# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py # # Check that we can recognize a shuffle mask for a zip instruction, and produce # G_ZIP1 or G_ZIP2 where appropriate. # # RUN: llc -mtriple aarch64 -run-pass=aarch64-postlegalizer-lowering -verify-machineinstrs %s -o - | FileCheck %s ... --- name: zip1_v2s32 alignment: 4 legalized: true tracksRegLiveness: true body: | bb.1.entry: liveins: $d0, $d1 ; CHECK-LABEL: name: zip1_v2s32 ; CHECK: liveins: $d0, $d1 ; CHECK: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $d0 ; CHECK: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $d1 ; CHECK: [[ZIP1_:%[0-9]+]]:_(<2 x s32>) = G_ZIP1 [[COPY]], [[COPY1]] ; CHECK: $d0 = COPY [[ZIP1_]](<2 x s32>) ; CHECK: RET_ReallyLR implicit $d0 %0:_(<2 x s32>) = COPY $d0 %1:_(<2 x s32>) = COPY $d1 %2:_(<2 x s32>) = G_SHUFFLE_VECTOR %0(<2 x s32>), %1, shufflemask(0, 2) $d0 = COPY %2(<2 x s32>) RET_ReallyLR implicit $d0 ... --- name: zip1_v2s64 alignment: 4 legalized: true tracksRegLiveness: true body: | bb.1.entry: liveins: $q0, $q1 ; CHECK-LABEL: name: zip1_v2s64 ; CHECK: liveins: $q0, $q1 ; CHECK: [[COPY:%[0-9]+]]:_(<2 x s64>) = COPY $q0 ; CHECK: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $q1 ; CHECK: [[ZIP1_:%[0-9]+]]:_(<2 x s64>) = G_ZIP1 [[COPY]], [[COPY1]] ; CHECK: $q0 = COPY [[ZIP1_]](<2 x s64>) ; CHECK: RET_ReallyLR implicit $q0 %0:_(<2 x s64>) = COPY $q0 %1:_(<2 x s64>) = COPY $q1 %2:_(<2 x s64>) = G_SHUFFLE_VECTOR %0(<2 x s64>), %1, shufflemask(0, 2) $q0 = COPY %2(<2 x s64>) RET_ReallyLR implicit $q0 ... --- name: zip1_v4s32 alignment: 4 legalized: true tracksRegLiveness: true body: | bb.1.entry: liveins: $q0, $q1 ; CHECK-LABEL: name: zip1_v4s32 ; CHECK: liveins: $q0, $q1 ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $q0 ; CHECK: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $q1 ; CHECK: [[ZIP1_:%[0-9]+]]:_(<4 x s32>) = G_ZIP1 [[COPY]], [[COPY1]] ; CHECK: $q0 = COPY [[ZIP1_]](<4 x s32>) ; CHECK: RET_ReallyLR implicit $q0 %0:_(<4 x s32>) = COPY $q0 %1:_(<4 x s32>) = COPY $q1 %2:_(<4 x s32>) = G_SHUFFLE_VECTOR %0(<4 x s32>), %1, shufflemask(0, 4, 1, 5) $q0 = COPY %2(<4 x s32>) RET_ReallyLR implicit $q0 ... --- name: zip2_v2s32 alignment: 4 legalized: true tracksRegLiveness: true body: | bb.1.entry: liveins: $d0, $d1 ; CHECK-LABEL: name: zip2_v2s32 ; CHECK: liveins: $d0, $d1 ; CHECK: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $d0 ; CHECK: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $d1 ; CHECK: [[ZIP2_:%[0-9]+]]:_(<2 x s32>) = G_ZIP2 [[COPY]], [[COPY1]] ; CHECK: $d0 = COPY [[ZIP2_]](<2 x s32>) ; CHECK: RET_ReallyLR implicit $d0 %0:_(<2 x s32>) = COPY $d0 %1:_(<2 x s32>) = COPY $d1 %2:_(<2 x s32>) = G_SHUFFLE_VECTOR %0(<2 x s32>), %1, shufflemask(1, 3) $d0 = COPY %2(<2 x s32>) RET_ReallyLR implicit $d0 ... --- name: zip2_v2s64 alignment: 4 legalized: true tracksRegLiveness: true body: | bb.1.entry: liveins: $q0, $q1 ; CHECK-LABEL: name: zip2_v2s64 ; CHECK: liveins: $q0, $q1 ; CHECK: [[COPY:%[0-9]+]]:_(<2 x s64>) = COPY $q0 ; CHECK: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $q1 ; CHECK: [[ZIP2_:%[0-9]+]]:_(<2 x s64>) = G_ZIP2 [[COPY]], [[COPY1]] ; CHECK: $q0 = COPY [[ZIP2_]](<2 x s64>) ; CHECK: RET_ReallyLR implicit $q0 %0:_(<2 x s64>) = COPY $q0 %1:_(<2 x s64>) = COPY $q1 %2:_(<2 x s64>) = G_SHUFFLE_VECTOR %0(<2 x s64>), %1, shufflemask(1, 3) $q0 = COPY %2(<2 x s64>) RET_ReallyLR implicit $q0 ... --- name: zip2_v4s32 alignment: 4 legalized: true tracksRegLiveness: true body: | bb.1.entry: liveins: $q0, $q1 ; CHECK-LABEL: name: zip2_v4s32 ; CHECK: liveins: $q0, $q1 ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $q0 ; CHECK: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $q1 ; CHECK: [[ZIP2_:%[0-9]+]]:_(<4 x s32>) = G_ZIP2 [[COPY]], [[COPY1]] ; CHECK: $q0 = COPY [[ZIP2_]](<4 x s32>) ; CHECK: RET_ReallyLR implicit $q0 %0:_(<4 x s32>) = COPY $q0 %1:_(<4 x s32>) = COPY $q1 %2:_(<4 x s32>) = G_SHUFFLE_VECTOR %0(<4 x s32>), %1, shufflemask(2, 6, 3, 7) $q0 = COPY %2(<4 x s32>) RET_ReallyLR implicit $q0 ... --- name: zip2_no_combine_idx_mismatch alignment: 4 legalized: true tracksRegLiveness: true body: | bb.1.entry: liveins: $q0, $q1 ; This will fail because it expects 3 to be the second element of the ; shuffle vector mask. ; ; CHECK-LABEL: name: zip2_no_combine_idx_mismatch ; CHECK: liveins: $q0, $q1 ; CHECK: [[COPY:%[0-9]+]]:_(<2 x s64>) = COPY $q0 ; CHECK: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $q1 ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 ; CHECK: [[EXT:%[0-9]+]]:_(<2 x s64>) = G_EXT [[COPY]], [[COPY1]], [[C]](s32) ; CHECK: $q0 = COPY [[EXT]](<2 x s64>) ; CHECK: RET_ReallyLR implicit $q0 %0:_(<2 x s64>) = COPY $q0 %1:_(<2 x s64>) = COPY $q1 %2:_(<2 x s64>) = G_SHUFFLE_VECTOR %0(<2 x s64>), %1, shufflemask(1, 2) $q0 = COPY %2(<2 x s64>) RET_ReallyLR implicit $q0 ... --- name: zip1_no_combine_idx_mismatch alignment: 4 legalized: true tracksRegLiveness: true body: | bb.1.entry: liveins: $q0, $q1 ; This will fail because it expects 2 to be the second element of the ; shuffle vector mask. ; ; CHECK-LABEL: name: zip1_no_combine_idx_mismatch ; CHECK: liveins: $q0, $q1 ; CHECK: [[COPY:%[0-9]+]]:_(<2 x s64>) = COPY $q0 ; CHECK: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $q1 ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 ; CHECK: [[EXT:%[0-9]+]]:_(<2 x s64>) = G_EXT [[COPY]], [[COPY1]], [[C]](s32) ; CHECK: $q0 = COPY [[EXT]](<2 x s64>) ; CHECK: RET_ReallyLR implicit $q0 %0:_(<2 x s64>) = COPY $q0 %1:_(<2 x s64>) = COPY $q1 %2:_(<2 x s64>) = G_SHUFFLE_VECTOR %0(<2 x s64>), %1, shufflemask(0, 1) $q0 = COPY %2(<2 x s64>) RET_ReallyLR implicit $q0 ... --- name: no_combine_first_elt_of_mask_must_be_zero_or_one alignment: 4 legalized: true tracksRegLiveness: true body: | bb.1.entry: liveins: $q0, $q1 ; zip1/zip2 must have 0 or 1 as the first element in the shuffle mask. ; ; CHECK-LABEL: name: no_combine_first_elt_of_mask_must_be_zero_or_one ; CHECK: liveins: $q0, $q1 ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $q0 ; CHECK: [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $q1 ; CHECK: [[SHUF:%[0-9]+]]:_(<4 x s32>) = G_SHUFFLE_VECTOR [[COPY]](<4 x s32>), [[COPY1]], shufflemask(3, 4, 1, 5) ; CHECK: $q0 = COPY [[SHUF]](<4 x s32>) ; CHECK: RET_ReallyLR implicit $q0 %0:_(<4 x s32>) = COPY $q0 %1:_(<4 x s32>) = COPY $q1 %2:_(<4 x s32>) = G_SHUFFLE_VECTOR %0(<4 x s32>), %1, shufflemask(3, 4, 1, 5) $q0 = COPY %2(<4 x s32>) RET_ReallyLR implicit $q0