Skip to content

Commit 8da3852

Browse files
committed
Revert "[AMDGPU] Generate test fix-wwm-vgpr-copy.ll (NFC)"
This reverts commit 5c8eb83.
1 parent 5c8eb83 commit 8da3852

File tree

1 file changed

+12
-55
lines changed

1 file changed

+12
-55
lines changed

llvm/test/CodeGen/AMDGPU/fix-wwm-vgpr-copy.ll

Lines changed: 12 additions & 55 deletions
Original file line numberDiff line numberDiff line change
@@ -1,37 +1,9 @@
1-
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
21
; RUN: llc -mtriple=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
32

43
; NOTE: llvm.amdgcn.wwm is deprecated, use llvm.amdgcn.strict.wwm instead.
54

65
; GCN-LABEL: wwm:
76
define amdgpu_hs void @wwm(i32 inreg %arg, ptr addrspace(8) inreg %buffer) {
8-
; GCN-LABEL: wwm:
9-
; GCN: ; %bb.0: ; %entry
10-
; GCN-NEXT: s_mov_b32 s7, s4
11-
; GCN-NEXT: s_mov_b32 s6, s3
12-
; GCN-NEXT: s_mov_b32 s5, s2
13-
; GCN-NEXT: s_mov_b32 s4, s1
14-
; GCN-NEXT: s_mov_b32 s1, 1
15-
; GCN-NEXT: v_mov_b32_e32 v0, 4
16-
; GCN-NEXT: s_not_b64 exec, exec
17-
; GCN-NEXT: v_mov_b32_e32 v0, 1
18-
; GCN-NEXT: s_not_b64 exec, exec
19-
; GCN-NEXT: s_or_saveexec_b64 s[2:3], -1
20-
; GCN-NEXT: v_lshlrev_b32_e32 v0, 2, v0
21-
; GCN-NEXT: s_mov_b64 exec, s[2:3]
22-
; GCN-NEXT: s_cmp_lg_u32 s0, 0
23-
; GCN-NEXT: v_mov_b32_e32 v1, v0
24-
; GCN-NEXT: s_cbranch_scc0 .LBB0_2
25-
; GCN-NEXT: ; %bb.1: ; %bb42
26-
; GCN-NEXT: s_mov_b32 s1, 0
27-
; GCN-NEXT: .LBB0_2: ; %bb602
28-
; GCN-NEXT: v_cmp_ne_u32_e32 vcc, s1, v1
29-
; GCN-NEXT: s_cbranch_vccnz .LBB0_4
30-
; GCN-NEXT: ; %bb.3: ; %bb49
31-
; GCN-NEXT: v_mov_b32_e32 v1, 1.0
32-
; GCN-NEXT: tbuffer_store_format_x v1, off, s[4:7], 1 format:[BUF_DATA_FORMAT_32,BUF_NUM_FORMAT_FLOAT] offset:4 glc
33-
; GCN-NEXT: .LBB0_4: ; %bb54
34-
; GCN-NEXT: s_endpgm
357
entry:
368
br label %work
379

@@ -51,10 +23,16 @@ bb54:
5123
ret void
5224

5325
work:
26+
; GCN: s_not_b64 exec, exec
27+
; GCN: v_mov_b32_e32 v[[tmp1189:[0-9]+]], 1
28+
; GCN: s_not_b64 exec, exec
5429
%tmp1189 = tail call i32 @llvm.amdgcn.set.inactive.i32(i32 4, i32 1)
5530

31+
; GCN: s_or_saveexec_b64 s[[[LO:[0-9]+]]:[[HI:[0-9]+]]], -1
32+
; GCN: v_lshlrev_b32_e32 v[[tmp1191:[0-9]+]], 2, v[[tmp1189]]
5633
%tmp1191 = mul i32 %tmp1189, 4
5734

35+
; GCN: s_mov_b64 exec, s[[[LO]]:[[HI]]]
5836
%tmp1196 = tail call i32 @llvm.amdgcn.wwm.i32(i32 %tmp1191)
5937

6038
%tmp34 = icmp eq i32 %arg, 0
@@ -63,33 +41,6 @@ work:
6341

6442
; GCN-LABEL: strict_wwm:
6543
define amdgpu_hs void @strict_wwm(i32 inreg %arg, ptr addrspace(8) inreg %buffer) {
66-
; GCN-LABEL: strict_wwm:
67-
; GCN: ; %bb.0: ; %entry
68-
; GCN-NEXT: s_mov_b32 s7, s4
69-
; GCN-NEXT: s_mov_b32 s6, s3
70-
; GCN-NEXT: s_mov_b32 s5, s2
71-
; GCN-NEXT: s_mov_b32 s4, s1
72-
; GCN-NEXT: s_mov_b32 s1, 1
73-
; GCN-NEXT: v_mov_b32_e32 v0, 4
74-
; GCN-NEXT: s_not_b64 exec, exec
75-
; GCN-NEXT: v_mov_b32_e32 v0, 1
76-
; GCN-NEXT: s_not_b64 exec, exec
77-
; GCN-NEXT: s_or_saveexec_b64 s[2:3], -1
78-
; GCN-NEXT: v_lshlrev_b32_e32 v0, 2, v0
79-
; GCN-NEXT: s_mov_b64 exec, s[2:3]
80-
; GCN-NEXT: s_cmp_lg_u32 s0, 0
81-
; GCN-NEXT: v_mov_b32_e32 v1, v0
82-
; GCN-NEXT: s_cbranch_scc0 .LBB1_2
83-
; GCN-NEXT: ; %bb.1: ; %bb42
84-
; GCN-NEXT: s_mov_b32 s1, 0
85-
; GCN-NEXT: .LBB1_2: ; %bb602
86-
; GCN-NEXT: v_cmp_ne_u32_e32 vcc, s1, v1
87-
; GCN-NEXT: s_cbranch_vccnz .LBB1_4
88-
; GCN-NEXT: ; %bb.3: ; %bb49
89-
; GCN-NEXT: v_mov_b32_e32 v1, 1.0
90-
; GCN-NEXT: tbuffer_store_format_x v1, off, s[4:7], 1 format:[BUF_DATA_FORMAT_32,BUF_NUM_FORMAT_FLOAT] offset:4 glc
91-
; GCN-NEXT: .LBB1_4: ; %bb54
92-
; GCN-NEXT: s_endpgm
9344
entry:
9445
br label %work
9546

@@ -109,10 +60,16 @@ bb54:
10960
ret void
11061

11162
work:
63+
; GCN: s_not_b64 exec, exec
64+
; GCN: v_mov_b32_e32 v[[tmp1189:[0-9]+]], 1
65+
; GCN: s_not_b64 exec, exec
11266
%tmp1189 = tail call i32 @llvm.amdgcn.set.inactive.i32(i32 4, i32 1)
11367

68+
; GCN: s_or_saveexec_b64 s[[[LO:[0-9]+]]:[[HI:[0-9]+]]], -1
69+
; GCN: v_lshlrev_b32_e32 v[[tmp1191:[0-9]+]], 2, v[[tmp1189]]
11470
%tmp1191 = mul i32 %tmp1189, 4
11571

72+
; GCN: s_mov_b64 exec, s[[[LO]]:[[HI]]]
11673
%tmp1196 = tail call i32 @llvm.amdgcn.strict.wwm.i32(i32 %tmp1191)
11774

11875
%tmp34 = icmp eq i32 %arg, 0

0 commit comments

Comments
 (0)