[llvm] r218668 - [AVX512] Added intrinsics for VPCMPEQB and VPCMPEQW.

[Juergen Ributzka]

In this case I would prefer that the V64 type is moved out of the common values section, because it pushed out the ARG type.
The ARG type is used quite often and by all targets, so I really want to keep it in the efficient encoding section.

> On Oct 14, 2014, at 7:55 AM, Robert Khasanov <rob.khasanov at gmail.com> wrote:
> 
> Hi Juergen,
> 
> AFAIK, X86 is the only target that uses IIT_V64. Since AVX-512 about 50-100 intrinsics will be enabled with this operand type.
> 
> Robert
> 
> 2014-10-09 21:58 GMT+04:00 Juergen Ributzka <juergen at apple.com <mailto:juergen at apple.com>>:
> Hi Robert,
> 
> I don’t think this comment is correct. We can only encode up to 8bit in the short encoding format.
> 
> Also: How common is IIT_V64? I am just wondering because X86 is the only target that uses it for now.
> 
> Thanks
> 
> Cheers,
> Juergen
> 
>> On Sep 30, 2014, at 4:32 AM, Robert Khasanov <rob.khasanov at gmail.com <mailto:rob.khasanov at gmail.com>> wrote:
>> 
>> -  // Common values should be encoded with 0-15.
>> +  // Common values should be encoded with 0-16.
>>   IIT_Done = 0,
>>   IIT_I1   = 1,
>>   IIT_I8   = 2,
>> @@ -489,23 +489,24 @@ enum IIT_Info {
>>   IIT_V8   = 11,
>>   IIT_V16  = 12,
>>   IIT_V32  = 13,
>> -  IIT_PTR  = 14,
>> -  IIT_ARG  = 15,
>> +  IIT_V64  = 14,
>> +  IIT_PTR  = 15,
>> +  IIT_ARG  = 16,
> 
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20141014/e038b2fb/attachment.html>