Synergy User Manual and Tutorial. - THE CORE MEMORY
Synergy User Manual and Tutorial. - THE CORE MEMORY
Synergy User Manual and Tutorial. - THE CORE MEMORY
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
<strong>Synergy</strong> <strong>User</strong> <strong>Manual</strong> <strong>and</strong> <strong>Tutorial</strong><br />
Amdahl’s law for parallelization states that the sequential fraction F s of a task that cannot<br />
be performed in parallel <strong>and</strong> the fraction F p = (1-F s ) that can gives the following formula<br />
for maximum speedup by N p processors:<br />
S<br />
=<br />
F<br />
s<br />
1<br />
1−<br />
F<br />
+<br />
N<br />
p<br />
s<br />
As N approaches infinity, the maximal speedup approaches 1/F s . As the (1-F s )/N p value<br />
becomes very small, the price paid for marginal performance increases. Assume that F s =<br />
0.06. Then F p = 1-F s = 0.94. For 4 processors:<br />
S<br />
1<br />
=<br />
1−<br />
0.06<br />
0.06 +<br />
4<br />
1<br />
=<br />
0.06 +<br />
0.94<br />
4<br />
1<br />
=<br />
=<br />
0.06 + 0.235<br />
1<br />
0.295<br />
= 3.3898<br />
The table below shows the run time, speedup, efficiency <strong>and</strong> cost for processors<br />
N p ={1,2,4,…,1024}, where F s = 0.06 <strong>and</strong> F p = 0.94. Notice that the speedup per<br />
additional processor is much less as N p increases, causing greater cost <strong>and</strong> less efficiency.<br />
The graphs show the effect on speedup (y-axis) with respect to F s (x-axis) with increasing<br />
N p .<br />
Processors(N p) 1 2 4 8 16 32 64 128 256 512 1024<br />
Run Time 1024.00 542.72 302.08 181.76 121.60 91.52 76.48 68.96 65.20 63.32 62.38<br />
Speedup 1.0000 1.8868 3.3898 5.6338 8.4211 11.1888 13.3891 14.8492 15.7055 16.1718 16.4155<br />
Efficiency 100.00% 94.34% 84.75% 70.42% 52.63% 34.97% 20.92% 11.60% 6.13% 3.16% 1.60%<br />
Cost 1.00 1.06 1.18 1.42 1.90 2.86 4.78 8.62 16.30 31.66 62.38<br />
4<br />
16<br />
16<br />
64<br />
70<br />
60<br />
3.8<br />
14<br />
50<br />
1<br />
1<br />
1<br />
F<br />
1 F 3.6<br />
4<br />
F<br />
1 F<br />
16<br />
12<br />
F<br />
1 F<br />
64<br />
40<br />
30<br />
3.4<br />
10<br />
20<br />
3.39 3.2<br />
0 0.01 0.02 0.03 0.04 0.05 0.06<br />
8.421<br />
8<br />
0 0.01 0.02 0.03 0.04 0.05 0.06<br />
13.389<br />
10<br />
0 0.01 0.02 0.03 0.04 0.05 0.06<br />
0 F<br />
0.06<br />
0 F<br />
0.06<br />
0 F<br />
0.06<br />
256<br />
300<br />
1200<br />
1.024 . 10 3<br />
250<br />
1000<br />
200<br />
800<br />
1<br />
1<br />
F<br />
1 F 150<br />
256<br />
F<br />
1 F 600<br />
1024<br />
100<br />
400<br />
50<br />
103<br />
200<br />
15.706<br />
0<br />
0 0.01 0.02 0.03 0.04 0.05 0.06<br />
16.416<br />
0<br />
0 0.01 0.02 0.03 0.04 0.05 0.06<br />
0 F<br />
0.06<br />
0 F<br />
0.06