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
|
;=========================== begin_copyright_notice ============================
;
; Copyright (C) 2022-2024 Intel Corporation
;
; SPDX-License-Identifier: MIT
;
;============================ end_copyright_notice =============================
;
; REQUIRES: llvm-14-plus
; RUN: igc_opt --opaque-pointers -enable-debugify -loop-gating -S < %s 2>&1 | FileCheck %s --check-prefixes=CHECK
; ------------------------------------------------
; GatingSimilarSamples
; ------------------------------------------------
; Test checks gating of similar sample instructions(to skip redundant) :
; New if-then block with gating condition is created
; Corresponding phi nodes(x,y,z) have fast math flag set
; Debug-info related check
; CHECK-NOT: WARNING
; CHECK: CheckModuleDebugify: PASS
define void @test(float %src1, float %src2, float %src3, ptr %dst) {
; CHECK-LABEL: @test(
; CHECK: [[MUL:%[A-z0-9]*]] = fmul float [[SRC1:%[A-z0-9]*]], [[SRC2:%[A-z0-9]*]]
; CHECK: [[A:%[A-z0-9]*]] = fsub float [[SRC1]], [[MUL]]
; CHECK: [[B:%[A-z0-9]*]] = fadd float [[SRC1]], [[MUL]]
; CHECK: [[TMP1:%[A-z0-9]*]] = call <4 x float> @llvm.genx.GenISA.sampleptr.v4f32.f32.p196609.p524293(float [[SRC1]], float [[SRC2]], float 0.000000e+00, float 1.000000e+00, float [[SRC2]], float [[SRC3:%[A-z0-9]*]], ptr addrspace(196609) null, ptr addrspace(524293) inttoptr (i64 5 to ptr addrspace(524293)), i32 0, i32 0, i32 0)
; CHECK: [[TMP2:%[A-z0-9]*]] = call <4 x float> @llvm.genx.GenISA.sampleptr.v4f32.f32.p196609.p524293(float [[B]], float [[A]], float [[SRC3]], float 1.000000e+00, float [[SRC2]], float [[SRC3]], ptr addrspace(196609) null, ptr addrspace(524293) inttoptr (i64 5 to ptr addrspace(524293)), i32 0, i32 0, i32 0)
; CHECK: [[TMP3:%[A-z0-9]*]] = extractelement <4 x float> [[TMP2]], i32 0
; CHECK: [[TMP4:%[A-z0-9]*]] = extractelement <4 x float> [[TMP2]], i32 1
; CHECK: [[TMP5:%[A-z0-9]*]] = extractelement <4 x float> [[TMP2]], i32 2
; CHECK: [[TMP6:%[A-z0-9]*]] = fcmp fast one float [[MUL]], 0.000000e+00
; CHECK: [[TMP7:%[A-z0-9]*]] = fcmp fast one float [[MUL]], 0.000000e+00
; CHECK: [[TMP8:%[A-z0-9]*]] = or i1 [[TMP6]], [[TMP7]]
; CHECK: br i1 [[TMP8]], label %[[TMP9:[A-z0-9]*]], label %[[TMP17:[A-z0-9]*]]
; CHECK [[TMP9]]:
; CHECK: [[TMP10:%[A-z0-9]*]] = call <4 x float> @llvm.genx.GenISA.sampleptr.v4f32.f32.p196609.p524293(float [[A]], float [[B]], float [[SRC3]], float 1.000000e+00, float [[SRC2]], float [[SRC3]], ptr addrspace(196609) null, ptr addrspace(524293) inttoptr (i64 5 to ptr addrspace(524293)), i32 0, i32 0, i32 0)
; CHECK: [[TMP11:%[A-z0-9]*]] = extractelement <4 x float> [[TMP10]], i32 0
; CHECK: [[TMP12:%[A-z0-9]*]] = extractelement <4 x float> [[TMP10]], i32 1
; CHECK: [[TMP13:%[A-z0-9]*]] = extractelement <4 x float> [[TMP10]], i32 2
; CHECK: [[TMP14:%[A-z0-9]*]] = fmul float [[TMP11]], 5.000000e-01
; CHECK: [[TMP15:%[A-z0-9]*]] = fmul float 5.000000e-01, [[TMP12]]
; CHECK: [[TMP16:%[A-z0-9]*]] = fmul float [[TMP13]], 5.000000e-01
; CHECK: br label %[[TMP17]]
; CHECK [[TMP17]]:
; Fast flags are not supported on phi for llvm9
; CHECK-PRE-LLVM-14: [[TMP18:%[A-z0-9]*]] = phi{{.*}} float [ [[TMP14]], %[[TMP9]] ], [ [[TMP3]], [[TMP0:%[A-z0-9]*]] ]
; CHECK-PRE-LLVM-14: [[TMP19:%[A-z0-9]*]] = phi{{.*}} float [ [[TMP15]], %[[TMP9]] ], [ [[TMP4]], [[TMP0]] ]
; CHECK-PRE-LLVM-14: [[TMP20:%[A-z0-9]*]] = phi{{.*}} float [ [[TMP16]], %[[TMP9]] ], [ [[TMP5]], [[TMP0]] ]
; CHECK: [[TMP18:%[A-z0-9]*]] = phi fast float [ [[TMP14]], %[[TMP9]] ], [ [[TMP3]], [[TMP0:%[A-z0-9]*]] ]
; CHECK: [[TMP19:%[A-z0-9]*]] = phi fast float [ [[TMP15]], %[[TMP9]] ], [ [[TMP4]], [[TMP0]] ]
; CHECK: [[TMP20:%[A-z0-9]*]] = phi fast float [ [[TMP16]], %[[TMP9]] ], [ [[TMP5]], [[TMP0]] ]
; CHECK: call void @llvm.genx.GenISA.OUTPUT.f32(float [[TMP18]], float [[TMP19]], float [[TMP20]], float 0.000000e+00, i32 1, i32 1, i32 15)
; CHECK: ret void
;
%mul = fmul float %src1, %src2
%a = fsub float %src1, %mul
%b = fadd float %src1, %mul
%1 = call <4 x float> @llvm.genx.GenISA.sampleptr.v4f32.f32.p196609.p524293(float %src1, float %src2, float 0.000000e+00, float 1.000000e+00, float %src2, float %src3, ptr addrspace(196609) null, ptr addrspace(524293) inttoptr (i64 5 to ptr addrspace(524293)), i32 0, i32 0, i32 0)
%2 = call <4 x float> @llvm.genx.GenISA.sampleptr.v4f32.f32.p196609.p524293(float %b, float %a, float %src3, float 1.000000e+00, float %src2, float %src3, ptr addrspace(196609) null, ptr addrspace(524293) inttoptr (i64 5 to ptr addrspace(524293)), i32 0, i32 0, i32 0)
%3 = extractelement <4 x float> %2, i32 0
%4 = extractelement <4 x float> %2, i32 1
%5 = extractelement <4 x float> %2, i32 2
%6 = call <4 x float> @llvm.genx.GenISA.sampleptr.v4f32.f32.p196609.p524293(float %a, float %b, float %src3, float 1.000000e+00, float %src2, float %src3, ptr addrspace(196609) null, ptr addrspace(524293) inttoptr (i64 5 to ptr addrspace(524293)), i32 0, i32 0, i32 0)
%7 = extractelement <4 x float> %6, i32 0
%8 = extractelement <4 x float> %6, i32 1
%9 = extractelement <4 x float> %6, i32 2
%10 = fmul float %7, 5.000000e-01
%11 = fmul float 5.000000e-01, %8
%12 = fmul float %9, 5.000000e-01
call void @llvm.genx.GenISA.OUTPUT.f32(float %10, float %11, float %12, float 0.000000e+00, i32 1, i32 1, i32 15)
ret void
}
declare <4 x float> @llvm.genx.GenISA.sampleptr.v4f32.f32.p196609.p524293(float, float, float, float, float, float, ptr addrspace(196609), ptr addrspace(524293), i32, i32, i32)
declare void @llvm.genx.GenISA.OUTPUT.f32(float, float, float, float, i32, i32, i32)
|
