Version 1.9 Build 1556

News	FAQ
Search	Home

Next: Comments Up: NOTE 232 - Parallelization Test Results Previous: Tests

Subsections

Results

Clark CLEAN Timings

The following table contains the timing data for the parallelized Clark CLEAN.

Processors Roadrunner Time(s) SGI Origin 200 Time(s)

1 1068 602,600,620

2 729,921 316,326,317

4 1645,1243,1157 208,214,211

Figure 1 plots the number of processors versus the median time to completion for the Clark CLEAN algorithm with 50000 iterations. Figure 2 shows the relative speedup of time versus processors for the median time to completion for the Clark CLEAN algorithm with 50000 iterations.

**Figure 1:** Clark CLEAN time to completion
$\begin{figure} \centering\epsfig{file=clean_time2complete.ps, height=4in, angle=-90}\end{figure}$

**Figure 2:** Clark CLEAN speedup

Parallel IO Tests

This table contains the timing data from the parallelized IO test. Note: timings on the Linux cluster with and without Pablo enabled were essentially the same.

Processors	Roadrunner Time(s)	Normalized SGI Origin 2000 Time(s)
		With Pablo Enabled	Without Pablo Enabled
1	3772,3815,3772	6932,5704,6127	5610,5613,5745
2	1922,1931,1914	2998,2914,2854	3080,2980,2830
4	1006,1013,1018	1547,1560,1500	1508,1505,1523
8	522,549,534	834,878,838	760,770,734
16	318,320,316	456,456,481	392,383,378
32	275,202,257	244,251,248	317,208,206
64	308,93,278	277,276,279	159,159,161
128			283,283,300

Figure 3 plots the number of processors versus the median time to completion for the parallel IO test. Figure 4 shows the relative speedup of time versus processors for the median time to completion for the parallel IO test.

Note: because of the heterogenous types of processors available at the NCSA the timings are adjusted to a normalized 175MHz R10000 chip. Absolute timings on the 225MHz chips were about 30% faster.

**Figure 3:** Parallel IO test time to completion
$\begin{figure} \centering\epsfig{file=pio_time2complete.ps, height=4in, angle=-90}\end{figure}$

**Figure 4:** Parallel IO test speedup

Next: Comments Up: NOTE 232 - Parallelization Test Results Previous: Tests Contents

Processors	Roadrunner Time(s)	SGI Origin 200 Time(s)
1	1068	602,600,620
2	729,921	316,326,317
4	1645,1243,1157	208,214,211