| blob | 538ce15979de88b1c035022f0ae369a53fb6acbc |
1 ; RUN: llc -mtriple=amdgcn--amdpal -mcpu=gfx1100 -verify-machineinstrs < %s | FileCheck %s
3 ; CHECK: .amdgpu_pal_metadata
4 ; CHECK-NEXT: ---
5 ; CHECK-NEXT: amdpal.pipelines:
6 ; CHECK-NEXT: - .api: Vulkan
7 ; CHECK-NEXT: .compute_registers:
8 ; CHECK-NEXT: .tg_size_en: true
9 ; CHECK-NEXT: .tgid_x_en: false
10 ; CHECK-NEXT: .tgid_y_en: false
11 ; CHECK-NEXT: .tgid_z_en: false
12 ; CHECK-NEXT: .tidig_comp_cnt: 0x1
13 ; CHECK-NEXT: .hardware_stages:
14 ; CHECK-NEXT: .cs:
15 ; CHECK-NEXT: .checksum_value: 0x9444d7d0
16 ; CHECK-NEXT: .debug_mode: 0
17 ; CHECK-NEXT: .excp_en: 0
18 ; CHECK-NEXT: .float_mode: 0xc0
19 ; CHECK-NEXT: .ieee_mode: true
20 ; CHECK-NEXT: .image_op: false
21 ; CHECK-NEXT: .lds_size: 0x200
22 ; CHECK-NEXT: .mem_ordered: true
23 ; CHECK-NEXT: .sgpr_limit: 0x6a
24 ; CHECK-NEXT: .threadgroup_dimensions:
25 ; CHECK-NEXT: - 0x1
26 ; CHECK-NEXT: - 0x400
27 ; CHECK-NEXT: - 0x1
28 ; CHECK-NEXT: .trap_present: false
29 ; CHECK-NEXT: .user_data_reg_map:
30 ; CHECK-NEXT: - 0x10000000
31 ; CHECK-NEXT: - 0xffffffff
32 ; CHECK-NEXT: - 0
33 ; CHECK-NEXT: - 0xffffffff
34 ; CHECK-NEXT: - 0xffffffff
35 ; CHECK-NEXT: - 0xffffffff
36 ; CHECK-NEXT: - 0xffffffff
37 ; CHECK-NEXT: - 0xffffffff
38 ; CHECK-NEXT: - 0xffffffff
39 ; CHECK-NEXT: - 0xffffffff
40 ; CHECK-NEXT: - 0xffffffff
41 ; CHECK-NEXT: - 0xffffffff
42 ; CHECK-NEXT: - 0xffffffff
43 ; CHECK-NEXT: - 0xffffffff
44 ; CHECK-NEXT: - 0xffffffff
45 ; CHECK-NEXT: - 0xffffffff
46 ; CHECK-NEXT: - 0xffffffff
47 ; CHECK-NEXT: - 0xffffffff
48 ; CHECK-NEXT: - 0xffffffff
49 ; CHECK-NEXT: - 0xffffffff
50 ; CHECK-NEXT: - 0xffffffff
51 ; CHECK-NEXT: - 0xffffffff
52 ; CHECK-NEXT: - 0xffffffff
53 ; CHECK-NEXT: - 0xffffffff
54 ; CHECK-NEXT: - 0xffffffff
55 ; CHECK-NEXT: - 0xffffffff
56 ; CHECK-NEXT: - 0xffffffff
57 ; CHECK-NEXT: - 0xffffffff
58 ; CHECK-NEXT: - 0xffffffff
59 ; CHECK-NEXT: - 0xffffffff
60 ; CHECK-NEXT: - 0xffffffff
61 ; CHECK-NEXT: - 0xffffffff
62 ; CHECK-NEXT: .user_sgprs: 0x3
63 ; CHECK-NEXT: .vgpr_limit: 0x100
64 ; CHECK-NEXT: .wavefront_size: 0x40
65 ; CHECK-NEXT: .wgp_mode: true
66 ; CHECK: .registers: {}
67 ; CHECK-NEXT: .shader_functions:
68 ; CHECK-NEXT: dynamic_stack:
69 ; CHECK-NEXT: .backend_stack_size: 0x10
70 ; CHECK-NEXT: .lds_size: 0
71 ; CHECK-NEXT: .sgpr_count: 0x22
72 ; CHECK-NEXT: .stack_frame_size_in_bytes: 0x10
73 ; CHECK-NEXT: .vgpr_count: 0x2
74 ; CHECK-NEXT: dynamic_stack_loop:
75 ; CHECK-NEXT: .backend_stack_size: 0x10
76 ; CHECK-NEXT: .lds_size: 0
77 ; CHECK-NEXT: .sgpr_count: 0x22
78 ; CHECK-NEXT: .stack_frame_size_in_bytes: 0x10
79 ; CHECK-NEXT: .vgpr_count: 0x3
80 ; CHECK-NEXT: multiple_stack:
81 ; CHECK-NEXT: .backend_stack_size: 0x24
82 ; CHECK-NEXT: .lds_size: 0
83 ; CHECK-NEXT: .sgpr_count: 0x21
84 ; CHECK-NEXT: .stack_frame_size_in_bytes: 0x24
85 ; CHECK-NEXT: .vgpr_count: 0x3
86 ; CHECK-NEXT: no_stack:
87 ; CHECK-NEXT: .backend_stack_size: 0
88 ; CHECK-NEXT: .lds_size: 0
89 ; CHECK-NEXT: .sgpr_count: 0x20
90 ; CHECK-NEXT: .stack_frame_size_in_bytes: 0
91 ; CHECK-NEXT: .vgpr_count: 0x1
92 ; CHECK-NEXT: no_stack_call:
93 ; CHECK-NEXT: .backend_stack_size: 0x10
94 ; CHECK-NEXT: .lds_size: 0
95 ; CHECK-NEXT: .sgpr_count: 0x22
96 ; CHECK-NEXT: .stack_frame_size_in_bytes: 0x10
97 ; CHECK-NEXT: .vgpr_count: 0x3
98 ; CHECK-NEXT: no_stack_extern_call:
99 ; CHECK-NEXT: .backend_stack_size: 0x10
100 ; CHECK-NEXT: .lds_size: 0
101 ; CHECK-NEXT: .sgpr_count: 0x29
102 ; CHECK-NEXT: .stack_frame_size_in_bytes: 0x10
103 ; CHECK-NEXT: .vgpr_count: 0x58
104 ; CHECK-NEXT: no_stack_extern_call_many_args:
105 ; CHECK-NEXT: .backend_stack_size: 0x90
106 ; CHECK-NEXT: .lds_size: 0
107 ; CHECK-NEXT: .sgpr_count: 0x29
108 ; CHECK-NEXT: .stack_frame_size_in_bytes: 0x90
109 ; CHECK-NEXT: .vgpr_count: 0x58
110 ; CHECK-NEXT: no_stack_indirect_call:
111 ; CHECK-NEXT: .backend_stack_size: 0x10
112 ; CHECK-NEXT: .lds_size: 0
113 ; CHECK-NEXT: .sgpr_count: 0x29
114 ; CHECK-NEXT: .stack_frame_size_in_bytes: 0x10
115 ; CHECK-NEXT: .vgpr_count: 0x58
116 ; CHECK-NEXT: simple_lds:
117 ; CHECK-NEXT: .backend_stack_size: 0
118 ; CHECK-NEXT: .lds_size: 0x100
119 ; CHECK-NEXT: .sgpr_count: 0x20
120 ; CHECK-NEXT: .stack_frame_size_in_bytes: 0
121 ; CHECK-NEXT: .vgpr_count: 0x1
122 ; CHECK-NEXT: simple_lds_recurse:
123 ; CHECK-NEXT: .backend_stack_size: 0x10
124 ; CHECK-NEXT: .lds_size: 0x100
125 ; CHECK-NEXT: .sgpr_count: 0x24
126 ; CHECK-NEXT: .stack_frame_size_in_bytes: 0x10
127 ; CHECK-NEXT: .vgpr_count: 0x29
128 ; CHECK-NEXT: simple_stack:
129 ; CHECK-NEXT: .backend_stack_size: 0x14
130 ; CHECK-NEXT: .lds_size: 0
131 ; CHECK-NEXT: .sgpr_count: 0x21
132 ; CHECK-NEXT: .stack_frame_size_in_bytes: 0x14
133 ; CHECK-NEXT: .vgpr_count: 0x2
134 ; CHECK-NEXT: simple_stack_call:
135 ; CHECK-NEXT: .backend_stack_size: 0x20
136 ; CHECK-NEXT: .lds_size: 0
137 ; CHECK-NEXT: .sgpr_count: 0x22
138 ; CHECK-NEXT: .stack_frame_size_in_bytes: 0x20
139 ; CHECK-NEXT: .vgpr_count: 0x4
140 ; CHECK-NEXT: simple_stack_extern_call:
141 ; CHECK-NEXT: .backend_stack_size: 0x20
142 ; CHECK-NEXT: .lds_size: 0
143 ; CHECK-NEXT: .sgpr_count: 0x29
144 ; CHECK-NEXT: .stack_frame_size_in_bytes: 0x20
145 ; CHECK-NEXT: .vgpr_count: 0x58
146 ; CHECK-NEXT: simple_stack_indirect_call:
147 ; CHECK-NEXT: .backend_stack_size: 0x20
148 ; CHECK-NEXT: .lds_size: 0
149 ; CHECK-NEXT: .sgpr_count: 0x29
150 ; CHECK-NEXT: .stack_frame_size_in_bytes: 0x20
151 ; CHECK-NEXT: .vgpr_count: 0x58
152 ; CHECK-NEXT: simple_stack_recurse:
153 ; CHECK-NEXT: .backend_stack_size: 0x20
154 ; CHECK-NEXT: .lds_size: 0
155 ; CHECK-NEXT: .sgpr_count: 0x24
156 ; CHECK-NEXT: .stack_frame_size_in_bytes: 0x20
157 ; CHECK-NEXT: .vgpr_count: 0x2a
158 ; CHECK:amdpal.version:
159 ; CHECK-NEXT: - 0x3
160 ; CHECK-NEXT: - 0
161 ; CHECK-NEXT:...
162 ; CHECK-NEXT: .end_amdgpu_pal_metadata
164 declare amdgpu_gfx float @extern_func(float) #0
165 declare amdgpu_gfx float @extern_func_many_args(<64 x float>) #0
167 @funcptr = external hidden unnamed_addr addrspace(4) constant ptr, align 4
169 define amdgpu_gfx float @no_stack(float %arg0) #0 {
170 %add = fadd float %arg0, 1.0
171 ret float %add
172 }
174 define amdgpu_gfx float @simple_stack(float %arg0) #0 {
175 %stack = alloca float, i32 4, align 4, addrspace(5)
176 store volatile float 2.0, ptr addrspace(5) %stack
177 %val = load volatile float, ptr addrspace(5) %stack
178 %add = fadd float %arg0, %val
179 ret float %add
180 }
182 define amdgpu_gfx float @multiple_stack(float %arg0) #0 {
183 %stack = alloca float, i32 4, align 4, addrspace(5)
184 store volatile float 2.0, ptr addrspace(5) %stack
185 %val = load volatile float, ptr addrspace(5) %stack
186 %add = fadd float %arg0, %val
187 %stack2 = alloca float, i32 4, align 4, addrspace(5)
188 store volatile float 2.0, ptr addrspace(5) %stack2
189 %val2 = load volatile float, ptr addrspace(5) %stack2
190 %add2 = fadd float %add, %val2
191 ret float %add2
192 }
194 define amdgpu_gfx float @dynamic_stack(float %arg0) #0 {
195 bb0:
196 %cmp = fcmp ogt float %arg0, 0.0
197 br i1 %cmp, label %bb1, label %bb2
199 bb1:
200 %stack = alloca float, i32 4, align 4, addrspace(5)
201 store volatile float 2.0, ptr addrspace(5) %stack
202 %val = load volatile float, ptr addrspace(5) %stack
203 %add = fadd float %arg0, %val
204 br label %bb2
206 bb2:
207 %res = phi float [ 0.0, %bb0 ], [ %add, %bb1 ]
208 ret float %res
209 }
211 define amdgpu_gfx float @dynamic_stack_loop(float %arg0) #0 {
212 bb0:
213 br label %bb1
215 bb1:
216 %ctr = phi i32 [ 0, %bb0 ], [ %newctr, %bb1 ]
217 %stack = alloca float, i32 4, align 4, addrspace(5)
218 store volatile float 2.0, ptr addrspace(5) %stack
219 %val = load volatile float, ptr addrspace(5) %stack
220 %add = fadd float %arg0, %val
221 %cmp = icmp sgt i32 %ctr, 0
222 %newctr = sub i32 %ctr, 1
223 br i1 %cmp, label %bb1, label %bb2
225 bb2:
226 ret float %add
227 }
229 define amdgpu_gfx float @no_stack_call(float %arg0) #0 {
230 %res = call amdgpu_gfx float @simple_stack(float %arg0)
231 ret float %res
232 }
234 define amdgpu_gfx float @simple_stack_call(float %arg0) #0 {
235 %stack = alloca float, i32 4, align 4, addrspace(5)
236 store volatile float 2.0, ptr addrspace(5) %stack
237 %val = load volatile float, ptr addrspace(5) %stack
238 %res = call amdgpu_gfx float @simple_stack(float %arg0)
239 %add = fadd float %res, %val
240 ret float %add
241 }
243 define amdgpu_gfx float @no_stack_extern_call(float %arg0) #0 {
244 %res = call amdgpu_gfx float @extern_func(float %arg0)
245 ret float %res
246 }
248 define amdgpu_gfx float @simple_stack_extern_call(float %arg0) #0 {
249 %stack = alloca float, i32 4, align 4, addrspace(5)
250 store volatile float 2.0, ptr addrspace(5) %stack
251 %val = load volatile float, ptr addrspace(5) %stack
252 %res = call amdgpu_gfx float @extern_func(float %arg0)
253 %add = fadd float %res, %val
254 ret float %add
255 }
257 define amdgpu_gfx float @no_stack_extern_call_many_args(<64 x float> %arg0) #0 {
258 %res = call amdgpu_gfx float @extern_func_many_args(<64 x float> %arg0)
259 ret float %res
260 }
262 define amdgpu_gfx float @no_stack_indirect_call(float %arg0) #0 {
263 %fptr = load ptr, ptr addrspace(4) @funcptr
264 call amdgpu_gfx void %fptr()
265 ret float %arg0
266 }
268 define amdgpu_gfx float @simple_stack_indirect_call(float %arg0) #0 {
269 %stack = alloca float, i32 4, align 4, addrspace(5)
270 store volatile float 2.0, ptr addrspace(5) %stack
271 %val = load volatile float, ptr addrspace(5) %stack
272 %fptr = load ptr, ptr addrspace(4) @funcptr
273 call amdgpu_gfx void %fptr()
274 %add = fadd float %arg0, %val
275 ret float %add
276 }
278 define amdgpu_gfx float @simple_stack_recurse(float %arg0) #0 {
279 %stack = alloca float, i32 4, align 4, addrspace(5)
280 store volatile float 2.0, ptr addrspace(5) %stack
281 %val = load volatile float, ptr addrspace(5) %stack
282 %res = call amdgpu_gfx float @simple_stack_recurse(float %arg0)
283 %add = fadd float %res, %val
284 ret float %add
285 }
287 @lds = internal addrspace(3) global [64 x float] undef
289 define amdgpu_gfx float @simple_lds(float %arg0) #0 {
290 %val = load float, ptr addrspace(3) @lds
291 ret float %val
292 }
294 define amdgpu_gfx float @simple_lds_recurse(float %arg0) #0 {
295 %val = load float, ptr addrspace(3) @lds
296 %res = call amdgpu_gfx float @simple_lds_recurse(float %val)
297 ret float %res
298 }
300 attributes #0 = { nounwind }
302 !amdgpu.pal.metadata.msgpack = !{!0}
304 !0 = !{!"\82\B0amdpal.pipelines\91\8A\A4.api\A6Vulkan\B2.compute_registers\85\AB.tg_size_en\C3\AA.tgid_x_en\C2\AA.tgid_y_en\C2\AA.tgid_z_en\C2\AF.tidig_comp_cnt\01\B0.hardware_stages\81\A3.cs\8C\AF.checksum_value\CE\94D\D7\D0\AB.debug_mode\00\AB.float_mode\CC\C0\A9.image_op\C2\AC.mem_ordered\C3\AB.sgpr_limitj\B7.threadgroup_dimensions\93\01\CD\04\00\01\AD.trap_present\00\B2.user_data_reg_map\DC\00 \CE\10\00\00\00\CE\FF\FF\FF\FF\00\CE\FF\FF\FF\FF\CE\FF\FF\FF\FF\CE\FF\FF\FF\FF\CE\FF\FF\FF\FF\CE\FF\FF\FF\FF\CE\FF\FF\FF\FF\CE\FF\FF\FF\FF\CE\FF\FF\FF\FF\CE\FF\FF\FF\FF\CE\FF\FF\FF\FF\CE\FF\FF\FF\FF\CE\FF\FF\FF\FF\CE\FF\FF\FF\FF\CE\FF\FF\FF\FF\CE\FF\FF\FF\FF\CE\FF\FF\FF\FF\CE\FF\FF\FF\FF\CE\FF\FF\FF\FF\CE\FF\FF\FF\FF\CE\FF\FF\FF\FF\CE\FF\FF\FF\FF\CE\FF\FF\FF\FF\CE\FF\FF\FF\FF\CE\FF\FF\FF\FF\CE\FF\FF\FF\FF\CE\FF\FF\FF\FF\CE\FF\FF\FF\FF\CE\FF\FF\FF\FF\CE\FF\FF\FF\FF\AB.user_sgprs\03\AB.vgpr_limit\CD\01\00\AF.wavefront_size@\B7.internal_pipeline_hash\92\CF\E7\10k\A6:\A6%\F7\CF\B2\1F\1A\D4{\DA\E1T\AA.registers\80\A8.shaders\81\A8.compute\82\B0.api_shader_hash\92\CF\E9Zn7}\1E\B9\E7\00\B1.hardware_mapping\91\A3.cs\B0.spill_threshold\CE\FF\FF\FF\FF\A5.type\A2Cs\B0.user_data_limit\01\AF.xgl_cache_info\82\B3.128_bit_cache_hash\92\CF\B4X\B8\11[\A4\88P\CF\A0;\B0\AF\FF\B4\BE\C0\AD.llpc_version\A461.1\AEamdpal.version\92\03\00"}
305 !1 = !{i32 7}