Improve comments for r211040

Added comment to clarify why we r211040 choose to bail out of fast isel instead
of generating a more complicated relocation, and fix mislabelled register in the
comments of the asan test case.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211052 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Louis Gerbarg
2014-06-16 20:31:50 +00:00
parent baf9879290
commit 163eb0990c
2 changed files with 6 additions and 3 deletions

View File

@@ -2709,7 +2709,10 @@ unsigned X86FastISel::TargetMaterializeConstant(const Constant *C) {
// Materialize addresses with LEA instructions.
if (isa<GlobalValue>(C)) {
//LEA can only handle 32 bit immediates
// LEA can only handle 32 bit immediates. Currently this happens pretty
// rarely, so rather than deal with it just bail out of fast isel. If any
// architectures endis up needing to use this path a lot then fast isel
// could get the address with a MOV64ri and use that to load the value.
if (TM.getRelocationModel() == Reloc::Static && Subtarget->is64Bit())
return false;

View File

@@ -9,7 +9,7 @@
; }
; with "clang++ -S -emit-llvm -fsanitize=address -O0 -g test.cc"
; First, argument variable "y" resides in %rdi:
; First, argument variable "y" resides in %rdx:
; CHECK: DEBUG_VALUE: bar:y <- RDX
; Then its address is stored in a location on a stack:
@@ -21,7 +21,7 @@
; CHECK: .Ldebug_loc{{[0-9]+}}:
; We expect two location ranges for the variable.
; First, it is stored in %rdi:
; First, it is stored in %rdx:
; CHECK: .Lset{{[0-9]+}} = .Lfunc_begin0-.Lfunc_begin0
; CHECK-NEXT: .quad .Lset{{[0-9]+}}
; CHECK-NEXT: .Lset{{[0-9]+}} = [[START_LABEL]]-.Lfunc_begin0