Announcement

Collapse
No announcement yet.

OpenACC Changes Merged Today For GCC 5

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • OpenACC Changes Merged Today For GCC 5

    Phoronix: OpenACC Changes Merged Today For GCC 5

    The OpenACC support was merged for GCC 5 on the final day of permitting new work for this major GPLv3 compiler update...

    http://www.phoronix.com/scan.php?pag...tage-3-Closing

  • #2
    Originally posted by phoronix View Post
    Phoronix: OpenACC Changes Merged Today For GCC 5

    The OpenACC support was merged for GCC 5 on the final day of permitting new work for this major GPLv3 compiler update...

    http://www.phoronix.com/scan.php?pag...tage-3-Closing
    So... if I understand this correctly (please correct me if I don't since I am a noob on this subject), what was commited to GCC is a common interface for different accelerators but the nvidia accelerator is still not there.

    If this is how it works, does anyone know if there is any plans to give OSS Radeon drivers their own accelerator in the future?

    Comment


    • #3
      Should this be called, openMP? or FSA?

      I think that, maybe was a better movo, to use HSA , because it have better performance than opencl...

      Comment


      • #4
        Originally posted by vein View Post
        So... if I understand this correctly (please correct me if I don't since I am a noob on this subject), what was commited to GCC is a common interface for different accelerators but the nvidia accelerator is still not there.
        Well, a common infrastructure for offloading (OpenACC or OpenMP to any accelerator) was committed to GCC a few month ago. Today's commit contains OpenACC-specific things (front-end, run-time, etc.) As I understand it, they plan to support OpenACC to Nvidia offloading, I don't know about OpenACC to Radeon or to Xeon Phi.

        Comment


        • #5
          @AMD devs
          What's the relationship between OpenACC and the new radeon HSA work?
          ## VGA ##
          AMD: X1950XTX, HD3870, HD5870
          Intel: GMA45, HD3000 (Core i5 2500K)

          Comment


          • #6
            Originally posted by Gumix View Post
            Well, a common infrastructure for offloading (OpenACC or OpenMP to any accelerator) was committed to GCC a few month ago. Today's commit contains OpenACC-specific things (front-end, run-time, etc.) As I understand it, they plan to support OpenACC to Nvidia offloading, I don't know about OpenACC to Radeon or to Xeon Phi.
            Actually, the nvptx offloading infracture has been committed too, it is just missing a few small patches to work properly. Those will be hopefully resolved during stage4.

            Comment


            • #7
              Originally posted by JakubJelinek View Post
              Actually, the nvptx offloading infracture has been committed too, it is just missing a few small patches to work properly. Those will be hopefully resolved during stage4.
              So, I understand that we could use our nVidia's GPUs with OpenACC in GCC 5. And in OpenMP? Will be supported the GPU?

              Comment

              Working...
              X