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
|
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
; RUN: opt -S -mtriple=amdgcn-unknown-amdhsa -passes=amdgpu-promote-alloca < %s | FileCheck %s
; Check that promoting an alloca to a vector form works correctly when a variable
; vector index is used.
define amdgpu_kernel void @non_constant_index(i32 %arg) {
; CHECK-LABEL: define amdgpu_kernel void @non_constant_index(
; CHECK-SAME: i32 [[ARG:%.*]]) {
; CHECK-NEXT: bb:
; CHECK-NEXT: br label [[BB1:%.*]]
; CHECK: bb1:
; CHECK-NEXT: br label [[BB1]]
; CHECK: bb2:
; CHECK-NEXT: br label [[BB3:%.*]]
; CHECK: bb3:
; CHECK-NEXT: [[PROMOTEALLOCA:%.*]] = phi <2 x float> [ [[TMP2:%.*]], [[BB3]] ], [ poison, [[BB2:%.*]] ]
; CHECK-NEXT: [[TMP0:%.*]] = insertelement <2 x float> [[PROMOTEALLOCA]], float 0.000000e+00, i32 [[ARG]]
; CHECK-NEXT: [[TMP1:%.*]] = add i32 [[ARG]], 1
; CHECK-NEXT: [[TMP2]] = insertelement <2 x float> [[TMP0]], float 0.000000e+00, i32 [[TMP1]]
; CHECK-NEXT: br label [[BB3]]
;
bb:
%i = alloca [2 x float], align 4, addrspace(5)
br label %bb1
bb1:
br label %bb1
bb2:
br label %bb3
bb3:
%i4 = getelementptr float, ptr addrspace(5) %i, i32 %arg
store <2 x float> zeroinitializer, ptr addrspace(5) %i4, align 8
br label %bb3
}
|
