maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/llvm/test/CodeGen/AMDGPU/llvm.r600.read.local.size.ll

185 lines
6.4 KiB
LLVM
Raw Normal View History

Enable FeatureFlatForGlobal on Volcanic Islands This switches to the workaround that HSA defaults to for the mesa path. This should be applied to the 4.0 branch. Patch by Vedran Miletić <vedran@miletic.net> llvm-svn: 292982
2017-01-25 06:02:15 +08:00
; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=GCN -check-prefix=SI-NOHSA -check-prefix=GCN-NOHSA -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=VI-NOHSA -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=FUNC %s
AMDGPU: Rework how private buffer passed for HSA If we know we have stack objects, we reserve the registers that the private buffer resource and wave offset are passed and use them directly. If not, reserve the last 5 SGPRs just in case we need to spill. After register allocation, try to pick the next available registers instead of the last SGPRs, and then insert copies from the inputs to the reserved registers in the progloue. This also only selectively enables all of the input registers which are really required instead of always enabling them. llvm-svn: 254331
2015-12-01 05:16:03 +08:00
; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
; FUNC-LABEL: {{^}}local_size_x:
; EG: MEM_RAT_CACHELESS STORE_RAW [[VAL:T[0-9]+\.X]]
ScheduleDAGInstrs: Rework schedule graph builder. Re-comitting with a change that avoids undefined uses getting put into the VRegUses list. The new algorithm remembers the uses encountered while walking backwards until a matching def is found. Contrary to the previous version this: - Works without LiveIntervals being available - Allows to increase the precision to subregisters/lanemasks (not used for now) The changes in the AMDGPU tests are necessary because the R600 scheduler is not stable with respect to the order of nodes in the ready queues. Differential Revision: http://reviews.llvm.org/D9068 llvm-svn: 254683
2015-12-04 09:51:19 +08:00
; EG: MOV * [[VAL]], KC0[1].Z
AMDGPU: Rework how private buffer passed for HSA If we know we have stack objects, we reserve the registers that the private buffer resource and wave offset are passed and use them directly. If not, reserve the last 5 SGPRs just in case we need to spill. After register allocation, try to pick the next available registers instead of the last SGPRs, and then insert copies from the inputs to the reserved registers in the progloue. This also only selectively enables all of the input registers which are really required instead of always enabling them. llvm-svn: 254331
2015-12-01 05:16:03 +08:00
; SI-NOHSA: s_load_dword [[VAL:s[0-9]+]], s[0:1], 0x6
; VI-NOHSA: s_load_dword [[VAL:s[0-9]+]], s[0:1], 0x18
; CI-HSA: s_load_dword [[XY:s[0-9]+]], s[4:5], 0x1
; VI-HSA: s_load_dword [[XY:s[0-9]+]], s[4:5], 0x4
; GCN: v_mov_b32_e32 [[VVAL:v[0-9]+]], [[VAL]]
; GCN: buffer_store_dword [[VVAL]]
define void @local_size_x(i32 addrspace(1)* %out) {
entry:
%0 = call i32 @llvm.r600.read.local.size.x() #0
store i32 %0, i32 addrspace(1)* %out
ret void
}
; FUNC-LABEL: {{^}}local_size_y:
; EG: MEM_RAT_CACHELESS STORE_RAW [[VAL:T[0-9]+\.X]]
ScheduleDAGInstrs: Rework schedule graph builder. Re-comitting with a change that avoids undefined uses getting put into the VRegUses list. The new algorithm remembers the uses encountered while walking backwards until a matching def is found. Contrary to the previous version this: - Works without LiveIntervals being available - Allows to increase the precision to subregisters/lanemasks (not used for now) The changes in the AMDGPU tests are necessary because the R600 scheduler is not stable with respect to the order of nodes in the ready queues. Differential Revision: http://reviews.llvm.org/D9068 llvm-svn: 254683
2015-12-04 09:51:19 +08:00
; EG: MOV * [[VAL]], KC0[1].W
AMDGPU: Rework how private buffer passed for HSA If we know we have stack objects, we reserve the registers that the private buffer resource and wave offset are passed and use them directly. If not, reserve the last 5 SGPRs just in case we need to spill. After register allocation, try to pick the next available registers instead of the last SGPRs, and then insert copies from the inputs to the reserved registers in the progloue. This also only selectively enables all of the input registers which are really required instead of always enabling them. llvm-svn: 254331
2015-12-01 05:16:03 +08:00
; SI-NOHSA: s_load_dword [[VAL:s[0-9]+]], s[0:1], 0x7
; VI-NOHSA: s_load_dword [[VAL:s[0-9]+]], s[0:1], 0x1c
; GCN: v_mov_b32_e32 [[VVAL:v[0-9]+]], [[VAL]]
; GCN: buffer_store_dword [[VVAL]]
define void @local_size_y(i32 addrspace(1)* %out) {
entry:
%0 = call i32 @llvm.r600.read.local.size.y() #0
store i32 %0, i32 addrspace(1)* %out
ret void
}
; FUNC-LABEL: {{^}}local_size_z:
; EG: MEM_RAT_CACHELESS STORE_RAW [[VAL:T[0-9]+\.X]]
ScheduleDAGInstrs: Rework schedule graph builder. Re-comitting with a change that avoids undefined uses getting put into the VRegUses list. The new algorithm remembers the uses encountered while walking backwards until a matching def is found. Contrary to the previous version this: - Works without LiveIntervals being available - Allows to increase the precision to subregisters/lanemasks (not used for now) The changes in the AMDGPU tests are necessary because the R600 scheduler is not stable with respect to the order of nodes in the ready queues. Differential Revision: http://reviews.llvm.org/D9068 llvm-svn: 254683
2015-12-04 09:51:19 +08:00
; EG: MOV * [[VAL]], KC0[2].X
AMDGPU: Rework how private buffer passed for HSA If we know we have stack objects, we reserve the registers that the private buffer resource and wave offset are passed and use them directly. If not, reserve the last 5 SGPRs just in case we need to spill. After register allocation, try to pick the next available registers instead of the last SGPRs, and then insert copies from the inputs to the reserved registers in the progloue. This also only selectively enables all of the input registers which are really required instead of always enabling them. llvm-svn: 254331
2015-12-01 05:16:03 +08:00
; SI-NOHSA: s_load_dword [[VAL:s[0-9]+]], s[0:1], 0x8
; VI-NOHSA: s_load_dword [[VAL:s[0-9]+]], s[0:1], 0x20
; GCN: v_mov_b32_e32 [[VVAL:v[0-9]+]], [[VAL]]
; GCN: buffer_store_dword [[VVAL]]
define void @local_size_z(i32 addrspace(1)* %out) {
entry:
%0 = call i32 @llvm.r600.read.local.size.z() #0
store i32 %0, i32 addrspace(1)* %out
ret void
}
; FUNC-LABEL: {{^}}local_size_xy:
; SI-NOHSA-DAG: s_load_dword [[X:s[0-9]+]], s[0:1], 0x6
; SI-NOHSA-DAG: s_load_dword [[Y:s[0-9]+]], s[0:1], 0x7
; VI-NOHSA-DAG: s_load_dword [[X:s[0-9]+]], s[0:1], 0x18
; VI-NOHSA-DAG: s_load_dword [[Y:s[0-9]+]], s[0:1], 0x1c
; GCN-DAG: v_mov_b32_e32 [[VY:v[0-9]+]], [[Y]]
; GCN: v_mul_u32_u24_e32 [[VAL:v[0-9]+]], [[X]], [[VY]]
; GCN: buffer_store_dword [[VAL]]
define void @local_size_xy(i32 addrspace(1)* %out) {
entry:
%x = call i32 @llvm.r600.read.local.size.x() #0
%y = call i32 @llvm.r600.read.local.size.y() #0
%val = mul i32 %x, %y
store i32 %val, i32 addrspace(1)* %out
ret void
}
; FUNC-LABEL: {{^}}local_size_xz:
; SI-NOHSA-DAG: s_load_dword [[X:s[0-9]+]], s[0:1], 0x6
; SI-NOHSA-DAG: s_load_dword [[Z:s[0-9]+]], s[0:1], 0x8
; VI-NOHSA-DAG: s_load_dword [[X:s[0-9]+]], s[0:1], 0x18
; VI-NOHSA-DAG: s_load_dword [[Z:s[0-9]+]], s[0:1], 0x20
; HSA-DAG: s_and_b32 [[X:s[0-9]+]], [[XY]], 0xffff
; GCN-DAG: v_mov_b32_e32 [[VZ:v[0-9]+]], [[Z]]
; GCN: v_mul_u32_u24_e32 [[VAL:v[0-9]+]], [[X]], [[VZ]]
; GCN: buffer_store_dword [[VAL]]
define void @local_size_xz(i32 addrspace(1)* %out) {
entry:
%x = call i32 @llvm.r600.read.local.size.x() #0
%z = call i32 @llvm.r600.read.local.size.z() #0
%val = mul i32 %x, %z
store i32 %val, i32 addrspace(1)* %out
ret void
}
; FUNC-LABEL: {{^}}local_size_yz:
; HSA: enable_sgpr_private_segment_buffer = 1
; HSA: enable_sgpr_dispatch_ptr = 1
; SI-NOHSA-DAG: s_load_dword [[Y:s[0-9]+]], s[0:1], 0x7
; SI-NOHSA-DAG: s_load_dword [[Z:s[0-9]+]], s[0:1], 0x8
; VI-NOHSA-DAG: s_load_dword [[Y:s[0-9]+]], s[0:1], 0x1c
; VI-NOHSA-DAG: s_load_dword [[Z:s[0-9]+]], s[0:1], 0x20
; GCN-DAG: v_mov_b32_e32 [[VZ:v[0-9]+]], [[Z]]
; GCN: v_mul_u32_u24_e32 [[VAL:v[0-9]+]], [[Y]], [[VZ]]
; GCN: buffer_store_dword [[VAL]]
define void @local_size_yz(i32 addrspace(1)* %out) {
entry:
%y = call i32 @llvm.r600.read.local.size.y() #0
%z = call i32 @llvm.r600.read.local.size.z() #0
%val = mul i32 %y, %z
store i32 %val, i32 addrspace(1)* %out
ret void
}
; FUNC-LABEL: {{^}}local_size_xyz:
; HSA: enable_sgpr_private_segment_buffer = 1
; HSA: enable_sgpr_dispatch_ptr = 1
; SI-NOHSA-DAG: s_load_dword [[X:s[0-9]+]], s[0:1], 0x6
; SI-NOHSA-DAG: s_load_dword [[Y:s[0-9]+]], s[0:1], 0x7
; SI-NOHSA-DAG: s_load_dword [[Z:s[0-9]+]], s[0:1], 0x8
; VI-NOHSA-DAG: s_load_dword [[X:s[0-9]+]], s[0:1], 0x18
; VI-NOHSA-DAG: s_load_dword [[Y:s[0-9]+]], s[0:1], 0x1c
; VI-NOHSA-DAG: s_load_dword [[Z:s[0-9]+]], s[0:1], 0x20
; GCN-DAG: v_mov_b32_e32 [[VY:v[0-9]+]], [[Y]]
; GCN-DAG: v_mov_b32_e32 [[VZ:v[0-9]+]], [[Z]]
; GCN: v_mad_u32_u24 [[VAL:v[0-9]+]], [[X]], [[VY]], [[VZ]]
; GCN: buffer_store_dword [[VAL]]
define void @local_size_xyz(i32 addrspace(1)* %out) {
entry:
%x = call i32 @llvm.r600.read.local.size.x() #0
%y = call i32 @llvm.r600.read.local.size.y() #0
%z = call i32 @llvm.r600.read.local.size.z() #0
%xy = mul i32 %x, %y
%xyz = add i32 %xy, %z
store i32 %xyz, i32 addrspace(1)* %out
ret void
}
; FUNC-LABEL: {{^}}local_size_x_known_bits:
; SI: s_load_dword [[VAL:s[0-9]+]], s[0:1], 0x6
; VI: s_load_dword [[VAL:s[0-9]+]], s[0:1], 0x18
; GCN-NOT: 0xffff
; GCN: v_mov_b32_e32 [[VVAL:v[0-9]+]], [[VAL]]
; GCN-NEXT: buffer_store_dword [[VVAL]]
define void @local_size_x_known_bits(i32 addrspace(1)* %out) {
entry:
%size = call i32 @llvm.r600.read.local.size.x() #0
%shl = shl i32 %size, 16
%shr = lshr i32 %shl, 16
store i32 %shr, i32 addrspace(1)* %out
ret void
}
; FUNC-LABEL: {{^}}local_size_y_known_bits:
; SI: s_load_dword [[VAL:s[0-9]+]], s[0:1], 0x7
; VI: s_load_dword [[VAL:s[0-9]+]], s[0:1], 0x1c
; GCN-NOT: 0xffff
; GCN: v_mov_b32_e32 [[VVAL:v[0-9]+]], [[VAL]]
; GCN-NEXT: buffer_store_dword [[VVAL]]
define void @local_size_y_known_bits(i32 addrspace(1)* %out) {
entry:
%size = call i32 @llvm.r600.read.local.size.y() #0
%shl = shl i32 %size, 16
%shr = lshr i32 %shl, 16
store i32 %shr, i32 addrspace(1)* %out
ret void
}
; FUNC-LABEL: {{^}}local_size_z_known_bits:
; SI: s_load_dword [[VAL:s[0-9]+]], s[0:1], 0x8
; VI: s_load_dword [[VAL:s[0-9]+]], s[0:1], 0x20
; GCN-NOT: 0xffff
; GCN: v_mov_b32_e32 [[VVAL:v[0-9]+]], [[VAL]]
; GCN-NEXT: buffer_store_dword [[VVAL]]
define void @local_size_z_known_bits(i32 addrspace(1)* %out) {
entry:
%size = call i32 @llvm.r600.read.local.size.z() #0
%shl = shl i32 %size, 16
%shr = lshr i32 %shl, 16
store i32 %shr, i32 addrspace(1)* %out
ret void
}
declare i32 @llvm.r600.read.local.size.x() #0
declare i32 @llvm.r600.read.local.size.y() #0
declare i32 @llvm.r600.read.local.size.z() #0
attributes #0 = { nounwind readnone }