1 ; RUN: opt -mtriple=amdgcn-- -S -separate-const-offset-from-gep -reassociate-geps-verify-no-dead-code -gvn < %s | FileCheck -check-prefix=IR %s
3 target datalayout = "e-p:32:32-p1:64:64-p2:64:64-p3:32:32-p4:64:64-p5:32:32-p24:64:64-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024-v2048:2048-n32:64"
5 @array = internal addrspace(4) constant [4096 x [32 x float]] zeroinitializer, align 4
7 ; IR-LABEL: @sum_of_array(
8 ; IR: [[BASE_PTR:%[a-zA-Z0-9]+]] = getelementptr [4096 x [32 x float]], [4096 x [32 x float]] addrspace(4)* @array, i64 0, i64 %{{[a-zA-Z0-9]+}}, i64 %{{[a-zA-Z0-9]+}}
9 ; IR: getelementptr inbounds float, float addrspace(4)* [[BASE_PTR]], i64 1
10 ; IR: getelementptr inbounds float, float addrspace(4)* [[BASE_PTR]], i64 32
11 ; IR: getelementptr inbounds float, float addrspace(4)* [[BASE_PTR]], i64 33
12 define amdgpu_kernel void @sum_of_array(i32 %x, i32 %y, float addrspace(1)* nocapture %output) {
13 %tmp = sext i32 %y to i64
14 %tmp1 = sext i32 %x to i64
15 %tmp2 = getelementptr inbounds [4096 x [32 x float]], [4096 x [32 x float]] addrspace(4)* @array, i64 0, i64 %tmp1, i64 %tmp
16 %tmp4 = load float, float addrspace(4)* %tmp2, align 4
17 %tmp5 = fadd float %tmp4, 0.000000e+00
19 %tmp7 = sext i32 %tmp6 to i64
20 %tmp8 = getelementptr inbounds [4096 x [32 x float]], [4096 x [32 x float]] addrspace(4)* @array, i64 0, i64 %tmp1, i64 %tmp7
21 %tmp10 = load float, float addrspace(4)* %tmp8, align 4
22 %tmp11 = fadd float %tmp5, %tmp10
23 %tmp12 = add i32 %x, 1
24 %tmp13 = sext i32 %tmp12 to i64
25 %tmp14 = getelementptr inbounds [4096 x [32 x float]], [4096 x [32 x float]] addrspace(4)* @array, i64 0, i64 %tmp13, i64 %tmp
26 %tmp16 = load float, float addrspace(4)* %tmp14, align 4
27 %tmp17 = fadd float %tmp11, %tmp16
28 %tmp18 = getelementptr inbounds [4096 x [32 x float]], [4096 x [32 x float]] addrspace(4)* @array, i64 0, i64 %tmp13, i64 %tmp7
29 %tmp20 = load float, float addrspace(4)* %tmp18, align 4
30 %tmp21 = fadd float %tmp17, %tmp20
31 store float %tmp21, float addrspace(1)* %output, align 4
35 @array2 = internal addrspace(4) constant [4096 x [4 x float]] zeroinitializer, align 4
37 ; Some of the indices go over the maximum mubuf offset, so don't split them.
39 ; IR-LABEL: @sum_of_array_over_max_mubuf_offset(
40 ; IR: [[BASE_PTR:%[a-zA-Z0-9]+]] = getelementptr [4096 x [4 x float]], [4096 x [4 x float]] addrspace(4)* @array2, i64 0, i64 %{{[a-zA-Z0-9]+}}, i64 %{{[a-zA-Z0-9]+}}
41 ; IR: getelementptr inbounds float, float addrspace(4)* [[BASE_PTR]], i64 255
43 ; IR: getelementptr inbounds [4096 x [4 x float]], [4096 x [4 x float]] addrspace(4)* @array2, i64 0, i64 %{{[a-zA-Z0-9]+}}, i64 %{{[a-zA-Z0-9]+}}
44 ; IR: getelementptr inbounds [4096 x [4 x float]], [4096 x [4 x float]] addrspace(4)* @array2, i64 0, i64 %{{[a-zA-Z0-9]+}}, i64 %{{[a-zA-Z0-9]+}}
45 define amdgpu_kernel void @sum_of_array_over_max_mubuf_offset(i32 %x, i32 %y, float addrspace(1)* nocapture %output) {
46 %tmp = sext i32 %y to i64
47 %tmp1 = sext i32 %x to i64
48 %tmp2 = getelementptr inbounds [4096 x [4 x float]], [4096 x [4 x float]] addrspace(4)* @array2, i64 0, i64 %tmp1, i64 %tmp
49 %tmp4 = load float, float addrspace(4)* %tmp2, align 4
50 %tmp5 = fadd float %tmp4, 0.000000e+00
51 %tmp6 = add i32 %y, 255
52 %tmp7 = sext i32 %tmp6 to i64
53 %tmp8 = getelementptr inbounds [4096 x [4 x float]], [4096 x [4 x float]] addrspace(4)* @array2, i64 0, i64 %tmp1, i64 %tmp7
54 %tmp10 = load float, float addrspace(4)* %tmp8, align 4
55 %tmp11 = fadd float %tmp5, %tmp10
56 %tmp12 = add i32 %x, 256
57 %tmp13 = sext i32 %tmp12 to i64
58 %tmp14 = getelementptr inbounds [4096 x [4 x float]], [4096 x [4 x float]] addrspace(4)* @array2, i64 0, i64 %tmp13, i64 %tmp
59 %tmp16 = load float, float addrspace(4)* %tmp14, align 4
60 %tmp17 = fadd float %tmp11, %tmp16
61 %tmp18 = getelementptr inbounds [4096 x [4 x float]], [4096 x [4 x float]] addrspace(4)* @array2, i64 0, i64 %tmp13, i64 %tmp7
62 %tmp20 = load float, float addrspace(4)* %tmp18, align 4
63 %tmp21 = fadd float %tmp17, %tmp20
64 store float %tmp21, float addrspace(1)* %output, align 4
69 @lds_array = internal addrspace(3) global [4096 x [4 x float]] undef, align 4
71 ; DS instructions have a larger immediate offset, so make sure these are OK.
72 ; IR-LABEL: @sum_of_lds_array_over_max_mubuf_offset(
73 ; IR: [[BASE_PTR:%[a-zA-Z0-9]+]] = getelementptr [4096 x [4 x float]], [4096 x [4 x float]] addrspace(3)* @lds_array, i32 0, i32 %{{[a-zA-Z0-9]+}}, i32 %{{[a-zA-Z0-9]+}}
74 ; IR: getelementptr inbounds float, float addrspace(3)* [[BASE_PTR]], i32 255
75 ; IR: getelementptr inbounds float, float addrspace(3)* [[BASE_PTR]], i32 16128
76 ; IR: getelementptr inbounds float, float addrspace(3)* [[BASE_PTR]], i32 16383
77 define amdgpu_kernel void @sum_of_lds_array_over_max_mubuf_offset(i32 %x, i32 %y, float addrspace(1)* nocapture %output) {
78 %tmp2 = getelementptr inbounds [4096 x [4 x float]], [4096 x [4 x float]] addrspace(3)* @lds_array, i32 0, i32 %x, i32 %y
79 %tmp4 = load float, float addrspace(3)* %tmp2, align 4
80 %tmp5 = fadd float %tmp4, 0.000000e+00
81 %tmp6 = add i32 %y, 255
82 %tmp8 = getelementptr inbounds [4096 x [4 x float]], [4096 x [4 x float]] addrspace(3)* @lds_array, i32 0, i32 %x, i32 %tmp6
83 %tmp10 = load float, float addrspace(3)* %tmp8, align 4
84 %tmp11 = fadd float %tmp5, %tmp10
85 %tmp12 = add i32 %x, 4032
86 %tmp14 = getelementptr inbounds [4096 x [4 x float]], [4096 x [4 x float]] addrspace(3)* @lds_array, i32 0, i32 %tmp12, i32 %y
87 %tmp16 = load float, float addrspace(3)* %tmp14, align 4
88 %tmp17 = fadd float %tmp11, %tmp16
89 %tmp18 = getelementptr inbounds [4096 x [4 x float]], [4096 x [4 x float]] addrspace(3)* @lds_array, i32 0, i32 %tmp12, i32 %tmp6
90 %tmp20 = load float, float addrspace(3)* %tmp18, align 4
91 %tmp21 = fadd float %tmp17, %tmp20
92 store float %tmp21, float addrspace(1)* %output, align 4
96 ; IR-LABEL: @keep_metadata(
97 ; IR: getelementptr {{.*}} !amdgpu.uniform
98 ; IR: getelementptr {{.*}} !amdgpu.uniform
99 ; IR: getelementptr {{.*}} !amdgpu.uniform
100 define amdgpu_ps <{ i32, i32, i32, i32, i32, float, float, float, float, float, float, float, float, float, float, float, float, float, float, float }> @keep_metadata([0 x <4 x i32>] addrspace(4)* inreg noalias dereferenceable(18446744073709551615), [0 x <8 x i32>] addrspace(4)* inreg noalias dereferenceable(18446744073709551615), [0 x <4 x i32>] addrspace(4)* inreg noalias dereferenceable(18446744073709551615), [0 x <8 x i32>] addrspace(4)* inreg noalias dereferenceable(18446744073709551615), float inreg, i32 inreg, <2 x i32>, <2 x i32>, <2 x i32>, <3 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, float, float, float, float, float, i32, i32, float, i32) #5 {
102 %22 = call nsz float @llvm.amdgcn.interp.mov(i32 2, i32 0, i32 0, i32 %5) #8
103 %23 = bitcast float %22 to i32
105 %25 = getelementptr [0 x <8 x i32>], [0 x <8 x i32>] addrspace(4)* %1, i32 0, i32 %24, !amdgpu.uniform !0
106 %26 = load <8 x i32>, <8 x i32> addrspace(4)* %25, align 32, !invariant.load !0
109 %29 = bitcast [0 x <8 x i32>] addrspace(4)* %1 to [0 x <4 x i32>] addrspace(4)*
110 %30 = getelementptr [0 x <4 x i32>], [0 x <4 x i32>] addrspace(4)* %29, i32 0, i32 %28, !amdgpu.uniform !0
111 %31 = load <4 x i32>, <4 x i32> addrspace(4)* %30, align 16, !invariant.load !0
112 %32 = call nsz <4 x float> @llvm.amdgcn.image.sample.v4f32.v2f32.v8i32(<2 x float> zeroinitializer, <8 x i32> %26, <4 x i32> %31, i32 15, i1 false, i1 false, i1 false, i1 false, i1 false) #8
113 %33 = extractelement <4 x float> %32, i32 0
114 %34 = extractelement <4 x float> %32, i32 1
115 %35 = extractelement <4 x float> %32, i32 2
116 %36 = extractelement <4 x float> %32, i32 3
117 %37 = bitcast float %4 to i32
118 %38 = insertvalue <{ i32, i32, i32, i32, i32, float, float, float, float, float, float, float, float, float, float, float, float, float, float, float }> undef, i32 %37, 4
119 %39 = insertvalue <{ i32, i32, i32, i32, i32, float, float, float, float, float, float, float, float, float, float, float, float, float, float, float }> %38, float %33, 5
120 %40 = insertvalue <{ i32, i32, i32, i32, i32, float, float, float, float, float, float, float, float, float, float, float, float, float, float, float }> %39, float %34, 6
121 %41 = insertvalue <{ i32, i32, i32, i32, i32, float, float, float, float, float, float, float, float, float, float, float, float, float, float, float }> %40, float %35, 7
122 %42 = insertvalue <{ i32, i32, i32, i32, i32, float, float, float, float, float, float, float, float, float, float, float, float, float, float, float }> %41, float %36, 8
123 %43 = insertvalue <{ i32, i32, i32, i32, i32, float, float, float, float, float, float, float, float, float, float, float, float, float, float, float }> %42, float %20, 19
124 ret <{ i32, i32, i32, i32, i32, float, float, float, float, float, float, float, float, float, float, float, float, float, float, float }> %43
127 ; Function Attrs: nounwind readnone speculatable
128 declare float @llvm.amdgcn.interp.mov(i32, i32, i32, i32) #6
130 ; Function Attrs: nounwind readonly
131 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) #7
136 attributes #5 = { "InitialPSInputAddr"="45175" }
137 attributes #6 = { nounwind readnone speculatable }
138 attributes #7 = { nounwind readonly }
139 attributes #8 = { nounwind readnone }