1 Reply Latest reply on Oct 13, 2009 4:11 PM by jim.conyngham@amd.com

    ACML-GPU on single-precision only GPUs

    black_jack
      Errors when running sgemm on HD 4570

      Hi,
      my question is very simple: is it possible to use the GPU acceleration for ACML on single-precision only GPUs?
      I ask this because I tried to run the sgemm C example on my notebook, equipped with a Mobile Radeon HD 4570, and I got the following error:

      ACML example: SGEMM call
      --------------------------------------------------------------
      [...]
      calclCompile returned ERROR (Operational error)
      No error
      ERROR: ILScanILBinary: Unsupported opcode for architecture
      ERROR: failed to compile compute kernels for DGEMM
      ERROR: Failed to initialize GPU(s) for SGEMM

      Warning: libCALBLAS is not reentrant!
      Multi-threaded use of this library is not supported.
      When GPU-accellerated ACML routines are called on multiple threads
      concurrently, the requests will be executed serially, even if multiple
      GPUs are present.
      [...]

      I recompiled the source code by myself with Windows SDK v.7, and the result is the same. This is very strange, because I used the Stream SDK (v1.4) with no problems...
      My OS is Vista x64 (SP2), and the ATI drivers are the latest supported by Dell (ver.8.634.3).

      Thanks,
      Giacomo

        • ACML-GPU on single-precision only GPUs

          Hello black_jack.

          Sorry you didn't get a reply to this message sooner.

          This is a bug that was recently discovered in ACMLGPU version 1.0 and will be fixed in version1.1.

          Version 1.0 was intended to allow acceleration of single-precision SGEMM with GPUS that have no double-precision hardware.  Unfortunately, there is a bug in the logic that attempts to compile and download the compute kernels in advance of their being needed.