[Bug c/64457] New: NVCC like features CUDA/OpenCL support for GCC to use GPU's to speed up compile jobs.

[Bug c/64457] New: NVCC like features CUDA/OpenCL support for GCC to use GPU's to speed up compile jobs.

[mike at michaellawrenceit dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64457

            Bug ID: 64457
           Summary: NVCC like features CUDA/OpenCL support for GCC to use
                    GPU's to speed up compile jobs.
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Severity: enhancement
          Priority: P3
         Component: c
          Assignee: unassigned at gcc dot gnu.org
          Reporter: mike at michaellawrenceit dot com

WISH list: add use of GPU for GCC ,

 Big beefy GPU Black/TESLA , etc, (Dream of winning the lotto too to pay for
 em too lol)

 add ability to use OpenCL or CUDA to run GCC on GPU cores , for IE Kernel
 compiles , Larger Jobs. (or have cuda/openCL like features built in that can
use such GPU's) 

 gen2 I-7 16 Gigs of ram, atm its quite slow as hell .... ADD in GPU might
 speed it up, Linux Kernel modules , if CUDA many Linux modules could be done
on GPU in dynamic parallels 30-60 minuets to compile now however GPU support
might speed up greatly. 

 GPU as CO-CPU's ?

 swap to console , LET GCC consume most of the GPU , to help Boost compile
 jobs.... =E2=80=8B

 Granted I like Games and MMO's but I do IT-Sec research , and many
 applications that use GPU
 IE pyrit or other Key crackers on CPU ......................

 One my know WEP is junk for security , Back when working on BS it took
 30-60 mins to break the keys, but the comic book store has WEP on its
 wifi...... I don't take or publish they key but it's a good benchmark for
 pyrit etc version updates , TIME: KEY , GPU I have the key again in often
 30 seconds or Less.  CPU it's going to be awhile.......

 460 M (mobile) isn't the fastest GPU for a laptop these days, (336 CUDA
 Cores )
 *however putting them to work for GCC under Linux/Windows would be a
 nicety... *
 *for the AMD RADEON fan-boy types opencl   etc... *

 (https://developer.nvidia.com/cuda-toolkit) CUDA *NVCC*  Like functionality
 for GCC built in as an optimization plugin IE like LLVM gold. etc.

 Programing ? ha My dyslexic hide is more of an Architect , python basics
 etc. however Run Gentoo or related boxes long enough, any speed ups in code
 ie LLVM or runtime hacks you can get  are appreciated. ( I see apps
 in my head like Legos but cant build them to save my ass less its a basic
  scripting.)


 *Emerge -av gcc   +cuda + opencl   ,   YES PLEASE.... *

 *O/S  Sabayon With Spike repo (Gentoo) *

 *Gentoo packages can take a While to build.  however if native NVIDA-CUDA/
 OpenCL (AMD etc) GPU/APU were built in compile times could fall quite
 dramatically... *

[Bug c/64457] NVCC like features CUDA/OpenCL support for GCC to use GPU's to speed up compile jobs.

[mike at michaellawrenceit dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64457

--- Comment #1 from Michael R. Lawrence <mike at michaellawrenceit dot com> ---
for Masters degree studies.  in IT security , many projects used for Forensics
use GPU, as well Gentoo or arch (with repos) has much tools needed for IT Sec
research and Management. 

being able to compile them faster a plus.

[Bug middle-end/64457] NVCC like features CUDA/OpenCL support for GCC to use GPU's to speed up compile jobs.

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64457

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|c                           |middle-end

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
I think this is already implemented as part of OpenMP 4.0 offloading which
includes support for nvptx (well nvptx is GCC 5).

[Bug middle-end/64457] NVCC like features CUDA/OpenCL support for GCC to use GPU's to speed up compile jobs.

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64457

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jakub at gcc dot gnu.org

--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Only OpenACC offloading to nvptx is in latest stages before merge (hopefully it
will merge into trunk in the first half of January), OpenMP offloading to nvptx
will need some further work.  But yeah, I think OpenMP or OpenACC is
preferrable over CUDA here, which is by design specific to a single hw
implementation.

[Bug middle-end/64457] NVCC like features CUDA/OpenCL support for GCC to use GPU's to speed up compile jobs.

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64457

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|---                         |FIXED
   Target Milestone|---                         |5.0

--- Comment #4 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Fixed in GCC 5+, see https://gcc.gnu.org/wiki/Offloading for the support
options and even how to do it.