File: add_i128.ll

package info (click to toggle)
llvm-toolchain-16 1%3A16.0.6-15~deb12u1
  • links: PTS, VCS
  • area: main
  • in suites: bookworm
  • size: 1,634,792 kB
  • sloc: cpp: 6,179,261; ansic: 1,216,205; asm: 741,319; python: 196,614; objc: 75,325; f90: 49,640; lisp: 32,396; pascal: 12,286; sh: 9,394; perl: 7,442; ml: 5,494; awk: 3,523; makefile: 2,723; javascript: 1,206; xml: 886; fortran: 581; cs: 573
file content (56 lines) | stat: -rw-r--r-- 2,037 bytes parent folder | download | duplicates (4)
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
52
53
54
55
56
; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s

; GCN-LABEL: {{^}}test_i128_vreg:
; GCN: v_add_i32_e32 v[[LO:[0-9]+]], vcc,
; GCN-NEXT: v_addc_u32_e32 v{{[0-9]+}}, vcc, v{{[0-9]+}}, v{{[0-9]+}}, vcc
; GCN-NEXT: v_addc_u32_e32 v{{[0-9]+}}, vcc, v{{[0-9]+}}, v{{[0-9]+}}, vcc
; GCN-NEXT: v_addc_u32_e32 v[[HI:[0-9]+]], vcc, v{{[0-9]+}}, v{{[0-9]+}}, vcc
; GCN: buffer_store_dwordx4 v[[[LO]]:[[HI]]],
define amdgpu_kernel void @test_i128_vreg(ptr addrspace(1) noalias %out, ptr addrspace(1) noalias %inA, ptr addrspace(1) noalias %inB) {
  %tid = call i32 @llvm.amdgcn.workitem.id.x() readnone
  %a_ptr = getelementptr i128, ptr addrspace(1) %inA, i32 %tid
  %b_ptr = getelementptr i128, ptr addrspace(1) %inB, i32 %tid
  %a = load i128, ptr addrspace(1) %a_ptr
  %b = load i128, ptr addrspace(1) %b_ptr
  %result = add i128 %a, %b
  store i128 %result, ptr addrspace(1) %out
  ret void
}

; Check that the SGPR add operand is correctly moved to a VGPR.
; GCN-LABEL: {{^}}sgpr_operand:
; GCN: s_add_u32
; GCN: s_addc_u32
; GCN: s_addc_u32
; GCN: s_addc_u32
define amdgpu_kernel void @sgpr_operand(ptr addrspace(1) noalias %out, ptr addrspace(1) noalias %in, i128 %a) {
  %foo = load i128, ptr addrspace(1) %in, align 8
  %result = add i128 %foo, %a
  store i128 %result, ptr addrspace(1) %out
  ret void
}

; GCN-LABEL: {{^}}sgpr_operand_reversed:
; GCN: s_add_u32
; GCN: s_addc_u32
; GCN: s_addc_u32
; GCN: s_addc_u32
define amdgpu_kernel void @sgpr_operand_reversed(ptr addrspace(1) noalias %out, ptr addrspace(1) noalias %in, i128 %a) {
  %foo = load i128, ptr addrspace(1) %in, align 8
  %result = add i128 %a, %foo
  store i128 %result, ptr addrspace(1) %out
  ret void
}

; GCN-LABEL: {{^}}test_sreg:
; GCN: s_add_u32
; GCN: s_addc_u32
; GCN: s_addc_u32
; GCN: s_addc_u32
define amdgpu_kernel void @test_sreg(ptr addrspace(1) noalias %out, i128 %a, i128 %b) {
  %result = add i128 %a, %b
  store i128 %result, ptr addrspace(1) %out
  ret void
}

declare i32 @llvm.amdgcn.workitem.id.x() readnone