File: sgpr-copy-duplicate-operand.ll

package info (click to toggle)
llvm-toolchain-11 1%3A11.0.1-2
  • links: PTS, VCS
  • area: main
  • in suites: bullseye
  • size: 995,808 kB
  • sloc: cpp: 4,767,656; ansic: 760,916; asm: 477,436; python: 170,940; objc: 69,804; lisp: 29,914; sh: 23,855; f90: 18,173; pascal: 7,551; perl: 7,471; ml: 5,603; awk: 3,489; makefile: 2,573; xml: 915; cs: 573; fortran: 503; javascript: 452
file content (19 lines) | stat: -rw-r--r-- 925 bytes parent folder | download | duplicates (15) 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
 
; RUN:  llc -amdgpu-scalarize-global-loads=false  -march=amdgcn -verify-machineinstrs< %s | FileCheck -check-prefix=SI %s
; RUN:  llc -amdgpu-scalarize-global-loads=false  -march=amdgcn -mcpu=tonga -verify-machineinstrs< %s | FileCheck -check-prefix=SI %s

; Copy VGPR -> SGPR used twice as an instruction operand, which is then
; used in an REG_SEQUENCE that also needs to be handled.

; SI-LABEL: {{^}}test_dup_operands:
; SI: v_add_{{[iu]}}32_e32
define amdgpu_kernel void @test_dup_operands(<2 x i32> addrspace(1)* noalias %out, <2 x i32> addrspace(1)* noalias %in) {
  %a = load <2 x i32>, <2 x i32> addrspace(1)* %in
  %lo = extractelement <2 x i32> %a, i32 0
  %hi = extractelement <2 x i32> %a, i32 1
  %add = add i32 %lo, %lo
  %vec0 = insertelement <2 x i32> undef, i32 %add, i32 0
  %vec1 = insertelement <2 x i32> %vec0, i32 %hi, i32 1
  store <2 x i32> %vec1, <2 x i32> addrspace(1)* %out, align 8
  ret void
}