1
0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2025-02-01 05:01:59 +01:00
Matt Arsenault 2df3daaca8 AMDGPU: Force skip branches over calls
Unfortunately the way SIInsertSkips works is backwards, and is
required for correctness. r338235 added handling of some special cases
where skipping is mandatory to avoid side effects if no lanes are
active. It conservatively handled asm correctly, but the same logic
needs to apply to calls.

Usually the call sequence code is larger than the skip threshold,
although the way the count is computed is really broken, so I'm not
sure if anything was likely to really hit this.

llvm-svn: 361202
2019-05-20 22:04:42 +00:00

68 lines
1.5 KiB
LLVM

; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=hawaii < %s | FileCheck -enable-var-scope -check-prefix=GCN %s
; A call should be skipped if all lanes are zero, since we don't know
; what side effects should be avoided inside the call.
define hidden void @func() #1 {
ret void
}
; GCN-LABEL: {{^}}if_call:
; GCN: s_and_saveexec_b64
; GCN-NEXT: ; mask branch [[END:BB[0-9]+_[0-9]+]]
; GCN-NEXT: s_cbranch_execz [[END]]
; GCN: s_swappc_b64
; GCN: [[END]]:
define void @if_call(i32 %flag) #0 {
%cc = icmp eq i32 %flag, 0
br i1 %cc, label %call, label %end
call:
call void @func()
br label %end
end:
ret void
}
; GCN-LABEL: {{^}}if_asm:
; GCN: s_and_saveexec_b64
; GCN-NEXT: ; mask branch [[END:BB[0-9]+_[0-9]+]]
; GCN-NEXT: s_cbranch_execz [[END]]
; GCN: ; sample asm
; GCN: [[END]]:
define void @if_asm(i32 %flag) #0 {
%cc = icmp eq i32 %flag, 0
br i1 %cc, label %call, label %end
call:
call void asm sideeffect "; sample asm", ""()
br label %end
end:
ret void
}
; GCN-LABEL: {{^}}if_call_kernel:
; GCN: s_and_saveexec_b64
; GCN-NEXT: ; mask branch [[END:BB[0-9]+_[0-9]+]]
; GCN-NEXT: s_cbranch_execz [[END]]
; GCN: s_swappc_b64
define amdgpu_kernel void @if_call_kernel() #0 {
%id = call i32 @llvm.amdgcn.workitem.id.x()
%cc = icmp eq i32 %id, 0
br i1 %cc, label %call, label %end
call:
call void @func()
br label %end
end:
ret void
}
declare i32 @llvm.amdgcn.workitem.id.x() #2
attributes #0 = { nounwind }
attributes #1 = { nounwind noinline }
attributes #2 = { nounwind readnone speculatable }