Level1Ramble: What's Up With Nvidia + ARM?
Added 2020-09-18 16:00:19 +0000 UTCWendell gives us his thoughts on Nvidia's recent acquisition of ARM. What does this mean for the future of x86?
-Grant
Comments
Damn ENTER key... Sorry. But consider INTEL's hundreds of instructions for each vector width, requiring new coding to support each new "architecture" (like having to learn to drive again because they added a lane to a highway). Worse design follows bad design...
Robert Bernecky
2020-09-18 21:24:45 +0000 UTCRe baggage: INTEL has been creating rubbish CPUs since the 8008, and never did learn. My worst favorite example is the vector instructions. IBM introduced the "Vector Facility" in 1985( https://en.wikipedia.org/wiki/IBM_3090). It featured a register that told you the number of vector registers that were on the system you were using, so you write a loop that would work, unchanged, on any system's vector hardware, regardless of its register count. Yes, it needed a loop, but
Robert Bernecky
2020-09-18 21:21:53 +0000 UTC
