RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-10-19 11:02:59 +02:00
Code Issues Projects Releases Wiki Activity

[DAG] Prevent NodeId pruning of TokenFactors in Instruction Selection.

Browse Source 
Summary:
Instruction Selection preserves relative orders of all nodes save
TokenFactors which we treat specially. As a result Node Ids for
TokenFactors may violate the topological ordering and should not be
considered as valid pruning candidates in predecessor search.

Fixes PR35316.

Reviewers: RKSimon, hfinkel

Subscribers: hiraditya, llvm-commits

Differential Revision: https://reviews.llvm.org/D42701

llvm-svn: 323880
This commit is contained in:
Nirav Dave 2018-01-31 15:23:17 +00:00
parent 079ff9a80c
commit 0e452c9531
2 changed files with 79 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp
View File

@ -2157,7 +2157,9 @@ static bool findNonImmUse(SDNode *Use, SDNode* Def, SDNode *ImmedUse,
while (!WorkList.empty()) {
Use = WorkList.back();
WorkList.pop_back();
if (Use->getNodeId() < Def->getNodeId() && Use->getNodeId() != -1)
// NodeId topological order of TokenFactors is not guaranteed. Do not skip.
if (Use->getOpcode() != ISD::TokenFactor &&
Use->getNodeId() < Def->getNodeId() && Use->getNodeId() != -1)
continue;
// Don't revisit nodes if we already scanned it and didn't fail, we know we

76
test/CodeGen/X86/pr35316.ll Normal file
View File

@ -0,0 +1,76 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-linux | FileCheck %s --check-prefix=CHECK
; This tests for a cyclic dependencies in the generated DAG.
@c = external local_unnamed_addr global i32, align 4
@a = external local_unnamed_addr global i32, align 4
@b = external local_unnamed_addr global i32, align 4
define void @main() {
; CHECK-LABEL: main:
; CHECK: # %bb.0: # %entry
; CHECK-NEXT: movl -{{[0-9]+}}(%rsp), %eax
; CHECK-NEXT: movl -{{[0-9]+}}(%rsp), %eax
; CHECK-NEXT: movl -{{[0-9]+}}(%rsp), %eax
; CHECK-NEXT: movl -{{[0-9]+}}(%rsp), %eax
; CHECK-NEXT: movl -{{[0-9]+}}(%rsp), %eax
; CHECK-NEXT: movl -{{[0-9]+}}(%rsp), %eax
; CHECK-NEXT: movl -{{[0-9]+}}(%rsp), %eax
; CHECK-NEXT: movl -{{[0-9]+}}(%rsp), %eax
; CHECK-NEXT: movl -{{[0-9]+}}(%rsp), %eax
; CHECK-NEXT: movl -{{[0-9]+}}(%rsp), %eax
; CHECK-NEXT: movl -{{[0-9]+}}(%rsp), %eax
; CHECK-NEXT: movl -{{[0-9]+}}(%rsp), %eax
; CHECK-NEXT: movl -{{[0-9]+}}(%rsp), %eax
; CHECK-NEXT: movl -{{[0-9]+}}(%rsp), %eax
; CHECK-NEXT: movl -{{[0-9]+}}(%rsp), %eax
; CHECK-NEXT: movl $0, {{.*}}(%rip)
; CHECK-NEXT: movl -{{[0-9]+}}(%rsp), %r8d
; CHECK-NEXT: movl -{{[0-9]+}}(%rsp), %edi
; CHECK-NEXT: movl -{{[0-9]+}}(%rsp), %esi
; CHECK-NEXT: movl -{{[0-9]+}}(%rsp), %eax
; CHECK-NEXT: cltd
; CHECK-NEXT: idivl {{.*}}(%rip)
; CHECK-NEXT: movl %eax, %ecx
; CHECK-NEXT: movl {{.*}}(%rip), %eax
; CHECK-NEXT: cltd
; CHECK-NEXT: idivl %esi
; CHECK-NEXT: andl %edi, %eax
; CHECK-NEXT: addl %ecx, %eax
; CHECK-NEXT: andl %r8d, %eax
; CHECK-NEXT: movl %eax, (%rax)
; CHECK-NEXT: retq
entry:
%e = alloca i32, align 4
%e.0.e.0.24 = load volatile i32, i32* %e, align 4
%e.0.e.0.25 = load volatile i32, i32* %e, align 4
%e.0.e.0.26 = load volatile i32, i32* %e, align 4
%e.0.e.0.27 = load volatile i32, i32* %e, align 4
%e.0.e.0.28 = load volatile i32, i32* %e, align 4
%e.0.e.0.29 = load volatile i32, i32* %e, align 4
%e.0.e.0.30 = load volatile i32, i32* %e, align 4
%e.0.e.0.31 = load volatile i32, i32* %e, align 4
%e.0.e.0.32 = load volatile i32, i32* %e, align 4
%e.0.e.0.33 = load volatile i32, i32* %e, align 4
%e.0.e.0.34 = load volatile i32, i32* %e, align 4
%e.0.e.0.35 = load volatile i32, i32* %e, align 4
%e.0.e.0.36 = load volatile i32, i32* %e, align 4
%e.0.e.0.37 = load volatile i32, i32* %e, align 4
%e.0.e.0.39 = load volatile i32, i32* %e, align 4
%tmp = load i32, i32* @a, align 4
store i32 0, i32* @b, align 4
%e.0.e.0.41 = load volatile i32, i32* %e, align 4
%add17 = add nsw i32 %e.0.e.0.41, 0
%e.0.e.0.42 = load volatile i32, i32* %e, align 4
%tmp1 = load i32, i32* @c, align 4
%e.0.e.0.43 = load volatile i32, i32* %e, align 4
%div = sdiv i32 %tmp1, %e.0.e.0.43
%and18 = and i32 %div, %e.0.e.0.42
%e.0.e.0.44 = load volatile i32, i32* %e, align 4
%div19 = sdiv i32 %e.0.e.0.44, %tmp
%add20 = add nsw i32 %div19, %and18
%and21 = and i32 %add20, %add17
store volatile i32 %and21, i32* undef, align 4
ret void
}