AMDGPU: Convert image intrinsic uses in tests

[mareko]

Acked.

[arsenm]

r298386

test/CodeGen/AMDGPU/coalescer-subrange-crash.ll

main_body:
   %tmp26 = insertelement <16 x i32> <i32 212739, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>, i32 %tmp25, i32 1
   %tmp27 = insertelement <16 x i32> %tmp26, i32 undef, i32 2
   %tmp28 = insertelement <16 x i32> %tmp27, i32 undef, i32 3
   %tmp29 = insertelement <16 x i32> %tmp28, i32 undef, i32 4
   %tmp30 = insertelement <16 x i32> %tmp29, i32 0, i32 5
   %tmp31 = insertelement <16 x i32> %tmp30, i32 undef, i32 6
   %tmp32 = insertelement <16 x i32> %tmp31, i32 undef, i32 7
   %tmp33 = insertelement <16 x i32> %tmp32, i32 undef, i32 8
-  %tmp34 = call <4 x float> @llvm.SI.image.sample.c.d.o.v16i32(<16 x i32> %tmp33, <8 x i32> undef, <4 x i32> undef, i32 15, i32 0, i32 0, i32 1, i32 0, i32 0, i32 0, i32 0)
+  %tmp33.bc = bitcast <16 x i32> %tmp33 to <16 x float>
+  %tmp34 = call <4 x float> @llvm.amdgcn.image.sample.c.d.o.v4f32.v16f32.v8i32(<16 x float> %tmp33.bc, <8 x i32> undef, <4 x i32> undef, i32 15, i1 false, i1 false, i1 false, i1 false, i1 true)
   %tmp35 = extractelement <4 x float> %tmp34, i32 0
   %tmp36 = bitcast float %tmp24 to i32
   %tmp37 = insertelement <16 x i32> <i32 212739, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>, i32 %tmp36, i32 1
   %tmp38 = insertelement <16 x i32> %tmp37, i32 undef, i32 2
   %tmp39 = insertelement <16 x i32> %tmp38, i32 undef, i32 3
   %tmp40 = insertelement <16 x i32> %tmp39, i32 undef, i32 4
   %tmp41 = insertelement <16 x i32> %tmp40, i32 0, i32 5
   %tmp42 = insertelement <16 x i32> %tmp41, i32 undef, i32 6
   %tmp43 = insertelement <16 x i32> %tmp42, i32 undef, i32 7
   %tmp44 = insertelement <16 x i32> %tmp43, i32 undef, i32 8
-  %tmp45 = call <4 x float> @llvm.SI.image.sample.c.d.o.v16i32(<16 x i32> %tmp44, <8 x i32> undef, <4 x i32> undef, i32 15, i32 0, i32 0, i32 1, i32 0, i32 0, i32 0, i32 0)
+  %tmp44.bc = bitcast <16 x i32> %tmp44 to <16 x float>
+  %tmp45 = call <4 x float> @llvm.amdgcn.image.sample.c.d.o.v4f32.v16f32.v8i32(<16 x float> %tmp44.bc, <8 x i32> undef, <4 x i32> undef, i32 15, i1 false, i1 false, i1 false, i1 false, i1 true)
   %tmp46 = extractelement <4 x float> %tmp45, i32 0
   %tmp47 = fmul float %tmp35, %tmp46
   %tmp48 = insertvalue <{ i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, float, float, float, float, float, float, float, float, float, float, float, float, float, float }> undef, float %tmp47, 14
   %tmp49 = insertvalue <{ i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, float, float, float, float, float, float, float, float, float, float, float, float, float, float }> %tmp48, float %arg21, 24
   ret <{ i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, float, float, float, float, float, float, float, float, float, float, float, float, float, float }> %tmp49
 }
 
-; Function Attrs: nounwind readnone
-declare float @llvm.SI.load.const(<16 x i8>, i32) #0
-
-; Function Attrs: nounwind readnone
-declare <4 x float> @llvm.SI.image.sample.c.d.o.v16i32(<16 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #0
-
-; Function Attrs: nounwind readnone
-declare float @llvm.amdgcn.interp.p1(float, i32, i32, i32) #0
-
-; Function Attrs: nounwind readnone
-declare float @llvm.amdgcn.interp.p2(float, float, i32, i32, i32) #0
-
-; Function Attrs: nounwind readnone
-declare float @llvm.amdgcn.interp.mov(i32, i32, i32, i32) #0
-
-attributes #0 = { nounwind readnone }
-attributes #1 = { nounwind }
+declare float @llvm.amdgcn.interp.p1(float, i32, i32, i32) #1
+declare float @llvm.amdgcn.interp.p2(float, float, i32, i32, i32) #1
+declare <4 x float> @llvm.amdgcn.image.sample.c.d.o.v4f32.v16f32.v8i32(<16 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #2
+
+attributes #0 = { nounwind }
+attributes #1 = { nounwind readnone }
+attributes #2 = { nounwind readonly }

test/CodeGen/AMDGPU/commute-shifts.ll

 ; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
 ; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
 
 ; GCN-LABEL: {{^}}main:
 ; SI: v_lshl_b32_e32 v{{[0-9]+}}, 1, v{{[0-9]+}}
 ; VI: v_lshlrev_b32_e64 v{{[0-9]+}}, v{{[0-9]+}}, 1
 define amdgpu_ps float @main(float %arg0, float %arg1) #0 {
 bb:
   %tmp = fptosi float %arg0 to i32
-  %tmp1 = call <4 x float> @llvm.SI.image.load.v4i32(<4 x i32> undef, <8 x i32> undef, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %tmp1 = call <4 x float> @llvm.amdgcn.image.load.v4f32.v4i32.v8i32(<4 x i32> undef, <8 x i32> undef, i32 15, i1 false, i1 false, i1 false, i1 false)
   %tmp2.f = extractelement <4 x float> %tmp1, i32 0
   %tmp2 = bitcast float %tmp2.f to i32
   %tmp3 = and i32 %tmp, 7
   %tmp4 = shl i32 1, %tmp3
   %tmp5 = and i32 %tmp2, %tmp4
   %tmp6 = icmp eq i32 %tmp5, 0
   %tmp7 = select i1 %tmp6, float 0.000000e+00, float %arg1
   %tmp8 = call <2 x half> @llvm.amdgcn.cvt.pkrtz(float undef, float %tmp7)
   %tmp9 = bitcast <2 x half> %tmp8 to float
   ret float %tmp9
 }
 
-declare <4 x float> @llvm.SI.image.load.v4i32(<4 x i32>, <8 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #1
 declare <2 x half> @llvm.amdgcn.cvt.pkrtz(float, float) #1
+declare <4 x float> @llvm.amdgcn.image.load.v4f32.v4i32.v8i32(<4 x i32>, <8 x i32>, i32, i1, i1, i1, i1) #2
 
 attributes #0 = { nounwind }
 attributes #1 = { nounwind readnone }
+attributes #2 = { nounwind readonly }

test/CodeGen/AMDGPU/else.ll

-; RUN: llc < %s -march=amdgcn -mcpu=verde -verify-machineinstrs | FileCheck %s
-; RUN: llc < %s -march=amdgcn -mcpu=tonga -verify-machineinstrs | FileCheck %s
+; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck %s
+; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck %s
 
 ; CHECK-LABEL: {{^}}else_no_execfix:
 ; CHECK: ; %Flow
 ; CHECK-NEXT: s_or_saveexec_b64 [[DST:s\[[0-9]+:[0-9]+\]]],
 ; CHECK-NEXT: s_xor_b64 exec, exec, [[DST]]
 ; CHECK-NEXT: ; mask branch
-define amdgpu_ps float @else_no_execfix(i32 %z, float %v) {
+define amdgpu_ps float @else_no_execfix(i32 %z, float %v) #0 {
 main_body:
   %cc = icmp sgt i32 %z, 5
   br i1 %cc, label %if, label %else
 
 if:
   %v.if = fmul float %v, 2.0
   br label %end
 

 ; CHECK: ; BB#0:
 ; CHECK-NEXT: s_mov_b64 [[INIT_EXEC:s\[[0-9]+:[0-9]+\]]], exec
 ; CHECK: ; %Flow
 ; CHECK-NEXT: s_or_saveexec_b64 [[DST:s\[[0-9]+:[0-9]+\]]],
 ; CHECK-NEXT: s_and_b64 exec, exec, [[INIT_EXEC]]
 ; CHECK-NEXT: s_and_b64 [[AND_INIT:s\[[0-9]+:[0-9]+\]]], exec, [[DST]]
 ; CHECK-NEXT: s_xor_b64 exec, exec, [[AND_INIT]]
 ; CHECK-NEXT: ; mask branch
-define amdgpu_ps void @else_execfix_leave_wqm(i32 %z, float %v) {
+define amdgpu_ps void @else_execfix_leave_wqm(i32 %z, float %v) #0 {
 main_body:
   %cc = icmp sgt i32 %z, 5
   br i1 %cc, label %if, label %else
 
 if:
   %v.if = fmul float %v, 2.0
   br label %end
 
 else:
   %c = fmul float %v, 3.0
-  %c.i = bitcast float %c to i32
-  %tex = call <4 x float> @llvm.SI.image.sample.i32(i32 %c.i, <8 x i32> undef, <4 x i32> undef, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %tex = call <4 x float> @llvm.amdgcn.image.sample.v4f32.f32.v8i32(float %c, <8 x i32> undef, <4 x i32> undef, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
   %v.else = extractelement <4 x float> %tex, i32 0
   br label %end
 
 end:
   %r = phi float [ %v.if, %if ], [ %v.else, %else ]
   call void @llvm.amdgcn.buffer.store.f32(float %r, <4 x i32> undef, i32 0, i32 0, i1 0, i1 0)
   ret void
 }
 
-declare void @llvm.amdgcn.buffer.store.f32(float, <4 x i32>, i32, i32, i1, i1) nounwind
+declare void @llvm.amdgcn.buffer.store.f32(float, <4 x i32>, i32, i32, i1, i1) #1
+declare <4 x float> @llvm.amdgcn.image.sample.v4f32.f32.v8i32(float, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #2
 
-declare <4 x float> @llvm.SI.image.sample.i32(i32, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) nounwind readnone
+attributes #0 = { nounwind }
+attributes #1 = { nounwind writeonly }
+attributes #2 = { nounwind readonly }

test/CodeGen/AMDGPU/llvm.amdgcn.ps.live.ll

-; RUN: llc -march=amdgcn -mcpu=SI -verify-machineinstrs < %s | FileCheck -check-prefix=CHECK %s
+; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=CHECK %s
 
 ; CHECK-LABEL: {{^}}test1:
 ; CHECK: v_cndmask_b32_e64 v0, 0, 1, exec
 ;
 ; Note: We could generate better code here if we recognized earlier that
 ; there is no WQM use and therefore llvm.amdgcn.ps.live is constant. However,
 ; the expectation is that the intrinsic will be used in non-trivial shaders,
 ; so such an optimization doesn't seem worth the effort.
-define amdgpu_ps float @test1() {
+define amdgpu_ps float @test1() #0 {
   %live = call i1 @llvm.amdgcn.ps.live()
   %live.32 = zext i1 %live to i32
   %r = bitcast i32 %live.32 to float
   ret float %r
 }
 
 ; CHECK-LABEL: {{^}}test2:
 ; CHECK: s_mov_b64 [[LIVE:s\[[0-9]+:[0-9]+\]]], exec
 ; CHECK-DAG: s_wqm_b64 exec, exec
 ; CHECK-DAG: v_cndmask_b32_e64 [[VAR:v[0-9]+]], 0, 1, [[LIVE]]
 ; CHECK: image_sample v0, [[VAR]],
-define amdgpu_ps float @test2() {
+define amdgpu_ps float @test2() #0 {
   %live = call i1 @llvm.amdgcn.ps.live()
   %live.32 = zext i1 %live to i32
-
-  %t = call <4 x float> @llvm.SI.image.sample.i32(i32 %live.32, <8 x i32> undef, <4 x i32> undef, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
-
+  %live.32.bc = bitcast i32 %live.32 to float
+  %t = call <4 x float> @llvm.amdgcn.image.sample.v4f32.f32.v8i32(float %live.32.bc, <8 x i32> undef, <4 x i32> undef, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
   %r = extractelement <4 x float> %t, i32 0
   ret float %r
 }
 
 ; CHECK-LABEL: {{^}}test3:
 ; CHECK: s_mov_b64 [[LIVE:s\[[0-9]+:[0-9]+\]]], exec
 ; CHECK-DAG: s_wqm_b64 exec, exec
 ; CHECK-DAG: s_xor_b64 [[HELPER:s\[[0-9]+:[0-9]+\]]], [[LIVE]], -1
 ; CHECK_DAG: s_and_saveexec_b64 [[SAVED:s\[[0-9]+:[0-9]+\]]], [[HELPER]]
 ; CHECK: ; %dead
-define amdgpu_ps float @test3(i32 %in) {
+define amdgpu_ps float @test3(i32 %in) #0 {
 entry:
   %live = call i1 @llvm.amdgcn.ps.live()
   br i1 %live, label %end, label %dead
 
 dead:
   %tc.dead = mul i32 %in, 2
   br label %end
 
 end:
   %tc = phi i32 [ %in, %entry ], [ %tc.dead, %dead ]
-  %t = call <4 x float> @llvm.SI.image.sample.i32(i32 %tc, <8 x i32> undef, <4 x i32> undef, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
-
+  %tc.bc = bitcast i32 %tc to float
+  %t = call <4 x float> @llvm.amdgcn.image.sample.v4f32.f32.v8i32(float %tc.bc, <8 x i32> undef, <4 x i32> undef, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false) #0
   %r = extractelement <4 x float> %t, i32 0
   ret float %r
 }
 
-declare i1 @llvm.amdgcn.ps.live() #0
-
-declare <4 x float> @llvm.SI.image.sample.i32(i32, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #0
+declare i1 @llvm.amdgcn.ps.live() #1
+declare <4 x float> @llvm.amdgcn.image.sample.v4f32.f32.v8i32(float, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #2
 
-attributes #0 = { nounwind readnone }
+attributes #0 = { nounwind }
+attributes #1 = { nounwind readnone }
+attributes #2 = { nounwind readonly }

test/CodeGen/AMDGPU/sgpr-copy.ll

main_body:
   %j.f.i4 = bitcast i32 %j.i2 to float
   %p1.i5 = call float @llvm.amdgcn.interp.p1(float %i.f.i3, i32 2, i32 1, i32 %arg3) #1
   %p2.i6 = call float @llvm.amdgcn.interp.p2(float %p1.i5, float %j.f.i4, i32 2, i32 1, i32 %arg3) #1
   %tmp45 = bitcast float %p2.i to i32
   %tmp46 = bitcast float %p2.i24 to i32
   %tmp47 = insertelement <2 x i32> undef, i32 %tmp45, i32 0
   %tmp48 = insertelement <2 x i32> %tmp47, i32 %tmp46, i32 1
   %tmp39.bc = bitcast <16 x i8> %tmp39 to <4 x i32>
-  %tmp49 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %tmp48, <8 x i32> %tmp37, <4 x i32> %tmp39.bc, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
-  %tmp50 = extractelement <4 x float> %tmp49, i32 2
+  %a.bc.i = bitcast <2 x i32> %tmp48 to <2 x float>
+  %tmp1 = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v2f32.v8i32(<2 x float> %a.bc.i, <8 x i32> %tmp37, <4 x i32> %tmp39.bc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %tmp50 = extractelement <4 x float> %tmp1, i32 2
   %tmp51 = call float @llvm.fabs.f32(float %tmp50)
   %tmp52 = fmul float %p2.i18, %p2.i18
   %tmp53 = fmul float %p2.i12, %p2.i12
   %tmp54 = fadd float %tmp53, %tmp52
   %tmp55 = fmul float %p2.i6, %p2.i6
   %tmp56 = fadd float %tmp54, %tmp55
   %tmp57 = call float @llvm.amdgcn.rsq.f32(float %tmp56)
   %tmp58 = fmul float %p2.i18, %tmp57
entry:
   %tmp24 = load <8 x i32>, <8 x i32> addrspace(2)* %tmp23, !tbaa !0
   %tmp25 = getelementptr [32 x <16 x i8>], [32 x <16 x i8>] addrspace(2)* %arg1, i64 0, i32 0
   %tmp26 = load <16 x i8>, <16 x i8> addrspace(2)* %tmp25, !tbaa !0
   %tmp27 = fcmp oeq float %tmp22, 0.000000e+00
   %tmp26.bc = bitcast <16 x i8> %tmp26 to <4 x i32>
   br i1 %tmp27, label %if, label %else
 
 if:                                               ; preds = %entry
-  %val.if = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> <i32 11, i32 13>, <8 x i32> %tmp24, <4 x i32> %tmp26.bc, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
-  %val.if.0 = extractelement <4 x float> %val.if, i32 0
-  %val.if.1 = extractelement <4 x float> %val.if, i32 1
-  %val.if.2 = extractelement <4 x float> %val.if, i32 2
+  %tmp1 = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v2f32.v8i32(<2 x float> <float 0x36D6000000000000, float 0x36DA000000000000>, <8 x i32> %tmp24, <4 x i32> %tmp26.bc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %val.if.0 = extractelement <4 x float> %tmp1, i32 0
+  %val.if.1 = extractelement <4 x float> %tmp1, i32 1
+  %val.if.2 = extractelement <4 x float> %tmp1, i32 2
   br label %endif
 
 else:                                             ; preds = %entry
-  %val.else = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> <i32 5, i32 7>, <8 x i32> %tmp24, <4 x i32> %tmp26.bc, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
-  %val.else.0 = extractelement <4 x float> %val.else, i32 0
-  %val.else.1 = extractelement <4 x float> %val.else, i32 1
-  %val.else.2 = extractelement <4 x float> %val.else, i32 2
+  %tmp2 = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v2f32.v8i32(<2 x float> <float 0x36C4000000000000, float 0x36CC000000000000>, <8 x i32> %tmp24, <4 x i32> %tmp26.bc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %val.else.0 = extractelement <4 x float> %tmp2, i32 0
+  %val.else.1 = extractelement <4 x float> %tmp2, i32 1
+  %val.else.2 = extractelement <4 x float> %tmp2, i32 2
   br label %endif
 
 endif:                                            ; preds = %else, %if
   %val.0 = phi float [ %val.if.0, %if ], [ %val.else.0, %else ]
   %val.1 = phi float [ %val.if.1, %if ], [ %val.else.1, %else ]
   %val.2 = phi float [ %val.if.2, %if ], [ %val.else.2, %else ]
   call void @llvm.amdgcn.exp.f32(i32 0, i32 15, float %val.0, float %val.1, float %val.2, float 0.000000e+00, i1 true, i1 true) #0
   ret void
bb:
   br i1 %tmp36, label %bb38, label %bb80
 
 bb38:                                             ; preds = %bb
   %tmp52 = bitcast float %p2.i to i32
   %tmp53 = bitcast float %p2.i6 to i32
   %tmp54 = insertelement <2 x i32> undef, i32 %tmp52, i32 0
   %tmp55 = insertelement <2 x i32> %tmp54, i32 %tmp53, i32 1
   %tmp56 = bitcast <8 x i32> %tmp26 to <8 x i32>
-  %tmp58 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %tmp55, <8 x i32> %tmp56, <4 x i32> %tmp28, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %a.bc.i = bitcast <2 x i32> %tmp55 to <2 x float>
+  %tmp2 = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v2f32.v8i32(<2 x float> %a.bc.i, <8 x i32> %tmp56, <4 x i32> %tmp28, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
   br label %bb71
 
 bb80:                                             ; preds = %bb
   %tmp81 = bitcast float %p2.i to i32
   %tmp82 = bitcast float %p2.i6 to i32
   %tmp82.2 = add i32 %tmp82, 1
   %tmp83 = insertelement <2 x i32> undef, i32 %tmp81, i32 0
   %tmp84 = insertelement <2 x i32> %tmp83, i32 %tmp82.2, i32 1
   %tmp85 = bitcast <8 x i32> %tmp26 to <8 x i32>
-  %tmp87 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %tmp84, <8 x i32> %tmp85, <4 x i32> %tmp28, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %a.bc.i1 = bitcast <2 x i32> %tmp84 to <2 x float>
+  %tmp3 = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v2f32.v8i32(<2 x float> %a.bc.i1, <8 x i32> %tmp85, <4 x i32> %tmp28, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
   br label %bb71
 
 bb71:                                             ; preds = %bb80, %bb38
-  %tmp72 = phi <4 x float> [ %tmp58, %bb38 ], [ %tmp87, %bb80 ]
+  %tmp72 = phi <4 x float> [ %tmp2, %bb38 ], [ %tmp3, %bb80 ]
   %tmp88 = extractelement <4 x float> %tmp72, i32 0
   call void @llvm.amdgcn.exp.f32(i32 0, i32 15, float %tmp88, float %tmp88, float %tmp88, float %tmp88, i1 true, i1 true) #0
   ret void
 }
 
 ; Check the the resource descriptor is stored in an sgpr.
 ; CHECK-LABEL: {{^}}mimg_srsrc_sgpr:
 ; CHECK: image_sample v{{[0-9]+}}, v[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}] dmask:0x1
 define amdgpu_ps void @mimg_srsrc_sgpr([34 x <8 x i32>] addrspace(2)* byval %arg) #0 {
 bb:
   %tid = call i32 @llvm.amdgcn.mbcnt.lo(i32 -1, i32 0) #0
   %tmp7 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %arg, i32 0, i32 %tid
   %tmp8 = load <8 x i32>, <8 x i32> addrspace(2)* %tmp7, align 32, !tbaa !0
-  %tmp9 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> <i32 1061158912, i32 1048576000>, <8 x i32> %tmp8, <4 x i32> undef, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
-  %tmp10 = extractelement <4 x float> %tmp9, i32 0
+  %tmp = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v2f32.v8i32(<2 x float> <float 7.500000e-01, float 2.500000e-01>, <8 x i32> %tmp8, <4 x i32> undef, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %tmp10 = extractelement <4 x float> %tmp, i32 0
   %tmp12 = call <2 x half> @llvm.amdgcn.cvt.pkrtz(float undef, float %tmp10)
   call void @llvm.amdgcn.exp.compr.v2f16(i32 0, i32 15, <2 x half> %tmp12, <2 x half> undef, i1 true, i1 true) #0
   ret void
 }
 
 ; Check the the sampler is stored in an sgpr.
 ; CHECK-LABEL: {{^}}mimg_ssamp_sgpr:
 ; CHECK: image_sample v{{[0-9]+}}, v[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}] dmask:0x1
 define amdgpu_ps void @mimg_ssamp_sgpr([17 x <4 x i32>] addrspace(2)* byval %arg) #0 {
 bb:
   %tid = call i32 @llvm.amdgcn.mbcnt.lo(i32 -1, i32 0) #0
   %tmp7 = getelementptr [17 x <4 x i32>], [17 x <4 x i32>] addrspace(2)* %arg, i32 0, i32 %tid
   %tmp8 = load <4 x i32>, <4 x i32> addrspace(2)* %tmp7, align 16, !tbaa !0
-  %tmp9 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> <i32 1061158912, i32 1048576000>, <8 x i32> undef, <4 x i32> %tmp8, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
-  %tmp10 = extractelement <4 x float> %tmp9, i32 0
+  %tmp = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v2f32.v8i32(<2 x float> <float 7.500000e-01, float 2.500000e-01>, <8 x i32> undef, <4 x i32> %tmp8, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %tmp10 = extractelement <4 x float> %tmp, i32 0
   %tmp12 = call <2 x half> @llvm.amdgcn.cvt.pkrtz(float %tmp10, float undef)
   call void @llvm.amdgcn.exp.compr.v2f16(i32 0, i32 15, <2 x half> %tmp12, <2 x half> undef, i1 true, i1 true) #0
   ret void
 }
 
 declare float @llvm.fabs.f32(float) #1
 declare float @llvm.amdgcn.rsq.f32(float) #1
 declare float @llvm.exp2.f32(float) #1
 declare float @llvm.pow.f32(float, float) #1
 declare i32 @llvm.amdgcn.mbcnt.lo(i32, i32) #1
 declare float @llvm.amdgcn.interp.p1(float, i32, i32, i32) #1
 declare float @llvm.amdgcn.interp.p2(float, float, i32, i32, i32) #1
 declare void @llvm.amdgcn.exp.f32(i32, i32, float, float, float, float, i1, i1) #0
 declare void @llvm.amdgcn.exp.compr.v2f16(i32, i32, <2 x half>, <2 x half>, i1, i1) #0
 declare <2 x half> @llvm.amdgcn.cvt.pkrtz(float, float) #1
-
-declare <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #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) #2
 declare float @llvm.SI.load.const(<16 x i8>, i32) #1
 
 attributes #0 = { nounwind }
 attributes #1 = { nounwind readnone }
+attributes #2 = { nounwind readonly }
 
 !0 = !{!1, !1, i64 0, i32 1}
 !1 = !{!"const", !2}
 !2 = !{!"tbaa root"}
 !3 = !{!1, !1, i64 0}

test/CodeGen/AMDGPU/si-lod-bias.ll

main_body:
   %tmp28 = bitcast float %tmp21 to i32
   %tmp29 = bitcast float %p2.i to i32
   %tmp30 = bitcast float %p2.i6 to i32
   %tmp31 = insertelement <4 x i32> undef, i32 %tmp28, i32 0
   %tmp32 = insertelement <4 x i32> %tmp31, i32 %tmp29, i32 1
   %tmp33 = insertelement <4 x i32> %tmp32, i32 %tmp30, i32 2
   %tmp34 = insertelement <4 x i32> %tmp33, i32 undef, i32 3
   %tmp25.bc = bitcast <16 x i8> %tmp25 to <4 x i32>
-  %tmp35 = call <4 x float> @llvm.SI.image.sample.b.v4i32(<4 x i32> %tmp34, <8 x i32> %tmp23, <4 x i32> %tmp25.bc, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %tmp34.bc = bitcast <4 x i32> %tmp34 to <4 x float>
+  %tmp35 = call <4 x float> @llvm.amdgcn.image.sample.b.v4f32.v4f32.v8i32(<4 x float> %tmp34.bc, <8 x i32> %tmp23, <4 x i32> %tmp25.bc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
   %tmp36 = extractelement <4 x float> %tmp35, i32 0
   %tmp37 = extractelement <4 x float> %tmp35, i32 1
   %tmp38 = extractelement <4 x float> %tmp35, i32 2
   %tmp39 = extractelement <4 x float> %tmp35, i32 3
   call void @llvm.amdgcn.exp.f32(i32 0, i32 15, float %tmp36, float %tmp37, float %tmp38, float %tmp39, i1 true, i1 true) #0
   ret void
 }
 
 declare float @llvm.amdgcn.interp.p1(float, i32, i32, i32) #1
 declare float @llvm.amdgcn.interp.p2(float, float, i32, i32, i32) #1
 declare void @llvm.amdgcn.exp.f32(i32, i32, float, float, float, float, i1, i1) #0
-
+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) #2
 declare float @llvm.SI.load.const(<16 x i8>, i32) #1
-declare <4 x float> @llvm.SI.image.sample.b.v4i32(<4 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #1
 
 attributes #0 = { nounwind }
 attributes #1 = { nounwind readnone }
+attributes #2 = { nounwind readonly }
 
 !0 = !{!1, !1, i64 0, i32 1}
 !1 = !{!"const", !2}
 !2 = !{!"tbaa root"}

test/CodeGen/AMDGPU/si-scheduler.ll

main_body:
   %p1.i5 = call float @llvm.amdgcn.interp.p1(float %i.f.i3, i32 1, i32 0, i32 %arg5) #1
   %p2.i6 = call float @llvm.amdgcn.interp.p2(float %p1.i5, float %j.f.i4, i32 1, i32 0, i32 %arg5) #1
   %tmp27 = bitcast float %p2.i to i32
   %tmp28 = bitcast float %p2.i6 to i32
   %tmp29 = insertelement <2 x i32> undef, i32 %tmp27, i32 0
   %tmp30 = insertelement <2 x i32> %tmp29, i32 %tmp28, i32 1
   %tmp22.bc = bitcast <32 x i8> %tmp22 to <8 x i32>
   %tmp24.bc = bitcast <16 x i8> %tmp24 to <4 x i32>
-  %tmp31 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %tmp30, <8 x i32> %tmp22.bc, <4 x i32> %tmp24.bc, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %tmp30.bc = bitcast <2 x i32> %tmp30 to <2 x float>
+  %tmp31 = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v2f32.v8i32(<2 x float> %tmp30.bc, <8 x i32> %tmp22.bc, <4 x i32> %tmp24.bc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+
   %tmp32 = extractelement <4 x float> %tmp31, i32 0
   %tmp33 = extractelement <4 x float> %tmp31, i32 1
   %tmp34 = extractelement <4 x float> %tmp31, i32 2
   %tmp35 = extractelement <4 x float> %tmp31, i32 3
   %tmp36 = call <2 x half> @llvm.amdgcn.cvt.pkrtz(float %tmp32, float %tmp33)
   %tmp38 = call <2 x half> @llvm.amdgcn.cvt.pkrtz(float %tmp34, float %tmp35)
   call void @llvm.amdgcn.exp.compr.v2f16(i32 0, i32 15, <2 x half> %tmp36, <2 x half> %tmp38, i1 true, i1 false) #0
   ret void
 }
 
 declare float @llvm.amdgcn.interp.p1(float, i32, i32, i32) #1
 declare float @llvm.amdgcn.interp.p2(float, float, i32, i32, i32) #1
 declare void @llvm.amdgcn.exp.compr.v2f16(i32, i32, <2 x half>, <2 x half>, i1, i1) #0
-
-declare <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #1
 declare <2 x half> @llvm.amdgcn.cvt.pkrtz(float, float) #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) #2
 
 attributes #0 = { nounwind }
 attributes #1 = { nounwind readnone }
+attributes #2 = { nounwind readonly }
 
 !0 = !{!1, !1, i64 0, i32 1}
 !1 = !{!"const", !2}
 !2 = !{!"tbaa root"}

test/CodeGen/AMDGPU/si-sgpr-spill.ll

 
 ENDIF:                                            ; preds = %LOOP
   %tmp237 = bitcast float %temp28.0 to i32
   %tmp238 = bitcast float %temp29.0 to i32
   %tmp239 = insertelement <8 x i32> %tmp223, i32 %tmp237, i32 4
   %tmp240 = insertelement <8 x i32> %tmp239, i32 %tmp238, i32 5
   %tmp241 = insertelement <8 x i32> %tmp240, i32 undef, i32 6
   %tmp242 = insertelement <8 x i32> %tmp241, i32 undef, i32 7
-  %tmp243 = call <4 x float> @llvm.SI.image.sample.d.v8i32(<8 x i32> %tmp242, <8 x i32> %tmp61, <4 x i32> %tmp63.bc, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %tmp242.bc = bitcast <8 x i32> %tmp242 to <8 x float>
+  %tmp243 = call <4 x float> @llvm.amdgcn.image.sample.d.v4f32.v8f32.v8i32(<8 x float> %tmp242.bc, <8 x i32> %tmp61, <4 x i32> %tmp63.bc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
   %tmp244 = extractelement <4 x float> %tmp243, i32 3
   %tmp245 = fcmp oge float %temp30.0, %tmp244
   %tmp246 = sext i1 %tmp245 to i32
   %tmp247 = bitcast i32 %tmp246 to float
   %tmp248 = bitcast float %tmp247 to i32
   %tmp249 = and i32 %tmp248, 1065353216
   %tmp250 = bitcast i32 %tmp249 to float
   %tmp251 = fmul float %tmp213, %tmp250
IF67:                                             ; preds = %LOOP65
   %tmp270 = insertelement <8 x i32> %tmp269, i32 %tmp264, i32 1
   %tmp271 = insertelement <8 x i32> %tmp270, i32 %tmp265, i32 2
   %tmp272 = insertelement <8 x i32> %tmp271, i32 %tmp266, i32 3
   %tmp273 = insertelement <8 x i32> %tmp272, i32 %tmp267, i32 4
   %tmp274 = insertelement <8 x i32> %tmp273, i32 %tmp268, i32 5
   %tmp275 = insertelement <8 x i32> %tmp274, i32 undef, i32 6
   %tmp276 = insertelement <8 x i32> %tmp275, i32 undef, i32 7
   %tmp67.bc = bitcast <16 x i8> %tmp67 to <4 x i32>
-  %tmp277 = call <4 x float> @llvm.SI.image.sample.d.v8i32(<8 x i32> %tmp276, <8 x i32> %tmp65, <4 x i32> %tmp67.bc, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %tmp276.bc = bitcast <8 x i32> %tmp276 to <8 x float>
+  %tmp277 = call <4 x float> @llvm.amdgcn.image.sample.d.v4f32.v8f32.v8i32(<8 x float> %tmp276.bc, <8 x i32> %tmp65, <4 x i32> %tmp67.bc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
   %tmp278 = extractelement <4 x float> %tmp277, i32 0
   %tmp279 = extractelement <4 x float> %tmp277, i32 1
   %tmp280 = extractelement <4 x float> %tmp277, i32 2
   %tmp281 = extractelement <4 x float> %tmp277, i32 3
   %tmp282 = fmul float %tmp281, %tmp46
   %tmp283 = bitcast float %tmp135 to i32
   %tmp284 = bitcast float %tmp181 to i32
   %tmp285 = bitcast float %tmp136 to i32
   %tmp286 = bitcast float %tmp182 to i32
   %tmp287 = bitcast float %temp28.1 to i32
   %tmp288 = bitcast float %temp29.1 to i32
   %tmp289 = insertelement <8 x i32> undef, i32 %tmp283, i32 0
   %tmp290 = insertelement <8 x i32> %tmp289, i32 %tmp284, i32 1
   %tmp291 = insertelement <8 x i32> %tmp290, i32 %tmp285, i32 2
   %tmp292 = insertelement <8 x i32> %tmp291, i32 %tmp286, i32 3
   %tmp293 = insertelement <8 x i32> %tmp292, i32 %tmp287, i32 4
   %tmp294 = insertelement <8 x i32> %tmp293, i32 %tmp288, i32 5
   %tmp295 = insertelement <8 x i32> %tmp294, i32 undef, i32 6
   %tmp296 = insertelement <8 x i32> %tmp295, i32 undef, i32 7
   %tmp83.bc = bitcast <16 x i8> %tmp83 to <4 x i32>
-  %tmp297 = call <4 x float> @llvm.SI.image.sample.d.v8i32(<8 x i32> %tmp296, <8 x i32> %tmp81, <4 x i32> %tmp83.bc, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %tmp296.bc = bitcast <8 x i32> %tmp296 to <8 x float>
+  %tmp297 = call <4 x float> @llvm.amdgcn.image.sample.d.v4f32.v8f32.v8i32(<8 x float> %tmp296.bc, <8 x i32> %tmp81, <4 x i32> %tmp83.bc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
   %tmp298 = extractelement <4 x float> %tmp297, i32 0
   %tmp299 = extractelement <4 x float> %tmp297, i32 1
   %tmp300 = extractelement <4 x float> %tmp297, i32 2
   %tmp301 = bitcast float %tmp135 to i32
   %tmp302 = bitcast float %tmp181 to i32
   %tmp303 = bitcast float %tmp136 to i32
   %tmp304 = bitcast float %tmp182 to i32
   %tmp305 = bitcast float %temp28.1 to i32
   %tmp306 = bitcast float %temp29.1 to i32
   %tmp307 = insertelement <8 x i32> undef, i32 %tmp301, i32 0
   %tmp308 = insertelement <8 x i32> %tmp307, i32 %tmp302, i32 1
   %tmp309 = insertelement <8 x i32> %tmp308, i32 %tmp303, i32 2
   %tmp310 = insertelement <8 x i32> %tmp309, i32 %tmp304, i32 3
   %tmp311 = insertelement <8 x i32> %tmp310, i32 %tmp305, i32 4
   %tmp312 = insertelement <8 x i32> %tmp311, i32 %tmp306, i32 5
   %tmp313 = insertelement <8 x i32> %tmp312, i32 undef, i32 6
   %tmp314 = insertelement <8 x i32> %tmp313, i32 undef, i32 7
   %tmp79.bc = bitcast <16 x i8> %tmp79 to <4 x i32>
-  %tmp315 = call <4 x float> @llvm.SI.image.sample.d.v8i32(<8 x i32> %tmp314, <8 x i32> %tmp77, <4 x i32> %tmp79.bc, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %tmp314.bc = bitcast <8 x i32> %tmp314 to <8 x float>
+  %tmp315 = call <4 x float> @llvm.amdgcn.image.sample.d.v4f32.v8f32.v8i32(<8 x float> %tmp314.bc, <8 x i32> %tmp77, <4 x i32> %tmp79.bc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
   %tmp316 = extractelement <4 x float> %tmp315, i32 0
   %tmp317 = extractelement <4 x float> %tmp315, i32 1
   %tmp318 = extractelement <4 x float> %tmp315, i32 2
   %tmp319 = fmul float %tmp316, %tmp22
   %tmp320 = fmul float %tmp317, %tmp23
   %tmp321 = fmul float %tmp318, %tmp24
   %tmp322 = fmul float %tmp298, %tmp25
   %tmp323 = fadd float %tmp322, %tmp319
IF67:                                             ; preds = %LOOP65
   %tmp337 = insertelement <8 x i32> undef, i32 %tmp331, i32 0
   %tmp338 = insertelement <8 x i32> %tmp337, i32 %tmp332, i32 1
   %tmp339 = insertelement <8 x i32> %tmp338, i32 %tmp333, i32 2
   %tmp340 = insertelement <8 x i32> %tmp339, i32 %tmp334, i32 3
   %tmp341 = insertelement <8 x i32> %tmp340, i32 %tmp335, i32 4
   %tmp342 = insertelement <8 x i32> %tmp341, i32 %tmp336, i32 5
   %tmp343 = insertelement <8 x i32> %tmp342, i32 undef, i32 6
   %tmp344 = insertelement <8 x i32> %tmp343, i32 undef, i32 7
-  %tmp345 = call <4 x float> @llvm.SI.image.sample.d.v8i32(<8 x i32> %tmp344, <8 x i32> %tmp61, <4 x i32> %tmp63.bc, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %tmp344.bc = bitcast <8 x i32> %tmp344 to <8 x float>
+  %tmp345 = call <4 x float> @llvm.amdgcn.image.sample.d.v4f32.v8f32.v8i32(<8 x float> %tmp344.bc, <8 x i32> %tmp61, <4 x i32> %tmp63.bc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
   %tmp346 = extractelement <4 x float> %tmp345, i32 0
   %tmp347 = extractelement <4 x float> %tmp345, i32 1
   %tmp348 = extractelement <4 x float> %tmp345, i32 2
   %tmp349 = fadd float %tmp346, -5.000000e-01
   %tmp350 = fadd float %tmp347, -5.000000e-01
   %tmp351 = fadd float %tmp348, -5.000000e-01
   %tmp352 = fmul float %tmp349, %tmp349
   %tmp353 = fmul float %tmp350, %tmp350
IF67:                                             ; preds = %LOOP65
   %tmp368 = insertelement <8 x i32> %tmp367, i32 %tmp362, i32 1
   %tmp369 = insertelement <8 x i32> %tmp368, i32 %tmp363, i32 2
   %tmp370 = insertelement <8 x i32> %tmp369, i32 %tmp364, i32 3
   %tmp371 = insertelement <8 x i32> %tmp370, i32 %tmp365, i32 4
   %tmp372 = insertelement <8 x i32> %tmp371, i32 %tmp366, i32 5
   %tmp373 = insertelement <8 x i32> %tmp372, i32 undef, i32 6
   %tmp374 = insertelement <8 x i32> %tmp373, i32 undef, i32 7
   %tmp71.bc = bitcast <16 x i8> %tmp71 to <4 x i32>
-  %tmp375 = call <4 x float> @llvm.SI.image.sample.d.v8i32(<8 x i32> %tmp374, <8 x i32> %tmp69, <4 x i32> %tmp71.bc, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %tmp374.bc = bitcast <8 x i32> %tmp374 to <8 x float>
+  %tmp375 = call <4 x float> @llvm.amdgcn.image.sample.d.v4f32.v8f32.v8i32(<8 x float> %tmp374.bc, <8 x i32> %tmp69, <4 x i32> %tmp71.bc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
   %tmp376 = extractelement <4 x float> %tmp375, i32 0
   %tmp377 = extractelement <4 x float> %tmp375, i32 1
   %tmp378 = extractelement <4 x float> %tmp375, i32 2
   %tmp379 = extractelement <4 x float> %tmp375, i32 3
   %tmp380 = fsub float -0.000000e+00, %p2.i90
   %tmp381 = fsub float -0.000000e+00, %p2.i84
   %tmp382 = fsub float -0.000000e+00, %p2.i78
   %tmp383 = fmul float %tmp358, %tmp380
IF67:                                             ; preds = %LOOP65
   %tmp422 = insertelement <8 x i32> %tmp421, i32 %tmp416, i32 1
   %tmp423 = insertelement <8 x i32> %tmp422, i32 %tmp417, i32 2
   %tmp424 = insertelement <8 x i32> %tmp423, i32 %tmp418, i32 3
   %tmp425 = insertelement <8 x i32> %tmp424, i32 %tmp419, i32 4
   %tmp426 = insertelement <8 x i32> %tmp425, i32 %tmp420, i32 5
   %tmp427 = insertelement <8 x i32> %tmp426, i32 undef, i32 6
   %tmp428 = insertelement <8 x i32> %tmp427, i32 undef, i32 7
   %tmp87.bc = bitcast <16 x i8> %tmp87 to <4 x i32>
-  %tmp429 = call <4 x float> @llvm.SI.image.sample.d.v8i32(<8 x i32> %tmp428, <8 x i32> %tmp85, <4 x i32> %tmp87.bc, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %tmp428.bc = bitcast <8 x i32> %tmp428 to <8 x float>
+  %tmp429 = call <4 x float> @llvm.amdgcn.image.sample.d.v4f32.v8f32.v8i32(<8 x float> %tmp428.bc, <8 x i32> %tmp85, <4 x i32> %tmp87.bc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
   %tmp430 = extractelement <4 x float> %tmp429, i32 0
   %tmp431 = extractelement <4 x float> %tmp429, i32 1
   %tmp432 = extractelement <4 x float> %tmp429, i32 2
   %tmp433 = fmul float %tmp47, %tmp410
   %tmp434 = fmul float %tmp48, %tmp410
   %tmp435 = fmul float %tmp49, %tmp410
   %tmp436 = fmul float %tmp50, %tmp412
   %tmp437 = fadd float %tmp436, %tmp433
IF67:                                             ; preds = %LOOP65
   %tmp463 = bitcast float %tmp462 to i32
   %tmp464 = bitcast float %tmp460 to i32
   %tmp465 = bitcast float %tmp456 to i32
   %tmp466 = insertelement <4 x i32> undef, i32 %tmp463, i32 0
   %tmp467 = insertelement <4 x i32> %tmp466, i32 %tmp464, i32 1
   %tmp468 = insertelement <4 x i32> %tmp467, i32 %tmp465, i32 2
   %tmp469 = insertelement <4 x i32> %tmp468, i32 undef, i32 3
   %tmp91.bc = bitcast <16 x i8> %tmp91 to <4 x i32>
-  %tmp470 = call <4 x float> @llvm.SI.image.sample.v4i32(<4 x i32> %tmp469, <8 x i32> %tmp89, <4 x i32> %tmp91.bc, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %tmp469.bc = bitcast <4 x i32> %tmp469 to <4 x float>
+  %tmp470 = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> %tmp469.bc, <8 x i32> %tmp89, <4 x i32> %tmp91.bc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false) #0
   %tmp471 = extractelement <4 x float> %tmp470, i32 0
   %tmp472 = extractelement <4 x float> %tmp470, i32 1
   %tmp473 = extractelement <4 x float> %tmp470, i32 2
   %tmp474 = fmul float %tmp430, %tmp471
   %tmp475 = fadd float %tmp474, %tmp328
   %tmp476 = fmul float %tmp431, %tmp472
   %tmp477 = fadd float %tmp476, %tmp329
   %tmp478 = fmul float %tmp432, %tmp473
IF67:                                             ; preds = %LOOP65
   %tmp564 = insertelement <8 x i32> %tmp563, i32 %tmp558, i32 1
   %tmp565 = insertelement <8 x i32> %tmp564, i32 %tmp559, i32 2
   %tmp566 = insertelement <8 x i32> %tmp565, i32 %tmp560, i32 3
   %tmp567 = insertelement <8 x i32> %tmp566, i32 %tmp561, i32 4
   %tmp568 = insertelement <8 x i32> %tmp567, i32 %tmp562, i32 5
   %tmp569 = insertelement <8 x i32> %tmp568, i32 undef, i32 6
   %tmp570 = insertelement <8 x i32> %tmp569, i32 undef, i32 7
   %tmp75.bc = bitcast <16 x i8> %tmp75 to <4 x i32>
-  %tmp571 = call <4 x float> @llvm.SI.image.sample.d.v8i32(<8 x i32> %tmp570, <8 x i32> %tmp73, <4 x i32> %tmp75.bc, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %tmp570.bc = bitcast <8 x i32> %tmp570 to <8 x float>
+  %tmp571 = call <4 x float> @llvm.amdgcn.image.sample.d.v4f32.v8f32.v8i32(<8 x float> %tmp570.bc, <8 x i32> %tmp73, <4 x i32> %tmp75.bc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
   %tmp572 = extractelement <4 x float> %tmp571, i32 0
   %tmp573 = extractelement <4 x float> %tmp571, i32 1
   %tmp574 = extractelement <4 x float> %tmp571, i32 2
   %tmp575 = fmul float %tmp572, %tmp43
   %tmp576 = fadd float %tmp575, %tmp552
   %tmp577 = fmul float %tmp573, %tmp44
   %tmp578 = fadd float %tmp577, %tmp554
   %tmp579 = fmul float %tmp574, %tmp45
   %tmp580 = fadd float %tmp579, %tmp556
   %tmp581 = call <2 x half> @llvm.amdgcn.cvt.pkrtz(float %tmp576, float %tmp578)
   %tmp583 = call <2 x half> @llvm.amdgcn.cvt.pkrtz(float %tmp580, float %tmp282)
   call void @llvm.amdgcn.exp.compr.v2f16(i32 0, i32 15, <2 x half> %tmp581, <2 x half> %tmp583, i1 true, i1 true) #0
   ret void
 
 ENDIF66:                                          ; preds = %LOOP65
   %tmp585 = bitcast float %temp28.1 to i32
   %tmp586 = bitcast float %temp29.1 to i32
   %tmp587 = insertelement <8 x i32> %tmp236, i32 %tmp585, i32 4
   %tmp588 = insertelement <8 x i32> %tmp587, i32 %tmp586, i32 5
   %tmp589 = insertelement <8 x i32> %tmp588, i32 undef, i32 6
   %tmp590 = insertelement <8 x i32> %tmp589, i32 undef, i32 7
-  %tmp591 = call <4 x float> @llvm.SI.image.sample.d.v8i32(<8 x i32> %tmp590, <8 x i32> %tmp61, <4 x i32> %tmp63.bc, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %tmp590.bc = bitcast <8 x i32> %tmp590 to <8 x float>
+  %tmp591 = call <4 x float> @llvm.amdgcn.image.sample.d.v4f32.v8f32.v8i32(<8 x float> %tmp590.bc, <8 x i32> %tmp61, <4 x i32> %tmp63.bc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
   %tmp592 = extractelement <4 x float> %tmp591, i32 3
   %tmp593 = fcmp oge float %temp30.1, %tmp592
   %tmp594 = sext i1 %tmp593 to i32
   %tmp595 = bitcast i32 %tmp594 to float
   %tmp596 = bitcast float %tmp595 to i32
   %tmp597 = and i32 %tmp596, 1065353216
   %tmp598 = bitcast i32 %tmp597 to float
   %tmp599 = fmul float 5.000000e-01, %temp32.0
   %tmp600 = fsub float -0.000000e+00, %tmp599
   %tmp601 = fmul float %tmp598, %temp32.0
   %tmp602 = fadd float %tmp601, %tmp600
   %tmp603 = fmul float %tmp213, %tmp602
   %tmp604 = fadd float %tmp603, %temp28.1
   %tmp605 = fmul float %tmp214, %tmp602
   %tmp606 = fadd float %tmp605, %temp29.1
   %tmp607 = fmul float %tmp215, %tmp602
   %tmp608 = fadd float %tmp607, %temp30.1
   %tmp609 = fadd float %temp24.1, 1.000000e+00
   %tmp610 = fmul float %temp32.0, 5.000000e-01
   br label %LOOP65
 }
 
-define amdgpu_ps void @main1([17 x <16 x i8>] addrspace(2)* byval %arg, [32 x <16 x i8>] addrspace(2)* byval %arg1, [16 x <8 x i32>] addrspace(2)* byval %arg2, float inreg %arg3, i32 inreg %arg4, <2 x i32> %arg5, <2 x i32> %arg6, <2 x i32> %arg7, <3 x i32> %arg8, <2 x i32> %arg9, <2 x i32> %arg10, <2 x i32> %arg11, float %arg12, float %arg13, float %arg14, float %arg15, float %arg16, float %arg17, float %arg18, float %arg19, float %arg20) {
+define amdgpu_ps void @main1([17 x <16 x i8>] addrspace(2)* byval %arg, [32 x <16 x i8>] addrspace(2)* byval %arg1, [16 x <8 x i32>] addrspace(2)* byval %arg2, float inreg %arg3, i32 inreg %arg4, <2 x i32> %arg5, <2 x i32> %arg6, <2 x i32> %arg7, <3 x i32> %arg8, <2 x i32> %arg9, <2 x i32> %arg10, <2 x i32> %arg11, float %arg12, float %arg13, float %arg14, float %arg15, float %arg16, float %arg17, float %arg18, float %arg19, float %arg20) #0 {
 main_body:
   %tmp = getelementptr [17 x <16 x i8>], [17 x <16 x i8>] addrspace(2)* %arg, i64 0, i32 0
   %tmp21 = load <16 x i8>, <16 x i8> addrspace(2)* %tmp, !tbaa !0
   %tmp22 = call float @llvm.SI.load.const(<16 x i8> %tmp21, i32 0)
   %tmp23 = call float @llvm.SI.load.const(<16 x i8> %tmp21, i32 4)
   %tmp24 = call float @llvm.SI.load.const(<16 x i8> %tmp21, i32 8)
   %tmp25 = call float @llvm.SI.load.const(<16 x i8> %tmp21, i32 12)
   %tmp26 = call float @llvm.SI.load.const(<16 x i8> %tmp21, i32 28)
main_body:
   %tmp218 = fmul float %., %tmp53
   %tmp219 = fmul float %arg13, %tmp46
   %tmp220 = fmul float %tmp196, %tmp47
   %tmp221 = bitcast float %p2.i132 to i32
   %tmp222 = bitcast float %p2.i126 to i32
   %tmp223 = insertelement <2 x i32> undef, i32 %tmp221, i32 0
   %tmp224 = insertelement <2 x i32> %tmp223, i32 %tmp222, i32 1
   %tmp132.bc = bitcast <16 x i8> %tmp132 to <4 x i32>
-  %tmp225 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %tmp224, <8 x i32> %tmp130, <4 x i32> %tmp132.bc, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %tmp224.bc = bitcast <2 x i32> %tmp224 to <2 x float>
+  %tmp225 = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v2f32.v8i32(<2 x float> %tmp224.bc, <8 x i32> %tmp130, <4 x i32> %tmp132.bc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
   %tmp226 = extractelement <4 x float> %tmp225, i32 0
   %tmp227 = extractelement <4 x float> %tmp225, i32 1
   %tmp228 = extractelement <4 x float> %tmp225, i32 2
   %tmp229 = extractelement <4 x float> %tmp225, i32 3
   %tmp230 = fmul float %tmp226, 0x4012611180000000
   %tmp231 = fmul float %tmp227, 0x4012611180000000
   %tmp232 = fmul float %tmp228, 0x4012611180000000
   %one.sub.a.i = fsub float 1.000000e+00, %tmp26
LOOP:                                             ; preds = %LOOP, %main_body
   %temp170.0 = phi float [ %tmp252, %main_body ], [ %tmp286, %LOOP ]
   %tmp276 = bitcast float %temp168.0 to i32
   %tmp277 = bitcast float %temp169.0 to i32
   %tmp278 = insertelement <4 x i32> undef, i32 %tmp276, i32 0
   %tmp279 = insertelement <4 x i32> %tmp278, i32 %tmp277, i32 1
   %tmp280 = insertelement <4 x i32> %tmp279, i32 0, i32 2
   %tmp281 = insertelement <4 x i32> %tmp280, i32 undef, i32 3
   %tmp148.bc = bitcast <16 x i8> %tmp148 to <4 x i32>
-  %tmp282 = call <4 x float> @llvm.SI.image.sample.l.v4i32(<4 x i32> %tmp281, <8 x i32> %tmp146, <4 x i32> %tmp148.bc, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %tmp281.bc = bitcast <4 x i32> %tmp281 to <4 x float>
+  %tmp282 = call <4 x float> @llvm.amdgcn.image.sample.l.v4f32.v4f32.v8i32(<4 x float> %tmp281.bc, <8 x i32> %tmp146, <4 x i32> %tmp148.bc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
   %tmp283 = extractelement <4 x float> %tmp282, i32 3
   %tmp284 = fadd float %temp168.0, %tmp273
   %tmp285 = fadd float %temp169.0, %tmp274
   %tmp286 = fadd float %temp170.0, %tmp275
   %tmp287 = fsub float -0.000000e+00, %tmp283
   %tmp288 = fadd float %tmp286, %tmp287
   %tmp289 = fcmp oge float 0.000000e+00, %tmp288
   %tmp290 = sext i1 %tmp289 to i32
IF189:                                            ; preds = %LOOP
   %tmp335 = fadd float %p2.i162, %tmp329
   %tmp336 = fadd float %p2.i156, %tmp331
   %tmp337 = fadd float %p2.i150, %tmp333
   %tmp338 = bitcast float %tmp334 to i32
   %tmp339 = bitcast float %tmp335 to i32
   %tmp340 = insertelement <2 x i32> undef, i32 %tmp338, i32 0
   %tmp341 = insertelement <2 x i32> %tmp340, i32 %tmp339, i32 1
   %tmp136.bc = bitcast <16 x i8> %tmp136 to <4 x i32>
-  %tmp342 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %tmp341, <8 x i32> %tmp134, <4 x i32> %tmp136.bc, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
-  %tmp343 = extractelement <4 x float> %tmp342, i32 0
-  %tmp344 = extractelement <4 x float> %tmp342, i32 1
-  %tmp345 = extractelement <4 x float> %tmp342, i32 2
-  %tmp346 = extractelement <4 x float> %tmp342, i32 3
+  %a.bc.i = bitcast <2 x i32> %tmp341 to <2 x float>
+  %tmp0 = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v2f32.v8i32(<2 x float> %a.bc.i, <8 x i32> %tmp134, <4 x i32> %tmp136.bc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %tmp343 = extractelement <4 x float> %tmp0, i32 0
+  %tmp344 = extractelement <4 x float> %tmp0, i32 1
+  %tmp345 = extractelement <4 x float> %tmp0, i32 2
+  %tmp346 = extractelement <4 x float> %tmp0, i32 3
   %tmp347 = fmul float %tmp343, %tmp22
   %tmp348 = fmul float %tmp344, %tmp23
   %tmp349 = fmul float %tmp345, %tmp24
   %tmp350 = fmul float %tmp346, %tmp25
   %tmp351 = fmul float %tmp347, %p2.i96
   %tmp352 = fmul float %tmp348, %p2.i90
   %tmp353 = fmul float %tmp349, %p2.i84
   %tmp354 = fmul float %tmp350, %p2.i78
IF189:                                            ; preds = %LOOP
   %one.sub.ac.i30 = fmul float %one.sub.a.i29, %tmp353
   %mul.i31 = fmul float %tmp345, %tmp353
   %result.i32 = fadd float %mul.i31, %one.sub.ac.i30
   %tmp358 = bitcast float %tmp336 to i32
   %tmp359 = bitcast float %tmp337 to i32
   %tmp360 = insertelement <2 x i32> undef, i32 %tmp358, i32 0
   %tmp361 = insertelement <2 x i32> %tmp360, i32 %tmp359, i32 1
   %tmp152.bc = bitcast <16 x i8> %tmp152 to <4 x i32>
-  %tmp362 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %tmp361, <8 x i32> %tmp150, <4 x i32> %tmp152.bc, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
-  %tmp363 = extractelement <4 x float> %tmp362, i32 2
+  %a.bc.i3 = bitcast <2 x i32> %tmp361 to <2 x float>
+  %tmp1 = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v2f32.v8i32(<2 x float> %a.bc.i3, <8 x i32> %tmp150, <4 x i32> %tmp152.bc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %tmp363 = extractelement <4 x float> %tmp1, i32 2
   %tmp364 = fmul float %result.i40, %result.i
   %tmp365 = fmul float %result.i36, %result.i44
   %tmp366 = fmul float %result.i32, %result.i42
   %tmp367 = fmul float %tmp354, %tmp229
   %tmp368 = bitcast float %tmp310 to i32
   %tmp369 = bitcast float %tmp311 to i32
   %tmp370 = insertelement <2 x i32> undef, i32 %tmp368, i32 0
   %tmp371 = insertelement <2 x i32> %tmp370, i32 %tmp369, i32 1
   %tmp140.bc = bitcast <16 x i8> %tmp140 to <4 x i32>
-  %tmp372 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %tmp371, <8 x i32> %tmp138, <4 x i32> %tmp140.bc, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
-  %tmp373 = extractelement <4 x float> %tmp372, i32 0
-  %tmp374 = extractelement <4 x float> %tmp372, i32 1
-  %tmp375 = extractelement <4 x float> %tmp372, i32 2
-  %tmp376 = extractelement <4 x float> %tmp372, i32 3
+  %a.bc.i2 = bitcast <2 x i32> %tmp371 to <2 x float>
+  %tmp2 = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v2f32.v8i32(<2 x float> %a.bc.i2, <8 x i32> %tmp138, <4 x i32> %tmp140.bc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %tmp373 = extractelement <4 x float> %tmp2, i32 0
+  %tmp374 = extractelement <4 x float> %tmp2, i32 1
+  %tmp375 = extractelement <4 x float> %tmp2, i32 2
+  %tmp376 = extractelement <4 x float> %tmp2, i32 3
   %tmp377 = fcmp olt float 0.000000e+00, %tmp375
   %tmp378 = sext i1 %tmp377 to i32
   %tmp379 = bitcast i32 %tmp378 to float
   %tmp380 = bitcast float %tmp379 to i32
   %tmp381 = icmp ne i32 %tmp380, 0
   %.224 = select i1 %tmp381, float %tmp374, float %tmp373
   %.225 = select i1 %tmp381, float %tmp376, float %tmp374
   %tmp382 = bitcast float %tmp320 to i32
   %tmp383 = bitcast float %tmp321 to i32
   %tmp384 = insertelement <2 x i32> undef, i32 %tmp382, i32 0
   %tmp385 = insertelement <2 x i32> %tmp384, i32 %tmp383, i32 1
   %tmp144.bc = bitcast <16 x i8> %tmp144 to <4 x i32>
-  %tmp386 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %tmp385, <8 x i32> %tmp142, <4 x i32> %tmp144.bc, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
-  %tmp387 = extractelement <4 x float> %tmp386, i32 0
-  %tmp388 = extractelement <4 x float> %tmp386, i32 1
-  %tmp389 = extractelement <4 x float> %tmp386, i32 2
-  %tmp390 = extractelement <4 x float> %tmp386, i32 3
+  %a.bc.i1 = bitcast <2 x i32> %tmp385 to <2 x float>
+  %tmp3 = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v2f32.v8i32(<2 x float> %a.bc.i1, <8 x i32> %tmp142, <4 x i32> %tmp144.bc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+  %tmp387 = extractelement <4 x float> %tmp3, i32 0
+  %tmp388 = extractelement <4 x float> %tmp3, i32 1
+  %tmp389 = extractelement <4 x float> %tmp3, i32 2
+  %tmp390 = extractelement <4 x float> %tmp3, i32 3
   %tmp391 = fcmp olt float 0.000000e+00, %tmp389
   %tmp392 = sext i1 %tmp391 to i32
   %tmp393 = bitcast i32 %tmp392 to float
   %tmp394 = bitcast float %tmp393 to i32
   %tmp395 = icmp ne i32 %tmp394, 0
   %temp112.1 = select i1 %tmp395, float %tmp388, float %tmp387
   %temp113.1 = select i1 %tmp395, float %tmp390, float %tmp388
   %tmp396 = fmul float %.224, 2.000000e+00
ENDIF197:                                         ; preds = %IF198, %IF189
   %temp14.0 = phi float [ %tmp465, %IF198 ], [ %tmp457, %IF189 ]
   %temp13.0 = phi float [ %tmp464, %IF198 ], [ %tmp456, %IF189 ]
   %temp12.0 = phi float [ %tmp463, %IF198 ], [ %tmp455, %IF189 ]
   %tmp466 = bitcast float %tmp219 to i32
   %tmp467 = bitcast float %tmp220 to i32
   %tmp468 = insertelement <2 x i32> undef, i32 %tmp466, i32 0
   %tmp469 = insertelement <2 x i32> %tmp468, i32 %tmp467, i32 1
   %tmp160.bc = bitcast <16 x i8> %tmp160 to <4 x i32>
-  %tmp470 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %tmp469, <8 x i32> %tmp158, <4 x i32> %tmp160.bc, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %tmp469.bc = bitcast <2 x i32> %tmp469 to <2 x float>
+  %tmp470 = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v2f32.v8i32(<2 x float> %tmp469.bc, <8 x i32> %tmp158, <4 x i32> %tmp160.bc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
   %tmp471 = extractelement <4 x float> %tmp470, i32 0
   %tmp472 = extractelement <4 x float> %tmp470, i32 1
   %tmp473 = extractelement <4 x float> %tmp470, i32 2
   %tmp474 = extractelement <4 x float> %tmp470, i32 3
   %tmp475 = fmul float %tmp471, %tmp39
   %tmp476 = fadd float %tmp475, %tmp40
   %tmp477 = fmul float %tmp472, %tmp39
   %tmp478 = fadd float %tmp477, %tmp40
   %tmp479 = fmul float %tmp473, %tmp39
   %tmp480 = fadd float %tmp479, %tmp40
   %tmp481 = fmul float %tmp474, %tmp41
   %tmp482 = fadd float %tmp481, %tmp42
   %tmp483 = bitcast float %p2.i144 to i32
   %tmp484 = bitcast float %p2.i138 to i32
   %tmp485 = insertelement <2 x i32> undef, i32 %tmp483, i32 0
   %tmp486 = insertelement <2 x i32> %tmp485, i32 %tmp484, i32 1
   %tmp156.bc = bitcast <16 x i8> %tmp156 to <4 x i32>
-  %tmp487 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> %tmp486, <8 x i32> %tmp154, <4 x i32> %tmp156.bc, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %tmp486.bc = bitcast <2 x i32> %tmp486 to <2 x float>
+  %tmp487 = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v2f32.v8i32(<2 x float> %tmp486.bc, <8 x i32> %tmp154, <4 x i32> %tmp156.bc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
   %tmp488 = extractelement <4 x float> %tmp487, i32 0
   %tmp489 = extractelement <4 x float> %tmp487, i32 1
   %tmp490 = extractelement <4 x float> %tmp487, i32 2
   %tmp491 = extractelement <4 x float> %tmp487, i32 3
   %tmp492 = fmul float %tmp491, 3.200000e+01
   %tmp493 = fadd float %tmp492, -1.600000e+01
   %tmp494 = call float @llvm.exp2.f32(float %tmp493)
   %tmp495 = fmul float %tmp488, %tmp494
ENDIF209:                                         ; preds = %ELSE214, %ELSE211, %ENDIF200
   %tmp653 = bitcast float %tmp642 to i32
   %tmp654 = bitcast float %tmp644 to i32
   %tmp655 = bitcast float 0.000000e+00 to i32
   %tmp656 = insertelement <4 x i32> undef, i32 %tmp653, i32 0
   %tmp657 = insertelement <4 x i32> %tmp656, i32 %tmp654, i32 1
   %tmp658 = insertelement <4 x i32> %tmp657, i32 %tmp655, i32 2
   %tmp659 = insertelement <4 x i32> %tmp658, i32 undef, i32 3
   %tmp128.bc = bitcast <16 x i8> %tmp128 to <4 x i32>
-  %tmp660 = call <4 x float> @llvm.SI.image.sample.l.v4i32(<4 x i32> %tmp659, <8 x i32> %tmp126, <4 x i32> %tmp128.bc, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %tmp659.bc = bitcast <4 x i32> %tmp659 to <4 x float>
+  %tmp660 = call <4 x float> @llvm.amdgcn.image.sample.l.v4f32.v4f32.v8i32(<4 x float> %tmp659.bc, <8 x i32> %tmp126, <4 x i32> %tmp128.bc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
   %tmp661 = extractelement <4 x float> %tmp660, i32 0
   %tmp662 = extractelement <4 x float> %tmp660, i32 1
   %tmp663 = bitcast float %tmp646 to i32
   %tmp664 = bitcast float %tmp648 to i32
   %tmp665 = bitcast float 0.000000e+00 to i32
   %tmp666 = insertelement <4 x i32> undef, i32 %tmp663, i32 0
   %tmp667 = insertelement <4 x i32> %tmp666, i32 %tmp664, i32 1
   %tmp668 = insertelement <4 x i32> %tmp667, i32 %tmp665, i32 2
   %tmp669 = insertelement <4 x i32> %tmp668, i32 undef, i32 3
-  %tmp670 = call <4 x float> @llvm.SI.image.sample.l.v4i32(<4 x i32> %tmp669, <8 x i32> %tmp126, <4 x i32> %tmp128.bc, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %tmp669.bc = bitcast <4 x i32> %tmp669 to <4 x float>
+  %tmp670 = call <4 x float> @llvm.amdgcn.image.sample.l.v4f32.v4f32.v8i32(<4 x float> %tmp669.bc, <8 x i32> %tmp126, <4 x i32> %tmp128.bc, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
   %tmp671 = extractelement <4 x float> %tmp670, i32 0
   %tmp672 = extractelement <4 x float> %tmp670, i32 1
   %tmp673 = fsub float -0.000000e+00, %tmp662
   %tmp674 = fadd float 1.000000e+00, %tmp673
   %tmp675 = fsub float -0.000000e+00, %tmp672
   %tmp676 = fadd float 1.000000e+00, %tmp675
   %tmp677 = fmul float %tmp674, 2.500000e-01
   %tmp678 = fmul float %tmp676, 2.500000e-01

 declare float @llvm.amdgcn.cubetc(float, float, float) #1
 declare float @llvm.amdgcn.cubema(float, float, float) #1
 declare float @llvm.amdgcn.interp.p1(float, i32, i32, i32) #1
 declare float @llvm.amdgcn.interp.p2(float, float, i32, i32, i32) #1
 declare i32 @llvm.amdgcn.mbcnt.lo(i32, i32) #1
 declare i32 @llvm.amdgcn.mbcnt.hi(i32, i32) #1
 declare void @llvm.amdgcn.exp.compr.v2f16(i32, i32, <2 x half>, <2 x half>, i1, i1) #0
 declare <2 x half> @llvm.amdgcn.cvt.pkrtz(float, float) #1
-
-declare <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #1
-declare <4 x float> @llvm.SI.image.sample.l.v4i32(<4 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #1
-declare <4 x float> @llvm.SI.image.sample.d.v8i32(<8 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #1
-declare <4 x float> @llvm.SI.image.sample.v4i32(<4 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #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) #2
+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) #2
+declare <4 x float> @llvm.amdgcn.image.sample.d.v4f32.v8f32.v8i32(<8 x float>, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #2
+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) #2
 declare float @llvm.SI.load.const(<16 x i8>, i32) #1
 
 attributes #0 = { nounwind }
 attributes #1 = { nounwind readnone }
+attributes #2 = { nounwind readonly }
 
 !0 = !{!1, !1, i64 0, i32 1}
 !1 = !{!"const", !2}
 !2 = !{!"tbaa root"}

test/CodeGen/AMDGPU/skip-if-dead.ll

 ; CHECK-NOT: branch
 
 ; CHECK: BB{{[0-9]+_[0-9]+}}: ; %bb8
 ; CHECK: buffer_store_dword
 
 ; CHECK: [[END]]:
 ; CHECK: s_or_b64 exec, exec
 ; CHECK: s_endpgm
-define amdgpu_ps void @if_after_kill_block(float %arg, float %arg1, <4 x i32> %arg2) #0 {
+define amdgpu_ps void @if_after_kill_block(float %arg, float %arg1, <4 x float> %arg2) #0 {
 bb:
   %tmp = fcmp ult float %arg1, 0.000000e+00
   br i1 %tmp, label %bb3, label %bb4
 
 bb3:                                              ; preds = %bb
   call void @llvm.AMDGPU.kill(float %arg)
   br label %bb4
 
 bb4:                                              ; preds = %bb3, %bb
-  %tmp5 = call <4 x float> @llvm.SI.image.sample.c.v4i32(<4 x i32> %arg2, <8 x i32> undef, <4 x i32> undef, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %tmp5 = call <4 x float> @llvm.amdgcn.image.sample.c.v4f32.v4f32.v8i32(<4 x float> %arg2, <8 x i32> undef, <4 x i32> undef, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
   %tmp6 = extractelement <4 x float> %tmp5, i32 0
   %tmp7 = fcmp une float %tmp6, 0.000000e+00
   br i1 %tmp7, label %bb8, label %bb9
 
 bb8:                                              ; preds = %bb9, %bb4
   store volatile i32 9, i32 addrspace(1)* undef
   ret void
 
 bb9:                                              ; preds = %bb4
   ret void
 }
 
+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 void @llvm.AMDGPU.kill(float) #0
-declare <4 x float> @llvm.SI.image.sample.c.v4i32(<4 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #1
-declare void @llvm.amdgcn.buffer.store.f32(float, <4 x i32>, i32, i32, i1, i1) nounwind
 
 attributes #0 = { nounwind }
-attributes #1 = { nounwind readnone }
+attributes #1 = { nounwind readonly }

test/CodeGen/AMDGPU/split-smrd.ll

bb2:                                              ; preds = %bb
   unreachable
 
 bb3:                                              ; preds = %bb
   %tmp4 = bitcast float %tmp to i32
   %tmp5 = add i32 %tmp4, 4
   %tmp6 = sext i32 %tmp5 to i64
   %tmp7 = getelementptr [34 x <8 x i32>], [34 x <8 x i32>] addrspace(2)* %arg, i64 0, i64 %tmp6
   %tmp8 = load <8 x i32>, <8 x i32> addrspace(2)* %tmp7, align 32, !tbaa !0
-  %tmp9 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> <i32 1061158912, i32 1048576000>, <8 x i32> %tmp8, <4 x i32> undef, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %tmp9 = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v2f32.v8i32(<2 x float> <float bitcast (i32 1061158912 to float), float bitcast (i32 1048576000 to float)>, <8 x i32> %tmp8, <4 x i32> undef, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
   %tmp10 = extractelement <4 x float> %tmp9, i32 0
   %tmp12 = call <2 x half> @llvm.amdgcn.cvt.pkrtz(float %tmp10, float undef)
   call void @llvm.amdgcn.exp.compr.v2f16(i32 0, i32 15, <2 x half> %tmp12, <2 x half> undef, i1 true, i1 true) #0
   ret void
 }
 
-declare void @llvm.amdgcn.exp.compr.v2f16(i32, i32, <2 x half>, <2 x half>, i1, i1) #0
 declare <2 x half> @llvm.amdgcn.cvt.pkrtz(float, float) #1
-
+declare void @llvm.amdgcn.exp.compr.v2f16(i32, i32, <2 x half>, <2 x half>, i1, i1) #0
+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) #2
 declare float @llvm.SI.load.const(<16 x i8>, i32) #1
-declare <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #1
 
 attributes #0 = { nounwind }
 attributes #1 = { nounwind readnone }
+attributes #2 = { nounwind readonly }
 
 !0 = !{!1, !1, i64 0, i32 1}
 !1 = !{!"const", !2}
 !2 = !{!"tbaa root"}

test/CodeGen/AMDGPU/subreg-coalescer-crash.ll

 ; RUN: llc -march=amdgcn -verify-machineinstrs -o - %s | FileCheck -check-prefix=GCN %s
 ; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs -o - %s | FileCheck -check-prefix=GCN %s
 
 ; GCN-LABEL:{{^}}row_filter_C1_D0:
-define void @row_filter_C1_D0() {
+define void @row_filter_C1_D0() #0 {
 entry:
   br i1 undef, label %for.inc.1, label %do.body.preheader
 
 do.body.preheader:                                ; preds = %entry
   %tmp = insertelement <4 x i32> zeroinitializer, i32 undef, i32 1
   br i1 undef, label %do.body56.1, label %do.body90
 
 do.body90:                                        ; preds = %do.body56.2, %do.body56.1, %do.body.preheader

 bb6:                                              ; preds = %bb1
   br i1 undef, label %bb7, label %bb4
 
 bb7:                                              ; preds = %bb6
   %tmp8 = fmul float undef, undef
   br label %bb4
 
 bb9:                                              ; preds = %bb2
-  %tmp10 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> undef, <8 x i32> undef, <4 x i32> undef, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %tmp10 = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v2f32.v8i32(<2 x float> undef, <8 x i32> undef, <4 x i32> undef, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
   %tmp11 = extractelement <4 x float> %tmp10, i32 1
   %tmp12 = extractelement <4 x float> %tmp10, i32 3
   br label %bb14
 
 bb13:                                             ; preds = %bb2
   br i1 undef, label %bb23, label %bb24
 
 bb14:                                             ; preds = %bb27, %bb24, %bb9

 
 bb26:                                             ; preds = %bb23
   br label %bb24
 
 bb27:                                             ; preds = %bb24
   br label %bb14
 }
 
+
 declare void @llvm.amdgcn.exp.f32(i32, i32, float, float, float, float, i1, i1) #0
-declare <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #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
 
 attributes #0 = { nounwind }
-attributes #1 = { nounwind readnone }
+attributes #1 = { nounwind readonly }

test/CodeGen/AMDGPU/undefined-subreg-liverange.ll

 ; overwritten with defined copy
 ; CHECK-LABEL: {{^}}valley_partially_undef_copy:
 define amdgpu_ps float @valley_partially_undef_copy() #0 {
 bb:
   %tmp = load volatile i32, i32 addrspace(1)* undef, align 4
   %tmp1 = load volatile i32, i32 addrspace(1)* undef, align 4
   %tmp2 = insertelement <4 x i32> undef, i32 %tmp1, i32 0
   %tmp3 = insertelement <4 x i32> %tmp2, i32 %tmp1, i32 1
-  %tmp4 = call <4 x float> @llvm.SI.image.sample.v4i32(<4 x i32> %tmp3, <8 x i32> undef, <4 x i32> undef, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %tmp3.cast = bitcast <4 x i32> %tmp3 to <4 x float>
+  %tmp4 = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> %tmp3.cast, <8 x i32> undef, <4 x i32> undef, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
   %tmp5 = extractelement <4 x float> %tmp4, i32 0
   %tmp6 = fmul float %tmp5, undef
   %tmp7 = fadd float %tmp6, %tmp6
   %tmp8 = insertelement <4 x i32> %tmp2, i32 %tmp, i32 1
   store <4 x i32> %tmp8, <4 x i32> addrspace(1)* undef, align 16
   store float %tmp7, float addrspace(1)* undef, align 4
   br label %bb9
 
define void @partially_undef_copy() #0 {
   %partially.undef.0 = insertelement <4 x i32> undef, i32 %tmp0, i32 0
   %partially.undef.1 = insertelement <4 x i32> %partially.undef.0, i32 %tmp1, i32 0
 
   store volatile <4 x i32> %partially.undef.1, <4 x i32> addrspace(1)* undef, align 16
   tail call void asm sideeffect "v_nop", "v={VGPR5_VGPR6_VGPR7_VGPR8}"(<4 x i32> %partially.undef.0)
   ret void
 }
 
-declare <4 x float> @llvm.SI.image.sample.v4i32(<4 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #1
-declare float @llvm.SI.image.sample.i32(i32, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #1
+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
 
 attributes #0 = { nounwind }
-attributes #1 = { nounwind readnone }
+attributes #1 = { nounwind readonly }

test/CodeGen/AMDGPU/unigine-liveness-crash.ll

 define amdgpu_ps <{ i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, float, float, float, float, float, float, float, float, float, float, float, float, float, float }> @main([17 x <16 x i8>] addrspace(2)* byval dereferenceable(18446744073709551615) %arg, [16 x <16 x i8>] addrspace(2)* byval dereferenceable(18446744073709551615) %arg1, [32 x <8 x i32>] addrspace(2)* byval dereferenceable(18446744073709551615) %arg2, [16 x <8 x i32>] addrspace(2)* byval dereferenceable(18446744073709551615) %arg3, [16 x <4 x i32>] addrspace(2)* byval dereferenceable(18446744073709551615) %arg4, float inreg %arg5, i32 inreg %arg6, <2 x i32> %arg7, <2 x i32> %arg8, <2 x i32> %arg9, <3 x i32> %arg10, <2 x i32> %arg11, <2 x i32> %arg12, <2 x i32> %arg13, float %arg14, float %arg15, float %arg16, float %arg17, float %arg18, i32 %arg19, i32 %arg20, float %arg21, i32 %arg22) #0 {
 main_body:
   %i.i = extractelement <2 x i32> %arg8, i32 0
   %j.i = extractelement <2 x i32> %arg8, i32 1
   %i.f.i = bitcast i32 %i.i to float
   %j.f.i = bitcast i32 %j.i to float
   %p1.i = call float @llvm.amdgcn.interp.p1(float %i.f.i, i32 3, i32 4, i32 %arg6) #2
   %p2.i = call float @llvm.amdgcn.interp.p2(float %p1.i, float %j.f.i, i32 3, i32 4, i32 %arg6) #2
-  %tmp23 = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> undef, <8 x i32> undef, <4 x i32> undef, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %tmp23 = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v2f32.v8i32(<2 x float> undef, <8 x i32> undef, <4 x i32> undef, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
+
   %tmp24 = extractelement <4 x float> %tmp23, i32 3
   %tmp25 = fmul float %tmp24, undef
   %tmp26 = fmul float undef, %p2.i
   %tmp27 = fadd float %tmp26, undef
   %tmp28 = bitcast float %tmp27 to i32
   %tmp29 = insertelement <4 x i32> undef, i32 %tmp28, i32 0
   %tmp30 = insertelement <4 x i32> %tmp29, i32 0, i32 1
   %tmp31 = insertelement <4 x i32> %tmp30, i32 undef, i32 2
-  %tmp32 = call <4 x float> @llvm.SI.image.sample.c.v4i32(<4 x i32> %tmp31, <8 x i32> undef, <4 x i32> undef, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %tmp31.cast = bitcast <4 x i32> %tmp31 to <4 x float>
+  %tmp32 = call <4 x float> @llvm.amdgcn.image.sample.c.v4f32.v4f32.v8i32(<4 x float> %tmp31.cast, <8 x i32> undef, <4 x i32> undef, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
   %tmp33 = extractelement <4 x float> %tmp32, i32 0
   %tmp34 = fadd float undef, %tmp33
   %tmp35 = fadd float %tmp34, undef
   %tmp36 = fadd float %tmp35, undef
   %tmp37 = fadd float %tmp36, undef
   %tmp38 = fadd float %tmp37, undef
-  %tmp39 = call <4 x float> @llvm.SI.image.sample.v4i32(<4 x i32> undef, <8 x i32> undef, <4 x i32> undef, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %tmp39 = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> undef, <8 x i32> undef, <4 x i32> undef, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
   %tmp40 = extractelement <4 x float> %tmp39, i32 0
   %tmp41 = extractelement <4 x float> %tmp39, i32 1
   %tmp42 = extractelement <4 x float> %tmp39, i32 2
   %tmp43 = extractelement <4 x float> %tmp39, i32 3
   %tmp44 = fmul float %tmp40, undef
   %tmp45 = fmul float %tmp41, undef
   %tmp46 = fmul float %tmp42, undef
   %tmp47 = fmul float %tmp43, undef
   %tmp48 = fadd float undef, %tmp44
   %tmp49 = fadd float undef, %tmp45
   %tmp50 = bitcast float %tmp27 to i32
   %tmp51 = bitcast float %tmp48 to i32
   %tmp52 = bitcast float %tmp49 to i32
   %tmp53 = insertelement <4 x i32> undef, i32 %tmp50, i32 0
   %tmp54 = insertelement <4 x i32> %tmp53, i32 %tmp51, i32 1
   %tmp55 = insertelement <4 x i32> %tmp54, i32 %tmp52, i32 2
-  %tmp56 = call <4 x float> @llvm.SI.image.sample.c.v4i32(<4 x i32> %tmp55, <8 x i32> undef, <4 x i32> undef, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %tmp55.cast = bitcast <4 x i32> %tmp55 to <4 x float>
+  %tmp56 = call <4 x float> @llvm.amdgcn.image.sample.c.v4f32.v4f32.v8i32(<4 x float> %tmp55.cast, <8 x i32> undef, <4 x i32> undef, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
   %tmp57 = extractelement <4 x float> %tmp56, i32 0
   %tmp58 = fadd float %tmp38, %tmp57
   %tmp59 = fadd float undef, %tmp46
   %tmp60 = fadd float undef, %tmp47
   %tmp61 = bitcast float %tmp59 to i32
   %tmp62 = bitcast float %tmp60 to i32
   %tmp63 = insertelement <4 x i32> undef, i32 %tmp61, i32 1
   %tmp64 = insertelement <4 x i32> %tmp63, i32 %tmp62, i32 2
-  %tmp65 = call <4 x float> @llvm.SI.image.sample.c.v4i32(<4 x i32> %tmp64, <8 x i32> undef, <4 x i32> undef, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %tmp64.cast = bitcast <4 x i32> %tmp64 to <4 x float>
+  %tmp65 = call <4 x float> @llvm.amdgcn.image.sample.c.v4f32.v4f32.v8i32(<4 x float> %tmp64.cast, <8 x i32> undef, <4 x i32> undef, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
   %tmp66 = extractelement <4 x float> %tmp65, i32 0
   %tmp67 = fadd float %tmp58, %tmp66
   %tmp68 = fmul float %tmp67, 1.250000e-01
   %tmp69 = fmul float %tmp68, undef
   %tmp70 = fcmp une float %tmp69, 0.000000e+00
   br i1 %tmp70, label %IF26, label %ENDIF25
 
 IF26:                                             ; preds = %main_body

 
 IF29:                                             ; preds = %LOOP
   %tmp84 = fmul float %.5, 3.125000e-02
   br label %ENDIF25
 
 ENDIF28:                                          ; preds = %LOOP
   %tmp85 = insertelement <4 x i32> %tmp72, i32 undef, i32 1
   %tmp86 = insertelement <4 x i32> %tmp85, i32 undef, i32 2
-  %tmp87 = call <4 x float> @llvm.SI.image.sample.c.v4i32(<4 x i32> %tmp86, <8 x i32> undef, <4 x i32> undef, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %tmp86.cast = bitcast <4 x i32> %tmp86 to <4 x float>
+  %tmp87 = call <4 x float> @llvm.amdgcn.image.sample.c.v4f32.v4f32.v8i32(<4 x float> %tmp86.cast, <8 x i32> undef, <4 x i32> undef, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false)
   %tmp88 = extractelement <4 x float> %tmp87, i32 0
   %tmp89 = fadd float undef, %tmp88
   br label %LOOP
 }
 
-; Function Attrs: nounwind readnone
-declare <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #1
-
-; Function Attrs: nounwind readnone
-declare <4 x float> @llvm.SI.image.sample.v4i32(<4 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #1
-
-; Function Attrs: nounwind readnone
-declare <4 x float> @llvm.SI.image.sample.c.v4i32(<4 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #1
-
-; Function Attrs: nounwind readnone
-declare float @llvm.amdgcn.interp.p1(float, i32, i32, i32) #1
-
-; Function Attrs: nounwind readnone
-declare float @llvm.amdgcn.interp.p2(float, float, i32, i32, i32) #1
-
-; Function Attrs: nounwind readnone
 declare float @llvm.minnum.f32(float, float) #1
-
-; Function Attrs: nounwind readnone
 declare float @llvm.maxnum.f32(float, float) #1
+declare float @llvm.amdgcn.interp.p1(float, i32, i32, i32) #1
+declare float @llvm.amdgcn.interp.p2(float, float, i32, i32, i32) #1
+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) #2
+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) #2
+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) #2
 
 attributes #0 = { nounwind "InitialPSInputAddr"="36983" "target-cpu"="tonga" }
 attributes #1 = { nounwind readnone }
-attributes #2 = { nounwind }
+attributes #2 = { nounwind readonly }
+attributes #3 = { nounwind }

test/CodeGen/AMDGPU/wqm.ll

 }
 
 ; Check that WQM is triggered by image samples and left untouched for loads...
 ;
 ;CHECK-LABEL: {{^}}test2:
 ;CHECK-NEXT: ; %main_body
 ;CHECK-NEXT: s_wqm_b64 exec, exec
 ;CHECK-NOT: exec
-define amdgpu_ps void @test2(<8 x i32> inreg %rsrc, <4 x i32> inreg %sampler, float addrspace(1)* inreg %ptr, <4 x i32> %c) {
+define amdgpu_ps void @test2(<8 x i32> inreg %rsrc, <4 x i32> inreg %sampler, float addrspace(1)* inreg %ptr, <4 x float> %c) {
 main_body:
-  %c.1 = call <4 x float> @llvm.SI.image.sample.v4i32(<4 x i32> %c, <8 x i32> %rsrc, <4 x i32> %sampler, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %c.1 = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> %c, <8 x i32> %rsrc, <4 x i32> %sampler, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false) #0
   %c.2 = bitcast <4 x float> %c.1 to <4 x i32>
   %c.3 = extractelement <4 x i32> %c.2, i32 0
   %gep = getelementptr float, float addrspace(1)* %ptr, i32 %c.3
   %data = load float, float addrspace(1)* %gep
   call void @llvm.amdgcn.exp.f32(i32 0, i32 15, float %data, float undef, float undef, float undef, i1 true, i1 true) #1
   ret void
 }
 
 ; ... but disabled for stores (and, in this simple case, not re-enabled).
 ;
 ;CHECK-LABEL: {{^}}test3:
 ;CHECK-NEXT: ; %main_body
 ;CHECK-NEXT: s_mov_b64 [[ORIG:s\[[0-9]+:[0-9]+\]]], exec
 ;CHECK-NEXT: s_wqm_b64 exec, exec
 ;CHECK: s_and_b64 exec, exec, [[ORIG]]
 ;CHECK: image_sample
 ;CHECK: store
 ;CHECK-NOT: exec
 ;CHECK: .size test3
-define amdgpu_ps <4 x float> @test3(<8 x i32> inreg %rsrc, <4 x i32> inreg %sampler, <4 x i32> %c) {
+define amdgpu_ps <4 x float> @test3(<8 x i32> inreg %rsrc, <4 x i32> inreg %sampler, <4 x float> %c) {
 main_body:
-  %tex = call <4 x float> @llvm.SI.image.sample.v4i32(<4 x i32> %c, <8 x i32> %rsrc, <4 x i32> %sampler, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %tex = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> %c, <8 x i32> %rsrc, <4 x i32> %sampler, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false) #0
   %tex.1 = bitcast <4 x float> %tex to <4 x i32>
   %tex.2 = extractelement <4 x i32> %tex.1, i32 0
 
   call void @llvm.amdgcn.buffer.store.v4f32(<4 x float> %tex, <4 x i32> undef, i32 %tex.2, i32 0, i1 0, i1 0)
 
   ret <4 x float> %tex
 }
 

 ;CHECK: s_wqm_b64 exec, exec
 ;CHECK: image_sample
 ;CHECK: image_sample
 define amdgpu_ps <4 x float> @test4(<8 x i32> inreg %rsrc, <4 x i32> inreg %sampler, float addrspace(1)* inreg %ptr, i32 %c, i32 %d, float %data) {
 main_body:
   %c.1 = mul i32 %c, %d
 
   call void @llvm.amdgcn.buffer.store.v4f32(<4 x float> undef, <4 x i32> undef, i32 %c.1, i32 0, i1 0, i1 0)
-
-  %tex = call <4 x float> @llvm.SI.image.sample.i32(i32 %c.1, <8 x i32> %rsrc, <4 x i32> %sampler, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
-  %tex.1 = bitcast <4 x float> %tex to <4 x i32>
-  %dtex = call <4 x float> @llvm.SI.image.sample.v4i32(<4 x i32> %tex.1, <8 x i32> %rsrc, <4 x i32> %sampler, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %c.1.bc = bitcast i32 %c.1 to float
+  %tex = call <4 x float> @llvm.amdgcn.image.sample.v4f32.f32.v8i32(float %c.1.bc, <8 x i32> %rsrc, <4 x i32> %sampler, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false) #0
+  %dtex = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> %tex, <8 x i32> %rsrc, <4 x i32> %sampler, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false) #0
   ret <4 x float> %dtex
 }
 
 ; Check a case of one branch of an if-else requiring WQM, the other requiring
 ; exact.
 ;
 ; Note: In this particular case, the save-and-restore could be avoided if the
 ; analysis understood that the two branches of the if-else are mutually

 ;CHECK: image_sample
 ;CHECK: image_sample
 define amdgpu_ps float @test_control_flow_0(<8 x i32> inreg %rsrc, <4 x i32> inreg %sampler, i32 %c, i32 %z, float %data) {
 main_body:
   %cmp = icmp eq i32 %z, 0
   br i1 %cmp, label %IF, label %ELSE
 
 IF:
-  %tex = call <4 x float> @llvm.SI.image.sample.i32(i32 %c, <8 x i32> %rsrc, <4 x i32> %sampler, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
-  %tex.1 = bitcast <4 x float> %tex to <4 x i32>
-  %dtex = call <4 x float> @llvm.SI.image.sample.v4i32(<4 x i32> %tex.1, <8 x i32> %rsrc, <4 x i32> %sampler, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %c.bc = bitcast i32 %c to float
+  %tex = call <4 x float> @llvm.amdgcn.image.sample.v4f32.f32.v8i32(float %c.bc, <8 x i32> %rsrc, <4 x i32> %sampler, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false) #0
+  %dtex = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> %tex, <8 x i32> %rsrc, <4 x i32> %sampler, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false) #0
   %data.if = extractelement <4 x float> %dtex, i32 0
   br label %END
 
 ELSE:
   call void @llvm.amdgcn.buffer.store.f32(float %data, <4 x i32> undef, i32 %c, i32 0, i1 0, i1 0)
   br label %END
 
 END:

 ;CHECK: v_mov_b32_e32 v0
 ;CHECK: ; return
 define amdgpu_ps float @test_control_flow_1(<8 x i32> inreg %rsrc, <4 x i32> inreg %sampler, i32 %c, i32 %z, float %data) {
 main_body:
   %cmp = icmp eq i32 %z, 0
   br i1 %cmp, label %ELSE, label %IF
 
 IF:
-  %tex = call <4 x float> @llvm.SI.image.sample.i32(i32 %c, <8 x i32> %rsrc, <4 x i32> %sampler, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
-  %tex.1 = bitcast <4 x float> %tex to <4 x i32>
-  %dtex = call <4 x float> @llvm.SI.image.sample.v4i32(<4 x i32> %tex.1, <8 x i32> %rsrc, <4 x i32> %sampler, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %c.bc = bitcast i32 %c to float
+  %tex = call <4 x float> @llvm.amdgcn.image.sample.v4f32.f32.v8i32(float %c.bc, <8 x i32> %rsrc, <4 x i32> %sampler, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false) #0
+  %dtex = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> %tex, <8 x i32> %rsrc, <4 x i32> %sampler, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false) #0
   %data.if = extractelement <4 x float> %dtex, i32 0
   br label %END
 
 ELSE:
   call void @llvm.amdgcn.buffer.store.f32(float %data, <4 x i32> undef, i32 %c, i32 0, i1 0, i1 0)
   br label %END
 
 END:
IF:
   br label %END
 
 ELSE:
   %coord.ELSE = mul i32 %coord, 4
   br label %END
 
 END:
   %coord.END = phi i32 [ %coord.IF, %IF ], [ %coord.ELSE, %ELSE ]
-  %tex = call <4 x float> @llvm.SI.image.sample.i32(i32 %coord.END, <8 x i32> %rsrc, <4 x i32> %sampler, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %coord.END.bc = bitcast i32 %coord.END to float
+  %tex = call <4 x float> @llvm.amdgcn.image.sample.v4f32.f32.v8i32(float %coord.END.bc, <8 x i32> %rsrc, <4 x i32> %sampler, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false) #0
   ret <4 x float> %tex
 }
 
 ; ... but only if they really do need it.
 ;
 ;CHECK-LABEL: {{^}}test_control_flow_3:
 ;CHECK-NEXT: ; %main_body
 ;CHECK-NEXT: s_mov_b64 [[ORIG:s\[[0-9]+:[0-9]+\]]], exec
 ;CHECK-NEXT: s_wqm_b64 exec, exec
 ;CHECK: image_sample
 ;CHECK: s_and_b64 exec, exec, [[ORIG]]
 ;CHECK: image_sample
 ;CHECK: v_cmp
 ;CHECK: store
-define amdgpu_ps float @test_control_flow_3(<8 x i32> inreg %rsrc, <4 x i32> inreg %sampler, i32 %idx, i32 %coord) {
+define amdgpu_ps float @test_control_flow_3(<8 x i32> inreg %rsrc, <4 x i32> inreg %sampler, i32 %idx, float %coord) {
 main_body:
-  %tex = call <4 x float> @llvm.SI.image.sample.i32(i32 %coord, <8 x i32> %rsrc, <4 x i32> %sampler, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
-  %tex.1 = bitcast <4 x float> %tex to <4 x i32>
-  %dtex = call <4 x float> @llvm.SI.image.sample.v4i32(<4 x i32> %tex.1, <8 x i32> %rsrc, <4 x i32> %sampler, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %tex = call <4 x float> @llvm.amdgcn.image.sample.v4f32.f32.v8i32(float %coord, <8 x i32> %rsrc, <4 x i32> %sampler, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false) #0
+  %dtex = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> %tex, <8 x i32> %rsrc, <4 x i32> %sampler, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false) #0
   %dtex.1 = extractelement <4 x float> %dtex, i32 0
-
   call void @llvm.amdgcn.buffer.store.f32(float %dtex.1, <4 x i32> undef, i32 %idx, i32 0, i1 0, i1 0)
 
   %cc = fcmp ogt float %dtex.1, 0.0
   br i1 %cc, label %IF, label %ELSE
 
 IF:
   %tex.IF = fmul float %dtex.1, 3.0
   br label %END

 ;CHECK: %IF
 ;CHECK: s_and_saveexec_b64 [[SAVE:s\[[0-9]+:[0-9]+\]]],  [[ORIG]]
 ;CHECK: load
 ;CHECK: store
 ;CHECK: s_mov_b64 exec, [[SAVE]]
 ;CHECK: %END
 ;CHECK: image_sample
 ;CHECK: image_sample
-define amdgpu_ps <4 x float> @test_control_flow_4(<8 x i32> inreg %rsrc, <4 x i32> inreg %sampler, i32 %coord, i32 %y, float %z) {
+define amdgpu_ps <4 x float> @test_control_flow_4(<8 x i32> inreg %rsrc, <4 x i32> inreg %sampler, float %coord, i32 %y, float %z) {
 main_body:
   %cond = icmp eq i32 %y, 0
   br i1 %cond, label %IF, label %END
 
 IF:
   %data = call float @llvm.amdgcn.buffer.load.f32(<4 x i32> undef, i32 0, i32 0, i1 0, i1 0)
   call void @llvm.amdgcn.buffer.store.f32(float %data, <4 x i32> undef, i32 1, i32 0, i1 0, i1 0)
   br label %END
 
 END:
-  %tex = call <4 x float> @llvm.SI.image.sample.i32(i32 %coord, <8 x i32> %rsrc, <4 x i32> %sampler, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
-  %tex.1 = bitcast <4 x float> %tex to <4 x i32>
-  %dtex = call <4 x float> @llvm.SI.image.sample.v4i32(<4 x i32> %tex.1, <8 x i32> %rsrc, <4 x i32> %sampler, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %tex = call <4 x float> @llvm.amdgcn.image.sample.v4f32.f32.v8i32(float %coord, <8 x i32> %rsrc, <4 x i32> %sampler, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false) #0
+  %dtex = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> %tex, <8 x i32> %rsrc, <4 x i32> %sampler, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false) #0
   ret <4 x float> %dtex
 }
 
 ; Kill is performed in WQM mode so that uniform kill behaves correctly ...
 ;
 ;CHECK-LABEL: {{^}}test_kill_0:
 ;CHECK-NEXT: ; %main_body
 ;CHECK-NEXT: s_mov_b64 [[ORIG:s\[[0-9]+:[0-9]+\]]], exec
 ;CHECK-NEXT: s_wqm_b64 exec, exec
 ;CHECK: s_and_b64 exec, exec, [[ORIG]]
 ;CHECK: image_sample
 ;CHECK: buffer_store_dword
 ;CHECK: s_wqm_b64 exec, exec
 ;CHECK: v_cmpx_
 ;CHECK: s_and_saveexec_b64 [[SAVE:s\[[0-9]+:[0-9]+\]]], [[ORIG]]
 ;CHECK: buffer_store_dword
 ;CHECK: s_mov_b64 exec, [[SAVE]]
 ;CHECK: image_sample
-define amdgpu_ps <4 x float> @test_kill_0(<8 x i32> inreg %rsrc, <4 x i32> inreg %sampler, float addrspace(1)* inreg %ptr, <2 x i32> %idx, <2 x float> %data, i32 %coord, i32 %coord2, float %z) {
+define amdgpu_ps <4 x float> @test_kill_0(<8 x i32> inreg %rsrc, <4 x i32> inreg %sampler, float addrspace(1)* inreg %ptr, <2 x i32> %idx, <2 x float> %data, float %coord, float %coord2, float %z) {
 main_body:
-  %tex = call <4 x float> @llvm.SI.image.sample.i32(i32 %coord, <8 x i32> %rsrc, <4 x i32> %sampler, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
-
+  %tex = call <4 x float> @llvm.amdgcn.image.sample.v4f32.f32.v8i32(float %coord, <8 x i32> %rsrc, <4 x i32> %sampler, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false) #0
   %idx.0 = extractelement <2 x i32> %idx, i32 0
   %data.0 = extractelement <2 x float> %data, i32 0
   call void @llvm.amdgcn.buffer.store.f32(float %data.0, <4 x i32> undef, i32 %idx.0, i32 0, i1 0, i1 0)
 
   call void @llvm.AMDGPU.kill(float %z)
 
   %idx.1 = extractelement <2 x i32> %idx, i32 1
   %data.1 = extractelement <2 x float> %data, i32 1
   call void @llvm.amdgcn.buffer.store.f32(float %data.1, <4 x i32> undef, i32 %idx.1, i32 0, i1 0, i1 0)
-
-  %tex2 = call <4 x float> @llvm.SI.image.sample.i32(i32 %coord2, <8 x i32> %rsrc, <4 x i32> %sampler, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
-  %tex2.1 = bitcast <4 x float> %tex2 to <4 x i32>
-  %dtex = call <4 x float> @llvm.SI.image.sample.v4i32(<4 x i32> %tex2.1, <8 x i32> %rsrc, <4 x i32> %sampler, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %tex2 = call <4 x float> @llvm.amdgcn.image.sample.v4f32.f32.v8i32(float %coord2, <8 x i32> %rsrc, <4 x i32> %sampler, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false) #0
+  %dtex = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> %tex2, <8 x i32> %rsrc, <4 x i32> %sampler, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false) #0
   %out = fadd <4 x float> %tex, %dtex
 
   ret <4 x float> %out
 }
 
 ; ... but only if WQM is necessary.
 ;
 ; CHECK-LABEL: {{^}}test_kill_1:
 ; CHECK-NEXT: ; %main_body
 ; CHECK: s_mov_b64 [[ORIG:s\[[0-9]+:[0-9]+\]]], exec
 ; CHECK: s_wqm_b64 exec, exec
 ; CHECK: image_sample
 ; CHECK: s_and_b64 exec, exec, [[ORIG]]
 ; CHECK: image_sample
 ; CHECK: buffer_store_dword
 ; CHECK-NOT: wqm
 ; CHECK: v_cmpx_
-define amdgpu_ps <4 x float> @test_kill_1(<8 x i32> inreg %rsrc, <4 x i32> inreg %sampler, i32 %idx, float %data, i32 %coord, i32 %coord2, float %z) {
+define amdgpu_ps <4 x float> @test_kill_1(<8 x i32> inreg %rsrc, <4 x i32> inreg %sampler, i32 %idx, float %data, float %coord, float %coord2, float %z) {
 main_body:
-  %tex = call <4 x float> @llvm.SI.image.sample.i32(i32 %coord, <8 x i32> %rsrc, <4 x i32> %sampler, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
-  %tex.1 = bitcast <4 x float> %tex to <4 x i32>
-  %dtex = call <4 x float> @llvm.SI.image.sample.v4i32(<4 x i32> %tex.1, <8 x i32> %rsrc, <4 x i32> %sampler, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %tex = call <4 x float> @llvm.amdgcn.image.sample.v4f32.f32.v8i32(float %coord, <8 x i32> %rsrc, <4 x i32> %sampler, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false) #0
+  %dtex = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> %tex, <8 x i32> %rsrc, <4 x i32> %sampler, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false) #0
 
   call void @llvm.amdgcn.buffer.store.f32(float %data, <4 x i32> undef, i32 0, i32 0, i1 0, i1 0)
 
   call void @llvm.AMDGPU.kill(float %z)
 
   ret <4 x float> %dtex
 }
 

 
 loop:
   %ctr.iv = phi float [ 0.0, %entry ], [ %ctr.next, %body ]
   %c.iv = phi <4 x float> [ %in, %entry ], [ %c.next, %body ]
   %cc = fcmp ogt float %ctr.iv, 7.0
   br i1 %cc, label %break, label %body
 
 body:
-  %c.i = bitcast <4 x float> %c.iv to <4 x i32>
-  %c.next = call <4 x float> @llvm.SI.image.sample.v4i32(<4 x i32> %c.i, <8 x i32> undef, <4 x i32> undef, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %c.next = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> %c.iv, <8 x i32> undef, <4 x i32> undef, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false) #0
   %ctr.next = fadd float %ctr.iv, 2.0
   br label %loop
 
 break:
   ret <4 x float> %c.iv
 }
 
 ; Only intrinsic stores need exact execution -- other stores do not have
entry:
 
   %s.gep = getelementptr [32 x i32], [32 x i32]* %array, i32 0, i32 0
   store volatile i32 %a, i32* %s.gep, align 4
 
   call void @llvm.amdgcn.buffer.store.f32(float %data, <4 x i32> undef, i32 1, i32 0, i1 0, i1 0)
 
   %c.gep = getelementptr [32 x i32], [32 x i32]* %array, i32 0, i32 %idx
   %c = load i32, i32* %c.gep, align 4
-
-  %t = call <4 x float> @llvm.SI.image.sample.i32(i32 %c, <8 x i32> undef, <4 x i32> undef, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
-
+  %c.bc = bitcast i32 %c to float
+  %t = call <4 x float> @llvm.amdgcn.image.sample.v4f32.f32.v8i32(float %c.bc, <8 x i32> undef, <4 x i32> undef, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false) #0
   call void @llvm.amdgcn.buffer.store.v4f32(<4 x float> %t, <4 x i32> undef, i32 0, i32 0, i1 0, i1 0)
 
   ret void
 }
 
 ; Must return to exact at the end of a non-void returning shader,
 ; otherwise the EXEC mask exported by the epilog will be wrong. This is true
 ; even if the shader has no kills, because a kill could have happened in a
 ; previous shader fragment.
 ;
 ; CHECK-LABEL: {{^}}test_nonvoid_return:
 ; CHECK: s_mov_b64 [[LIVE:s\[[0-9]+:[0-9]+\]]], exec
 ; CHECK: s_wqm_b64 exec, exec
 ;
 ; CHECK: s_and_b64 exec, exec, [[LIVE]]
 ; CHECK-NOT: exec
 define amdgpu_ps <4 x float> @test_nonvoid_return() nounwind {
-  %tex = call <4 x float> @llvm.SI.image.sample.v4i32(<4 x i32> undef, <8 x i32> undef, <4 x i32> undef, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
-  %tex.i = bitcast <4 x float> %tex to <4 x i32>
-  %dtex = call <4 x float> @llvm.SI.image.sample.v4i32(<4 x i32> %tex.i, <8 x i32> undef, <4 x i32> undef, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %tex = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> undef, <8 x i32> undef, <4 x i32> undef, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false) #0
+  %dtex = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> %tex, <8 x i32> undef, <4 x i32> undef, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false) #0
   ret <4 x float> %dtex
 }
 
 ; CHECK-LABEL: {{^}}test_nonvoid_return_unreachable:
 ; CHECK: s_mov_b64 [[LIVE:s\[[0-9]+:[0-9]+\]]], exec
 ; CHECK: s_wqm_b64 exec, exec
 ;
 ; CHECK: s_and_b64 exec, exec, [[LIVE]]
 ; CHECK-NOT: exec
 define amdgpu_ps <4 x float> @test_nonvoid_return_unreachable(i32 inreg %c) nounwind {
 entry:
-  %tex = call <4 x float> @llvm.SI.image.sample.v4i32(<4 x i32> undef, <8 x i32> undef, <4 x i32> undef, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
-  %tex.i = bitcast <4 x float> %tex to <4 x i32>
-  %dtex = call <4 x float> @llvm.SI.image.sample.v4i32(<4 x i32> %tex.i, <8 x i32> undef, <4 x i32> undef, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
-
+  %tex = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> undef, <8 x i32> undef, <4 x i32> undef, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false) #0
+  %dtex = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v4f32.v8i32(<4 x float> %tex, <8 x i32> undef, <4 x i32> undef, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false) #0
   %cc = icmp sgt i32 %c, 0
   br i1 %cc, label %if, label %else
 
 if:
   store volatile <4 x float> %dtex, <4 x float> addrspace(1)* undef
   unreachable
 
 else:

 ; CHECK: image_sample
 ; CHECK: ; %end
 define amdgpu_ps <4 x float> @test_scc(i32 inreg %sel, i32 %idx) #1 {
 main_body:
   %cc = icmp sgt i32 %sel, 0
   br i1 %cc, label %if, label %else
 
 if:
-  %r.if = call <4 x float> @llvm.SI.image.sample.i32(i32 0, <8 x i32> undef, <4 x i32> undef, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %r.if = call <4 x float> @llvm.amdgcn.image.sample.v4f32.f32.v8i32(float 0.0, <8 x i32> undef, <4 x i32> undef, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false) #0
   br label %end
 
 else:
-  %r.else = call <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32> <i32 0, i32 1>, <8 x i32> undef, <4 x i32> undef, i32 15, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0)
+  %r.else = call <4 x float> @llvm.amdgcn.image.sample.v4f32.v2f32.v8i32(<2 x float> <float 0.0, float bitcast (i32 1 to float)>, <8 x i32> undef, <4 x i32> undef, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false) #0
   br label %end
 
 end:
   %r = phi <4 x float> [ %r.if, %if ], [ %r.else, %else ]
-
   call void @llvm.amdgcn.buffer.store.f32(float 1.0, <4 x i32> undef, i32 %idx, i32 0, i1 0, i1 0)
-
   ret <4 x float> %r
 }
 
 declare void @llvm.amdgcn.exp.f32(i32, i32, float, float, float, float, i1, i1) #1
 declare void @llvm.amdgcn.image.store.v4f32.v4i32.v8i32(<4 x float>, <4 x i32>, <8 x i32>, i32, i1, i1, i1, i1) #1
-declare void @llvm.amdgcn.buffer.store.f32(float, <4 x i32>, i32, i32, i1, i1) #1
-declare void @llvm.amdgcn.buffer.store.v4f32(<4 x float>, <4 x i32>, i32, i32, i1, i1) #1
-
-declare <4 x float> @llvm.amdgcn.image.load.v4f32.v4i32.v8i32(<4 x i32>, <8 x i32>, i32, i1, i1, i1, i1) #2
-declare float @llvm.amdgcn.buffer.load.f32(<4 x i32>, i32, i32, i1, i1) #2
-
-declare <4 x float> @llvm.SI.image.sample.i32(i32, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #3
-declare <4 x float> @llvm.SI.image.sample.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #3
-declare <4 x float> @llvm.SI.image.sample.v4i32(<4 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) #3
-
+declare void @llvm.amdgcn.buffer.store.f32(float, <4 x i32>, i32, i32, i1, i1) #2
+declare void @llvm.amdgcn.buffer.store.v4f32(<4 x float>, <4 x i32>, i32, i32, i1, i1) #2
+declare <4 x float> @llvm.amdgcn.image.load.v4f32.v4i32.v8i32(<4 x i32>, <8 x i32>, i32, i1, i1, i1, i1) #3
+declare float @llvm.amdgcn.buffer.load.f32(<4 x i32>, i32, i32, i1, i1) #3
+declare <4 x float> @llvm.amdgcn.image.sample.v4f32.f32.v8i32(float, <8 x i32>, <4 x i32>, i32, i1, i1, i1, i1, i1) #3
+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) #3
+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) #3
 declare void @llvm.AMDGPU.kill(float) #1
 
 attributes #1 = { nounwind }
 attributes #2 = { nounwind readonly }
 attributes #3 = { nounwind readnone }
 attributes #4 = { "amdgpu-ps-wqm-outputs" }