AMD AOCC 1.2 Compiler Released For Zen Systems, Brings FLANG & Retpolines

[phoronix]

Phoronix: AMD AOCC 1.2 Compiler Released For Zen Systems, Brings FLANG & Retpolines

AMD has released a new update to their AMD Optimizing C/C++ Compiler (AOCC)...

http://www.phoronix.com/scan.php?page=news_item&px=AMD-AOCC-1.2-Released

[timofonic]

A closed source fork of LLVM under NDA? Well done, AMD.

Are you trying to imitate Intel's ICC? Amazing, but at least ICC has it's own codebase

This is amazing for Open Source. BSD/MIT predators love that. Sony PlayStation4's OrbitOS (based on FreeBSD) is nothing compared to this

[pegasus]

Looks like it is time to add some fortran code to benchmarks ... I'll look around if there's anything modest that could be useful.

[xiando]

>Extends LLVM 6.0 trunk (llvm.org, 30th Nov) with enhancements and optimizations
Dear AMD, why don't you simply submit these enhancements and optimizations to LLVM? Slamming your own brand on LLVM and calling it your own product and requiring a NDA for good measure seems like a rather rude move.

[boboviz]

A closed source fork of LLVM under NDA? Well done, AMD.

I think they will do like CodeXL. After some very buggy and poor initial releases, when the code is sufficient mature, they put it in open source..

[bridgman]

A closed source fork of LLVM under NDA? Well done, AMD.

Are you sure ? I looked through all the download materials and didn't see any kind of NDA. There is a standard EULA but that is not the same as an NDA.

BTW I think the LLVM changes do all go upstream - AOCC is just a fast path for release.

[timofonic]

Are you sure ? I looked through all the download materials and didn't see any kind of NDA. There is a standard EULA but that is not the same as an NDA.

BTW I think the LLVM changes do all go upstream - AOCC is just a fast path for release.

Access to the code and such requires NDA, so it's under NDA. No user requires to sign the NDA but to read an ugly EULA (another thing I dislike).

In my open source software mindset, it requires a NDA to be useful and minimally trustful.

LLVM changes maybe go upstream (you think it, not 100% confirmed), but the project is a lot more than that. I fail to see the strategy behind AOCC too, just to get some performance improvements probably only useful for AMD platforms (is AOCC used in PlayStation4 and Xbox One games?).

A proprietary compiler might make sense for Intel because they have +90% market and can waste money/resources on crazy twisted evil plans, instead on having *A LOT* better and more open (without ME/PSP crap stuff) products (this is some kind of fake oligopoly, it's mostly a monopoly in the desktop world). I think AMD is not in the place to waste their *A LOT MORE LIMITED* resources, I even think AMD Company should improve a lot more on using them a lot more efficiently. I consider it's a waste for AMD to waste their resurces in proprietary shit, IMHO.

I hope RISC-V becomes extremely sucessful someday

[bridgman]

Access to the code and such requires NDA, so it's under NDA. No user requires to sign the NDA but to read an ugly EULA (another thing I dislike). In my open source software mindset, it requires a NDA to be useful and minimally trustful.

Ahh, OK - it's a binary release but you are assuming that source code would be available under NDA (actually a source code license). I don't know if that is actually the case but it seems like a reasonable assumption.

LLVM changes maybe go upstream (you think it, not 100% confirmed), but the project is a lot more than that. I fail to see the strategy behind AOCC too, just to get some performance improvements probably only useful for AMD platforms (is AOCC used in PlayStation4 and Xbox One games?).

A proprietary compiler might make sense for Intel because they have +90% market and can waste money/resources on crazy twisted evil plans, instead on having *A LOT* better and more open (without ME/PSP crap stuff) products (this is some kind of fake oligopoly, it's mostly a monopoly in the desktop world). I think AMD is not in the place to waste their *A LOT MORE LIMITED* resources, I even think AMD Company should improve a lot more on using them a lot more efficiently. I consider it's a waste for AMD to waste their resurces in proprietary shit, IMHO.

When you say "a proprietary compiler" do you mean "binary releases" ? If so then I don't really understand your logic - binary releases are a lot less effort than open source (ie the opposite of being wasteful), and I suspect that is what drove the decision to start with binary releases in the first place.

"Just because something is evil doesn't mean it takes more time"

I hope RISC-V becomes extremely sucessful someday

You want to bring the proprietary world to RISC-V as well ?