[llvm-commits] [llvm] r95171 - /llvm/trunk/lib/Target/X86/AsmPrinter/X86MCInstLower.cpp

Dale Johannesen dalej at apple.com
Tue Feb 2 16:56:44 PST 2010


On Feb 2, 2010, at 4:48 PMPST, John McCall wrote:

> On Feb 2, 2010, at 4:36 PM, Dale Johannesen wrote:
>> Author: johannes
>> Date: Tue Feb  2 18:36:40 2010
>> New Revision: 95171
>>
>> URL: http://llvm.org/viewvc/llvm-project?rev=95171&view=rev
>> Log:
>> Print FPImm a less kludgy way; APFloat.toString seems
>> to have some problems anyway.
>
> Can you elaborate?  I mean, I think you're better off staying in hex  
> anyway, but if there are obvious problems in APFloat::toString I'd  
> be happy to fix them.

It hit the assertion at line 3417 when converting a number which was  
double 873.1834 in the source, and appeared to be represented  
correctly as an APFloat.   The 2nd and 3rd parameters to toString were  
0.  I didn't investigate it farther than that.

If all the bits matter, e.g. when generating assembly code, hex is  
better and almost necessary.  This is a comment.




More information about the llvm-commits mailing list