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/callee-special-input-vgprs.ll

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

899 lines
33 KiB
LLVM
Raw Normal View History

AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=kaveri -mattr=-code-object-v3 -enable-ipra=0 -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,VARABI %s
; RUN: llc -amdgpu-fixed-function-abi -mtriple=amdgcn-amd-amdhsa -mcpu=kaveri -mattr=-code-object-v3 -enable-ipra=0 -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,FIXEDABI %s
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
; GCN-LABEL: {{^}}use_workitem_id_x:
; GCN: s_waitcnt
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; VARABI: v_and_b32_e32 [[ID:v[0-9]+]], 0x3ff, v0
; FIXEDABI: v_and_b32_e32 [[ID:v[0-9]+]], 0x3ff, v31
[AMDGPU] Test update. NFC. llvm-svn: 366715
2019-07-23 02:08:53 +08:00
; GCN-NEXT: {{flat|global}}_store_dword v{{\[[0-9]:[0-9]+\]}}, [[ID]]
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
; GCN-NEXT: s_waitcnt
; GCN-NEXT: s_setpc_b64
define void @use_workitem_id_x() #1 {
%val = call i32 @llvm.amdgcn.workitem.id.x()
store volatile i32 %val, i32 addrspace(1)* undef
ret void
}
; GCN-LABEL: {{^}}use_workitem_id_y:
; GCN: s_waitcnt
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; VARABI: v_bfe_u32 [[ID:v[0-9]+]], v0, 10, 10
; FIXEDABI: v_bfe_u32 [[ID:v[0-9]+]], v31, 10, 10
[AMDGPU] Test update. NFC. llvm-svn: 366715
2019-07-23 02:08:53 +08:00
; GCN-NEXT: {{flat|global}}_store_dword v{{\[[0-9]:[0-9]+\]}}, [[ID]]
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
; GCN-NEXT: s_waitcnt
; GCN-NEXT: s_setpc_b64
define void @use_workitem_id_y() #1 {
%val = call i32 @llvm.amdgcn.workitem.id.y()
store volatile i32 %val, i32 addrspace(1)* undef
ret void
}
; GCN-LABEL: {{^}}use_workitem_id_z:
; GCN: s_waitcnt
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; VARABI: v_bfe_u32 [[ID:v[0-9]+]], v0, 20, 10
; FIXEDABI: v_bfe_u32 [[ID:v[0-9]+]], v31, 20, 10
[AMDGPU] Test update. NFC. llvm-svn: 366715
2019-07-23 02:08:53 +08:00
; GCN-NEXT: {{flat|global}}_store_dword v{{\[[0-9]:[0-9]+\]}}, [[ID]]
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
; GCN-NEXT: s_waitcnt
; GCN-NEXT: s_setpc_b64
define void @use_workitem_id_z() #1 {
%val = call i32 @llvm.amdgcn.workitem.id.z()
store volatile i32 %val, i32 addrspace(1)* undef
ret void
}
; GCN-LABEL: {{^}}use_workitem_id_xy:
; GCN: s_waitcnt
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; VARABI-DAG: v_and_b32_e32 [[IDX:v[0-9]+]], 0x3ff, v0
; VARABI-DAG: v_bfe_u32 [[IDY:v[0-9]+]], v0, 10, 10
; FIXEDABI-DAG: v_and_b32_e32 [[IDX:v[0-9]+]], 0x3ff, v31
; FIXEDABI-DAG: v_bfe_u32 [[IDY:v[0-9]+]], v31, 10, 10
[AMDGPU] Test update. NFC. llvm-svn: 366715
2019-07-23 02:08:53 +08:00
; GCN-DAG: {{flat|global}}_store_dword v{{\[[0-9]:[0-9]+\]}}, [[IDX]]
; GCN-DAG: {{flat|global}}_store_dword v{{\[[0-9]:[0-9]+\]}}, [[IDY]]
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
; GCN-NEXT: s_waitcnt
; GCN-NEXT: s_setpc_b64
define void @use_workitem_id_xy() #1 {
%val0 = call i32 @llvm.amdgcn.workitem.id.x()
%val1 = call i32 @llvm.amdgcn.workitem.id.y()
store volatile i32 %val0, i32 addrspace(1)* undef
store volatile i32 %val1, i32 addrspace(1)* undef
ret void
}
; GCN-LABEL: {{^}}use_workitem_id_xyz:
; GCN: s_waitcnt
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; VARABI-DAG: v_and_b32_e32 [[IDX:v[0-9]+]], 0x3ff, v0
; VARABI-DAG: v_bfe_u32 [[IDY:v[0-9]+]], v0, 10, 10
; VARABI-DAG: v_bfe_u32 [[IDZ:v[0-9]+]], v0, 20, 10
; FIXEDABI-DAG: v_and_b32_e32 [[IDX:v[0-9]+]], 0x3ff, v31
; FIXEDABI-DAG: v_bfe_u32 [[IDY:v[0-9]+]], v31, 10, 10
; FIXEDABI-DAG: v_bfe_u32 [[IDZ:v[0-9]+]], v31, 20, 10
[AMDGPU] Test update. NFC. llvm-svn: 366715
2019-07-23 02:08:53 +08:00
; GCN-DAG: {{flat|global}}_store_dword v{{\[[0-9]:[0-9]+\]}}, [[IDX]]
; GCN-DAG: {{flat|global}}_store_dword v{{\[[0-9]:[0-9]+\]}}, [[IDY]]
; GCN-DAG: {{flat|global}}_store_dword v{{\[[0-9]:[0-9]+\]}}, [[IDZ]]
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
; GCN-NEXT: s_waitcnt
; GCN-NEXT: s_setpc_b64
define void @use_workitem_id_xyz() #1 {
%val0 = call i32 @llvm.amdgcn.workitem.id.x()
%val1 = call i32 @llvm.amdgcn.workitem.id.y()
%val2 = call i32 @llvm.amdgcn.workitem.id.z()
store volatile i32 %val0, i32 addrspace(1)* undef
store volatile i32 %val1, i32 addrspace(1)* undef
store volatile i32 %val2, i32 addrspace(1)* undef
ret void
}
; GCN-LABEL: {{^}}use_workitem_id_xz:
; GCN: s_waitcnt
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; VARABI-DAG: v_and_b32_e32 [[IDX:v[0-9]+]], 0x3ff, v0
; VARABI-DAG: v_bfe_u32 [[IDZ:v[0-9]+]], v0, 20, 10
; FIXEDABI-DAG: v_and_b32_e32 [[IDX:v[0-9]+]], 0x3ff, v31
; FIXEDABI-DAG: v_bfe_u32 [[IDZ:v[0-9]+]], v31, 20, 10
[AMDGPU] Test update. NFC. llvm-svn: 366715
2019-07-23 02:08:53 +08:00
; GCN-DAG: {{flat|global}}_store_dword v{{\[[0-9]:[0-9]+\]}}, [[IDX]]
; GCN-DAG: {{flat|global}}_store_dword v{{\[[0-9]:[0-9]+\]}}, [[IDZ]]
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
; GCN-NEXT: s_waitcnt
; GCN-NEXT: s_setpc_b64
define void @use_workitem_id_xz() #1 {
%val0 = call i32 @llvm.amdgcn.workitem.id.x()
%val1 = call i32 @llvm.amdgcn.workitem.id.z()
store volatile i32 %val0, i32 addrspace(1)* undef
store volatile i32 %val1, i32 addrspace(1)* undef
ret void
}
; GCN-LABEL: {{^}}use_workitem_id_yz:
; GCN: s_waitcnt
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; VARABI-DAG: v_bfe_u32 [[IDY:v[0-9]+]], v0, 10, 10
; VARABI-DAG: v_bfe_u32 [[IDZ:v[0-9]+]], v0, 20, 10
; FIXEDABI-DAG: v_bfe_u32 [[IDY:v[0-9]+]], v31, 10, 10
; FIXEDABI-DAG: v_bfe_u32 [[IDZ:v[0-9]+]], v31, 20, 10
[AMDGPU] Test update. NFC. llvm-svn: 366715
2019-07-23 02:08:53 +08:00
; GCN-DAG: {{flat|global}}_store_dword v{{\[[0-9]:[0-9]+\]}}, [[IDY]]
; GCN-DAG: {{flat|global}}_store_dword v{{\[[0-9]:[0-9]+\]}}, [[IDZ]]
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
; GCN-NEXT: s_waitcnt
; GCN-NEXT: s_setpc_b64
define void @use_workitem_id_yz() #1 {
%val0 = call i32 @llvm.amdgcn.workitem.id.y()
%val1 = call i32 @llvm.amdgcn.workitem.id.z()
store volatile i32 %val0, i32 addrspace(1)* undef
store volatile i32 %val1, i32 addrspace(1)* undef
ret void
}
; GCN-LABEL: {{^}}kern_indirect_use_workitem_id_x:
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; VARABI: enable_vgpr_workitem_id = 0
; FIXEDABI: enable_vgpr_workitem_id = 2
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; FIXEDA-NOT: v0
; VARABI-NOT: v31
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
; GCN: s_swappc_b64
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; FIXEDABI-NOT: v0
; VARABI-NOT: v31
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
define amdgpu_kernel void @kern_indirect_use_workitem_id_x() #1 {
call void @use_workitem_id_x()
ret void
}
; GCN-LABEL: {{^}}kern_indirect_use_workitem_id_y:
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; VARABI: enable_vgpr_workitem_id = 1
; FIXEDABI: enable_vgpr_workitem_id = 2
; FIXEDABI-NOT: v0
; FIXEDABI-NOT: v1
; VARABI-NOT: v31
; VARABI: v_lshlrev_b32_e32 v0, 10, v1
; FIXEDABI-DAG: v_lshlrev_b32_e32 [[TMP1:v[0-9]+]], 10, v1
; FIXEDABI-DAG: v_lshlrev_b32_e32 [[TMP0:v[0-9]+]], 20, v2
; FIXEDABI: v_or_b32_e32 [[TMP2:v[0-9]+]], v0, [[TMP1]]
; FIXEDABI: v_or_b32_e32 v31, [[TMP2]], [[TMP0]]
; FIXEDABI-NOT: v0
; FIXEDABI-NOT: v1
; VARABI-NOT: v31
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
; GCN: s_swappc_b64
define amdgpu_kernel void @kern_indirect_use_workitem_id_y() #1 {
call void @use_workitem_id_y()
ret void
}
; GCN-LABEL: {{^}}kern_indirect_use_workitem_id_z:
; GCN: enable_vgpr_workitem_id = 2
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; VARABI-NOT: v0
; VARABI-NOT: v2
; VARABI: v_lshlrev_b32_e32 v0, 20, v2
; VARABI-NOT: v0
; VARABI-NOT: v1
; FIXEDABI-DAG: v_lshlrev_b32_e32 [[TMP1:v[0-9]+]], 10, v1
; FIXEDABI-DAG: v_lshlrev_b32_e32 [[TMP0:v[0-9]+]], 20, v2
; FIXEDABI: v_or_b32_e32 [[TMP2:v[0-9]+]], v0, [[TMP1]]
; FIXEDABI: v_or_b32_e32 v31, [[TMP2]], [[TMP0]]
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
; GCN: s_swappc_b64
define amdgpu_kernel void @kern_indirect_use_workitem_id_z() #1 {
call void @use_workitem_id_z()
ret void
}
[AMDGPU] Packed thread ids in function call ABI Differential Revision: https://reviews.llvm.org/D63851 llvm-svn: 364619
2019-06-28 09:52:13 +08:00
; GCN-LABEL: {{^}}kern_indirect_use_workitem_id_xy:
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; VARABI-NOT: v0
; VARABI-NOT: v1
; VARABI: v_lshlrev_b32_e32 [[IDY:v[0-9]+]], 10, v1
; VARABI: v_or_b32_e32 v0, v0, [[IDY]]
; VARABI-NOT: v0
; VARABI-NOT: v1
; FIXEDABI-DAG: v_lshlrev_b32_e32 [[TMP1:v[0-9]+]], 10, v1
; FIXEDABI-DAG: v_lshlrev_b32_e32 [[TMP0:v[0-9]+]], 20, v2
; FIXEDABI: v_or_b32_e32 [[TMP2:v[0-9]+]], v0, [[TMP1]]
; FIXEDABI: v_or_b32_e32 v31, [[TMP2]], [[TMP0]]
[AMDGPU] Packed thread ids in function call ABI Differential Revision: https://reviews.llvm.org/D63851 llvm-svn: 364619
2019-06-28 09:52:13 +08:00
; GCN: s_swappc_b64
define amdgpu_kernel void @kern_indirect_use_workitem_id_xy() #1 {
call void @use_workitem_id_xy()
ret void
}
; GCN-LABEL: {{^}}kern_indirect_use_workitem_id_xz:
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; VARABI-NOT: v0
; VARABI-NOT: v2
; VARABI: v_lshlrev_b32_e32 [[IDZ:v[0-9]+]], 20, v2
; VARABI: v_or_b32_e32 v0, v0, [[IDZ]]
; VARABI-NOT: v0
; VARABI-NOT: v2
; FIXEDABI-DAG: v_lshlrev_b32_e32 [[TMP1:v[0-9]+]], 10, v1
; FIXEDABI-DAG: v_lshlrev_b32_e32 [[TMP0:v[0-9]+]], 20, v2
; FIXEDABI: v_or_b32_e32 [[TMP2:v[0-9]+]], v0, [[TMP1]]
; FIXEDABI: v_or_b32_e32 v31, [[TMP2]], [[TMP0]]
[AMDGPU] Packed thread ids in function call ABI Differential Revision: https://reviews.llvm.org/D63851 llvm-svn: 364619
2019-06-28 09:52:13 +08:00
; GCN: s_swappc_b64
define amdgpu_kernel void @kern_indirect_use_workitem_id_xz() #1 {
call void @use_workitem_id_xz()
ret void
}
; GCN-LABEL: {{^}}kern_indirect_use_workitem_id_yz:
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; VARABI-NOT: v1
; VARABI-NOT: v2
; VARABI-DAG: v_lshlrev_b32_e32 [[IDY:v[0-9]+]], 10, v1
; VARABI-DAG: v_lshlrev_b32_e32 [[IDZ:v[0-9]+]], 20, v2
; VARABI: v_or_b32_e32 v0, [[IDY]], [[IDZ]]
; VARABI-NOT: v1
; VARABI-NOT: v2
; FIXEDABI-DAG: v_lshlrev_b32_e32 [[TMP1:v[0-9]+]], 10, v1
; FIXEDABI-DAG: v_lshlrev_b32_e32 [[TMP0:v[0-9]+]], 20, v2
; FIXEDABI: v_or_b32_e32 [[TMP2:v[0-9]+]], v0, [[TMP1]]
; FIXEDABI: v_or_b32_e32 v31, [[TMP2]], [[TMP0]]
[AMDGPU] Packed thread ids in function call ABI Differential Revision: https://reviews.llvm.org/D63851 llvm-svn: 364619
2019-06-28 09:52:13 +08:00
; GCN: s_swappc_b64
define amdgpu_kernel void @kern_indirect_use_workitem_id_yz() #1 {
call void @use_workitem_id_yz()
ret void
}
; GCN-LABEL: {{^}}kern_indirect_use_workitem_id_xyz:
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; VARABI-NOT: v0
; VARABI-NOT: v1
; VARABI-NOT: v2
; VARABI-DAG: v_lshlrev_b32_e32 [[IDY:v[0-9]+]], 10, v1
; VARABI-DAG: v_lshlrev_b32_e32 [[IDZ:v[0-9]+]], 20, v2
; VARABI-DAG: v_or_b32_e32 v0, v0, [[IDY]]
; VARABI-DAG: v_or_b32_e32 v0, v0, [[IDZ]]
; VARABI-NOT: v0
; VARABI-NOT: v1
; VARABI-NOT: v2
; FIXEDABI-DAG: v_lshlrev_b32_e32 [[TMP1:v[0-9]+]], 10, v1
; FIXEDABI-DAG: v_lshlrev_b32_e32 [[TMP0:v[0-9]+]], 20, v2
; FIXEDABI: v_or_b32_e32 [[TMP2:v[0-9]+]], v0, [[TMP1]]
; FIXEDABI: v_or_b32_e32 v31, [[TMP2]], [[TMP0]]
[AMDGPU] Packed thread ids in function call ABI Differential Revision: https://reviews.llvm.org/D63851 llvm-svn: 364619
2019-06-28 09:52:13 +08:00
; GCN: s_swappc_b64
define amdgpu_kernel void @kern_indirect_use_workitem_id_xyz() #1 {
call void @use_workitem_id_xyz()
ret void
}
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
; GCN-LABEL: {{^}}func_indirect_use_workitem_id_x:
; GCN-NOT: v0
; GCN: s_swappc_b64
; GCN-NOT: v0
define void @func_indirect_use_workitem_id_x() #1 {
call void @use_workitem_id_x()
ret void
}
; GCN-LABEL: {{^}}func_indirect_use_workitem_id_y:
; GCN-NOT: v0
; GCN: s_swappc_b64
; GCN-NOT: v0
define void @func_indirect_use_workitem_id_y() #1 {
call void @use_workitem_id_y()
ret void
}
; GCN-LABEL: {{^}}func_indirect_use_workitem_id_z:
; GCN-NOT: v0
; GCN: s_swappc_b64
; GCN-NOT: v0
define void @func_indirect_use_workitem_id_z() #1 {
call void @use_workitem_id_z()
ret void
}
; GCN-LABEL: {{^}}other_arg_use_workitem_id_x:
; GCN: s_waitcnt
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; VARABI-DAG: v_and_b32_e32 [[ID:v[0-9]+]], 0x3ff, v1
; FIXEDABI-DAG: v_and_b32_e32 [[ID:v[0-9]+]], 0x3ff, v31
[AMDGPU] Test update. NFC. llvm-svn: 366715
2019-07-23 02:08:53 +08:00
; GCN-DAG: {{flat|global}}_store_dword v{{\[[0-9]+:[0-9]+\]}}, v0
; GCN-DAG: {{flat|global}}_store_dword v{{\[[0-9]+:[0-9]+\]}}, [[ID]]
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
define void @other_arg_use_workitem_id_x(i32 %arg0) #1 {
%val = call i32 @llvm.amdgcn.workitem.id.x()
store volatile i32 %arg0, i32 addrspace(1)* undef
store volatile i32 %val, i32 addrspace(1)* undef
ret void
}
; GCN-LABEL: {{^}}other_arg_use_workitem_id_y:
; GCN: s_waitcnt
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; VARABI-DAG: v_bfe_u32 [[ID:v[0-9]+]], v1, 10, 10
; FIXEDABI-DAG: v_bfe_u32 [[ID:v[0-9]+]], v31, 10, 10
[AMDGPU] Test update. NFC. llvm-svn: 366715
2019-07-23 02:08:53 +08:00
; GCN-DAG: {{flat|global}}_store_dword v{{\[[0-9]+:[0-9]+\]}}, v0
; GCN-DAG: {{flat|global}}_store_dword v{{\[[0-9]+:[0-9]+\]}}, [[ID]]
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
define void @other_arg_use_workitem_id_y(i32 %arg0) #1 {
%val = call i32 @llvm.amdgcn.workitem.id.y()
store volatile i32 %arg0, i32 addrspace(1)* undef
store volatile i32 %val, i32 addrspace(1)* undef
ret void
}
; GCN-LABEL: {{^}}other_arg_use_workitem_id_z:
; GCN: s_waitcnt
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; VARABI-DAG: v_bfe_u32 [[ID:v[0-9]+]], v1, 20, 10
; FIXEDABI-DAG: v_bfe_u32 [[ID:v[0-9]+]], v31, 20, 10
[AMDGPU] Test update. NFC. llvm-svn: 366715
2019-07-23 02:08:53 +08:00
; GCN-DAG: {{flat|global}}_store_dword v{{\[[0-9]+:[0-9]+\]}}, v0
; GCN-DAG: {{flat|global}}_store_dword v{{\[[0-9]+:[0-9]+\]}}, [[ID]]
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
define void @other_arg_use_workitem_id_z(i32 %arg0) #1 {
%val = call i32 @llvm.amdgcn.workitem.id.z()
store volatile i32 %arg0, i32 addrspace(1)* undef
store volatile i32 %val, i32 addrspace(1)* undef
ret void
}
; GCN-LABEL: {{^}}kern_indirect_other_arg_use_workitem_id_x:
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; VARABI: enable_vgpr_workitem_id = 0
; FIXEDABI: enable_vgpr_workitem_id = 2
; VARABI: v_mov_b32_e32 v1, v0
; VARABI: v_mov_b32_e32 v0, 0x22b
; FIXEDABI-DAG: v_lshlrev_b32_e32 [[TMP1:v[0-9]+]], 10, v1
; FIXEDABI-DAG: v_lshlrev_b32_e32 [[TMP0:v[0-9]+]], 20, v2
; FIXEDABI: v_or_b32_e32 [[TMP2:v[0-9]+]], v0, [[TMP1]]
; FIXEDABI: v_or_b32_e32 v31, [[TMP2]], [[TMP0]]
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
; GCN: s_swappc_b64
define amdgpu_kernel void @kern_indirect_other_arg_use_workitem_id_x() #1 {
call void @other_arg_use_workitem_id_x(i32 555)
ret void
}
; GCN-LABEL: {{^}}kern_indirect_other_arg_use_workitem_id_y:
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; VARABI: enable_vgpr_workitem_id = 1
; VARABI: v_lshlrev_b32_e32 v1, 10, v1
; VARABI-NOT: v1
; VARABI: v_mov_b32_e32 v0, 0x22b
; VARABI-NOT: v1
; VARABI: s_swappc_b64
; VARABI-NOT: v0
; FIXEDABI: enable_vgpr_workitem_id = 2
; FIXEDABI-DAG: v_lshlrev_b32_e32 [[TMP1:v[0-9]+]], 10, v1
; FIXEDABI-DAG: v_lshlrev_b32_e32 [[TMP0:v[0-9]+]], 20, v2
; FIXEDABI: v_or_b32_e32 [[TMP2:v[0-9]+]], v0, [[TMP1]]
; FIXEDABI: v_or_b32_e32 v31, [[TMP2]], [[TMP0]]
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
define amdgpu_kernel void @kern_indirect_other_arg_use_workitem_id_y() #1 {
call void @other_arg_use_workitem_id_y(i32 555)
ret void
}
; GCN-LABEL: {{^}}kern_indirect_other_arg_use_workitem_id_z:
; GCN: enable_vgpr_workitem_id = 2
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; VARABI-DAG: v_mov_b32_e32 v0, 0x22b
; VARABI-DAG: v_lshlrev_b32_e32 v1, 20, v2
; VARABI: s_swappc_b64
; VARABI-NOT: v0
; FIXEDABI-DAG: v_lshlrev_b32_e32 [[TMP1:v[0-9]+]], 10, v1
; FIXEDABI-DAG: v_lshlrev_b32_e32 [[TMP0:v[0-9]+]], 20, v2
; FIXEDABI: v_or_b32_e32 [[TMP2:v[0-9]+]], v0, [[TMP1]]
; FIXEDABI: v_or_b32_e32 v31, [[TMP2]], [[TMP0]]
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
define amdgpu_kernel void @kern_indirect_other_arg_use_workitem_id_z() #1 {
call void @other_arg_use_workitem_id_z(i32 555)
ret void
}
; GCN-LABEL: {{^}}too_many_args_use_workitem_id_x:
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; VARABI: buffer_load_dword v32, off, s[0:3], s32{{$}}
; VARABI: v_and_b32_e32 v32, 0x3ff, v32
; VARABI: {{flat|global}}_store_dword v{{\[[0-9]+:[0-9]+]}}, v32
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
[AMDGPU] Introduce more scratch registers in the ABI. The AMDGPU target has a convention that defined all VGPRs (execept the initial 32 argument registers) as callee-saved. This convention is not efficient always, esp. when the callee requiring more registers, ended up emitting a large number of spills, even though its caller requires only a few. This patch revises the ABI by introducing more scratch registers that a callee can freely use. The 256 vgpr registers now become: 32 argument registers 112 scratch registers and 112 callee saved registers. The scratch registers and the CSRs are intermixed at regular intervals (a split boundary of 8) to obtain a better occupancy. Reviewers: arsenm, t-tye, rampitec, b-sumner, mjbedy, tpr Reviewed By: arsenm, t-tye Differential Revision: https://reviews.llvm.org/D76356
2020-03-27 15:46:51 +08:00
; VARABI: s_waitcnt
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; VARABI-NEXT: s_setpc_b64
; FIXEDABI: v_and_b32_e32 v31, 0x3ff, v31
; FIXEDABI: buffer_load_dword v{{[0-9]+}}, off, s[0:3], s32{{$}}
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
define void @too_many_args_use_workitem_id_x(
i32 %arg0, i32 %arg1, i32 %arg2, i32 %arg3, i32 %arg4, i32 %arg5, i32 %arg6, i32 %arg7,
i32 %arg8, i32 %arg9, i32 %arg10, i32 %arg11, i32 %arg12, i32 %arg13, i32 %arg14, i32 %arg15,
i32 %arg16, i32 %arg17, i32 %arg18, i32 %arg19, i32 %arg20, i32 %arg21, i32 %arg22, i32 %arg23,
i32 %arg24, i32 %arg25, i32 %arg26, i32 %arg27, i32 %arg28, i32 %arg29, i32 %arg30, i32 %arg31) #1 {
%val = call i32 @llvm.amdgcn.workitem.id.x()
store volatile i32 %val, i32 addrspace(1)* undef
store volatile i32 %arg0, i32 addrspace(1)* undef
store volatile i32 %arg1, i32 addrspace(1)* undef
store volatile i32 %arg2, i32 addrspace(1)* undef
store volatile i32 %arg3, i32 addrspace(1)* undef
store volatile i32 %arg4, i32 addrspace(1)* undef
store volatile i32 %arg5, i32 addrspace(1)* undef
store volatile i32 %arg6, i32 addrspace(1)* undef
store volatile i32 %arg7, i32 addrspace(1)* undef
store volatile i32 %arg8, i32 addrspace(1)* undef
store volatile i32 %arg9, i32 addrspace(1)* undef
store volatile i32 %arg10, i32 addrspace(1)* undef
store volatile i32 %arg11, i32 addrspace(1)* undef
store volatile i32 %arg12, i32 addrspace(1)* undef
store volatile i32 %arg13, i32 addrspace(1)* undef
store volatile i32 %arg14, i32 addrspace(1)* undef
store volatile i32 %arg15, i32 addrspace(1)* undef
store volatile i32 %arg16, i32 addrspace(1)* undef
store volatile i32 %arg17, i32 addrspace(1)* undef
store volatile i32 %arg18, i32 addrspace(1)* undef
store volatile i32 %arg19, i32 addrspace(1)* undef
store volatile i32 %arg20, i32 addrspace(1)* undef
store volatile i32 %arg21, i32 addrspace(1)* undef
store volatile i32 %arg22, i32 addrspace(1)* undef
store volatile i32 %arg23, i32 addrspace(1)* undef
store volatile i32 %arg24, i32 addrspace(1)* undef
store volatile i32 %arg25, i32 addrspace(1)* undef
store volatile i32 %arg26, i32 addrspace(1)* undef
store volatile i32 %arg27, i32 addrspace(1)* undef
store volatile i32 %arg28, i32 addrspace(1)* undef
store volatile i32 %arg29, i32 addrspace(1)* undef
store volatile i32 %arg30, i32 addrspace(1)* undef
store volatile i32 %arg31, i32 addrspace(1)* undef
ret void
}
; GCN-LABEL: {{^}}kern_call_too_many_args_use_workitem_id_x:
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; VARABI: enable_vgpr_workitem_id = 0
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
[AMDGPU] Add Scratch Wave Offset to Scratch Buffer Descriptor in entry functions Add the scratch wave offset to the scratch buffer descriptor (SRSrc) in the entry function prologue. This allows us to removes the scratch wave offset register from the calling convention ABI. As part of this change, allow the use of an inline constant zero for the SOffset of MUBUF instructions accessing the stack in entry functions when a frame pointer is not requested/required. Entry functions with calls still need to set up the calling convention ABI stack pointer register, and reference it in order to address arguments of called functions. The ABI stack pointer register remains unswizzled, but is now wave-relative instead of queue-relative. Non-entry functions also use an inline constant zero SOffset for wave-relative scratch access, but continue to use the stack and frame pointers as before. When the stack or frame pointer is converted to a swizzled offset it is now scaled directly, as the scratch wave offset no longer needs to be subtracted first. Update llvm/docs/AMDGPUUsage.rst to reflect these changes to the calling convention. Tags: #llvm Differential Revision: https://reviews.llvm.org/D75138
2020-01-22 06:27:57 +08:00
; VARABI: s_mov_b32 s32, 0
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; VARABI: buffer_store_dword v0, off, s[0:3], s32{{$}}
; VARABI: s_swappc_b64
; FIXEDABI: enable_vgpr_workitem_id = 2
[AMDGPU] Add Scratch Wave Offset to Scratch Buffer Descriptor in entry functions Add the scratch wave offset to the scratch buffer descriptor (SRSrc) in the entry function prologue. This allows us to removes the scratch wave offset register from the calling convention ABI. As part of this change, allow the use of an inline constant zero for the SOffset of MUBUF instructions accessing the stack in entry functions when a frame pointer is not requested/required. Entry functions with calls still need to set up the calling convention ABI stack pointer register, and reference it in order to address arguments of called functions. The ABI stack pointer register remains unswizzled, but is now wave-relative instead of queue-relative. Non-entry functions also use an inline constant zero SOffset for wave-relative scratch access, but continue to use the stack and frame pointers as before. When the stack or frame pointer is converted to a swizzled offset it is now scaled directly, as the scratch wave offset no longer needs to be subtracted first. Update llvm/docs/AMDGPUUsage.rst to reflect these changes to the calling convention. Tags: #llvm Differential Revision: https://reviews.llvm.org/D75138
2020-01-22 06:27:57 +08:00
; FIXEDABI-DAG: s_mov_b32 s32, 0
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; FIXEDABI-DAG: v_mov_b32_e32 [[K:v[0-9]+]], 0x140{{$}}
; FIXEDABI-DAG: v_lshlrev_b32_e32 [[TMP1:v[0-9]+]], 10, v1
; FIXEDABI-DAG: v_lshlrev_b32_e32 [[TMP0:v[0-9]+]], 20, v2
; FIXEDABI-DAG: v_or_b32_e32 [[TMP2:v[0-9]+]], v0, [[TMP1]]
; FIXEDABI-DAG: v_or_b32_e32 v31, [[TMP2]], [[TMP0]]
; FIXEDABI: buffer_store_dword [[K]], off, s[0:3], s32{{$}}
; FIXEDABI: s_swappc_b64
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
define amdgpu_kernel void @kern_call_too_many_args_use_workitem_id_x() #1 {
call void @too_many_args_use_workitem_id_x(
i32 10, i32 20, i32 30, i32 40,
i32 50, i32 60, i32 70, i32 80,
i32 90, i32 100, i32 110, i32 120,
i32 130, i32 140, i32 150, i32 160,
i32 170, i32 180, i32 190, i32 200,
i32 210, i32 220, i32 230, i32 240,
i32 250, i32 260, i32 270, i32 280,
i32 290, i32 300, i32 310, i32 320)
ret void
}
; GCN-LABEL: {{^}}func_call_too_many_args_use_workitem_id_x:
[AMDGPU] Move frame pointer from s34 to s33 Remove the gap left between the stack pointer (s32) and frame pointer (s34) now that the scratch wave offset is no longer a part of the calling convention ABI. Update llvm/docs/AMDGPUUsage.rst to reflect the change. Tags: #llvm Differential Revision: https://reviews.llvm.org/D75657
2020-03-04 09:39:47 +08:00
; VARABI: s_mov_b32 s33, s32
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; VARABI: buffer_store_dword v1, off, s[0:3], s32{{$}}
; Touching the workitem id register is not necessary.
; FIXEDABI-NOT: v31
; FIXEDABI: v_mov_b32_e32 [[K:v[0-9]+]], 0x140{{$}}
; FIXEDABI-NOT: v31
; FIXEDABI: buffer_store_dword [[K]], off, s[0:3], s32{{$}}
; FIXEDABI-NOT: v31
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
; GCN: s_swappc_b64
define void @func_call_too_many_args_use_workitem_id_x(i32 %arg0) #1 {
store volatile i32 %arg0, i32 addrspace(1)* undef
call void @too_many_args_use_workitem_id_x(
i32 10, i32 20, i32 30, i32 40,
i32 50, i32 60, i32 70, i32 80,
i32 90, i32 100, i32 110, i32 120,
i32 130, i32 140, i32 150, i32 160,
i32 170, i32 180, i32 190, i32 200,
i32 210, i32 220, i32 230, i32 240,
i32 250, i32 260, i32 270, i32 280,
i32 290, i32 300, i32 310, i32 320)
ret void
}
; Requires loading and storing to stack slot.
; GCN-LABEL: {{^}}too_many_args_call_too_many_args_use_workitem_id_x:
AMDGPU: Make s34 the FP register Make the FP register callee saved. This is tricky because now the FP needs to be spilled in the prolog relative to the incoming SP register, rather than the frame register used throughout the rest of the function. I don't like how this bypassess the standard mechanism for CSR spills just to get the correct insert point. I may look for a better solution, since all CSR VGPRs may also need to have all lanes activated. Another option might be to make getFrameIndexReference change the base register if the frame index is a CSR, and then try to figure out the right insertion point in emitProlog. If there is a free VGPR lane available for SGPR spilling, try to use it for the FP. If that would require intrtoducing a new VGPR spill, try to use a free call clobbered SGPR. Only fallback to introducing a new VGPR spill as a last resort. This also doesn't attempt to handle SGPR spilling with scalar stores. llvm-svn: 365372
2019-07-09 03:03:38 +08:00
; GCN-DAG: s_add_u32 s32, s32, 0x400{{$}}
[AMDGPU] Introduce more scratch registers in the ABI. The AMDGPU target has a convention that defined all VGPRs (execept the initial 32 argument registers) as callee-saved. This convention is not efficient always, esp. when the callee requiring more registers, ended up emitting a large number of spills, even though its caller requires only a few. This patch revises the ABI by introducing more scratch registers that a callee can freely use. The 256 vgpr registers now become: 32 argument registers 112 scratch registers and 112 callee saved registers. The scratch registers and the CSRs are intermixed at regular intervals (a split boundary of 8) to obtain a better occupancy. Reviewers: arsenm, t-tye, rampitec, b-sumner, mjbedy, tpr Reviewed By: arsenm, t-tye Differential Revision: https://reviews.llvm.org/D76356
2020-03-27 15:46:51 +08:00
; GCN-DAG: buffer_store_dword v40, off, s[0:3], s32 offset:4 ; 4-byte Folded Spill
[AMDGPU] Move frame pointer from s34 to s33 Remove the gap left between the stack pointer (s32) and frame pointer (s34) now that the scratch wave offset is no longer a part of the calling convention ABI. Update llvm/docs/AMDGPUUsage.rst to reflect the change. Tags: #llvm Differential Revision: https://reviews.llvm.org/D75657
2020-03-04 09:39:47 +08:00
; GCN-DAG: buffer_load_dword v32, off, s[0:3], s33{{$}}
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
AMDGPU: Don't fix emergency stack slot at offset 0 This forced the caller to be aware of this, which is an ugly ABI feature. Partially reverts r295877. The original reasons for doing this are mostly fixed. Alloca is now in a non-0 address space, so it should be OK to have 0 as a valid pointer. Since we treat the absolute address as the pointer value, this part only really needed to apply to kernels. Since r357093, we avoid the need to increment/decrement the offset register in more cases, and since r354816 the scavenger can fail without spilling, so it's less critical that we try to avoid an offset that fits in the MUBUF offset. Restrict to callable functions for now to split this into 2 steps to limit thte number of test updates and in case anything breaks. llvm-svn: 362665
2019-06-06 06:37:50 +08:00
; GCN: buffer_store_dword v32, off, s[0:3], s32{{$}}
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
; GCN: s_swappc_b64
; GCN: s_sub_u32 s32, s32, 0x400{{$}}
[AMDGPU] Introduce more scratch registers in the ABI. The AMDGPU target has a convention that defined all VGPRs (execept the initial 32 argument registers) as callee-saved. This convention is not efficient always, esp. when the callee requiring more registers, ended up emitting a large number of spills, even though its caller requires only a few. This patch revises the ABI by introducing more scratch registers that a callee can freely use. The 256 vgpr registers now become: 32 argument registers 112 scratch registers and 112 callee saved registers. The scratch registers and the CSRs are intermixed at regular intervals (a split boundary of 8) to obtain a better occupancy. Reviewers: arsenm, t-tye, rampitec, b-sumner, mjbedy, tpr Reviewed By: arsenm, t-tye Differential Revision: https://reviews.llvm.org/D76356
2020-03-27 15:46:51 +08:00
; GCN: buffer_load_dword v40, off, s[0:3], s32 offset:4 ; 4-byte Folded Reload
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
; GCN: s_setpc_b64
define void @too_many_args_call_too_many_args_use_workitem_id_x(
i32 %arg0, i32 %arg1, i32 %arg2, i32 %arg3, i32 %arg4, i32 %arg5, i32 %arg6, i32 %arg7,
i32 %arg8, i32 %arg9, i32 %arg10, i32 %arg11, i32 %arg12, i32 %arg13, i32 %arg14, i32 %arg15,
i32 %arg16, i32 %arg17, i32 %arg18, i32 %arg19, i32 %arg20, i32 %arg21, i32 %arg22, i32 %arg23,
i32 %arg24, i32 %arg25, i32 %arg26, i32 %arg27, i32 %arg28, i32 %arg29, i32 %arg30, i32 %arg31) #1 {
call void @too_many_args_use_workitem_id_x(
i32 %arg0, i32 %arg1, i32 %arg2, i32 %arg3, i32 %arg4, i32 %arg5, i32 %arg6, i32 %arg7,
i32 %arg8, i32 %arg9, i32 %arg10, i32 %arg11, i32 %arg12, i32 %arg13, i32 %arg14, i32 %arg15,
i32 %arg16, i32 %arg17, i32 %arg18, i32 %arg19, i32 %arg20, i32 %arg21, i32 %arg22, i32 %arg23,
i32 %arg24, i32 %arg25, i32 %arg26, i32 %arg27, i32 %arg28, i32 %arg29, i32 %arg30, i32 %arg31)
ret void
}
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; var abi stack layout:
AMDGPU: Don't fix emergency stack slot at offset 0 This forced the caller to be aware of this, which is an ugly ABI feature. Partially reverts r295877. The original reasons for doing this are mostly fixed. Alloca is now in a non-0 address space, so it should be OK to have 0 as a valid pointer. Since we treat the absolute address as the pointer value, this part only really needed to apply to kernels. Since r357093, we avoid the need to increment/decrement the offset register in more cases, and since r354816 the scavenger can fail without spilling, so it's less critical that we try to avoid an offset that fits in the MUBUF offset. Restrict to callable functions for now to split this into 2 steps to limit thte number of test updates and in case anything breaks. llvm-svn: 362665
2019-06-06 06:37:50 +08:00
; frame[0] = byval arg32
; frame[1] = stack passed workitem ID x
; frame[2] = VGPR spill slot
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
; GCN-LABEL: {{^}}too_many_args_use_workitem_id_x_byval:
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; VARABI: buffer_load_dword v32, off, s[0:3], s32 offset:4
; VARABI-NEXT: s_waitcnt
; VARABI-NEXT: v_and_b32_e32 v32, 0x3ff, v32
; VARABI-NEXT: {{flat|global}}_store_dword v{{\[[0-9]+:[0-9]+\]}}, v32
; VARABI: buffer_load_dword v0, off, s[0:3], s32{{$}}
; VARABI: s_setpc_b64
; FIXEDABI: v_and_b32_e32 v31, 0x3ff, v31
; FIXEDABI-NEXT: {{flat|global}}_store_dword v{{\[[0-9]+:[0-9]+\]}}, v31
; FIXEDABI: buffer_load_dword v0, off, s[0:3], s32{{$}}
; FIXEDABI: {{flat|global}}_store_dword v{{\[[0-9]+:[0-9]+\]}}, v0
; FIXEDABI: buffer_load_dword v0, off, s[0:3], s32 offset:4{{$}}
; FIXEDABI: s_setpc_b64
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
define void @too_many_args_use_workitem_id_x_byval(
i32 %arg0, i32 %arg1, i32 %arg2, i32 %arg3, i32 %arg4, i32 %arg5, i32 %arg6, i32 %arg7,
i32 %arg8, i32 %arg9, i32 %arg10, i32 %arg11, i32 %arg12, i32 %arg13, i32 %arg14, i32 %arg15,
i32 %arg16, i32 %arg17, i32 %arg18, i32 %arg19, i32 %arg20, i32 %arg21, i32 %arg22, i32 %arg23,
[AMDGPU] Switch to the new addr space mapping by default This requires corresponding clang change. Differential Revision: https://reviews.llvm.org/D40955 llvm-svn: 324101
2018-02-03 00:07:16 +08:00
i32 %arg24, i32 %arg25, i32 %arg26, i32 %arg27, i32 %arg28, i32 %arg29, i32 %arg30, i32 %arg31, i32 addrspace(5)* byval %arg32) #1 {
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
%val = call i32 @llvm.amdgcn.workitem.id.x()
store volatile i32 %val, i32 addrspace(1)* undef
store volatile i32 %arg0, i32 addrspace(1)* undef
store volatile i32 %arg1, i32 addrspace(1)* undef
store volatile i32 %arg2, i32 addrspace(1)* undef
store volatile i32 %arg3, i32 addrspace(1)* undef
store volatile i32 %arg4, i32 addrspace(1)* undef
store volatile i32 %arg5, i32 addrspace(1)* undef
store volatile i32 %arg6, i32 addrspace(1)* undef
store volatile i32 %arg7, i32 addrspace(1)* undef
store volatile i32 %arg8, i32 addrspace(1)* undef
store volatile i32 %arg9, i32 addrspace(1)* undef
store volatile i32 %arg10, i32 addrspace(1)* undef
store volatile i32 %arg11, i32 addrspace(1)* undef
store volatile i32 %arg12, i32 addrspace(1)* undef
store volatile i32 %arg13, i32 addrspace(1)* undef
store volatile i32 %arg14, i32 addrspace(1)* undef
store volatile i32 %arg15, i32 addrspace(1)* undef
store volatile i32 %arg16, i32 addrspace(1)* undef
store volatile i32 %arg17, i32 addrspace(1)* undef
store volatile i32 %arg18, i32 addrspace(1)* undef
store volatile i32 %arg19, i32 addrspace(1)* undef
store volatile i32 %arg20, i32 addrspace(1)* undef
store volatile i32 %arg21, i32 addrspace(1)* undef
store volatile i32 %arg22, i32 addrspace(1)* undef
store volatile i32 %arg23, i32 addrspace(1)* undef
store volatile i32 %arg24, i32 addrspace(1)* undef
store volatile i32 %arg25, i32 addrspace(1)* undef
store volatile i32 %arg26, i32 addrspace(1)* undef
store volatile i32 %arg27, i32 addrspace(1)* undef
store volatile i32 %arg28, i32 addrspace(1)* undef
store volatile i32 %arg29, i32 addrspace(1)* undef
store volatile i32 %arg30, i32 addrspace(1)* undef
store volatile i32 %arg31, i32 addrspace(1)* undef
[AMDGPU] Switch to the new addr space mapping by default This requires corresponding clang change. Differential Revision: https://reviews.llvm.org/D40955 llvm-svn: 324101
2018-02-03 00:07:16 +08:00
%private = load volatile i32, i32 addrspace(5)* %arg32
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
ret void
}
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; var abi stack layout:
AMDGPU: Don't fix emergency stack slot at offset 0 This forced the caller to be aware of this, which is an ugly ABI feature. Partially reverts r295877. The original reasons for doing this are mostly fixed. Alloca is now in a non-0 address space, so it should be OK to have 0 as a valid pointer. Since we treat the absolute address as the pointer value, this part only really needed to apply to kernels. Since r357093, we avoid the need to increment/decrement the offset register in more cases, and since r354816 the scavenger can fail without spilling, so it's less critical that we try to avoid an offset that fits in the MUBUF offset. Restrict to callable functions for now to split this into 2 steps to limit thte number of test updates and in case anything breaks. llvm-svn: 362665
2019-06-06 06:37:50 +08:00
; sp[0] = byval
; sp[1] = ??
; sp[2] = stack passed workitem ID x
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
; GCN-LABEL: {{^}}kern_call_too_many_args_use_workitem_id_x_byval:
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; VARABI: enable_vgpr_workitem_id = 0
[AMDGPU] Add Scratch Wave Offset to Scratch Buffer Descriptor in entry functions Add the scratch wave offset to the scratch buffer descriptor (SRSrc) in the entry function prologue. This allows us to removes the scratch wave offset register from the calling convention ABI. As part of this change, allow the use of an inline constant zero for the SOffset of MUBUF instructions accessing the stack in entry functions when a frame pointer is not requested/required. Entry functions with calls still need to set up the calling convention ABI stack pointer register, and reference it in order to address arguments of called functions. The ABI stack pointer register remains unswizzled, but is now wave-relative instead of queue-relative. Non-entry functions also use an inline constant zero SOffset for wave-relative scratch access, but continue to use the stack and frame pointers as before. When the stack or frame pointer is converted to a swizzled offset it is now scaled directly, as the scratch wave offset no longer needs to be subtracted first. Update llvm/docs/AMDGPUUsage.rst to reflect these changes to the calling convention. Tags: #llvm Differential Revision: https://reviews.llvm.org/D75138
2020-01-22 06:27:57 +08:00
; VARABI: v_mov_b32_e32 [[K:v[0-9]+]], 0x3e7{{$}}
; VARABI: buffer_store_dword [[K]], off, s[0:3], 0 offset:4
; VARABI: buffer_load_dword [[RELOAD_BYVAL:v[0-9]+]], off, s[0:3], 0 offset:4
; VARABI: s_movk_i32 s32, 0x400{{$}}
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; VARABI-NOT: s32
; VARABI: buffer_store_dword v0, off, s[0:3], s32 offset:4
; VARABI: buffer_store_dword [[RELOAD_BYVAL]], off, s[0:3], s32{{$}}
; VARABI: v_mov_b32_e32 [[RELOAD_BYVAL]],
; VARABI: s_swappc_b64
[AMDGPU] Add Scratch Wave Offset to Scratch Buffer Descriptor in entry functions Add the scratch wave offset to the scratch buffer descriptor (SRSrc) in the entry function prologue. This allows us to removes the scratch wave offset register from the calling convention ABI. As part of this change, allow the use of an inline constant zero for the SOffset of MUBUF instructions accessing the stack in entry functions when a frame pointer is not requested/required. Entry functions with calls still need to set up the calling convention ABI stack pointer register, and reference it in order to address arguments of called functions. The ABI stack pointer register remains unswizzled, but is now wave-relative instead of queue-relative. Non-entry functions also use an inline constant zero SOffset for wave-relative scratch access, but continue to use the stack and frame pointers as before. When the stack or frame pointer is converted to a swizzled offset it is now scaled directly, as the scratch wave offset no longer needs to be subtracted first. Update llvm/docs/AMDGPUUsage.rst to reflect these changes to the calling convention. Tags: #llvm Differential Revision: https://reviews.llvm.org/D75138
2020-01-22 06:27:57 +08:00
; FIXEDABI: v_mov_b32_e32 [[K0:v[0-9]+]], 0x3e7
; FIXEDABI: s_movk_i32 s32, 0x400{{$}}
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; FIXEDABI: v_mov_b32_e32 [[K1:v[0-9]+]], 0x140
[Scheduling] Create the missing dependency edges for store cluster If it is load cluster, we don't need to create the dependency edges(SUb->reg) from SUb to SUa as they both depend on the base register "reg" +-------+ +----> reg | | +---+---+ | ^ | | | | | | | +---+---+ | | SUa | Load 0(reg) | +---+---+ | ^ | | | | | +---+---+ +----+ SUb | Load 4(reg) +-------+ But if it is store cluster, we need to create it as follow shows to avoid the instruction store depend on scheduled in-between SUb and SUa. +-------+ +----> reg | | +---+---+ | ^ | | Missing +-------+ | | +-------------------->+ y | | | | +---+---+ | +---+-+-+ ^ | | SUa | Store x 0(reg) | | +---+---+ | | ^ | | | +------------------------+ | | | | +---+--++ +----+ SUb | Store y 4(reg) +-------+ Reviewed By: evandro, arsenm, rampitec, foad, fhahn Differential Revision: https://reviews.llvm.org/D72031
2020-08-07 12:53:37 +08:00
; FIXEDABI: buffer_store_dword [[K0]], off, s[0:3], 0 offset:4{{$}}
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; FIXEDABI: buffer_store_dword [[K1]], off, s[0:3], s32{{$}}
; FIXME: Why this reload?
[AMDGPU] Add Scratch Wave Offset to Scratch Buffer Descriptor in entry functions Add the scratch wave offset to the scratch buffer descriptor (SRSrc) in the entry function prologue. This allows us to removes the scratch wave offset register from the calling convention ABI. As part of this change, allow the use of an inline constant zero for the SOffset of MUBUF instructions accessing the stack in entry functions when a frame pointer is not requested/required. Entry functions with calls still need to set up the calling convention ABI stack pointer register, and reference it in order to address arguments of called functions. The ABI stack pointer register remains unswizzled, but is now wave-relative instead of queue-relative. Non-entry functions also use an inline constant zero SOffset for wave-relative scratch access, but continue to use the stack and frame pointers as before. When the stack or frame pointer is converted to a swizzled offset it is now scaled directly, as the scratch wave offset no longer needs to be subtracted first. Update llvm/docs/AMDGPUUsage.rst to reflect these changes to the calling convention. Tags: #llvm Differential Revision: https://reviews.llvm.org/D75138
2020-01-22 06:27:57 +08:00
; FIXEDABI: buffer_load_dword [[RELOAD:v[0-9]+]], off, s[0:3], 0 offset:4{{$}}
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; FIXEDABI-DAG: v_lshlrev_b32_e32 [[TMP1:v[0-9]+]], 10, v1
; FIXEDABI-DAG: v_lshlrev_b32_e32 [[TMP0:v[0-9]+]], 20, v2
; FIXEDABI-DAG: v_or_b32_e32 [[TMP2:v[0-9]+]], v0, [[TMP1]]
; FIXEDABI: v_or_b32_e32 v31, [[TMP2]], [[TMP0]]
; FIXEDABI-NOT: s32
; FIXEDABI: buffer_store_dword [[RELOAD]], off, s[0:3], s32 offset:4
; FIXEDABI: s_swappc_b64
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
define amdgpu_kernel void @kern_call_too_many_args_use_workitem_id_x_byval() #1 {
[AMDGPU] Switch to the new addr space mapping by default This requires corresponding clang change. Differential Revision: https://reviews.llvm.org/D40955 llvm-svn: 324101
2018-02-03 00:07:16 +08:00
%alloca = alloca i32, align 4, addrspace(5)
store volatile i32 999, i32 addrspace(5)* %alloca
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
call void @too_many_args_use_workitem_id_x_byval(
i32 10, i32 20, i32 30, i32 40,
i32 50, i32 60, i32 70, i32 80,
i32 90, i32 100, i32 110, i32 120,
i32 130, i32 140, i32 150, i32 160,
i32 170, i32 180, i32 190, i32 200,
i32 210, i32 220, i32 230, i32 240,
i32 250, i32 260, i32 270, i32 280,
i32 290, i32 300, i32 310, i32 320,
[AMDGPU] Switch to the new addr space mapping by default This requires corresponding clang change. Differential Revision: https://reviews.llvm.org/D40955 llvm-svn: 324101
2018-02-03 00:07:16 +08:00
i32 addrspace(5)* %alloca)
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
ret void
}
; GCN-LABEL: {{^}}func_call_too_many_args_use_workitem_id_x_byval:
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; VARABI: v_mov_b32_e32 [[K:v[0-9]+]], 0x3e7{{$}}
[AMDGPU] Move frame pointer from s34 to s33 Remove the gap left between the stack pointer (s32) and frame pointer (s34) now that the scratch wave offset is no longer a part of the calling convention ABI. Update llvm/docs/AMDGPUUsage.rst to reflect the change. Tags: #llvm Differential Revision: https://reviews.llvm.org/D75657
2020-03-04 09:39:47 +08:00
; VARABI: buffer_store_dword [[K]], off, s[0:3], s33{{$}}
; VARABI: buffer_load_dword [[RELOAD_BYVAL:v[0-9]+]], off, s[0:3], s33{{$}}
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; VARABI: buffer_store_dword v0, off, s[0:3], s32 offset:4
; VARABI: buffer_store_dword [[RELOAD_BYVAL]], off, s[0:3], s32{{$}}
; VARABI: v_mov_b32_e32 [[RELOAD_BYVAL]],
; VARABI: s_swappc_b64
; FIXED-ABI-NOT: v31
; FIXEDABI: v_mov_b32_e32 [[K0:v[0-9]+]], 0x3e7{{$}}
; FIXEDABI: v_mov_b32_e32 [[K1:v[0-9]+]], 0x140{{$}}
[Scheduling] Create the missing dependency edges for store cluster If it is load cluster, we don't need to create the dependency edges(SUb->reg) from SUb to SUa as they both depend on the base register "reg" +-------+ +----> reg | | +---+---+ | ^ | | | | | | | +---+---+ | | SUa | Load 0(reg) | +---+---+ | ^ | | | | | +---+---+ +----+ SUb | Load 4(reg) +-------+ But if it is store cluster, we need to create it as follow shows to avoid the instruction store depend on scheduled in-between SUb and SUa. +-------+ +----> reg | | +---+---+ | ^ | | Missing +-------+ | | +-------------------->+ y | | | | +---+---+ | +---+-+-+ ^ | | SUa | Store x 0(reg) | | +---+---+ | | ^ | | | +------------------------+ | | | | +---+--++ +----+ SUb | Store y 4(reg) +-------+ Reviewed By: evandro, arsenm, rampitec, foad, fhahn Differential Revision: https://reviews.llvm.org/D72031
2020-08-07 12:53:37 +08:00
; FIXEDABI: buffer_store_dword [[K0]], off, s[0:3], s33{{$}}
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; FIXEDABI: buffer_store_dword [[K1]], off, s[0:3], s32{{$}}
[AMDGPU] Move frame pointer from s34 to s33 Remove the gap left between the stack pointer (s32) and frame pointer (s34) now that the scratch wave offset is no longer a part of the calling convention ABI. Update llvm/docs/AMDGPUUsage.rst to reflect the change. Tags: #llvm Differential Revision: https://reviews.llvm.org/D75657
2020-03-04 09:39:47 +08:00
; FIXEDABI: buffer_load_dword [[RELOAD_BYVAL:v[0-9]+]], off, s[0:3], s33{{$}}
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; FIXED-ABI-NOT: v31
; FIXEDABI: buffer_store_dword [[RELOAD_BYVAL]], off, s[0:3], s32 offset:4{{$}}
; FIXED-ABI-NOT: v31
; FIXEDABI: s_swappc_b64
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
define void @func_call_too_many_args_use_workitem_id_x_byval() #1 {
[AMDGPU] Switch to the new addr space mapping by default This requires corresponding clang change. Differential Revision: https://reviews.llvm.org/D40955 llvm-svn: 324101
2018-02-03 00:07:16 +08:00
%alloca = alloca i32, align 4, addrspace(5)
store volatile i32 999, i32 addrspace(5)* %alloca
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
call void @too_many_args_use_workitem_id_x_byval(
i32 10, i32 20, i32 30, i32 40,
i32 50, i32 60, i32 70, i32 80,
i32 90, i32 100, i32 110, i32 120,
i32 130, i32 140, i32 150, i32 160,
i32 170, i32 180, i32 190, i32 200,
i32 210, i32 220, i32 230, i32 240,
i32 250, i32 260, i32 270, i32 280,
i32 290, i32 300, i32 310, i32 320,
[AMDGPU] Switch to the new addr space mapping by default This requires corresponding clang change. Differential Revision: https://reviews.llvm.org/D40955 llvm-svn: 324101
2018-02-03 00:07:16 +08:00
i32 addrspace(5)* %alloca)
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
ret void
}
; GCN-LABEL: {{^}}too_many_args_use_workitem_id_xyz:
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; VARABI-NOT: buffer_load_dword v{{[0-9]+}}, off, s[0:3], s32{{$}}
; VARABI: buffer_load_dword v32, off, s[0:3], s32{{$}}
; VARABI-NOT: buffer_load_dword
; VARABI: v_and_b32_e32 [[AND_X:v[0-9]+]], 0x3ff, v32
; VARABI-NOT: buffer_load_dword
; VARABI: {{flat|global}}_store_dword v{{\[[0-9]+:[0-9]+]}}, [[AND_X]]
; VARABI-NOT: buffer_load_dword
; VARABI: v_bfe_u32 [[BFE_Y:v[0-9]+]], v32, 10, 10
; VARABI-NEXT: v_bfe_u32 [[BFE_Z:v[0-9]+]], v32, 20, 10
; VARABI-NEXT: {{flat|global}}_store_dword v{{\[[0-9]+:[0-9]+]}}, [[BFE_Y]]
; VARABI-NEXT: {{flat|global}}_store_dword v{{\[[0-9]+:[0-9]+]}}, [[BFE_Z]]
[AMDGPU] Introduce more scratch registers in the ABI. The AMDGPU target has a convention that defined all VGPRs (execept the initial 32 argument registers) as callee-saved. This convention is not efficient always, esp. when the callee requiring more registers, ended up emitting a large number of spills, even though its caller requires only a few. This patch revises the ABI by introducing more scratch registers that a callee can freely use. The 256 vgpr registers now become: 32 argument registers 112 scratch registers and 112 callee saved registers. The scratch registers and the CSRs are intermixed at regular intervals (a split boundary of 8) to obtain a better occupancy. Reviewers: arsenm, t-tye, rampitec, b-sumner, mjbedy, tpr Reviewed By: arsenm, t-tye Differential Revision: https://reviews.llvm.org/D76356
2020-03-27 15:46:51 +08:00
; VARABI: s_waitcnt
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; VARABI-NEXT: s_setpc_b64
; FIXEDABI: v_and_b32_e32 [[AND_X:v[0-9]+]], 0x3ff, v31
; FIXEDABI-NOT: buffer_load_dword
; FIXEDABI: {{flat|global}}_store_dword v{{\[[0-9]+:[0-9]+]}}, [[AND_X]]
; FIXEDABI-NOT: buffer_load_dword
; FIXEDABI: v_bfe_u32 [[BFE_Y:v[0-9]+]], v31, 10, 10
; FIXEDABI-NEXT: v_bfe_u32 [[BFE_Z:v[0-9]+]], v31, 20, 10
; FIXEDABI-NEXT: {{flat|global}}_store_dword v{{\[[0-9]+:[0-9]+]}}, [[BFE_Y]]
; FIXEDABI-NEXT: {{flat|global}}_store_dword v{{\[[0-9]+:[0-9]+]}}, [[BFE_Z]]
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
define void @too_many_args_use_workitem_id_xyz(
i32 %arg0, i32 %arg1, i32 %arg2, i32 %arg3, i32 %arg4, i32 %arg5, i32 %arg6, i32 %arg7,
i32 %arg8, i32 %arg9, i32 %arg10, i32 %arg11, i32 %arg12, i32 %arg13, i32 %arg14, i32 %arg15,
i32 %arg16, i32 %arg17, i32 %arg18, i32 %arg19, i32 %arg20, i32 %arg21, i32 %arg22, i32 %arg23,
i32 %arg24, i32 %arg25, i32 %arg26, i32 %arg27, i32 %arg28, i32 %arg29, i32 %arg30, i32 %arg31) #1 {
%val0 = call i32 @llvm.amdgcn.workitem.id.x()
store volatile i32 %val0, i32 addrspace(1)* undef
%val1 = call i32 @llvm.amdgcn.workitem.id.y()
store volatile i32 %val1, i32 addrspace(1)* undef
%val2 = call i32 @llvm.amdgcn.workitem.id.z()
store volatile i32 %val2, i32 addrspace(1)* undef
store volatile i32 %arg0, i32 addrspace(1)* undef
store volatile i32 %arg1, i32 addrspace(1)* undef
store volatile i32 %arg2, i32 addrspace(1)* undef
store volatile i32 %arg3, i32 addrspace(1)* undef
store volatile i32 %arg4, i32 addrspace(1)* undef
store volatile i32 %arg5, i32 addrspace(1)* undef
store volatile i32 %arg6, i32 addrspace(1)* undef
store volatile i32 %arg7, i32 addrspace(1)* undef
store volatile i32 %arg8, i32 addrspace(1)* undef
store volatile i32 %arg9, i32 addrspace(1)* undef
store volatile i32 %arg10, i32 addrspace(1)* undef
store volatile i32 %arg11, i32 addrspace(1)* undef
store volatile i32 %arg12, i32 addrspace(1)* undef
store volatile i32 %arg13, i32 addrspace(1)* undef
store volatile i32 %arg14, i32 addrspace(1)* undef
store volatile i32 %arg15, i32 addrspace(1)* undef
store volatile i32 %arg16, i32 addrspace(1)* undef
store volatile i32 %arg17, i32 addrspace(1)* undef
store volatile i32 %arg18, i32 addrspace(1)* undef
store volatile i32 %arg19, i32 addrspace(1)* undef
store volatile i32 %arg20, i32 addrspace(1)* undef
store volatile i32 %arg21, i32 addrspace(1)* undef
store volatile i32 %arg22, i32 addrspace(1)* undef
store volatile i32 %arg23, i32 addrspace(1)* undef
store volatile i32 %arg24, i32 addrspace(1)* undef
store volatile i32 %arg25, i32 addrspace(1)* undef
store volatile i32 %arg26, i32 addrspace(1)* undef
store volatile i32 %arg27, i32 addrspace(1)* undef
store volatile i32 %arg28, i32 addrspace(1)* undef
store volatile i32 %arg29, i32 addrspace(1)* undef
store volatile i32 %arg30, i32 addrspace(1)* undef
store volatile i32 %arg31, i32 addrspace(1)* undef
ret void
}
; GCN-LABEL: {{^}}kern_call_too_many_args_use_workitem_id_xyz:
; GCN: enable_vgpr_workitem_id = 2
[AMDGPU] Add Scratch Wave Offset to Scratch Buffer Descriptor in entry functions Add the scratch wave offset to the scratch buffer descriptor (SRSrc) in the entry function prologue. This allows us to removes the scratch wave offset register from the calling convention ABI. As part of this change, allow the use of an inline constant zero for the SOffset of MUBUF instructions accessing the stack in entry functions when a frame pointer is not requested/required. Entry functions with calls still need to set up the calling convention ABI stack pointer register, and reference it in order to address arguments of called functions. The ABI stack pointer register remains unswizzled, but is now wave-relative instead of queue-relative. Non-entry functions also use an inline constant zero SOffset for wave-relative scratch access, but continue to use the stack and frame pointers as before. When the stack or frame pointer is converted to a swizzled offset it is now scaled directly, as the scratch wave offset no longer needs to be subtracted first. Update llvm/docs/AMDGPUUsage.rst to reflect these changes to the calling convention. Tags: #llvm Differential Revision: https://reviews.llvm.org/D75138
2020-01-22 06:27:57 +08:00
; GCN-DAG: s_mov_b32 s32, 0
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; GCN-DAG: v_lshlrev_b32_e32 [[TMP1:v[0-9]+]], 10, v1
; GCN-DAG: v_lshlrev_b32_e32 [[TMP0:v[0-9]+]], 20, v2
; GCN-DAG: v_or_b32_e32 [[TMP2:v[0-9]+]], v0, [[TMP1]]
; VARABI-DAG: v_or_b32_e32 [[PACKEDID:v[0-9]+]], [[TMP2]], [[TMP0]]
; VARABI: buffer_store_dword [[PACKEDID]], off, s[0:3], s32{{$}}
; FIXEDABI-DAG: v_or_b32_e32 v31, [[TMP2]], [[TMP0]]
; FIXEDABI-DAG: v_mov_b32_e32 [[K:v[0-9]+]], 0x140
; FIXEDABI: buffer_store_dword [[K]], off, s[0:3], s32{{$}}
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
; GCN: s_swappc_b64
define amdgpu_kernel void @kern_call_too_many_args_use_workitem_id_xyz() #1 {
call void @too_many_args_use_workitem_id_xyz(
i32 10, i32 20, i32 30, i32 40,
i32 50, i32 60, i32 70, i32 80,
i32 90, i32 100, i32 110, i32 120,
i32 130, i32 140, i32 150, i32 160,
i32 170, i32 180, i32 190, i32 200,
i32 210, i32 220, i32 230, i32 240,
i32 250, i32 260, i32 270, i32 280,
i32 290, i32 300, i32 310, i32 320)
ret void
}
AMDGPU: Add flag to used fixed function ABI Pass all arguments to every function, rather than only passing the minimum set of inputs needed for the call graph.
2020-03-12 04:13:52 +08:00
; Var abi: workitem ID X in register, yz on stack
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
; v31 = workitem ID X
[AMDGPU] Packed thread ids in function call ABI Differential Revision: https://reviews.llvm.org/D63851 llvm-svn: 364619
2019-06-28 09:52:13 +08:00
; frame[0] = workitem { Z, Y, X }
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
; GCN-LABEL: {{^}}too_many_args_use_workitem_id_x_stack_yz:
[AMDGPU] Packed thread ids in function call ABI Differential Revision: https://reviews.llvm.org/D63851 llvm-svn: 364619
2019-06-28 09:52:13 +08:00
; GCN-DAG: v_and_b32_e32 [[IDX:v[0-9]+]], 0x3ff, v31
[AMDGPU] Test update. NFC. llvm-svn: 366715
2019-07-23 02:08:53 +08:00
; GCN-DAG: {{flat|global}}_store_dword v[0:1], [[IDX]]
[AMDGPU] Packed thread ids in function call ABI Differential Revision: https://reviews.llvm.org/D63851 llvm-svn: 364619
2019-06-28 09:52:13 +08:00
; GCN-DAG: v_bfe_u32 [[IDY:v[0-9]+]], v31, 10, 10
[AMDGPU] Test update. NFC. llvm-svn: 366715
2019-07-23 02:08:53 +08:00
; GCN-DAG: {{flat|global}}_store_dword v{{\[[0-9]+:[0-9]+]}}, [[IDY]]
[AMDGPU] Packed thread ids in function call ABI Differential Revision: https://reviews.llvm.org/D63851 llvm-svn: 364619
2019-06-28 09:52:13 +08:00
; GCN-DAG: v_bfe_u32 [[IDZ:v[0-9]+]], v31, 20, 10
[AMDGPU] Test update. NFC. llvm-svn: 366715
2019-07-23 02:08:53 +08:00
; GCN-DAG: {{flat|global}}_store_dword v{{\[[0-9]+:[0-9]+]}}, [[IDZ]]
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
; GCN: s_waitcnt
; GCN-NEXT: s_setpc_b64
[AMDGPU] Introduce more scratch registers in the ABI. The AMDGPU target has a convention that defined all VGPRs (execept the initial 32 argument registers) as callee-saved. This convention is not efficient always, esp. when the callee requiring more registers, ended up emitting a large number of spills, even though its caller requires only a few. This patch revises the ABI by introducing more scratch registers that a callee can freely use. The 256 vgpr registers now become: 32 argument registers 112 scratch registers and 112 callee saved registers. The scratch registers and the CSRs are intermixed at regular intervals (a split boundary of 8) to obtain a better occupancy. Reviewers: arsenm, t-tye, rampitec, b-sumner, mjbedy, tpr Reviewed By: arsenm, t-tye Differential Revision: https://reviews.llvm.org/D76356
2020-03-27 15:46:51 +08:00
; GCN: ScratchSize: 0
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
define void @too_many_args_use_workitem_id_x_stack_yz(
i32 %arg0, i32 %arg1, i32 %arg2, i32 %arg3, i32 %arg4, i32 %arg5, i32 %arg6, i32 %arg7,
i32 %arg8, i32 %arg9, i32 %arg10, i32 %arg11, i32 %arg12, i32 %arg13, i32 %arg14, i32 %arg15,
i32 %arg16, i32 %arg17, i32 %arg18, i32 %arg19, i32 %arg20, i32 %arg21, i32 %arg22, i32 %arg23,
i32 %arg24, i32 %arg25, i32 %arg26, i32 %arg27, i32 %arg28, i32 %arg29, i32 %arg30) #1 {
%val0 = call i32 @llvm.amdgcn.workitem.id.x()
store volatile i32 %val0, i32 addrspace(1)* undef
%val1 = call i32 @llvm.amdgcn.workitem.id.y()
store volatile i32 %val1, i32 addrspace(1)* undef
%val2 = call i32 @llvm.amdgcn.workitem.id.z()
store volatile i32 %val2, i32 addrspace(1)* undef
store volatile i32 %arg0, i32 addrspace(1)* undef
store volatile i32 %arg1, i32 addrspace(1)* undef
store volatile i32 %arg2, i32 addrspace(1)* undef
store volatile i32 %arg3, i32 addrspace(1)* undef
store volatile i32 %arg4, i32 addrspace(1)* undef
store volatile i32 %arg5, i32 addrspace(1)* undef
store volatile i32 %arg6, i32 addrspace(1)* undef
store volatile i32 %arg7, i32 addrspace(1)* undef
store volatile i32 %arg8, i32 addrspace(1)* undef
store volatile i32 %arg9, i32 addrspace(1)* undef
store volatile i32 %arg10, i32 addrspace(1)* undef
store volatile i32 %arg11, i32 addrspace(1)* undef
store volatile i32 %arg12, i32 addrspace(1)* undef
store volatile i32 %arg13, i32 addrspace(1)* undef
store volatile i32 %arg14, i32 addrspace(1)* undef
store volatile i32 %arg15, i32 addrspace(1)* undef
store volatile i32 %arg16, i32 addrspace(1)* undef
store volatile i32 %arg17, i32 addrspace(1)* undef
store volatile i32 %arg18, i32 addrspace(1)* undef
store volatile i32 %arg19, i32 addrspace(1)* undef
store volatile i32 %arg20, i32 addrspace(1)* undef
store volatile i32 %arg21, i32 addrspace(1)* undef
store volatile i32 %arg22, i32 addrspace(1)* undef
store volatile i32 %arg23, i32 addrspace(1)* undef
store volatile i32 %arg24, i32 addrspace(1)* undef
store volatile i32 %arg25, i32 addrspace(1)* undef
store volatile i32 %arg26, i32 addrspace(1)* undef
store volatile i32 %arg27, i32 addrspace(1)* undef
store volatile i32 %arg28, i32 addrspace(1)* undef
store volatile i32 %arg29, i32 addrspace(1)* undef
store volatile i32 %arg30, i32 addrspace(1)* undef
ret void
}
; GCN-LABEL: {{^}}kern_call_too_many_args_use_workitem_id_x_stack_yz:
; GCN: enable_vgpr_workitem_id = 2
[AMDGPU] Packed thread ids in function call ABI Differential Revision: https://reviews.llvm.org/D63851 llvm-svn: 364619
2019-06-28 09:52:13 +08:00
; GCN-NOT: v0
; GCN-DAG: v_lshlrev_b32_e32 v1, 10, v1
; GCN-DAG: v_or_b32_e32 v0, v0, v1
; GCN-DAG: v_lshlrev_b32_e32 v2, 20, v2
; GCN-DAG: v_or_b32_e32 v31, v0, v2
[AMDGPU] Add Scratch Wave Offset to Scratch Buffer Descriptor in entry functions Add the scratch wave offset to the scratch buffer descriptor (SRSrc) in the entry function prologue. This allows us to removes the scratch wave offset register from the calling convention ABI. As part of this change, allow the use of an inline constant zero for the SOffset of MUBUF instructions accessing the stack in entry functions when a frame pointer is not requested/required. Entry functions with calls still need to set up the calling convention ABI stack pointer register, and reference it in order to address arguments of called functions. The ABI stack pointer register remains unswizzled, but is now wave-relative instead of queue-relative. Non-entry functions also use an inline constant zero SOffset for wave-relative scratch access, but continue to use the stack and frame pointers as before. When the stack or frame pointer is converted to a swizzled offset it is now scaled directly, as the scratch wave offset no longer needs to be subtracted first. Update llvm/docs/AMDGPUUsage.rst to reflect these changes to the calling convention. Tags: #llvm Differential Revision: https://reviews.llvm.org/D75138
2020-01-22 06:27:57 +08:00
; GCN: s_mov_b32 s32, 0
AMDGPU: Pass special input registers to functions llvm-svn: 309998
2017-08-04 07:00:29 +08:00
; GCN: s_swappc_b64
define amdgpu_kernel void @kern_call_too_many_args_use_workitem_id_x_stack_yz() #1 {
call void @too_many_args_use_workitem_id_x_stack_yz(
i32 10, i32 20, i32 30, i32 40,
i32 50, i32 60, i32 70, i32 80,
i32 90, i32 100, i32 110, i32 120,
i32 130, i32 140, i32 150, i32 160,
i32 170, i32 180, i32 190, i32 200,
i32 210, i32 220, i32 230, i32 240,
i32 250, i32 260, i32 270, i32 280,
i32 290, i32 300, i32 310)
ret void
}
declare i32 @llvm.amdgcn.workitem.id.x() #0
declare i32 @llvm.amdgcn.workitem.id.y() #0
declare i32 @llvm.amdgcn.workitem.id.z() #0
attributes #0 = { nounwind readnone speculatable }
attributes #1 = { nounwind noinline }