Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-and.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-and.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-and.mir @@ -435,28 +435,67 @@ name: test_and_v3s16 body: | bb.0: - + liveins: $vgpr0_vgpr1_vgpr2, $vgpr3_vgpr4_vgpr5 ; CHECK-LABEL: name: test_and_v3s16 + ; CHECK: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + ; CHECK: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + ; CHECK: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) + ; CHECK: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>) ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) - ; CHECK: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) - ; CHECK: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) - ; CHECK: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) - ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 - ; CHECK: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 + ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; CHECK: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0 ; CHECK: [[AND:%[0-9]+]]:_(<4 x s16>) = G_AND [[INSERT]], [[INSERT1]] - ; CHECK: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[AND]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) - ; CHECK: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) - ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 - ; CHECK: $vgpr0_vgpr1 = COPY [[INSERT2]](<4 x s16>) - %0:_(<3 x s16>) = G_IMPLICIT_DEF - %1:_(<3 x s16>) = G_IMPLICIT_DEF - %2:_(<3 x s16>) = G_AND %0, %1 - %4:_(<4 x s16>) = G_IMPLICIT_DEF - %5:_(<4 x s16>) = G_INSERT %4, %2, 0 - $vgpr0_vgpr1 = COPY %5 + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[AND]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; CHECK: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; CHECK: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; CHECK: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; CHECK: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV4]](<3 x s16>), 0 + ; CHECK: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) + ; CHECK: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV12]](<2 x s16>) + ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 + ; CHECK: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) + ; CHECK: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV13]](<2 x s16>) + ; CHECK: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) + ; CHECK: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV8]](<3 x s16>), 0 + ; CHECK: [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT3]](<4 x s16>) + ; CHECK: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV14]](<2 x s16>) + ; CHECK: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) + ; CHECK: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV15]](<2 x s16>) + ; CHECK: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) + ; CHECK: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535 + ; CHECK: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) + ; CHECK: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C1]] + ; CHECK: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) + ; CHECK: [[AND2:%[0-9]+]]:_(s32) = G_AND [[COPY3]], [[C1]] + ; CHECK: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C]](s32) + ; CHECK: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND1]], [[SHL]] + ; CHECK: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) + ; CHECK: [[COPY4:%[0-9]+]]:_(s32) = COPY [[BITCAST1]](s32) + ; CHECK: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY4]], [[C1]] + ; CHECK: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST2]](s32) + ; CHECK: [[AND4:%[0-9]+]]:_(s32) = G_AND [[COPY5]], [[C1]] + ; CHECK: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND4]], [[C]](s32) + ; CHECK: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND3]], [[SHL1]] + ; CHECK: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) + ; CHECK: [[COPY6:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) + ; CHECK: [[AND5:%[0-9]+]]:_(s32) = G_AND [[COPY6]], [[C1]] + ; CHECK: [[COPY7:%[0-9]+]]:_(s32) = COPY [[BITCAST3]](s32) + ; CHECK: [[AND6:%[0-9]+]]:_(s32) = G_AND [[COPY7]], [[C1]] + ; CHECK: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND6]], [[C]](s32) + ; CHECK: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND5]], [[SHL2]] + ; CHECK: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32) + ; CHECK: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>) + ; CHECK: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) + %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + %1:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + %2:_(<3 x s16>), %3:_(<3 x s16>) = G_UNMERGE_VALUES %0 + %4:_(<3 x s16>), %5:_(<3 x s16>) = G_UNMERGE_VALUES %1 + %6:_(<3 x s16>) = G_AND %2, %4 + %7:_(<3 x s16>) = G_IMPLICIT_DEF + %8:_(<6 x s16>) = G_CONCAT_VECTORS %6, %7 + $vgpr0_vgpr1_vgpr2 = COPY %8 + ... --- Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-ashr.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-ashr.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-ashr.mir @@ -750,26 +750,25 @@ name: test_ashr_v3s16_v3s16 body: | bb.0: - liveins: $vgpr0_vgpr1, $vgpr2_vgpr3 - + liveins: $vgpr0_vgpr1_vgpr2, $vgpr3_vgpr4_vgpr5 ; SI-LABEL: name: test_ashr_v3s16_v3s16 - ; SI: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 - ; SI: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3 - ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0 - ; SI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY1]](<4 x s16>), 0 + ; SI: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + ; SI: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + ; SI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) + ; SI: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>) ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; SI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; SI: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) - ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0 - ; SI: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; SI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) + ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0 + ; SI: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; SI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>) ; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) - ; SI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; SI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>) ; SI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535 ; SI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST2]](s32) @@ -801,32 +800,68 @@ ; SI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND5]], [[SHL1]] ; SI: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) - ; SI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; SI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT2]](<3 x s16>), 0 - ; SI: $vgpr0_vgpr1 = COPY [[INSERT2]](<4 x s16>) + ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 + ; SI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; SI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; SI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; SI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; SI: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) + ; SI: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV12]](<2 x s16>) + ; SI: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32) + ; SI: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV13]](<2 x s16>) + ; SI: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST7]], [[C]](s32) + ; SI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV8]](<3 x s16>), 0 + ; SI: [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT3]](<4 x s16>) + ; SI: [[BITCAST8:%[0-9]+]]:_(s32) = G_BITCAST [[UV14]](<2 x s16>) + ; SI: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST8]], [[C]](s32) + ; SI: [[BITCAST9:%[0-9]+]]:_(s32) = G_BITCAST [[UV15]](<2 x s16>) + ; SI: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST9]], [[C]](s32) + ; SI: [[COPY11:%[0-9]+]]:_(s32) = COPY [[BITCAST6]](s32) + ; SI: [[AND6:%[0-9]+]]:_(s32) = G_AND [[COPY11]], [[C1]] + ; SI: [[COPY12:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) + ; SI: [[AND7:%[0-9]+]]:_(s32) = G_AND [[COPY12]], [[C1]] + ; SI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C]](s32) + ; SI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL2]] + ; SI: [[BITCAST10:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32) + ; SI: [[COPY13:%[0-9]+]]:_(s32) = COPY [[BITCAST7]](s32) + ; SI: [[AND8:%[0-9]+]]:_(s32) = G_AND [[COPY13]], [[C1]] + ; SI: [[COPY14:%[0-9]+]]:_(s32) = COPY [[BITCAST8]](s32) + ; SI: [[AND9:%[0-9]+]]:_(s32) = G_AND [[COPY14]], [[C1]] + ; SI: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND9]], [[C]](s32) + ; SI: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND8]], [[SHL3]] + ; SI: [[BITCAST11:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32) + ; SI: [[COPY15:%[0-9]+]]:_(s32) = COPY [[LSHR6]](s32) + ; SI: [[AND10:%[0-9]+]]:_(s32) = G_AND [[COPY15]], [[C1]] + ; SI: [[COPY16:%[0-9]+]]:_(s32) = COPY [[BITCAST9]](s32) + ; SI: [[AND11:%[0-9]+]]:_(s32) = G_AND [[COPY16]], [[C1]] + ; SI: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND11]], [[C]](s32) + ; SI: [[OR4:%[0-9]+]]:_(s32) = G_OR [[AND10]], [[SHL4]] + ; SI: [[BITCAST12:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32) + ; SI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST10]](<2 x s16>), [[BITCAST11]](<2 x s16>), [[BITCAST12]](<2 x s16>) + ; SI: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) ; VI-LABEL: name: test_ashr_v3s16_v3s16 - ; VI: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 - ; VI: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3 - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0 - ; VI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY1]](<4 x s16>), 0 + ; VI: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + ; VI: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) + ; VI: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>) ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; VI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; VI: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; VI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0 - ; VI: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; VI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) + ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0 + ; VI: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; VI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>) ; VI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; VI: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) - ; VI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; VI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>) ; VI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; VI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; VI: [[ASHR:%[0-9]+]]:_(s16) = G_ASHR [[TRUNC]], [[TRUNC3]](s16) @@ -843,46 +878,109 @@ ; VI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; VI: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) - ; VI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; VI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT2]](<3 x s16>), 0 - ; VI: $vgpr0_vgpr1 = COPY [[INSERT2]](<4 x s16>) + ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 + ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; VI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; VI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; VI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; VI: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) + ; VI: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV12]](<2 x s16>) + ; VI: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32) + ; VI: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV13]](<2 x s16>) + ; VI: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST7]], [[C]](s32) + ; VI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV8]](<3 x s16>), 0 + ; VI: [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT3]](<4 x s16>) + ; VI: [[BITCAST8:%[0-9]+]]:_(s32) = G_BITCAST [[UV14]](<2 x s16>) + ; VI: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST8]], [[C]](s32) + ; VI: [[BITCAST9:%[0-9]+]]:_(s32) = G_BITCAST [[UV15]](<2 x s16>) + ; VI: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST9]], [[C]](s32) + ; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535 + ; VI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST6]](s32) + ; VI: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C2]] + ; VI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) + ; VI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY3]], [[C2]] + ; VI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32) + ; VI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL2]] + ; VI: [[BITCAST10:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32) + ; VI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[BITCAST7]](s32) + ; VI: [[AND2:%[0-9]+]]:_(s32) = G_AND [[COPY4]], [[C2]] + ; VI: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST8]](s32) + ; VI: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY5]], [[C2]] + ; VI: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32) + ; VI: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL3]] + ; VI: [[BITCAST11:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32) + ; VI: [[COPY6:%[0-9]+]]:_(s32) = COPY [[LSHR6]](s32) + ; VI: [[AND4:%[0-9]+]]:_(s32) = G_AND [[COPY6]], [[C2]] + ; VI: [[COPY7:%[0-9]+]]:_(s32) = COPY [[BITCAST9]](s32) + ; VI: [[AND5:%[0-9]+]]:_(s32) = G_AND [[COPY7]], [[C2]] + ; VI: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32) + ; VI: [[OR4:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL4]] + ; VI: [[BITCAST12:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32) + ; VI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST10]](<2 x s16>), [[BITCAST11]](<2 x s16>), [[BITCAST12]](<2 x s16>) + ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) ; GFX9-LABEL: name: test_ashr_v3s16_v3s16 - ; GFX9: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 - ; GFX9: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3 - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0 - ; GFX9: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY1]](<4 x s16>), 0 + ; GFX9: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + ; GFX9: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) + ; GFX9: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>) ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; GFX9: [[EXTRACT2:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 0 - ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; GFX9: [[EXTRACT3:%[0-9]+]]:_(s16) = G_EXTRACT [[INSERT1]](<4 x s16>), 32 - ; GFX9: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0 - ; GFX9: [[EXTRACT4:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT2]](<4 x s16>), 0 - ; GFX9: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0 - ; GFX9: [[EXTRACT5:%[0-9]+]]:_(s16) = G_EXTRACT [[INSERT3]](<4 x s16>), 32 - ; GFX9: [[ASHR:%[0-9]+]]:_(<2 x s16>) = G_ASHR [[EXTRACT2]], [[EXTRACT4]](<2 x s16>) - ; GFX9: [[ASHR1:%[0-9]+]]:_(s16) = G_ASHR [[EXTRACT3]], [[EXTRACT5]](s16) + ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; GFX9: [[EXTRACT:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 0 + ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; GFX9: [[EXTRACT1:%[0-9]+]]:_(s16) = G_EXTRACT [[INSERT1]](<4 x s16>), 32 + ; GFX9: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0 + ; GFX9: [[EXTRACT2:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT2]](<4 x s16>), 0 + ; GFX9: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0 + ; GFX9: [[EXTRACT3:%[0-9]+]]:_(s16) = G_EXTRACT [[INSERT3]](<4 x s16>), 32 + ; GFX9: [[ASHR:%[0-9]+]]:_(<2 x s16>) = G_ASHR [[EXTRACT]], [[EXTRACT2]](<2 x s16>) + ; GFX9: [[ASHR1:%[0-9]+]]:_(s16) = G_ASHR [[EXTRACT1]], [[EXTRACT3]](s16) ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) - ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) - ; GFX9: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; GFX9: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; GFX9: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV4]](<3 x s16>), 0 ; GFX9: [[INSERT5:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT4]], [[ASHR]](<2 x s16>), 0 ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT5]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) - ; GFX9: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) - ; GFX9: [[INSERT6:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV4]](<3 x s16>), 0 + ; GFX9: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9: [[INSERT6:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV8]](<3 x s16>), 0 ; GFX9: [[INSERT7:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT6]], [[ASHR1]](s16), 32 ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT7]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) - ; GFX9: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) - ; GFX9: [[INSERT8:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV8]](<3 x s16>), 0 - ; GFX9: $vgpr0_vgpr1 = COPY [[INSERT8]](<4 x s16>) - %0:_(<4 x s16>) = COPY $vgpr0_vgpr1 - %1:_(<4 x s16>) = COPY $vgpr2_vgpr3 - %2:_(<3 x s16>) = G_EXTRACT %0, 0 - %3:_(<3 x s16>) = G_EXTRACT %1, 0 - %4:_(<3 x s16>) = G_ASHR %2, %3 - %5:_(<4 x s16>) = G_IMPLICIT_DEF - %6:_(<4 x s16>) = G_INSERT %5, %4, 0 - $vgpr0_vgpr1 = COPY %6 + ; GFX9: [[UV12:%[0-9]+]]:_(<3 x s16>), [[UV13:%[0-9]+]]:_(<3 x s16>), [[UV14:%[0-9]+]]:_(<3 x s16>), [[UV15:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; GFX9: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX9: [[CONCAT_VECTORS3:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX9: [[UV16:%[0-9]+]]:_(<3 x s16>), [[UV17:%[0-9]+]]:_(<3 x s16>), [[UV18:%[0-9]+]]:_(<3 x s16>), [[UV19:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS3]](<12 x s16>) + ; GFX9: [[INSERT8:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV12]](<3 x s16>), 0 + ; GFX9: [[UV20:%[0-9]+]]:_(<2 x s16>), [[UV21:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT8]](<4 x s16>) + ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV20]](<2 x s16>) + ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 + ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) + ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV21]](<2 x s16>) + ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) + ; GFX9: [[INSERT9:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV16]](<3 x s16>), 0 + ; GFX9: [[UV22:%[0-9]+]]:_(<2 x s16>), [[UV23:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT9]](<4 x s16>) + ; GFX9: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV22]](<2 x s16>) + ; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) + ; GFX9: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV23]](<2 x s16>) + ; GFX9: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) + ; GFX9: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) + ; GFX9: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) + ; GFX9: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY2]](s32), [[COPY3]](s32) + ; GFX9: [[COPY4:%[0-9]+]]:_(s32) = COPY [[BITCAST1]](s32) + ; GFX9: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST2]](s32) + ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY4]](s32), [[COPY5]](s32) + ; GFX9: [[COPY6:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) + ; GFX9: [[COPY7:%[0-9]+]]:_(s32) = COPY [[BITCAST3]](s32) + ; GFX9: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY6]](s32), [[COPY7]](s32) + ; GFX9: [[CONCAT_VECTORS4:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[BUILD_VECTOR_TRUNC2]](<2 x s16>) + ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS4]](<6 x s16>) + %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + %1:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + %2:_(<3 x s16>), %3:_(<3 x s16>) = G_UNMERGE_VALUES %0 + %4:_(<3 x s16>), %5:_(<3 x s16>) = G_UNMERGE_VALUES %1 + %6:_(<3 x s16>) = G_ASHR %2, %4 + %7:_(<3 x s16>) = G_IMPLICIT_DEF + %8:_(<6 x s16>) = G_CONCAT_VECTORS %6, %7 + $vgpr0_vgpr1_vgpr2 = COPY %8 + ... --- Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fadd.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fadd.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fadd.mir @@ -323,148 +323,32 @@ ... --- -name: test_fadd_v3s16 +name: test_faaddd_v3s16 body: | - bb.0.entry: - liveins: $vgpr0, $vgpr1 + bb.0: + liveins: $vgpr0_vgpr1_vgpr2, $vgpr3_vgpr4_vgpr5 + ; CHECK-LABEL: name: test_or_v3s16 + ; CHECK: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + ; CHECK: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + ; CHECK: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) + ; CHECK: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>) + ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; CHECK: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0 + ; CHECK: [[OR:%[0-9]+]]:_(<4 x s16>) = G_OR [[INSERT]], [[INSERT1]] + ; CHECK: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[OR]](<4 x s16>), 0 + ; CHECK: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[EXTRACT]](<3 x s16>), [[EXTRACT1]](<3 x s16>) + ; CHECK: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>) + %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + %1:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + %2:_(<3 x s16>), %3:_(<3 x s16>) = G_UNMERGE_VALUES %0 + %4:_(<3 x s16>), %5:_(<3 x s16>) = G_UNMERGE_VALUES %1 + %6:_(<3 x s16>) = G_FADD %2, %4 + %7:_(<3 x s16>) = G_IMPLICIT_DEF + %8:_(<6 x s16>) = G_CONCAT_VECTORS %6, %7 + $vgpr0_vgpr1_vgpr2 = COPY %8 - ; SI-LABEL: name: test_fadd_v3s16 - ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) - ; SI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) - ; SI: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) - ; SI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) - ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 - ; SI: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) - ; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) - ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 - ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) - ; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) - ; SI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) - ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 - ; SI: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; SI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) - ; SI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) - ; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) - ; SI: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) - ; SI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) - ; SI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) - ; SI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) - ; SI: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) - ; SI: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC3]](s16) - ; SI: [[FADD:%[0-9]+]]:_(s32) = G_FADD [[FPEXT]], [[FPEXT1]] - ; SI: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD]](s32) - ; SI: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16) - ; SI: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC4]](s16) - ; SI: [[FADD1:%[0-9]+]]:_(s32) = G_FADD [[FPEXT2]], [[FPEXT3]] - ; SI: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD1]](s32) - ; SI: [[FPEXT4:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC2]](s16) - ; SI: [[FPEXT5:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC5]](s16) - ; SI: [[FADD2:%[0-9]+]]:_(s32) = G_FADD [[FPEXT4]], [[FPEXT5]] - ; SI: [[FPTRUNC2:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD2]](s32) - ; SI: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[FPTRUNC]](s16) - ; SI: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[FPTRUNC1]](s16) - ; SI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32) - ; SI: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]] - ; SI: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) - ; SI: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[FPTRUNC2]](s16) - ; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 - ; SI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32) - ; SI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] - ; SI: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; SI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) - ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS2]](<4 x s16>), 0 - ; SI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) - ; VI-LABEL: name: test_fadd_v3s16 - ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) - ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) - ; VI: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) - ; VI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 - ; VI: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) - ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) - ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 - ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) - ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) - ; VI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) - ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 - ; VI: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; VI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) - ; VI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) - ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) - ; VI: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) - ; VI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) - ; VI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) - ; VI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) - ; VI: [[FADD:%[0-9]+]]:_(s16) = G_FADD [[TRUNC]], [[TRUNC3]] - ; VI: [[FADD1:%[0-9]+]]:_(s16) = G_FADD [[TRUNC1]], [[TRUNC4]] - ; VI: [[FADD2:%[0-9]+]]:_(s16) = G_FADD [[TRUNC2]], [[TRUNC5]] - ; VI: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[FADD]](s16) - ; VI: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[FADD1]](s16) - ; VI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32) - ; VI: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]] - ; VI: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) - ; VI: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[FADD2]](s16) - ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 - ; VI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32) - ; VI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] - ; VI: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; VI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS2]](<4 x s16>), 0 - ; VI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) - ; GFX9-LABEL: name: test_fadd_v3s16 - ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) - ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) - ; GFX9: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) - ; GFX9: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 - ; GFX9: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) - ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 - ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) - ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) - ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; GFX9: [[COPY:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) - ; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) - ; GFX9: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY]](s32), [[COPY1]](s32) - ; GFX9: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST1]](s32) - ; GFX9: [[DEF3:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF - ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY2]](s32), [[DEF3]](s32) - ; GFX9: [[DEF4:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 - ; GFX9: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; GFX9: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) - ; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) - ; GFX9: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) - ; GFX9: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) - ; GFX9: [[COPY3:%[0-9]+]]:_(s32) = COPY [[BITCAST2]](s32) - ; GFX9: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) - ; GFX9: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY3]](s32), [[COPY4]](s32) - ; GFX9: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST3]](s32) - ; GFX9: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY5]](s32), [[DEF3]](s32) - ; GFX9: [[FADD:%[0-9]+]]:_(<2 x s16>) = G_FADD [[BUILD_VECTOR_TRUNC]], [[BUILD_VECTOR_TRUNC2]] - ; GFX9: [[FADD1:%[0-9]+]]:_(<2 x s16>) = G_FADD [[BUILD_VECTOR_TRUNC1]], [[BUILD_VECTOR_TRUNC3]] - ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[FADD]](<2 x s16>), [[FADD1]](<2 x s16>), [[DEF4]](<2 x s16>) - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS2]](<6 x s16>), 0 - ; GFX9: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) - %0:_(<3 x s16>) = G_IMPLICIT_DEF - %1:_(<3 x s16>) = G_IMPLICIT_DEF - %2:_(<3 x s16>) = G_FADD %0, %1 - S_NOP 0, implicit %2 ... --- Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fma.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fma.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fma.mir @@ -390,44 +390,42 @@ name: test_fma_v3s16 body: | bb.0: + liveins: $vgpr0_vgpr1_vgpr2, $vgpr3_vgpr4_vgpr5, $vgpr6_vgpr7_vgpr8 ; SI-LABEL: name: test_fma_v3s16 + ; SI: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + ; SI: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + ; SI: [[COPY2:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr6_vgpr7_vgpr8 + ; SI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) + ; SI: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>) + ; SI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY2]](<6 x s16>) ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) - ; SI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) - ; SI: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) - ; SI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) - ; SI: [[DEF3:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) - ; SI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) - ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 - ; SI: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV12]](<2 x s16>) + ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; SI: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>) ; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV13]](<2 x s16>) + ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>) ; SI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 - ; SI: [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; SI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV14]](<2 x s16>) + ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0 + ; SI: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; SI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) ; SI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; SI: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) - ; SI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV15]](<2 x s16>) + ; SI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) ; SI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; SI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) - ; SI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 - ; SI: [[UV16:%[0-9]+]]:_(<2 x s16>), [[UV17:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) - ; SI: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV16]](<2 x s16>) + ; SI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV4]](<3 x s16>), 0 + ; SI: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) + ; SI: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) ; SI: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST4]](s32) ; SI: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32) ; SI: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32) - ; SI: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV17]](<2 x s16>) + ; SI: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) ; SI: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST5]](s32) ; SI: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32) ; SI: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) @@ -455,46 +453,81 @@ ; SI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32) ; SI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; SI: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; SI: [[CONCAT_VECTORS3:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>) - ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS3]](<4 x s16>), 0 - ; SI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) + ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>) + ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 + ; SI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; SI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; SI: [[UV12:%[0-9]+]]:_(<3 x s16>), [[UV13:%[0-9]+]]:_(<3 x s16>), [[UV14:%[0-9]+]]:_(<3 x s16>), [[UV15:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; SI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; SI: [[UV16:%[0-9]+]]:_(<2 x s16>), [[UV17:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT3]](<4 x s16>) + ; SI: [[BITCAST8:%[0-9]+]]:_(s32) = G_BITCAST [[UV16]](<2 x s16>) + ; SI: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST8]], [[C]](s32) + ; SI: [[BITCAST9:%[0-9]+]]:_(s32) = G_BITCAST [[UV17]](<2 x s16>) + ; SI: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST9]], [[C]](s32) + ; SI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV12]](<3 x s16>), 0 + ; SI: [[UV18:%[0-9]+]]:_(<2 x s16>), [[UV19:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT4]](<4 x s16>) + ; SI: [[BITCAST10:%[0-9]+]]:_(s32) = G_BITCAST [[UV18]](<2 x s16>) + ; SI: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST10]], [[C]](s32) + ; SI: [[BITCAST11:%[0-9]+]]:_(s32) = G_BITCAST [[UV19]](<2 x s16>) + ; SI: [[LSHR9:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST11]], [[C]](s32) + ; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535 + ; SI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[BITCAST8]](s32) + ; SI: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY3]], [[C2]] + ; SI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR6]](s32) + ; SI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY4]], [[C2]] + ; SI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32) + ; SI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL2]] + ; SI: [[BITCAST12:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32) + ; SI: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST9]](s32) + ; SI: [[AND2:%[0-9]+]]:_(s32) = G_AND [[COPY5]], [[C2]] + ; SI: [[COPY6:%[0-9]+]]:_(s32) = COPY [[BITCAST10]](s32) + ; SI: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY6]], [[C2]] + ; SI: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32) + ; SI: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL3]] + ; SI: [[BITCAST13:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32) + ; SI: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR8]](s32) + ; SI: [[AND4:%[0-9]+]]:_(s32) = G_AND [[COPY7]], [[C2]] + ; SI: [[COPY8:%[0-9]+]]:_(s32) = COPY [[BITCAST11]](s32) + ; SI: [[AND5:%[0-9]+]]:_(s32) = G_AND [[COPY8]], [[C2]] + ; SI: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32) + ; SI: [[OR4:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL4]] + ; SI: [[BITCAST14:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32) + ; SI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST12]](<2 x s16>), [[BITCAST13]](<2 x s16>), [[BITCAST14]](<2 x s16>) + ; SI: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) ; VI-LABEL: name: test_fma_v3s16 + ; VI: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + ; VI: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + ; VI: [[COPY2:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr6_vgpr7_vgpr8 + ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) + ; VI: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>) + ; VI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY2]](<6 x s16>) ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) - ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) - ; VI: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) - ; VI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) - ; VI: [[DEF3:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) - ; VI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 - ; VI: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV12]](<2 x s16>) + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; VI: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV13]](<2 x s16>) + ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>) ; VI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 - ; VI: [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; VI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV14]](<2 x s16>) + ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0 + ; VI: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; VI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) ; VI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; VI: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) - ; VI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV15]](<2 x s16>) + ; VI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) ; VI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; VI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) - ; VI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 - ; VI: [[UV16:%[0-9]+]]:_(<2 x s16>), [[UV17:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) - ; VI: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV16]](<2 x s16>) + ; VI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV4]](<3 x s16>), 0 + ; VI: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) + ; VI: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) ; VI: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST4]](s32) ; VI: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32) ; VI: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32) - ; VI: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV17]](<2 x s16>) + ; VI: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) ; VI: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST5]](s32) ; VI: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32) ; VI: [[FMA:%[0-9]+]]:_(s16) = G_FMA [[TRUNC]], [[TRUNC3]], [[TRUNC6]] @@ -510,66 +543,132 @@ ; VI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32) ; VI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; VI: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; VI: [[CONCAT_VECTORS3:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>) - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS3]](<4 x s16>), 0 - ; VI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) + ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>) + ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 + ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; VI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; VI: [[UV12:%[0-9]+]]:_(<3 x s16>), [[UV13:%[0-9]+]]:_(<3 x s16>), [[UV14:%[0-9]+]]:_(<3 x s16>), [[UV15:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; VI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; VI: [[UV16:%[0-9]+]]:_(<2 x s16>), [[UV17:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT3]](<4 x s16>) + ; VI: [[BITCAST8:%[0-9]+]]:_(s32) = G_BITCAST [[UV16]](<2 x s16>) + ; VI: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST8]], [[C]](s32) + ; VI: [[BITCAST9:%[0-9]+]]:_(s32) = G_BITCAST [[UV17]](<2 x s16>) + ; VI: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST9]], [[C]](s32) + ; VI: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV12]](<3 x s16>), 0 + ; VI: [[UV18:%[0-9]+]]:_(<2 x s16>), [[UV19:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT4]](<4 x s16>) + ; VI: [[BITCAST10:%[0-9]+]]:_(s32) = G_BITCAST [[UV18]](<2 x s16>) + ; VI: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST10]], [[C]](s32) + ; VI: [[BITCAST11:%[0-9]+]]:_(s32) = G_BITCAST [[UV19]](<2 x s16>) + ; VI: [[LSHR9:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST11]], [[C]](s32) + ; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535 + ; VI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[BITCAST8]](s32) + ; VI: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY3]], [[C2]] + ; VI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR6]](s32) + ; VI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY4]], [[C2]] + ; VI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32) + ; VI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL2]] + ; VI: [[BITCAST12:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32) + ; VI: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST9]](s32) + ; VI: [[AND2:%[0-9]+]]:_(s32) = G_AND [[COPY5]], [[C2]] + ; VI: [[COPY6:%[0-9]+]]:_(s32) = COPY [[BITCAST10]](s32) + ; VI: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY6]], [[C2]] + ; VI: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32) + ; VI: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL3]] + ; VI: [[BITCAST13:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32) + ; VI: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR8]](s32) + ; VI: [[AND4:%[0-9]+]]:_(s32) = G_AND [[COPY7]], [[C2]] + ; VI: [[COPY8:%[0-9]+]]:_(s32) = COPY [[BITCAST11]](s32) + ; VI: [[AND5:%[0-9]+]]:_(s32) = G_AND [[COPY8]], [[C2]] + ; VI: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32) + ; VI: [[OR4:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL4]] + ; VI: [[BITCAST14:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32) + ; VI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST12]](<2 x s16>), [[BITCAST13]](<2 x s16>), [[BITCAST14]](<2 x s16>) + ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) ; GFX9-LABEL: name: test_fma_v3s16 + ; GFX9: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + ; GFX9: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + ; GFX9: [[COPY2:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr6_vgpr7_vgpr8 + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) + ; GFX9: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>) + ; GFX9: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY2]](<6 x s16>) ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) - ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) - ; GFX9: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) - ; GFX9: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) - ; GFX9: [[DEF3:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF3]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) - ; GFX9: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 - ; GFX9: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV12]](<2 x s16>) + ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; GFX9: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) - ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV13]](<2 x s16>) + ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>) ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; GFX9: [[COPY:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) - ; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) - ; GFX9: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY]](s32), [[COPY1]](s32) - ; GFX9: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST1]](s32) - ; GFX9: [[DEF4:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF - ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY2]](s32), [[DEF4]](s32) - ; GFX9: [[DEF5:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 - ; GFX9: [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; GFX9: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV14]](<2 x s16>) + ; GFX9: [[COPY3:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) + ; GFX9: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) + ; GFX9: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY3]](s32), [[COPY4]](s32) + ; GFX9: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST1]](s32) + ; GFX9: [[DEF1:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF + ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY5]](s32), [[DEF1]](s32) + ; GFX9: [[DEF2:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF + ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0 + ; GFX9: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; GFX9: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) ; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) - ; GFX9: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV15]](<2 x s16>) + ; GFX9: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) ; GFX9: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) - ; GFX9: [[COPY3:%[0-9]+]]:_(s32) = COPY [[BITCAST2]](s32) - ; GFX9: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) - ; GFX9: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY3]](s32), [[COPY4]](s32) - ; GFX9: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST3]](s32) - ; GFX9: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY5]](s32), [[DEF4]](s32) - ; GFX9: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 - ; GFX9: [[UV16:%[0-9]+]]:_(<2 x s16>), [[UV17:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) - ; GFX9: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV16]](<2 x s16>) + ; GFX9: [[COPY6:%[0-9]+]]:_(s32) = COPY [[BITCAST2]](s32) + ; GFX9: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) + ; GFX9: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY6]](s32), [[COPY7]](s32) + ; GFX9: [[COPY8:%[0-9]+]]:_(s32) = COPY [[BITCAST3]](s32) + ; GFX9: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY8]](s32), [[DEF1]](s32) + ; GFX9: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV4]](<3 x s16>), 0 + ; GFX9: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) + ; GFX9: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) ; GFX9: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32) - ; GFX9: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV17]](<2 x s16>) + ; GFX9: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) ; GFX9: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32) - ; GFX9: [[COPY6:%[0-9]+]]:_(s32) = COPY [[BITCAST4]](s32) - ; GFX9: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) - ; GFX9: [[BUILD_VECTOR_TRUNC4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY6]](s32), [[COPY7]](s32) - ; GFX9: [[COPY8:%[0-9]+]]:_(s32) = COPY [[BITCAST5]](s32) - ; GFX9: [[BUILD_VECTOR_TRUNC5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY8]](s32), [[DEF4]](s32) + ; GFX9: [[COPY9:%[0-9]+]]:_(s32) = COPY [[BITCAST4]](s32) + ; GFX9: [[COPY10:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) + ; GFX9: [[BUILD_VECTOR_TRUNC4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY9]](s32), [[COPY10]](s32) + ; GFX9: [[COPY11:%[0-9]+]]:_(s32) = COPY [[BITCAST5]](s32) + ; GFX9: [[BUILD_VECTOR_TRUNC5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY11]](s32), [[DEF1]](s32) ; GFX9: [[FMA:%[0-9]+]]:_(<2 x s16>) = G_FMA [[BUILD_VECTOR_TRUNC]], [[BUILD_VECTOR_TRUNC2]], [[BUILD_VECTOR_TRUNC4]] ; GFX9: [[FMA1:%[0-9]+]]:_(<2 x s16>) = G_FMA [[BUILD_VECTOR_TRUNC1]], [[BUILD_VECTOR_TRUNC3]], [[BUILD_VECTOR_TRUNC5]] - ; GFX9: [[CONCAT_VECTORS3:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[FMA]](<2 x s16>), [[FMA1]](<2 x s16>), [[DEF5]](<2 x s16>) - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS3]](<6 x s16>), 0 - ; GFX9: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) - %0:_(<3 x s16>) = G_IMPLICIT_DEF - %1:_(<3 x s16>) = G_IMPLICIT_DEF - %2:_(<3 x s16>) = G_IMPLICIT_DEF - %3:_(<3 x s16>) = G_FMA %0, %1, %2 - S_NOP 0, implicit %3 + ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[FMA]](<2 x s16>), [[FMA1]](<2 x s16>), [[DEF2]](<2 x s16>) + ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<6 x s16>), 0 + ; GFX9: [[DEF3:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF3]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX9: [[UV12:%[0-9]+]]:_(<3 x s16>), [[UV13:%[0-9]+]]:_(<3 x s16>), [[UV14:%[0-9]+]]:_(<3 x s16>), [[UV15:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; GFX9: [[UV16:%[0-9]+]]:_(<2 x s16>), [[UV17:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT3]](<4 x s16>) + ; GFX9: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV16]](<2 x s16>) + ; GFX9: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32) + ; GFX9: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV17]](<2 x s16>) + ; GFX9: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST7]], [[C]](s32) + ; GFX9: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV12]](<3 x s16>), 0 + ; GFX9: [[UV18:%[0-9]+]]:_(<2 x s16>), [[UV19:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT4]](<4 x s16>) + ; GFX9: [[BITCAST8:%[0-9]+]]:_(s32) = G_BITCAST [[UV18]](<2 x s16>) + ; GFX9: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST8]], [[C]](s32) + ; GFX9: [[BITCAST9:%[0-9]+]]:_(s32) = G_BITCAST [[UV19]](<2 x s16>) + ; GFX9: [[LSHR9:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST9]], [[C]](s32) + ; GFX9: [[COPY12:%[0-9]+]]:_(s32) = COPY [[BITCAST6]](s32) + ; GFX9: [[COPY13:%[0-9]+]]:_(s32) = COPY [[LSHR6]](s32) + ; GFX9: [[BUILD_VECTOR_TRUNC6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY12]](s32), [[COPY13]](s32) + ; GFX9: [[COPY14:%[0-9]+]]:_(s32) = COPY [[BITCAST7]](s32) + ; GFX9: [[COPY15:%[0-9]+]]:_(s32) = COPY [[BITCAST8]](s32) + ; GFX9: [[BUILD_VECTOR_TRUNC7:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY14]](s32), [[COPY15]](s32) + ; GFX9: [[COPY16:%[0-9]+]]:_(s32) = COPY [[LSHR8]](s32) + ; GFX9: [[COPY17:%[0-9]+]]:_(s32) = COPY [[BITCAST9]](s32) + ; GFX9: [[BUILD_VECTOR_TRUNC8:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY16]](s32), [[COPY17]](s32) + ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC6]](<2 x s16>), [[BUILD_VECTOR_TRUNC7]](<2 x s16>), [[BUILD_VECTOR_TRUNC8]](<2 x s16>) + ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) + %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + %1:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + %2:_(<6 x s16>) = COPY $vgpr6_vgpr7_vgpr8 + %3:_(<3 x s16>), %4:_(<3 x s16>) = G_UNMERGE_VALUES %0 + %5:_(<3 x s16>), %6:_(<3 x s16>) = G_UNMERGE_VALUES %1 + %7:_(<3 x s16>), %8:_(<3 x s16>) = G_UNMERGE_VALUES %2 + %9:_(<3 x s16>) = G_FMA %3, %5, %7 + %10:_(<3 x s16>) = G_IMPLICIT_DEF + %11:_(<6 x s16>) = G_CONCAT_VECTORS %9, %10 + $vgpr0_vgpr1_vgpr2 = COPY %11 + ... --- Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fmaxnum.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fmaxnum.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fmaxnum.mir @@ -388,31 +388,30 @@ name: test_fmaxnum_v3s16 body: | bb.0: - liveins: $vgpr0_vgpr1, $vgpr2_vgpr3 - + liveins: $vgpr0_vgpr1_vgpr2, $vgpr3_vgpr4_vgpr5 ; SI-LABEL: name: test_fmaxnum_v3s16 - ; SI: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 - ; SI: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3 - ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0 - ; SI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY1]](<4 x s16>), 0 + ; SI: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + ; SI: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + ; SI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) + ; SI: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>) ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; SI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; SI: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; SI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0 - ; SI: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; SI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) + ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0 + ; SI: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; SI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>) ; SI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; SI: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) - ; SI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; SI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>) ; SI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; SI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; SI: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) @@ -438,32 +437,69 @@ ; SI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; SI: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) - ; SI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; SI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT2]](<3 x s16>), 0 - ; SI: $vgpr0_vgpr1 = COPY [[INSERT2]](<4 x s16>) + ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 + ; SI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; SI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; SI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; SI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; SI: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) + ; SI: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV12]](<2 x s16>) + ; SI: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32) + ; SI: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV13]](<2 x s16>) + ; SI: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST7]], [[C]](s32) + ; SI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV8]](<3 x s16>), 0 + ; SI: [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT3]](<4 x s16>) + ; SI: [[BITCAST8:%[0-9]+]]:_(s32) = G_BITCAST [[UV14]](<2 x s16>) + ; SI: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST8]], [[C]](s32) + ; SI: [[BITCAST9:%[0-9]+]]:_(s32) = G_BITCAST [[UV15]](<2 x s16>) + ; SI: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST9]], [[C]](s32) + ; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535 + ; SI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST6]](s32) + ; SI: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C2]] + ; SI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) + ; SI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY3]], [[C2]] + ; SI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32) + ; SI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL2]] + ; SI: [[BITCAST10:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32) + ; SI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[BITCAST7]](s32) + ; SI: [[AND2:%[0-9]+]]:_(s32) = G_AND [[COPY4]], [[C2]] + ; SI: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST8]](s32) + ; SI: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY5]], [[C2]] + ; SI: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32) + ; SI: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL3]] + ; SI: [[BITCAST11:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32) + ; SI: [[COPY6:%[0-9]+]]:_(s32) = COPY [[LSHR6]](s32) + ; SI: [[AND4:%[0-9]+]]:_(s32) = G_AND [[COPY6]], [[C2]] + ; SI: [[COPY7:%[0-9]+]]:_(s32) = COPY [[BITCAST9]](s32) + ; SI: [[AND5:%[0-9]+]]:_(s32) = G_AND [[COPY7]], [[C2]] + ; SI: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32) + ; SI: [[OR4:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL4]] + ; SI: [[BITCAST12:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32) + ; SI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST10]](<2 x s16>), [[BITCAST11]](<2 x s16>), [[BITCAST12]](<2 x s16>) + ; SI: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) ; VI-LABEL: name: test_fmaxnum_v3s16 - ; VI: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 - ; VI: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3 - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0 - ; VI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY1]](<4 x s16>), 0 + ; VI: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + ; VI: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) + ; VI: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>) ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; VI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; VI: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; VI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0 - ; VI: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; VI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) + ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0 + ; VI: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; VI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>) ; VI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; VI: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) - ; VI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; VI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>) ; VI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; VI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; VI: [[FCANONICALIZE:%[0-9]+]]:_(s16) = G_FCANONICALIZE [[TRUNC]] @@ -486,38 +522,101 @@ ; VI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; VI: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) - ; VI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; VI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT2]](<3 x s16>), 0 - ; VI: $vgpr0_vgpr1 = COPY [[INSERT2]](<4 x s16>) + ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 + ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; VI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; VI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; VI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; VI: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) + ; VI: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV12]](<2 x s16>) + ; VI: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32) + ; VI: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV13]](<2 x s16>) + ; VI: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST7]], [[C]](s32) + ; VI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV8]](<3 x s16>), 0 + ; VI: [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT3]](<4 x s16>) + ; VI: [[BITCAST8:%[0-9]+]]:_(s32) = G_BITCAST [[UV14]](<2 x s16>) + ; VI: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST8]], [[C]](s32) + ; VI: [[BITCAST9:%[0-9]+]]:_(s32) = G_BITCAST [[UV15]](<2 x s16>) + ; VI: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST9]], [[C]](s32) + ; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535 + ; VI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST6]](s32) + ; VI: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C2]] + ; VI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) + ; VI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY3]], [[C2]] + ; VI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32) + ; VI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL2]] + ; VI: [[BITCAST10:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32) + ; VI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[BITCAST7]](s32) + ; VI: [[AND2:%[0-9]+]]:_(s32) = G_AND [[COPY4]], [[C2]] + ; VI: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST8]](s32) + ; VI: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY5]], [[C2]] + ; VI: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32) + ; VI: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL3]] + ; VI: [[BITCAST11:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32) + ; VI: [[COPY6:%[0-9]+]]:_(s32) = COPY [[LSHR6]](s32) + ; VI: [[AND4:%[0-9]+]]:_(s32) = G_AND [[COPY6]], [[C2]] + ; VI: [[COPY7:%[0-9]+]]:_(s32) = COPY [[BITCAST9]](s32) + ; VI: [[AND5:%[0-9]+]]:_(s32) = G_AND [[COPY7]], [[C2]] + ; VI: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32) + ; VI: [[OR4:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL4]] + ; VI: [[BITCAST12:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32) + ; VI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST10]](<2 x s16>), [[BITCAST11]](<2 x s16>), [[BITCAST12]](<2 x s16>) + ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) ; GFX9-LABEL: name: test_fmaxnum_v3s16 - ; GFX9: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 - ; GFX9: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3 - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0 - ; GFX9: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY1]](<4 x s16>), 0 + ; GFX9: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + ; GFX9: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) + ; GFX9: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>) ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0 - ; GFX9: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; GFX9: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; GFX9: [[FCANONICALIZE:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[UV]] - ; GFX9: [[FCANONICALIZE1:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[UV2]] + ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0 + ; GFX9: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; GFX9: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; GFX9: [[FCANONICALIZE:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[UV4]] + ; GFX9: [[FCANONICALIZE1:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[UV6]] ; GFX9: [[FMAXNUM_IEEE:%[0-9]+]]:_(<2 x s16>) = G_FMAXNUM_IEEE [[FCANONICALIZE]], [[FCANONICALIZE1]] - ; GFX9: [[FCANONICALIZE2:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[UV1]] - ; GFX9: [[FCANONICALIZE3:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[UV3]] + ; GFX9: [[FCANONICALIZE2:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[UV5]] + ; GFX9: [[FCANONICALIZE3:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[UV7]] ; GFX9: [[FMAXNUM_IEEE1:%[0-9]+]]:_(<2 x s16>) = G_FMAXNUM_IEEE [[FCANONICALIZE2]], [[FCANONICALIZE3]] ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[FMAXNUM_IEEE]](<2 x s16>), [[FMAXNUM_IEEE1]](<2 x s16>) ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[CONCAT_VECTORS]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) - ; GFX9: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) - ; GFX9: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV4]](<3 x s16>), 0 - ; GFX9: $vgpr0_vgpr1 = COPY [[INSERT2]](<4 x s16>) - %0:_(<4 x s16>) = COPY $vgpr0_vgpr1 - %1:_(<4 x s16>) = COPY $vgpr2_vgpr3 - %2:_(<3 x s16>) = G_EXTRACT %0, 0 - %3:_(<3 x s16>) = G_EXTRACT %1, 0 - %4:_(<3 x s16>) = G_FMAXNUM %2, %3 - %5:_(<4 x s16>) = G_IMPLICIT_DEF - %6:_(<4 x s16>) = G_INSERT %5, %4, 0 - $vgpr0_vgpr1 = COPY %6 + ; GFX9: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX9: [[UV12:%[0-9]+]]:_(<3 x s16>), [[UV13:%[0-9]+]]:_(<3 x s16>), [[UV14:%[0-9]+]]:_(<3 x s16>), [[UV15:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; GFX9: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV8]](<3 x s16>), 0 + ; GFX9: [[UV16:%[0-9]+]]:_(<2 x s16>), [[UV17:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) + ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV16]](<2 x s16>) + ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 + ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) + ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV17]](<2 x s16>) + ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) + ; GFX9: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV12]](<3 x s16>), 0 + ; GFX9: [[UV18:%[0-9]+]]:_(<2 x s16>), [[UV19:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT3]](<4 x s16>) + ; GFX9: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV18]](<2 x s16>) + ; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) + ; GFX9: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV19]](<2 x s16>) + ; GFX9: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) + ; GFX9: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) + ; GFX9: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) + ; GFX9: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY2]](s32), [[COPY3]](s32) + ; GFX9: [[COPY4:%[0-9]+]]:_(s32) = COPY [[BITCAST1]](s32) + ; GFX9: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST2]](s32) + ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY4]](s32), [[COPY5]](s32) + ; GFX9: [[COPY6:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) + ; GFX9: [[COPY7:%[0-9]+]]:_(s32) = COPY [[BITCAST3]](s32) + ; GFX9: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY6]](s32), [[COPY7]](s32) + ; GFX9: [[CONCAT_VECTORS3:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[BUILD_VECTOR_TRUNC2]](<2 x s16>) + ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS3]](<6 x s16>) + %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + %1:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + %2:_(<3 x s16>), %3:_(<3 x s16>) = G_UNMERGE_VALUES %0 + %4:_(<3 x s16>), %5:_(<3 x s16>) = G_UNMERGE_VALUES %1 + %6:_(<3 x s16>) = G_FMAXNUM %2, %4 + %7:_(<3 x s16>) = G_IMPLICIT_DEF + %8:_(<6 x s16>) = G_CONCAT_VECTORS %6, %7 + $vgpr0_vgpr1_vgpr2 = COPY %8 + ... --- Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fminnum.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fminnum.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fminnum.mir @@ -388,31 +388,30 @@ name: test_fminnum_v3s16 body: | bb.0: - liveins: $vgpr0_vgpr1, $vgpr2_vgpr3 - + liveins: $vgpr0_vgpr1_vgpr2, $vgpr3_vgpr4_vgpr5 ; SI-LABEL: name: test_fminnum_v3s16 - ; SI: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 - ; SI: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3 - ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0 - ; SI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY1]](<4 x s16>), 0 + ; SI: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + ; SI: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + ; SI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) + ; SI: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>) ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; SI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; SI: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; SI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0 - ; SI: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; SI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) + ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0 + ; SI: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; SI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>) ; SI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; SI: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) - ; SI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; SI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>) ; SI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; SI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; SI: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) @@ -438,32 +437,69 @@ ; SI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; SI: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) - ; SI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; SI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT2]](<3 x s16>), 0 - ; SI: $vgpr0_vgpr1 = COPY [[INSERT2]](<4 x s16>) + ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 + ; SI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; SI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; SI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; SI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; SI: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) + ; SI: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV12]](<2 x s16>) + ; SI: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32) + ; SI: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV13]](<2 x s16>) + ; SI: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST7]], [[C]](s32) + ; SI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV8]](<3 x s16>), 0 + ; SI: [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT3]](<4 x s16>) + ; SI: [[BITCAST8:%[0-9]+]]:_(s32) = G_BITCAST [[UV14]](<2 x s16>) + ; SI: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST8]], [[C]](s32) + ; SI: [[BITCAST9:%[0-9]+]]:_(s32) = G_BITCAST [[UV15]](<2 x s16>) + ; SI: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST9]], [[C]](s32) + ; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535 + ; SI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST6]](s32) + ; SI: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C2]] + ; SI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) + ; SI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY3]], [[C2]] + ; SI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32) + ; SI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL2]] + ; SI: [[BITCAST10:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32) + ; SI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[BITCAST7]](s32) + ; SI: [[AND2:%[0-9]+]]:_(s32) = G_AND [[COPY4]], [[C2]] + ; SI: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST8]](s32) + ; SI: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY5]], [[C2]] + ; SI: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32) + ; SI: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL3]] + ; SI: [[BITCAST11:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32) + ; SI: [[COPY6:%[0-9]+]]:_(s32) = COPY [[LSHR6]](s32) + ; SI: [[AND4:%[0-9]+]]:_(s32) = G_AND [[COPY6]], [[C2]] + ; SI: [[COPY7:%[0-9]+]]:_(s32) = COPY [[BITCAST9]](s32) + ; SI: [[AND5:%[0-9]+]]:_(s32) = G_AND [[COPY7]], [[C2]] + ; SI: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32) + ; SI: [[OR4:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL4]] + ; SI: [[BITCAST12:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32) + ; SI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST10]](<2 x s16>), [[BITCAST11]](<2 x s16>), [[BITCAST12]](<2 x s16>) + ; SI: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) ; VI-LABEL: name: test_fminnum_v3s16 - ; VI: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 - ; VI: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3 - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0 - ; VI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY1]](<4 x s16>), 0 + ; VI: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + ; VI: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) + ; VI: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>) ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; VI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; VI: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; VI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0 - ; VI: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; VI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) + ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0 + ; VI: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; VI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>) ; VI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; VI: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) - ; VI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; VI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>) ; VI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; VI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; VI: [[FCANONICALIZE:%[0-9]+]]:_(s16) = G_FCANONICALIZE [[TRUNC]] @@ -486,38 +522,101 @@ ; VI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; VI: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) - ; VI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; VI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT2]](<3 x s16>), 0 - ; VI: $vgpr0_vgpr1 = COPY [[INSERT2]](<4 x s16>) + ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 + ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; VI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; VI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; VI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; VI: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) + ; VI: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV12]](<2 x s16>) + ; VI: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32) + ; VI: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV13]](<2 x s16>) + ; VI: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST7]], [[C]](s32) + ; VI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV8]](<3 x s16>), 0 + ; VI: [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT3]](<4 x s16>) + ; VI: [[BITCAST8:%[0-9]+]]:_(s32) = G_BITCAST [[UV14]](<2 x s16>) + ; VI: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST8]], [[C]](s32) + ; VI: [[BITCAST9:%[0-9]+]]:_(s32) = G_BITCAST [[UV15]](<2 x s16>) + ; VI: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST9]], [[C]](s32) + ; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535 + ; VI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST6]](s32) + ; VI: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C2]] + ; VI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) + ; VI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY3]], [[C2]] + ; VI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32) + ; VI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL2]] + ; VI: [[BITCAST10:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32) + ; VI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[BITCAST7]](s32) + ; VI: [[AND2:%[0-9]+]]:_(s32) = G_AND [[COPY4]], [[C2]] + ; VI: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST8]](s32) + ; VI: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY5]], [[C2]] + ; VI: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32) + ; VI: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL3]] + ; VI: [[BITCAST11:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32) + ; VI: [[COPY6:%[0-9]+]]:_(s32) = COPY [[LSHR6]](s32) + ; VI: [[AND4:%[0-9]+]]:_(s32) = G_AND [[COPY6]], [[C2]] + ; VI: [[COPY7:%[0-9]+]]:_(s32) = COPY [[BITCAST9]](s32) + ; VI: [[AND5:%[0-9]+]]:_(s32) = G_AND [[COPY7]], [[C2]] + ; VI: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32) + ; VI: [[OR4:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL4]] + ; VI: [[BITCAST12:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32) + ; VI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST10]](<2 x s16>), [[BITCAST11]](<2 x s16>), [[BITCAST12]](<2 x s16>) + ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) ; GFX9-LABEL: name: test_fminnum_v3s16 - ; GFX9: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 - ; GFX9: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3 - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0 - ; GFX9: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY1]](<4 x s16>), 0 + ; GFX9: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + ; GFX9: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) + ; GFX9: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>) ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0 - ; GFX9: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; GFX9: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; GFX9: [[FCANONICALIZE:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[UV]] - ; GFX9: [[FCANONICALIZE1:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[UV2]] + ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0 + ; GFX9: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; GFX9: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; GFX9: [[FCANONICALIZE:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[UV4]] + ; GFX9: [[FCANONICALIZE1:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[UV6]] ; GFX9: [[FMINNUM_IEEE:%[0-9]+]]:_(<2 x s16>) = G_FMINNUM_IEEE [[FCANONICALIZE]], [[FCANONICALIZE1]] - ; GFX9: [[FCANONICALIZE2:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[UV1]] - ; GFX9: [[FCANONICALIZE3:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[UV3]] + ; GFX9: [[FCANONICALIZE2:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[UV5]] + ; GFX9: [[FCANONICALIZE3:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[UV7]] ; GFX9: [[FMINNUM_IEEE1:%[0-9]+]]:_(<2 x s16>) = G_FMINNUM_IEEE [[FCANONICALIZE2]], [[FCANONICALIZE3]] ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[FMINNUM_IEEE]](<2 x s16>), [[FMINNUM_IEEE1]](<2 x s16>) ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[CONCAT_VECTORS]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) - ; GFX9: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) - ; GFX9: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV4]](<3 x s16>), 0 - ; GFX9: $vgpr0_vgpr1 = COPY [[INSERT2]](<4 x s16>) - %0:_(<4 x s16>) = COPY $vgpr0_vgpr1 - %1:_(<4 x s16>) = COPY $vgpr2_vgpr3 - %2:_(<3 x s16>) = G_EXTRACT %0, 0 - %3:_(<3 x s16>) = G_EXTRACT %1, 0 - %4:_(<3 x s16>) = G_FMINNUM %2, %3 - %5:_(<4 x s16>) = G_IMPLICIT_DEF - %6:_(<4 x s16>) = G_INSERT %5, %4, 0 - $vgpr0_vgpr1 = COPY %6 + ; GFX9: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX9: [[UV12:%[0-9]+]]:_(<3 x s16>), [[UV13:%[0-9]+]]:_(<3 x s16>), [[UV14:%[0-9]+]]:_(<3 x s16>), [[UV15:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; GFX9: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV8]](<3 x s16>), 0 + ; GFX9: [[UV16:%[0-9]+]]:_(<2 x s16>), [[UV17:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) + ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV16]](<2 x s16>) + ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 + ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) + ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV17]](<2 x s16>) + ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) + ; GFX9: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV12]](<3 x s16>), 0 + ; GFX9: [[UV18:%[0-9]+]]:_(<2 x s16>), [[UV19:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT3]](<4 x s16>) + ; GFX9: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV18]](<2 x s16>) + ; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) + ; GFX9: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV19]](<2 x s16>) + ; GFX9: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) + ; GFX9: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) + ; GFX9: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) + ; GFX9: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY2]](s32), [[COPY3]](s32) + ; GFX9: [[COPY4:%[0-9]+]]:_(s32) = COPY [[BITCAST1]](s32) + ; GFX9: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST2]](s32) + ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY4]](s32), [[COPY5]](s32) + ; GFX9: [[COPY6:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) + ; GFX9: [[COPY7:%[0-9]+]]:_(s32) = COPY [[BITCAST3]](s32) + ; GFX9: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY6]](s32), [[COPY7]](s32) + ; GFX9: [[CONCAT_VECTORS3:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[BUILD_VECTOR_TRUNC2]](<2 x s16>) + ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS3]](<6 x s16>) + %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + %1:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + %2:_(<3 x s16>), %3:_(<3 x s16>) = G_UNMERGE_VALUES %0 + %4:_(<3 x s16>), %5:_(<3 x s16>) = G_UNMERGE_VALUES %1 + %6:_(<3 x s16>) = G_FMINNUM %2, %4 + %7:_(<3 x s16>) = G_IMPLICIT_DEF + %8:_(<6 x s16>) = G_CONCAT_VECTORS %6, %7 + $vgpr0_vgpr1_vgpr2 = COPY %8 + ... --- Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fmul.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fmul.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fmul.mir @@ -325,145 +325,30 @@ name: test_fmul_v3s16 body: | bb.0: - liveins: $vgpr0, $vgpr1 - ; SI-LABEL: name: test_fmul_v3s16 - ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) - ; SI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) - ; SI: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) - ; SI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) - ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 - ; SI: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) - ; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) - ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 - ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) - ; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) - ; SI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) - ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 - ; SI: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; SI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) - ; SI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) - ; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) - ; SI: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) - ; SI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) - ; SI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) - ; SI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) - ; SI: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) - ; SI: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC3]](s16) - ; SI: [[FMUL:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT]], [[FPEXT1]] - ; SI: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL]](s32) - ; SI: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16) - ; SI: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC4]](s16) - ; SI: [[FMUL1:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT2]], [[FPEXT3]] - ; SI: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL1]](s32) - ; SI: [[FPEXT4:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC2]](s16) - ; SI: [[FPEXT5:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC5]](s16) - ; SI: [[FMUL2:%[0-9]+]]:_(s32) = G_FMUL [[FPEXT4]], [[FPEXT5]] - ; SI: [[FPTRUNC2:%[0-9]+]]:_(s16) = G_FPTRUNC [[FMUL2]](s32) - ; SI: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[FPTRUNC]](s16) - ; SI: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[FPTRUNC1]](s16) - ; SI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32) - ; SI: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]] - ; SI: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) - ; SI: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[FPTRUNC2]](s16) - ; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 - ; SI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32) - ; SI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] - ; SI: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; SI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) - ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS2]](<4 x s16>), 0 - ; SI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) - ; VI-LABEL: name: test_fmul_v3s16 - ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) - ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) - ; VI: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) - ; VI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 - ; VI: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) - ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) - ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 - ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) - ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) - ; VI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) - ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 - ; VI: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; VI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) - ; VI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) - ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) - ; VI: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) - ; VI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) - ; VI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) - ; VI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) - ; VI: [[FMUL:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC]], [[TRUNC3]] - ; VI: [[FMUL1:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC1]], [[TRUNC4]] - ; VI: [[FMUL2:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC2]], [[TRUNC5]] - ; VI: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[FMUL]](s16) - ; VI: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[FMUL1]](s16) - ; VI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32) - ; VI: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]] - ; VI: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) - ; VI: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[FMUL2]](s16) - ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 - ; VI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32) - ; VI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] - ; VI: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; VI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS2]](<4 x s16>), 0 - ; VI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) - ; GFX9-LABEL: name: test_fmul_v3s16 - ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) - ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) - ; GFX9: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) - ; GFX9: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 - ; GFX9: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) - ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 - ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) - ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) - ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; GFX9: [[COPY:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) - ; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) - ; GFX9: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY]](s32), [[COPY1]](s32) - ; GFX9: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST1]](s32) - ; GFX9: [[DEF3:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF - ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY2]](s32), [[DEF3]](s32) - ; GFX9: [[DEF4:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 - ; GFX9: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; GFX9: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) - ; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) - ; GFX9: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) - ; GFX9: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) - ; GFX9: [[COPY3:%[0-9]+]]:_(s32) = COPY [[BITCAST2]](s32) - ; GFX9: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) - ; GFX9: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY3]](s32), [[COPY4]](s32) - ; GFX9: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST3]](s32) - ; GFX9: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY5]](s32), [[DEF3]](s32) - ; GFX9: [[FMUL:%[0-9]+]]:_(<2 x s16>) = G_FMUL [[BUILD_VECTOR_TRUNC]], [[BUILD_VECTOR_TRUNC2]] - ; GFX9: [[FMUL1:%[0-9]+]]:_(<2 x s16>) = G_FMUL [[BUILD_VECTOR_TRUNC1]], [[BUILD_VECTOR_TRUNC3]] - ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[FMUL]](<2 x s16>), [[FMUL1]](<2 x s16>), [[DEF4]](<2 x s16>) - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS2]](<6 x s16>), 0 - ; GFX9: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) - %0:_(<3 x s16>) = G_IMPLICIT_DEF - %1:_(<3 x s16>) = G_IMPLICIT_DEF - %2:_(<3 x s16>) = G_FMUL %0, %1 - S_NOP 0, implicit %2 + liveins: $vgpr0_vgpr1_vgpr2, $vgpr3_vgpr4_vgpr5 + ; CHECK-LABEL: name: test_or_v3s16 + ; CHECK: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + ; CHECK: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + ; CHECK: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) + ; CHECK: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>) + ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; CHECK: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0 + ; CHECK: [[OR:%[0-9]+]]:_(<4 x s16>) = G_OR [[INSERT]], [[INSERT1]] + ; CHECK: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[OR]](<4 x s16>), 0 + ; CHECK: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[EXTRACT]](<3 x s16>), [[EXTRACT1]](<3 x s16>) + ; CHECK: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>) + %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + %1:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + %2:_(<3 x s16>), %3:_(<3 x s16>) = G_UNMERGE_VALUES %0 + %4:_(<3 x s16>), %5:_(<3 x s16>) = G_UNMERGE_VALUES %1 + %6:_(<3 x s16>) = G_FMUL %2, %4 + %7:_(<3 x s16>) = G_IMPLICIT_DEF + %8:_(<6 x s16>) = G_CONCAT_VECTORS %6, %7 + $vgpr0_vgpr1_vgpr2 = COPY %8 + ... --- Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fpext.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fpext.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fpext.mir @@ -66,21 +66,19 @@ name: test_fpext_v3f16_to_v3f32 body: | bb.0: - liveins: $vgpr0 - + liveins: $vgpr0_vgpr1_vgpr2 ; CHECK-LABEL: name: test_fpext_v3f16_to_v3f32 + ; CHECK: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + ; CHECK: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) - ; CHECK: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) - ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 - ; CHECK: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; CHECK: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) + ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; CHECK: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; CHECK: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) ; CHECK: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CHECK: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; CHECK: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; CHECK: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) + ; CHECK: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) ; CHECK: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; CHECK: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; CHECK: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) @@ -88,9 +86,10 @@ ; CHECK: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC2]](s16) ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[FPEXT]](s32), [[FPEXT1]](s32), [[FPEXT2]](s32) ; CHECK: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>) - %0:_(<3 x s16>) = G_IMPLICIT_DEF - %1:_(<3 x s32>) = G_FPEXT %0 - $vgpr0_vgpr1_vgpr2 = COPY %1 + %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + %1:_(<3 x s16>), %2:_(<3 x s16>) = G_UNMERGE_VALUES %0 + %3:_(<3 x s32>) = G_FPEXT %1 + $vgpr0_vgpr1_vgpr2 = COPY %3 ... --- Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fsub.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fsub.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fsub.mir @@ -369,156 +369,30 @@ --- name: test_fsub_v3s16 body: | - bb.0.entry: - liveins: $vgpr0, $vgpr1 + bb.0: + liveins: $vgpr0_vgpr1_vgpr2, $vgpr3_vgpr4_vgpr5 + ; CHECK-LABEL: name: test_or_v3s16 + ; CHECK: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + ; CHECK: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + ; CHECK: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) + ; CHECK: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>) + ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; CHECK: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0 + ; CHECK: [[OR:%[0-9]+]]:_(<4 x s16>) = G_OR [[INSERT]], [[INSERT1]] + ; CHECK: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[OR]](<4 x s16>), 0 + ; CHECK: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 0 + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[EXTRACT]](<3 x s16>), [[EXTRACT1]](<3 x s16>) + ; CHECK: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>) + %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + %1:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + %2:_(<3 x s16>), %3:_(<3 x s16>) = G_UNMERGE_VALUES %0 + %4:_(<3 x s16>), %5:_(<3 x s16>) = G_UNMERGE_VALUES %1 + %6:_(<3 x s16>) = G_FSUB %2, %4 + %7:_(<3 x s16>) = G_IMPLICIT_DEF + %8:_(<6 x s16>) = G_CONCAT_VECTORS %6, %7 + $vgpr0_vgpr1_vgpr2 = COPY %8 - ; SI-LABEL: name: test_fsub_v3s16 - ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) - ; SI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) - ; SI: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) - ; SI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) - ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 - ; SI: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) - ; SI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) - ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 - ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) - ; SI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) - ; SI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) - ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 - ; SI: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; SI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) - ; SI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) - ; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) - ; SI: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) - ; SI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) - ; SI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) - ; SI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) - ; SI: [[FNEG:%[0-9]+]]:_(s16) = G_FNEG [[TRUNC3]] - ; SI: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) - ; SI: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[FNEG]](s16) - ; SI: [[FADD:%[0-9]+]]:_(s32) = G_FADD [[FPEXT]], [[FPEXT1]] - ; SI: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD]](s32) - ; SI: [[FNEG1:%[0-9]+]]:_(s16) = G_FNEG [[TRUNC4]] - ; SI: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC1]](s16) - ; SI: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[FNEG1]](s16) - ; SI: [[FADD1:%[0-9]+]]:_(s32) = G_FADD [[FPEXT2]], [[FPEXT3]] - ; SI: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD1]](s32) - ; SI: [[FNEG2:%[0-9]+]]:_(s16) = G_FNEG [[TRUNC5]] - ; SI: [[FPEXT4:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC2]](s16) - ; SI: [[FPEXT5:%[0-9]+]]:_(s32) = G_FPEXT [[FNEG2]](s16) - ; SI: [[FADD2:%[0-9]+]]:_(s32) = G_FADD [[FPEXT4]], [[FPEXT5]] - ; SI: [[FPTRUNC2:%[0-9]+]]:_(s16) = G_FPTRUNC [[FADD2]](s32) - ; SI: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[FPTRUNC]](s16) - ; SI: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[FPTRUNC1]](s16) - ; SI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32) - ; SI: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]] - ; SI: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) - ; SI: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[FPTRUNC2]](s16) - ; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 - ; SI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32) - ; SI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] - ; SI: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; SI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) - ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS2]](<4 x s16>), 0 - ; SI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) - ; VI-LABEL: name: test_fsub_v3s16 - ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) - ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) - ; VI: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) - ; VI: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 - ; VI: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) - ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) - ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 - ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) - ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) - ; VI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) - ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 - ; VI: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; VI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) - ; VI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) - ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) - ; VI: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) - ; VI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) - ; VI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) - ; VI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) - ; VI: [[FNEG:%[0-9]+]]:_(s16) = G_FNEG [[TRUNC3]] - ; VI: [[FADD:%[0-9]+]]:_(s16) = G_FADD [[TRUNC]], [[FNEG]] - ; VI: [[FNEG1:%[0-9]+]]:_(s16) = G_FNEG [[TRUNC4]] - ; VI: [[FADD1:%[0-9]+]]:_(s16) = G_FADD [[TRUNC1]], [[FNEG1]] - ; VI: [[FNEG2:%[0-9]+]]:_(s16) = G_FNEG [[TRUNC5]] - ; VI: [[FADD2:%[0-9]+]]:_(s16) = G_FADD [[TRUNC2]], [[FNEG2]] - ; VI: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[FADD]](s16) - ; VI: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[FADD1]](s16) - ; VI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32) - ; VI: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]] - ; VI: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) - ; VI: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[FADD2]](s16) - ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 - ; VI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32) - ; VI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] - ; VI: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; VI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS2]](<4 x s16>), 0 - ; VI: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) - ; GFX9-LABEL: name: test_fsub_v3s16 - ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) - ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) - ; GFX9: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) - ; GFX9: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 - ; GFX9: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) - ; GFX9: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) - ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 - ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) - ; GFX9: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) - ; GFX9: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) - ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 - ; GFX9: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; GFX9: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) - ; GFX9: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) - ; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) - ; GFX9: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) - ; GFX9: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) - ; GFX9: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) - ; GFX9: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) - ; GFX9: [[FNEG:%[0-9]+]]:_(s16) = G_FNEG [[TRUNC3]] - ; GFX9: [[FADD:%[0-9]+]]:_(s16) = G_FADD [[TRUNC]], [[FNEG]] - ; GFX9: [[FNEG1:%[0-9]+]]:_(s16) = G_FNEG [[TRUNC4]] - ; GFX9: [[FADD1:%[0-9]+]]:_(s16) = G_FADD [[TRUNC1]], [[FNEG1]] - ; GFX9: [[FNEG2:%[0-9]+]]:_(s16) = G_FNEG [[TRUNC5]] - ; GFX9: [[FADD2:%[0-9]+]]:_(s16) = G_FADD [[TRUNC2]], [[FNEG2]] - ; GFX9: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[FADD]](s16) - ; GFX9: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[FADD1]](s16) - ; GFX9: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT]](s32), [[ANYEXT1]](s32) - ; GFX9: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[FADD2]](s16) - ; GFX9: [[DEF3:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF - ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT2]](s32), [[DEF3]](s32) - ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>) - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS2]](<4 x s16>), 0 - ; GFX9: S_NOP 0, implicit [[EXTRACT]](<3 x s16>) - %0:_(<3 x s16>) = G_IMPLICIT_DEF - %1:_(<3 x s16>) = G_IMPLICIT_DEF - %2:_(<3 x s16>) = G_FSUB %0, %1 - S_NOP 0, implicit %2 ... --- Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-intrinsic-round.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-intrinsic-round.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-intrinsic-round.mir @@ -616,20 +616,19 @@ name: test_intrinsic_round_v3s16 body: | bb.0: - liveins: $vgpr0_vgpr1 - + liveins: $vgpr0_vgpr1_vgpr2 ; GFX6-LABEL: name: test_intrinsic_round_v3s16 - ; GFX6: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 - ; GFX6: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0 + ; GFX6: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + ; GFX6: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) ; GFX6: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX6: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; GFX6: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; GFX6: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; GFX6: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; GFX6: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; GFX6: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) ; GFX6: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX6: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX6: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX6: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; GFX6: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; GFX6: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) ; GFX6: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX6: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX6: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16) @@ -704,21 +703,58 @@ ; GFX6: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; GFX6: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32) ; GFX6: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>) - ; GFX6: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; GFX6: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0 - ; GFX6: $vgpr0_vgpr1 = COPY [[INSERT1]](<4 x s16>) + ; GFX6: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 + ; GFX6: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX6: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX6: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX6: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; GFX6: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; GFX6: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) + ; GFX6: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32) + ; GFX6: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) + ; GFX6: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32) + ; GFX6: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV4]](<3 x s16>), 0 + ; GFX6: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) + ; GFX6: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) + ; GFX6: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32) + ; GFX6: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) + ; GFX6: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST7]], [[C]](s32) + ; GFX6: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535 + ; GFX6: [[COPY1:%[0-9]+]]:_(s32) = COPY [[BITCAST4]](s32) + ; GFX6: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C6]] + ; GFX6: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) + ; GFX6: [[AND4:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C6]] + ; GFX6: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND4]], [[C]](s32) + ; GFX6: [[OR5:%[0-9]+]]:_(s32) = G_OR [[AND3]], [[SHL2]] + ; GFX6: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR5]](s32) + ; GFX6: [[COPY3:%[0-9]+]]:_(s32) = COPY [[BITCAST5]](s32) + ; GFX6: [[AND5:%[0-9]+]]:_(s32) = G_AND [[COPY3]], [[C6]] + ; GFX6: [[COPY4:%[0-9]+]]:_(s32) = COPY [[BITCAST6]](s32) + ; GFX6: [[AND6:%[0-9]+]]:_(s32) = G_AND [[COPY4]], [[C6]] + ; GFX6: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND6]], [[C]](s32) + ; GFX6: [[OR6:%[0-9]+]]:_(s32) = G_OR [[AND5]], [[SHL3]] + ; GFX6: [[BITCAST9:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR6]](s32) + ; GFX6: [[COPY5:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) + ; GFX6: [[AND7:%[0-9]+]]:_(s32) = G_AND [[COPY5]], [[C6]] + ; GFX6: [[COPY6:%[0-9]+]]:_(s32) = COPY [[BITCAST7]](s32) + ; GFX6: [[AND8:%[0-9]+]]:_(s32) = G_AND [[COPY6]], [[C6]] + ; GFX6: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND8]], [[C]](s32) + ; GFX6: [[OR7:%[0-9]+]]:_(s32) = G_OR [[AND7]], [[SHL4]] + ; GFX6: [[BITCAST10:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR7]](s32) + ; GFX6: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST8]](<2 x s16>), [[BITCAST9]](<2 x s16>), [[BITCAST10]](<2 x s16>) + ; GFX6: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) ; GFX8-LABEL: name: test_intrinsic_round_v3s16 - ; GFX8: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 - ; GFX8: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0 + ; GFX8: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + ; GFX8: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) ; GFX8: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX8: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; GFX8: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; GFX8: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; GFX8: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; GFX8: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; GFX8: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) ; GFX8: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX8: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX8: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX8: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; GFX8: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; GFX8: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) ; GFX8: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX8: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX8: [[INTRINSIC_TRUNC:%[0-9]+]]:_(s16) = G_INTRINSIC_TRUNC [[TRUNC]] @@ -763,21 +799,58 @@ ; GFX8: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; GFX8: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32) ; GFX8: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>) - ; GFX8: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; GFX8: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0 - ; GFX8: $vgpr0_vgpr1 = COPY [[INSERT1]](<4 x s16>) + ; GFX8: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 + ; GFX8: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX8: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX8: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX8: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; GFX8: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; GFX8: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) + ; GFX8: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32) + ; GFX8: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) + ; GFX8: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32) + ; GFX8: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV4]](<3 x s16>), 0 + ; GFX8: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) + ; GFX8: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) + ; GFX8: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32) + ; GFX8: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) + ; GFX8: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST7]], [[C]](s32) + ; GFX8: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535 + ; GFX8: [[COPY1:%[0-9]+]]:_(s32) = COPY [[BITCAST4]](s32) + ; GFX8: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C6]] + ; GFX8: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) + ; GFX8: [[AND4:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C6]] + ; GFX8: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND4]], [[C]](s32) + ; GFX8: [[OR5:%[0-9]+]]:_(s32) = G_OR [[AND3]], [[SHL2]] + ; GFX8: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR5]](s32) + ; GFX8: [[COPY3:%[0-9]+]]:_(s32) = COPY [[BITCAST5]](s32) + ; GFX8: [[AND5:%[0-9]+]]:_(s32) = G_AND [[COPY3]], [[C6]] + ; GFX8: [[COPY4:%[0-9]+]]:_(s32) = COPY [[BITCAST6]](s32) + ; GFX8: [[AND6:%[0-9]+]]:_(s32) = G_AND [[COPY4]], [[C6]] + ; GFX8: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND6]], [[C]](s32) + ; GFX8: [[OR6:%[0-9]+]]:_(s32) = G_OR [[AND5]], [[SHL3]] + ; GFX8: [[BITCAST9:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR6]](s32) + ; GFX8: [[COPY5:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) + ; GFX8: [[AND7:%[0-9]+]]:_(s32) = G_AND [[COPY5]], [[C6]] + ; GFX8: [[COPY6:%[0-9]+]]:_(s32) = COPY [[BITCAST7]](s32) + ; GFX8: [[AND8:%[0-9]+]]:_(s32) = G_AND [[COPY6]], [[C6]] + ; GFX8: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND8]], [[C]](s32) + ; GFX8: [[OR7:%[0-9]+]]:_(s32) = G_OR [[AND7]], [[SHL4]] + ; GFX8: [[BITCAST10:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR7]](s32) + ; GFX8: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST8]](<2 x s16>), [[BITCAST9]](<2 x s16>), [[BITCAST10]](<2 x s16>) + ; GFX8: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) ; GFX9-LABEL: name: test_intrinsic_round_v3s16 - ; GFX9: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0 + ; GFX9: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; GFX9: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; GFX9: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) ; GFX9: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX9: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) ; GFX9: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX9: [[INTRINSIC_TRUNC:%[0-9]+]]:_(s16) = G_INTRINSIC_TRUNC [[TRUNC]] @@ -818,15 +891,39 @@ ; GFX9: [[DEF1:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT2]](s32), [[DEF1]](s32) ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>) - ; GFX9: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0 - ; GFX9: $vgpr0_vgpr1 = COPY [[INSERT1]](<4 x s16>) - %0:_(<4 x s16>) = COPY $vgpr0_vgpr1 - %1:_(<3 x s16>) = G_EXTRACT %0, 0 - %2:_(<3 x s16>) = G_INTRINSIC_ROUND %1 - %3:_(<4 x s16>) = G_IMPLICIT_DEF - %4:_(<4 x s16>) = G_INSERT %3, %2, 0 - $vgpr0_vgpr1 = COPY %4 + ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 + ; GFX9: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX9: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; GFX9: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; GFX9: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) + ; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) + ; GFX9: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) + ; GFX9: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) + ; GFX9: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV4]](<3 x s16>), 0 + ; GFX9: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) + ; GFX9: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) + ; GFX9: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32) + ; GFX9: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) + ; GFX9: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32) + ; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY [[BITCAST2]](s32) + ; GFX9: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) + ; GFX9: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY1]](s32), [[COPY2]](s32) + ; GFX9: [[COPY3:%[0-9]+]]:_(s32) = COPY [[BITCAST3]](s32) + ; GFX9: [[COPY4:%[0-9]+]]:_(s32) = COPY [[BITCAST4]](s32) + ; GFX9: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY3]](s32), [[COPY4]](s32) + ; GFX9: [[COPY5:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) + ; GFX9: [[COPY6:%[0-9]+]]:_(s32) = COPY [[BITCAST5]](s32) + ; GFX9: [[BUILD_VECTOR_TRUNC4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY5]](s32), [[COPY6]](s32) + ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC2]](<2 x s16>), [[BUILD_VECTOR_TRUNC3]](<2 x s16>), [[BUILD_VECTOR_TRUNC4]](<2 x s16>) + ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) + %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + %1:_(<3 x s16>), %2:_(<3 x s16>) = G_UNMERGE_VALUES %0 + %3:_(<3 x s16>) = G_INTRINSIC_ROUND %1 + %4:_(<3 x s16>) = G_IMPLICIT_DEF + %5:_(<6 x s16>) = G_CONCAT_VECTORS %3, %4 + $vgpr0_vgpr1_vgpr2 = COPY %5 ... --- Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-lshr.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-lshr.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-lshr.mir @@ -748,6 +748,242 @@ $vgpr0_vgpr1 = COPY %6 ... +--- +name: test_ashr_v3s16_v3s16 +body: | + bb.0: + liveins: $vgpr0_vgpr1_vgpr2, $vgpr3_vgpr4_vgpr5 + ; SI-LABEL: name: test_ashr_v3s16_v3s16 + ; SI: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + ; SI: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + ; SI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) + ; SI: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>) + ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; SI: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) + ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 + ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) + ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) + ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) + ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0 + ; SI: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; SI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>) + ; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) + ; SI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>) + ; SI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) + ; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535 + ; SI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST2]](s32) + ; SI: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C1]] + ; SI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) + ; SI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY3]], [[C1]] + ; SI: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[AND1]], [[AND]](s32) + ; SI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) + ; SI: [[AND2:%[0-9]+]]:_(s32) = G_AND [[COPY4]], [[C1]] + ; SI: [[COPY5:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) + ; SI: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY5]], [[C1]] + ; SI: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[AND3]], [[AND2]](s32) + ; SI: [[COPY6:%[0-9]+]]:_(s32) = COPY [[BITCAST3]](s32) + ; SI: [[AND4:%[0-9]+]]:_(s32) = G_AND [[COPY6]], [[C1]] + ; SI: [[COPY7:%[0-9]+]]:_(s32) = COPY [[BITCAST1]](s32) + ; SI: [[AND5:%[0-9]+]]:_(s32) = G_AND [[COPY7]], [[C1]] + ; SI: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[AND5]], [[AND4]](s32) + ; SI: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) + ; SI: [[AND6:%[0-9]+]]:_(s32) = G_AND [[COPY8]], [[C1]] + ; SI: [[COPY9:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32) + ; SI: [[AND7:%[0-9]+]]:_(s32) = G_AND [[COPY9]], [[C1]] + ; SI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C]](s32) + ; SI: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL]] + ; SI: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) + ; SI: [[COPY10:%[0-9]+]]:_(s32) = COPY [[LSHR6]](s32) + ; SI: [[AND8:%[0-9]+]]:_(s32) = G_AND [[COPY10]], [[C1]] + ; SI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 + ; SI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C]](s32) + ; SI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND8]], [[SHL1]] + ; SI: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) + ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) + ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 + ; SI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; SI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; SI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; SI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; SI: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) + ; SI: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV12]](<2 x s16>) + ; SI: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32) + ; SI: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV13]](<2 x s16>) + ; SI: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST7]], [[C]](s32) + ; SI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV8]](<3 x s16>), 0 + ; SI: [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT3]](<4 x s16>) + ; SI: [[BITCAST8:%[0-9]+]]:_(s32) = G_BITCAST [[UV14]](<2 x s16>) + ; SI: [[LSHR9:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST8]], [[C]](s32) + ; SI: [[BITCAST9:%[0-9]+]]:_(s32) = G_BITCAST [[UV15]](<2 x s16>) + ; SI: [[LSHR10:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST9]], [[C]](s32) + ; SI: [[COPY11:%[0-9]+]]:_(s32) = COPY [[BITCAST6]](s32) + ; SI: [[AND9:%[0-9]+]]:_(s32) = G_AND [[COPY11]], [[C1]] + ; SI: [[COPY12:%[0-9]+]]:_(s32) = COPY [[LSHR7]](s32) + ; SI: [[AND10:%[0-9]+]]:_(s32) = G_AND [[COPY12]], [[C1]] + ; SI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND10]], [[C]](s32) + ; SI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND9]], [[SHL2]] + ; SI: [[BITCAST10:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32) + ; SI: [[COPY13:%[0-9]+]]:_(s32) = COPY [[BITCAST7]](s32) + ; SI: [[AND11:%[0-9]+]]:_(s32) = G_AND [[COPY13]], [[C1]] + ; SI: [[COPY14:%[0-9]+]]:_(s32) = COPY [[BITCAST8]](s32) + ; SI: [[AND12:%[0-9]+]]:_(s32) = G_AND [[COPY14]], [[C1]] + ; SI: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND12]], [[C]](s32) + ; SI: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND11]], [[SHL3]] + ; SI: [[BITCAST11:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32) + ; SI: [[COPY15:%[0-9]+]]:_(s32) = COPY [[LSHR9]](s32) + ; SI: [[AND13:%[0-9]+]]:_(s32) = G_AND [[COPY15]], [[C1]] + ; SI: [[COPY16:%[0-9]+]]:_(s32) = COPY [[BITCAST9]](s32) + ; SI: [[AND14:%[0-9]+]]:_(s32) = G_AND [[COPY16]], [[C1]] + ; SI: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND14]], [[C]](s32) + ; SI: [[OR4:%[0-9]+]]:_(s32) = G_OR [[AND13]], [[SHL4]] + ; SI: [[BITCAST12:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32) + ; SI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST10]](<2 x s16>), [[BITCAST11]](<2 x s16>), [[BITCAST12]](<2 x s16>) + ; SI: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) + ; VI-LABEL: name: test_ashr_v3s16_v3s16 + ; VI: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + ; VI: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) + ; VI: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>) + ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; VI: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) + ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) + ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 + ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) + ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) + ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) + ; VI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) + ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) + ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0 + ; VI: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; VI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>) + ; VI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) + ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) + ; VI: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) + ; VI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>) + ; VI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) + ; VI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) + ; VI: [[LSHR4:%[0-9]+]]:_(s16) = G_LSHR [[TRUNC]], [[TRUNC3]](s16) + ; VI: [[LSHR5:%[0-9]+]]:_(s16) = G_LSHR [[TRUNC1]], [[TRUNC4]](s16) + ; VI: [[LSHR6:%[0-9]+]]:_(s16) = G_LSHR [[TRUNC2]], [[TRUNC5]](s16) + ; VI: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[LSHR4]](s16) + ; VI: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[LSHR5]](s16) + ; VI: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32) + ; VI: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]] + ; VI: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) + ; VI: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[LSHR6]](s16) + ; VI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 + ; VI: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32) + ; VI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] + ; VI: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) + ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) + ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 + ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; VI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; VI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; VI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; VI: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) + ; VI: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV12]](<2 x s16>) + ; VI: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32) + ; VI: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV13]](<2 x s16>) + ; VI: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST7]], [[C]](s32) + ; VI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV8]](<3 x s16>), 0 + ; VI: [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT3]](<4 x s16>) + ; VI: [[BITCAST8:%[0-9]+]]:_(s32) = G_BITCAST [[UV14]](<2 x s16>) + ; VI: [[LSHR9:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST8]], [[C]](s32) + ; VI: [[BITCAST9:%[0-9]+]]:_(s32) = G_BITCAST [[UV15]](<2 x s16>) + ; VI: [[LSHR10:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST9]], [[C]](s32) + ; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535 + ; VI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST6]](s32) + ; VI: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C2]] + ; VI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR7]](s32) + ; VI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY3]], [[C2]] + ; VI: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32) + ; VI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL2]] + ; VI: [[BITCAST10:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32) + ; VI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[BITCAST7]](s32) + ; VI: [[AND2:%[0-9]+]]:_(s32) = G_AND [[COPY4]], [[C2]] + ; VI: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST8]](s32) + ; VI: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY5]], [[C2]] + ; VI: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32) + ; VI: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL3]] + ; VI: [[BITCAST11:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32) + ; VI: [[COPY6:%[0-9]+]]:_(s32) = COPY [[LSHR9]](s32) + ; VI: [[AND4:%[0-9]+]]:_(s32) = G_AND [[COPY6]], [[C2]] + ; VI: [[COPY7:%[0-9]+]]:_(s32) = COPY [[BITCAST9]](s32) + ; VI: [[AND5:%[0-9]+]]:_(s32) = G_AND [[COPY7]], [[C2]] + ; VI: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32) + ; VI: [[OR4:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL4]] + ; VI: [[BITCAST12:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32) + ; VI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST10]](<2 x s16>), [[BITCAST11]](<2 x s16>), [[BITCAST12]](<2 x s16>) + ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) + ; GFX9-LABEL: name: test_ashr_v3s16_v3s16 + ; GFX9: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + ; GFX9: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) + ; GFX9: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>) + ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; GFX9: [[EXTRACT:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 0 + ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; GFX9: [[EXTRACT1:%[0-9]+]]:_(s16) = G_EXTRACT [[INSERT1]](<4 x s16>), 32 + ; GFX9: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0 + ; GFX9: [[EXTRACT2:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT2]](<4 x s16>), 0 + ; GFX9: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0 + ; GFX9: [[EXTRACT3:%[0-9]+]]:_(s16) = G_EXTRACT [[INSERT3]](<4 x s16>), 32 + ; GFX9: [[LSHR:%[0-9]+]]:_(<2 x s16>) = G_LSHR [[EXTRACT]], [[EXTRACT2]](<2 x s16>) + ; GFX9: [[LSHR1:%[0-9]+]]:_(s16) = G_LSHR [[EXTRACT1]], [[EXTRACT3]](s16) + ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX9: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; GFX9: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV4]](<3 x s16>), 0 + ; GFX9: [[INSERT5:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT4]], [[LSHR]](<2 x s16>), 0 + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT5]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX9: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9: [[INSERT6:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV8]](<3 x s16>), 0 + ; GFX9: [[INSERT7:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT6]], [[LSHR1]](s16), 32 + ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT7]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX9: [[UV12:%[0-9]+]]:_(<3 x s16>), [[UV13:%[0-9]+]]:_(<3 x s16>), [[UV14:%[0-9]+]]:_(<3 x s16>), [[UV15:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; GFX9: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX9: [[CONCAT_VECTORS3:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX9: [[UV16:%[0-9]+]]:_(<3 x s16>), [[UV17:%[0-9]+]]:_(<3 x s16>), [[UV18:%[0-9]+]]:_(<3 x s16>), [[UV19:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS3]](<12 x s16>) + ; GFX9: [[INSERT8:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV12]](<3 x s16>), 0 + ; GFX9: [[UV20:%[0-9]+]]:_(<2 x s16>), [[UV21:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT8]](<4 x s16>) + ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV20]](<2 x s16>) + ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 + ; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) + ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV21]](<2 x s16>) + ; GFX9: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) + ; GFX9: [[INSERT9:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV16]](<3 x s16>), 0 + ; GFX9: [[UV22:%[0-9]+]]:_(<2 x s16>), [[UV23:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT9]](<4 x s16>) + ; GFX9: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV22]](<2 x s16>) + ; GFX9: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) + ; GFX9: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV23]](<2 x s16>) + ; GFX9: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) + ; GFX9: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) + ; GFX9: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) + ; GFX9: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY2]](s32), [[COPY3]](s32) + ; GFX9: [[COPY4:%[0-9]+]]:_(s32) = COPY [[BITCAST1]](s32) + ; GFX9: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST2]](s32) + ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY4]](s32), [[COPY5]](s32) + ; GFX9: [[COPY6:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) + ; GFX9: [[COPY7:%[0-9]+]]:_(s32) = COPY [[BITCAST3]](s32) + ; GFX9: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY6]](s32), [[COPY7]](s32) + ; GFX9: [[CONCAT_VECTORS4:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[BUILD_VECTOR_TRUNC2]](<2 x s16>) + ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS4]](<6 x s16>) + %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + %1:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + %2:_(<3 x s16>), %3:_(<3 x s16>) = G_UNMERGE_VALUES %0 + %4:_(<3 x s16>), %5:_(<3 x s16>) = G_UNMERGE_VALUES %1 + %6:_(<3 x s16>) = G_LSHR %2, %4 + %7:_(<3 x s16>) = G_IMPLICIT_DEF + %8:_(<6 x s16>) = G_CONCAT_VECTORS %6, %7 + $vgpr0_vgpr1_vgpr2 = COPY %8 +... + --- name: test_lshr_v4s16_v4s16 body: | Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-or.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-or.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-or.mir @@ -435,28 +435,67 @@ name: test_or_v3s16 body: | bb.0: - + liveins: $vgpr0_vgpr1_vgpr2, $vgpr3_vgpr4_vgpr5 ; CHECK-LABEL: name: test_or_v3s16 + ; CHECK: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + ; CHECK: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + ; CHECK: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) + ; CHECK: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>) ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) - ; CHECK: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) - ; CHECK: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) - ; CHECK: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) - ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 - ; CHECK: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 + ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; CHECK: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0 ; CHECK: [[OR:%[0-9]+]]:_(<4 x s16>) = G_OR [[INSERT]], [[INSERT1]] - ; CHECK: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[OR]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) - ; CHECK: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) - ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 - ; CHECK: $vgpr0_vgpr1 = COPY [[INSERT2]](<4 x s16>) - %0:_(<3 x s16>) = G_IMPLICIT_DEF - %1:_(<3 x s16>) = G_IMPLICIT_DEF - %2:_(<3 x s16>) = G_OR %0, %1 - %4:_(<4 x s16>) = G_IMPLICIT_DEF - %5:_(<4 x s16>) = G_INSERT %4, %2, 0 - $vgpr0_vgpr1 = COPY %5 + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[OR]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; CHECK: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; CHECK: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; CHECK: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; CHECK: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV4]](<3 x s16>), 0 + ; CHECK: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) + ; CHECK: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV12]](<2 x s16>) + ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 + ; CHECK: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) + ; CHECK: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV13]](<2 x s16>) + ; CHECK: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) + ; CHECK: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV8]](<3 x s16>), 0 + ; CHECK: [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT3]](<4 x s16>) + ; CHECK: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV14]](<2 x s16>) + ; CHECK: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) + ; CHECK: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV15]](<2 x s16>) + ; CHECK: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) + ; CHECK: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535 + ; CHECK: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) + ; CHECK: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C1]] + ; CHECK: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) + ; CHECK: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY3]], [[C1]] + ; CHECK: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32) + ; CHECK: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]] + ; CHECK: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) + ; CHECK: [[COPY4:%[0-9]+]]:_(s32) = COPY [[BITCAST1]](s32) + ; CHECK: [[AND2:%[0-9]+]]:_(s32) = G_AND [[COPY4]], [[C1]] + ; CHECK: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST2]](s32) + ; CHECK: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY5]], [[C1]] + ; CHECK: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32) + ; CHECK: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]] + ; CHECK: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32) + ; CHECK: [[COPY6:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) + ; CHECK: [[AND4:%[0-9]+]]:_(s32) = G_AND [[COPY6]], [[C1]] + ; CHECK: [[COPY7:%[0-9]+]]:_(s32) = COPY [[BITCAST3]](s32) + ; CHECK: [[AND5:%[0-9]+]]:_(s32) = G_AND [[COPY7]], [[C1]] + ; CHECK: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32) + ; CHECK: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]] + ; CHECK: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32) + ; CHECK: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>) + ; CHECK: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) + %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + %1:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + %2:_(<3 x s16>), %3:_(<3 x s16>) = G_UNMERGE_VALUES %0 + %4:_(<3 x s16>), %5:_(<3 x s16>) = G_UNMERGE_VALUES %1 + %6:_(<3 x s16>) = G_OR %2, %4 + %7:_(<3 x s16>) = G_IMPLICIT_DEF + %8:_(<6 x s16>) = G_CONCAT_VECTORS %6, %7 + $vgpr0_vgpr1_vgpr2 = COPY %8 + ... --- Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-saddo.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-saddo.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-saddo.mir @@ -209,26 +209,25 @@ name: test_saddo_v3s16 body: | bb.0: - liveins: $vgpr0_vgpr1, $vgpr2_vgpr3 - + liveins: $vgpr0_vgpr1_vgpr2, $vgpr3_vgpr4_vgpr5 ; CHECK-LABEL: name: test_saddo_v3s16 - ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 - ; CHECK: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr1_vgpr2 - ; CHECK: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0 - ; CHECK: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0 + ; CHECK: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + ; CHECK: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + ; CHECK: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) + ; CHECK: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>) ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; CHECK: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; CHECK: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; CHECK: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; CHECK: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CHECK: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) - ; CHECK: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; CHECK: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; CHECK: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; CHECK: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0 - ; CHECK: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; CHECK: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) + ; CHECK: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0 + ; CHECK: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; CHECK: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>) ; CHECK: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) - ; CHECK: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; CHECK: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>) ; CHECK: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; CHECK: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) ; CHECK: [[COPY3:%[0-9]+]]:_(s32) = COPY [[BITCAST2]](s32) @@ -254,38 +253,110 @@ ; CHECK: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]] ; CHECK: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) - ; CHECK: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 + ; CHECK: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 ; CHECK: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 0 - ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0 - ; CHECK: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) - ; CHECK: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) + ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; CHECK: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) + ; CHECK: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) ; CHECK: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32) - ; CHECK: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) + ; CHECK: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) ; CHECK: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST7]], [[C]](s32) - ; CHECK: [[COPY11:%[0-9]+]]:_(s32) = COPY [[BITCAST6]](s32) - ; CHECK: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY11]], [[C1]] - ; CHECK: [[COPY12:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) - ; CHECK: [[AND4:%[0-9]+]]:_(s32) = G_AND [[COPY12]], [[C1]] - ; CHECK: [[COPY13:%[0-9]+]]:_(s32) = COPY [[BITCAST7]](s32) - ; CHECK: [[AND5:%[0-9]+]]:_(s32) = G_AND [[COPY13]], [[C1]] - ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[AND3]](s32), [[AND4]](s32), [[AND5]](s32) - ; CHECK: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT2]](<3 x s16>), 0 - ; CHECK: [[DEF1:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF - ; CHECK: [[INSERT4:%[0-9]+]]:_(<4 x s32>) = G_INSERT [[DEF1]], [[BUILD_VECTOR]](<3 x s32>), 0 - ; CHECK: $vgpr0_vgpr1 = COPY [[INSERT3]](<4 x s16>) - ; CHECK: $vgpr2_vgpr3_vgpr4_vgpr5 = COPY [[INSERT4]](<4 x s32>) - %0:_(<4 x s16>) = COPY $vgpr0_vgpr1 - %1:_(<4 x s16>) = COPY $vgpr1_vgpr2 - %2:_(<3 x s16>) = G_EXTRACT %0(<4 x s16>), 0 - %3:_(<3 x s16>) = G_EXTRACT %0(<4 x s16>), 0 - %4:_(<3 x s16>), %5:_(<3 x s1>) = G_SADDO %2, %3 - %6:_(<3 x s32>) = G_ZEXT %3 - %7:_(<4 x s16>) = G_IMPLICIT_DEF - %8:_(<4 x s16>) = G_INSERT %7, %4(<3 x s16>), 0 - %9:_(<4 x s32>) = G_IMPLICIT_DEF - %10:_(<4 x s32>) = G_INSERT %9, %6(<3 x s32>), 0 - $vgpr0_vgpr1 = COPY %8 - $vgpr2_vgpr3_vgpr4_vgpr5 = COPY %10 + ; CHECK: [[COPY11:%[0-9]+]]:_(s32) = COPY [[ADD]](s32) + ; CHECK: [[SEXT_INREG:%[0-9]+]]:_(s32) = G_SEXT_INREG [[COPY11]], 16 + ; CHECK: [[COPY12:%[0-9]+]]:_(s32) = COPY [[BITCAST6]](s32) + ; CHECK: [[SEXT_INREG1:%[0-9]+]]:_(s32) = G_SEXT_INREG [[COPY12]], 16 + ; CHECK: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[SEXT_INREG]](s32), [[SEXT_INREG1]] + ; CHECK: [[COPY13:%[0-9]+]]:_(s32) = COPY [[ADD1]](s32) + ; CHECK: [[SEXT_INREG2:%[0-9]+]]:_(s32) = G_SEXT_INREG [[COPY13]], 16 + ; CHECK: [[COPY14:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) + ; CHECK: [[SEXT_INREG3:%[0-9]+]]:_(s32) = G_SEXT_INREG [[COPY14]], 16 + ; CHECK: [[ICMP1:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[SEXT_INREG2]](s32), [[SEXT_INREG3]] + ; CHECK: [[COPY15:%[0-9]+]]:_(s32) = COPY [[ADD2]](s32) + ; CHECK: [[SEXT_INREG4:%[0-9]+]]:_(s32) = G_SEXT_INREG [[COPY15]], 16 + ; CHECK: [[COPY16:%[0-9]+]]:_(s32) = COPY [[BITCAST7]](s32) + ; CHECK: [[SEXT_INREG5:%[0-9]+]]:_(s32) = G_SEXT_INREG [[COPY16]], 16 + ; CHECK: [[ICMP2:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[SEXT_INREG4]](s32), [[SEXT_INREG5]] + ; CHECK: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0 + ; CHECK: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT3]](<4 x s16>) + ; CHECK: [[BITCAST8:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) + ; CHECK: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST8]], [[C]](s32) + ; CHECK: [[BITCAST9:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) + ; CHECK: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST9]], [[C]](s32) + ; CHECK: [[COPY17:%[0-9]+]]:_(s32) = COPY [[BITCAST8]](s32) + ; CHECK: [[SEXT_INREG6:%[0-9]+]]:_(s32) = G_SEXT_INREG [[COPY17]], 16 + ; CHECK: [[SEXT:%[0-9]+]]:_(s32) = G_SEXT [[C3]](s16) + ; CHECK: [[ICMP3:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[SEXT_INREG6]](s32), [[SEXT]] + ; CHECK: [[COPY18:%[0-9]+]]:_(s32) = COPY [[LSHR6]](s32) + ; CHECK: [[SEXT_INREG7:%[0-9]+]]:_(s32) = G_SEXT_INREG [[COPY18]], 16 + ; CHECK: [[SEXT1:%[0-9]+]]:_(s32) = G_SEXT [[C3]](s16) + ; CHECK: [[ICMP4:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[SEXT_INREG7]](s32), [[SEXT1]] + ; CHECK: [[COPY19:%[0-9]+]]:_(s32) = COPY [[BITCAST9]](s32) + ; CHECK: [[SEXT_INREG8:%[0-9]+]]:_(s32) = G_SEXT_INREG [[COPY19]], 16 + ; CHECK: [[SEXT2:%[0-9]+]]:_(s32) = G_SEXT [[C3]](s16) + ; CHECK: [[ICMP5:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[SEXT_INREG8]](s32), [[SEXT2]] + ; CHECK: [[XOR:%[0-9]+]]:_(s1) = G_XOR [[ICMP3]], [[ICMP]] + ; CHECK: [[XOR1:%[0-9]+]]:_(s1) = G_XOR [[ICMP4]], [[ICMP1]] + ; CHECK: [[XOR2:%[0-9]+]]:_(s1) = G_XOR [[ICMP5]], [[ICMP2]] + ; CHECK: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[XOR]](s1) + ; CHECK: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[XOR1]](s1) + ; CHECK: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[XOR2]](s1) + ; CHECK: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; CHECK: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; CHECK: [[UV12:%[0-9]+]]:_(<3 x s16>), [[UV13:%[0-9]+]]:_(<3 x s16>), [[UV14:%[0-9]+]]:_(<3 x s16>), [[UV15:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; CHECK: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; CHECK: [[UV16:%[0-9]+]]:_(<2 x s16>), [[UV17:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT4]](<4 x s16>) + ; CHECK: [[BITCAST10:%[0-9]+]]:_(s32) = G_BITCAST [[UV16]](<2 x s16>) + ; CHECK: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST10]], [[C]](s32) + ; CHECK: [[BITCAST11:%[0-9]+]]:_(s32) = G_BITCAST [[UV17]](<2 x s16>) + ; CHECK: [[LSHR9:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST11]], [[C]](s32) + ; CHECK: [[INSERT5:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV12]](<3 x s16>), 0 + ; CHECK: [[UV18:%[0-9]+]]:_(<2 x s16>), [[UV19:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT5]](<4 x s16>) + ; CHECK: [[BITCAST12:%[0-9]+]]:_(s32) = G_BITCAST [[UV18]](<2 x s16>) + ; CHECK: [[LSHR10:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST12]], [[C]](s32) + ; CHECK: [[BITCAST13:%[0-9]+]]:_(s32) = G_BITCAST [[UV19]](<2 x s16>) + ; CHECK: [[LSHR11:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST13]], [[C]](s32) + ; CHECK: [[COPY20:%[0-9]+]]:_(s32) = COPY [[BITCAST10]](s32) + ; CHECK: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY20]], [[C1]] + ; CHECK: [[COPY21:%[0-9]+]]:_(s32) = COPY [[LSHR8]](s32) + ; CHECK: [[AND4:%[0-9]+]]:_(s32) = G_AND [[COPY21]], [[C1]] + ; CHECK: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND4]], [[C]](s32) + ; CHECK: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND3]], [[SHL2]] + ; CHECK: [[BITCAST14:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32) + ; CHECK: [[COPY22:%[0-9]+]]:_(s32) = COPY [[BITCAST11]](s32) + ; CHECK: [[AND5:%[0-9]+]]:_(s32) = G_AND [[COPY22]], [[C1]] + ; CHECK: [[COPY23:%[0-9]+]]:_(s32) = COPY [[BITCAST12]](s32) + ; CHECK: [[AND6:%[0-9]+]]:_(s32) = G_AND [[COPY23]], [[C1]] + ; CHECK: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND6]], [[C]](s32) + ; CHECK: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND5]], [[SHL3]] + ; CHECK: [[BITCAST15:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32) + ; CHECK: [[COPY24:%[0-9]+]]:_(s32) = COPY [[LSHR10]](s32) + ; CHECK: [[AND7:%[0-9]+]]:_(s32) = G_AND [[COPY24]], [[C1]] + ; CHECK: [[COPY25:%[0-9]+]]:_(s32) = COPY [[BITCAST13]](s32) + ; CHECK: [[AND8:%[0-9]+]]:_(s32) = G_AND [[COPY25]], [[C1]] + ; CHECK: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND8]], [[C]](s32) + ; CHECK: [[OR4:%[0-9]+]]:_(s32) = G_OR [[AND7]], [[SHL4]] + ; CHECK: [[BITCAST16:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32) + ; CHECK: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST14]](<2 x s16>), [[BITCAST15]](<2 x s16>), [[BITCAST16]](<2 x s16>) + ; CHECK: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 1 + ; CHECK: [[COPY26:%[0-9]+]]:_(s32) = COPY [[ANYEXT]](s32) + ; CHECK: [[AND9:%[0-9]+]]:_(s32) = G_AND [[COPY26]], [[C4]] + ; CHECK: [[COPY27:%[0-9]+]]:_(s32) = COPY [[ANYEXT1]](s32) + ; CHECK: [[AND10:%[0-9]+]]:_(s32) = G_AND [[COPY27]], [[C4]] + ; CHECK: [[COPY28:%[0-9]+]]:_(s32) = COPY [[ANYEXT2]](s32) + ; CHECK: [[AND11:%[0-9]+]]:_(s32) = G_AND [[COPY28]], [[C4]] + ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[AND9]](s32), [[AND10]](s32), [[AND11]](s32) + ; CHECK: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) + ; CHECK: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>) + %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + %1:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + %2:_(<3 x s16>), %3:_(<3 x s16>) = G_UNMERGE_VALUES %0 + %4:_(<3 x s16>), %5:_(<3 x s16>) = G_UNMERGE_VALUES %1 + %6:_(<3 x s16>), %7:_(<3 x s1>) = G_SADDO %2, %4 + %8:_(<3 x s16>) = G_IMPLICIT_DEF + %9:_(<6 x s16>) = G_CONCAT_VECTORS %6, %8 + %10:_(<3 x s32>) = G_ZEXT %7 + $vgpr0_vgpr1_vgpr2 = COPY %9 + $vgpr0_vgpr1_vgpr2 = COPY %10 ... --- Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-select.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-select.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-select.mir @@ -387,35 +387,72 @@ name: test_select_v3s16 body: | bb.0: - liveins: $vgpr0, $vgpr1_vgpr2, $vgpr3_vgpr4 + liveins: $vgpr0_vgpr1_vgpr2, $vgpr3_vgpr4_vgpr5, $vgpr6 ; CHECK-LABEL: name: test_select_v3s16 - ; CHECK: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 - ; CHECK: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr1_vgpr2 - ; CHECK: [[COPY2:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr3_vgpr4 + ; CHECK: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + ; CHECK: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + ; CHECK: [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr6 ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 - ; CHECK: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(ne), [[COPY]](s32), [[C]] - ; CHECK: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY1]](<4 x s16>), 0 - ; CHECK: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY2]](<4 x s16>), 0 + ; CHECK: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(ne), [[COPY2]](s32), [[C]] + ; CHECK: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) + ; CHECK: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>) ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; CHECK: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0 + ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; CHECK: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0 ; CHECK: [[SELECT:%[0-9]+]]:_(<4 x s16>) = G_SELECT [[ICMP]](s1), [[INSERT]], [[INSERT1]] ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[SELECT]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) - ; CHECK: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) - ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 - ; CHECK: $vgpr0_vgpr1 = COPY [[INSERT2]](<4 x s16>) - %0:_(s32) = COPY $vgpr0 - %1:_(<4 x s16>) = COPY $vgpr1_vgpr2 - %2:_(<4 x s16>) = COPY $vgpr3_vgpr4 - %4:_(s32) = G_CONSTANT i32 0 - %5:_(s1) = G_ICMP intpred(ne), %0, %4 - - %6:_(<3 x s16>) = G_EXTRACT %1, 0 - %7:_(<3 x s16>) = G_EXTRACT %2, 0 - %8:_(<3 x s16>) = G_SELECT %5, %6, %7 - %9:_(<4 x s16>) = G_IMPLICIT_DEF - %10:_(<4 x s16>) = G_INSERT %9, %8, 0 - $vgpr0_vgpr1 = COPY %10 + ; CHECK: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; CHECK: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; CHECK: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; CHECK: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV4]](<3 x s16>), 0 + ; CHECK: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) + ; CHECK: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV12]](<2 x s16>) + ; CHECK: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 + ; CHECK: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C1]](s32) + ; CHECK: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV13]](<2 x s16>) + ; CHECK: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C1]](s32) + ; CHECK: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV8]](<3 x s16>), 0 + ; CHECK: [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT3]](<4 x s16>) + ; CHECK: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV14]](<2 x s16>) + ; CHECK: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C1]](s32) + ; CHECK: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV15]](<2 x s16>) + ; CHECK: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C1]](s32) + ; CHECK: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535 + ; CHECK: [[COPY3:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) + ; CHECK: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY3]], [[C2]] + ; CHECK: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) + ; CHECK: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY4]], [[C2]] + ; CHECK: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C1]](s32) + ; CHECK: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]] + ; CHECK: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) + ; CHECK: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST1]](s32) + ; CHECK: [[AND2:%[0-9]+]]:_(s32) = G_AND [[COPY5]], [[C2]] + ; CHECK: [[COPY6:%[0-9]+]]:_(s32) = COPY [[BITCAST2]](s32) + ; CHECK: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY6]], [[C2]] + ; CHECK: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C1]](s32) + ; CHECK: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]] + ; CHECK: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) + ; CHECK: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) + ; CHECK: [[AND4:%[0-9]+]]:_(s32) = G_AND [[COPY7]], [[C2]] + ; CHECK: [[COPY8:%[0-9]+]]:_(s32) = COPY [[BITCAST3]](s32) + ; CHECK: [[AND5:%[0-9]+]]:_(s32) = G_AND [[COPY8]], [[C2]] + ; CHECK: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C1]](s32) + ; CHECK: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]] + ; CHECK: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32) + ; CHECK: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>) + ; CHECK: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) + %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + %1:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + %2:_(s32) = COPY $vgpr6 + %3:_(s32) = G_CONSTANT i32 0 + %4:_(s1) = G_ICMP intpred(ne), %2, %3 + %5:_(<3 x s16>), %6:_(<3 x s16>) = G_UNMERGE_VALUES %0 + %7:_(<3 x s16>), %8:_(<3 x s16>) = G_UNMERGE_VALUES %1 + %9:_(<3 x s16>) = G_SELECT %4, %5, %7 + %10:_(<3 x s16>) = G_IMPLICIT_DEF + %11:_(<6 x s16>) = G_CONCAT_VECTORS %9, %10 + $vgpr0_vgpr1_vgpr2 = COPY %11 ... @@ -1247,62 +1284,102 @@ bb.0: liveins: $vgpr0_vgpr1_vgpr2, $vgpr3_vgpr4_vgpr5, $vgpr6_vgpr7_vgpr8, $vgpr9_vgpr10_vgpr11 ; CHECK-LABEL: name: test_vselect_v3s16 - ; CHECK: [[COPY:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2 - ; CHECK: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr3_vgpr4_vgpr5 + ; CHECK: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + ; CHECK: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 ; CHECK: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr6_vgpr7_vgpr8 ; CHECK: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr9_vgpr10_vgpr11 - ; CHECK: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<3 x s32>) - ; CHECK: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>) + ; CHECK: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>) + ; CHECK: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>) ; CHECK: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(ne), [[UV]](s32), [[UV3]] ; CHECK: [[ICMP1:%[0-9]+]]:_(s1) = G_ICMP intpred(ne), [[UV1]](s32), [[UV4]] ; CHECK: [[ICMP2:%[0-9]+]]:_(s1) = G_ICMP intpred(ne), [[UV2]](s32), [[UV5]] - ; CHECK: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>) - ; CHECK: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[UV6]](s32) - ; CHECK: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[UV7]](s32) - ; CHECK: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[UV8]](s32) - ; CHECK: [[UV9:%[0-9]+]]:_(s32), [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>) - ; CHECK: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[UV9]](s32) - ; CHECK: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[UV10]](s32) - ; CHECK: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[UV11]](s32) + ; CHECK: [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) + ; CHECK: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>) + ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV6]](<3 x s16>), 0 + ; CHECK: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; CHECK: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) + ; CHECK: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) + ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 + ; CHECK: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) + ; CHECK: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) + ; CHECK: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) + ; CHECK: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) + ; CHECK: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) + ; CHECK: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV8]](<3 x s16>), 0 + ; CHECK: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; CHECK: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV12]](<2 x s16>) + ; CHECK: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) + ; CHECK: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) + ; CHECK: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) + ; CHECK: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV13]](<2 x s16>) + ; CHECK: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) + ; CHECK: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; CHECK: [[SELECT:%[0-9]+]]:_(s16) = G_SELECT [[ICMP]](s1), [[TRUNC]], [[TRUNC3]] ; CHECK: [[SELECT1:%[0-9]+]]:_(s16) = G_SELECT [[ICMP1]](s1), [[TRUNC1]], [[TRUNC4]] ; CHECK: [[SELECT2:%[0-9]+]]:_(s16) = G_SELECT [[ICMP2]](s1), [[TRUNC2]], [[TRUNC5]] ; CHECK: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[SELECT]](s16) ; CHECK: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[SELECT1]](s16) - ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CHECK: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32) ; CHECK: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]] - ; CHECK: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) + ; CHECK: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) ; CHECK: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[SELECT2]](s16) ; CHECK: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 ; CHECK: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32) ; CHECK: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] - ; CHECK: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) - ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>) + ; CHECK: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) ; CHECK: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; CHECK: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; CHECK: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV12]](<2 x s16>) - ; CHECK: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) - ; CHECK: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV13]](<2 x s16>) - ; CHECK: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) - ; CHECK: [[COPY4:%[0-9]+]]:_(s32) = COPY [[BITCAST2]](s32) - ; CHECK: [[COPY5:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) - ; CHECK: [[COPY6:%[0-9]+]]:_(s32) = COPY [[BITCAST3]](s32) - ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32) - ; CHECK: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>) - %0:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2 - %1:_(<3 x s32>) = COPY $vgpr3_vgpr4_vgpr5 + ; CHECK: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; CHECK: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; CHECK: [[UV14:%[0-9]+]]:_(<3 x s16>), [[UV15:%[0-9]+]]:_(<3 x s16>), [[UV16:%[0-9]+]]:_(<3 x s16>), [[UV17:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; CHECK: [[UV18:%[0-9]+]]:_(<2 x s16>), [[UV19:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) + ; CHECK: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV18]](<2 x s16>) + ; CHECK: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32) + ; CHECK: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV19]](<2 x s16>) + ; CHECK: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST7]], [[C]](s32) + ; CHECK: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV14]](<3 x s16>), 0 + ; CHECK: [[UV20:%[0-9]+]]:_(<2 x s16>), [[UV21:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT3]](<4 x s16>) + ; CHECK: [[BITCAST8:%[0-9]+]]:_(s32) = G_BITCAST [[UV20]](<2 x s16>) + ; CHECK: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST8]], [[C]](s32) + ; CHECK: [[BITCAST9:%[0-9]+]]:_(s32) = G_BITCAST [[UV21]](<2 x s16>) + ; CHECK: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST9]], [[C]](s32) + ; CHECK: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535 + ; CHECK: [[COPY4:%[0-9]+]]:_(s32) = COPY [[BITCAST6]](s32) + ; CHECK: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY4]], [[C2]] + ; CHECK: [[COPY5:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) + ; CHECK: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY5]], [[C2]] + ; CHECK: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32) + ; CHECK: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL2]] + ; CHECK: [[BITCAST10:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32) + ; CHECK: [[COPY6:%[0-9]+]]:_(s32) = COPY [[BITCAST7]](s32) + ; CHECK: [[AND2:%[0-9]+]]:_(s32) = G_AND [[COPY6]], [[C2]] + ; CHECK: [[COPY7:%[0-9]+]]:_(s32) = COPY [[BITCAST8]](s32) + ; CHECK: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY7]], [[C2]] + ; CHECK: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32) + ; CHECK: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL3]] + ; CHECK: [[BITCAST11:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32) + ; CHECK: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR6]](s32) + ; CHECK: [[AND4:%[0-9]+]]:_(s32) = G_AND [[COPY8]], [[C2]] + ; CHECK: [[COPY9:%[0-9]+]]:_(s32) = COPY [[BITCAST9]](s32) + ; CHECK: [[AND5:%[0-9]+]]:_(s32) = G_AND [[COPY9]], [[C2]] + ; CHECK: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32) + ; CHECK: [[OR4:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL4]] + ; CHECK: [[BITCAST12:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32) + ; CHECK: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST10]](<2 x s16>), [[BITCAST11]](<2 x s16>), [[BITCAST12]](<2 x s16>) + ; CHECK: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) + %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + %1:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 %2:_(<3 x s32>) = COPY $vgpr6_vgpr7_vgpr8 %3:_(<3 x s32>) = COPY $vgpr9_vgpr10_vgpr11 - %4:_(<3 x s1>) = G_ICMP intpred(ne), %0, %1 - %5:_(<3 x s16>) = G_TRUNC %2 - %6:_(<3 x s16>) = G_TRUNC %3 - %7:_(<3 x s16>) = G_SELECT %4, %5, %6 - %8:_(<3 x s32>) = G_ANYEXT %7 - $vgpr0_vgpr1_vgpr2 = COPY %8 - + %4:_(<3 x s1>) = G_ICMP intpred(ne), %2, %3 + %5:_(<3 x s16>), %6:_(<3 x s16>) = G_UNMERGE_VALUES %0 + %7:_(<3 x s16>), %8:_(<3 x s16>) = G_UNMERGE_VALUES %1 + %9:_(<3 x s16>) = G_SELECT %4, %5, %7 + %10:_(<3 x s16>) = G_IMPLICIT_DEF + %11:_(<6 x s16>) = G_CONCAT_VECTORS %9, %10 + $vgpr0_vgpr1_vgpr2 = COPY %11 ... --- Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-sext-inreg.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-sext-inreg.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-sext-inreg.mir @@ -691,18 +691,17 @@ name: test_sext_inreg_v3s16_1 body: | bb.0: - liveins: $vgpr0_vgpr1 - + liveins: $vgpr0_vgpr1_vgpr2 ; GFX9-LABEL: name: test_sext_inreg_v3s16_1 - ; GFX9: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0 + ; GFX9: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; GFX9: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; GFX9: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) - ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) ; GFX9: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) @@ -727,21 +726,45 @@ ; GFX9: [[SHL2:%[0-9]+]]:_(<2 x s16>) = G_SHL [[DEF2]], [[BUILD_VECTOR_TRUNC4]](<2 x s16>) ; GFX9: [[ASHR2:%[0-9]+]]:_(<2 x s16>) = G_ASHR [[SHL2]], [[BUILD_VECTOR_TRUNC4]](<2 x s16>) ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[ASHR]](<2 x s16>), [[ASHR1]](<2 x s16>), [[ASHR2]](<2 x s16>) - ; GFX9: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<6 x s16>), 0 - ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[COPY]], [[EXTRACT1]](<3 x s16>), 0 - ; GFX9: $vgpr0_vgpr1 = COPY [[INSERT1]](<4 x s16>) + ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<6 x s16>), 0 + ; GFX9: [[DEF3:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF3]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX9: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; GFX9: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; GFX9: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) + ; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) + ; GFX9: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) + ; GFX9: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) + ; GFX9: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV4]](<3 x s16>), 0 + ; GFX9: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) + ; GFX9: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) + ; GFX9: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32) + ; GFX9: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) + ; GFX9: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32) + ; GFX9: [[COPY9:%[0-9]+]]:_(s32) = COPY [[BITCAST2]](s32) + ; GFX9: [[COPY10:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) + ; GFX9: [[BUILD_VECTOR_TRUNC5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY9]](s32), [[COPY10]](s32) + ; GFX9: [[COPY11:%[0-9]+]]:_(s32) = COPY [[BITCAST3]](s32) + ; GFX9: [[COPY12:%[0-9]+]]:_(s32) = COPY [[BITCAST4]](s32) + ; GFX9: [[BUILD_VECTOR_TRUNC6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY11]](s32), [[COPY12]](s32) + ; GFX9: [[COPY13:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) + ; GFX9: [[COPY14:%[0-9]+]]:_(s32) = COPY [[BITCAST5]](s32) + ; GFX9: [[BUILD_VECTOR_TRUNC7:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY13]](s32), [[COPY14]](s32) + ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC5]](<2 x s16>), [[BUILD_VECTOR_TRUNC6]](<2 x s16>), [[BUILD_VECTOR_TRUNC7]](<2 x s16>) + ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) ; GFX8-LABEL: name: test_sext_inreg_v3s16_1 - ; GFX8: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 - ; GFX8: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0 + ; GFX8: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + ; GFX8: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) ; GFX8: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX8: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; GFX8: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; GFX8: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; GFX8: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; GFX8: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; GFX8: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) ; GFX8: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; GFX8: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX8: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; GFX8: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; GFX8: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; GFX8: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) ; GFX8: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; GFX8: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX8: [[C1:%[0-9]+]]:_(s16) = G_CONSTANT i16 15 @@ -762,19 +785,56 @@ ; GFX8: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL4]] ; GFX8: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) ; GFX8: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>) - ; GFX8: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; GFX8: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[COPY]], [[EXTRACT1]](<3 x s16>), 0 - ; GFX8: $vgpr0_vgpr1 = COPY [[INSERT1]](<4 x s16>) + ; GFX8: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 + ; GFX8: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX8: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX8: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX8: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; GFX8: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; GFX8: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) + ; GFX8: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32) + ; GFX8: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) + ; GFX8: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32) + ; GFX8: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV4]](<3 x s16>), 0 + ; GFX8: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) + ; GFX8: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) + ; GFX8: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32) + ; GFX8: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) + ; GFX8: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST7]], [[C]](s32) + ; GFX8: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535 + ; GFX8: [[COPY1:%[0-9]+]]:_(s32) = COPY [[BITCAST4]](s32) + ; GFX8: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C3]] + ; GFX8: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) + ; GFX8: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C3]] + ; GFX8: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32) + ; GFX8: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL5]] + ; GFX8: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32) + ; GFX8: [[COPY3:%[0-9]+]]:_(s32) = COPY [[BITCAST5]](s32) + ; GFX8: [[AND2:%[0-9]+]]:_(s32) = G_AND [[COPY3]], [[C3]] + ; GFX8: [[COPY4:%[0-9]+]]:_(s32) = COPY [[BITCAST6]](s32) + ; GFX8: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY4]], [[C3]] + ; GFX8: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32) + ; GFX8: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL6]] + ; GFX8: [[BITCAST9:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32) + ; GFX8: [[COPY5:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) + ; GFX8: [[AND4:%[0-9]+]]:_(s32) = G_AND [[COPY5]], [[C3]] + ; GFX8: [[COPY6:%[0-9]+]]:_(s32) = COPY [[BITCAST7]](s32) + ; GFX8: [[AND5:%[0-9]+]]:_(s32) = G_AND [[COPY6]], [[C3]] + ; GFX8: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32) + ; GFX8: [[OR4:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL7]] + ; GFX8: [[BITCAST10:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32) + ; GFX8: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST8]](<2 x s16>), [[BITCAST9]](<2 x s16>), [[BITCAST10]](<2 x s16>) + ; GFX8: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) ; GFX6-LABEL: name: test_sext_inreg_v3s16_1 - ; GFX6: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 - ; GFX6: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0 + ; GFX6: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + ; GFX6: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) ; GFX6: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX6: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; GFX6: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; GFX6: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; GFX6: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; GFX6: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; GFX6: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) ; GFX6: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; GFX6: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) - ; GFX6: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; GFX6: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) ; GFX6: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; GFX6: [[COPY1:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) ; GFX6: [[SEXT_INREG:%[0-9]+]]:_(s32) = G_SEXT_INREG [[COPY1]], 1 @@ -797,14 +857,51 @@ ; GFX6: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]] ; GFX6: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) ; GFX6: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>) - ; GFX6: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; GFX6: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[COPY]], [[EXTRACT1]](<3 x s16>), 0 - ; GFX6: $vgpr0_vgpr1 = COPY [[INSERT1]](<4 x s16>) - %0:_(<4 x s16>) = COPY $vgpr0_vgpr1 - %1:_(<3 x s16>) = G_EXTRACT %0, 0 - %2:_(<3 x s16>) = G_SEXT_INREG %1, 1 - %3:_(<4 x s16>) = G_INSERT %0, %2, 0 - $vgpr0_vgpr1 = COPY %3 + ; GFX6: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 + ; GFX6: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX6: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX6: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX6: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; GFX6: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; GFX6: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) + ; GFX6: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32) + ; GFX6: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) + ; GFX6: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32) + ; GFX6: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV4]](<3 x s16>), 0 + ; GFX6: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) + ; GFX6: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) + ; GFX6: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32) + ; GFX6: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) + ; GFX6: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST7]], [[C]](s32) + ; GFX6: [[COPY7:%[0-9]+]]:_(s32) = COPY [[BITCAST4]](s32) + ; GFX6: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY7]], [[C1]] + ; GFX6: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) + ; GFX6: [[AND4:%[0-9]+]]:_(s32) = G_AND [[COPY8]], [[C1]] + ; GFX6: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND4]], [[C]](s32) + ; GFX6: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND3]], [[SHL2]] + ; GFX6: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32) + ; GFX6: [[COPY9:%[0-9]+]]:_(s32) = COPY [[BITCAST5]](s32) + ; GFX6: [[AND5:%[0-9]+]]:_(s32) = G_AND [[COPY9]], [[C1]] + ; GFX6: [[COPY10:%[0-9]+]]:_(s32) = COPY [[BITCAST6]](s32) + ; GFX6: [[AND6:%[0-9]+]]:_(s32) = G_AND [[COPY10]], [[C1]] + ; GFX6: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND6]], [[C]](s32) + ; GFX6: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND5]], [[SHL3]] + ; GFX6: [[BITCAST9:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32) + ; GFX6: [[COPY11:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) + ; GFX6: [[AND7:%[0-9]+]]:_(s32) = G_AND [[COPY11]], [[C1]] + ; GFX6: [[COPY12:%[0-9]+]]:_(s32) = COPY [[BITCAST7]](s32) + ; GFX6: [[AND8:%[0-9]+]]:_(s32) = G_AND [[COPY12]], [[C1]] + ; GFX6: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND8]], [[C]](s32) + ; GFX6: [[OR4:%[0-9]+]]:_(s32) = G_OR [[AND7]], [[SHL4]] + ; GFX6: [[BITCAST10:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32) + ; GFX6: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST8]](<2 x s16>), [[BITCAST9]](<2 x s16>), [[BITCAST10]](<2 x s16>) + ; GFX6: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) + %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + %1:_(<3 x s16>), %2:_(<3 x s16>) = G_UNMERGE_VALUES %0 + %3:_(<3 x s16>) = G_SEXT_INREG %1, 1 + %4:_(<3 x s16>) = G_IMPLICIT_DEF + %5:_(<6 x s16>) = G_CONCAT_VECTORS %3, %4 + $vgpr0_vgpr1_vgpr2 = COPY %5 ... --- Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-shl.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-shl.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-shl.mir @@ -600,26 +600,25 @@ name: test_shl_v3s16_v3s16 body: | bb.0: - liveins: $vgpr0_vgpr1, $vgpr2_vgpr3 - + liveins: $vgpr0_vgpr1_vgpr2, $vgpr3_vgpr4_vgpr5 ; SI-LABEL: name: test_shl_v3s16_v3s16 - ; SI: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 - ; SI: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3 - ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0 - ; SI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY1]](<4 x s16>), 0 + ; SI: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + ; SI: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + ; SI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) + ; SI: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>) ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; SI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; SI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; SI: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; SI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; SI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; SI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) - ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; SI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; SI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0 - ; SI: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; SI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) + ; SI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0 + ; SI: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; SI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>) ; SI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) - ; SI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; SI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>) ; SI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; SI: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535 ; SI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST2]](s32) @@ -648,32 +647,68 @@ ; SI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND5]], [[SHL4]] ; SI: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) ; SI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) - ; SI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; SI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT2]](<3 x s16>), 0 - ; SI: $vgpr0_vgpr1 = COPY [[INSERT2]](<4 x s16>) + ; SI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 + ; SI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; SI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; SI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; SI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; SI: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) + ; SI: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV12]](<2 x s16>) + ; SI: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32) + ; SI: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV13]](<2 x s16>) + ; SI: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST7]], [[C]](s32) + ; SI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV8]](<3 x s16>), 0 + ; SI: [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT3]](<4 x s16>) + ; SI: [[BITCAST8:%[0-9]+]]:_(s32) = G_BITCAST [[UV14]](<2 x s16>) + ; SI: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST8]], [[C]](s32) + ; SI: [[BITCAST9:%[0-9]+]]:_(s32) = G_BITCAST [[UV15]](<2 x s16>) + ; SI: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST9]], [[C]](s32) + ; SI: [[COPY11:%[0-9]+]]:_(s32) = COPY [[BITCAST6]](s32) + ; SI: [[AND6:%[0-9]+]]:_(s32) = G_AND [[COPY11]], [[C1]] + ; SI: [[COPY12:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) + ; SI: [[AND7:%[0-9]+]]:_(s32) = G_AND [[COPY12]], [[C1]] + ; SI: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C]](s32) + ; SI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL5]] + ; SI: [[BITCAST10:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32) + ; SI: [[COPY13:%[0-9]+]]:_(s32) = COPY [[BITCAST7]](s32) + ; SI: [[AND8:%[0-9]+]]:_(s32) = G_AND [[COPY13]], [[C1]] + ; SI: [[COPY14:%[0-9]+]]:_(s32) = COPY [[BITCAST8]](s32) + ; SI: [[AND9:%[0-9]+]]:_(s32) = G_AND [[COPY14]], [[C1]] + ; SI: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[AND9]], [[C]](s32) + ; SI: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND8]], [[SHL6]] + ; SI: [[BITCAST11:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32) + ; SI: [[COPY15:%[0-9]+]]:_(s32) = COPY [[LSHR6]](s32) + ; SI: [[AND10:%[0-9]+]]:_(s32) = G_AND [[COPY15]], [[C1]] + ; SI: [[COPY16:%[0-9]+]]:_(s32) = COPY [[BITCAST9]](s32) + ; SI: [[AND11:%[0-9]+]]:_(s32) = G_AND [[COPY16]], [[C1]] + ; SI: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[AND11]], [[C]](s32) + ; SI: [[OR4:%[0-9]+]]:_(s32) = G_OR [[AND10]], [[SHL7]] + ; SI: [[BITCAST12:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32) + ; SI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST10]](<2 x s16>), [[BITCAST11]](<2 x s16>), [[BITCAST12]](<2 x s16>) + ; SI: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) ; VI-LABEL: name: test_shl_v3s16_v3s16 - ; VI: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 - ; VI: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3 - ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0 - ; VI: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY1]](<4 x s16>), 0 + ; VI: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + ; VI: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + ; VI: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) + ; VI: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>) ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; VI: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; VI: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; VI: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; VI: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; VI: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; VI: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; VI: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; VI: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; VI: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; VI: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; VI: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0 - ; VI: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; VI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) + ; VI: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0 + ; VI: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; VI: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>) ; VI: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; VI: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; VI: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) - ; VI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; VI: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>) ; VI: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; VI: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; VI: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[TRUNC]], [[TRUNC3]](s16) @@ -690,46 +725,109 @@ ; VI: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL4]] ; VI: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) ; VI: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) - ; VI: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; VI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT2]](<3 x s16>), 0 - ; VI: $vgpr0_vgpr1 = COPY [[INSERT2]](<4 x s16>) + ; VI: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 + ; VI: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; VI: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; VI: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; VI: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; VI: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) + ; VI: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV12]](<2 x s16>) + ; VI: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32) + ; VI: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV13]](<2 x s16>) + ; VI: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST7]], [[C]](s32) + ; VI: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV8]](<3 x s16>), 0 + ; VI: [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT3]](<4 x s16>) + ; VI: [[BITCAST8:%[0-9]+]]:_(s32) = G_BITCAST [[UV14]](<2 x s16>) + ; VI: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST8]], [[C]](s32) + ; VI: [[BITCAST9:%[0-9]+]]:_(s32) = G_BITCAST [[UV15]](<2 x s16>) + ; VI: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST9]], [[C]](s32) + ; VI: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535 + ; VI: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST6]](s32) + ; VI: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C2]] + ; VI: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) + ; VI: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY3]], [[C2]] + ; VI: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32) + ; VI: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL5]] + ; VI: [[BITCAST10:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32) + ; VI: [[COPY4:%[0-9]+]]:_(s32) = COPY [[BITCAST7]](s32) + ; VI: [[AND2:%[0-9]+]]:_(s32) = G_AND [[COPY4]], [[C2]] + ; VI: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST8]](s32) + ; VI: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY5]], [[C2]] + ; VI: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32) + ; VI: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL6]] + ; VI: [[BITCAST11:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32) + ; VI: [[COPY6:%[0-9]+]]:_(s32) = COPY [[LSHR6]](s32) + ; VI: [[AND4:%[0-9]+]]:_(s32) = G_AND [[COPY6]], [[C2]] + ; VI: [[COPY7:%[0-9]+]]:_(s32) = COPY [[BITCAST9]](s32) + ; VI: [[AND5:%[0-9]+]]:_(s32) = G_AND [[COPY7]], [[C2]] + ; VI: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32) + ; VI: [[OR4:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL7]] + ; VI: [[BITCAST12:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32) + ; VI: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST10]](<2 x s16>), [[BITCAST11]](<2 x s16>), [[BITCAST12]](<2 x s16>) + ; VI: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) ; GFX9-LABEL: name: test_shl_v3s16_v3s16 - ; GFX9: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 - ; GFX9: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3 - ; GFX9: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0 - ; GFX9: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY1]](<4 x s16>), 0 + ; GFX9: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + ; GFX9: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) + ; GFX9: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>) ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; GFX9: [[EXTRACT2:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 0 - ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; GFX9: [[EXTRACT3:%[0-9]+]]:_(s16) = G_EXTRACT [[INSERT1]](<4 x s16>), 32 - ; GFX9: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0 - ; GFX9: [[EXTRACT4:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT2]](<4 x s16>), 0 - ; GFX9: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0 - ; GFX9: [[EXTRACT5:%[0-9]+]]:_(s16) = G_EXTRACT [[INSERT3]](<4 x s16>), 32 - ; GFX9: [[SHL:%[0-9]+]]:_(<2 x s16>) = G_SHL [[EXTRACT2]], [[EXTRACT4]](<2 x s16>) - ; GFX9: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[EXTRACT3]], [[EXTRACT5]](s16) + ; GFX9: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; GFX9: [[EXTRACT:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 0 + ; GFX9: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; GFX9: [[EXTRACT1:%[0-9]+]]:_(s16) = G_EXTRACT [[INSERT1]](<4 x s16>), 32 + ; GFX9: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0 + ; GFX9: [[EXTRACT2:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT2]](<4 x s16>), 0 + ; GFX9: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0 + ; GFX9: [[EXTRACT3:%[0-9]+]]:_(s16) = G_EXTRACT [[INSERT3]](<4 x s16>), 32 + ; GFX9: [[SHL:%[0-9]+]]:_(<2 x s16>) = G_SHL [[EXTRACT]], [[EXTRACT2]](<2 x s16>) + ; GFX9: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[EXTRACT1]], [[EXTRACT3]](s16) ; GFX9: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) - ; GFX9: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) - ; GFX9: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; GFX9: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; GFX9: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV4]](<3 x s16>), 0 ; GFX9: [[INSERT5:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT4]], [[SHL]](<2 x s16>), 0 ; GFX9: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT5]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) - ; GFX9: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) - ; GFX9: [[INSERT6:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV4]](<3 x s16>), 0 + ; GFX9: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; GFX9: [[INSERT6:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV8]](<3 x s16>), 0 ; GFX9: [[INSERT7:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT6]], [[SHL1]](s16), 32 ; GFX9: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT7]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) - ; GFX9: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) - ; GFX9: [[INSERT8:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV8]](<3 x s16>), 0 - ; GFX9: $vgpr0_vgpr1 = COPY [[INSERT8]](<4 x s16>) - %0:_(<4 x s16>) = COPY $vgpr0_vgpr1 - %1:_(<4 x s16>) = COPY $vgpr2_vgpr3 - %2:_(<3 x s16>) = G_EXTRACT %0, 0 - %3:_(<3 x s16>) = G_EXTRACT %1, 0 - %4:_(<3 x s16>) = G_SHL %2, %3 - %5:_(<4 x s16>) = G_IMPLICIT_DEF - %6:_(<4 x s16>) = G_INSERT %5, %4, 0 - $vgpr0_vgpr1 = COPY %6 + ; GFX9: [[UV12:%[0-9]+]]:_(<3 x s16>), [[UV13:%[0-9]+]]:_(<3 x s16>), [[UV14:%[0-9]+]]:_(<3 x s16>), [[UV15:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) + ; GFX9: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; GFX9: [[CONCAT_VECTORS3:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; GFX9: [[UV16:%[0-9]+]]:_(<3 x s16>), [[UV17:%[0-9]+]]:_(<3 x s16>), [[UV18:%[0-9]+]]:_(<3 x s16>), [[UV19:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS3]](<12 x s16>) + ; GFX9: [[INSERT8:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV12]](<3 x s16>), 0 + ; GFX9: [[UV20:%[0-9]+]]:_(<2 x s16>), [[UV21:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT8]](<4 x s16>) + ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV20]](<2 x s16>) + ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 + ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) + ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV21]](<2 x s16>) + ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) + ; GFX9: [[INSERT9:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV16]](<3 x s16>), 0 + ; GFX9: [[UV22:%[0-9]+]]:_(<2 x s16>), [[UV23:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT9]](<4 x s16>) + ; GFX9: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV22]](<2 x s16>) + ; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) + ; GFX9: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV23]](<2 x s16>) + ; GFX9: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) + ; GFX9: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) + ; GFX9: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) + ; GFX9: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY2]](s32), [[COPY3]](s32) + ; GFX9: [[COPY4:%[0-9]+]]:_(s32) = COPY [[BITCAST1]](s32) + ; GFX9: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST2]](s32) + ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY4]](s32), [[COPY5]](s32) + ; GFX9: [[COPY6:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) + ; GFX9: [[COPY7:%[0-9]+]]:_(s32) = COPY [[BITCAST3]](s32) + ; GFX9: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY6]](s32), [[COPY7]](s32) + ; GFX9: [[CONCAT_VECTORS4:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[BUILD_VECTOR_TRUNC2]](<2 x s16>) + ; GFX9: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS4]](<6 x s16>) + %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + %1:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + %2:_(<3 x s16>), %3:_(<3 x s16>) = G_UNMERGE_VALUES %0 + %4:_(<3 x s16>), %5:_(<3 x s16>) = G_UNMERGE_VALUES %1 + %6:_(<3 x s16>) = G_SHL %2, %4 + %7:_(<3 x s16>) = G_IMPLICIT_DEF + %8:_(<6 x s16>) = G_CONCAT_VECTORS %6, %7 + $vgpr0_vgpr1_vgpr2 = COPY %8 + ... --- Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-ssubo.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-ssubo.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-ssubo.mir @@ -209,26 +209,25 @@ name: test_ssubo_v3s16 body: | bb.0: - liveins: $vgpr0_vgpr1, $vgpr2_vgpr3 - + liveins: $vgpr0_vgpr1_vgpr2, $vgpr3_vgpr4_vgpr5 ; CHECK-LABEL: name: test_ssubo_v3s16 - ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 - ; CHECK: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr1_vgpr2 - ; CHECK: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0 - ; CHECK: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0 + ; CHECK: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + ; CHECK: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + ; CHECK: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) + ; CHECK: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>) ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; CHECK: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; CHECK: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; CHECK: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; CHECK: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CHECK: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) - ; CHECK: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; CHECK: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; CHECK: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; CHECK: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0 - ; CHECK: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; CHECK: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) + ; CHECK: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0 + ; CHECK: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; CHECK: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>) ; CHECK: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) - ; CHECK: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; CHECK: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>) ; CHECK: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; CHECK: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) ; CHECK: [[COPY3:%[0-9]+]]:_(s32) = COPY [[BITCAST2]](s32) @@ -254,38 +253,110 @@ ; CHECK: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]] ; CHECK: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) - ; CHECK: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 + ; CHECK: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 ; CHECK: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 0 - ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0 - ; CHECK: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) - ; CHECK: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) + ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; CHECK: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) + ; CHECK: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) ; CHECK: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32) - ; CHECK: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) + ; CHECK: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) ; CHECK: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST7]], [[C]](s32) - ; CHECK: [[COPY11:%[0-9]+]]:_(s32) = COPY [[BITCAST6]](s32) - ; CHECK: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY11]], [[C1]] - ; CHECK: [[COPY12:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) - ; CHECK: [[AND4:%[0-9]+]]:_(s32) = G_AND [[COPY12]], [[C1]] - ; CHECK: [[COPY13:%[0-9]+]]:_(s32) = COPY [[BITCAST7]](s32) - ; CHECK: [[AND5:%[0-9]+]]:_(s32) = G_AND [[COPY13]], [[C1]] - ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[AND3]](s32), [[AND4]](s32), [[AND5]](s32) - ; CHECK: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT2]](<3 x s16>), 0 - ; CHECK: [[DEF1:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF - ; CHECK: [[INSERT4:%[0-9]+]]:_(<4 x s32>) = G_INSERT [[DEF1]], [[BUILD_VECTOR]](<3 x s32>), 0 - ; CHECK: $vgpr0_vgpr1 = COPY [[INSERT3]](<4 x s16>) - ; CHECK: $vgpr2_vgpr3_vgpr4_vgpr5 = COPY [[INSERT4]](<4 x s32>) - %0:_(<4 x s16>) = COPY $vgpr0_vgpr1 - %1:_(<4 x s16>) = COPY $vgpr1_vgpr2 - %2:_(<3 x s16>) = G_EXTRACT %0(<4 x s16>), 0 - %3:_(<3 x s16>) = G_EXTRACT %0(<4 x s16>), 0 - %4:_(<3 x s16>), %5:_(<3 x s1>) = G_SSUBO %2, %3 - %6:_(<3 x s32>) = G_ZEXT %3 - %7:_(<4 x s16>) = G_IMPLICIT_DEF - %8:_(<4 x s16>) = G_INSERT %7, %4(<3 x s16>), 0 - %9:_(<4 x s32>) = G_IMPLICIT_DEF - %10:_(<4 x s32>) = G_INSERT %9, %6(<3 x s32>), 0 - $vgpr0_vgpr1 = COPY %8 - $vgpr2_vgpr3_vgpr4_vgpr5 = COPY %10 + ; CHECK: [[COPY11:%[0-9]+]]:_(s32) = COPY [[SUB]](s32) + ; CHECK: [[SEXT_INREG:%[0-9]+]]:_(s32) = G_SEXT_INREG [[COPY11]], 16 + ; CHECK: [[COPY12:%[0-9]+]]:_(s32) = COPY [[BITCAST6]](s32) + ; CHECK: [[SEXT_INREG1:%[0-9]+]]:_(s32) = G_SEXT_INREG [[COPY12]], 16 + ; CHECK: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[SEXT_INREG]](s32), [[SEXT_INREG1]] + ; CHECK: [[COPY13:%[0-9]+]]:_(s32) = COPY [[SUB1]](s32) + ; CHECK: [[SEXT_INREG2:%[0-9]+]]:_(s32) = G_SEXT_INREG [[COPY13]], 16 + ; CHECK: [[COPY14:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) + ; CHECK: [[SEXT_INREG3:%[0-9]+]]:_(s32) = G_SEXT_INREG [[COPY14]], 16 + ; CHECK: [[ICMP1:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[SEXT_INREG2]](s32), [[SEXT_INREG3]] + ; CHECK: [[COPY15:%[0-9]+]]:_(s32) = COPY [[SUB2]](s32) + ; CHECK: [[SEXT_INREG4:%[0-9]+]]:_(s32) = G_SEXT_INREG [[COPY15]], 16 + ; CHECK: [[COPY16:%[0-9]+]]:_(s32) = COPY [[BITCAST7]](s32) + ; CHECK: [[SEXT_INREG5:%[0-9]+]]:_(s32) = G_SEXT_INREG [[COPY16]], 16 + ; CHECK: [[ICMP2:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[SEXT_INREG4]](s32), [[SEXT_INREG5]] + ; CHECK: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0 + ; CHECK: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT3]](<4 x s16>) + ; CHECK: [[BITCAST8:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) + ; CHECK: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST8]], [[C]](s32) + ; CHECK: [[BITCAST9:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) + ; CHECK: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST9]], [[C]](s32) + ; CHECK: [[COPY17:%[0-9]+]]:_(s32) = COPY [[BITCAST8]](s32) + ; CHECK: [[SEXT_INREG6:%[0-9]+]]:_(s32) = G_SEXT_INREG [[COPY17]], 16 + ; CHECK: [[SEXT:%[0-9]+]]:_(s32) = G_SEXT [[C3]](s16) + ; CHECK: [[ICMP3:%[0-9]+]]:_(s1) = G_ICMP intpred(sgt), [[SEXT_INREG6]](s32), [[SEXT]] + ; CHECK: [[COPY18:%[0-9]+]]:_(s32) = COPY [[LSHR6]](s32) + ; CHECK: [[SEXT_INREG7:%[0-9]+]]:_(s32) = G_SEXT_INREG [[COPY18]], 16 + ; CHECK: [[SEXT1:%[0-9]+]]:_(s32) = G_SEXT [[C3]](s16) + ; CHECK: [[ICMP4:%[0-9]+]]:_(s1) = G_ICMP intpred(sgt), [[SEXT_INREG7]](s32), [[SEXT1]] + ; CHECK: [[COPY19:%[0-9]+]]:_(s32) = COPY [[BITCAST9]](s32) + ; CHECK: [[SEXT_INREG8:%[0-9]+]]:_(s32) = G_SEXT_INREG [[COPY19]], 16 + ; CHECK: [[SEXT2:%[0-9]+]]:_(s32) = G_SEXT [[C3]](s16) + ; CHECK: [[ICMP5:%[0-9]+]]:_(s1) = G_ICMP intpred(sgt), [[SEXT_INREG8]](s32), [[SEXT2]] + ; CHECK: [[XOR:%[0-9]+]]:_(s1) = G_XOR [[ICMP3]], [[ICMP]] + ; CHECK: [[XOR1:%[0-9]+]]:_(s1) = G_XOR [[ICMP4]], [[ICMP1]] + ; CHECK: [[XOR2:%[0-9]+]]:_(s1) = G_XOR [[ICMP5]], [[ICMP2]] + ; CHECK: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[XOR]](s1) + ; CHECK: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[XOR1]](s1) + ; CHECK: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[XOR2]](s1) + ; CHECK: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; CHECK: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; CHECK: [[UV12:%[0-9]+]]:_(<3 x s16>), [[UV13:%[0-9]+]]:_(<3 x s16>), [[UV14:%[0-9]+]]:_(<3 x s16>), [[UV15:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; CHECK: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; CHECK: [[UV16:%[0-9]+]]:_(<2 x s16>), [[UV17:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT4]](<4 x s16>) + ; CHECK: [[BITCAST10:%[0-9]+]]:_(s32) = G_BITCAST [[UV16]](<2 x s16>) + ; CHECK: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST10]], [[C]](s32) + ; CHECK: [[BITCAST11:%[0-9]+]]:_(s32) = G_BITCAST [[UV17]](<2 x s16>) + ; CHECK: [[LSHR9:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST11]], [[C]](s32) + ; CHECK: [[INSERT5:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV12]](<3 x s16>), 0 + ; CHECK: [[UV18:%[0-9]+]]:_(<2 x s16>), [[UV19:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT5]](<4 x s16>) + ; CHECK: [[BITCAST12:%[0-9]+]]:_(s32) = G_BITCAST [[UV18]](<2 x s16>) + ; CHECK: [[LSHR10:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST12]], [[C]](s32) + ; CHECK: [[BITCAST13:%[0-9]+]]:_(s32) = G_BITCAST [[UV19]](<2 x s16>) + ; CHECK: [[LSHR11:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST13]], [[C]](s32) + ; CHECK: [[COPY20:%[0-9]+]]:_(s32) = COPY [[BITCAST10]](s32) + ; CHECK: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY20]], [[C1]] + ; CHECK: [[COPY21:%[0-9]+]]:_(s32) = COPY [[LSHR8]](s32) + ; CHECK: [[AND4:%[0-9]+]]:_(s32) = G_AND [[COPY21]], [[C1]] + ; CHECK: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND4]], [[C]](s32) + ; CHECK: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND3]], [[SHL2]] + ; CHECK: [[BITCAST14:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32) + ; CHECK: [[COPY22:%[0-9]+]]:_(s32) = COPY [[BITCAST11]](s32) + ; CHECK: [[AND5:%[0-9]+]]:_(s32) = G_AND [[COPY22]], [[C1]] + ; CHECK: [[COPY23:%[0-9]+]]:_(s32) = COPY [[BITCAST12]](s32) + ; CHECK: [[AND6:%[0-9]+]]:_(s32) = G_AND [[COPY23]], [[C1]] + ; CHECK: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND6]], [[C]](s32) + ; CHECK: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND5]], [[SHL3]] + ; CHECK: [[BITCAST15:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32) + ; CHECK: [[COPY24:%[0-9]+]]:_(s32) = COPY [[LSHR10]](s32) + ; CHECK: [[AND7:%[0-9]+]]:_(s32) = G_AND [[COPY24]], [[C1]] + ; CHECK: [[COPY25:%[0-9]+]]:_(s32) = COPY [[BITCAST13]](s32) + ; CHECK: [[AND8:%[0-9]+]]:_(s32) = G_AND [[COPY25]], [[C1]] + ; CHECK: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND8]], [[C]](s32) + ; CHECK: [[OR4:%[0-9]+]]:_(s32) = G_OR [[AND7]], [[SHL4]] + ; CHECK: [[BITCAST16:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32) + ; CHECK: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST14]](<2 x s16>), [[BITCAST15]](<2 x s16>), [[BITCAST16]](<2 x s16>) + ; CHECK: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 1 + ; CHECK: [[COPY26:%[0-9]+]]:_(s32) = COPY [[ANYEXT]](s32) + ; CHECK: [[AND9:%[0-9]+]]:_(s32) = G_AND [[COPY26]], [[C4]] + ; CHECK: [[COPY27:%[0-9]+]]:_(s32) = COPY [[ANYEXT1]](s32) + ; CHECK: [[AND10:%[0-9]+]]:_(s32) = G_AND [[COPY27]], [[C4]] + ; CHECK: [[COPY28:%[0-9]+]]:_(s32) = COPY [[ANYEXT2]](s32) + ; CHECK: [[AND11:%[0-9]+]]:_(s32) = G_AND [[COPY28]], [[C4]] + ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[AND9]](s32), [[AND10]](s32), [[AND11]](s32) + ; CHECK: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) + ; CHECK: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>) + %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + %1:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + %2:_(<3 x s16>), %3:_(<3 x s16>) = G_UNMERGE_VALUES %0 + %4:_(<3 x s16>), %5:_(<3 x s16>) = G_UNMERGE_VALUES %1 + %6:_(<3 x s16>), %7:_(<3 x s1>) = G_SSUBO %2, %4 + %8:_(<3 x s16>) = G_IMPLICIT_DEF + %9:_(<6 x s16>) = G_CONCAT_VECTORS %6, %8 + %10:_(<3 x s32>) = G_ZEXT %7 + $vgpr0_vgpr1_vgpr2 = COPY %9 + $vgpr0_vgpr1_vgpr2 = COPY %10 ... --- Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-uaddo.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-uaddo.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-uaddo.mir @@ -169,31 +169,30 @@ name: test_uaddo_v3s16 body: | bb.0: - liveins: $vgpr0_vgpr1, $vgpr2_vgpr3 - + liveins: $vgpr0_vgpr1_vgpr2, $vgpr3_vgpr4_vgpr5 ; CHECK-LABEL: name: test_uaddo_v3s16 - ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 - ; CHECK: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr1_vgpr2 - ; CHECK: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0 - ; CHECK: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0 + ; CHECK: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + ; CHECK: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + ; CHECK: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) + ; CHECK: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>) ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; CHECK: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; CHECK: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; CHECK: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; CHECK: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; CHECK: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CHECK: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; CHECK: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; CHECK: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; CHECK: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; CHECK: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; CHECK: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; CHECK: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0 - ; CHECK: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; CHECK: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) + ; CHECK: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0 + ; CHECK: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; CHECK: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>) ; CHECK: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; CHECK: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; CHECK: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) - ; CHECK: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; CHECK: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>) ; CHECK: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; CHECK: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; CHECK: [[ADD:%[0-9]+]]:_(s16) = G_ADD [[TRUNC]], [[TRUNC3]] @@ -210,38 +209,80 @@ ; CHECK: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; CHECK: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) - ; CHECK: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0 - ; CHECK: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) - ; CHECK: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) + ; CHECK: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 + ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0 + ; CHECK: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) + ; CHECK: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) + ; CHECK: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST6]](s32) ; CHECK: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32) - ; CHECK: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) + ; CHECK: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32) + ; CHECK: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) + ; CHECK: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST7]](s32) ; CHECK: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST7]], [[C]](s32) + ; CHECK: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(ult), [[ADD]](s16), [[TRUNC6]] + ; CHECK: [[ICMP1:%[0-9]+]]:_(s1) = G_ICMP intpred(ult), [[ADD1]](s16), [[TRUNC7]] + ; CHECK: [[ICMP2:%[0-9]+]]:_(s1) = G_ICMP intpred(ult), [[ADD2]](s16), [[TRUNC8]] + ; CHECK: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[ICMP]](s1) + ; CHECK: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[ICMP1]](s1) + ; CHECK: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[ICMP2]](s1) + ; CHECK: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; CHECK: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; CHECK: [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>), [[UV12:%[0-9]+]]:_(<3 x s16>), [[UV13:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; CHECK: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; CHECK: [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT3]](<4 x s16>) + ; CHECK: [[BITCAST8:%[0-9]+]]:_(s32) = G_BITCAST [[UV14]](<2 x s16>) + ; CHECK: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST8]], [[C]](s32) + ; CHECK: [[BITCAST9:%[0-9]+]]:_(s32) = G_BITCAST [[UV15]](<2 x s16>) + ; CHECK: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST9]], [[C]](s32) + ; CHECK: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV10]](<3 x s16>), 0 + ; CHECK: [[UV16:%[0-9]+]]:_(<2 x s16>), [[UV17:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT4]](<4 x s16>) + ; CHECK: [[BITCAST10:%[0-9]+]]:_(s32) = G_BITCAST [[UV16]](<2 x s16>) + ; CHECK: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST10]], [[C]](s32) + ; CHECK: [[BITCAST11:%[0-9]+]]:_(s32) = G_BITCAST [[UV17]](<2 x s16>) + ; CHECK: [[LSHR9:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST11]], [[C]](s32) ; CHECK: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535 - ; CHECK: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST6]](s32) + ; CHECK: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST8]](s32) ; CHECK: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C2]] - ; CHECK: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) + ; CHECK: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR6]](s32) ; CHECK: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY3]], [[C2]] - ; CHECK: [[COPY4:%[0-9]+]]:_(s32) = COPY [[BITCAST7]](s32) + ; CHECK: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32) + ; CHECK: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL2]] + ; CHECK: [[BITCAST12:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32) + ; CHECK: [[COPY4:%[0-9]+]]:_(s32) = COPY [[BITCAST9]](s32) ; CHECK: [[AND2:%[0-9]+]]:_(s32) = G_AND [[COPY4]], [[C2]] - ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[AND]](s32), [[AND1]](s32), [[AND2]](s32) - ; CHECK: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT2]](<3 x s16>), 0 - ; CHECK: [[DEF1:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF - ; CHECK: [[INSERT4:%[0-9]+]]:_(<4 x s32>) = G_INSERT [[DEF1]], [[BUILD_VECTOR]](<3 x s32>), 0 - ; CHECK: $vgpr0_vgpr1 = COPY [[INSERT3]](<4 x s16>) - ; CHECK: $vgpr2_vgpr3_vgpr4_vgpr5 = COPY [[INSERT4]](<4 x s32>) - %0:_(<4 x s16>) = COPY $vgpr0_vgpr1 - %1:_(<4 x s16>) = COPY $vgpr1_vgpr2 - %2:_(<3 x s16>) = G_EXTRACT %0(<4 x s16>), 0 - %3:_(<3 x s16>) = G_EXTRACT %0(<4 x s16>), 0 - %4:_(<3 x s16>), %5:_(<3 x s1>) = G_UADDO %2, %3 - %6:_(<3 x s32>) = G_ZEXT %3 - %7:_(<4 x s16>) = G_IMPLICIT_DEF - %8:_(<4 x s16>) = G_INSERT %7, %4(<3 x s16>), 0 - %9:_(<4 x s32>) = G_IMPLICIT_DEF - %10:_(<4 x s32>) = G_INSERT %9, %6(<3 x s32>), 0 - $vgpr0_vgpr1 = COPY %8 - $vgpr2_vgpr3_vgpr4_vgpr5 = COPY %10 + ; CHECK: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST10]](s32) + ; CHECK: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY5]], [[C2]] + ; CHECK: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32) + ; CHECK: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL3]] + ; CHECK: [[BITCAST13:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32) + ; CHECK: [[COPY6:%[0-9]+]]:_(s32) = COPY [[LSHR8]](s32) + ; CHECK: [[AND4:%[0-9]+]]:_(s32) = G_AND [[COPY6]], [[C2]] + ; CHECK: [[COPY7:%[0-9]+]]:_(s32) = COPY [[BITCAST11]](s32) + ; CHECK: [[AND5:%[0-9]+]]:_(s32) = G_AND [[COPY7]], [[C2]] + ; CHECK: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32) + ; CHECK: [[OR4:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL4]] + ; CHECK: [[BITCAST14:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32) + ; CHECK: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST12]](<2 x s16>), [[BITCAST13]](<2 x s16>), [[BITCAST14]](<2 x s16>) + ; CHECK: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 1 + ; CHECK: [[COPY8:%[0-9]+]]:_(s32) = COPY [[ANYEXT]](s32) + ; CHECK: [[AND6:%[0-9]+]]:_(s32) = G_AND [[COPY8]], [[C3]] + ; CHECK: [[COPY9:%[0-9]+]]:_(s32) = COPY [[ANYEXT1]](s32) + ; CHECK: [[AND7:%[0-9]+]]:_(s32) = G_AND [[COPY9]], [[C3]] + ; CHECK: [[COPY10:%[0-9]+]]:_(s32) = COPY [[ANYEXT2]](s32) + ; CHECK: [[AND8:%[0-9]+]]:_(s32) = G_AND [[COPY10]], [[C3]] + ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[AND6]](s32), [[AND7]](s32), [[AND8]](s32) + ; CHECK: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) + ; CHECK: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>) + %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + %1:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + %2:_(<3 x s16>), %3:_(<3 x s16>) = G_UNMERGE_VALUES %0 + %4:_(<3 x s16>), %5:_(<3 x s16>) = G_UNMERGE_VALUES %1 + %6:_(<3 x s16>), %7:_(<3 x s1>) = G_UADDO %2, %4 + %8:_(<3 x s16>) = G_IMPLICIT_DEF + %9:_(<6 x s16>) = G_CONCAT_VECTORS %6, %8 + %10:_(<3 x s32>) = G_ZEXT %7 + $vgpr0_vgpr1_vgpr2 = COPY %9 + $vgpr0_vgpr1_vgpr2 = COPY %10 ... --- Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-usubo.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-usubo.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-usubo.mir @@ -173,31 +173,30 @@ name: test_usubo_v3s16 body: | bb.0: - liveins: $vgpr0_vgpr1, $vgpr2_vgpr3 - + liveins: $vgpr0_vgpr1_vgpr2, $vgpr3_vgpr4_vgpr5 ; CHECK-LABEL: name: test_usubo_v3s16 - ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 - ; CHECK: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr1_vgpr2 - ; CHECK: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0 - ; CHECK: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0 + ; CHECK: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + ; CHECK: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + ; CHECK: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) + ; CHECK: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>) ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 - ; CHECK: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) - ; CHECK: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) + ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; CHECK: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>) + ; CHECK: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) ; CHECK: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 ; CHECK: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; CHECK: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) - ; CHECK: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) + ; CHECK: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) ; CHECK: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) ; CHECK: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) - ; CHECK: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0 - ; CHECK: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) - ; CHECK: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) + ; CHECK: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0 + ; CHECK: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>) + ; CHECK: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>) ; CHECK: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) ; CHECK: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) ; CHECK: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) - ; CHECK: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) + ; CHECK: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>) ; CHECK: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) ; CHECK: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) ; CHECK: [[SUB:%[0-9]+]]:_(s16) = G_SUB [[TRUNC]], [[TRUNC3]] @@ -214,38 +213,89 @@ ; CHECK: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] ; CHECK: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) - ; CHECK: [[EXTRACT2:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 - ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0 - ; CHECK: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) - ; CHECK: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) + ; CHECK: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0 + ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; CHECK: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) + ; CHECK: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>) + ; CHECK: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST6]](s32) ; CHECK: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32) - ; CHECK: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) + ; CHECK: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32) + ; CHECK: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>) + ; CHECK: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST7]](s32) ; CHECK: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST7]], [[C]](s32) + ; CHECK: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0 + ; CHECK: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT3]](<4 x s16>) + ; CHECK: [[BITCAST8:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>) + ; CHECK: [[TRUNC9:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST8]](s32) + ; CHECK: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST8]], [[C]](s32) + ; CHECK: [[TRUNC10:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR6]](s32) + ; CHECK: [[BITCAST9:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>) + ; CHECK: [[TRUNC11:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST9]](s32) + ; CHECK: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST9]], [[C]](s32) + ; CHECK: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(ult), [[TRUNC6]](s16), [[TRUNC9]] + ; CHECK: [[ICMP1:%[0-9]+]]:_(s1) = G_ICMP intpred(ult), [[TRUNC7]](s16), [[TRUNC10]] + ; CHECK: [[ICMP2:%[0-9]+]]:_(s1) = G_ICMP intpred(ult), [[TRUNC8]](s16), [[TRUNC11]] + ; CHECK: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[ICMP]](s1) + ; CHECK: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[ICMP1]](s1) + ; CHECK: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[ICMP2]](s1) + ; CHECK: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; CHECK: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; CHECK: [[UV12:%[0-9]+]]:_(<3 x s16>), [[UV13:%[0-9]+]]:_(<3 x s16>), [[UV14:%[0-9]+]]:_(<3 x s16>), [[UV15:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; CHECK: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0 + ; CHECK: [[UV16:%[0-9]+]]:_(<2 x s16>), [[UV17:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT4]](<4 x s16>) + ; CHECK: [[BITCAST10:%[0-9]+]]:_(s32) = G_BITCAST [[UV16]](<2 x s16>) + ; CHECK: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST10]], [[C]](s32) + ; CHECK: [[BITCAST11:%[0-9]+]]:_(s32) = G_BITCAST [[UV17]](<2 x s16>) + ; CHECK: [[LSHR9:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST11]], [[C]](s32) + ; CHECK: [[INSERT5:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV12]](<3 x s16>), 0 + ; CHECK: [[UV18:%[0-9]+]]:_(<2 x s16>), [[UV19:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT5]](<4 x s16>) + ; CHECK: [[BITCAST12:%[0-9]+]]:_(s32) = G_BITCAST [[UV18]](<2 x s16>) + ; CHECK: [[LSHR10:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST12]], [[C]](s32) + ; CHECK: [[BITCAST13:%[0-9]+]]:_(s32) = G_BITCAST [[UV19]](<2 x s16>) + ; CHECK: [[LSHR11:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST13]], [[C]](s32) ; CHECK: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535 - ; CHECK: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST6]](s32) + ; CHECK: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST10]](s32) ; CHECK: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C2]] - ; CHECK: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) + ; CHECK: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR8]](s32) ; CHECK: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY3]], [[C2]] - ; CHECK: [[COPY4:%[0-9]+]]:_(s32) = COPY [[BITCAST7]](s32) + ; CHECK: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32) + ; CHECK: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL2]] + ; CHECK: [[BITCAST14:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32) + ; CHECK: [[COPY4:%[0-9]+]]:_(s32) = COPY [[BITCAST11]](s32) ; CHECK: [[AND2:%[0-9]+]]:_(s32) = G_AND [[COPY4]], [[C2]] - ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[AND]](s32), [[AND1]](s32), [[AND2]](s32) - ; CHECK: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT2]](<3 x s16>), 0 - ; CHECK: [[DEF1:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF - ; CHECK: [[INSERT4:%[0-9]+]]:_(<4 x s32>) = G_INSERT [[DEF1]], [[BUILD_VECTOR]](<3 x s32>), 0 - ; CHECK: $vgpr0_vgpr1 = COPY [[INSERT3]](<4 x s16>) - ; CHECK: $vgpr2_vgpr3_vgpr4_vgpr5 = COPY [[INSERT4]](<4 x s32>) - %0:_(<4 x s16>) = COPY $vgpr0_vgpr1 - %1:_(<4 x s16>) = COPY $vgpr1_vgpr2 - %2:_(<3 x s16>) = G_EXTRACT %0(<4 x s16>), 0 - %3:_(<3 x s16>) = G_EXTRACT %0(<4 x s16>), 0 - %4:_(<3 x s16>), %5:_(<3 x s1>) = G_USUBO %2, %3 - %6:_(<3 x s32>) = G_ZEXT %3 - %7:_(<4 x s16>) = G_IMPLICIT_DEF - %8:_(<4 x s16>) = G_INSERT %7, %4(<3 x s16>), 0 - %9:_(<4 x s32>) = G_IMPLICIT_DEF - %10:_(<4 x s32>) = G_INSERT %9, %6(<3 x s32>), 0 - $vgpr0_vgpr1 = COPY %8 - $vgpr2_vgpr3_vgpr4_vgpr5 = COPY %10 + ; CHECK: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST12]](s32) + ; CHECK: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY5]], [[C2]] + ; CHECK: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32) + ; CHECK: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL3]] + ; CHECK: [[BITCAST15:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32) + ; CHECK: [[COPY6:%[0-9]+]]:_(s32) = COPY [[LSHR10]](s32) + ; CHECK: [[AND4:%[0-9]+]]:_(s32) = G_AND [[COPY6]], [[C2]] + ; CHECK: [[COPY7:%[0-9]+]]:_(s32) = COPY [[BITCAST13]](s32) + ; CHECK: [[AND5:%[0-9]+]]:_(s32) = G_AND [[COPY7]], [[C2]] + ; CHECK: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32) + ; CHECK: [[OR4:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL4]] + ; CHECK: [[BITCAST16:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32) + ; CHECK: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST14]](<2 x s16>), [[BITCAST15]](<2 x s16>), [[BITCAST16]](<2 x s16>) + ; CHECK: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 1 + ; CHECK: [[COPY8:%[0-9]+]]:_(s32) = COPY [[ANYEXT]](s32) + ; CHECK: [[AND6:%[0-9]+]]:_(s32) = G_AND [[COPY8]], [[C3]] + ; CHECK: [[COPY9:%[0-9]+]]:_(s32) = COPY [[ANYEXT1]](s32) + ; CHECK: [[AND7:%[0-9]+]]:_(s32) = G_AND [[COPY9]], [[C3]] + ; CHECK: [[COPY10:%[0-9]+]]:_(s32) = COPY [[ANYEXT2]](s32) + ; CHECK: [[AND8:%[0-9]+]]:_(s32) = G_AND [[COPY10]], [[C3]] + ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[AND6]](s32), [[AND7]](s32), [[AND8]](s32) + ; CHECK: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) + ; CHECK: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>) + %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + %1:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + %2:_(<3 x s16>), %3:_(<3 x s16>) = G_UNMERGE_VALUES %0 + %4:_(<3 x s16>), %5:_(<3 x s16>) = G_UNMERGE_VALUES %1 + %6:_(<3 x s16>), %7:_(<3 x s1>) = G_USUBO %2, %4 + %8:_(<3 x s16>) = G_IMPLICIT_DEF + %9:_(<6 x s16>) = G_CONCAT_VECTORS %6, %8 + %10:_(<3 x s32>) = G_ZEXT %7 + $vgpr0_vgpr1_vgpr2 = COPY %9 + $vgpr0_vgpr1_vgpr2 = COPY %10 ... --- Index: llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-xor.mir =================================================================== --- llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-xor.mir +++ llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-xor.mir @@ -435,28 +435,66 @@ name: test_xor_v3s16 body: | bb.0: - + liveins: $vgpr0_vgpr1_vgpr2, $vgpr3_vgpr4_vgpr5 ; CHECK-LABEL: name: test_xor_v3s16 + ; CHECK: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + ; CHECK: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + ; CHECK: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) + ; CHECK: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>) ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) - ; CHECK: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) - ; CHECK: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF - ; CHECK: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) - ; CHECK: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) - ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0 - ; CHECK: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0 + ; CHECK: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0 + ; CHECK: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0 ; CHECK: [[XOR:%[0-9]+]]:_(<4 x s16>) = G_XOR [[INSERT]], [[INSERT1]] - ; CHECK: [[CONCAT_VECTORS2:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[XOR]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>) - ; CHECK: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS2]](<12 x s16>) - ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV8]](<3 x s16>), 0 - ; CHECK: $vgpr0_vgpr1 = COPY [[INSERT2]](<4 x s16>) - %0:_(<3 x s16>) = G_IMPLICIT_DEF - %1:_(<3 x s16>) = G_IMPLICIT_DEF - %2:_(<3 x s16>) = G_XOR %0, %1 - %4:_(<4 x s16>) = G_IMPLICIT_DEF - %5:_(<4 x s16>) = G_INSERT %4, %2, 0 - $vgpr0_vgpr1 = COPY %5 + ; CHECK: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[XOR]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; CHECK: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>) + ; CHECK: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF + ; CHECK: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>) + ; CHECK: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>), [[UV10:%[0-9]+]]:_(<3 x s16>), [[UV11:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>) + ; CHECK: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV4]](<3 x s16>), 0 + ; CHECK: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>) + ; CHECK: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV12]](<2 x s16>) + ; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 + ; CHECK: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) + ; CHECK: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV13]](<2 x s16>) + ; CHECK: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) + ; CHECK: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV8]](<3 x s16>), 0 + ; CHECK: [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT3]](<4 x s16>) + ; CHECK: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV14]](<2 x s16>) + ; CHECK: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) + ; CHECK: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV15]](<2 x s16>) + ; CHECK: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) + ; CHECK: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535 + ; CHECK: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) + ; CHECK: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C1]] + ; CHECK: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) + ; CHECK: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY3]], [[C1]] + ; CHECK: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32) + ; CHECK: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]] + ; CHECK: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) + ; CHECK: [[COPY4:%[0-9]+]]:_(s32) = COPY [[BITCAST1]](s32) + ; CHECK: [[AND2:%[0-9]+]]:_(s32) = G_AND [[COPY4]], [[C1]] + ; CHECK: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST2]](s32) + ; CHECK: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY5]], [[C1]] + ; CHECK: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32) + ; CHECK: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]] + ; CHECK: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) + ; CHECK: [[COPY6:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) + ; CHECK: [[AND4:%[0-9]+]]:_(s32) = G_AND [[COPY6]], [[C1]] + ; CHECK: [[COPY7:%[0-9]+]]:_(s32) = COPY [[BITCAST3]](s32) + ; CHECK: [[AND5:%[0-9]+]]:_(s32) = G_AND [[COPY7]], [[C1]] + ; CHECK: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32) + ; CHECK: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]] + ; CHECK: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32) + ; CHECK: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>) + ; CHECK: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>) + %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 + %1:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5 + %2:_(<3 x s16>), %3:_(<3 x s16>) = G_UNMERGE_VALUES %0 + %4:_(<3 x s16>), %5:_(<3 x s16>) = G_UNMERGE_VALUES %1 + %6:_(<3 x s16>) = G_XOR %2, %4 + %7:_(<3 x s16>) = G_IMPLICIT_DEF + %8:_(<6 x s16>) = G_CONCAT_VECTORS %6, %7 + $vgpr0_vgpr1_vgpr2 = COPY %8 ... ---