8 Replies Latest reply on Jul 15, 2009 9:15 PM by hagen

    SDK 1.5??

    ryta1203

      The FirestreamSDK 1.4 was released 3/12/2009 and going with AMDs estimated 3 month release timeframe the 1.5 SDK would seem to be overdue... any idea on when AMD might be releasing 1.5?

      Is AMD waiting to have a working version of OpenCL first? Again, will AMD continue to support Brook+ (it seriuosly doesn't look this way)? What about IL (I'm assuming so since OpenCL will most likely be built on top of CAL/IL)?

        • SDK 1.5??
          riza.guntur

          I'm also curious about this issue.

          I'm on my way to make scription using Brook+ and introducing GPGPU technologies to my University fellow for 2nd term 2009.

          It's more confusing now, continue doing my research using Brook+ or switch to CAL/IL.

            • SDK 1.5??
              dukeleto

              one more request for a bit more information:

              it would be great to have a realistic date for when we can expect SDK 1.5, with OpenCL!

              • SDK 1.5??
                ryta1203

                 

                Originally posted by: riza.guntur I'm also curious about this issue.

                I'm on my way to make scription using Brook+ and introducing GPGPU technologies to my University fellow for 2nd term 2009.

                It's more confusing now, continue doing my research using Brook+ or switch to CAL/IL.

                I think that using and getting comfortable with Brook+ is a great way to get used to the Streaming Environment; however, you will get much greater control with CAL/IL. You could get even greater control with CAL/ISA if AMD would fix the assembler.

                  • SDK 1.5??
                    riza.guntur

                    What do you mean by much greater control?

                    How much optimization could we get using CAL/IL or CAL/ISA compared to Brook+?

                     

                      • SDK 1.5??
                        alorenz

                        Well opencl ( according to some ) was to be release with the 1.4 sdk.  However this didn't happen.  Presently the release of onpencl will probably be with 1.5 ?, and this should be soon, as the June 30 release date has passed.  I have not found any anouncement on the forum if there is to be a delay.   As such it should be soon....

                        Also I think opencl (by virtue of being an open standard) will be the right way to go.  Both Amd and Nvidia claim support for this standard, and Nvidia has aready released their opencl development.  This will finally give a good comparison of the power of the respective gpu in solving real world problems fft, mpeg4 encoding, ...

                        If there is  a significant delay, I think Amd might lose whatever market share they have of the gpu/compute market.   Folding at home app was great, but not mcuh since then.

                         

                        • SDK 1.5??
                          ryta1203

                           

                          Originally posted by: riza.guntur What do you mean by much greater control?

                           

                          How much optimization could we get using CAL/IL or CAL/ISA compared to Brook+?

                           

                           

                          The closer you get to the hardware (lower level language) the more control you will have. By control I mean just that... if you code in Brook+ it's  only going to create CAL/IL code that it has been coded to create, if you code directly in CAL/IL then you have more control of what functions are being called/used from CAL and what instructions are being written for the IL. Besides, I doubt Brook+ will be supported much longer.

                            • SDK 1.5??
                              ryta1203

                              So any word on 1.5 yet?

                                • SDK 1.5??
                                  hagen

                                  Instead of patching brook to create sdk 1.4, AMD should have focused their energy on openCL.  Brook+ is an example of why new should not be mixed with old.  It is an odd hybrid of mismatched computing concepts.

                                  Now users like us are stuck.  Should I continue to code in brook+ or wait for openCL?  If I have the option, I would not want to code in brook+.  But applications can't wait, so it looks like when I wake up tomorrow I'll have to write another brook+ code :-( 

                                  Or just cut the loss and switch to Nvidia?