I have tried CTM on Firestream, it was not a good tool for most programmer. CAL may be different, maybe. AMD GPU's SP is 5D SIMD based, which means if you need all ALU working, you should program your codes based on SIMD and then as what CUDA did. AMD's 4870 was supposed to provide 1.2TFlops, but how to achieve it is a big problem. Li, Bo ----- Original Message ----- From: "Kozin, I (Igor)" <[EMAIL PROTECTED]> To: "Li, Bo" <[EMAIL PROTECTED]> Cc: "Beowulf" <beowulf@beowulf.org> Sent: Thursday, August 28, 2008 12:34 AM Subject: RE: [Beowulf] gpgpu
> Have you tried ACML on Firestream? > >> -----Original Message----- >> From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] > On >> Behalf Of Li, Bo >> Sent: 27 August 2008 01:50 >> To: Mikhail Kuzminsky; Vincent Diepeveen >> Cc: Beowulf >> Subject: Re: [Beowulf] gpgpu >> >> Hello, >> IMHO, it is better to call the BLAS or similiar libarary rather than >> programing you own functions. And CUDA treats the GPU as a cluster, so > .CU >> is not working as our normal codes. If you have got to many matrix or >> vector computation, it is better to use Brook+/CAL, which can show > great >> power of AMD gpu. >> Regards, >> Li, Bo > _______________________________________________ Beowulf mailing list, Beowulf@beowulf.org To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
