Skip to content

Commit 0fe914e

Browse files
chinmayddchinmaydd
committed
[AMDGPU] Update fsub.f32 test
1 parent ae8adc5 commit 0fe914e

File tree

1 file changed

+219
-0
lines changed

1 file changed

+219
-0
lines changed

‎llvm/test/CodeGen/AMDGPU/strict_fsub.f32.ll‎

Lines changed: 219 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -8,6 +8,9 @@
88
; RUN: llc -global-isel= -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -amdgpu-enable-delay-alu=0 < %s | FileCheck -check-prefixes=GFX10PLUS,GFX11 %s
99
; RUN: llc -global-isel=1 -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -amdgpu-enable-delay-alu=0 < %s | FileCheck -check-prefixes=GFX10PLUS,GFX11 %s
1010

11+
; RUN: llc -global-isel=0 -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1200 < %s | FileCheck -check-prefixes=GFX12-SDAG %s
12+
; RUN: llc -global-isel=1 -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1200 < %s | FileCheck -check-prefixes=GFX12-GISEL %s
13+
1114
define float @v_constained_fsub_f32_fpexcept_strict(float %x, float %y) #0 {
1215
; GCN-LABEL: v_constained_fsub_f32_fpexcept_strict:
1316
; GCN: ; %bb.0:
@@ -20,6 +23,26 @@ define float @v_constained_fsub_f32_fpexcept_strict(float %x, float %y) #0 {
2023
; GFX10PLUS-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
2124
; GFX10PLUS-NEXT: v_sub_f32_e32 v0, v0, v1
2225
; GFX10PLUS-NEXT: s_setpc_b64 s[30:31]
26+
;
27+
; GFX12-SDAG-LABEL: v_constained_fsub_f32_fpexcept_strict:
28+
; GFX12-SDAG: ; %bb.0:
29+
; GFX12-SDAG-NEXT: s_wait_loadcnt_dscnt 0x0
30+
; GFX12-SDAG-NEXT: s_wait_expcnt 0x0
31+
; GFX12-SDAG-NEXT: s_wait_samplecnt 0x0
32+
; GFX12-SDAG-NEXT: s_wait_bvhcnt 0x0
33+
; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0
34+
; GFX12-SDAG-NEXT: v_sub_f32_e32 v0, v0, v1
35+
; GFX12-SDAG-NEXT: s_setpc_b64 s[30:31]
36+
;
37+
; GFX12-GISEL-LABEL: v_constained_fsub_f32_fpexcept_strict:
38+
; GFX12-GISEL: ; %bb.0:
39+
; GFX12-GISEL-NEXT: s_wait_loadcnt_dscnt 0x0
40+
; GFX12-GISEL-NEXT: s_wait_expcnt 0x0
41+
; GFX12-GISEL-NEXT: s_wait_samplecnt 0x0
42+
; GFX12-GISEL-NEXT: s_wait_bvhcnt 0x0
43+
; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0
44+
; GFX12-GISEL-NEXT: v_sub_f32_e32 v0, v0, v1
45+
; GFX12-GISEL-NEXT: s_setpc_b64 s[30:31]
2346
%val = call float @llvm.experimental.constrained.fsub.f32(float %x, float %y, metadata !"round.tonearest", metadata !"fpexcept.strict")
2447
ret float %val
2548
}
@@ -36,6 +59,26 @@ define float @v_constained_fsub_f32_fpexcept_ignore(float %x, float %y) #0 {
3659
; GFX10PLUS-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
3760
; GFX10PLUS-NEXT: v_sub_f32_e32 v0, v0, v1
3861
; GFX10PLUS-NEXT: s_setpc_b64 s[30:31]
62+
;
63+
; GFX12-SDAG-LABEL: v_constained_fsub_f32_fpexcept_ignore:
64+
; GFX12-SDAG: ; %bb.0:
65+
; GFX12-SDAG-NEXT: s_wait_loadcnt_dscnt 0x0
66+
; GFX12-SDAG-NEXT: s_wait_expcnt 0x0
67+
; GFX12-SDAG-NEXT: s_wait_samplecnt 0x0
68+
; GFX12-SDAG-NEXT: s_wait_bvhcnt 0x0
69+
; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0
70+
; GFX12-SDAG-NEXT: v_sub_f32_e32 v0, v0, v1
71+
; GFX12-SDAG-NEXT: s_setpc_b64 s[30:31]
72+
;
73+
; GFX12-GISEL-LABEL: v_constained_fsub_f32_fpexcept_ignore:
74+
; GFX12-GISEL: ; %bb.0:
75+
; GFX12-GISEL-NEXT: s_wait_loadcnt_dscnt 0x0
76+
; GFX12-GISEL-NEXT: s_wait_expcnt 0x0
77+
; GFX12-GISEL-NEXT: s_wait_samplecnt 0x0
78+
; GFX12-GISEL-NEXT: s_wait_bvhcnt 0x0
79+
; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0
80+
; GFX12-GISEL-NEXT: v_sub_f32_e32 v0, v0, v1
81+
; GFX12-GISEL-NEXT: s_setpc_b64 s[30:31]
3982
%val = call float @llvm.experimental.constrained.fsub.f32(float %x, float %y, metadata !"round.tonearest", metadata !"fpexcept.ignore")
4083
ret float %val
4184
}
@@ -52,6 +95,26 @@ define float @v_constained_fsub_f32_fpexcept_maytrap(float %x, float %y) #0 {
5295
; GFX10PLUS-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
5396
; GFX10PLUS-NEXT: v_sub_f32_e32 v0, v0, v1
5497
; GFX10PLUS-NEXT: s_setpc_b64 s[30:31]
98+
;
99+
; GFX12-SDAG-LABEL: v_constained_fsub_f32_fpexcept_maytrap:
100+
; GFX12-SDAG: ; %bb.0:
101+
; GFX12-SDAG-NEXT: s_wait_loadcnt_dscnt 0x0
102+
; GFX12-SDAG-NEXT: s_wait_expcnt 0x0
103+
; GFX12-SDAG-NEXT: s_wait_samplecnt 0x0
104+
; GFX12-SDAG-NEXT: s_wait_bvhcnt 0x0
105+
; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0
106+
; GFX12-SDAG-NEXT: v_sub_f32_e32 v0, v0, v1
107+
; GFX12-SDAG-NEXT: s_setpc_b64 s[30:31]
108+
;
109+
; GFX12-GISEL-LABEL: v_constained_fsub_f32_fpexcept_maytrap:
110+
; GFX12-GISEL: ; %bb.0:
111+
; GFX12-GISEL-NEXT: s_wait_loadcnt_dscnt 0x0
112+
; GFX12-GISEL-NEXT: s_wait_expcnt 0x0
113+
; GFX12-GISEL-NEXT: s_wait_samplecnt 0x0
114+
; GFX12-GISEL-NEXT: s_wait_bvhcnt 0x0
115+
; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0
116+
; GFX12-GISEL-NEXT: v_sub_f32_e32 v0, v0, v1
117+
; GFX12-GISEL-NEXT: s_setpc_b64 s[30:31]
55118
%val = call float @llvm.experimental.constrained.fsub.f32(float %x, float %y, metadata !"round.tonearest", metadata !"fpexcept.maytrap")
56119
ret float %val
57120
}
@@ -76,6 +139,26 @@ define <2 x float> @v_constained_fsub_v2f32_fpexcept_strict(<2 x float> %x, <2 x
76139
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
77140
; GFX11-NEXT: v_dual_sub_f32 v0, v0, v2 :: v_dual_sub_f32 v1, v1, v3
78141
; GFX11-NEXT: s_setpc_b64 s[30:31]
142+
;
143+
; GFX12-SDAG-LABEL: v_constained_fsub_v2f32_fpexcept_strict:
144+
; GFX12-SDAG: ; %bb.0:
145+
; GFX12-SDAG-NEXT: s_wait_loadcnt_dscnt 0x0
146+
; GFX12-SDAG-NEXT: s_wait_expcnt 0x0
147+
; GFX12-SDAG-NEXT: s_wait_samplecnt 0x0
148+
; GFX12-SDAG-NEXT: s_wait_bvhcnt 0x0
149+
; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0
150+
; GFX12-SDAG-NEXT: v_dual_sub_f32 v0, v0, v2 :: v_dual_sub_f32 v1, v1, v3
151+
; GFX12-SDAG-NEXT: s_setpc_b64 s[30:31]
152+
;
153+
; GFX12-GISEL-LABEL: v_constained_fsub_v2f32_fpexcept_strict:
154+
; GFX12-GISEL: ; %bb.0:
155+
; GFX12-GISEL-NEXT: s_wait_loadcnt_dscnt 0x0
156+
; GFX12-GISEL-NEXT: s_wait_expcnt 0x0
157+
; GFX12-GISEL-NEXT: s_wait_samplecnt 0x0
158+
; GFX12-GISEL-NEXT: s_wait_bvhcnt 0x0
159+
; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0
160+
; GFX12-GISEL-NEXT: v_dual_sub_f32 v0, v0, v2 :: v_dual_sub_f32 v1, v1, v3
161+
; GFX12-GISEL-NEXT: s_setpc_b64 s[30:31]
79162
%val = call <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float> %x, <2 x float> %y, metadata !"round.tonearest", metadata !"fpexcept.strict")
80163
ret <2 x float> %val
81164
}
@@ -100,6 +183,26 @@ define <2 x float> @v_constained_fsub_v2f32_fpexcept_ignore(<2 x float> %x, <2 x
100183
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
101184
; GFX11-NEXT: v_dual_sub_f32 v0, v0, v2 :: v_dual_sub_f32 v1, v1, v3
102185
; GFX11-NEXT: s_setpc_b64 s[30:31]
186+
;
187+
; GFX12-SDAG-LABEL: v_constained_fsub_v2f32_fpexcept_ignore:
188+
; GFX12-SDAG: ; %bb.0:
189+
; GFX12-SDAG-NEXT: s_wait_loadcnt_dscnt 0x0
190+
; GFX12-SDAG-NEXT: s_wait_expcnt 0x0
191+
; GFX12-SDAG-NEXT: s_wait_samplecnt 0x0
192+
; GFX12-SDAG-NEXT: s_wait_bvhcnt 0x0
193+
; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0
194+
; GFX12-SDAG-NEXT: v_dual_sub_f32 v0, v0, v2 :: v_dual_sub_f32 v1, v1, v3
195+
; GFX12-SDAG-NEXT: s_setpc_b64 s[30:31]
196+
;
197+
; GFX12-GISEL-LABEL: v_constained_fsub_v2f32_fpexcept_ignore:
198+
; GFX12-GISEL: ; %bb.0:
199+
; GFX12-GISEL-NEXT: s_wait_loadcnt_dscnt 0x0
200+
; GFX12-GISEL-NEXT: s_wait_expcnt 0x0
201+
; GFX12-GISEL-NEXT: s_wait_samplecnt 0x0
202+
; GFX12-GISEL-NEXT: s_wait_bvhcnt 0x0
203+
; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0
204+
; GFX12-GISEL-NEXT: v_dual_sub_f32 v0, v0, v2 :: v_dual_sub_f32 v1, v1, v3
205+
; GFX12-GISEL-NEXT: s_setpc_b64 s[30:31]
103206
%val = call <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float> %x, <2 x float> %y, metadata !"round.tonearest", metadata !"fpexcept.ignore")
104207
ret <2 x float> %val
105208
}
@@ -124,6 +227,26 @@ define <2 x float> @v_constained_fsub_v2f32_fpexcept_maytrap(<2 x float> %x, <2
124227
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
125228
; GFX11-NEXT: v_dual_sub_f32 v0, v0, v2 :: v_dual_sub_f32 v1, v1, v3
126229
; GFX11-NEXT: s_setpc_b64 s[30:31]
230+
;
231+
; GFX12-SDAG-LABEL: v_constained_fsub_v2f32_fpexcept_maytrap:
232+
; GFX12-SDAG: ; %bb.0:
233+
; GFX12-SDAG-NEXT: s_wait_loadcnt_dscnt 0x0
234+
; GFX12-SDAG-NEXT: s_wait_expcnt 0x0
235+
; GFX12-SDAG-NEXT: s_wait_samplecnt 0x0
236+
; GFX12-SDAG-NEXT: s_wait_bvhcnt 0x0
237+
; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0
238+
; GFX12-SDAG-NEXT: v_dual_sub_f32 v0, v0, v2 :: v_dual_sub_f32 v1, v1, v3
239+
; GFX12-SDAG-NEXT: s_setpc_b64 s[30:31]
240+
;
241+
; GFX12-GISEL-LABEL: v_constained_fsub_v2f32_fpexcept_maytrap:
242+
; GFX12-GISEL: ; %bb.0:
243+
; GFX12-GISEL-NEXT: s_wait_loadcnt_dscnt 0x0
244+
; GFX12-GISEL-NEXT: s_wait_expcnt 0x0
245+
; GFX12-GISEL-NEXT: s_wait_samplecnt 0x0
246+
; GFX12-GISEL-NEXT: s_wait_bvhcnt 0x0
247+
; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0
248+
; GFX12-GISEL-NEXT: v_dual_sub_f32 v0, v0, v2 :: v_dual_sub_f32 v1, v1, v3
249+
; GFX12-GISEL-NEXT: s_setpc_b64 s[30:31]
127250
%val = call <2 x float> @llvm.experimental.constrained.fsub.v2f32(<2 x float> %x, <2 x float> %y, metadata !"round.tonearest", metadata !"fpexcept.maytrap")
128251
ret <2 x float> %val
129252
}
@@ -151,6 +274,28 @@ define <3 x float> @v_constained_fsub_v3f32_fpexcept_strict(<3 x float> %x, <3 x
151274
; GFX11-NEXT: v_dual_sub_f32 v0, v0, v3 :: v_dual_sub_f32 v1, v1, v4
152275
; GFX11-NEXT: v_sub_f32_e32 v2, v2, v5
153276
; GFX11-NEXT: s_setpc_b64 s[30:31]
277+
;
278+
; GFX12-SDAG-LABEL: v_constained_fsub_v3f32_fpexcept_strict:
279+
; GFX12-SDAG: ; %bb.0:
280+
; GFX12-SDAG-NEXT: s_wait_loadcnt_dscnt 0x0
281+
; GFX12-SDAG-NEXT: s_wait_expcnt 0x0
282+
; GFX12-SDAG-NEXT: s_wait_samplecnt 0x0
283+
; GFX12-SDAG-NEXT: s_wait_bvhcnt 0x0
284+
; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0
285+
; GFX12-SDAG-NEXT: v_dual_sub_f32 v0, v0, v3 :: v_dual_sub_f32 v1, v1, v4
286+
; GFX12-SDAG-NEXT: v_sub_f32_e32 v2, v2, v5
287+
; GFX12-SDAG-NEXT: s_setpc_b64 s[30:31]
288+
;
289+
; GFX12-GISEL-LABEL: v_constained_fsub_v3f32_fpexcept_strict:
290+
; GFX12-GISEL: ; %bb.0:
291+
; GFX12-GISEL-NEXT: s_wait_loadcnt_dscnt 0x0
292+
; GFX12-GISEL-NEXT: s_wait_expcnt 0x0
293+
; GFX12-GISEL-NEXT: s_wait_samplecnt 0x0
294+
; GFX12-GISEL-NEXT: s_wait_bvhcnt 0x0
295+
; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0
296+
; GFX12-GISEL-NEXT: v_dual_sub_f32 v0, v0, v3 :: v_dual_sub_f32 v1, v1, v4
297+
; GFX12-GISEL-NEXT: v_sub_f32_e32 v2, v2, v5
298+
; GFX12-GISEL-NEXT: s_setpc_b64 s[30:31]
154299
%val = call <3 x float> @llvm.experimental.constrained.fsub.v3f32(<3 x float> %x, <3 x float> %y, metadata !"round.tonearest", metadata !"fpexcept.strict")
155300
ret <3 x float> %val
156301
}
@@ -166,6 +311,20 @@ define amdgpu_ps float @s_constained_fsub_f32_fpexcept_strict(float inreg %x, fl
166311
; GFX10PLUS: ; %bb.0:
167312
; GFX10PLUS-NEXT: v_sub_f32_e64 v0, s2, s3
168313
; GFX10PLUS-NEXT: ; return to shader part epilog
314+
;
315+
; GFX12-SDAG-LABEL: s_constained_fsub_f32_fpexcept_strict:
316+
; GFX12-SDAG: ; %bb.0:
317+
; GFX12-SDAG-NEXT: s_sub_f32 s0, s2, s3
318+
; GFX12-SDAG-NEXT: s_delay_alu instid0(SALU_CYCLE_3)
319+
; GFX12-SDAG-NEXT: v_mov_b32_e32 v0, s0
320+
; GFX12-SDAG-NEXT: ; return to shader part epilog
321+
;
322+
; GFX12-GISEL-LABEL: s_constained_fsub_f32_fpexcept_strict:
323+
; GFX12-GISEL: ; %bb.0:
324+
; GFX12-GISEL-NEXT: s_sub_f32 s0, s2, s3
325+
; GFX12-GISEL-NEXT: s_delay_alu instid0(SALU_CYCLE_3)
326+
; GFX12-GISEL-NEXT: v_mov_b32_e32 v0, s0
327+
; GFX12-GISEL-NEXT: ; return to shader part epilog
169328
%val = call float @llvm.experimental.constrained.fsub.f32(float %x, float %y, metadata !"round.tonearest", metadata !"fpexcept.strict")
170329
ret float %val
171330
}
@@ -182,6 +341,26 @@ define float @v_constained_fsub_f32_fpexcept_strict_fabs_lhs(float %x, float %y)
182341
; GFX10PLUS-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
183342
; GFX10PLUS-NEXT: v_sub_f32_e64 v0, |v0|, v1
184343
; GFX10PLUS-NEXT: s_setpc_b64 s[30:31]
344+
;
345+
; GFX12-SDAG-LABEL: v_constained_fsub_f32_fpexcept_strict_fabs_lhs:
346+
; GFX12-SDAG: ; %bb.0:
347+
; GFX12-SDAG-NEXT: s_wait_loadcnt_dscnt 0x0
348+
; GFX12-SDAG-NEXT: s_wait_expcnt 0x0
349+
; GFX12-SDAG-NEXT: s_wait_samplecnt 0x0
350+
; GFX12-SDAG-NEXT: s_wait_bvhcnt 0x0
351+
; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0
352+
; GFX12-SDAG-NEXT: v_sub_f32_e64 v0, |v0|, v1
353+
; GFX12-SDAG-NEXT: s_setpc_b64 s[30:31]
354+
;
355+
; GFX12-GISEL-LABEL: v_constained_fsub_f32_fpexcept_strict_fabs_lhs:
356+
; GFX12-GISEL: ; %bb.0:
357+
; GFX12-GISEL-NEXT: s_wait_loadcnt_dscnt 0x0
358+
; GFX12-GISEL-NEXT: s_wait_expcnt 0x0
359+
; GFX12-GISEL-NEXT: s_wait_samplecnt 0x0
360+
; GFX12-GISEL-NEXT: s_wait_bvhcnt 0x0
361+
; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0
362+
; GFX12-GISEL-NEXT: v_sub_f32_e64 v0, |v0|, v1
363+
; GFX12-GISEL-NEXT: s_setpc_b64 s[30:31]
185364
%fabs.x = call float @llvm.fabs.f32(float %x) #0
186365
%val = call float @llvm.experimental.constrained.fsub.f32(float %fabs.x, float %y, metadata !"round.tonearest", metadata !"fpexcept.strict")
187366
ret float %val
@@ -199,6 +378,26 @@ define float @v_constained_fsub_f32_fpexcept_strict_fabs_rhs(float %x, float %y)
199378
; GFX10PLUS-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
200379
; GFX10PLUS-NEXT: v_sub_f32_e64 v0, v0, |v1|
201380
; GFX10PLUS-NEXT: s_setpc_b64 s[30:31]
381+
;
382+
; GFX12-SDAG-LABEL: v_constained_fsub_f32_fpexcept_strict_fabs_rhs:
383+
; GFX12-SDAG: ; %bb.0:
384+
; GFX12-SDAG-NEXT: s_wait_loadcnt_dscnt 0x0
385+
; GFX12-SDAG-NEXT: s_wait_expcnt 0x0
386+
; GFX12-SDAG-NEXT: s_wait_samplecnt 0x0
387+
; GFX12-SDAG-NEXT: s_wait_bvhcnt 0x0
388+
; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0
389+
; GFX12-SDAG-NEXT: v_sub_f32_e64 v0, v0, |v1|
390+
; GFX12-SDAG-NEXT: s_setpc_b64 s[30:31]
391+
;
392+
; GFX12-GISEL-LABEL: v_constained_fsub_f32_fpexcept_strict_fabs_rhs:
393+
; GFX12-GISEL: ; %bb.0:
394+
; GFX12-GISEL-NEXT: s_wait_loadcnt_dscnt 0x0
395+
; GFX12-GISEL-NEXT: s_wait_expcnt 0x0
396+
; GFX12-GISEL-NEXT: s_wait_samplecnt 0x0
397+
; GFX12-GISEL-NEXT: s_wait_bvhcnt 0x0
398+
; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0
399+
; GFX12-GISEL-NEXT: v_sub_f32_e64 v0, v0, |v1|
400+
; GFX12-GISEL-NEXT: s_setpc_b64 s[30:31]
202401
%fabs.y = call float @llvm.fabs.f32(float %y) #0
203402
%val = call float @llvm.experimental.constrained.fsub.f32(float %x, float %fabs.y, metadata !"round.tonearest", metadata !"fpexcept.strict")
204403
ret float %val
@@ -216,6 +415,26 @@ define float @v_constained_fsub_f32_fpexcept_strict_fneg_fabs_lhs(float %x, floa
216415
; GFX10PLUS-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
217416
; GFX10PLUS-NEXT: v_sub_f32_e64 v0, -|v0|, v1
218417
; GFX10PLUS-NEXT: s_setpc_b64 s[30:31]
418+
;
419+
; GFX12-SDAG-LABEL: v_constained_fsub_f32_fpexcept_strict_fneg_fabs_lhs:
420+
; GFX12-SDAG: ; %bb.0:
421+
; GFX12-SDAG-NEXT: s_wait_loadcnt_dscnt 0x0
422+
; GFX12-SDAG-NEXT: s_wait_expcnt 0x0
423+
; GFX12-SDAG-NEXT: s_wait_samplecnt 0x0
424+
; GFX12-SDAG-NEXT: s_wait_bvhcnt 0x0
425+
; GFX12-SDAG-NEXT: s_wait_kmcnt 0x0
426+
; GFX12-SDAG-NEXT: v_sub_f32_e64 v0, -|v0|, v1
427+
; GFX12-SDAG-NEXT: s_setpc_b64 s[30:31]
428+
;
429+
; GFX12-GISEL-LABEL: v_constained_fsub_f32_fpexcept_strict_fneg_fabs_lhs:
430+
; GFX12-GISEL: ; %bb.0:
431+
; GFX12-GISEL-NEXT: s_wait_loadcnt_dscnt 0x0
432+
; GFX12-GISEL-NEXT: s_wait_expcnt 0x0
433+
; GFX12-GISEL-NEXT: s_wait_samplecnt 0x0
434+
; GFX12-GISEL-NEXT: s_wait_bvhcnt 0x0
435+
; GFX12-GISEL-NEXT: s_wait_kmcnt 0x0
436+
; GFX12-GISEL-NEXT: v_sub_f32_e64 v0, -|v0|, v1
437+
; GFX12-GISEL-NEXT: s_setpc_b64 s[30:31]
219438
%fabs.x = call float @llvm.fabs.f32(float %x) #0
220439
%neg.fabs.x = fneg float %fabs.x
221440
%val = call float @llvm.experimental.constrained.fsub.f32(float %neg.fabs.x, float %y, metadata !"round.tonearest", metadata !"fpexcept.strict")

0 commit comments

Comments
 (0)