[Patch]Enable double to float shrinking optimizations for binary functions like 'fmin/fmax'
yjiang at apple.com
Mon Dec 16 09:25:29 PST 2013
On Dec 13, 2013, at 6:55 PM, Yi Jiang wrote:
> Hi Weiming,
> Thank you for your comments. Refactoring sounds like a good idea. Please see the new patch:
> On Dec 13, 2013, at 10:30 AM, Zhao, Weiming wrote:
>> + Chad
>> Hi Yi,
>> Thanks for the patch.
>> This transformation looks good for me (similar to the cases of floor/ceil).
>> One comment:
>> In EmitBinaryFloatFnCall, the code to append "f"/"l" suffix is the same as EmitUnaryFloatFnCall. Could you consider refactoring them?
>> -----Original Message-----
>> From: llvm-commits-bounces at cs.uiuc.edu [mailto:llvm-commits-bounces at cs.uiuc.edu] On Behalf Of Yi Jiang
>> Sent: Thursday, December 12, 2013 4:29 PM
>> To: LLVM Commits
>> Subject: Re: [Patch]Enable double to float shrinking optimizations for binary functions like 'fmin/fmax'
>> Ping ...
>> On Dec 11, 2013, at 4:45 PM, Yi Jiang <yjiang at apple.com> wrote:
>>> This patch is to enable double to float shrinking optimizations for binary functions like 'fmin/fmax':
>>> fmin((double)floatval1, (double)floatval2) -> (double)fmin(floatval1, floatval2)
>>> This patch will only cover fmin/fmax but the utility functions could allow us to add more functions in the future. Any comments will be highly appreciated.
>> llvm-commits mailing list
>> llvm-commits at cs.uiuc.edu
More information about the llvm-commits