Hello,
I'm try to run clAmdBlasTune on a FireProv8800 and I get segfaults or weird incomplete output like:
<pre>
C:\Program Files (x86)\AMD\clAmdBlas\bin64>env AMD_CLBLAS_STORAGE_PATH=F:\clAm
lasTune .\clAmdBlasTune.exe --double
GEMV is being tuned, progress: 25.00% SYMV is being tuned, progress: 25.00% GE
is being tuned, progress: 6.25% TRMM is being tuned, progress: 6.25% TRSM i
being tuned, progress: 25.00% SYRK is being tuned, progress: 12.50% SYR2K is b
ng tuned, progress: 12.50%
C:\Program Files (x86)\AMD\clAmdBlas\bin64>
</pre>
After a non-segfaulting run, I get a file named Cypress.kdb in F:\clAmlasTune as set in the env variable.
The problem is that my application runs the same with or w/o it. Same time.
There's also the problem that running the code on the CPU is faster than on the GPU, but I'm hoping that's just the transfer times.
Can anyone shed some light into how to use this tool properly?
