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
; RUN: opt < %s -mtriple=amdgcn-amd-amdhsa -passes=instcombine -S | FileCheck %s
; Optimize the intrinsic away if the argument is uniform.
define i32 @test_constant() {
; CHECK-LABEL: @test_constant(
; CHECK-NEXT: ret i32 99
;
%call = call i32 @llvm.amdgcn.permlane64.i32(i32 99)
ret i32 %call
}
define i32 @test_bitcast_f32_to_i32_permlane64(float %val) {
; CHECK-LABEL: @test_bitcast_f32_to_i32_permlane64(
; CHECK-NEXT: [[VAL:%.*]] = call float @llvm.amdgcn.permlane64.f32(float [[VAL1:%.*]])
; CHECK-NEXT: [[BITCAST:%.*]] = bitcast float [[VAL]] to i32
; CHECK-NEXT: ret i32 [[BITCAST]]
;
%bitcast = bitcast float %val to i32
%result = call i32 @llvm.amdgcn.permlane64.i32(i32 %bitcast)
ret i32 %result
}
define i32 @test_bitcast_f32_to_i32_permlane64_convergencetokenn(float %val) convergent {
; CHECK-LABEL: @test_bitcast_f32_to_i32_permlane64_convergencetokenn(
; CHECK-NEXT: [[T:%.*]] = call token @llvm.experimental.convergence.entry()
; CHECK-NEXT: [[VAL:%.*]] = call float @llvm.amdgcn.permlane64.f32(float [[VAL1:%.*]]) [ "convergencectrl"(token [[T]]) ]
; CHECK-NEXT: [[BITCAST:%.*]] = bitcast float [[VAL]] to i32
; CHECK-NEXT: ret i32 [[BITCAST]]
;
%t = call token @llvm.experimental.convergence.entry()
%bitcast = bitcast float %val to i32
%result = call i32 @llvm.amdgcn.permlane64.i32(i32 %bitcast) [ "convergencectrl"(token %t) ]
ret i32 %result
}
declare i32 @llvm.amdgcn.permlane64.i32(i32)
|
