Re: SSE Level 3 drop in gemm

[Camm Maguire <camm@enhanced.com>]

Greetings!

R Clint Whaley <rwhaley@cs.utk.edu> writes:

> Camm,
> 
> >I guess my problem is that make mmutstcase pre=c nb=?? mmrout=... only
> >compiles the kernel once, so that xsmmtst fails to link, with an
> >undefined reference to the bX routine.
> 
> Ah, I see.  The complex test is cmmutstcase, mmutstcase is the real only.
> The names are not well chosen, to say the least, but I've just not got
> around to cleaning them up.  Page 15 of atlas_contrib.ps shows the procedure
> for complex tests . . .
> 

Thanks Clint!  That's it.  Sorry, I should have read more carefully.
Kernel works real and complex.

Now I have a different issue.  My kernel likes nb=56 the best.  Atlas
standard likes nb=64.  And this is what I get in sMMRES:

intech20:/mnt/i19/f/debian/mm/atlas/tmp/atlas-3.1.2D/tune/blas/gemm/Linux_fpic$ cat res/sMMRES
MULADD  LAT  NB  MU  NU  KU  FFTCH  IFTCH  NFTCH    MFLOP
     0    2  64   5   1  64      0      5      1   371.46
16
ATL_sgemm_SSE.c "CM"
     1    1  64   2   2  64      0      4      1   617.61

At nb=56, the performance is just under 700.

Also, any sugegstions on the unrolling issue?  If I put in macros to
unroll k at differing levels depending on KB, would that confuse the
search engine?  Should install faster than having 3 different k unroll
kernels to time.

Take care,


> Sorry for the confusion,
> Clint
> 
> 

-- 
Camm Maguire			     			camm@enhanced.com
==========================================================================
"The earth is but one country, and mankind its citizens."  --  Baha'u'llah