maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity

AMDGPU/GlobalISel: Fix flat load/store of pointer types 

llvm-svn: 367513
This commit is contained in:
Matt Arsenault 2019-08-01 03:57:42 +00:00
parent 57495268ac
commit 9952f46407
8 changed files with 117 additions and 104 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
llvm/lib/Target/AMDGPU/AMDGPU.td
View File

@ -10,6 +10,15 @@ include "llvm/TableGen/SearchableTable.td"
include "llvm/Target/Target.td"
include "AMDGPUFeatures.td"
def p0 : PtrValueType<i64, 0>;
def p1 : PtrValueType<i64, 1>;
def p2 : PtrValueType<i32, 2>;
def p3 : PtrValueType<i32, 3>;
def p4 : PtrValueType<i64, 4>;
def p5 : PtrValueType<i32, 5>;
def p6 : PtrValueType<i32, 6>;
class BoolToList<bit Value> {
list<int> ret = !if(Value, [1]<int>, []<int>);
}

4
llvm/lib/Target/AMDGPU/AMDGPUGISel.td
View File

@ -12,10 +12,6 @@
include "AMDGPU.td"
def p0 : PtrValueType<i64, 0>;
def p1 : PtrValueType<i64, 1>;
def p4 : PtrValueType<i64, 4>;
def sd_vsrc0 : ComplexPattern<i32, 1, "">;
def gi_vsrc0 :
GIComplexOperandMatcher<s32, "selectVSRC0">,

4
llvm/lib/Target/AMDGPU/FLATInstructions.td
View File

@ -786,7 +786,7 @@ def : FlatLoadAtomicPat <FLAT_LOAD_DWORDX2, atomic_load_64_flat, i64>;
def : FlatStorePat <FLAT_STORE_BYTE, truncstorei8_flat, i32>;
def : FlatStorePat <FLAT_STORE_SHORT, truncstorei16_flat, i32>;
foreach vt = [i32, f32, v2i16, v2f16] in {
foreach vt = [i32, f32, v2i16, v2f16, p2, p3, p5, p6] in {
def : FlatLoadPat <FLAT_LOAD_DWORD, load_flat, vt>;
def : FlatStorePat <FLAT_STORE_DWORD, store_flat, vt>;
}
@ -867,7 +867,7 @@ def : FlatLoadSignedPat <GLOBAL_LOAD_USHORT, zextloadi16_global, i32>;
def : FlatLoadSignedPat <GLOBAL_LOAD_SSHORT, sextloadi16_global, i32>;
def : FlatLoadSignedPat <GLOBAL_LOAD_USHORT, load_global, i16>;
foreach vt = [i32, f32, v2i16, v2f16] in {
foreach vt = [i32, f32, v2i16, v2f16, p2, p3, p5, p6] in {
def : FlatLoadSignedPat <GLOBAL_LOAD_DWORD, load_global, vt>;
def : FlatStoreSignedPat <GLOBAL_STORE_DWORD, store_global, vt, VGPR_32>;
}

4
llvm/lib/Target/AMDGPU/SIRegisterInfo.td
View File

@ -352,7 +352,7 @@ def TTMP0_TTMP1_TTMP2_TTMP3_TTMP4_TTMP5_TTMP6_TTMP7_TTMP8_TTMP9_TTMP10_TTMP11_TT
// VGPR 32-bit registers
// i16/f16 only on VI+
def VGPR_32 : RegisterClass<"AMDGPU", [i32, f32, i16, f16, v2i16, v2f16], 32,
def VGPR_32 : RegisterClass<"AMDGPU", [i32, f32, i16, f16, v2i16, v2f16, p2, p3, p5, p6], 32,
(add (sequence "VGPR%u", 0, 255))> {
let AllocationPriority = 1;
let Size = 32;
@ -586,7 +586,7 @@ def SReg_1024 : RegisterClass<"AMDGPU", [v32i32, v32f32], 32,
}
// Register class for all vector registers (VGPRs + Interploation Registers)
def VReg_64 : RegisterClass<"AMDGPU", [i64, f64, v2i32, v2f32, v4f16, v4i16], 32,
def VReg_64 : RegisterClass<"AMDGPU", [i64, f64, v2i32, v2f32, v4f16, v4i16, p0, p1, p4], 32,
(add VGPR_64)> {
let Size = 64;

50
llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-load-flat.mir
View File

@ -428,24 +428,25 @@ body: |
; GFX7-LABEL: name: load_flat_p3_from_4
; GFX7: liveins: $vgpr0_vgpr1
; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
; GFX7: [[LOAD:%[0-9]+]]:vgpr_32(p3) = G_LOAD [[COPY]](p1) :: (load 4)
; GFX7: $vgpr0 = COPY [[LOAD]](p3)
; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX7: [[FLAT_LOAD_DWORD:%[0-9]+]]:vgpr_32 = FLAT_LOAD_DWORD [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 4)
; GFX7: $vgpr0 = COPY [[FLAT_LOAD_DWORD]]
; GFX8-LABEL: name: load_flat_p3_from_4
; GFX8: liveins: $vgpr0_vgpr1
; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
; GFX8: [[LOAD:%[0-9]+]]:vgpr_32(p3) = G_LOAD [[COPY]](p1) :: (load 4)
; GFX8: $vgpr0 = COPY [[LOAD]](p3)
; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX8: [[FLAT_LOAD_DWORD:%[0-9]+]]:vgpr_32 = FLAT_LOAD_DWORD [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 4)
; GFX8: $vgpr0 = COPY [[FLAT_LOAD_DWORD]]
; GFX9-LABEL: name: load_flat_p3_from_4
; GFX9: liveins: $vgpr0_vgpr1
; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:vgpr_32(p3) = G_LOAD [[COPY]](p1) :: (load 4)
; GFX9: $vgpr0 = COPY [[LOAD]](p3)
; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX9: [[FLAT_LOAD_DWORD:%[0-9]+]]:vgpr_32 = FLAT_LOAD_DWORD [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 4)
; GFX9: $vgpr0 = COPY [[FLAT_LOAD_DWORD]]
; GFX10-LABEL: name: load_flat_p3_from_4
; GFX10: liveins: $vgpr0_vgpr1
; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
; GFX10: [[LOAD:%[0-9]+]]:vgpr_32(p3) = G_LOAD [[COPY]](p1) :: (load 4)
; GFX10: $vgpr0 = COPY [[LOAD]](p3)
; GFX10: $vcc_hi = IMPLICIT_DEF
; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX10: [[FLAT_LOAD_DWORD:%[0-9]+]]:vgpr_32 = FLAT_LOAD_DWORD [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 4)
; GFX10: $vgpr0 = COPY [[FLAT_LOAD_DWORD]]
%0:vgpr(p1) = COPY $vgpr0_vgpr1
%1:vgpr(p3) = G_LOAD %0 :: (load 4, align 4, addrspace 0)
$vgpr0 = COPY %1
@ -465,24 +466,25 @@ body: |
; GFX7-LABEL: name: load_flat_p1_from_8
; GFX7: liveins: $vgpr0_vgpr1
; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
; GFX7: [[LOAD:%[0-9]+]]:vreg_64(p1) = G_LOAD [[COPY]](p1) :: (load 8)
; GFX7: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX7: [[FLAT_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = FLAT_LOAD_DWORDX2 [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 8)
; GFX7: $vgpr0_vgpr1 = COPY [[FLAT_LOAD_DWORDX2_]]
; GFX8-LABEL: name: load_flat_p1_from_8
; GFX8: liveins: $vgpr0_vgpr1
; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
; GFX8: [[LOAD:%[0-9]+]]:vreg_64(p1) = G_LOAD [[COPY]](p1) :: (load 8)
; GFX8: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX8: [[FLAT_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = FLAT_LOAD_DWORDX2 [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 8)
; GFX8: $vgpr0_vgpr1 = COPY [[FLAT_LOAD_DWORDX2_]]
; GFX9-LABEL: name: load_flat_p1_from_8
; GFX9: liveins: $vgpr0_vgpr1
; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:vreg_64(p1) = G_LOAD [[COPY]](p1) :: (load 8)
; GFX9: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX9: [[FLAT_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = FLAT_LOAD_DWORDX2 [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 8)
; GFX9: $vgpr0_vgpr1 = COPY [[FLAT_LOAD_DWORDX2_]]
; GFX10-LABEL: name: load_flat_p1_from_8
; GFX10: liveins: $vgpr0_vgpr1
; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
; GFX10: [[LOAD:%[0-9]+]]:vreg_64(p1) = G_LOAD [[COPY]](p1) :: (load 8)
; GFX10: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
; GFX10: $vcc_hi = IMPLICIT_DEF
; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX10: [[FLAT_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = FLAT_LOAD_DWORDX2 [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 8)
; GFX10: $vgpr0_vgpr1 = COPY [[FLAT_LOAD_DWORDX2_]]
%0:vgpr(p1) = COPY $vgpr0_vgpr1
%1:vgpr(p1) = G_LOAD %0 :: (load 8, align 8, addrspace 0)
$vgpr0_vgpr1 = COPY %1

50
llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-load-global.mir
View File

@ -434,24 +434,25 @@ body: |
; GFX7-LABEL: name: load_global_p3_from_4
; GFX7: liveins: $vgpr0_vgpr1
; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
; GFX7: [[LOAD:%[0-9]+]]:vgpr_32(p3) = G_LOAD [[COPY]](p1) :: (load 4, addrspace 1)
; GFX7: $vgpr0 = COPY [[LOAD]](p3)
; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX7: [[FLAT_LOAD_DWORD:%[0-9]+]]:vgpr_32 = FLAT_LOAD_DWORD [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 4, addrspace 1)
; GFX7: $vgpr0 = COPY [[FLAT_LOAD_DWORD]]
; GFX8-LABEL: name: load_global_p3_from_4
; GFX8: liveins: $vgpr0_vgpr1
; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
; GFX8: [[LOAD:%[0-9]+]]:vgpr_32(p3) = G_LOAD [[COPY]](p1) :: (load 4, addrspace 1)
; GFX8: $vgpr0 = COPY [[LOAD]](p3)
; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX8: [[FLAT_LOAD_DWORD:%[0-9]+]]:vgpr_32 = FLAT_LOAD_DWORD [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 4, addrspace 1)
; GFX8: $vgpr0 = COPY [[FLAT_LOAD_DWORD]]
; GFX9-LABEL: name: load_global_p3_from_4
; GFX9: liveins: $vgpr0_vgpr1
; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:vgpr_32(p3) = G_LOAD [[COPY]](p1) :: (load 4, addrspace 1)
; GFX9: $vgpr0 = COPY [[LOAD]](p3)
; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX9: [[GLOBAL_LOAD_DWORD:%[0-9]+]]:vgpr_32 = GLOBAL_LOAD_DWORD [[COPY]], 0, 0, 0, 0, implicit $exec :: (load 4, addrspace 1)
; GFX9: $vgpr0 = COPY [[GLOBAL_LOAD_DWORD]]
; GFX10-LABEL: name: load_global_p3_from_4
; GFX10: liveins: $vgpr0_vgpr1
; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
; GFX10: [[LOAD:%[0-9]+]]:vgpr_32(p3) = G_LOAD [[COPY]](p1) :: (load 4, addrspace 1)
; GFX10: $vgpr0 = COPY [[LOAD]](p3)
; GFX10: $vcc_hi = IMPLICIT_DEF
; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX10: [[GLOBAL_LOAD_DWORD:%[0-9]+]]:vgpr_32 = GLOBAL_LOAD_DWORD [[COPY]], 0, 0, 0, 0, implicit $exec :: (load 4, addrspace 1)
; GFX10: $vgpr0 = COPY [[GLOBAL_LOAD_DWORD]]
%0:vgpr(p1) = COPY $vgpr0_vgpr1
%1:vgpr(p3) = G_LOAD %0 :: (load 4, align 4, addrspace 1)
$vgpr0 = COPY %1
@ -471,24 +472,25 @@ body: |
; GFX7-LABEL: name: load_global_p1_from_8
; GFX7: liveins: $vgpr0_vgpr1
; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
; GFX7: [[LOAD:%[0-9]+]]:vreg_64(p1) = G_LOAD [[COPY]](p1) :: (load 8, addrspace 1)
; GFX7: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX7: [[FLAT_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = FLAT_LOAD_DWORDX2 [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 8, addrspace 1)
; GFX7: $vgpr0_vgpr1 = COPY [[FLAT_LOAD_DWORDX2_]]
; GFX8-LABEL: name: load_global_p1_from_8
; GFX8: liveins: $vgpr0_vgpr1
; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
; GFX8: [[LOAD:%[0-9]+]]:vreg_64(p1) = G_LOAD [[COPY]](p1) :: (load 8, addrspace 1)
; GFX8: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX8: [[FLAT_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = FLAT_LOAD_DWORDX2 [[COPY]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (load 8, addrspace 1)
; GFX8: $vgpr0_vgpr1 = COPY [[FLAT_LOAD_DWORDX2_]]
; GFX9-LABEL: name: load_global_p1_from_8
; GFX9: liveins: $vgpr0_vgpr1
; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
; GFX9: [[LOAD:%[0-9]+]]:vreg_64(p1) = G_LOAD [[COPY]](p1) :: (load 8, addrspace 1)
; GFX9: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX9: [[GLOBAL_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = GLOBAL_LOAD_DWORDX2 [[COPY]], 0, 0, 0, 0, implicit $exec :: (load 8, addrspace 1)
; GFX9: $vgpr0_vgpr1 = COPY [[GLOBAL_LOAD_DWORDX2_]]
; GFX10-LABEL: name: load_global_p1_from_8
; GFX10: liveins: $vgpr0_vgpr1
; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
; GFX10: [[LOAD:%[0-9]+]]:vreg_64(p1) = G_LOAD [[COPY]](p1) :: (load 8, addrspace 1)
; GFX10: $vgpr0_vgpr1 = COPY [[LOAD]](p1)
; GFX10: $vcc_hi = IMPLICIT_DEF
; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX10: [[GLOBAL_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = GLOBAL_LOAD_DWORDX2 [[COPY]], 0, 0, 0, 0, implicit $exec :: (load 8, addrspace 1)
; GFX10: $vgpr0_vgpr1 = COPY [[GLOBAL_LOAD_DWORDX2_]]
%0:vgpr(p1) = COPY $vgpr0_vgpr1
%1:vgpr(p1) = G_LOAD %0 :: (load 8, align 8, addrspace 1)
$vgpr0_vgpr1 = COPY %1

50
llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-store-flat.mir
View File

@ -537,24 +537,25 @@ body: |
; GFX7-LABEL: name: store_flat_p1
; GFX7: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
; GFX7: [[COPY1:%[0-9]+]]:vgpr(p1) = COPY $vgpr2_vgpr3
; GFX7: G_STORE [[COPY1]](p1), [[COPY]](p1) :: (store 8)
; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX7: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
; GFX7: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 8)
; GFX8-LABEL: name: store_flat_p1
; GFX8: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
; GFX8: [[COPY1:%[0-9]+]]:vgpr(p1) = COPY $vgpr2_vgpr3
; GFX8: G_STORE [[COPY1]](p1), [[COPY]](p1) :: (store 8)
; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX8: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
; GFX8: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 8)
; GFX9-LABEL: name: store_flat_p1
; GFX9: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
; GFX9: [[COPY1:%[0-9]+]]:vgpr(p1) = COPY $vgpr2_vgpr3
; GFX9: G_STORE [[COPY1]](p1), [[COPY]](p1) :: (store 8)
; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX9: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
; GFX9: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 8)
; GFX10-LABEL: name: store_flat_p1
; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
; GFX10: [[COPY1:%[0-9]+]]:vgpr(p1) = COPY $vgpr2_vgpr3
; GFX10: G_STORE [[COPY1]](p1), [[COPY]](p1) :: (store 8)
; GFX10: $vcc_hi = IMPLICIT_DEF
; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX10: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
; GFX10: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 8)
%0:vgpr(p1) = COPY $vgpr0_vgpr1
%1:vgpr(p1) = COPY $vgpr2_vgpr3
G_STORE %1, %0 :: (store 8, align 8, addrspace 0)
@ -611,24 +612,25 @@ body: |
; GFX7-LABEL: name: store_flat_p3
; GFX7: liveins: $vgpr0_vgpr1, $vgpr2
; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
; GFX7: [[COPY1:%[0-9]+]]:vgpr(p3) = COPY $vgpr2
; GFX7: G_STORE [[COPY1]](p3), [[COPY]](p1) :: (store 4)
; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; GFX7: FLAT_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 4)
; GFX8-LABEL: name: store_flat_p3
; GFX8: liveins: $vgpr0_vgpr1, $vgpr2
; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
; GFX8: [[COPY1:%[0-9]+]]:vgpr(p3) = COPY $vgpr2
; GFX8: G_STORE [[COPY1]](p3), [[COPY]](p1) :: (store 4)
; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; GFX8: FLAT_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 4)
; GFX9-LABEL: name: store_flat_p3
; GFX9: liveins: $vgpr0_vgpr1, $vgpr2
; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
; GFX9: [[COPY1:%[0-9]+]]:vgpr(p3) = COPY $vgpr2
; GFX9: G_STORE [[COPY1]](p3), [[COPY]](p1) :: (store 4)
; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX9: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; GFX9: FLAT_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 4)
; GFX10-LABEL: name: store_flat_p3
; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
; GFX10: [[COPY1:%[0-9]+]]:vgpr(p3) = COPY $vgpr2
; GFX10: G_STORE [[COPY1]](p3), [[COPY]](p1) :: (store 4)
; GFX10: $vcc_hi = IMPLICIT_DEF
; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; GFX10: FLAT_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 4)
%0:vgpr(p1) = COPY $vgpr0_vgpr1
%1:vgpr(p3) = COPY $vgpr2
G_STORE %1, %0 :: (store 4, align 4, addrspace 0)

50
llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-store-global.mir
View File

@ -537,24 +537,25 @@ body: |
; GFX7-LABEL: name: store_global_p1
; GFX7: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
; GFX7: [[COPY1:%[0-9]+]]:vgpr(p1) = COPY $vgpr2_vgpr3
; GFX7: G_STORE [[COPY1]](p1), [[COPY]](p1) :: (store 8, addrspace 1)
; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX7: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
; GFX7: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 8, addrspace 1)
; GFX8-LABEL: name: store_global_p1
; GFX8: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
; GFX8: [[COPY1:%[0-9]+]]:vgpr(p1) = COPY $vgpr2_vgpr3
; GFX8: G_STORE [[COPY1]](p1), [[COPY]](p1) :: (store 8, addrspace 1)
; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX8: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
; GFX8: FLAT_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 8, addrspace 1)
; GFX9-LABEL: name: store_global_p1
; GFX9: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
; GFX9: [[COPY1:%[0-9]+]]:vgpr(p1) = COPY $vgpr2_vgpr3
; GFX9: G_STORE [[COPY1]](p1), [[COPY]](p1) :: (store 8, addrspace 1)
; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX9: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
; GFX9: GLOBAL_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec :: (store 8, addrspace 1)
; GFX10-LABEL: name: store_global_p1
; GFX10: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
; GFX10: [[COPY1:%[0-9]+]]:vgpr(p1) = COPY $vgpr2_vgpr3
; GFX10: G_STORE [[COPY1]](p1), [[COPY]](p1) :: (store 8, addrspace 1)
; GFX10: $vcc_hi = IMPLICIT_DEF
; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX10: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
; GFX10: GLOBAL_STORE_DWORDX2 [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec :: (store 8, addrspace 1)
%0:vgpr(p1) = COPY $vgpr0_vgpr1
%1:vgpr(p1) = COPY $vgpr2_vgpr3
G_STORE %1, %0 :: (store 8, align 8, addrspace 1)
@ -611,24 +612,25 @@ body: |
; GFX7-LABEL: name: store_global_p3
; GFX7: liveins: $vgpr0_vgpr1, $vgpr2
; GFX7: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
; GFX7: [[COPY1:%[0-9]+]]:vgpr(p3) = COPY $vgpr2
; GFX7: G_STORE [[COPY1]](p3), [[COPY]](p1) :: (store 4, addrspace 1)
; GFX7: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX7: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; GFX7: FLAT_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 4, addrspace 1)
; GFX8-LABEL: name: store_global_p3
; GFX8: liveins: $vgpr0_vgpr1, $vgpr2
; GFX8: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
; GFX8: [[COPY1:%[0-9]+]]:vgpr(p3) = COPY $vgpr2
; GFX8: G_STORE [[COPY1]](p3), [[COPY]](p1) :: (store 4, addrspace 1)
; GFX8: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX8: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; GFX8: FLAT_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 4, addrspace 1)
; GFX9-LABEL: name: store_global_p3
; GFX9: liveins: $vgpr0_vgpr1, $vgpr2
; GFX9: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
; GFX9: [[COPY1:%[0-9]+]]:vgpr(p3) = COPY $vgpr2
; GFX9: G_STORE [[COPY1]](p3), [[COPY]](p1) :: (store 4, addrspace 1)
; GFX9: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX9: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; GFX9: GLOBAL_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec :: (store 4, addrspace 1)
; GFX10-LABEL: name: store_global_p3
; GFX10: liveins: $vgpr0_vgpr1, $vgpr2
; GFX10: [[COPY:%[0-9]+]]:vgpr(p1) = COPY $vgpr0_vgpr1
; GFX10: [[COPY1:%[0-9]+]]:vgpr(p3) = COPY $vgpr2
; GFX10: G_STORE [[COPY1]](p3), [[COPY]](p1) :: (store 4, addrspace 1)
; GFX10: $vcc_hi = IMPLICIT_DEF
; GFX10: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; GFX10: GLOBAL_STORE_DWORD [[COPY]], [[COPY1]], 0, 0, 0, 0, implicit $exec :: (store 4, addrspace 1)
%0:vgpr(p1) = COPY $vgpr0_vgpr1
%1:vgpr(p3) = COPY $vgpr2
G_STORE %1, %0 :: (store 4, align 4, addrspace 1)