[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Athlon 1.2Ghz results == P4 1.5



Greetings!  This is fantastic!

Just wondering if in principle the p4 can be made this good.  My
understanding is definitely no with the normal x87 cpu, as the Athlon
can do 2x here but the p4 1x.  Theoretical peaks for p4 SSE2 and
athlon x87 cpu should be the same, right?  Those SSE2 instructions are
big, and no consideration of instruction decoding has at least entered
my mind in writing anything, but I would still be very surprised if
such considerations could produce an extra 30%.  My suspicion is that
the P4 SSE2 attainable peak is simply less than that of the Athlon x87
cpu primarily due to the instruction size and the load this puts on
the decoder, but that's just a guess.  Anyone have some real
information here?

Take care,

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

> Guys,
> 
> I include below the new stuff on a 1.2Ghz Athlon with DDR memory.  This is
> Julian's kernel, with Peters mods, plus the new cacheedge, and we are talking
> over 80% of peak in out-of-cache performance on a 1.2Ghz machine.  If
> you compare these numbers in double precision (real and complex), you will
> find that the Athlon wins big for small-medium size probs, and ties for large
> with a 1.5Ghz P4.  It gets very close to the performance of a 800Mhz IA64
> as well.  I am impressed.  Of course, the P4 cleans up the place in single,
> since it is using SSE for double the theoretical peak . . .
> 
> You'll notice that the double complex performance is much better than the
> last time I posted numbers; it was a problem with cacheedge setting screwing
> it up before . . .
> 
> Cheers,
> Clint
> 
> Results on 1.2Ghz Athlon
> 3.3.2  : old ATLAS version, using generated kernel
> 3.3.11 : ATLAS + Julian/Peter + CE
> 
>               100    200    300    400    500    600    700    800    900   1000
>            ====== ====== ====== ====== ====== ====== ====== ====== ====== ======
> 3.3.11 sMM 1612.9 1710.3 1800.0 1828.6 1851.9 1963.6 1905.6 1896.3 1970.3 1941.7
> 3.3.11 sLU  781.6 1147.2 1323.0 1381.2 1456.1 1554.8 1593.6 1586.1 1618.7 1624.8
> 
> 3.3.11 cMM 1771.4 1792.0 1878.3 1896.3 1886.8 1898.9 1905.6 1887.6 1924.8 1900.2
> 3.3.11 cLU 1068.3 1381.2 1438.2 1504.5 1549.2 1599.0 1603.8 1624.6 1660.8 1687.1
> 
> 3.3.2  dMM 1136.4 1271.8 1388.6 1280.0 1315.8 1393.5 1372.0 1383.8 1429.4 1418.4
> 3.3.11 dMM 1470.6 1550.0 1800.0 1828.6 1785.7 1878.3 1960.0 1861.8 1944.0 1904.8
> 3.3.2  dLU  676.0  841.3  914.1  982.7  970.8 1027.3 1054.3 1065.7 1116.3 1110.3
> 3.3.11 dLU  714.6  952.4 1092.9 1188.5 1323.8 1307.5 1370.5 1391.9 1566.4 1448.2
> 
> 3.3.11 zMM 1503.0 1600.0 1728.0 1765.5 1554.4 1838.3 1841.6 1820.4 1869.2 1834.5
> 3.3.11 zLU  947.4 1087.3 1227.7 1311.6 1359.5 1439.1 1428.4 1483.3 1530.0 1540.8
> 
> 

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