llvm-project/llvm/test/CodeGen/AMDGPU/branch-relaxation.ll

546 lines
14 KiB
LLVM
Raw Normal View History

; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs -amdgpu-s-branch-bits=4 < %s | FileCheck -check-prefix=GCN %s
; FIXME: We should use llvm-mc for this, but we can't even parse our own output.
; See PR33579.
; RUN: llc -march=amdgcn -verify-machineinstrs -amdgpu-s-branch-bits=4 -o %t.o -filetype=obj %s
; RUN: llvm-readobj -r %t.o | FileCheck --check-prefix=OBJ %s
; OBJ: Relocations [
; OBJ-NEXT: ]
; Restrict maximum branch to between +7 and -8 dwords
; Used to emit an always 4 byte instruction. Inline asm always assumes
; each instruction is the maximum size.
declare void @llvm.amdgcn.s.sleep(i32) #0
declare i32 @llvm.amdgcn.workitem.id.x() #1
; GCN-LABEL: {{^}}uniform_conditional_max_short_forward_branch:
; GCN: s_load_dword [[CND:s[0-9]+]]
; GCN: s_cmp_eq_u32 [[CND]], 0
; GCN-NEXT: s_cbranch_scc1 [[BB3:BB[0-9]+_[0-9]+]]
; GCN-NEXT: ; %bb.1: ; %bb2
; GCN-NEXT: ;;#ASMSTART
; GCN-NEXT: v_nop_e64
; GCN-NEXT: v_nop_e64
; GCN-NEXT: v_nop_e64
; GCN-NEXT: ;;#ASMEND
; GCN-NEXT: s_sleep 0
; GCN-NEXT: [[BB3]]: ; %bb3
; GCN: v_mov_b32_e32 [[V_CND:v[0-9]+]], [[CND]]
; GCN: buffer_store_dword [[V_CND]]
; GCN: s_endpgm
define amdgpu_kernel void @uniform_conditional_max_short_forward_branch(i32 addrspace(1)* %arg, i32 %cnd) #0 {
bb:
%cmp = icmp eq i32 %cnd, 0
br i1 %cmp, label %bb3, label %bb2 ; +8 dword branch
bb2:
; 24 bytes
call void asm sideeffect
"v_nop_e64
v_nop_e64
v_nop_e64", ""() #0
call void @llvm.amdgcn.s.sleep(i32 0)
br label %bb3
bb3:
store volatile i32 %cnd, i32 addrspace(1)* %arg
ret void
}
; GCN-LABEL: {{^}}uniform_conditional_min_long_forward_branch:
; GCN: s_load_dword [[CND:s[0-9]+]]
; GCN: s_cmp_eq_u32 [[CND]], 0
; GCN-NEXT: s_cbranch_scc0 [[LONGBB:BB[0-9]+_[0-9]+]]
; GCN-NEXT: [[LONG_JUMP:BB[0-9]+_[0-9]+]]: ; %bb0
; GCN-NEXT: s_getpc_b64 vcc
; GCN-NEXT: s_add_u32 vcc_lo, vcc_lo, [[ENDBB:BB[0-9]+_[0-9]+]]-([[LONG_JUMP]]+4)
; GCN-NEXT: s_addc_u32 vcc_hi, vcc_hi, 0
; GCN-NEXT: s_setpc_b64 vcc
; GCN-NEXT: [[LONGBB]]:
; GCN-NEXT: ;;#ASMSTART
; GCN: v_nop_e64
; GCN: v_nop_e64
; GCN: v_nop_e64
; GCN: v_nop_e64
; GCN-NEXT: ;;#ASMEND
; GCN-NEXT: [[ENDBB]]:
; GCN: v_mov_b32_e32 [[V_CND:v[0-9]+]], [[CND]]
; GCN: buffer_store_dword [[V_CND]]
; GCN: s_endpgm
define amdgpu_kernel void @uniform_conditional_min_long_forward_branch(i32 addrspace(1)* %arg, i32 %cnd) #0 {
bb0:
%cmp = icmp eq i32 %cnd, 0
br i1 %cmp, label %bb3, label %bb2 ; +9 dword branch
bb2:
; 32 bytes
call void asm sideeffect
"v_nop_e64
v_nop_e64
v_nop_e64
v_nop_e64", ""() #0
br label %bb3
bb3:
store volatile i32 %cnd, i32 addrspace(1)* %arg
ret void
}
; GCN-LABEL: {{^}}uniform_conditional_min_long_forward_vcnd_branch:
; GCN: s_load_dword [[CND:s[0-9]+]]
; GCN-DAG: v_mov_b32_e32 [[V_CND:v[0-9]+]], [[CND]]
; GCN-DAG: v_cmp_eq_f32_e64 [[UNMASKED:s\[[0-9]+:[0-9]+\]]], [[CND]], 0
; GCN-DAG: s_and_b64 vcc, exec, [[UNMASKED]]
; GCN: s_cbranch_vccz [[LONGBB:BB[0-9]+_[0-9]+]]
; GCN-NEXT: [[LONG_JUMP:BB[0-9]+_[0-9]+]]: ; %bb0
; GCN-NEXT: s_getpc_b64 vcc
; GCN-NEXT: s_add_u32 vcc_lo, vcc_lo, [[ENDBB:BB[0-9]+_[0-9]+]]-([[LONG_JUMP]]+4)
; GCN-NEXT: s_addc_u32 vcc_hi, vcc_hi, 0
; GCN-NEXT: s_setpc_b64 vcc
; GCN-NEXT: [[LONGBB]]:
; GCN: v_nop_e64
; GCN: v_nop_e64
; GCN: v_nop_e64
; GCN: v_nop_e64
; GCN: [[ENDBB]]:
; GCN: buffer_store_dword [[V_CND]]
; GCN: s_endpgm
define amdgpu_kernel void @uniform_conditional_min_long_forward_vcnd_branch(float addrspace(1)* %arg, float %cnd) #0 {
bb0:
%cmp = fcmp oeq float %cnd, 0.0
br i1 %cmp, label %bb3, label %bb2 ; + 8 dword branch
bb2:
call void asm sideeffect " ; 32 bytes
v_nop_e64
v_nop_e64
v_nop_e64
v_nop_e64", ""() #0
br label %bb3
bb3:
store volatile float %cnd, float addrspace(1)* %arg
ret void
}
; GCN-LABEL: {{^}}min_long_forward_vbranch:
; GCN: buffer_load_dword
; GCN: v_cmp_ne_u32_e32 vcc, 0, v{{[0-9]+}}
; GCN: s_and_saveexec_b64 [[SAVE:s\[[0-9]+:[0-9]+\]]], vcc
; GCN: v_nop_e64
; GCN: v_nop_e64
; GCN: v_nop_e64
; GCN: v_nop_e64
; GCN: s_or_b64 exec, exec, [[SAVE]]
; GCN: buffer_store_dword
; GCN: s_endpgm
define amdgpu_kernel void @min_long_forward_vbranch(i32 addrspace(1)* %arg) #0 {
bb:
%tid = call i32 @llvm.amdgcn.workitem.id.x()
%tid.ext = zext i32 %tid to i64
%gep = getelementptr inbounds i32, i32 addrspace(1)* %arg, i64 %tid.ext
%load = load volatile i32, i32 addrspace(1)* %gep
%cmp = icmp eq i32 %load, 0
br i1 %cmp, label %bb3, label %bb2 ; + 8 dword branch
bb2:
call void asm sideeffect " ; 32 bytes
v_nop_e64
v_nop_e64
v_nop_e64
v_nop_e64", ""() #0
br label %bb3
bb3:
store volatile i32 %load, i32 addrspace(1)* %gep
ret void
}
; GCN-LABEL: {{^}}long_backward_sbranch:
; GCN: s_mov_b32 [[LOOPIDX:s[0-9]+]], 0{{$}}
; GCN: [[LOOPBB:BB[0-9]+_[0-9]+]]: ; %bb2
; GCN-NEXT: ; =>This Inner Loop Header: Depth=1
; GCN-NEXT: s_add_i32 [[INC:s[0-9]+]], [[LOOPIDX]], 1
; GCN-NEXT: s_cmp_lt_i32 [[INC]], 10
; GCN-NEXT: ;;#ASMSTART
; GCN-NEXT: v_nop_e64
; GCN-NEXT: v_nop_e64
; GCN-NEXT: v_nop_e64
; GCN-NEXT: ;;#ASMEND
; GCN-NEXT: s_cbranch_scc0 [[ENDBB:BB[0-9]+_[0-9]+]]
; GCN-NEXT: [[LONG_JUMP:BB[0-9]+_[0-9]+]]: ; %bb2
; GCN-NEXT: ; in Loop: Header=[[LOOPBB]] Depth=1
; GCN-NEXT: s_getpc_b64 vcc
; GCN-NEXT: s_sub_u32 vcc_lo, vcc_lo, ([[LONG_JUMP]]+4)-[[LOOPBB]]
; GCN-NEXT: s_subb_u32 vcc_hi, vcc_hi, 0
; GCN-NEXT: s_setpc_b64 vcc
; GCN-NEXT: [[ENDBB]]:
; GCN-NEXT: s_endpgm
define amdgpu_kernel void @long_backward_sbranch(i32 addrspace(1)* %arg) #0 {
bb:
br label %bb2
bb2:
%loop.idx = phi i32 [ 0, %bb ], [ %inc, %bb2 ]
; 24 bytes
call void asm sideeffect
"v_nop_e64
v_nop_e64
v_nop_e64", ""() #0
%inc = add nsw i32 %loop.idx, 1 ; add cost 4
%cmp = icmp slt i32 %inc, 10 ; condition cost = 8
br i1 %cmp, label %bb2, label %bb3 ; -
bb3:
ret void
}
; Requires expansion of unconditional branch from %bb2 to %bb4 (and
; expansion of conditional branch from %bb to %bb3.
; GCN-LABEL: {{^}}uniform_unconditional_min_long_forward_branch:
; GCN: s_cmp_eq_u32
; GCN-NEXT: s_cbranch_scc0 [[BB2:BB[0-9]+_[0-9]+]]
; GCN-NEXT: [[LONG_JUMP0:BB[0-9]+_[0-9]+]]: ; %bb0
; GCN-NEXT: s_getpc_b64 vcc
; GCN-NEXT: s_add_u32 vcc_lo, vcc_lo, [[BB3:BB[0-9]_[0-9]+]]-([[LONG_JUMP0]]+4)
; GCN-NEXT: s_addc_u32 vcc_hi, vcc_hi, 0{{$}}
; GCN-NEXT: s_setpc_b64 vcc
; GCN-NEXT: [[BB2]]: ; %bb2
; GCN: v_mov_b32_e32 [[BB2_K:v[0-9]+]], 17
; GCN: buffer_store_dword [[BB2_K]]
; GCN-NEXT: [[LONG_JUMP1:BB[0-9]+_[0-9]+]]: ; %bb2
; GCN-NEXT: s_getpc_b64 vcc
; GCN-NEXT: s_add_u32 vcc_lo, vcc_lo, [[BB4:BB[0-9]_[0-9]+]]-([[LONG_JUMP1]]+4)
; GCN-NEXT: s_addc_u32 vcc_hi, vcc_hi, 0{{$}}
; GCN-NEXT: s_setpc_b64 vcc
; GCN: [[BB3]]: ; %bb3
; GCN: v_nop_e64
; GCN: v_nop_e64
; GCN: v_nop_e64
; GCN: v_nop_e64
; GCN: ;;#ASMEND
; GCN-NEXT: [[BB4]]: ; %bb4
; GCN: v_mov_b32_e32 [[BB4_K:v[0-9]+]], 63
; GCN: buffer_store_dword [[BB4_K]]
; GCN-NEXT: s_endpgm
; GCN-NEXT: .Lfunc_end{{[0-9]+}}:
define amdgpu_kernel void @uniform_unconditional_min_long_forward_branch(i32 addrspace(1)* %arg, i32 %arg1) {
bb0:
%tmp = icmp ne i32 %arg1, 0
br i1 %tmp, label %bb2, label %bb3
bb2:
store volatile i32 17, i32 addrspace(1)* undef
br label %bb4
bb3:
; 32 byte asm
call void asm sideeffect
"v_nop_e64
v_nop_e64
v_nop_e64
v_nop_e64", ""() #0
br label %bb4
bb4:
store volatile i32 63, i32 addrspace(1)* %arg
ret void
}
; GCN-LABEL: {{^}}uniform_unconditional_min_long_backward_branch:
; GCN-NEXT: ; %bb.0: ; %entry
; GCN-NEXT: [[LOOP:BB[0-9]_[0-9]+]]: ; %loop
; GCN-NEXT: ; =>This Inner Loop Header: Depth=1
; GCN-NEXT: ;;#ASMSTART
; GCN-NEXT: v_nop_e64
; GCN-NEXT: v_nop_e64
; GCN-NEXT: v_nop_e64
; GCN-NEXT: v_nop_e64
; GCN-NEXT: ;;#ASMEND
; GCN-NEXT: [[LONGBB:BB[0-9]+_[0-9]+]]: ; %loop
; GCN-NEXT: ; in Loop: Header=[[LOOP]] Depth=1
; GCN-NEXT: s_getpc_b64 vcc
; GCN-NEXT: s_sub_u32 vcc_lo, vcc_lo, ([[LONGBB]]+4)-[[LOOP]]
; GCN-NEXT: s_subb_u32 vcc_hi, vcc_hi, 0{{$}}
; GCN-NEXT: s_setpc_b64 vcc
; GCN-NEXT .Lfunc_end{{[0-9]+}}:
define amdgpu_kernel void @uniform_unconditional_min_long_backward_branch(i32 addrspace(1)* %arg, i32 %arg1) {
entry:
br label %loop
loop:
; 32 byte asm
call void asm sideeffect
"v_nop_e64
v_nop_e64
v_nop_e64
v_nop_e64", ""() #0
br label %loop
}
; Expansion of branch from %bb1 to %bb3 introduces need to expand
; branch from %bb0 to %bb2
; GCN-LABEL: {{^}}expand_requires_expand:
; GCN-NEXT: ; %bb.0: ; %bb0
; GCN: s_load_dword
; GCN: s_cmp_lt_i32 s{{[0-9]+}}, 0{{$}}
; GCN-NEXT: s_cbranch_scc0 [[BB1:BB[0-9]+_[0-9]+]]
; GCN-NEXT: [[LONGBB0:BB[0-9]+_[0-9]+]]: ; %bb0
; GCN-NEXT: s_getpc_b64 vcc
; GCN-NEXT: s_add_u32 vcc_lo, vcc_lo, [[BB2:BB[0-9]_[0-9]+]]-([[LONGBB0]]+4)
; GCN-NEXT: s_addc_u32 vcc_hi, vcc_hi, 0{{$}}
; GCN-NEXT: s_setpc_b64 vcc
; GCN-NEXT: [[BB1]]: ; %bb1
; GCN-NEXT: s_load_dword
; GCN-NEXT: s_waitcnt lgkmcnt(0)
; GCN-NEXT: s_cmp_eq_u32 s{{[0-9]+}}, 3{{$}}
; GCN-NEXT: s_cbranch_scc0 [[BB2:BB[0-9]_[0-9]+]]
; GCN-NEXT: [[LONGBB1:BB[0-9]+_[0-9]+]]: ; %bb1
; GCN-NEXT: s_getpc_b64 vcc
; GCN-NEXT: s_add_u32 vcc_lo, vcc_lo, [[BB3:BB[0-9]+_[0-9]+]]-([[LONGBB1]]+4)
; GCN-NEXT: s_addc_u32 vcc_hi, vcc_hi, 0{{$}}
; GCN-NEXT: s_setpc_b64 vcc
; GCN-NEXT: [[BB2]]: ; %bb2
; GCN-NEXT: ;;#ASMSTART
; GCN-NEXT: v_nop_e64
; GCN-NEXT: v_nop_e64
; GCN-NEXT: v_nop_e64
; GCN-NEXT: v_nop_e64
; GCN-NEXT: ;;#ASMEND
; GCN-NEXT: [[BB3]]: ; %bb3
; GCN-NEXT: ;;#ASMSTART
; GCN-NEXT: v_nop_e64
; GCN-NEXT: ;;#ASMEND
; GCN-NEXT: ;;#ASMSTART
; GCN-NEXT: v_nop_e64
; GCN-NEXT: ;;#ASMEND
; GCN-NEXT: s_endpgm
define amdgpu_kernel void @expand_requires_expand(i32 %cond0) #0 {
bb0:
%tmp = tail call i32 @llvm.amdgcn.workitem.id.x() #0
%cmp0 = icmp slt i32 %cond0, 0
br i1 %cmp0, label %bb2, label %bb1
bb1:
%val = load volatile i32, i32 addrspace(2)* undef
%cmp1 = icmp eq i32 %val, 3
br i1 %cmp1, label %bb3, label %bb2
bb2:
call void asm sideeffect
"v_nop_e64
v_nop_e64
v_nop_e64
v_nop_e64", ""() #0
br label %bb3
bb3:
; These NOPs prevent tail-duplication-based outlining
; from firing, which defeats the need to expand the branches and this test.
call void asm sideeffect
"v_nop_e64", ""() #0
call void asm sideeffect
"v_nop_e64", ""() #0
ret void
}
; Requires expanding of required skip branch.
; GCN-LABEL: {{^}}uniform_inside_divergent:
; GCN: v_cmp_gt_u32_e32 vcc, 16, v{{[0-9]+}}
; GCN-NEXT: s_and_saveexec_b64 [[MASK:s\[[0-9]+:[0-9]+\]]], vcc
; GCN-NEXT: ; mask branch [[ENDIF:BB[0-9]+_[0-9]+]]
; GCN-NEXT: s_cbranch_execnz [[IF:BB[0-9]+_[0-9]+]]
; GCN-NEXT: [[LONGBB:BB[0-9]+_[0-9]+]]: ; %entry
; GCN-NEXT: s_getpc_b64 vcc
; GCN-NEXT: s_add_u32 vcc_lo, vcc_lo, [[BB2:BB[0-9]_[0-9]+]]-([[LONGBB]]+4)
; GCN-NEXT: s_addc_u32 vcc_hi, vcc_hi, 0{{$}}
; GCN-NEXT: s_setpc_b64 vcc
; GCN-NEXT: [[IF]]: ; %if
; GCN: buffer_store_dword
; GCN: s_cmp_lg_u32
; GCN: s_cbranch_scc1 [[ENDIF]]
; GCN-NEXT: ; %bb.2: ; %if_uniform
; GCN: buffer_store_dword
; GCN-NEXT: [[ENDIF]]: ; %endif
; GCN-NEXT: s_or_b64 exec, exec, [[MASK]]
; GCN-NEXT: s_sleep 5
; GCN-NEXT: s_endpgm
define amdgpu_kernel void @uniform_inside_divergent(i32 addrspace(1)* %out, i32 %cond) #0 {
entry:
%tid = call i32 @llvm.amdgcn.workitem.id.x()
%d_cmp = icmp ult i32 %tid, 16
br i1 %d_cmp, label %if, label %endif
if:
store i32 0, i32 addrspace(1)* %out
%u_cmp = icmp eq i32 %cond, 0
br i1 %u_cmp, label %if_uniform, label %endif
if_uniform:
store i32 1, i32 addrspace(1)* %out
br label %endif
endif:
; layout can remove the split branch if it can copy the return block.
; This call makes the return block long enough that it doesn't get copied.
call void @llvm.amdgcn.s.sleep(i32 5);
ret void
}
; si_mask_branch
; GCN-LABEL: {{^}}analyze_mask_branch:
; GCN: v_cmp_lt_f32_e32 vcc
; GCN-NEXT: s_and_saveexec_b64 [[MASK:s\[[0-9]+:[0-9]+\]]], vcc
; GCN-NEXT: ; mask branch [[RET:BB[0-9]+_[0-9]+]]
[Dominators] Include infinite loops in PostDominatorTree Summary: This patch teaches PostDominatorTree about infinite loops. It is built on top of D29705 by @dberlin which includes a very detailed motivation for this change. What's new is that the patch also teaches the incremental updater how to deal with reverse-unreachable regions and how to properly maintain and verify tree roots. Before that, the incremental algorithm sometimes ended up preserving reverse-unreachable regions after updates that wouldn't appear in the tree if it was constructed from scratch on the same CFG. This patch makes the following assumptions: - A sequence of updates should produce the same tree as a recalculating it. - Any sequence of the same updates should lead to the same tree. - Siblings and roots are unordered. The last two properties are essential to efficiently perform batch updates in the future. When it comes to the first one, we can decide later that the consistency between freshly built tree and an updated one doesn't matter match, as there are many correct ways to pick roots in infinite loops, and to relax this assumption. That should enable us to recalculate postdominators less frequently. This patch is pretty conservative when it comes to incremental updates on reverse-unreachable regions and ends up recalculating the whole tree in many cases. It should be possible to improve the performance in many cases, if we decide that it's important enough. That being said, my experiments showed that reverse-unreachable are very rare in the IR emitted by clang when bootstrapping clang. Here are the statistics I collected by analyzing IR between passes and after each removePredecessor call: ``` # functions: 52283 # samples: 337609 # reverse unreachable BBs: 216022 # BBs: 247840796 Percent reverse-unreachable: 0.08716159869015269 % Max(PercRevUnreachable) in a function: 87.58620689655172 % # > 25 % samples: 471 ( 0.1395104988314885 % samples ) ... in 145 ( 0.27733680163724345 % functions ) ``` Most of the reverse-unreachable regions come from invalid IR where it wouldn't be possible to construct a PostDomTree anyway. I would like to commit this patch in the next week in order to be able to complete the work that depends on it before the end of my internship, so please don't wait long to voice your concerns :). Reviewers: dberlin, sanjoy, grosser, brzycki, davide, chandlerc, hfinkel Reviewed By: dberlin Subscribers: nhaehnle, javed.absar, kparzysz, uabelho, jlebar, hiraditya, llvm-commits, dberlin, david2050 Differential Revision: https://reviews.llvm.org/D35851 llvm-svn: 310940
2017-08-16 02:14:57 +08:00
; GCN-NEXT: [[LOOP_BODY:BB[0-9]+_[0-9]+]]: ; %loop_body
; GCN: ;;#ASMSTART
; GCN: v_nop_e64
; GCN: v_nop_e64
; GCN: v_nop_e64
; GCN: v_nop_e64
; GCN: v_nop_e64
; GCN: v_nop_e64
; GCN: ;;#ASMEND
; GCN-NEXT: [[LONGBB:BB[0-9]+_[0-9]+]]: ; %loop_body
; GCN-NEXT: ; in Loop: Header=[[LOOP_BODY]] Depth=1
; GCN-NEXT: s_getpc_b64 vcc
; GCN-NEXT: s_sub_u32 vcc_lo, vcc_lo, ([[LONGBB]]+4)-[[LOOP_BODY]]
; GCN-NEXT: s_subb_u32 vcc_hi, vcc_hi, 0
; GCN-NEXT: s_setpc_b64 vcc
[Dominators] Include infinite loops in PostDominatorTree Summary: This patch teaches PostDominatorTree about infinite loops. It is built on top of D29705 by @dberlin which includes a very detailed motivation for this change. What's new is that the patch also teaches the incremental updater how to deal with reverse-unreachable regions and how to properly maintain and verify tree roots. Before that, the incremental algorithm sometimes ended up preserving reverse-unreachable regions after updates that wouldn't appear in the tree if it was constructed from scratch on the same CFG. This patch makes the following assumptions: - A sequence of updates should produce the same tree as a recalculating it. - Any sequence of the same updates should lead to the same tree. - Siblings and roots are unordered. The last two properties are essential to efficiently perform batch updates in the future. When it comes to the first one, we can decide later that the consistency between freshly built tree and an updated one doesn't matter match, as there are many correct ways to pick roots in infinite loops, and to relax this assumption. That should enable us to recalculate postdominators less frequently. This patch is pretty conservative when it comes to incremental updates on reverse-unreachable regions and ends up recalculating the whole tree in many cases. It should be possible to improve the performance in many cases, if we decide that it's important enough. That being said, my experiments showed that reverse-unreachable are very rare in the IR emitted by clang when bootstrapping clang. Here are the statistics I collected by analyzing IR between passes and after each removePredecessor call: ``` # functions: 52283 # samples: 337609 # reverse unreachable BBs: 216022 # BBs: 247840796 Percent reverse-unreachable: 0.08716159869015269 % Max(PercRevUnreachable) in a function: 87.58620689655172 % # > 25 % samples: 471 ( 0.1395104988314885 % samples ) ... in 145 ( 0.27733680163724345 % functions ) ``` Most of the reverse-unreachable regions come from invalid IR where it wouldn't be possible to construct a PostDomTree anyway. I would like to commit this patch in the next week in order to be able to complete the work that depends on it before the end of my internship, so please don't wait long to voice your concerns :). Reviewers: dberlin, sanjoy, grosser, brzycki, davide, chandlerc, hfinkel Reviewed By: dberlin Subscribers: nhaehnle, javed.absar, kparzysz, uabelho, jlebar, hiraditya, llvm-commits, dberlin, david2050 Differential Revision: https://reviews.llvm.org/D35851 llvm-svn: 310940
2017-08-16 02:14:57 +08:00
; GCN-NEXT: [[RET]]: ; %ret
; GCN-NEXT: s_or_b64 exec, exec, [[MASK]]
; GCN: buffer_store_dword
; GCN-NEXT: s_endpgm
define amdgpu_kernel void @analyze_mask_branch() #0 {
entry:
%reg = call float asm sideeffect "v_mov_b32_e64 $0, 0", "=v"()
%cmp0 = fcmp ogt float %reg, 0.000000e+00
br i1 %cmp0, label %loop, label %ret
loop:
%phi = phi float [ 0.000000e+00, %loop_body ], [ 1.000000e+00, %entry ]
call void asm sideeffect
"v_nop_e64
v_nop_e64", ""() #0
%cmp1 = fcmp olt float %phi, 8.0
br i1 %cmp1, label %loop_body, label %ret
loop_body:
call void asm sideeffect
"v_nop_e64
v_nop_e64
v_nop_e64
v_nop_e64", ""() #0
br label %loop
ret:
store volatile i32 7, i32 addrspace(1)* undef
ret void
}
; GCN-LABEL: {{^}}long_branch_hang:
; GCN: s_cmp_lt_i32 s{{[0-9]+}}, 6
Codegen: Make chains from trellis-shaped CFGs Lay out trellis-shaped CFGs optimally. A trellis of the shape below: A B |\ /| | \ / | | X | | / \ | |/ \| C D would be laid out A; B->C ; D by the current layout algorithm. Now we identify trellises and lay them out either A->C; B->D or A->D; B->C. This scales with an increasing number of predecessors. A trellis is a a group of 2 or more predecessor blocks that all have the same successors. because of this we can tail duplicate to extend existing trellises. As an example consider the following CFG: B D F H / \ / \ / \ / \ A---C---E---G---Ret Where A,C,E,G are all small (Currently 2 instructions). The CFG preserving layout is then A,B,C,D,E,F,G,H,Ret. The current code will copy C into B, E into D and G into F and yield the layout A,C,B(C),E,D(E),F(G),G,H,ret define void @straight_test(i32 %tag) { entry: br label %test1 test1: ; A %tagbit1 = and i32 %tag, 1 %tagbit1eq0 = icmp eq i32 %tagbit1, 0 br i1 %tagbit1eq0, label %test2, label %optional1 optional1: ; B call void @a() br label %test2 test2: ; C %tagbit2 = and i32 %tag, 2 %tagbit2eq0 = icmp eq i32 %tagbit2, 0 br i1 %tagbit2eq0, label %test3, label %optional2 optional2: ; D call void @b() br label %test3 test3: ; E %tagbit3 = and i32 %tag, 4 %tagbit3eq0 = icmp eq i32 %tagbit3, 0 br i1 %tagbit3eq0, label %test4, label %optional3 optional3: ; F call void @c() br label %test4 test4: ; G %tagbit4 = and i32 %tag, 8 %tagbit4eq0 = icmp eq i32 %tagbit4, 0 br i1 %tagbit4eq0, label %exit, label %optional4 optional4: ; H call void @d() br label %exit exit: ret void } here is the layout after D27742: straight_test: # @straight_test ; ... Prologue elided ; BB#0: # %entry ; A (merged with test1) ; ... More prologue elided mr 30, 3 andi. 3, 30, 1 bc 12, 1, .LBB0_2 ; BB#1: # %test2 ; C rlwinm. 3, 30, 0, 30, 30 beq 0, .LBB0_3 b .LBB0_4 .LBB0_2: # %optional1 ; B (copy of C) bl a nop rlwinm. 3, 30, 0, 30, 30 bne 0, .LBB0_4 .LBB0_3: # %test3 ; E rlwinm. 3, 30, 0, 29, 29 beq 0, .LBB0_5 b .LBB0_6 .LBB0_4: # %optional2 ; D (copy of E) bl b nop rlwinm. 3, 30, 0, 29, 29 bne 0, .LBB0_6 .LBB0_5: # %test4 ; G rlwinm. 3, 30, 0, 28, 28 beq 0, .LBB0_8 b .LBB0_7 .LBB0_6: # %optional3 ; F (copy of G) bl c nop rlwinm. 3, 30, 0, 28, 28 beq 0, .LBB0_8 .LBB0_7: # %optional4 ; H bl d nop .LBB0_8: # %exit ; Ret ld 30, 96(1) # 8-byte Folded Reload addi 1, 1, 112 ld 0, 16(1) mtlr 0 blr The tail-duplication has produced some benefit, but it has also produced a trellis which is not laid out optimally. With this patch, we improve the layouts of such trellises, and decrease the cost calculation for tail-duplication accordingly. This patch produces the layout A,C,E,G,B,D,F,H,Ret. This layout does have back edges, which is a negative, but it has a bigger compensating positive, which is that it handles the case where there are long strings of skipped blocks much better than the original layout. Both layouts handle runs of executed blocks equally well. Branch prediction also improves if there is any correlation between subsequent optional blocks. Here is the resulting concrete layout: straight_test: # @straight_test ; BB#0: # %entry ; A (merged with test1) mr 30, 3 andi. 3, 30, 1 bc 12, 1, .LBB0_4 ; BB#1: # %test2 ; C rlwinm. 3, 30, 0, 30, 30 bne 0, .LBB0_5 .LBB0_2: # %test3 ; E rlwinm. 3, 30, 0, 29, 29 bne 0, .LBB0_6 .LBB0_3: # %test4 ; G rlwinm. 3, 30, 0, 28, 28 bne 0, .LBB0_7 b .LBB0_8 .LBB0_4: # %optional1 ; B (Copy of C) bl a nop rlwinm. 3, 30, 0, 30, 30 beq 0, .LBB0_2 .LBB0_5: # %optional2 ; D (Copy of E) bl b nop rlwinm. 3, 30, 0, 29, 29 beq 0, .LBB0_3 .LBB0_6: # %optional3 ; F (Copy of G) bl c nop rlwinm. 3, 30, 0, 28, 28 beq 0, .LBB0_8 .LBB0_7: # %optional4 ; H bl d nop .LBB0_8: # %exit Differential Revision: https://reviews.llvm.org/D28522 llvm-svn: 295223
2017-02-16 03:49:14 +08:00
; GCN-NEXT: s_cbranch_scc1 {{BB[0-9]+_[0-9]+}}
; GCN-NEXT: s_branch [[LONG_BR_0:BB[0-9]+_[0-9]+]]
; GCN-NEXT: BB{{[0-9]+_[0-9]+}}:
; GCN: s_add_u32 vcc_lo, vcc_lo, [[LONG_BR_DEST0:BB[0-9]+_[0-9]+]]-(
; GCN: s_setpc_b64
; GCN-NEXT: [[LONG_BR_0]]:
; GCN-DAG: v_cmp_lt_i32
; GCN-DAG: v_cmp_gt_i32
; GCN: s_cbranch_vccnz
; GCN: s_setpc_b64
; GCN: s_setpc_b64
; GCN: [[LONG_BR_DEST0]]
; GCN: s_cbranch_vccz
; GCN: s_setpc_b64
; GCN: s_endpgm
define amdgpu_kernel void @long_branch_hang(i32 addrspace(1)* nocapture %arg, i32 %arg1, i32 %arg2, i32 %arg3, i32 %arg4, i64 %arg5) #0 {
bb:
%tmp = icmp slt i32 %arg2, 9
%tmp6 = icmp eq i32 %arg1, 0
%tmp7 = icmp sgt i32 %arg4, 0
%tmp8 = icmp sgt i32 %arg4, 5
br i1 %tmp8, label %bb9, label %bb13
bb9: ; preds = %bb
%tmp10 = and i1 %tmp7, %tmp
%tmp11 = icmp slt i32 %arg3, %arg4
%tmp12 = or i1 %tmp11, %tmp7
br i1 %tmp12, label %bb19, label %bb14
bb13: ; preds = %bb
call void asm sideeffect
"v_nop_e64
v_nop_e64
v_nop_e64
v_nop_e64", ""() #0
br i1 %tmp6, label %bb19, label %bb14
bb14: ; preds = %bb13, %bb9
%tmp15 = icmp slt i32 %arg3, %arg4
%tmp16 = or i1 %tmp15, %tmp
%tmp17 = and i1 %tmp6, %tmp16
%tmp18 = zext i1 %tmp17 to i32
br label %bb19
bb19: ; preds = %bb14, %bb13, %bb9
%tmp20 = phi i32 [ undef, %bb9 ], [ undef, %bb13 ], [ %tmp18, %bb14 ]
%tmp21 = getelementptr inbounds i32, i32 addrspace(1)* %arg, i64 %arg5
store i32 %tmp20, i32 addrspace(1)* %tmp21, align 4
ret void
}
attributes #0 = { nounwind }
attributes #1 = { nounwind readnone }