i860: Re: modern stuff

[Ken Seefried]

>the i860 found at least a little niche on graphics boards, so somehow
>not a complete failure ;-)

I'd be mildly surprised if Intel ever made enough from selling i860s
as GPUs to cover the cost of developing and marketing them.  At the
time, Intel was pushing them as their RISC processor, and put a lot
into the program.  Going to take over the world and all that.  Maybe
not a 'complete' failure...just mostly.

From: Chuck Guzis <cclist at sydex.com>
>On 10/26/18 6:10 AM, Gordon Henderson via cctalk wrote:
>
>> However it was a royal PITA to code for although a 32-bit CPU, it would
>> read memory 64 bits at a time (actually 128 IIRC to satisfy the cache),
>> with half that 64-bit word being an instruction for the integer unit and
>> half for the floating point unit, so you effectively had to build a
>> floating point pipeline by hand coded instructions, so 8 (I think)
>> instructions to load the pipeline, then each subsequent instruction
>> would feed another value into the pipe, then another 8 at the end to
>> empty it. Great for big matrix operations, rubbish for a single add of 2
>> FP numbers.
>
>My impression of the i860 was that it might have been fun for about 2
>weeks for which to code assembly, but after that, you'd really start
>looking hard for an HLL to do the dirty work for you.  While there's a
>sense of accomplishment over looking at a page of painfully
>hand-optimized code that manages to keep everything busy with no
>"bubbles", you begin to wonder if there isn't a better way to spend your
>life.

It wasn't fun for the whole 2 weeks.  And the i860 is Yet Another
example of Intel claiming their compilers were going to be so smart
that all the architectural complexity/warts will never be noticed.
Wrong, and they didn't learn and said the same thing about Itanium.
The interrupt stall issue that Gordon pointed out was so bad they were
basically relegated to single-task software in the end.

KJ