IA64 timings

[R Clint Whaley <rwhaley@cs.utk.edu>]

Dan,

Are you still under NDA on the IA64?  If not, do you have MKL numbers for
this machine?  

You asked a while back whether anyone has gotten any decent results on this
machine.  Well, I finally have.  I'm getting roughly 70% of theoretical
peak now.  On the 666Mhz Itanium that Compaq has at their testdrive, the best
time I've seen on a full DGEMM is around 1845Mflop.  I include more complete
timings below.  The performance could probably be tweaked further, but
this will be good enough until after the release.

This code should be coming out in the final developer release within the
next couple of days, and as a stable release by Xmas.

Anyway, if anyone has seen MKL or other tuned IA64 numbers, I'd appreciate
the comparison point.

I'm too tired to do the other precisions, but initial timings seem to indicate
we will break 2Gflop for single . . .

Cheers,
Clint

666Mhz IA64 Itanium, double precision

             100    200    300    400    500    600    700    800    900   1000
           =====  =====  =====  =====  =====  =====  =====  =====  =====  =====
dMM REF     52.5   51.8   52.3   51.1   50.0   43.5   38.1   35.2   32.9   31.7
dMM ATL    866.3 1247.9 1472.4 1566.6 1570.6 1708.0 1645.1 1730.3 1710.2 1741.5
             
             144    288    432    576    720    864   1008   1152   1296   1440
           =====  =====  =====  =====  =====  =====  =====  =====  =====  =====
dMM       1278.5 1505.3 1618.8 1709.1 1742.6 1742.6 1667.1 1761.0 1769.1 1833.1
                          
                          
                          GEMM   SYMM   SYRK  SYR2K   TRMM   TRSM
                         =====  =====  =====  =====  =====  =====
d REF     500             48.5   73.0   51.9   61.1   52.4   48.0
d ATLAS   100            802.5  774.6  425.2  620.4  554.1  358.8
d ATLAS   500           1610.1 1201.9 1462.9 1462.9 1082.5  816.6
d ATLAS  1000           1746.0 1259.5 1357.6 1605.0 1297.8  977.1