nils.hoelscher
/
llvm-for-llvmta
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity
llvm-for-llvmta/test/Transforms/LowerMatrixIntrinsics/transpose-float-row-major.ll

112 lines
7.8 KiB
LLVM
Raw Permalink Normal View History

first commit
2022-04-25 10:02:23 +02:00
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt -lower-matrix-intrinsics -matrix-default-layout=row-major -S < %s | FileCheck --check-prefix=RM %s
define <8 x float> @transpose(<8 x float> %a) {
; RM-LABEL: @transpose(
; RM-NEXT: entry:
; RM-NEXT: [[SPLIT:%.*]] = shufflevector <8 x float> [[A:%.*]], <8 x float> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
; RM-NEXT: [[SPLIT1:%.*]] = shufflevector <8 x float> [[A]], <8 x float> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
; RM-NEXT: [[TMP0:%.*]] = extractelement <4 x float> [[SPLIT]], i64 0
; RM-NEXT: [[TMP1:%.*]] = insertelement <2 x float> undef, float [[TMP0]], i64 0
; RM-NEXT: [[TMP2:%.*]] = extractelement <4 x float> [[SPLIT1]], i64 0
; RM-NEXT: [[TMP3:%.*]] = insertelement <2 x float> [[TMP1]], float [[TMP2]], i64 1
; RM-NEXT: [[TMP4:%.*]] = extractelement <4 x float> [[SPLIT]], i64 1
; RM-NEXT: [[TMP5:%.*]] = insertelement <2 x float> undef, float [[TMP4]], i64 0
; RM-NEXT: [[TMP6:%.*]] = extractelement <4 x float> [[SPLIT1]], i64 1
; RM-NEXT: [[TMP7:%.*]] = insertelement <2 x float> [[TMP5]], float [[TMP6]], i64 1
; RM-NEXT: [[TMP8:%.*]] = extractelement <4 x float> [[SPLIT]], i64 2
; RM-NEXT: [[TMP9:%.*]] = insertelement <2 x float> undef, float [[TMP8]], i64 0
; RM-NEXT: [[TMP10:%.*]] = extractelement <4 x float> [[SPLIT1]], i64 2
; RM-NEXT: [[TMP11:%.*]] = insertelement <2 x float> [[TMP9]], float [[TMP10]], i64 1
; RM-NEXT: [[TMP12:%.*]] = extractelement <4 x float> [[SPLIT]], i64 3
; RM-NEXT: [[TMP13:%.*]] = insertelement <2 x float> undef, float [[TMP12]], i64 0
; RM-NEXT: [[TMP14:%.*]] = extractelement <4 x float> [[SPLIT1]], i64 3
; RM-NEXT: [[TMP15:%.*]] = insertelement <2 x float> [[TMP13]], float [[TMP14]], i64 1
; RM-NEXT: [[TMP16:%.*]] = shufflevector <2 x float> [[TMP3]], <2 x float> [[TMP7]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
; RM-NEXT: [[TMP17:%.*]] = shufflevector <2 x float> [[TMP11]], <2 x float> [[TMP15]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
; RM-NEXT: [[TMP18:%.*]] = shufflevector <4 x float> [[TMP16]], <4 x float> [[TMP17]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
; RM-NEXT: ret <8 x float> [[TMP18]]
;
entry:
%c = call <8 x float> @llvm.matrix.transpose(<8 x float> %a, i32 2, i32 4)
ret <8 x float> %c
}
declare <8 x float> @llvm.matrix.transpose(<8 x float>, i32, i32)
define <8 x float> @transpose_single_column(<8 x float> %a) {
; RM-LABEL: @transpose_single_column(
; RM-NEXT: entry:
; RM-NEXT: [[SPLIT:%.*]] = shufflevector <8 x float> [[A:%.*]], <8 x float> poison, <1 x i32> zeroinitializer
; RM-NEXT: [[SPLIT1:%.*]] = shufflevector <8 x float> [[A]], <8 x float> poison, <1 x i32> <i32 1>
; RM-NEXT: [[SPLIT2:%.*]] = shufflevector <8 x float> [[A]], <8 x float> poison, <1 x i32> <i32 2>
; RM-NEXT: [[SPLIT3:%.*]] = shufflevector <8 x float> [[A]], <8 x float> poison, <1 x i32> <i32 3>
; RM-NEXT: [[SPLIT4:%.*]] = shufflevector <8 x float> [[A]], <8 x float> poison, <1 x i32> <i32 4>
; RM-NEXT: [[SPLIT5:%.*]] = shufflevector <8 x float> [[A]], <8 x float> poison, <1 x i32> <i32 5>
; RM-NEXT: [[SPLIT6:%.*]] = shufflevector <8 x float> [[A]], <8 x float> poison, <1 x i32> <i32 6>
; RM-NEXT: [[SPLIT7:%.*]] = shufflevector <8 x float> [[A]], <8 x float> poison, <1 x i32> <i32 7>
; RM-NEXT: [[TMP0:%.*]] = extractelement <1 x float> [[SPLIT]], i64 0
; RM-NEXT: [[TMP1:%.*]] = insertelement <8 x float> undef, float [[TMP0]], i64 0
; RM-NEXT: [[TMP2:%.*]] = extractelement <1 x float> [[SPLIT1]], i64 0
; RM-NEXT: [[TMP3:%.*]] = insertelement <8 x float> [[TMP1]], float [[TMP2]], i64 1
; RM-NEXT: [[TMP4:%.*]] = extractelement <1 x float> [[SPLIT2]], i64 0
; RM-NEXT: [[TMP5:%.*]] = insertelement <8 x float> [[TMP3]], float [[TMP4]], i64 2
; RM-NEXT: [[TMP6:%.*]] = extractelement <1 x float> [[SPLIT3]], i64 0
; RM-NEXT: [[TMP7:%.*]] = insertelement <8 x float> [[TMP5]], float [[TMP6]], i64 3
; RM-NEXT: [[TMP8:%.*]] = extractelement <1 x float> [[SPLIT4]], i64 0
; RM-NEXT: [[TMP9:%.*]] = insertelement <8 x float> [[TMP7]], float [[TMP8]], i64 4
; RM-NEXT: [[TMP10:%.*]] = extractelement <1 x float> [[SPLIT5]], i64 0
; RM-NEXT: [[TMP11:%.*]] = insertelement <8 x float> [[TMP9]], float [[TMP10]], i64 5
; RM-NEXT: [[TMP12:%.*]] = extractelement <1 x float> [[SPLIT6]], i64 0
; RM-NEXT: [[TMP13:%.*]] = insertelement <8 x float> [[TMP11]], float [[TMP12]], i64 6
; RM-NEXT: [[TMP14:%.*]] = extractelement <1 x float> [[SPLIT7]], i64 0
; RM-NEXT: [[TMP15:%.*]] = insertelement <8 x float> [[TMP13]], float [[TMP14]], i64 7
; RM-NEXT: ret <8 x float> [[TMP15]]
;
entry:
%c = call <8 x float> @llvm.matrix.transpose(<8 x float> %a, i32 8, i32 1)
ret <8 x float> %c
}
declare <12 x float> @llvm.matrix.transpose.v12f32(<12 x float>, i32, i32)
define <12 x float> @transpose_float_3x4(<12 x float> %a) {
; RM-LABEL: @transpose_float_3x4(
; RM-NEXT: entry:
; RM-NEXT: [[SPLIT:%.*]] = shufflevector <12 x float> [[A:%.*]], <12 x float> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
; RM-NEXT: [[SPLIT1:%.*]] = shufflevector <12 x float> [[A]], <12 x float> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
; RM-NEXT: [[SPLIT2:%.*]] = shufflevector <12 x float> [[A]], <12 x float> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
; RM-NEXT: [[TMP0:%.*]] = extractelement <4 x float> [[SPLIT]], i64 0
; RM-NEXT: [[TMP1:%.*]] = insertelement <3 x float> undef, float [[TMP0]], i64 0
; RM-NEXT: [[TMP2:%.*]] = extractelement <4 x float> [[SPLIT1]], i64 0
; RM-NEXT: [[TMP3:%.*]] = insertelement <3 x float> [[TMP1]], float [[TMP2]], i64 1
; RM-NEXT: [[TMP4:%.*]] = extractelement <4 x float> [[SPLIT2]], i64 0
; RM-NEXT: [[TMP5:%.*]] = insertelement <3 x float> [[TMP3]], float [[TMP4]], i64 2
; RM-NEXT: [[TMP6:%.*]] = extractelement <4 x float> [[SPLIT]], i64 1
; RM-NEXT: [[TMP7:%.*]] = insertelement <3 x float> undef, float [[TMP6]], i64 0
; RM-NEXT: [[TMP8:%.*]] = extractelement <4 x float> [[SPLIT1]], i64 1
; RM-NEXT: [[TMP9:%.*]] = insertelement <3 x float> [[TMP7]], float [[TMP8]], i64 1
; RM-NEXT: [[TMP10:%.*]] = extractelement <4 x float> [[SPLIT2]], i64 1
; RM-NEXT: [[TMP11:%.*]] = insertelement <3 x float> [[TMP9]], float [[TMP10]], i64 2
; RM-NEXT: [[TMP12:%.*]] = extractelement <4 x float> [[SPLIT]], i64 2
; RM-NEXT: [[TMP13:%.*]] = insertelement <3 x float> undef, float [[TMP12]], i64 0
; RM-NEXT: [[TMP14:%.*]] = extractelement <4 x float> [[SPLIT1]], i64 2
; RM-NEXT: [[TMP15:%.*]] = insertelement <3 x float> [[TMP13]], float [[TMP14]], i64 1
; RM-NEXT: [[TMP16:%.*]] = extractelement <4 x float> [[SPLIT2]], i64 2
; RM-NEXT: [[TMP17:%.*]] = insertelement <3 x float> [[TMP15]], float [[TMP16]], i64 2
; RM-NEXT: [[TMP18:%.*]] = extractelement <4 x float> [[SPLIT]], i64 3
; RM-NEXT: [[TMP19:%.*]] = insertelement <3 x float> undef, float [[TMP18]], i64 0
; RM-NEXT: [[TMP20:%.*]] = extractelement <4 x float> [[SPLIT1]], i64 3
; RM-NEXT: [[TMP21:%.*]] = insertelement <3 x float> [[TMP19]], float [[TMP20]], i64 1
; RM-NEXT: [[TMP22:%.*]] = extractelement <4 x float> [[SPLIT2]], i64 3
; RM-NEXT: [[TMP23:%.*]] = insertelement <3 x float> [[TMP21]], float [[TMP22]], i64 2
; RM-NEXT: [[TMP24:%.*]] = shufflevector <3 x float> [[TMP5]], <3 x float> [[TMP11]], <6 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5>
; RM-NEXT: [[TMP25:%.*]] = shufflevector <3 x float> [[TMP17]], <3 x float> [[TMP23]], <6 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5>
; RM-NEXT: [[TMP26:%.*]] = shufflevector <6 x float> [[TMP24]], <6 x float> [[TMP25]], <12 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>
; RM-NEXT: ret <12 x float> [[TMP26]]
;
entry:
%c = call <12 x float> @llvm.matrix.transpose.v12f32(<12 x float> %a, i32 3, i32 4)
ret <12 x float> %c
}