Debian Codegraph @

vc-intrinsics/GenXIntrinsics/test/SimdCFLowering/opaque_ptrs/update_mask_masked_gather4.ll

raw
1 ;=========================== begin_copyright_notice ============================
2 ;
3 ; Copyright (C) 2024 Intel Corporation
4 ;
5 ; SPDX-License-Identifier: MIT
6 ;
7 ;============================ end_copyright_notice =============================
8
9 ; REQUIRES: opaque-pointers
10 ; RUN: opt -passes=cmsimdcflowering -S < %s | FileCheck %s
11
12 @g1 = internal global <64 x i32> undef
13
14 ; CHECK: @EM = internal global <32 x i1>
15
16 define dso_local dllexport void @test(<32 x i16> %cond1, <32 x i16> %cond2, <32 x i32> %addrs, <32 x i1> %pred) {
17 %g = alloca <64 x i32>, align 512
18 %cmp1 = icmp ne <32 x i16> %cond1, zeroinitializer
19 %call = call i1 @llvm.genx.simdcf.any.v32i1(<32 x i1> %cmp1)
20 br i1 %call, label %if.then, label %if.end
21
22 ; CHECK-LABEL: if.then:
23 if.then:
24 ; CHECK: [[EM_LOAD1:%.*]] = load <32 x i1>, ptr @EM
25 ; CHECK-NEXT: [[EM_UPDATE1:%.*]] = and <32 x i1> %pred, [[EM_LOAD1]]
26 ; CHECK-NEXT: [[CALL1:%.*]] = call <64 x i32> @llvm.genx.gather4.masked.scaled2.v64i32.v32i32.v32i1(i32 12, i16 0, i32 254, i32 0, <32 x i32> %addrs, <32 x i1> [[EM_UPDATE1]])
27 %call1 = call <64 x i32> @llvm.genx.gather4.masked.scaled2.v64i32.v32i32.v32i1(i32 12, i16 0, i32 254, i32 0, <32 x i32> %addrs, <32 x i1> %pred)
28 ; CHECK: [[EM_LOAD2:%.*]] = load <32 x i1>, ptr @EM
29 ; CHECK-NEXT: [[EM_UPDATE2:%.*]] = and <32 x i1> %pred, [[EM_LOAD2]]
30 ; CHECK-NEXT: [[CHENNELEM:%.*]] = shufflevector <32 x i1> [[EM_UPDATE2]], <32 x i1> undef, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
31 ; CHECK-NEXT: [[CALL1_SIMDCFPRED1:%.*]] = select <64 x i1> [[CHENNELEM]], <64 x i32> [[CALL1]]
32 store <64 x i32> %call1, ptr %g
33 %cmp2 = icmp ne <32 x i16> %cond2, zeroinitializer
34 %nest = call i1 @llvm.genx.simdcf.any.v32i1(<32 x i1> %cmp2)
35 br i1 %nest, label %if.then2, label %if.end2
36
37 ; CHECK-LABEL: if.then2:
38 if.then2:
39 ; CHECK: [[EM_LOAD3:%.*]] = load <32 x i1>, ptr @EM
40 ; CHECK-NEXT: [[EM_UPDATE2:%.*]] = and <32 x i1> %pred, [[EM_LOAD3]]
41 ; CHECK-NEXT: [[CHENNELEM2:%.*]] = shufflevector <32 x i1> [[EM_UPDATE2]], <32 x i1> undef, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
42 ; CHECK-NEXT: %call1.simdcfpred7 = select <64 x i1> [[CHENNELEM2]], <64 x i32> [[CALL1]]
43 store <64 x i32> %call1, ptr %g
44 br label %if.end2
45
46 if.end2:
47 br label %if.end
48
49 if.end:
50 %ld = load <64 x i32>, ptr %g
51 store <64 x i32> %ld, ptr @g1
52 ret void
53 }
54
55 declare <64 x i32> @llvm.genx.gather4.masked.scaled2.v64i32.v32i32.v32i1(i32, i16, i32, i32, <32 x i32>, <32 x i1>)
56 declare <64 x i32> @llvm.genx.wrregioni.v64i32.v16i32.i16.i1(<64 x i32> %load, <64 x i32> %call, i32, i32, i32, i16, i32, i1)
57 declare i1 @llvm.genx.simdcf.any.v32i1(<32 x i1>)