Fused Multiply-Add and HSA

Article: AMD's Analyst Update
By: Richard H. (no-spam-please.delete@this.gmail.com), February 7, 2012 7:29 pm
Room: Moderated Discussions
David Kanter wrote:
"Software that adopts FMA can theoretically double compute performance"

Please explain why. On Intel's processors, a SIMD multiply instruction plus a SIMD add instruction can be completed every clock. I don't understand how a Fused Multiply-Add (FMA) increases peak performance at all compared to what Intel already does (not even counting 128-bit SIMD vs 256-bit SIMD).

I'd like to hear people's opinions about AMD's Heterogeneous System Architecture (HSA). I am struggling to understand how it is possible to have an ISA-agnostic interface (therefore needing a Just-In-Time compiler) without paying a big performance penalty. The whole purpose of heterogeneous computing is to get more performance so why would it make sense to use a Just-In-Time compiler?

It seems like the vast majority of software vendors that make use of GPUs for computing only support Nvidia GPUs (for example, Adobe Photoshop, Adobe Premiere and Neat Video). What does HSA provide that could change this situation?
 Next Post in Thread >
TopicPosted ByDate
Fused Multiply-Add and HSARichard H.02/07/12 07:29 PM
  Fused Multiply-Add and HSADavid Kanter02/07/12 11:54 PM
    "CUDA is on its way out"pk02/09/12 05:37 AM
      "CUDA is on its way out"David Kanter02/09/12 11:49 AM
        "CUDA is on its way out"Bryan Catanzaro02/09/12 12:44 PM
          "CUDA is on its way out"David Kanter02/09/12 05:48 PM
            "CUDA is on its way out"Anon02/09/12 07:11 PM
              "CUDA is on its way out"David Kanter02/10/12 12:00 AM
                "CUDA is on its way out"Michael S02/10/12 04:54 AM
                "CUDA is on its way out"Anon02/10/12 06:42 PM
                  Proprietary SW interfaces and hardware vendorsDavid Kanter02/10/12 09:07 PM
                    Proprietary SW interfaces and hardware vendorsAnon02/11/12 04:08 AM
                      Proprietary SW interfaces and hardware vendorsAnon02/11/12 04:12 AM
                      Proprietary SW interfaces and hardware vendorsDel02/11/12 08:39 AM
                  "CUDA is on its way out"Del02/11/12 08:33 AM
                "NVIDIA is on its way out"Brendan02/10/12 10:42 PM
                  "NVIDIA is on its way out"anonymous02/11/12 11:37 AM
              "CUDA is on its way out"Del02/11/12 08:22 AM
                "CUDA is on its way out"Anon02/11/12 04:11 PM
                  ispc on GPUsMatt Pharr02/12/12 10:02 AM
                    ispc on GPUsAnon02/13/12 03:34 PM
          "CUDA is on its way out"Erwin Coumans02/09/12 09:39 PM
            "CUDA is on its way out"Brett02/10/12 12:54 PM
              "CUDA is on its way out"David Kanter02/10/12 04:00 PM
              "CUDA is on its way out"Anon02/10/12 06:45 PM
                "CUDA is on its way out"Brett02/10/12 08:20 PM
                  "CUDA is on its way out"David Kanter02/10/12 08:58 PM
                    "CUDA is on its way out"Brett02/11/12 07:15 AM
                      Macroscalarnone02/11/12 08:40 AM
        "CUDA is on its way out"Anon02/10/12 06:39 PM
          "CUDA is on its way out"David Kanter02/10/12 09:19 PM
            "CUDA is on its way out"Anon02/11/12 04:33 AM
          "CUDA is on its way out"NN02/14/12 04:24 PM
            "CUDA is on its way out"Anon02/15/12 04:16 PM
        "CUDA is on its way out"bakaneko02/11/12 05:31 AM
      "CUDA is on its way out"Aaron Spink02/10/12 12:29 AM
  Fused Multiply-Add and HSAEduardoS02/08/12 03:11 PM
    Fused Multiply-Add and HSAFoo_02/09/12 06:24 AM
Reply to this Topic
Name:
Email:
Topic:
Body: No Text
How do you spell blue?