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
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
|
;=========================== begin_copyright_notice ============================
;
; Copyright (C) 2025 Intel Corporation
;
; SPDX-License-Identifier: MIT
;
;============================ end_copyright_notice =============================
; REQUIRES: llvm-14-plus, regkeys
; RUN: igc_opt --opaque-pointers %s -S -inputocl -igc-ldstcombine -regkey=EnableLdStCombine=1 \
; RUN: -platformbmg \
; RUN: | FileCheck %s
; This is to test that symbolic expression can handle sext/zext instructions
target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v16:16:16-v24:32:32-v32:32:32-v48:64:64-v64:64:64-v96:128:128-v128:128:128-v192:256:256-v256:256:256-v512:512:512-v1024:1024:1024-n8:16:32"
target triple = "spir64-unknown-unknown"
; Function Attrs: convergent nounwind
define spir_kernel void @test_st(<2 x i32> addrspace(1)* %d, <4 x i32> addrspace(1)* %s, <8 x i32> %r0, <8 x i32> %payloadHeader, <3 x i32> %enqueuedLocalSize, i16 %localIdX, i16 %localIdY, i16 %localIdZ) #0 {
entry:
%payloadHeader.scalar = extractelement <8 x i32> %payloadHeader, i32 0
%enqueuedLocalSize.scalar = extractelement <3 x i32> %enqueuedLocalSize, i32 0
%r0.scalar17 = extractelement <8 x i32> %r0, i32 1
%mul.i.i.i = mul i32 %enqueuedLocalSize.scalar, %r0.scalar17
%localIdX2 = zext i16 %localIdX to i32
%add.i.i.i = add i32 %mul.i.i.i, %localIdX2
%add4.i.i.i = add i32 %add.i.i.i, %payloadHeader.scalar
%idx0 = sext i32 %add4.i.i.i to i64
%mul = shl nuw nsw i64 %idx0, 1
%arrayidx = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %s, i64 %mul
%0 = load <4 x i32>, <4 x i32> addrspace(1)* %arrayidx, align 16
%vecinit1.assembled.vect36 = shufflevector <4 x i32> %0, <4 x i32> undef, <2 x i32> <i32 0, i32 1>
%vecinit4.assembled.vect37 = shufflevector <4 x i32> %0, <4 x i32> undef, <2 x i32> <i32 2, i32 3>
; case for sext, merge two <2 x i32> stores into one <4 x i32> store
; CHECK-LABEL: define spir_kernel void @test_st
; CHECK: load <4 x i32>,
; CHECK: [[TMP1:%.*]] = insertvalue %__StructSOALayout_ undef, <2 x i32> %{{.*}}, 0
; CHECK: [[TMP2:%.*]] = insertvalue %__StructSOALayout_ [[TMP1]], <2 x i32> %{{.*}}, 1
; CHECK: [[TMP3:%.*]] = call <4 x i32> @llvm.genx.GenISA.bitcastfromstruct.v4i32.__StructSOALayout_(%__StructSOALayout_ [[TMP2]])
; CHECK: call void @llvm.genx.GenISA.PredicatedStore.p1.v4i32(ptr addrspace(1) %{{.*}}, <4 x i32> [[TMP3]], i64 8, i1 true)
%arrayidx5 = getelementptr inbounds <2 x i32>, <2 x i32> addrspace(1)* %d, i64 %idx0
call void @llvm.genx.GenISA.PredicatedStore.p1v2i32.v2i32(<2 x i32> addrspace(1)* %arrayidx5, <2 x i32> %vecinit1.assembled.vect36, i64 8, i1 true)
%add = add nsw i32 %add4.i.i.i, 1
%idx1 = sext i32 %add to i64
%arrayidx6 = getelementptr inbounds <2 x i32>, <2 x i32> addrspace(1)* %d, i64 %idx1
call void @llvm.genx.GenISA.PredicatedStore.p1v2i32.v2i32(<2 x i32> addrspace(1)* %arrayidx6, <2 x i32> %vecinit4.assembled.vect37, i64 8, i1 true)
; case for zext, merge two <2 x i32> stores into one <4 x i32> store
; CHECK: [[TMP11:%.*]] = insertvalue %__StructSOALayout_ undef, <2 x i32> %{{.*}}, 0
; CHECK: [[TMP21:%.*]] = insertvalue %__StructSOALayout_ [[TMP11]], <2 x i32> %{{.*}}, 1
; CHECK: [[TMP31:%.*]] = call <4 x i32> @llvm.genx.GenISA.bitcastfromstruct.v4i32.__StructSOALayout_(%__StructSOALayout_ [[TMP21]])
; CHECK: call void @llvm.genx.GenISA.PredicatedStore.p1.v4i32(ptr addrspace(1) %{{.*}}, <4 x i32> [[TMP31]], i64 8, i1 true)
%add.1 = add i32 %add4.i.i.i, 10
%idx0.1 = zext i32 %add.1 to i64
%arrayidx5.1 = getelementptr inbounds <2 x i32>, <2 x i32> addrspace(1)* %d, i64 %idx0.1
call void @llvm.genx.GenISA.PredicatedStore.p1v2i32.v2i32(<2 x i32> addrspace(1)* %arrayidx5.1, <2 x i32> %vecinit1.assembled.vect36, i64 8, i1 true)
%add.1.1 = add nsw i32 %add.1, 1
%idx1.1 = zext i32 %add.1.1 to i64
%arrayidx6.1 = getelementptr inbounds <2 x i32>, <2 x i32> addrspace(1)* %d, i64 %idx1.1
call void @llvm.genx.GenISA.PredicatedStore.p1v2i32.v2i32(<2 x i32> addrspace(1)* %arrayidx6.1, <2 x i32> %vecinit4.assembled.vect37, i64 8, i1 true)
; case for zext+sext, unsafe to merge, keep two <2 x i32> stores
; CHECK: call void @llvm.genx.GenISA.PredicatedStore.p1v2i32.v2i32(ptr addrspace(1) %{{.*}}, <2 x i32>
; CHECK: call void @llvm.genx.GenISA.PredicatedStore.p1v2i32.v2i32(ptr addrspace(1) %{{.*}}, <2 x i32>
; CHECK: ret void
%add.2 = add i32 %add4.i.i.i, 20
%idx0.2 = zext i32 %add.2 to i64
%arrayidx5.2 = getelementptr inbounds <2 x i32>, <2 x i32> addrspace(1)* %d, i64 %idx0.2
call void @llvm.genx.GenISA.PredicatedStore.p1v2i32.v2i32(<2 x i32> addrspace(1)* %arrayidx5.2, <2 x i32> %vecinit1.assembled.vect36, i64 8, i1 true)
%add.2.1 = add nsw i32 %add.1, 1
%idx1.2 = sext i32 %add.2.1 to i64
%arrayidx6.2 = getelementptr inbounds <2 x i32>, <2 x i32> addrspace(1)* %d, i64 %idx1.2
call void @llvm.genx.GenISA.PredicatedStore.p1v2i32.v2i32(<2 x i32> addrspace(1)* %arrayidx6.2, <2 x i32> %vecinit4.assembled.vect37, i64 8, i1 true)
ret void
}
declare void @llvm.genx.GenISA.PredicatedStore.p1v2i32.v2i32(<2 x i32> addrspace(1)*, <2 x i32>, i64, i1)
|
