Re:The first conformant M1 GPU driver
by saagarjha on Wednesday August 23, 2023 @02:40AM
(#37232544)
I’m sure they appreciate the work but Apple has internal documentation on their chips. It’s not like they make a black box and forget how they designed them when it comes to writing software for them.
[ Parent
| on HN ]
Re:The first conformant M1 GPU driver
by ironlenny on Wednesday August 23, 2023 @02:43AM
(#37232563)
Some times an outsider's attempt to reverse engineer your product brings new insights.
Re:The first conformant M1 GPU driver
by saagarjha on Wednesday August 23, 2023 @02:44AM
(#37232571)
That is true, but one does not simply implement a fully conformant bit shuffle instruction by accident.
Re:The first conformant M1 GPU driver
by dev_tty01 on Wednesday August 23, 2023 @10:45AM
(#37236832)
Or sometimes an instruction isn't in the compiler because the hardware has some strange corner case failures so they aren't using it. I would be nervous about depending upon an unused instruction.
The article claims that Apple's compiler doesn't use the instruction, so it seems like the compiler author did indeed forget or never learn about it.