blob: b74e8c39a97aaf3beadb150f242746bb1b8f3b7d (
plain)
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
57
|
; RUN: llc -verify-machineinstrs -O0 -mtriple=spirv-vulkan-unknown %s -o - | FileCheck %s
; RUN: %if spirv-tools %{ llc -O0 -mtriple=spirv-vulkan-unknown %s -o - -filetype=obj | spirv-val %}
; Test lowering to spir-v backend for various types and scalar/vector
; CHECK: OpCapability GroupNonUniformArithmetic
; CHECK-DAG: %[[#f16:]] = OpTypeFloat 16
; CHECK-DAG: %[[#f32:]] = OpTypeFloat 32
; CHECK-DAG: %[[#uint:]] = OpTypeInt 32 0
; CHECK-DAG: %[[#v4_half:]] = OpTypeVector %[[#f16]] 4
; CHECK-DAG: %[[#scope:]] = OpConstant %[[#uint]] 3
; CHECK-LABEL: Begin function test_float
; CHECK: %[[#fexpr:]] = OpFunctionParameter %[[#f32]]
define float @test_float(float %fexpr) {
entry:
; CHECK: %[[#fret:]] = OpGroupNonUniformFMax %[[#f32]] %[[#scope]] Reduce %[[#fexpr]]
%0 = call float @llvm.spv.wave.reduce.max.f32(float %fexpr)
ret float %0
}
; CHECK-LABEL: Begin function test_int_signed
; CHECK: %[[#iexpr:]] = OpFunctionParameter %[[#uint]]
define i32 @test_int_signed(i32 %iexpr) {
entry:
; CHECK: %[[#iret:]] = OpGroupNonUniformSMax %[[#uint]] %[[#scope]] Reduce %[[#iexpr]]
%0 = call i32 @llvm.spv.wave.reduce.max.i32(i32 %iexpr)
ret i32 %0
}
; CHECK-LABEL: Begin function test_int_unsigned
; CHECK: %[[#iexpr:]] = OpFunctionParameter %[[#uint]]
define i32 @test_int_unsigned(i32 %iexpr) {
entry:
; CHECK: %[[#iret:]] = OpGroupNonUniformUMax %[[#uint]] %[[#scope]] Reduce %[[#iexpr]]
%0 = call i32 @llvm.spv.wave.reduce.umax.i32(i32 %iexpr)
ret i32 %0
}
; CHECK-LABEL: Begin function test_vhalf
; CHECK: %[[#vbexpr:]] = OpFunctionParameter %[[#v4_half]]
define <4 x half> @test_vhalf(<4 x half> %vbexpr) {
entry:
; CHECK: %[[#vhalfret:]] = OpGroupNonUniformFMax %[[#v4_half]] %[[#scope]] Reduce %[[#vbexpr]]
%0 = call <4 x half> @llvm.spv.wave.reduce.max.v4half(<4 x half> %vbexpr)
ret <4 x half> %0
}
declare float @llvm.spv.wave.reduce.max.f32(float)
declare i32 @llvm.spv.wave.reduce.max.i32(i32)
declare <4 x half> @llvm.spv.wave.reduce.max.v4half(<4 x half>)
declare float @llvm.spv.wave.reduce.umax.f32(float)
declare i32 @llvm.spv.wave.reduce.umax.i32(i32)
declare <4 x half> @llvm.spv.wave.reduce.umax.v4half(<4 x half>)
|
