Skip to content

Commit 2d6d723

Browse files
jayfoadjayfoad
authored
[AMDGPU] Add some more GFX12 test coverage (llvm#120581)
1 parent f70ab7d commit 2d6d723

18 files changed

+3565
-55
lines changed

‎llvm/test/CodeGen/AMDGPU/GlobalISel/atomicrmw_uinc_wrap.ll

Lines changed: 567 additions & 0 deletions
Large diffs are not rendered by default.

‎llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-scalar-float-sop1.mir

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,6 @@
11
# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
22
# RUN: llc -mtriple=amdgcn -mcpu=gfx1150 -run-pass=instruction-select -verify-machineinstrs -o - %s | FileCheck -check-prefix=GFX1150 %s
3+
# RUN: llc -mtriple=amdgcn -mcpu=gfx1200 -run-pass=instruction-select -verify-machineinstrs -o - %s | FileCheck -check-prefix=GFX1150 %s
34

45
---
56
name: sitofp_i32_to_f32

‎llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-scalar-float-sop2.mir

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,6 @@
11
# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
22
# RUN: llc -mtriple=amdgcn -mcpu=gfx1150 -run-pass=instruction-select -verify-machineinstrs -o - %s | FileCheck -check-prefix=GFX1150 %s
3+
# RUN: llc -mtriple=amdgcn -mcpu=gfx1200 -run-pass=instruction-select -verify-machineinstrs -o - %s | FileCheck -check-prefix=GFX1150 %s
34

45
---
56
name: fadd_f32

‎llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-scalar-float-sopc.mir

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,6 @@
11
# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
22
# RUN: llc -mtriple=amdgcn -mcpu=gfx1150 -run-pass=instruction-select -verify-machineinstrs -o - %s | FileCheck -check-prefix=GFX1150 %s
3+
# RUN: llc -mtriple=amdgcn -mcpu=gfx1200 -run-pass=instruction-select -verify-machineinstrs -o - %s | FileCheck -check-prefix=GFX1150 %s
34

45
---
56
name: f32_olt

‎llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.interp.inreg.ll

Lines changed: 168 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -1,6 +1,8 @@
11
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
22
; RUN: llc -global-isel -mtriple=amdgcn -mcpu=gfx1100 -mattr=+real-true16 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX11,GFX11-TRUE16 %s
33
; RUN: llc -global-isel -mtriple=amdgcn -mcpu=gfx1100 -mattr=-real-true16 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX11,GFX11-FAKE16 %s
4+
; RUN: llc -global-isel -mtriple=amdgcn -mcpu=gfx1200 -mattr=+real-true16 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX12,GFX12-TRUE16 %s
5+
; RUN: llc -global-isel -mtriple=amdgcn -mcpu=gfx1200 -mattr=-real-true16 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX12,GFX12-FAKE16 %s
46

57
define amdgpu_ps void @v_interp_f32(float inreg %i, float inreg %j, i32 inreg %m0) #0 {
68
; GFX11-LABEL: v_interp_f32:
@@ -21,6 +23,25 @@ define amdgpu_ps void @v_interp_f32(float inreg %i, float inreg %j, i32 inreg %m
2123
; GFX11-NEXT: v_interp_p2_f32 v4, v1, v4, v5 wait_exp:7
2224
; GFX11-NEXT: exp mrt0 v3, v2, v5, v4 done
2325
; GFX11-NEXT: s_endpgm
26+
;
27+
; GFX12-LABEL: v_interp_f32:
28+
; GFX12: ; %bb.0: ; %main_body
29+
; GFX12-NEXT: s_mov_b32 s3, exec_lo
30+
; GFX12-NEXT: s_wqm_b32 exec_lo, exec_lo
31+
; GFX12-NEXT: s_mov_b32 m0, s2
32+
; GFX12-NEXT: ds_param_load v0, attr0.y wait_va_vdst:15 wait_vm_vsrc:1
33+
; GFX12-NEXT: ds_param_load v1, attr1.x wait_va_vdst:15 wait_vm_vsrc:1
34+
; GFX12-NEXT: s_mov_b32 exec_lo, s3
35+
; GFX12-NEXT: v_mov_b32_e32 v2, s0
36+
; GFX12-NEXT: v_mov_b32_e32 v4, s1
37+
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
38+
; GFX12-NEXT: v_interp_p10_f32 v3, v0, v2, v0 wait_exp:1
39+
; GFX12-NEXT: v_interp_p10_f32 v2, v1, v2, v1 wait_exp:0
40+
; GFX12-NEXT: v_interp_p2_f32 v5, v0, v4, v3 wait_exp:7
41+
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1)
42+
; GFX12-NEXT: v_interp_p2_f32 v4, v1, v4, v5 wait_exp:7
43+
; GFX12-NEXT: export mrt0 v3, v2, v5, v4 done
44+
; GFX12-NEXT: s_endpgm
2445
main_body:
2546
%p0 = call float @llvm.amdgcn.lds.param.load(i32 1, i32 0, i32 %m0)
2647
%p1 = call float @llvm.amdgcn.lds.param.load(i32 0, i32 1, i32 %m0)
@@ -57,6 +78,31 @@ define amdgpu_ps void @v_interp_f32_many(float inreg %i, float inreg %j, i32 inr
5778
; GFX11-NEXT: v_interp_p2_f32 v4, v3, v5, v4 wait_exp:7
5879
; GFX11-NEXT: exp mrt0 v6, v7, v8, v4 done
5980
; GFX11-NEXT: s_endpgm
81+
;
82+
; GFX12-LABEL: v_interp_f32_many:
83+
; GFX12: ; %bb.0: ; %main_body
84+
; GFX12-NEXT: s_mov_b32 s3, exec_lo
85+
; GFX12-NEXT: s_wqm_b32 exec_lo, exec_lo
86+
; GFX12-NEXT: s_mov_b32 m0, s2
87+
; GFX12-NEXT: ds_param_load v0, attr0.x wait_va_vdst:15 wait_vm_vsrc:1
88+
; GFX12-NEXT: ds_param_load v1, attr1.x wait_va_vdst:15 wait_vm_vsrc:1
89+
; GFX12-NEXT: ds_param_load v2, attr2.x wait_va_vdst:15 wait_vm_vsrc:1
90+
; GFX12-NEXT: ds_param_load v3, attr3.x wait_va_vdst:15 wait_vm_vsrc:1
91+
; GFX12-NEXT: s_mov_b32 exec_lo, s3
92+
; GFX12-NEXT: v_dual_mov_b32 v4, s0 :: v_dual_mov_b32 v5, s1
93+
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_4)
94+
; GFX12-NEXT: v_interp_p10_f32 v6, v0, v4, v0 wait_exp:3
95+
; GFX12-NEXT: v_interp_p10_f32 v7, v1, v4, v1 wait_exp:2
96+
; GFX12-NEXT: v_interp_p10_f32 v8, v2, v4, v2 wait_exp:1
97+
; GFX12-NEXT: v_interp_p10_f32 v4, v3, v4, v3 wait_exp:0
98+
; GFX12-NEXT: v_interp_p2_f32 v6, v0, v5, v6 wait_exp:7
99+
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
100+
; GFX12-NEXT: v_interp_p2_f32 v7, v1, v5, v7 wait_exp:7
101+
; GFX12-NEXT: v_interp_p2_f32 v8, v2, v5, v8 wait_exp:7
102+
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_4)
103+
; GFX12-NEXT: v_interp_p2_f32 v4, v3, v5, v4 wait_exp:7
104+
; GFX12-NEXT: export mrt0 v6, v7, v8, v4 done
105+
; GFX12-NEXT: s_endpgm
60106
main_body:
61107
%p0 = call float @llvm.amdgcn.lds.param.load(i32 0, i32 0, i32 %m0)
62108
%p1 = call float @llvm.amdgcn.lds.param.load(i32 0, i32 1, i32 %m0)
@@ -99,6 +145,31 @@ define amdgpu_ps void @v_interp_f32_many_vm(ptr addrspace(1) %ptr, i32 inreg %m0
99145
; GFX11-NEXT: v_interp_p2_f32 v0, v5, v1, v0 wait_exp:7
100146
; GFX11-NEXT: exp mrt0 v6, v7, v8, v0 done
101147
; GFX11-NEXT: s_endpgm
148+
;
149+
; GFX12-LABEL: v_interp_f32_many_vm:
150+
; GFX12: ; %bb.0: ; %main_body
151+
; GFX12-NEXT: global_load_b64 v[0:1], v[0:1], off offset:4
152+
; GFX12-NEXT: s_mov_b32 m0, s0
153+
; GFX12-NEXT: s_mov_b32 s0, exec_lo
154+
; GFX12-NEXT: s_wqm_b32 exec_lo, exec_lo
155+
; GFX12-NEXT: ds_param_load v2, attr0.x wait_va_vdst:15 wait_vm_vsrc:1
156+
; GFX12-NEXT: ds_param_load v3, attr1.x wait_va_vdst:15 wait_vm_vsrc:1
157+
; GFX12-NEXT: ds_param_load v4, attr2.x wait_va_vdst:15 wait_vm_vsrc:1
158+
; GFX12-NEXT: ds_param_load v5, attr3.x wait_va_vdst:15 wait_vm_vsrc:1
159+
; GFX12-NEXT: s_mov_b32 exec_lo, s0
160+
; GFX12-NEXT: s_wait_loadcnt 0x0
161+
; GFX12-NEXT: v_interp_p10_f32 v6, v2, v0, v2 wait_exp:3
162+
; GFX12-NEXT: v_interp_p10_f32 v7, v3, v0, v3 wait_exp:2
163+
; GFX12-NEXT: v_interp_p10_f32 v8, v4, v0, v4 wait_exp:1
164+
; GFX12-NEXT: v_interp_p10_f32 v0, v5, v0, v5 wait_exp:0
165+
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
166+
; GFX12-NEXT: v_interp_p2_f32 v6, v2, v1, v6 wait_exp:7
167+
; GFX12-NEXT: v_interp_p2_f32 v7, v3, v1, v7 wait_exp:7
168+
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
169+
; GFX12-NEXT: v_interp_p2_f32 v8, v4, v1, v8 wait_exp:7
170+
; GFX12-NEXT: v_interp_p2_f32 v0, v5, v1, v0 wait_exp:7
171+
; GFX12-NEXT: export mrt0 v6, v7, v8, v0 done
172+
; GFX12-NEXT: s_endpgm
102173
main_body:
103174
%i.ptr = getelementptr float, ptr addrspace(1) %ptr, i32 1
104175
%i = load float, ptr addrspace(1) %i.ptr, align 4
@@ -156,6 +227,42 @@ define amdgpu_ps half @v_interp_f16(float inreg %i, float inreg %j, i32 inreg %m
156227
; GFX11-FAKE16-NEXT: v_interp_p2_f16_f32 v0, v1, v2, v0 op_sel:[1,0,0,0] wait_exp:7
157228
; GFX11-FAKE16-NEXT: v_add_f16_e32 v0, v3, v0
158229
; GFX11-FAKE16-NEXT: ; return to shader part epilog
230+
;
231+
; GFX12-TRUE16-LABEL: v_interp_f16:
232+
; GFX12-TRUE16: ; %bb.0: ; %main_body
233+
; GFX12-TRUE16-NEXT: s_mov_b32 s3, exec_lo
234+
; GFX12-TRUE16-NEXT: s_wqm_b32 exec_lo, exec_lo
235+
; GFX12-TRUE16-NEXT: s_mov_b32 m0, s2
236+
; GFX12-TRUE16-NEXT: ds_param_load v1, attr0.x wait_va_vdst:15 wait_vm_vsrc:1
237+
; GFX12-TRUE16-NEXT: s_mov_b32 exec_lo, s3
238+
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v0, s0
239+
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v2, s1
240+
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
241+
; GFX12-TRUE16-NEXT: v_interp_p10_f16_f32 v3, v1.l, v0, v1.l wait_exp:0
242+
; GFX12-TRUE16-NEXT: v_interp_p10_f16_f32 v4, v1.h, v0, v1.h wait_exp:7
243+
; GFX12-TRUE16-NEXT: v_interp_p2_f16_f32 v0.l, v1.l, v2, v3 wait_exp:7
244+
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
245+
; GFX12-TRUE16-NEXT: v_interp_p2_f16_f32 v0.h, v1.h, v2, v4 wait_exp:7
246+
; GFX12-TRUE16-NEXT: v_add_f16_e32 v0.l, v0.l, v0.h
247+
; GFX12-TRUE16-NEXT: ; return to shader part epilog
248+
;
249+
; GFX12-FAKE16-LABEL: v_interp_f16:
250+
; GFX12-FAKE16: ; %bb.0: ; %main_body
251+
; GFX12-FAKE16-NEXT: s_mov_b32 s3, exec_lo
252+
; GFX12-FAKE16-NEXT: s_wqm_b32 exec_lo, exec_lo
253+
; GFX12-FAKE16-NEXT: s_mov_b32 m0, s2
254+
; GFX12-FAKE16-NEXT: ds_param_load v1, attr0.x wait_va_vdst:15 wait_vm_vsrc:1
255+
; GFX12-FAKE16-NEXT: s_mov_b32 exec_lo, s3
256+
; GFX12-FAKE16-NEXT: v_mov_b32_e32 v0, s0
257+
; GFX12-FAKE16-NEXT: v_mov_b32_e32 v2, s1
258+
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
259+
; GFX12-FAKE16-NEXT: v_interp_p10_f16_f32 v3, v1, v0, v1 wait_exp:0
260+
; GFX12-FAKE16-NEXT: v_interp_p10_f16_f32 v0, v1, v0, v1 op_sel:[1,0,1,0] wait_exp:7
261+
; GFX12-FAKE16-NEXT: v_interp_p2_f16_f32 v3, v1, v2, v3 wait_exp:7
262+
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
263+
; GFX12-FAKE16-NEXT: v_interp_p2_f16_f32 v0, v1, v2, v0 op_sel:[1,0,0,0] wait_exp:7
264+
; GFX12-FAKE16-NEXT: v_add_f16_e32 v0, v3, v0
265+
; GFX12-FAKE16-NEXT: ; return to shader part epilog
159266
main_body:
160267
%p0 = call float @llvm.amdgcn.lds.param.load(i32 0, i32 0, i32 %m0)
161268
%l_p0 = call float @llvm.amdgcn.interp.inreg.p10.f16(float %p0, float %i, float %p0, i1 0)
@@ -202,6 +309,42 @@ define amdgpu_ps half @v_interp_rtz_f16(float inreg %i, float inreg %j, i32 inre
202309
; GFX11-FAKE16-NEXT: v_interp_p2_rtz_f16_f32 v0, v1, v2, v0 op_sel:[1,0,0,0] wait_exp:7
203310
; GFX11-FAKE16-NEXT: v_add_f16_e32 v0, v3, v0
204311
; GFX11-FAKE16-NEXT: ; return to shader part epilog
312+
;
313+
; GFX12-TRUE16-LABEL: v_interp_rtz_f16:
314+
; GFX12-TRUE16: ; %bb.0: ; %main_body
315+
; GFX12-TRUE16-NEXT: s_mov_b32 s3, exec_lo
316+
; GFX12-TRUE16-NEXT: s_wqm_b32 exec_lo, exec_lo
317+
; GFX12-TRUE16-NEXT: s_mov_b32 m0, s2
318+
; GFX12-TRUE16-NEXT: ds_param_load v1, attr0.x wait_va_vdst:15 wait_vm_vsrc:1
319+
; GFX12-TRUE16-NEXT: s_mov_b32 exec_lo, s3
320+
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v0, s0
321+
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v2, s1
322+
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
323+
; GFX12-TRUE16-NEXT: v_interp_p10_rtz_f16_f32 v3, v1.l, v0, v1.l wait_exp:0
324+
; GFX12-TRUE16-NEXT: v_interp_p10_rtz_f16_f32 v4, v1.h, v0, v1.h wait_exp:7
325+
; GFX12-TRUE16-NEXT: v_interp_p2_rtz_f16_f32 v0.l, v1.l, v2, v3 wait_exp:7
326+
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
327+
; GFX12-TRUE16-NEXT: v_interp_p2_rtz_f16_f32 v0.h, v1.h, v2, v4 wait_exp:7
328+
; GFX12-TRUE16-NEXT: v_add_f16_e32 v0.l, v0.l, v0.h
329+
; GFX12-TRUE16-NEXT: ; return to shader part epilog
330+
;
331+
; GFX12-FAKE16-LABEL: v_interp_rtz_f16:
332+
; GFX12-FAKE16: ; %bb.0: ; %main_body
333+
; GFX12-FAKE16-NEXT: s_mov_b32 s3, exec_lo
334+
; GFX12-FAKE16-NEXT: s_wqm_b32 exec_lo, exec_lo
335+
; GFX12-FAKE16-NEXT: s_mov_b32 m0, s2
336+
; GFX12-FAKE16-NEXT: ds_param_load v1, attr0.x wait_va_vdst:15 wait_vm_vsrc:1
337+
; GFX12-FAKE16-NEXT: s_mov_b32 exec_lo, s3
338+
; GFX12-FAKE16-NEXT: v_mov_b32_e32 v0, s0
339+
; GFX12-FAKE16-NEXT: v_mov_b32_e32 v2, s1
340+
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
341+
; GFX12-FAKE16-NEXT: v_interp_p10_rtz_f16_f32 v3, v1, v0, v1 wait_exp:0
342+
; GFX12-FAKE16-NEXT: v_interp_p10_rtz_f16_f32 v0, v1, v0, v1 op_sel:[1,0,1,0] wait_exp:7
343+
; GFX12-FAKE16-NEXT: v_interp_p2_rtz_f16_f32 v3, v1, v2, v3 wait_exp:7
344+
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
345+
; GFX12-FAKE16-NEXT: v_interp_p2_rtz_f16_f32 v0, v1, v2, v0 op_sel:[1,0,0,0] wait_exp:7
346+
; GFX12-FAKE16-NEXT: v_add_f16_e32 v0, v3, v0
347+
; GFX12-FAKE16-NEXT: ; return to shader part epilog
205348
main_body:
206349
%p0 = call float @llvm.amdgcn.lds.param.load(i32 0, i32 0, i32 %m0)
207350
%l_p0 = call float @llvm.amdgcn.interp.p10.rtz.f16(float %p0, float %i, float %p0, i1 0)
@@ -237,6 +380,31 @@ define amdgpu_ps half @v_interp_f16_imm_params(float inreg %i, float inreg %j) #
237380
; GFX11-FAKE16-NEXT: v_cvt_f16_f32_e32 v1, v1
238381
; GFX11-FAKE16-NEXT: v_add_f16_e32 v0, v1, v0
239382
; GFX11-FAKE16-NEXT: ; return to shader part epilog
383+
;
384+
; GFX12-TRUE16-LABEL: v_interp_f16_imm_params:
385+
; GFX12-TRUE16: ; %bb.0: ; %main_body
386+
; GFX12-TRUE16-NEXT: v_dual_mov_b32 v1, s0 :: v_dual_mov_b32 v2, 0
387+
; GFX12-TRUE16-NEXT: v_mov_b16_e32 v0.l, 0
388+
; GFX12-TRUE16-NEXT: v_mov_b32_e32 v3, s1
389+
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
390+
; GFX12-TRUE16-NEXT: v_interp_p10_f16_f32 v1, v0.l, v1, v0.l wait_exp:7
391+
; GFX12-TRUE16-NEXT: v_interp_p2_f16_f32 v0.l, v0.l, v3, v2 wait_exp:7
392+
; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
393+
; GFX12-TRUE16-NEXT: v_cvt_f16_f32_e32 v0.h, v1
394+
; GFX12-TRUE16-NEXT: v_add_f16_e32 v0.l, v0.h, v0.l
395+
; GFX12-TRUE16-NEXT: ; return to shader part epilog
396+
;
397+
; GFX12-FAKE16-LABEL: v_interp_f16_imm_params:
398+
; GFX12-FAKE16: ; %bb.0: ; %main_body
399+
; GFX12-FAKE16-NEXT: v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v1, s0
400+
; GFX12-FAKE16-NEXT: v_mov_b32_e32 v2, s1
401+
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
402+
; GFX12-FAKE16-NEXT: v_interp_p10_f16_f32 v1, v0, v1, v0 wait_exp:7
403+
; GFX12-FAKE16-NEXT: v_interp_p2_f16_f32 v0, v0, v2, v0 wait_exp:7
404+
; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
405+
; GFX12-FAKE16-NEXT: v_cvt_f16_f32_e32 v1, v1
406+
; GFX12-FAKE16-NEXT: v_add_f16_e32 v0, v1, v0
407+
; GFX12-FAKE16-NEXT: ; return to shader part epilog
240408
main_body:
241409
%l_p0 = call float @llvm.amdgcn.interp.inreg.p10.f16(float 0.0, float %i, float 0.0, i1 0)
242410
%l_p1 = call half @llvm.amdgcn.interp.inreg.p2.f16(float 0.0, float %j, float 0.0, i1 0)

0 commit comments

Comments
 (0)