[llvm] r192631 - Revert part of a fix from 2010, changes since then:
Eric Christopher
echristo at gmail.com
Mon Oct 14 14:52:27 PDT 2013
Author: echristo
Date: Mon Oct 14 16:52:26 2013
New Revision: 192631
URL: http://llvm.org/viewvc/llvm-project?rev=192631&view=rev
Log:
Revert part of a fix from 2010, changes since then:
a) x86-64 TLS has been documented
b) the code path should use movq for the correct relocation
to be generated.
I've also added a fixme for the test case that we should improve
the code generated, it should look something like is documented
in the tls abi document.
Modified:
llvm/trunk/lib/Target/X86/X86InstrCompiler.td
llvm/trunk/test/CodeGen/X86/x86-64-tls-1.ll
Modified: llvm/trunk/lib/Target/X86/X86InstrCompiler.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86InstrCompiler.td?rev=192631&r1=192630&r2=192631&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86InstrCompiler.td (original)
+++ llvm/trunk/lib/Target/X86/X86InstrCompiler.td Mon Oct 14 16:52:26 2013
@@ -969,7 +969,7 @@ def : Pat<(store (i64 (X86Wrapper tblock
// tls has some funny stuff here...
// This corresponds to movabs $foo at tpoff, %rax
def : Pat<(i64 (X86Wrapper tglobaltlsaddr :$dst)),
- (MOV64ri tglobaltlsaddr :$dst)>;
+ (MOV64ri32 tglobaltlsaddr :$dst)>;
// This corresponds to add $foo at tpoff, %rax
def : Pat<(add GR64:$src1, (X86Wrapper tglobaltlsaddr :$dst)),
(ADD64ri32 GR64:$src1, tglobaltlsaddr :$dst)>;
Modified: llvm/trunk/test/CodeGen/X86/x86-64-tls-1.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/x86-64-tls-1.ll?rev=192631&r1=192630&r2=192631&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/x86-64-tls-1.ll (original)
+++ llvm/trunk/test/CodeGen/X86/x86-64-tls-1.ll Mon Oct 14 16:52:26 2013
@@ -1,6 +1,10 @@
; RUN: llc < %s -mtriple=x86_64-unknown-linux-gnu | FileCheck %s
@tm_nest_level = internal thread_local global i32 0
define i64 @z() nounwind {
-; CHECK: movabsq $tm_nest_level at TPOFF, %rcx
+; FIXME: The codegen here is primative at best and could be much better.
+; The add and the moves can be folded together.
+; CHECK: movq $tm_nest_level at TPOFF, %rcx
+; CHECK: movq %fs:0, %rax
+; CHECK: addl %ecx, %eax
ret i64 and (i64 ptrtoint (i32* @tm_nest_level to i64), i64 100)
}
More information about the llvm-commits
mailing list