[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Remote access to VSX and VMX128 capable machines?

Konstantinos Margaritis wrote:
> On Wed, 04 Jun 2014 06:45:54 +0900
> Hiroyuki Yamamoto <yama1066@gmail.com> wrote:
>> Hi,
>> VMX128 doesn't seem compatible with VMX on OPcode level as follows:
>> http://biallas.net/doc/vmx128/vmx128.txt
>> https://sourceware.org/ml/binutils/2007-03/msg00075.html
> Hi,
> Thanks for the reply, I've been reading the info in those links myself,
> from what I understand the VMX128 opcodes are an extention, that is,
> they are added on top of the existing VMX opcodes, isn't that correct?
> If not, then you're right and I will go forth and fix the data on the
> tables. Also, I noticed something else, those opcodes, while on 2007
> binutils, are not in the current version of binutils-and I guess that
> there are no respective intrinsics in gcc for that matter. Anyone by
> any chance knows why? If VMX128 is a completely dead and obsolete SIMD
> unit then perhaps I should not include it, it has been very hard
> finding documentation on it as it, the only things I could find were
> these:
> http://download.microsoft.com/download/2/2/0/2208E622-4365-4C00-A6E7-9C3A0190EAAA/Case_Studies_in_VMX128_Optimization_US.zip
> and
> http://download.microsoft.com/download/e/3/7/e3791b88-c41d-4a49-a24c-e4ccfbca4843/Using%20VMX%20and%20SSE%20Instructions%20for%20Performance.zip
> However, these include some -probably informative- docs on
> xds.xbox.com, which does not exist anymore. So the links you mentioned
> and the above are the only info I could find on vmx128.
> Anyway, thanks a lot for your time.

Although I don't check that there is overlap on current opcode yet,
VMX128 instructions use extra 2 bits for distinguishing 128 registers.

Best regards,
Hiroyuki Yamamoto
A75D B285 7050 4BF9 AEDA  91AC 3A10 59C6 5203 04DC

Reply to: