News for hackers. Stuff that matters.
Re:RISC-V on the Verge of Broad Adoption by IshKebab on Thursday February 14, 2019 @05:08PM (#19166540)
I'm still unconvinced by the modular nature of the ISA. Are we going to end up shipping fat binaries with a dozen binaries for all the different variants that are allowed?

On x86-64 there's a reasonable minimum supported set of instructions, and you only really need runtime CPU detection for the newer vector extensions.

Re:RISC-V on the Verge of Broad Adoption by audunw on Thursday February 14, 2019 @06:07PM (#19166957)
I wouldn't expect server/desktop/phone variants of RISC-V to drop any of the current extensions of the ISA. Why would you, if gate count isn't what you're optimising for?

There could be further, more advanced extension that may not be in every processor, but that's just the same as x86

The modularity in the ISA as it is now, is something that's more geared towards embedded devices or FPGAs, where you'll be compiling code specifically for that target anyway.

Re:RISC-V on the Verge of Broad Adoption by kabacha on Thursday February 14, 2019 @11:52PM (#19168535)
Easy solution: Just GPL it - every extension of RISC-V has to be open as well.
Re:RISC-V on the Verge of Broad Adoption by IshKebab on Friday February 15, 2019 @02:12PM (#19173776)
I don't see how that is a solution. The problem is the variability in what the hardware implements, not the availability of extension specifications.