Displaying 3 results from an estimated 3 matches for "r130801".
2011 Jun 22
0
[LLVMdev] Register class proliferation
...egister aliases, not classes. Register aliasing is a serious problem with the current implementation. The postRA scheduler is *cubic* in the size of the alias list. You dramatically improved compile time with this simple fix:
------------------------------------------------------------------------
r130801 | stoklund | 2011-05-03 15:31:24 -0700 (Tue, 03 May 2011) | 10 lines
Mark ultra-super-registers QQQQ as call-clobbered instead of the D sub-registers.
And the new regalloc is not yet designed to scale with the number of register aliases. (But that could be fixed under some limited definition of al...
2011 Jun 21
2
[LLVMdev] Register class proliferation
On Jun 21, 2011, at 9:23 AM, Jim Grosbach wrote:
>
> On Jun 21, 2011, at 8:51 AM, Jakob Stoklund Olesen wrote:
>
>> In the past, I've seen some pushback on the list against adding more register classes. You can see it in the code as well, TargetLowering::getRegClassForInlineAsmConstraint() returns a vector of registers instead of a real register class.
>>
>> What
2011 Jun 22
2
[LLVMdev] Register class proliferation
...issues.
> Register aliasing is a serious problem with the current implementation. The postRA scheduler is *cubic* in the size of the alias list. You dramatically improved compile time with this simple fix:
>
> ------------------------------------------------------------------------
> r130801 | stoklund | 2011-05-03 15:31:24 -0700 (Tue, 03 May 2011) | 10 lines
> Mark ultra-super-registers QQQQ as call-clobbered instead of the D sub-registers.
Completely unintended, I might add.
Our representation of call clobbered registers is bad enough that it deserves special attention. Every ca...