[PATCH] Serialize the floating point immediate machine operands.

[Alex L]

Thanks,

Yeah, the debug info stuff is isn't great here, but it's the only way I
could
generate the FP operands on X86. However, I can write tests that utilize
another target's instruction set which allows FP operands in normal
instructions.

My updated patch now includes tests that don't use the debug information
and utilize the NVPTX instruction set to test the floating point operands
directly.

Alex


2015-07-30 17:54 GMT-07:00 Duncan P. N. Exon Smith <dexonsmith at apple.com>:

>
> > On 2015-Jul-30, at 17:41, Alex L <arphaman at gmail.com> wrote:
> >
> > The updated patch below uses the 'parseConstantValue' function from the
> AsmParser
> > to parse the floating point constants. This radically simplifies the
> parsing of floating
> > point machine operands for the MI parser.
> > Also, the updated patch can now handle all the floating point types.
> >
> > Cheers,
> > Alex
> >
> > <0001-MIR-Serialization-Serialize-the-floating-point-opera.patch>
>
> Looks much better.
>
> We usually try to minimize which testcases need debug info.  Do
> you really need it here?  If you just need to exercise `DBG_VALUE`
> with a float, can you put that in a debug-info-specific test (and
> then strip the debug info out of the others)?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20150731/d316ae32/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0001-MIR-Serialization-Serialize-the-floating-point-opera.patch
Type: application/octet-stream
Size: 12385 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20150731/d316ae32/attachment.obj>