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
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
|
;=========================== begin_copyright_notice ============================
;
; Copyright (C) 2024 Intel Corporation
;
; SPDX-License-Identifier: MIT
;
;============================ end_copyright_notice =============================
; RUN: igc_opt --typed-pointers -igc-wave-shuffle-index-sinking -S < %s | FileCheck %s
; ------------------------------------------------
; WaveShuffleIndexSinking
;
; Test distributive property for commutative mul instruction, and also test unhoistable add with constant operand since add does not distribute over mul
; However, without any preceding anchor, add i32 %ws, 1 is hoistable since it is just an add with a constant
; ------------------------------------------------
define void @test_compare_cases(i32* %dst0, i32* %dst1, i32* %dst2, i32* %dst3, i32* %dst4, i32* %dst5, i32 %a, i32 %b) {
; CHECK: [[HOISTED:%.*]] = mul i32 %a, 2
; CHECK: [[WS0:%.*]] = call i32 @llvm.genx.GenISA.WaveShuffleIndex.i32(i32 [[HOISTED]], i32 0, i32 0)
%ws0 = call i32 @llvm.genx.GenISA.WaveShuffleIndex.i32(i32 %a, i32 0, i32 0)
; CHECK: [[ANCHOR0_HOISTED:%.*]] = mul i32 %b, 2
; CHECK: [[ANCHOR0_WS0:%.*]] = add i32 [[ANCHOR0_HOISTED]], [[WS0]]
%add0 = add i32 %b, %ws0
%mul0 = mul i32 %add0, 2
; CHECK: [[ANCHOR1_WS0:%.*]] = add i32 [[ANCHOR0_WS0]], 1
%add0_2 = add i32 %mul0, 1
; CHECK: store i32 [[ANCHOR1_WS0]], i32* %dst0
store i32 %add0_2, i32* %dst0
; CHECK: [[WS1:%.*]] = call i32 @llvm.genx.GenISA.WaveShuffleIndex.i32(i32 [[HOISTED]], i32 1, i32 0)
%ws1 = call i32 @llvm.genx.GenISA.WaveShuffleIndex.i32(i32 %a, i32 1, i32 0)
; CHECK: [[ANCHOR0_WS1:%.*]] = add i32 [[ANCHOR0_HOISTED]], [[WS1]]
%add1 = add i32 %b, %ws1
%mul1 = mul i32 %add1, 2
; CHECK: [[ANCHOR1_WS1:%.*]] = add i32 [[ANCHOR0_WS1]], 1
%add1_2 = add i32 %mul1, 1
; CHECK: store i32 [[ANCHOR1_WS1]], i32* %dst1
store i32 %add1_2, i32* %dst1
; CHECK: [[WS2:%.*]] = call i32 @llvm.genx.GenISA.WaveShuffleIndex.i32(i32 [[HOISTED]], i32 2, i32 0)
%ws2 = call i32 @llvm.genx.GenISA.WaveShuffleIndex.i32(i32 %a, i32 2, i32 0)
; CHECK: [[ANCHOR0_WS2:%.*]] = add i32 [[ANCHOR0_HOISTED]], [[WS2]]
%add2 = add i32 %b, %ws2
%mul2 = mul i32 %add2, 2
; CHECK: [[ANCHOR1_WS2:%.*]] = add i32 [[ANCHOR0_WS2]], 1
%add2_2 = add i32 %mul2, 1
; CHECK: store i32 [[ANCHOR1_WS2]], i32* %dst2
store i32 %add2_2, i32* %dst2
; CHECK: [[WS3:%.*]] = call i32 @llvm.genx.GenISA.WaveShuffleIndex.i32(i32 [[HOISTED]], i32 3, i32 0)
%ws3 = call i32 @llvm.genx.GenISA.WaveShuffleIndex.i32(i32 %a, i32 3, i32 0)
; CHECK: [[ANCHOR0_WS3:%.*]] = add i32 [[ANCHOR0_HOISTED]], [[WS3]]
%add3 = add i32 %b, %ws3
%mul3 = mul i32 %add3, 2
; CHECK: [[ANCHOR1_WS3:%.*]] = add i32 [[ANCHOR0_WS3]], 1
%add3_2 = add i32 %mul3, 1
; CHECK: store i32 [[ANCHOR1_WS3]], i32* %dst3
store i32 %add3_2, i32* %dst3
; CHECK: [[HOISTED2:%.*]] = add i32 %a, 4
; CHECK: [[WS4:%.*]] = call i32 @llvm.genx.GenISA.WaveShuffleIndex.i32(i32 [[HOISTED2]], i32 4, i32 0)
%ws4 = call i32 @llvm.genx.GenISA.WaveShuffleIndex.i32(i32 %a, i32 4, i32 0)
%add4 = add i32 %ws4, 4
; CHECK: store i32 [[WS4]], i32* %dst4
store i32 %add4, i32* %dst4
; CHECK: [[WS5:%.*]] = call i32 @llvm.genx.GenISA.WaveShuffleIndex.i32(i32 [[HOISTED2]], i32 5, i32 0)
%ws5 = call i32 @llvm.genx.GenISA.WaveShuffleIndex.i32(i32 %a, i32 5, i32 0)
%add5 = add i32 %ws5, 4
; CHECK: store i32 [[WS5]], i32* %dst5
store i32 %add5, i32* %dst5
ret void
}
; Function Attrs: convergent nounwind readnone
declare i32 @llvm.genx.GenISA.WaveShuffleIndex.i32(i32, i32, i32) #0
attributes #0 = { convergent nounwind readnone }
|
