vendor/llvm/llvm-trunk-r300890

10 files changed, 1354 insertions, 29 deletions

diff --git a/test/Transforms/InstCombine/2008-01-13-NoBitCastAttributes.ll b/test/Transforms/InstCombine/2008-01-13-NoBitCastAttributes.ll
deleted file mode 100644
index 510a68c3437e..000000000000
--- a/test/Transforms/InstCombine/2008-01-13-NoBitCastAttributes.ll
+++ /dev/null
@@ -1,15 +0,0 @@
-; RUN: opt < %s -instcombine -S | grep bitcast | count 2
-
-define signext i32 @b(i32* inreg  %x)   {
-	ret i32 0
-}
-
-define void @c(...) {
-	ret void
-}
-
-define void @g(i32* %y) {
-	call i32 bitcast (i32 (i32*)* @b to i32 (i32)*)( i32 zeroext  0 )		; <i32>:2 [#uses=0]
-	call void bitcast (void (...)* @c to void (i32*)*)( i32* sret  null )
-	ret void
-}
diff --git a/test/Transforms/InstCombine/amdgcn-demanded-vector-elts.ll b/test/Transforms/InstCombine/amdgcn-demanded-vector-elts.ll
index 888f51bf939d..0c4842c15988 100644
--- a/test/Transforms/InstCombine/amdgcn-demanded-vector-elts.ll
+++ b/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 }
 
diff --git a/test/Transforms/InstCombine/call-cast-attrs.ll b/test/Transforms/InstCombine/call-cast-attrs.ll
new file mode 100644
index 000000000000..ddaf90c3e74f
--- /dev/null
+++ b/test/Transforms/InstCombine/call-cast-attrs.ll
@@ -0,0 +1,29 @@
+; RUN: opt < %s -instcombine -S | FileCheck %s
+
+define signext i32 @b(i32* inreg %x)   {
+  ret i32 0
+}
+
+define void @c(...) {
+  ret void
+}
+
+declare void @useit(i32)
+
+define void @d(i32 %x, ...) {
+  call void @useit(i32 %x)
+  ret void
+}
+
+define void @g(i32* %y) {
+  call i32 bitcast (i32 (i32*)* @b to i32 (i32)*)(i32 zeroext 0)
+  call void bitcast (void (...)* @c to void (i32*)*)(i32* %y)
+  call void bitcast (void (...)* @c to void (i32*)*)(i32* sret %y)
+  call void bitcast (void (i32, ...)* @d to void (i32, i32*)*)(i32 0, i32* sret %y)
+  ret void
+}
+; CHECK-LABEL: define void @g(i32* %y)
+; CHECK: call i32 bitcast (i32 (i32*)* @b to i32 (i32)*)(i32 zeroext 0)
+; CHECK: call void (...) @c(i32* %y)
+; CHECK: call void bitcast (void (...)* @c to void (i32*)*)(i32* sret %y)
+; CHECK: call void bitcast (void (i32, ...)* @d to void (i32, i32*)*)(i32 0, i32* sret %y)
diff --git a/test/Transforms/InstCombine/constant-fold-math.ll b/test/Transforms/InstCombine/constant-fold-math.ll
index 50cd6070896e..27578387f827 100644
--- a/test/Transforms/InstCombine/constant-fold-math.ll
+++ b/test/Transforms/InstCombine/constant-fold-math.ll
@@ -45,4 +45,22 @@ define double @constant_fold_fmuladd_f64() #0 {
   ret double %x
 }
 
+; PR32177
+
+; CHECK-LABEL: @constant_fold_frem_f32
+; CHECK-NEXT: ret float 0x41A61B2000000000
+define float @constant_fold_frem_f32() #0 {
+  %x = frem float 0x43cbfcd960000000, 0xc1e2b34a00000000
+  ret float %x
+}
+
+; PR3316
+
+; CHECK-LABEL: @constant_fold_frem_f64
+; CHECK-NEXT: ret double 0.000000e+00
+define double @constant_fold_frem_f64() {
+  %x = frem double 0x43E0000000000000, 1.000000e+00
+  ret double %x
+}
+
 attributes #0 = { nounwind readnone }
diff --git a/test/Transforms/InstCombine/div-shift.ll b/test/Transforms/InstCombine/div-shift.ll
index 517313ed8e4e..b5a65048fda0 100644
--- a/test/Transforms/InstCombine/div-shift.ll
+++ b/test/Transforms/InstCombine/div-shift.ll
@@ -16,6 +16,21 @@ entry:
   ret i32 %d
 }
 
+define <2 x i32> @t1vec(<2 x i16> %x, <2 x i32> %y) {
+; CHECK-LABEL: @t1vec(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[CONV:%.*]] = zext <2 x i16> [[X:%.*]] to <2 x i32>
+; CHECK-NEXT:    [[TMP0:%.*]] = add <2 x i32> [[Y:%.*]], <i32 1, i32 1>
+; CHECK-NEXT:    [[D:%.*]] = lshr <2 x i32> [[CONV]], [[TMP0]]
+; CHECK-NEXT:    ret <2 x i32> [[D]]
+;
+entry:
+  %conv = zext <2 x i16> %x to <2 x i32>
+  %s = shl <2 x i32> <i32 2, i32 2>, %y
+  %d = sdiv <2 x i32> %conv, %s
+  ret <2 x i32> %d
+}
+
 ; rdar://11721329
 define i64 @t2(i64 %x, i32 %y) {
 ; CHECK-LABEL: @t2(
diff --git a/test/Transforms/InstCombine/div.ll b/test/Transforms/InstCombine/div.ll
index a037607267ac..796fce020fd3 100644
--- a/test/Transforms/InstCombine/div.ll
+++ b/test/Transforms/InstCombine/div.ll
@@ -225,6 +225,16 @@ define i32 @test19(i32 %x) {
   ret i32 %A
 }
 
+define <2 x i32> @test19vec(<2 x i32> %x) {
+; CHECK-LABEL: @test19vec(
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp eq <2 x i32> [[X:%.*]], <i32 1, i32 1>
+; CHECK-NEXT:    [[A:%.*]] = zext <2 x i1> [[TMP1]] to <2 x i32>
+; CHECK-NEXT:    ret <2 x i32> [[A]]
+;
+  %A = udiv <2 x i32> <i32 1, i32 1>, %x
+  ret <2 x i32> %A
+}
+
 define i32 @test20(i32 %x) {
 ; CHECK-LABEL: @test20(
 ; CHECK-NEXT:    [[TMP1:%.*]] = add i32 %x, 1
@@ -236,6 +246,17 @@ define i32 @test20(i32 %x) {
   ret i32 %A
 }
 
+define <2 x i32> @test20vec(<2 x i32> %x) {
+; CHECK-LABEL: @test20vec(
+; CHECK-NEXT:    [[TMP1:%.*]] = add <2 x i32> [[X:%.*]], <i32 1, i32 1>
+; CHECK-NEXT:    [[TMP2:%.*]] = icmp ult <2 x i32> [[TMP1]], <i32 3, i32 3>
+; CHECK-NEXT:    [[A:%.*]] = select <2 x i1> [[TMP2]], <2 x i32> [[X]], <2 x i32> zeroinitializer
+; CHECK-NEXT:    ret <2 x i32> [[A]]
+;
+  %A = sdiv <2 x i32> <i32 1, i32 1>, %x
+  ret <2 x i32> %A
+}
+
 define i32 @test21(i32 %a) {
 ; CHECK-LABEL: @test21(
 ; CHECK-NEXT:    [[DIV:%.*]] = sdiv i32 %a, 3
@@ -388,6 +409,17 @@ define i32 @test35(i32 %A) {
   ret i32 %mul
 }
 
+define <2 x i32> @test35vec(<2 x i32> %A) {
+; CHECK-LABEL: @test35vec(
+; CHECK-NEXT:    [[AND:%.*]] = and <2 x i32> [[A:%.*]], <i32 2147483647, i32 2147483647>
+; CHECK-NEXT:    [[MUL:%.*]] = udiv exact <2 x i32> [[AND]], <i32 2147483647, i32 2147483647>
+; CHECK-NEXT:    ret <2 x i32> [[MUL]]
+;
+  %and = and <2 x i32> %A, <i32 2147483647, i32 2147483647>
+  %mul = sdiv exact <2 x i32> %and, <i32 2147483647, i32 2147483647>
+  ret <2 x i32> %mul
+}
+
 define i32 @test36(i32 %A) {
 ; CHECK-LABEL: @test36(
 ; CHECK-NEXT:    [[AND:%.*]] = and i32 %A, 2147483647
@@ -400,13 +432,10 @@ define i32 @test36(i32 %A) {
   ret i32 %mul
 }
 
-; FIXME: Vector should get same transform as scalar.
-
 define <2 x i32> @test36vec(<2 x i32> %A) {
 ; CHECK-LABEL: @test36vec(
-; CHECK-NEXT:    [[AND:%.*]] = and <2 x i32> %A, <i32 2147483647, i32 2147483647>
-; CHECK-NEXT:    [[SHL:%.*]] = shl nuw nsw <2 x i32> <i32 1, i32 1>, %A
-; CHECK-NEXT:    [[MUL:%.*]] = sdiv exact <2 x i32> [[AND]], [[SHL]]
+; CHECK-NEXT:    [[AND:%.*]] = and <2 x i32> [[A:%.*]], <i32 2147483647, i32 2147483647>
+; CHECK-NEXT:    [[MUL:%.*]] = lshr exact <2 x i32> [[AND]], [[A]]
 ; CHECK-NEXT:    ret <2 x i32> [[MUL]]
 ;
   %and = and <2 x i32> %A, <i32 2147483647, i32 2147483647>
diff --git a/test/Transforms/InstCombine/pr32686.ll b/test/Transforms/InstCombine/pr32686.ll
new file mode 100644
index 000000000000..b2d2aff2fde8
--- /dev/null
+++ b/test/Transforms/InstCombine/pr32686.ll
@@ -0,0 +1,23 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt -S -instcombine %s | FileCheck %s
+
+@a = external global i8
+@b = external global i32
+
+define void @tinkywinky() {
+; CHECK-LABEL: @tinkywinky(
+; CHECK-NEXT:    [[PATATINO:%.*]] = load i8, i8* @a, align 1
+; CHECK-NEXT:    [[TOBOOL:%.*]] = icmp eq i8 [[PATATINO]], 0
+; CHECK-NEXT:    [[TMP1:%.*]] = zext i1 [[TOBOOL]] to i32
+; CHECK-NEXT:    [[OR1:%.*]] = or i32 [[TMP1]], or (i32 zext (i1 icmp ne (i32* bitcast (i8* @a to i32*), i32* @b) to i32), i32 2)
+; CHECK-NEXT:    store i32 [[OR1]], i32* @b, align 4
+; CHECK-NEXT:    ret void
+;
+  %patatino = load i8, i8* @a
+  %tobool = icmp ne i8 %patatino, 0
+  %lnot = xor i1 %tobool, true
+  %lnot.ext = zext i1 %lnot to i32
+  %or = or i32 xor (i32 zext (i1 icmp ne (i32* bitcast (i8* @a to i32*), i32* @b) to i32), i32 2), %lnot.ext
+  store i32 %or, i32* @b, align 4
+  ret void
+}
diff --git a/test/Transforms/InstCombine/rem.ll b/test/Transforms/InstCombine/rem.ll
index 7a7a134db9c5..86a3580189fd 100644
--- a/test/Transforms/InstCombine/rem.ll
+++ b/test/Transforms/InstCombine/rem.ll
@@ -1,3 +1,4 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt < %s -instcombine -S | FileCheck %s
 
 define i64 @rem_signed(i64 %x1, i64 %y2) {
@@ -571,3 +572,24 @@ rem.is.unsafe:
   ret i32 0
 }
 
+define i32 @test22(i32 %A) {
+; CHECK-LABEL: @test22(
+; CHECK-NEXT:    [[AND:%.*]] = and i32 [[A:%.*]], 2147483647
+; CHECK-NEXT:    [[MUL:%.*]] = urem i32 [[AND]], 2147483647
+; CHECK-NEXT:    ret i32 [[MUL]]
+;
+  %and = and i32 %A, 2147483647
+  %mul = srem i32 %and, 2147483647
+  ret i32 %mul
+}
+
+define <2 x i32> @test23(<2 x i32> %A) {
+; CHECK-LABEL: @test23(
+; CHECK-NEXT:    [[AND:%.*]] = and <2 x i32> [[A:%.*]], <i32 2147483647, i32 2147483647>
+; CHECK-NEXT:    [[MUL:%.*]] = urem <2 x i32> [[AND]], <i32 2147483647, i32 2147483647>
+; CHECK-NEXT:    ret <2 x i32> [[MUL]]
+;
+  %and = and <2 x i32> %A, <i32 2147483647, i32 2147483647>
+  %mul = srem <2 x i32> %and, <i32 2147483647, i32 2147483647>
+  ret <2 x i32> %mul
+}
diff --git a/test/Transforms/InstCombine/shift.ll b/test/Transforms/InstCombine/shift.ll
index 60ba35557f70..d5f489280a03 100644
--- a/test/Transforms/InstCombine/shift.ll
+++ b/test/Transforms/InstCombine/shift.ll
@@ -1268,3 +1268,23 @@ define <2 x i64> @test_64_splat_vec(<2 x i32> %t) {
   ret <2 x i64> %shl
 }
 
+define <2 x i8> @ashr_demanded_bits_splat(<2 x i8> %x) {
+; CHECK-LABEL: @ashr_demanded_bits_splat(
+; CHECK-NEXT:    [[SHR:%.*]] = ashr <2 x i8> %x, <i8 7, i8 7>
+; CHECK-NEXT:    ret <2 x i8> [[SHR]]
+;
+  %and = and <2 x i8> %x, <i8 128, i8 128>
+  %shr = ashr <2 x i8> %and, <i8 7, i8 7>
+  ret <2 x i8> %shr
+}
+
+define <2 x i8> @lshr_demanded_bits_splat(<2 x i8> %x) {
+; CHECK-LABEL: @lshr_demanded_bits_splat(
+; CHECK-NEXT:    [[SHR:%.*]] = lshr <2 x i8> %x, <i8 7, i8 7>
+; CHECK-NEXT:    ret <2 x i8> [[SHR]]
+;
+  %and = and <2 x i8> %x, <i8 128, i8 128>
+  %shr = lshr <2 x i8> %and, <i8 7, i8 7>
+  ret <2 x i8> %shr
+}
+
diff --git a/test/Transforms/InstCombine/vector-casts.ll b/test/Transforms/InstCombine/vector-casts.ll
index 643ab6c5348f..2197c250ace2 100644
--- a/test/Transforms/InstCombine/vector-casts.ll
+++ b/test/Transforms/InstCombine/vector-casts.ll
@@ -15,9 +15,9 @@ define <2 x i1> @test1(<2 x i64> %a) {
 ; The ashr turns into an lshr.
 define <2 x i64> @test2(<2 x i64> %a) {
 ; CHECK-LABEL: @test2(
-; CHECK-NEXT:    [[B:%.*]] = and <2 x i64> %a, <i64 65535, i64 65535>
-; CHECK-NEXT:    [[T:%.*]] = lshr <2 x i64> [[B]], <i64 1, i64 1>
-; CHECK-NEXT:    ret <2 x i64> [[T]]
+; CHECK-NEXT:    [[B:%.*]] = and <2 x i64> %a, <i64 65534, i64 65534>
+; CHECK-NEXT:    [[TMP1:%.*]] = lshr exact <2 x i64> [[B]], <i64 1, i64 1>
+; CHECK-NEXT:    ret <2 x i64> [[TMP1]]
 ;
   %b = and <2 x i64> %a, <i64 65535, i64 65535>
   %t = ashr <2 x i64> %b, <i64 1, i64 1>