Zachary Turner via llvm-dev
2017-May-17 14:38 UTC
[llvm-dev] PSA: Parallel STL algorithms available in LLVM
Yes, I would hate for some library implementer to either interpret the standard differently or simply not consider the issue of recursive parallelism at all and end up with an implementation that doesn't support it (not that unlikely considering it went 1.5 years through committee as you said and the topic never came up). On Wed, May 17, 2017 at 7:25 AM Bryce Lelbach <balelbach at lbl.gov> wrote:> On Fri, May 12, 2017 at 9:19 AM, Zachary Turner <zturner at google.com> > wrote: > > Even without a concrete use case, I agree that it's absolutely imperative > > for the standard to require this of a conforming implementation. It's > going > > to be the source of so many problems otherwise > > To confirm - what you'd like is clarification in the standard that > recursive parallelism is supported? > > I think this is feasible; I'd suggest a non-normative note. I could > write a short committee paper on this (targeting C++20) for the next > meeting. > > -- > Bryce Adelstein Lelbach aka wash > Lawrence Berkeley National Laboratory > ISO C++ Committee Member > CppCon and C++Now Program Chair > > Compiler ICE Hunter > -- >-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20170517/02ac38e9/attachment-0001.html>
Zachary Turner via llvm-dev
2017-May-17 14:56 UTC
[llvm-dev] PSA: Parallel STL algorithms available in LLVM
When you say a "non normative note" does this imply that a library implementer would be free to ignore the note? On Wed, May 17, 2017 at 7:38 AM Zachary Turner <zturner at google.com> wrote:> Yes, I would hate for some library implementer to either interpret the > standard differently or simply not consider the issue of recursive > parallelism at all and end up with an implementation that doesn't support > it (not that unlikely considering it went 1.5 years through committee as > you said and the topic never came up). > > > On Wed, May 17, 2017 at 7:25 AM Bryce Lelbach <balelbach at lbl.gov> wrote: > >> On Fri, May 12, 2017 at 9:19 AM, Zachary Turner <zturner at google.com> >> wrote: >> > Even without a concrete use case, I agree that it's absolutely >> imperative >> > for the standard to require this of a conforming implementation. It's >> going >> > to be the source of so many problems otherwise >> >> To confirm - what you'd like is clarification in the standard that >> recursive parallelism is supported? >> >> I think this is feasible; I'd suggest a non-normative note. I could >> write a short committee paper on this (targeting C++20) for the next >> meeting. >> >> -- >> Bryce Adelstein Lelbach aka wash >> Lawrence Berkeley National Laboratory >> ISO C++ Committee Member >> CppCon and C++Now Program Chair >> >> Compiler ICE Hunter >> -- >> >-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20170517/67d4175c/attachment.html>
Marshall Clow via llvm-dev
2017-May-25 00:35 UTC
[llvm-dev] PSA: Parallel STL algorithms available in LLVM
On Wed, May 17, 2017 at 8:56 AM, Zachary Turner via llvm-dev < llvm-dev at lists.llvm.org> wrote:> When you say a "non normative note" does this imply that a library > implementer would be free to ignore the note?Yes. That's what "non-normative" means. -- Marshall P.S. I expect to land all the new algorithms (the non-parallel versions!) in libc++ this week or next week. P.P.S. I would be happy to work with other people on the infastructure for libc++ that Bryce described. That idea (having a replaceable parallelism engine that could be replaced by rebuilding libc++) is what I had in mind. -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20170524/bd78d111/attachment.html>