Skip to content

Commit 16a639a

Browse files
krzysz00easyonaadit
krzysz00
authored and
easyonaadit
committed
Reapply "[AMDGPU][GlobalISel] Fix load/store of pointer vectors, buffer.*.pN (llvm#110714)" v2 (llvm#111708)
This adds `-disable-gisel-legality-check` to some gfx6 and gfx7 test lines to prevent behavior mismatches between debug and release builds The first attempted reapply was llvm#111059 This reverts commit e075dcf.
1 parent 910e562 commit 16a639a

File tree

2 files changed

+95
-72
lines changed

2 files changed

+95
-72
lines changed

‎llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-load-global-old-legalization.mir‎

Lines changed: 65 additions & 44 deletions
Original file line numberDiff line numberDiff line change
@@ -1,7 +1,7 @@
11
# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
2-
# RUN: llc -mtriple=amdgcn -mcpu=tahiti -run-pass=instruction-select -verify-machineinstrs -global-isel-abort=0 -o - %s | FileCheck -check-prefix=GFX6 %s
3-
# RUN: llc -mtriple=amdgcn -mcpu=hawaii -run-pass=instruction-select -verify-machineinstrs -global-isel-abort=0 -o - %s | FileCheck -check-prefix=GFX7 %s
4-
# RUN: llc -mtriple=amdgcn -mcpu=hawaii -mattr=+flat-for-global -run-pass=instruction-select -verify-machineinstrs -global-isel-abort=0 -o - %s | FileCheck -check-prefix=GFX7-FLAT %s
2+
# RUN: llc -mtriple=amdgcn -mcpu=tahiti -run-pass=instruction-select -verify-machineinstrs -global-isel-abort=0 -disable-gisel-legality-check -o - %s | FileCheck -check-prefix=GFX6 %s
3+
# RUN: llc -mtriple=amdgcn -mcpu=hawaii -run-pass=instruction-select -verify-machineinstrs -global-isel-abort=0 -disable-gisel-legality-check -o - %s | FileCheck -check-prefix=GFX7 %s
4+
# RUN: llc -mtriple=amdgcn -mcpu=hawaii -mattr=+flat-for-global -run-pass=instruction-select -verify-machineinstrs -global-isel-abort=0 -disable-gisel-legality-check -o - %s | FileCheck -check-prefix=GFX7-FLAT %s
55
# RUN: llc -mtriple=amdgcn -mcpu=fiji -run-pass=instruction-select -verify-machineinstrs -global-isel-abort=0 -o - %s | FileCheck -check-prefix=GFX8 %s
66
# RUN: llc -mtriple=amdgcn -mcpu=gfx900 -run-pass=instruction-select -verify-machineinstrs -global-isel-abort=0 -o - %s | FileCheck -check-prefix=GFX9 %s
77
# RUN: llc -mtriple=amdgcn -mcpu=gfx1010 -run-pass=instruction-select -verify-machineinstrs -global-isel-abort=0 -o - %s | FileCheck -check-prefix=GFX10 %s
@@ -598,61 +598,72 @@ body: |
598598
; GFX6-LABEL: name: load_global_v2p1
599599
; GFX6: liveins: $vgpr0_vgpr1
600600
; GFX6-NEXT: {{ $}}
601-
; GFX6-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
602-
; GFX6-NEXT: [[LOAD:%[0-9]+]]:vgpr(<2 x p1>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 4, addrspace 1)
603-
; GFX6-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x p1>)
601+
; GFX6-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
602+
; GFX6-NEXT: [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 0
603+
; GFX6-NEXT: [[S_MOV_B32_1:%[0-9]+]]:sreg_32 = S_MOV_B32 61440
604+
; GFX6-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sreg_64 = REG_SEQUENCE [[S_MOV_B32_]], %subreg.sub0, [[S_MOV_B32_1]], %subreg.sub1
605+
; GFX6-NEXT: [[S_MOV_B64_:%[0-9]+]]:sreg_64 = S_MOV_B64 0
606+
; GFX6-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[S_MOV_B64_]], %subreg.sub0_sub1, [[REG_SEQUENCE]], %subreg.sub2_sub3
607+
; GFX6-NEXT: [[BUFFER_LOAD_DWORDX4_ADDR64_:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_ADDR64 [[COPY]], [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec :: (load (<2 x s64>), align 4, addrspace 1)
608+
; GFX6-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUFFER_LOAD_DWORDX4_ADDR64_]]
604609
;
605610
; GFX7-LABEL: name: load_global_v2p1
606611
; GFX7: liveins: $vgpr0_vgpr1
607612
; GFX7-NEXT: {{ $}}
608-
; GFX7-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
609-
; GFX7-NEXT: [[LOAD:%[0-9]+]]:vgpr(<2 x p1>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 4, addrspace 1)
610-
; GFX7-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x p1>)
613+
; GFX7-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
614+
; GFX7-NEXT: [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 0
615+
; GFX7-NEXT: [[S_MOV_B32_1:%[0-9]+]]:sreg_32 = S_MOV_B32 61440
616+
; GFX7-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sreg_64 = REG_SEQUENCE [[S_MOV_B32_]], %subreg.sub0, [[S_MOV_B32_1]], %subreg.sub1
617+
; GFX7-NEXT: [[S_MOV_B64_:%[0-9]+]]:sreg_64 = S_MOV_B64 0
618+
; GFX7-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[S_MOV_B64_]], %subreg.sub0_sub1, [[REG_SEQUENCE]], %subreg.sub2_sub3
619+
; GFX7-NEXT: [[BUFFER_LOAD_DWORDX4_ADDR64_:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_ADDR64 [[COPY]], [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec :: (load (<2 x s64>), align 4, addrspace 1)
620+
; GFX7-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUFFER_LOAD_DWORDX4_ADDR64_]]
611621
;
612622
; GFX7-FLAT-LABEL: name: load_global_v2p1
613623
; GFX7-FLAT: liveins: $vgpr0_vgpr1
614624
; GFX7-FLAT-NEXT: {{ $}}
615-
; GFX7-FLAT-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
616-
; GFX7-FLAT-NEXT: [[LOAD:%[0-9]+]]:vgpr(<2 x p1>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 4, addrspace 1)
617-
; GFX7-FLAT-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x p1>)
625+
; GFX7-FLAT-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
626+
; GFX7-FLAT-NEXT: [[FLAT_LOAD_DWORDX4_:%[0-9]+]]:vreg_128 = FLAT_LOAD_DWORDX4 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<2 x s64>), align 4, addrspace 1)
627+
; GFX7-FLAT-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[FLAT_LOAD_DWORDX4_]]
618628
;
619629
; GFX8-LABEL: name: load_global_v2p1
620630
; GFX8: liveins: $vgpr0_vgpr1
621631
; GFX8-NEXT: {{ $}}
622-
; GFX8-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
623-
; GFX8-NEXT: [[LOAD:%[0-9]+]]:vgpr(<2 x p1>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 4, addrspace 1)
624-
; GFX8-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x p1>)
632+
; GFX8-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
633+
; GFX8-NEXT: [[FLAT_LOAD_DWORDX4_:%[0-9]+]]:vreg_128 = FLAT_LOAD_DWORDX4 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<2 x s64>), align 4, addrspace 1)
634+
; GFX8-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[FLAT_LOAD_DWORDX4_]]
625635
;
626636
; GFX9-LABEL: name: load_global_v2p1
627637
; GFX9: liveins: $vgpr0_vgpr1
628638
; GFX9-NEXT: {{ $}}
629-
; GFX9-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
630-
; GFX9-NEXT: [[LOAD:%[0-9]+]]:vgpr(<2 x p1>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 4, addrspace 1)
631-
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x p1>)
639+
; GFX9-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
640+
; GFX9-NEXT: [[GLOBAL_LOAD_DWORDX4_:%[0-9]+]]:vreg_128 = GLOBAL_LOAD_DWORDX4 [[COPY]], 0, 0, implicit $exec :: (load (<2 x s64>), align 4, addrspace 1)
641+
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[GLOBAL_LOAD_DWORDX4_]]
632642
;
633643
; GFX10-LABEL: name: load_global_v2p1
634644
; GFX10: liveins: $vgpr0_vgpr1
635645
; GFX10-NEXT: {{ $}}
636-
; GFX10-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
637-
; GFX10-NEXT: [[LOAD:%[0-9]+]]:vgpr(<2 x p1>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 4, addrspace 1)
638-
; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x p1>)
646+
; GFX10-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
647+
; GFX10-NEXT: [[GLOBAL_LOAD_DWORDX4_:%[0-9]+]]:vreg_128 = GLOBAL_LOAD_DWORDX4 [[COPY]], 0, 0, implicit $exec :: (load (<2 x s64>), align 4, addrspace 1)
648+
; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[GLOBAL_LOAD_DWORDX4_]]
639649
;
640650
; GFX11-LABEL: name: load_global_v2p1
641651
; GFX11: liveins: $vgpr0_vgpr1
642652
; GFX11-NEXT: {{ $}}
643-
; GFX11-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
644-
; GFX11-NEXT: [[LOAD:%[0-9]+]]:vgpr(<2 x p1>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 4, addrspace 1)
645-
; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x p1>)
653+
; GFX11-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
654+
; GFX11-NEXT: [[GLOBAL_LOAD_DWORDX4_:%[0-9]+]]:vreg_128 = GLOBAL_LOAD_DWORDX4 [[COPY]], 0, 0, implicit $exec :: (load (<2 x s64>), align 4, addrspace 1)
655+
; GFX11-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[GLOBAL_LOAD_DWORDX4_]]
646656
;
647657
; GFX12-LABEL: name: load_global_v2p1
648658
; GFX12: liveins: $vgpr0_vgpr1
649659
; GFX12-NEXT: {{ $}}
650-
; GFX12-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
651-
; GFX12-NEXT: [[LOAD:%[0-9]+]]:vgpr(<2 x p1>) = G_LOAD [[COPY]](p1) :: (load (<2 x p1>), align 4, addrspace 1)
652-
; GFX12-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<2 x p1>)
660+
; GFX12-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
661+
; GFX12-NEXT: [[GLOBAL_LOAD_DWORDX4_:%[0-9]+]]:vreg_128 = GLOBAL_LOAD_DWORDX4 [[COPY]], 0, 0, implicit $exec :: (load (<2 x s64>), align 4, addrspace 1)
662+
; GFX12-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[GLOBAL_LOAD_DWORDX4_]]
653663
%0:vgpr(p1) = COPY $vgpr0_vgpr1
654-
%1:vgpr(<2 x p1>) = G_LOAD %0 :: (load (<2 x p1>), align 4, addrspace 1)
655-
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %1
664+
%1:vgpr(<2 x s64>) = G_LOAD %0 :: (load (<2 x s64>), align 4, addrspace 1)
665+
%2:vgpr(<2 x p1>) = G_BITCAST %1(<2 x s64>)
666+
$vgpr0_vgpr1_vgpr2_vgpr3 = COPY %2
656667
657668
...
658669

@@ -671,21 +682,21 @@ body: |
671682
; GFX6: liveins: $vgpr0_vgpr1
672683
; GFX6-NEXT: {{ $}}
673684
; GFX6-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
674-
; GFX6-NEXT: [[LOAD:%[0-9]+]]:vgpr(s128) = G_LOAD [[COPY]](p1) :: (load (s128), align 4, addrspace 1)
685+
; GFX6-NEXT: [[LOAD:%[0-9]+]]:vreg_128(s128) = G_LOAD [[COPY]](p1) :: (load (s128), align 4, addrspace 1)
675686
; GFX6-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](s128)
676687
;
677688
; GFX7-LABEL: name: load_global_s128
678689
; GFX7: liveins: $vgpr0_vgpr1
679690
; GFX7-NEXT: {{ $}}
680691
; GFX7-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
681-
; GFX7-NEXT: [[LOAD:%[0-9]+]]:vgpr(s128) = G_LOAD [[COPY]](p1) :: (load (s128), align 4, addrspace 1)
692+
; GFX7-NEXT: [[LOAD:%[0-9]+]]:vreg_128(s128) = G_LOAD [[COPY]](p1) :: (load (s128), align 4, addrspace 1)
682693
; GFX7-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](s128)
683694
;
684695
; GFX7-FLAT-LABEL: name: load_global_s128
685696
; GFX7-FLAT: liveins: $vgpr0_vgpr1
686697
; GFX7-FLAT-NEXT: {{ $}}
687698
; GFX7-FLAT-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
688-
; GFX7-FLAT-NEXT: [[LOAD:%[0-9]+]]:vgpr(s128) = G_LOAD [[COPY]](p1) :: (load (s128), align 4, addrspace 1)
699+
; GFX7-FLAT-NEXT: [[LOAD:%[0-9]+]]:vreg_128(s128) = G_LOAD [[COPY]](p1) :: (load (s128), align 4, addrspace 1)
689700
; GFX7-FLAT-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](s128)
690701
;
691702
; GFX8-LABEL: name: load_global_s128
@@ -979,21 +990,21 @@ body: |
979990
; GFX6: liveins: $vgpr0_vgpr1
980991
; GFX6-NEXT: {{ $}}
981992
; GFX6-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
982-
; GFX6-NEXT: [[LOAD:%[0-9]+]]:vgpr(<2 x p3>) = G_LOAD [[COPY]](p1) :: (load (<2 x p3>), addrspace 1)
993+
; GFX6-NEXT: [[LOAD:%[0-9]+]]:vreg_64(<2 x p3>) = G_LOAD [[COPY]](p1) :: (load (<2 x p3>), addrspace 1)
983994
; GFX6-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x p3>)
984995
;
985996
; GFX7-LABEL: name: load_global_v2p3
986997
; GFX7: liveins: $vgpr0_vgpr1
987998
; GFX7-NEXT: {{ $}}
988999
; GFX7-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
989-
; GFX7-NEXT: [[LOAD:%[0-9]+]]:vgpr(<2 x p3>) = G_LOAD [[COPY]](p1) :: (load (<2 x p3>), addrspace 1)
1000+
; GFX7-NEXT: [[LOAD:%[0-9]+]]:vreg_64(<2 x p3>) = G_LOAD [[COPY]](p1) :: (load (<2 x p3>), addrspace 1)
9901001
; GFX7-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x p3>)
9911002
;
9921003
; GFX7-FLAT-LABEL: name: load_global_v2p3
9931004
; GFX7-FLAT: liveins: $vgpr0_vgpr1
9941005
; GFX7-FLAT-NEXT: {{ $}}
9951006
; GFX7-FLAT-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
996-
; GFX7-FLAT-NEXT: [[LOAD:%[0-9]+]]:vgpr(<2 x p3>) = G_LOAD [[COPY]](p1) :: (load (<2 x p3>), addrspace 1)
1007+
; GFX7-FLAT-NEXT: [[LOAD:%[0-9]+]]:vreg_64(<2 x p3>) = G_LOAD [[COPY]](p1) :: (load (<2 x p3>), addrspace 1)
9971008
; GFX7-FLAT-NEXT: $vgpr0_vgpr1 = COPY [[LOAD]](<2 x p3>)
9981009
;
9991010
; GFX8-LABEL: name: load_global_v2p3
@@ -1214,23 +1225,33 @@ body: |
12141225
; GFX6-LABEL: name: load_global_v8s16
12151226
; GFX6: liveins: $vgpr0_vgpr1
12161227
; GFX6-NEXT: {{ $}}
1217-
; GFX6-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
1218-
; GFX6-NEXT: [[LOAD:%[0-9]+]]:vgpr(<8 x s16>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), align 4, addrspace 1)
1219-
; GFX6-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<8 x s16>)
1228+
; GFX6-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
1229+
; GFX6-NEXT: [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 0
1230+
; GFX6-NEXT: [[S_MOV_B32_1:%[0-9]+]]:sreg_32 = S_MOV_B32 61440
1231+
; GFX6-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sreg_64 = REG_SEQUENCE [[S_MOV_B32_]], %subreg.sub0, [[S_MOV_B32_1]], %subreg.sub1
1232+
; GFX6-NEXT: [[S_MOV_B64_:%[0-9]+]]:sreg_64 = S_MOV_B64 0
1233+
; GFX6-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[S_MOV_B64_]], %subreg.sub0_sub1, [[REG_SEQUENCE]], %subreg.sub2_sub3
1234+
; GFX6-NEXT: [[BUFFER_LOAD_DWORDX4_ADDR64_:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_ADDR64 [[COPY]], [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec :: (load (<8 x s16>), align 4, addrspace 1)
1235+
; GFX6-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUFFER_LOAD_DWORDX4_ADDR64_]]
12201236
;
12211237
; GFX7-LABEL: name: load_global_v8s16
12221238
; GFX7: liveins: $vgpr0_vgpr1
12231239
; GFX7-NEXT: {{ $}}
1224-
; GFX7-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
1225-
; GFX7-NEXT: [[LOAD:%[0-9]+]]:vgpr(<8 x s16>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), align 4, addrspace 1)
1226-
; GFX7-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<8 x s16>)
1240+
; GFX7-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
1241+
; GFX7-NEXT: [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 0
1242+
; GFX7-NEXT: [[S_MOV_B32_1:%[0-9]+]]:sreg_32 = S_MOV_B32 61440
1243+
; GFX7-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sreg_64 = REG_SEQUENCE [[S_MOV_B32_]], %subreg.sub0, [[S_MOV_B32_1]], %subreg.sub1
1244+
; GFX7-NEXT: [[S_MOV_B64_:%[0-9]+]]:sreg_64 = S_MOV_B64 0
1245+
; GFX7-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[S_MOV_B64_]], %subreg.sub0_sub1, [[REG_SEQUENCE]], %subreg.sub2_sub3
1246+
; GFX7-NEXT: [[BUFFER_LOAD_DWORDX4_ADDR64_:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_ADDR64 [[COPY]], [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec :: (load (<8 x s16>), align 4, addrspace 1)
1247+
; GFX7-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUFFER_LOAD_DWORDX4_ADDR64_]]
12271248
;
12281249
; GFX7-FLAT-LABEL: name: load_global_v8s16
12291250
; GFX7-FLAT: liveins: $vgpr0_vgpr1
12301251
; GFX7-FLAT-NEXT: {{ $}}
1231-
; GFX7-FLAT-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
1232-
; GFX7-FLAT-NEXT: [[LOAD:%[0-9]+]]:vgpr(<8 x s16>) = G_LOAD [[COPY]](p1) :: (load (<8 x s16>), align 4, addrspace 1)
1233-
; GFX7-FLAT-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[LOAD]](<8 x s16>)
1252+
; GFX7-FLAT-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
1253+
; GFX7-FLAT-NEXT: [[FLAT_LOAD_DWORDX4_:%[0-9]+]]:vreg_128 = FLAT_LOAD_DWORDX4 [[COPY]], 0, 0, implicit $exec, implicit $flat_scr :: (load (<8 x s16>), align 4, addrspace 1)
1254+
; GFX7-FLAT-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[FLAT_LOAD_DWORDX4_]]
12341255
;
12351256
; GFX8-LABEL: name: load_global_v8s16
12361257
; GFX8: liveins: $vgpr0_vgpr1

0 commit comments

Comments
 (0)