IBM To Transition Their z/OS, POWER + AIX Compilers To Being LLVM/Clang-Based

[phoronix]

Phoronix: IBM To Transition Their z/OS, POWER + AIX Compilers To Being LLVM/Clang-Based

IBM last week made the bold announce that they will be transitioning to LLVM/Clang-based compilers across their hardware portfolio for C, C++, and Fortran compilation...

http://www.phoronix.com/scan.php?page=news_item&px=IBM-Will-Use-LLVM-Clang-Stack

[Ironmask]

Cue Stallman angrily jumping up and down screeching that this is the death of freedom.

[CommunityMember]

Cue Stallman angrily jumping up and down screeching that this is the death of freedom.

LLVM may not be the specific type of freedom that some individuals may make a case for, but there is little argument the LLVM license is more free than the existing IBM offerings (which are all closed source proprietary).

At one time the IBM compilers were either the only choice, or at least the better choice, for the IBM platforms due to better code generation or extensions, but with the improvements in LLVM it makes sense to be able to work together on (especially) the front end rather than for IBM to build teams to (essentially) duplicate that effort moving forward. Which will over time make LLVM better for everyone.

[Delgarde]

Maybe a little late to be relevant for me... we've historically had quite a few customers running AIX/Power, but most of them have moved to Intel/Linux in recent years, and the rest are likely to follow next time they buy new hardware. So while it might be interesting to see how much effort is required to switch our AIX build to Clang, I suspect we'll never need to.

[skeevy420]

Michael

IBM, last week, made the bold announcement that

I added commas and, well, just don't think that "made the bold announce that" makes very much sense.

That's quite a bold plan, but not too surprising considering all of the other ISVs and IHVs that have been embracing the LLVM compiler infrastructure for years. Among countless other examples include AMD's "AOCC" compiler that is based on LLVM Clang, Intel has various compilers based on LLVM, and even Microsoft has supported LLVM in various use-cases.

Tweaks to make it easier to read.

As part of this effort, IBM is working on improving the upstream LLVM/Clang support for POWER and Z hardware with better code generation and other improvements. As part of this broad effort, IBM's long-standing XL C/C++ compiler is integrating

I have the same problem of repeating a phrase once it gets stuck in my head and then not catching it when proofreading. I'd either add "also" between "is integrating" and turn the 2nd one into a semicolon or add broad to the first one and tweak the second one to "Also part of the effort is...."

At least you know we do read the articles.

[uid313]

Is there any reason for IBM to continue pursuing z/OS and AIX other than supporting existing customers?

I wish IBM would abandon z/OS and AIX in favor of focusing their everything on Linux.

[kpedersen]

I wish IBM would abandon z/OS and AIX in favor of focusing their everything on Linux.

They pretty much are. I would imagine that this might even be a partial step to get legacy customers to migrate to LLVM/clang, and solve their compilation issues; then in another 10 years go the rest of the way to Linux. Solving compiler differences *and* platform differences in one chew might be a bit too much.

I do like AIX though. Nothing has changed, virtually zero "gotchas" when developing for it. I am hoping IBM will open-source it in the end (for playing / preservation reasons) rather than just take it to the grave.

[edwaleni]

Is there any reason for IBM to continue pursuing z/OS and AIX other than supporting existing customers?

I wish IBM would abandon z/OS and AIX in favor of focusing their everything on Linux.

Because there are still use cases where z/OS is superior to Linux.

[jabl]

Is there any reason for IBM to continue pursuing z/OS and AIX other than supporting existing customers?

Isn't that a pretty big reason? As long as those business units continue operating at a handsome profit, why should they abandon them?

I wish IBM would abandon z/OS and AIX in favor of focusing their everything on Linux.

To an extent, sure, it would be more efficient. OTOH monoculture isn't good either.

[Hugh]

Interesting. Much of their compiler work has been done in IBM's Toronto Lab. I've seen a number of the lab folks at the Toronto LLVM meetup.

If I remember correctly, the lab has been supporting LLVM for Power for a number of years.