This is a discussion on Re: [UNSURE] Re: [UNSURE] Re: [UNSURE] Re: Hardware TPU? - VMS ; Chris Zach wrote: > Johnny Billquist wrote: > >> I believe that you told the compiler to generate code for CIS, just as >> you can tell some compilers wether or not to generate code for FPP, >> FIS or ...
Chris Zach wrote:
> Johnny Billquist wrote:
>> I believe that you told the compiler to generate code for CIS, just as
>> you can tell some compilers wether or not to generate code for FPP,
>> FIS or EIS.
> Correct. COBOL81 for example has a /CIS switch and appropriate
> libraries. With it, a lot of the Cobol math operations run a *lot*
> faster. In fact I recall apps on my 11/23+ with CIS ran much quicker
> than the 11/73.
Thanks for checking it out. Nice to be right for once... :-)
> There was also the CUS option that you could run on an 11/03 with the
> microcode WCS11 module. I think I have that code somewhere; I still have
> that board in the shed. Cute stuff, but I think CUS only ran on the
> DIBOL-11 app.
Did it download special microcode into the WCS11 then? Or how did that
work? Did it expect a certain microcode to already be loaded?
Do you know if something similar existed for the WCS of the 11/60?
>> So yes, you'd need to generate a separate image for those who had CIS
>> and for those who hadn't.
> Also correct. It's the same as compiling FORTRAN77 apps with the /FPP
> (could F77 run /FIS, I remember F4P could). Those apps would only run on
> a system with an FPP. Otherwise you had to link the apps to the local
> libs which really boosted your app's memory usage and slowed it down a lot.
No, F77 can't generate code for the FIS.
F4 still can, though (looking at V4.8 now, which is the latest version).