This is the mail archive of the libc-alpha@sourceware.org mailing list for the glibc project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCHv4 6/8] Convert _Complex tangent functions to generated code



On 08/19/2016 03:18 PM, Joseph Myers wrote:
> On Thu, 18 Aug 2016, Paul E. Murphy wrote:
> 
>> This is fairly straight fowards.  m68k overrides are
>> updated to use the framework, and thus are simplified
>> a bit.
> 
> There are no m68k overrides for these functions, so this description is 
> incorrect, as is the ChangeLog entry:
> 
>> 	* sysdeps/m68k/m680x0/fpu/s_ctanh.c: Refactor into ...
>> 	* sysdeps/m68k/m680x0/fpu/s_ctanh_template.c: New file.
>> 	* sysdeps/m68k/m680x0/fpu/s_ctanhf.c: Removed.
>> 	* sysdeps/m68k/m680x0/fpu/s_ctanhl.c: Removed.

I've removed both sections from the commit and ChangeLog.


>> -	      den = (1.0 - absx) * (1.0 + absx);
>> -	      if (den == -0.0)
>> -		den = 0.0;
>> +	      den = (1 - absx) * (1 + absx);
>> +	      if (den == -0)
>> +		den = 0;
> 
> Saying -0 as an integer is fairly nonsensical, since there is no such 
> integer as distinct from 0.
> 
> Saying == 0, as is done in cacos, is fine here; it's clear enough that 
> this code is converting -0 to +0 (and tests == -0.0 and == 0.0 mean the 
> same thing anyway).  Then catan and catanh should be consistent, so change 
> catanh as well
> 
>> +	      den = (1 - absx) * (1 + absx);
>> +	      if (den == M_LIT (-0.0))
>> +		den = 0;
> 
> so it also uses == 0 in the comparison.
> 
> Patches 5 and 6 are OK with those changes.

I have update both comparisons to "== 0".

Committed as f6d3a72 and d5602ce.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]