RPCSX/llvm
1
0
Fork 0
mirror of https://github.com/RPCSX/llvm.git synced 2024-11-24 04:09:45 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

[X86] Don't fold away the memory operand of an xchg.

Browse Source 
xchg with a mem operand has different locking semantics. If we unfold it
into a xchg r,r we will loose the implicit lock. Likewise we never want
to fold a register xchg into a memory one as it would be a lot slower.

This triggers during LLVM selfhost.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@304163 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Benjamin Kramer 2017-05-29 16:25:20 +00:00
parent 98be03e2e2
commit 3259254476
2 changed files with 45 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

37
test/CodeGen/X86/xchg-nofold.ll Normal file
View File

@ -0,0 +1,37 @@
; RUN: llc -mtriple=x86_64-linux-gnu < %s | FileCheck %s
%"struct.std::atomic" = type { %"struct.std::atomic_bool" }
%"struct.std::atomic_bool" = type { %"struct.std::__atomic_base" }
%"struct.std::__atomic_base" = type { i8 }
; CHECK-LABEL: _Z3fooRSt6atomicIbEb
define zeroext i1 @_Z3fooRSt6atomicIbEb(%"struct.std::atomic"* nocapture dereferenceable(1) %a, i1 returned zeroext %b) nounwind {
entry:
%frombool.i.i = zext i1 %b to i8
%_M_i.i.i = getelementptr inbounds %"struct.std::atomic", %"struct.std::atomic"* %a, i64 0, i32 0, i32 0, i32 0
%0 = ptrtoint i8* %_M_i.i.i to i64
%1 = lshr i64 %0, 3
%2 = add i64 %1, 2147450880
%3 = inttoptr i64 %2 to i8*
%4 = load i8, i8* %3
%5 = icmp ne i8 %4, 0
br i1 %5, label %6, label %11
; <label>:6: ; preds = %entry
%7 = and i64 %0, 7
%8 = trunc i64 %7 to i8
%9 = icmp sge i8 %8, %4
br i1 %9, label %10, label %11
; <label>:10: ; preds = %6
call void @__asan_report_store1(i64 %0)
call void asm sideeffect "", ""()
unreachable
; <label>:11: ; preds = %6, %entry
store atomic i8 %frombool.i.i, i8* %_M_i.i.i seq_cst, align 1
; CHECK: xchgb %{{.*}}, (%{{.*}})
ret i1 %b
}
declare void @__asan_report_store1(i64)

9
utils/TableGen/X86FoldTablesEmitter.cpp
View File

@ -77,7 +77,14 @@ const char *const NoFoldSet[] = {
"TCRETURNri64",
"TCRETURNmi64", // Special dealing (in X86InstrCompiler.td under
"TCRETURNri", // "tailcall stuff" section).
"TCRETURNmi"
"TCRETURNmi",
// Never fold XCHG, the register and memory forms have different locking
// semantics.
"XCHG8rr", "XCHG8rm",
"XCHG16rr", "XCHG16rm",
"XCHG32rr", "XCHG32rm",
"XCHG64rr", "XCHG64rm",
// Different calculations of the folded operand between
// memory and register forms (folding is illegal).