1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
|
; RUN: llc < %s -march=amdgcn -mcpu=SI -verify-machineinstrs | FileCheck %s
; RUN: llc < %s -march=amdgcn -mcpu=tonga -verify-machineinstrs | FileCheck %s
; CHECK: {{^}}inline_asm:
; CHECK: s_endpgm
; CHECK: s_endpgm
define void @inline_asm(i32 addrspace(1)* %out) {
entry:
store i32 5, i32 addrspace(1)* %out
call void asm sideeffect "s_endpgm", ""()
ret void
}
; CHECK: {{^}}inline_asm_shader:
; CHECK: s_endpgm
; CHECK: s_endpgm
define amdgpu_ps void @inline_asm_shader() {
entry:
call void asm sideeffect "s_endpgm", ""()
ret void
}
; CHECK: {{^}}branch_on_asm:
; Make sure inline assembly is treted as divergent.
; CHECK: s_mov_b32 s{{[0-9]+}}, 0
; CHECK: s_and_saveexec_b64
define void @branch_on_asm(i32 addrspace(1)* %out) {
%zero = call i32 asm "s_mov_b32 $0, 0", "=s"()
%cmp = icmp eq i32 %zero, 0
br i1 %cmp, label %if, label %endif
if:
store i32 0, i32 addrspace(1)* %out
br label %endif
endif:
ret void
}
; CHECK: {{^}}v_cmp_asm:
; CHECK: v_mov_b32_e32 [[SRC:v[0-9]+]], s{{[0-9]+}}
; CHECK: v_cmp_ne_i32_e64 s{{\[}}[[MASK_LO:[0-9]+]]:[[MASK_HI:[0-9]+]]{{\]}}, 0, [[SRC]]
; CHECK-DAG: v_mov_b32_e32 v[[V_LO:[0-9]+]], s[[MASK_LO]]
; CHECK-DAG: v_mov_b32_e32 v[[V_HI:[0-9]+]], s[[MASK_HI]]
; CHECK: buffer_store_dwordx2 v{{\[}}[[V_LO]]:[[V_HI]]{{\]}}
define void @v_cmp_asm(i64 addrspace(1)* %out, i32 %in) {
%sgpr = tail call i64 asm "v_cmp_ne_i32_e64 $0, 0, $1", "=s,v"(i32 %in)
store i64 %sgpr, i64 addrspace(1)* %out
ret void
}
|
