1
0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-22 18:54:02 +01:00
llvm-mirror/test/CodeGen/AMDGPU/control-flow-fastregalloc.ll
Sebastian Neubauer f4093e4216 [AMDGPU] Save VGPR of whole wave when spilling
Spilling SGPRs to scratch uses a temporary VGPR. LLVM currently cannot
determine if a VGPR is used in other lanes or not, so we need to save
all lanes of the VGPR. We even need to save the VGPR if it is marked as
dead.

The generated code depends on two things:
- Can we scavenge an SGPR to save EXEC?
- And can we scavenge a VGPR?

If we can scavenge an SGPR, we
- save EXEC into the SGPR
- set the needed lane mask
- save the temporary VGPR
- write the spilled SGPR into VGPR lanes
- save the VGPR again to the target stack slot
- restore the VGPR
- restore EXEC

If we were not able to scavenge an SGPR, we do the same operations, but
everytime the temporary VGPR is written to memory, we
- write VGPR to memory
- flip exec (s_not exec, exec)
- write VGPR again (previously inactive lanes)

Surprisingly often, we are able to scavenge an SGPR, even though we are
at the brink of running out of SGPRs.
Scavenging a VGPR does not have a great effect (saves three instructions
if no SGPR was scavenged), but we need to know if the VGPR we use is
live before or not, otherwise the machine verifier complains.

Differential Revision: https://reviews.llvm.org/D96336
2021-04-12 11:01:38 +02:00

273 lines
12 KiB
LLVM

; RUN: llc -O0 -mtriple=amdgcn--amdhsa -march=amdgcn --amdhsa-code-object-version=2 -amdgpu-spill-sgpr-to-vgpr=0 -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=VMEM -check-prefix=GCN %s
; RUN: llc -O0 -mtriple=amdgcn--amdhsa -march=amdgcn --amdhsa-code-object-version=2 -amdgpu-spill-sgpr-to-vgpr=1 -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=VGPR -check-prefix=GCN %s
; Verify registers used for tracking exec mask changes when all
; registers are spilled at the end of the block. The SGPR spill
; placement relative to the exec modifications are important.
; FIXME: This checks with SGPR to VGPR spilling disabled, but this may
; not work correctly in cases where no workitems take a branch.
; GCN-LABEL: {{^}}divergent_if_endif:
; VGPR: workitem_private_segment_byte_size = 12{{$}}
; GCN: {{^}}; %bb.0:
; GCN: s_mov_b32 m0, -1
; GCN: ds_read_b32 [[LOAD0:v[0-9]+]]
; Spill load
; GCN: buffer_store_dword [[LOAD0]], off, s[0:3], 0 offset:[[LOAD0_OFFSET:[0-9]+]] ; 4-byte Folded Spill
; GCN: v_cmp_eq_u32_e64 [[CMP0:s\[[0-9]+:[0-9]\]]], s{{[0-9]+}}, v0
; Spill saved exec
; GCN: s_mov_b64 s{{\[}}[[SAVEEXEC_LO:[0-9]+]]:[[SAVEEXEC_HI:[0-9]+]]{{\]}}, exec
; VGPR: v_writelane_b32 [[SPILL_VGPR:v[0-9]+]], s[[SAVEEXEC_LO]], [[SAVEEXEC_LO_LANE:[0-9]+]]
; VGPR: v_writelane_b32 [[SPILL_VGPR]], s[[SAVEEXEC_HI]], [[SAVEEXEC_HI_LANE:[0-9]+]]
; VMEM: v_writelane_b32 v[[V_SAVEEXEC:[0-9]+]], s[[SAVEEXEC_LO]], 0
; VMEM: v_writelane_b32 v[[V_SAVEEXEC]], s[[SAVEEXEC_HI]], 1
; VMEM: buffer_store_dword v[[V_SAVEEXEC]], off, s[0:3], 0 offset:[[V_EXEC_SPILL_OFFSET:[0-9]+]] ; 4-byte Folded Spill
; GCN: s_and_b64 s{{\[}}[[ANDEXEC_LO:[0-9]+]]:[[ANDEXEC_HI:[0-9]+]]{{\]}}, s{{\[}}[[SAVEEXEC_LO]]:[[SAVEEXEC_HI]]{{\]}}, [[CMP0]]
; GCN: s_mov_b64 exec, s{{\[}}[[ANDEXEC_LO]]:[[ANDEXEC_HI]]{{\]}}
; GCN: s_cbranch_execz [[ENDIF:BB[0-9]+_[0-9]+]]
; GCN: ; %bb.{{[0-9]+}}: ; %if
; GCN: buffer_load_dword [[RELOAD_LOAD0:v[0-9]+]], off, s[0:3], 0 offset:[[LOAD0_OFFSET]] ; 4-byte Folded Reload
; GCN: s_mov_b32 m0, -1
; GCN: ds_read_b32 [[LOAD1:v[0-9]+]]
; GCN: s_waitcnt vmcnt(0) lgkmcnt(0)
; Spill val register
; GCN: v_add_i32_e32 [[VAL:v[0-9]+]], vcc, [[LOAD1]], [[RELOAD_LOAD0]]
; GCN: buffer_store_dword [[VAL]], off, s[0:3], 0 offset:[[VAL_OFFSET:[0-9]+]] ; 4-byte Folded Spill
; VMEM: [[ENDIF]]:
; Reload and restore exec mask
; VGPR: v_readlane_b32 s[[S_RELOAD_SAVEEXEC_LO:[0-9]+]], [[SPILL_VGPR]], [[SAVEEXEC_LO_LANE]]
; VGPR: v_readlane_b32 s[[S_RELOAD_SAVEEXEC_HI:[0-9]+]], [[SPILL_VGPR]], [[SAVEEXEC_HI_LANE]]
; VMEM: buffer_load_dword v[[V_RELOAD_SAVEEXEC:[0-9]+]], off, s[0:3], 0 offset:[[V_EXEC_SPILL_OFFSET]] ; 4-byte Folded Reload
; VMEM: s_waitcnt vmcnt(0)
; VMEM: v_readlane_b32 s[[S_RELOAD_SAVEEXEC_LO:[0-9]+]], v[[V_RELOAD_SAVEEXEC]], 0
; VMEM: v_readlane_b32 s[[S_RELOAD_SAVEEXEC_HI:[0-9]+]], v[[V_RELOAD_SAVEEXEC]], 1
; GCN: s_or_b64 exec, exec, s{{\[}}[[S_RELOAD_SAVEEXEC_LO]]:[[S_RELOAD_SAVEEXEC_HI]]{{\]}}
; Restore val
; GCN: buffer_load_dword [[RELOAD_VAL:v[0-9]+]], off, s[0:3], 0 offset:[[VAL_OFFSET]] ; 4-byte Folded Reload
; GCN: flat_store_dword v{{\[[0-9]+:[0-9]+\]}}, [[RELOAD_VAL]]
define amdgpu_kernel void @divergent_if_endif(i32 addrspace(1)* %out) #0 {
entry:
%tid = call i32 @llvm.amdgcn.workitem.id.x()
%load0 = load volatile i32, i32 addrspace(3)* undef
%cmp0 = icmp eq i32 %tid, 0
br i1 %cmp0, label %if, label %endif
if:
%load1 = load volatile i32, i32 addrspace(3)* undef
%val = add i32 %load0, %load1
br label %endif
endif:
%tmp4 = phi i32 [ %val, %if ], [ 0, %entry ]
store i32 %tmp4, i32 addrspace(1)* %out
ret void
}
; GCN-LABEL: {{^}}divergent_loop:
; VGPR: workitem_private_segment_byte_size = 16{{$}}
; GCN: {{^}}; %bb.0:
; GCN-DAG: s_mov_b32 m0, -1
; GCN-DAG: v_mov_b32_e32 [[PTR0:v[0-9]+]], 0{{$}}
; GCN: ds_read_b32 [[LOAD0:v[0-9]+]], [[PTR0]]
; GCN: v_cmp_eq_u32_e64 [[CMP0:s\[[0-9]+:[0-9]+\]]], s{{[0-9]+}}, v0
; Spill load
; GCN: buffer_store_dword [[LOAD0]], off, s[0:3], 0 offset:[[LOAD0_OFFSET:[0-9]+]] ; 4-byte Folded Spill
; GCN: s_mov_b64 s{{\[}}[[SAVEEXEC_LO:[0-9]+]]:[[SAVEEXEC_HI:[0-9]+]]{{\]}}, exec
; Spill saved exec
; VGPR: v_writelane_b32 [[SPILL_VGPR:v[0-9]+]], s[[SAVEEXEC_LO]], [[SAVEEXEC_LO_LANE:[0-9]+]]
; VGPR: v_writelane_b32 [[SPILL_VGPR]], s[[SAVEEXEC_HI]], [[SAVEEXEC_HI_LANE:[0-9]+]]
; VMEM: v_writelane_b32 v[[V_SAVEEXEC:[0-9]+]], s[[SAVEEXEC_LO]], 0
; VMEM: v_writelane_b32 v[[V_SAVEEXEC]], s[[SAVEEXEC_HI]], 1
; VMEM: buffer_store_dword v[[V_SAVEEXEC]], off, s[0:3], 0 offset:[[V_EXEC_SPILL_OFFSET:[0-9]+]] ; 4-byte Folded Spill
; GCN: s_and_b64 s{{\[}}[[ANDEXEC_LO:[0-9]+]]:[[ANDEXEC_HI:[0-9]+]]{{\]}}, s{{\[}}[[SAVEEXEC_LO:[0-9]+]]:[[SAVEEXEC_HI:[0-9]+]]{{\]}}, [[CMP0]]
; GCN: s_mov_b64 exec, s{{\[}}[[ANDEXEC_LO]]:[[ANDEXEC_HI]]{{\]}}
; GCN-NEXT: s_cbranch_execz [[END:BB[0-9]+_[0-9]+]]
; GCN: [[LOOP:BB[0-9]+_[0-9]+]]:
; GCN: buffer_load_dword v[[VAL_LOOP_RELOAD:[0-9]+]], off, s[0:3], 0 offset:[[LOAD0_OFFSET]] ; 4-byte Folded Reload
; GCN: v_subrev_i32_e32 [[VAL_LOOP:v[0-9]+]], vcc, v{{[0-9]+}}, v[[VAL_LOOP_RELOAD]]
; GCN: s_cmp_lg_u32
; VMEM: buffer_store_dword
; VMEM: buffer_store_dword
; VMEM: buffer_store_dword
; GCN: buffer_store_dword [[VAL_LOOP]], off, s[0:3], 0 offset:{{[0-9]+}} ; 4-byte Folded Spill
; GCN-NEXT: s_cbranch_scc1 [[LOOP]]
; GCN: buffer_store_dword [[VAL_LOOP]], off, s[0:3], 0 offset:[[VAL_SUB_OFFSET:[0-9]+]] ; 4-byte Folded Spill
; GCN: [[END]]:
; VGPR: v_readlane_b32 s[[S_RELOAD_SAVEEXEC_LO:[0-9]+]], [[SPILL_VGPR]], [[SAVEEXEC_LO_LANE]]
; VGPR: v_readlane_b32 s[[S_RELOAD_SAVEEXEC_HI:[0-9]+]], [[SPILL_VGPR]], [[SAVEEXEC_HI_LANE]]
; VMEM: buffer_load_dword v[[V_RELOAD_SAVEEXEC:[0-9]+]], off, s[0:3], 0 offset:[[V_EXEC_SPILL_OFFSET]] ; 4-byte Folded Reload
; VMEM: s_waitcnt vmcnt(0)
; VMEM: v_readlane_b32 s[[S_RELOAD_SAVEEXEC_LO:[0-9]+]], v[[V_RELOAD_SAVEEXEC]], 0
; VMEM: v_readlane_b32 s[[S_RELOAD_SAVEEXEC_HI:[0-9]+]], v[[V_RELOAD_SAVEEXEC]], 1
; GCN: s_or_b64 exec, exec, s{{\[}}[[S_RELOAD_SAVEEXEC_LO]]:[[S_RELOAD_SAVEEXEC_HI]]{{\]}}
; GCN: buffer_load_dword v[[VAL_END:[0-9]+]], off, s[0:3], 0 offset:[[VAL_SUB_OFFSET]] ; 4-byte Folded Reload
; GCN: flat_store_dword v{{\[[0-9]+:[0-9]+\]}}, v[[VAL_END]]
define amdgpu_kernel void @divergent_loop(i32 addrspace(1)* %out) #0 {
entry:
%tid = call i32 @llvm.amdgcn.workitem.id.x()
%load0 = load volatile i32, i32 addrspace(3)* null
%cmp0 = icmp eq i32 %tid, 0
br i1 %cmp0, label %loop, label %end
loop:
%i = phi i32 [ %i.inc, %loop ], [ 0, %entry ]
%val = phi i32 [ %val.sub, %loop ], [ %load0, %entry ]
%load1 = load volatile i32, i32 addrspace(3)* undef
%i.inc = add i32 %i, 1
%val.sub = sub i32 %val, %load1
%cmp1 = icmp ne i32 %i, 256
br i1 %cmp1, label %loop, label %end
end:
%tmp4 = phi i32 [ %val.sub, %loop ], [ 0, %entry ]
store i32 %tmp4, i32 addrspace(1)* %out
ret void
}
; GCN-LABEL: {{^}}divergent_if_else_endif:
; GCN: {{^}}; %bb.0:
; GCN-DAG: s_mov_b32 m0, -1
; GCN-DAG: v_mov_b32_e32 [[PTR0:v[0-9]+]], 0{{$}}
; GCN: ds_read_b32 [[LOAD0:v[0-9]+]], [[PTR0]]
; Spill load
; GCN: buffer_store_dword [[LOAD0]], off, s[0:3], 0 offset:[[LOAD0_OFFSET:[0-9]+]] ; 4-byte Folded Spill
; GCN: s_mov_b32 [[ZERO:s[0-9]+]], 0
; GCN: v_cmp_ne_u32_e64 [[CMP0:s\[[0-9]+:[0-9]\]]], [[ZERO]], v0
; GCN: s_mov_b64 s{{\[}}[[SAVEEXEC_LO:[0-9]+]]:[[SAVEEXEC_HI:[0-9]+]]{{\]}}, exec
; GCN: s_and_b64 s{{\[}}[[ANDEXEC_LO:[0-9]+]]:[[ANDEXEC_HI:[0-9]+]]{{\]}}, s{{\[}}[[SAVEEXEC_LO:[0-9]+]]:[[SAVEEXEC_HI:[0-9]+]]{{\]}}, [[CMP0]]
; GCN: s_xor_b64 s{{\[}}[[SAVEEXEC_LO]]:[[SAVEEXEC_HI]]{{\]}}, s{{\[}}[[ANDEXEC_LO]]:[[ANDEXEC_HI]]{{\]}}, s{{\[}}[[SAVEEXEC_LO]]:[[SAVEEXEC_HI]]{{\]}}
; Spill saved exec
; VGPR: v_writelane_b32 [[SPILL_VGPR:v[0-9]+]], s[[SAVEEXEC_LO]], [[SAVEEXEC_LO_LANE:[0-9]+]]
; VGPR: v_writelane_b32 [[SPILL_VGPR]], s[[SAVEEXEC_HI]], [[SAVEEXEC_HI_LANE:[0-9]+]]
; VMEM: v_writelane_b32 v[[V_SAVEEXEC:[0-9]+]], s[[SAVEEXEC_LO]], 0
; VMEM: v_writelane_b32 v[[V_SAVEEXEC]], s[[SAVEEXEC_HI]], 1
; VMEM: buffer_store_dword v[[V_SAVEEXEC]], off, s[0:3], 0 offset:[[SAVEEXEC_OFFSET:[0-9]+]] ; 4-byte Folded Spill
; GCN: s_mov_b64 exec, [[CMP0]]
; FIXME: It makes no sense to put this skip here
; GCN: s_cbranch_execz [[FLOW:BB[0-9]+_[0-9]+]]
; GCN-NEXT: s_branch [[ELSE:BB[0-9]+_[0-9]+]]
; GCN: [[FLOW]]: ; %Flow
; VGPR: v_readlane_b32 s[[FLOW_S_RELOAD_SAVEEXEC_LO:[0-9]+]], [[SPILL_VGPR]], [[SAVEEXEC_LO_LANE]]
; VGPR: v_readlane_b32 s[[FLOW_S_RELOAD_SAVEEXEC_HI:[0-9]+]], [[SPILL_VGPR]], [[SAVEEXEC_HI_LANE]]
; VMEM: buffer_load_dword v[[FLOW_V_RELOAD_SAVEEXEC:[0-9]+]], off, s[0:3], 0 offset:[[SAVEEXEC_OFFSET]]
; VMEM: s_waitcnt vmcnt(0)
; VMEM: v_readlane_b32 s[[FLOW_S_RELOAD_SAVEEXEC_LO:[0-9]+]], v[[FLOW_V_RELOAD_SAVEEXEC]], 0
; VMEM: v_readlane_b32 s[[FLOW_S_RELOAD_SAVEEXEC_HI:[0-9]+]], v[[FLOW_V_RELOAD_SAVEEXEC]], 1
; GCN: s_or_saveexec_b64 s{{\[}}[[FLOW_S_RELOAD_SAVEEXEC_LO_SAVEEXEC:[0-9]+]]:[[FLOW_S_RELOAD_SAVEEXEC_HI_SAVEEXEC:[0-9]+]]{{\]}}, s{{\[}}[[FLOW_S_RELOAD_SAVEEXEC_LO]]:[[FLOW_S_RELOAD_SAVEEXEC_HI]]{{\]}}
; Regular spill value restored after exec modification
; GCN: buffer_load_dword [[FLOW_VAL:v[0-9]+]], off, s[0:3], 0 offset:[[FLOW_VAL_OFFSET:[0-9]+]] ; 4-byte Folded Reload
; Followed by spill
; GCN: buffer_store_dword [[FLOW_VAL]], off, s[0:3], 0 offset:[[RESULT_OFFSET:[0-9]+]] ; 4-byte Folded Spill
; GCN: s_and_b64 s{{\[}}[[FLOW_AND_EXEC_LO:[0-9]+]]:[[FLOW_AND_EXEC_HI:[0-9]+]]{{\]}}, exec, s{{\[}}[[FLOW_S_RELOAD_SAVEEXEC_LO_SAVEEXEC]]:[[FLOW_S_RELOAD_SAVEEXEC_HI_SAVEEXEC]]{{\]}}
; Spill saved exec
; VGPR: v_writelane_b32 [[SPILL_VGPR]], s[[FLOW_AND_EXEC_LO]], [[FLOW_SAVEEXEC_LO_LANE:[0-9]+]]
; VGPR: v_writelane_b32 [[SPILL_VGPR]], s[[FLOW_AND_EXEC_HI]], [[FLOW_SAVEEXEC_HI_LANE:[0-9]+]]
; VMEM: v_writelane_b32 v[[FLOW_V_SAVEEXEC:[0-9]+]], s[[FLOW_AND_EXEC_LO]], 0
; VMEM: v_writelane_b32 v[[FLOW_V_SAVEEXEC]], s[[FLOW_AND_EXEC_HI]], 1
; VMEM: buffer_store_dword v[[FLOW_V_SAVEEXEC]], off, s[0:3], 0 offset:[[FLOW_SAVEEXEC_OFFSET:[0-9]+]] ; 4-byte Folded Spill
; GCN: s_xor_b64 exec, exec, s{{\[}}[[FLOW_AND_EXEC_LO]]:[[FLOW_AND_EXEC_HI]]{{\]}}
; GCN-NEXT: s_cbranch_execz [[ENDIF:BB[0-9]+_[0-9]+]]
; GCN: ; %bb.{{[0-9]+}}: ; %if
; GCN: buffer_load_dword v[[LOAD0_RELOAD:[0-9]+]], off, s[0:3], 0 offset:[[LOAD0_OFFSET]] ; 4-byte Folded Reload
; GCN: ds_read_b32
; GCN: v_add_i32_e32 [[ADD:v[0-9]+]], vcc, v{{[0-9]+}}, v[[LOAD0_RELOAD]]
; GCN: buffer_store_dword [[ADD]], off, s[0:3], 0 offset:[[RESULT_OFFSET]] ; 4-byte Folded Spill
; GCN-NEXT: s_branch [[ENDIF:BB[0-9]+_[0-9]+]]
; GCN: [[ELSE]]: ; %else
; GCN: buffer_load_dword v[[LOAD0_RELOAD:[0-9]+]], off, s[0:3], 0 offset:[[LOAD0_OFFSET]] ; 4-byte Folded Reload
; GCN: v_subrev_i32_e32 [[SUB:v[0-9]+]], vcc, v{{[0-9]+}}, v[[LOAD0_RELOAD]]
; GCN: buffer_store_dword [[ADD]], off, s[0:3], 0 offset:[[FLOW_RESULT_OFFSET:[0-9]+]] ; 4-byte Folded Spill
; GCN-NEXT: s_branch [[FLOW]]
; GCN: [[ENDIF]]:
; VGPR: v_readlane_b32 s[[S_RELOAD_SAVEEXEC_LO:[0-9]+]], [[SPILL_VGPR]], [[FLOW_SAVEEXEC_LO_LANE]]
; VGPR: v_readlane_b32 s[[S_RELOAD_SAVEEXEC_HI:[0-9]+]], [[SPILL_VGPR]], [[FLOW_SAVEEXEC_HI_LANE]]
; VMEM: buffer_load_dword v[[V_RELOAD_SAVEEXEC:[0-9]+]], off, s[0:3], 0 offset:[[FLOW_SAVEEXEC_OFFSET]] ; 4-byte Folded Reload
; VMEM: s_waitcnt vmcnt(0)
; VMEM: v_readlane_b32 s[[S_RELOAD_SAVEEXEC_LO:[0-9]+]], v[[V_RELOAD_SAVEEXEC]], 0
; VMEM: v_readlane_b32 s[[S_RELOAD_SAVEEXEC_HI:[0-9]+]], v[[V_RELOAD_SAVEEXEC]], 1
; GCN: s_or_b64 exec, exec, s{{\[}}[[S_RELOAD_SAVEEXEC_LO]]:[[S_RELOAD_SAVEEXEC_HI]]{{\]}}
; GCN: buffer_load_dword v[[RESULT:[0-9]+]], off, s[0:3], 0 offset:[[RESULT_OFFSET]] ; 4-byte Folded Reload
; GCN: flat_store_dword v{{\[[0-9]+:[0-9]+\]}}, v[[RESULT]]
define amdgpu_kernel void @divergent_if_else_endif(i32 addrspace(1)* %out) #0 {
entry:
%tid = call i32 @llvm.amdgcn.workitem.id.x()
%load0 = load volatile i32, i32 addrspace(3)* null
%cmp0 = icmp eq i32 %tid, 0
br i1 %cmp0, label %if, label %else
if:
%load1 = load volatile i32, i32 addrspace(3)* undef
%val0 = add i32 %load0, %load1
br label %endif
else:
%load2 = load volatile i32, i32 addrspace(3)* undef
%val1 = sub i32 %load0, %load2
br label %endif
endif:
%result = phi i32 [ %val0, %if ], [ %val1, %else ]
store i32 %result, i32 addrspace(1)* %out
ret void
}
declare i32 @llvm.amdgcn.workitem.id.x() #1
attributes #0 = { nounwind }
attributes #1 = { nounwind readnone }