c33c2348a5919beb6ac1c5830622850d8f 100644 (file)
--- a/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp
@@ -1641,7 +1641,52 @@ Value *InstCombiner::SimplifyDemandedVectorElts(Value *V, APInt DemandedElts,
       UndefElts.setHighBits(VWidth / 2);
       break;
     case Intrinsic::amdgcn_buffer_load:
-    case Intrinsic::amdgcn_buffer_load_format: {
+    case Intrinsic::amdgcn_buffer_load_format:
+    case Intrinsic::amdgcn_image_sample:
+    case Intrinsic::amdgcn_image_sample_cl:
+    case Intrinsic::amdgcn_image_sample_d:
+    case Intrinsic::amdgcn_image_sample_d_cl:
+    case Intrinsic::amdgcn_image_sample_l:
+    case Intrinsic::amdgcn_image_sample_b:
+    case Intrinsic::amdgcn_image_sample_b_cl:
+    case Intrinsic::amdgcn_image_sample_lz:
+    case Intrinsic::amdgcn_image_sample_cd:
+    case Intrinsic::amdgcn_image_sample_cd_cl:
+
+    case Intrinsic::amdgcn_image_sample_c:
+    case Intrinsic::amdgcn_image_sample_c_cl:
+    case Intrinsic::amdgcn_image_sample_c_d:
+    case Intrinsic::amdgcn_image_sample_c_d_cl:
+    case Intrinsic::amdgcn_image_sample_c_l:
+    case Intrinsic::amdgcn_image_sample_c_b:
+    case Intrinsic::amdgcn_image_sample_c_b_cl:
+    case Intrinsic::amdgcn_image_sample_c_lz:
+    case Intrinsic::amdgcn_image_sample_c_cd:
+    case Intrinsic::amdgcn_image_sample_c_cd_cl:
+
+    case Intrinsic::amdgcn_image_sample_o:
+    case Intrinsic::amdgcn_image_sample_cl_o:
+    case Intrinsic::amdgcn_image_sample_d_o:
+    case Intrinsic::amdgcn_image_sample_d_cl_o:
+    case Intrinsic::amdgcn_image_sample_l_o:
+    case Intrinsic::amdgcn_image_sample_b_o:
+    case Intrinsic::amdgcn_image_sample_b_cl_o:
+    case Intrinsic::amdgcn_image_sample_lz_o:
+    case Intrinsic::amdgcn_image_sample_cd_o:
+    case Intrinsic::amdgcn_image_sample_cd_cl_o:
+
+    case Intrinsic::amdgcn_image_sample_c_o:
+    case Intrinsic::amdgcn_image_sample_c_cl_o:
+    case Intrinsic::amdgcn_image_sample_c_d_o:
+    case Intrinsic::amdgcn_image_sample_c_d_cl_o:
+    case Intrinsic::amdgcn_image_sample_c_l_o:
+    case Intrinsic::amdgcn_image_sample_c_b_o:
+    case Intrinsic::amdgcn_image_sample_c_b_cl_o:
+    case Intrinsic::amdgcn_image_sample_c_lz_o:
+    case Intrinsic::amdgcn_image_sample_c_cd_o:
+    case Intrinsic::amdgcn_image_sample_c_cd_cl_o:
+
+    case Intrinsic::amdgcn_image_getlod: {
       if (VWidth == 1 || !DemandedElts.isMask())
         return nullptr;
 
@@ -1656,8 +1701,17 @@ Value *InstCombiner::SimplifyDemandedVectorElts(Value *V, APInt DemandedElts,
       Type *NewTy = (NewNumElts == 1) ? EltTy :
         VectorType::get(EltTy, NewNumElts);
 
-      Function *NewIntrin = Intrinsic::getDeclaration(M, II->getIntrinsicID(),
-                                                      NewTy);
+      auto IID = II->getIntrinsicID();
+
+      bool IsBuffer = IID == Intrinsic::amdgcn_buffer_load ||
+                      IID == Intrinsic::amdgcn_buffer_load_format;
+
+      Function *NewIntrin = IsBuffer ?
+        Intrinsic::getDeclaration(M, IID, NewTy) :
+        // Samplers have 3 mangled types.
+        Intrinsic::getDeclaration(M, IID,
+                                  { NewTy, II->getArgOperand(0)->getType(),
+                                      II->getArgOperand(1)->getType()});
 
       SmallVector<Value *, 5> Args;
       for (unsigned I = 0, E = II->getNumArgOperands(); I != E; ++I)
@@ -1669,6 +1723,29 @@ Value *InstCombiner::SimplifyDemandedVectorElts(Value *V, APInt DemandedElts,
       CallInst *NewCall = Builder->CreateCall(NewIntrin, Args);
       NewCall->takeName(II);
       NewCall->copyMetadata(*II);
+
+      if (!IsBuffer) {
+        ConstantInt *DMask = dyn_cast<ConstantInt>(NewCall->getArgOperand(3));
+        if (DMask) {
+          unsigned DMaskVal = DMask->getZExtValue() & 0xf;
+
+          unsigned PopCnt = 0;
+          unsigned NewDMask = 0;
+          for (unsigned I = 0; I < 4; ++I) {
+            const unsigned Bit = 1 << I;
+            if (!!(DMaskVal & Bit)) {
+              if (++PopCnt > NewNumElts)
+                break;
+
+              NewDMask |= Bit;
+            }
+          }
+
+          NewCall->setArgOperand(3, ConstantInt::get(DMask->getType(), NewDMask));
+        }
+      }
+
+
       if (NewNumElts == 1) {
         return Builder->CreateInsertElement(UndefValue::get(V->getType()),
                                             NewCall, static_cast<uint64_t>(0));

diff --git a/llvm/test/Transforms/InstCombine/amdgcn-demanded-vector-elts.ll b/llvm/test/Transforms/InstCombine/amdgcn-demanded-vector-elts.ll
index 888f51bf939dd10d86e36311fb3e96f268fcb9a8..0c4842c159880dd47f2b214e36a3196dcd767714 100644 (file)
--- a/llvm/test/Transforms/InstCombine/amdgcn-demanded-vector-elts.ll
+++ b/llvm/test/Transforms/InstCombine/amdgcn-demanded-vector-elts.ll
@@ -227,6 +227,12 @@ define amdgpu_ps float @preserve_metadata_extract_elt0_buffer_load_v2f32(<4 x i3
   ret float %elt0
 }
 
+declare float @llvm.amdgcn.buffer.load.f32(<4 x i32>, i32, i32, i1, i1) #1
+declare <1 x float> @llvm.amdgcn.buffer.load.v1f32(<4 x i32>, i32, i32, i1, i1) #1
+declare <2 x float> @llvm.amdgcn.buffer.load.v2f32(<4 x i32>, i32, i32, i1, i1) #1
+declare <3 x float> @llvm.amdgcn.buffer.load.v3f32(<4 x i32>, i32, i32, i1, i1) #1
+declare <4 x float> @llvm.amdgcn.buffer.load.v4f32(<4 x i32>, i32, i32, i1, i1) #1
+
 ; --------------------------------------------------------------------
 ; llvm.amdgcn.buffer.load.format
 ; --------------------------------------------------------------------
@@ -304,18 +310,1196 @@ define i16 @extract_lo16_0_bitcast_buffer_load_format_v4f32(i32 %arg) #0 {
   ret i16 %tmp2
 }
 
-declare float @llvm.amdgcn.buffer.load.f32(<4 x i32>, i32, i32, i1, i1) #1
-declare <1 x float> @llvm.amdgcn.buffer.load.v1f32(<4 x i32>, i32, i32, i1, i1) #1
-declare <2 x float> @llvm.amdgcn.buffer.load.v2f32(<4 x i32>, i32, i32, i1, i1) #1
-declare <3 x float> @llvm.amdgcn.buffer.load.v3f32(<4 x i32>, i32, i32, i1, i1) #1
-declare <4 x float> @llvm.amdgcn.buffer.load.v4f32(<4 x i32>, i32, i32, i1, i1) #1
-
 declare float @llvm.amdgcn.buffer.load.format.f32(<4 x i32>, i32, i32, i1, i1) #1
 declare <1 x float> @llvm.amdgcn.buffer.load.format.v1f32(<4 x i32>, i32, i32, i1, i1) #1
 declare <2 x float> @llvm.amdgcn.buffer.load.format.v2f32(<4 x i32>, i32, i32, i1, i1) #1
 declare <3 x float> @llvm.amdgcn.buffer.load.format.v3f32(<4 x i32>, i32, i32, i1, i1) #1
 declare <4 x float> @llvm.amdgcn.buffer.load.format.v4f32(<4 x i32>, i32, i32, i1, i1) #1
 
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+; CHECK-LABEL: @extract_elt0_image_sample_v4f32_v4f32_v4i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.f32.v4f32.v4i32(<4 x float> %vaddr, <4 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_v4f32_v4f32_v4i32(<4 x float> %vaddr, <4 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v4i32(<4 x float> %vaddr, <4 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+; CHECK-LABEL: @extract_elt0_image_sample_v4f32_v2f32_v4i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.f32.v2f32.v8i32(<2 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_v4f32_v2f32_v4i32(<2 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v2f32.v8i32(<2 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+; CHECK-LABEL: @extract_elt0_invalid_dmask_image_sample_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 %dmask, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_invalid_dmask_image_sample_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc, i32 %dmask) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 %dmask, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+; FIXME: Should really fold to undef
+; CHECK-LABEL: @extract_elt0_dmask_0000_image_sample_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 0, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_dmask_0000_image_sample_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 0, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+; CHECK-LABEL: @extract_elt0_dmask_0001_image_sample_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_dmask_0001_image_sample_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+; FIXME: Should really fold to undef
+; CHECK-LABEL: @extract_elt0_dmask_0010_image_sample_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 2, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_dmask_0010_image_sample_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 2, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+; FIXME: Should really fold to undef
+; CHECK-LABEL: @extract_elt0_dmask_0100_image_sample_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 4, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_dmask_0100_image_sample_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 4, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+; FIXME: Should really fold to undef
+; CHECK-LABEL: @extract_elt0_dmask_1000_image_sample_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 8, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_dmask_1000_image_sample_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 8, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+; CHECK-LABEL: @extract_elt0_dmask_1001_image_sample_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_dmask_1001_image_sample_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 9, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+; CHECK-LABEL: @extract_elt0_dmask_0011_image_sample_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_dmask_0011_image_sample_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 3, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+; CHECK-LABEL: @extract_elt0_dmask_0111_image_sample_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_dmask_0111_image_sample_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 7, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+; CHECK-LABEL: @extract_elt0_elt1_dmask_0001_image_sample_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call <2 x float> @llvm.amdgcn.image.sample.v2f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret <2 x float> %data
+define amdgpu_ps <2 x float> @extract_elt0_elt1_dmask_0001_image_sample_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %shuf = shufflevector <4 x float> %data, <4 x float> undef, <2 x i32> <i32 0, i32 1>
+  ret <2 x float> %shuf
+}
+
+; CHECK-LABEL: @extract_elt0_elt1_dmask_0011_image_sample_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call <2 x float> @llvm.amdgcn.image.sample.v2f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 3, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret <2 x float> %data
+define amdgpu_ps <2 x float> @extract_elt0_elt1_dmask_0011_image_sample_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 3, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %shuf = shufflevector <4 x float> %data, <4 x float> undef, <2 x i32> <i32 0, i32 1>
+  ret <2 x float> %shuf
+}
+
+; CHECK-LABEL: @extract_elt0_elt1_dmask_0111_image_sample_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call <2 x float> @llvm.amdgcn.image.sample.v2f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 3, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret <2 x float> %data
+define amdgpu_ps <2 x float> @extract_elt0_elt1_dmask_0111_image_sample_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 7, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %shuf = shufflevector <4 x float> %data, <4 x float> undef, <2 x i32> <i32 0, i32 1>
+  ret <2 x float> %shuf
+}
+
+; CHECK-LABEL: @extract_elt0_elt1_dmask_0101_image_sample_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call <2 x float> @llvm.amdgcn.image.sample.v2f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 5, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret <2 x float> %data
+define amdgpu_ps <2 x float> @extract_elt0_elt1_dmask_0101_image_sample_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 5, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %shuf = shufflevector <4 x float> %data, <4 x float> undef, <2 x i32> <i32 0, i32 1>
+  ret <2 x float> %shuf
+}
+
+; CHECK-LABEL: @extract_elt0_elt1_elt2_dmask_0001_image_sample_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: %shuf = shufflevector <4 x float> %data, <4 x float> undef, <3 x i32> <i32 0, i32 1, i32 2>
+; CHECK-NEXT: ret <3 x float> %shuf
+define amdgpu_ps <3 x float> @extract_elt0_elt1_elt2_dmask_0001_image_sample_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %shuf = shufflevector <4 x float> %data, <4 x float> undef, <3 x i32> <i32 0, i32 1, i32 2>
+  ret <3 x float> %shuf
+}
+
+; CHECK-LABEL: @extract_elt0_elt1_elt2_dmask_0011_image_sample_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 3, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: %shuf = shufflevector <4 x float> %data, <4 x float> undef, <3 x i32> <i32 0, i32 1, i32 2>
+; CHECK-NEXT: ret <3 x float> %shuf
+define amdgpu_ps <3 x float> @extract_elt0_elt1_elt2_dmask_0011_image_sample_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 3, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %shuf = shufflevector <4 x float> %data, <4 x float> undef, <3 x i32> <i32 0, i32 1, i32 2>
+  ret <3 x float> %shuf
+}
+
+; CHECK-LABEL: @extract_elt0_elt1_elt2_dmask_0101_image_sample_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 5, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: %shuf = shufflevector <4 x float> %data, <4 x float> undef, <3 x i32> <i32 0, i32 1, i32 2>
+; CHECK-NEXT: ret <3 x float> %shuf
+define amdgpu_ps <3 x float> @extract_elt0_elt1_elt2_dmask_0101_image_sample_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 5, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %shuf = shufflevector <4 x float> %data, <4 x float> undef, <3 x i32> <i32 0, i32 1, i32 2>
+  ret <3 x float> %shuf
+}
+
+; CHECK-LABEL: @extract_elt0_elt1_elt2_dmask_0111_image_sample_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 7, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: %shuf = shufflevector <4 x float> %data, <4 x float> undef, <3 x i32> <i32 0, i32 1, i32 2>
+; CHECK-NEXT: ret <3 x float> %shuf
+define amdgpu_ps <3 x float> @extract_elt0_elt1_elt2_dmask_0111_image_sample_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 7, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %shuf = shufflevector <4 x float> %data, <4 x float> undef, <3 x i32> <i32 0, i32 1, i32 2>
+  ret <3 x float> %shuf
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+declare <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v4i32(<4 x float>, <4 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+declare <4 x float> @llvm.amdgcn.image.sample.v4f32.v2f32.v8i32(<2 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.cl
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_cl_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.cl.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_cl_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.cl.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.cl.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.d
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_d_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.d.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_d_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.d.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.d.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.d.cl
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_d_cl_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.d.cl.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_d_cl_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.d.cl.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.d.cl.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.l
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_l_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.l.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_l_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.l.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.l.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.b
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_b_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.b.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_b_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.b.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.b.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.b.cl
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_b_cl_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.b.cl.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_b_cl_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.b.cl.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.b.cl.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.lz
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_lz_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.lz.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_lz_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.lz.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.lz.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.cd
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_cd_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.cd.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_cd_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.cd.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.cd.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.cd.cl
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_cd_cl_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.cd.cl.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_cd_cl_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.cd.cl.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.cd.cl.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.c
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_c_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.c.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_c_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.c.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+; CHECK-LABEL: @extract_elt0_image_sample_c_v4f32_v4f32_v4i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.c.f32.v4f32.v4i32(<4 x float> %vaddr, <4 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_c_v4f32_v4f32_v4i32(<4 x float> %vaddr, <4 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.c.v4f32.v4f32.v4i32(<4 x float> %vaddr, <4 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+; CHECK-LABEL: @extract_elt0_image_sample_c_v4f32_v2f32_v4i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.c.f32.v2f32.v8i32(<2 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_c_v4f32_v2f32_v4i32(<2 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.c.v4f32.v2f32.v8i32(<2 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.c.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+declare <4 x float> @llvm.amdgcn.image.sample.c.v4f32.v4f32.v4i32(<4 x float>, <4 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+declare <4 x float> @llvm.amdgcn.image.sample.c.v4f32.v2f32.v8i32(<2 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.c.cl
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_c_cl_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.c.cl.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_c_cl_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.c.cl.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.c.cl.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.c.d
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_c_d_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.c.d.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_c_d_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.c.d.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.c.d.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.c.d.cl
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_c_d_cl_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.c.d.cl.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_c_d_cl_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.c.d.cl.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.c.d.cl.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.c.l
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_c_l_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.c.l.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_c_l_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.c.l.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.c.l.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.c.b
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_c_b_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.c.b.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_c_b_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.c.b.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.c.b.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.c.b.cl
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_c_b_cl_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.c.b.cl.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_c_b_cl_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.c.b.cl.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.c.b.cl.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.c.lz
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_c_lz_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.c.lz.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_c_lz_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.c.lz.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.c.lz.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.c.cd
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_c_cd_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.c.cd.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_c_cd_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.c.cd.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.c.cd.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.c.cd.cl
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_c_cd_cl_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.c.cd.cl.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_c_cd_cl_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.c.cd.cl.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.c.cd.cl.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.o
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_o_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.o.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_o_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+; CHECK-LABEL: @extract_elt0_image_sample_o_v4f32_v4f32_v4i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.o.f32.v4f32.v4i32(<4 x float> %vaddr, <4 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_o_v4f32_v4f32_v4i32(<4 x float> %vaddr, <4 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.o.v4f32.v4f32.v4i32(<4 x float> %vaddr, <4 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+; CHECK-LABEL: @extract_elt0_image_sample_o_v4f32_v2f32_v4i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.o.f32.v2f32.v8i32(<2 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_o_v4f32_v2f32_v4i32(<2 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.o.v4f32.v2f32.v8i32(<2 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.o.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+declare <4 x float> @llvm.amdgcn.image.sample.o.v4f32.v4f32.v4i32(<4 x float>, <4 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+declare <4 x float> @llvm.amdgcn.image.sample.o.v4f32.v2f32.v8i32(<2 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.cl.o
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_cl_o_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.cl.o.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_cl_o_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.cl.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.cl.o.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.d.o
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_d_o_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.d.o.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_d_o_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.d.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.d.o.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.d.cl.o
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_d_cl_o_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.d.cl.o.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_d_cl_o_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.d.cl.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.d.cl.o.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.l.o
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_l_o_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.l.o.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_l_o_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.l.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.l.o.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.b.o
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_b_o_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.b.o.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_b_o_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.b.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.b.o.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.b.cl.o
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_b_cl_o_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.b.cl.o.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_b_cl_o_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.b.cl.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.b.cl.o.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.lz.o
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_lz_o_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.lz.o.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_lz_o_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.lz.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.lz.o.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.cd.o
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_cd_o_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.cd.o.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_cd_o_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.cd.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.cd.o.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.cd.cl.o
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_cd_cl_o_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.cd.cl.o.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_cd_cl_o_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.cd.cl.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.cd.cl.o.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.c.o
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_c_o_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.c.o.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_c_o_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.c.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+; CHECK-LABEL: @extract_elt0_image_sample_c_o_v4f32_v4f32_v4i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.c.o.f32.v4f32.v4i32(<4 x float> %vaddr, <4 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_c_o_v4f32_v4f32_v4i32(<4 x float> %vaddr, <4 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.c.o.v4f32.v4f32.v4i32(<4 x float> %vaddr, <4 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+; CHECK-LABEL: @extract_elt0_image_sample_c_o_v4f32_v2f32_v4i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.c.o.f32.v2f32.v8i32(<2 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_c_o_v4f32_v2f32_v4i32(<2 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.c.o.v4f32.v2f32.v8i32(<2 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.c.o.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+declare <4 x float> @llvm.amdgcn.image.sample.c.o.v4f32.v4f32.v4i32(<4 x float>, <4 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+declare <4 x float> @llvm.amdgcn.image.sample.c.o.v4f32.v2f32.v8i32(<2 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.c.cl.o
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_c_cl_o_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.c.cl.o.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_c_cl_o_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.c.cl.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.c.cl.o.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.c.d.o
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_c_d_o_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.c.d.o.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_c_d_o_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.c.d.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.c.d.o.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.c.d.cl.o
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_c_d_cl_o_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.c.d.cl.o.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_c_d_cl_o_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.c.d.cl.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.c.d.cl.o.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.c.l.o
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_c_l_o_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.c.l.o.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_c_l_o_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.c.l.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.c.l.o.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.c.b.o
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_c_b_o_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.c.b.o.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_c_b_o_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.c.b.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.c.b.o.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.c.b.cl.o
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_c_b_cl_o_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.c.b.cl.o.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_c_b_cl_o_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.c.b.cl.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.c.b.cl.o.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.c.lz.o
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_c_lz_o_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.c.lz.o.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_c_lz_o_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.c.lz.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.c.lz.o.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.c.cd.o
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_c_cd_o_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.c.cd.o.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_c_cd_o_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.c.cd.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.c.cd.o.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.sample.c.cd.cl.o
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_sample_c_cd_cl_o_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.sample.c.cd.cl.o.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_sample_c_cd_cl_o_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.sample.c.cd.cl.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.sample.c.cd.cl.o.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.gather4
+; --------------------------------------------------------------------
+
+; Don't handle gather4*
+
+; CHECK-LABEL: @extract_elt0_image_gather4_v4f32_v4f32_v8i32(
+; CHECK: %data = call <4 x float> @llvm.amdgcn.image.gather4.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i3
+define amdgpu_ps float @extract_elt0_image_gather4_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.gather4.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+; CHECK-LABEL: @extract_elt0_image_gather4_v4f32_v4f32_v4i32(
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.image.gather4.v4f32.v4f32.v4i32(<4 x float> %vaddr, <4 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+define amdgpu_ps float @extract_elt0_image_gather4_v4f32_v4f32_v4i32(<4 x float> %vaddr, <4 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.gather4.v4f32.v4f32.v4i32(<4 x float> %vaddr, <4 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+; CHECK-LABEL: @extract_elt0_image_gather4_v4f32_v2f32_v4i32(
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.image.gather4.v4f32.v2f32.v8i32(<2 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+define amdgpu_ps float @extract_elt0_image_gather4_v4f32_v2f32_v4i32(<2 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.gather4.v4f32.v2f32.v8i32(<2 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.gather4.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+declare <4 x float> @llvm.amdgcn.image.gather4.v4f32.v4f32.v4i32(<4 x float>, <4 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+declare <4 x float> @llvm.amdgcn.image.gather4.v4f32.v2f32.v8i32(<2 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.gather4.cl
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_gather4_cl_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.image.gather4.cl.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+define amdgpu_ps float @extract_elt0_image_gather4_cl_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.gather4.cl.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.gather4.cl.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.gather4.l
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_gather4_l_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.image.gather4.l.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+define amdgpu_ps float @extract_elt0_image_gather4_l_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.gather4.l.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.gather4.l.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.gather4.b
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_gather4_b_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.image.gather4.b.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+define amdgpu_ps float @extract_elt0_image_gather4_b_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.gather4.b.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.gather4.b.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.gather4.b.cl
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_gather4_b_cl_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.image.gather4.b.cl.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+define amdgpu_ps float @extract_elt0_image_gather4_b_cl_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.gather4.b.cl.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.gather4.b.cl.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.gather4.lz
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_gather4_lz_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.image.gather4.lz.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+define amdgpu_ps float @extract_elt0_image_gather4_lz_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.gather4.lz.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.gather4.lz.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.gather4.o
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_gather4_o_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.image.gather4.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %gather4r, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+define amdgpu_ps float @extract_elt0_image_gather4_o_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %gather4r, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.gather4.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %gather4r, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+; CHECK-LABEL: @extract_elt0_image_gather4_o_v4f32_v4f32_v4i32(
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.image.gather4.o.v4f32.v4f32.v4i32(<4 x float> %vaddr, <4 x i32> %gather4r, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+define amdgpu_ps float @extract_elt0_image_gather4_o_v4f32_v4f32_v4i32(<4 x float> %vaddr, <4 x i32> inreg %gather4r, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.gather4.o.v4f32.v4f32.v4i32(<4 x float> %vaddr, <4 x i32> %gather4r, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+; CHECK-LABEL: @extract_elt0_image_gather4_o_v4f32_v2f32_v4i32(
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.image.gather4.o.v4f32.v2f32.v8i32(<2 x float> %vaddr, <8 x i32> %gather4r, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+define amdgpu_ps float @extract_elt0_image_gather4_o_v4f32_v2f32_v4i32(<2 x float> %vaddr, <8 x i32> inreg %gather4r, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.gather4.o.v4f32.v2f32.v8i32(<2 x float> %vaddr, <8 x i32> %gather4r, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.gather4.o.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+declare <4 x float> @llvm.amdgcn.image.gather4.o.v4f32.v4f32.v4i32(<4 x float>, <4 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+declare <4 x float> @llvm.amdgcn.image.gather4.o.v4f32.v2f32.v8i32(<2 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.gather4.cl.o
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_gather4_cl_o_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.image.gather4.cl.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %gather4r, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+define amdgpu_ps float @extract_elt0_image_gather4_cl_o_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %gather4r, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.gather4.cl.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %gather4r, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.gather4.cl.o.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.gather4.l.o
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_gather4_l_o_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.image.gather4.l.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %gather4r, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+define amdgpu_ps float @extract_elt0_image_gather4_l_o_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %gather4r, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.gather4.l.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %gather4r, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.gather4.l.o.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.gather4.b.o
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_gather4_b_o_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.image.gather4.b.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %gather4r, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+define amdgpu_ps float @extract_elt0_image_gather4_b_o_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %gather4r, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.gather4.b.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %gather4r, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.gather4.b.o.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.gather4.b.cl.o
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_gather4_b_cl_o_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.image.gather4.b.cl.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %gather4r, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+define amdgpu_ps float @extract_elt0_image_gather4_b_cl_o_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %gather4r, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.gather4.b.cl.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %gather4r, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.gather4.b.cl.o.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.gather4.lz.o
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_gather4_lz_o_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.image.gather4.lz.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %gather4r, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+define amdgpu_ps float @extract_elt0_image_gather4_lz_o_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %gather4r, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.gather4.lz.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %gather4r, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.gather4.lz.o.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.gather4.c.o
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_gather4_c_o_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.image.gather4.c.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %gather4r, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+define amdgpu_ps float @extract_elt0_image_gather4_c_o_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %gather4r, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.gather4.c.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %gather4r, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+; CHECK-LABEL: @extract_elt0_image_gather4_c_o_v4f32_v4f32_v4i32(
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.image.gather4.c.o.v4f32.v4f32.v4i32(<4 x float> %vaddr, <4 x i32> %gather4r, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+define amdgpu_ps float @extract_elt0_image_gather4_c_o_v4f32_v4f32_v4i32(<4 x float> %vaddr, <4 x i32> inreg %gather4r, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.gather4.c.o.v4f32.v4f32.v4i32(<4 x float> %vaddr, <4 x i32> %gather4r, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+; CHECK-LABEL: @extract_elt0_image_gather4_c_o_v4f32_v2f32_v4i32(
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.image.gather4.c.o.v4f32.v2f32.v8i32(<2 x float> %vaddr, <8 x i32> %gather4r, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+define amdgpu_ps float @extract_elt0_image_gather4_c_o_v4f32_v2f32_v4i32(<2 x float> %vaddr, <8 x i32> inreg %gather4r, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.gather4.c.o.v4f32.v2f32.v8i32(<2 x float> %vaddr, <8 x i32> %gather4r, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.gather4.c.o.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+declare <4 x float> @llvm.amdgcn.image.gather4.c.o.v4f32.v4f32.v4i32(<4 x float>, <4 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+declare <4 x float> @llvm.amdgcn.image.gather4.c.o.v4f32.v2f32.v8i32(<2 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.gather4.c.cl.o
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_gather4_c_cl_o_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.image.gather4.c.cl.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %gather4r, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+define amdgpu_ps float @extract_elt0_image_gather4_c_cl_o_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %gather4r, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.gather4.c.cl.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %gather4r, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.gather4.c.cl.o.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.gather4.c.l.o
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_gather4_c_l_o_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.image.gather4.c.l.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %gather4r, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+define amdgpu_ps float @extract_elt0_image_gather4_c_l_o_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %gather4r, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.gather4.c.l.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %gather4r, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.gather4.c.l.o.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.gather4.c.b.o
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_gather4_c_b_o_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.image.gather4.c.b.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %gather4r, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+define amdgpu_ps float @extract_elt0_image_gather4_c_b_o_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %gather4r, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.gather4.c.b.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %gather4r, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.gather4.c.b.o.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.gather4.c.b.cl.o
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_gather4_c_b_cl_o_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.image.gather4.c.b.cl.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %gather4r, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+define amdgpu_ps float @extract_elt0_image_gather4_c_b_cl_o_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %gather4r, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.gather4.c.b.cl.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %gather4r, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.gather4.c.b.cl.o.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.gather4.c.lz.o
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_gather4_c_lz_o_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call <4 x float> @llvm.amdgcn.image.gather4.c.lz.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %gather4r, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+define amdgpu_ps float @extract_elt0_image_gather4_c_lz_o_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %gather4r, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.gather4.c.lz.o.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %gather4r, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.gather4.c.lz.o.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
+; --------------------------------------------------------------------
+; llvm.amdgcn.image.getlod
+; --------------------------------------------------------------------
+
+; CHECK-LABEL: @extract_elt0_image_getlod_v4f32_v4f32_v8i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.getlod.f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_getlod_v4f32_v4f32_v8i32(<4 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.getlod.v4f32.v4f32.v8i32(<4 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+; CHECK-LABEL: @extract_elt0_image_getlod_v4f32_v4f32_v4i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.getlod.f32.v4f32.v4i32(<4 x float> %vaddr, <4 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_getlod_v4f32_v4f32_v4i32(<4 x float> %vaddr, <4 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.getlod.v4f32.v4f32.v4i32(<4 x float> %vaddr, <4 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+; CHECK-LABEL: @extract_elt0_image_getlod_v4f32_v2f32_v4i32(
+; CHECK-NEXT: %data = call float @llvm.amdgcn.image.getlod.f32.v2f32.v8i32(<2 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 1, i1 false, i1 false, i1 false, i1 false, i1 false)
+; CHECK-NEXT: ret float %data
+define amdgpu_ps float @extract_elt0_image_getlod_v4f32_v2f32_v4i32(<2 x float> %vaddr, <8 x i32> inreg %sampler, <4 x i32> inreg %rsrc) #0 {
+  %data = call <4 x float> @llvm.amdgcn.image.getlod.v4f32.v2f32.v8i32(<2 x float> %vaddr, <8 x i32> %sampler, <4 x i32> %rsrc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %elt0 = extractelement <4 x float> %data, i32 0
+  ret float %elt0
+}
+
+declare <4 x float> @llvm.amdgcn.image.getlod.v4f32.v4f32.v8i32(<4 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+declare <4 x float> @llvm.amdgcn.image.getlod.v4f32.v4f32.v4i32(<4 x float>, <4 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+declare <4 x float> @llvm.amdgcn.image.getlod.v4f32.v2f32.v8i32(<2 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #1
+
 attributes #0 = { nounwind }
 attributes #1 = { nounwind readonly }