[llvm-commits] [llvm] r66922 - in /llvm/trunk: lib/Target/X86/X86InstrInfo.td test/CodeGen/X86/tls13.ll test/CodeGen/X86/tls14.ll

Chris Lattner clattner at apple.com
Fri Mar 13 13:25:19 PDT 2009


On Mar 13, 2009, at 11:52 AM, Dan Gohman wrote:

>
> On Mar 13, 2009, at 11:37 AM, Rafael Espindola wrote:
>
>> Author: rafael
>> Date: Fri Mar 13 13:37:06 2009
>> New Revision: 66922
>>
>> URL: http://llvm.org/viewvc/llvm-project?rev=66922&view=rev
>> Log:
>> Improve sext and zext of TLS variables.
>
> Hi Rafael,
>
> x86 has a large number of instructions which have
> a load or store folded into them. Instead of making a
> TLS version of every one of them, would it make sense to
> extend the x86 address matching code from 4 to 5 operands,
> adding an operand for a TLS base? This would allow any
> instruction with an address operand to be able to talk
> directly to TLS storage.

Instead of making it TLS specific, I'd just make that operand "specify  
the segment".  Even though we do nothing with FS or ES or DS etc, it  
would be nice to be able to have a clean model where the instructions  
think about these machine level things, and isel lowering handles the  
translation of "TLS to GS"

-Chris 



More information about the llvm-commits mailing list