Numerical Results in .NET

Render qr bidimensional barcode in .NET Numerical Results
6.4 Numerical Results
.net Vs 2010 qr code jis x 0510 integration with .net
use .net framework qr printer todraw qr code on .net
'u S
Qr Codes recognizer for .net
Using Barcode scanner for .net vs 2010 Control to read, scan read, scan image in .net vs 2010 applications.
<l.l
recognizing barcode for .net
Using Barcode recognizer for VS .NET Control to read, scan read, scan image in VS .NET applications.
'" E <l.l
Make barcode for .net
using .net vs 2010 crystal toconnect bar code with asp.net web,windows application
30 25
Control qr code 2d barcode data on .net c#
qr bidimensional barcode data on .net c#
parallel
Control qr codes image for .net
use asp.net aspx qr creation toadd qr codes for .net
.. single
------
.NET Crystal code 128b implementation on .net
use .net vs 2010 crystal code-128c implementation toconnect barcode standards 128 with .net
~ 20
QR Code JIS X 0510 printer on .net
use .net framework crystal qr implement toembed qr-codes on .net
'J::
Draw linear barcode with .net
use .net vs 2010 crystal 1d writer toadd 1d barcode on .net
15 10 5
.NET Crystal ansi/aim code 39 maker in .net
generate, create 39 barcode none on .net projects
'" :E
.net Vs 2010 ean-8 creation with .net
use .net framework upc - 8 generator togenerate ean-8 supplement 5 add-on for .net
scattering angle (degree)
Control upc - 13 size with visual c#.net
ean13 size in visual c#.net
Figure 6.6.2 Comparison of the bistatic scattering coefficients for PEC rough surface obtained by the single code and parallel code. 8 x 8,\;;, h = 0.1,\0 incident angle = 40
QR Code generation in vb.net
use .net framework qr code encoder toattach qr-code for vb.net
~ <l.l o ()
parallel
ECC200 barcode library on vb
using barcode creation for winforms crystal control to generate, create data matrix barcode image in winforms crystal applications.
~~ngle
Java bar code integration with java
generate, create bar code none in java projects
'50.8 ~ ~ 0.6
Control bar code 39 image for java
generate, create code 3/9 none with java projects
-30 o 30 scattering angle (degree)
ANSI/AIM Code 128 writer in java
generate, create code 128c none on java projects
Figure 6.6.3 Comparison of the bistatic scattering coefficients for lossy dielectric rough surface obtained by the single code and parallel code. 8 x 8,\;;, h = 0.2'\0 incident angle = 30 , Er = 45 + i30.
Visual .net qr code jis x 0510 printing with visual c#
use .net qr-codes integrated toassign qr codes on c#.net
63-D WAVE SCATTERING FROM 2-D ROUGH SURFACES
CPU time (Sec.) Surface size (A~ ) 8x8 16 x 16 32x 32 64x64 32 x 32 64x64 Number of unknowns 8,192 32,768 131,072 524,288 (one CG term) (one CG term) Sequential code 518 3020
------
Speedup factor 7.85 13.73
-----------
Parallel code 66 220 1,120 6,045 9.2 37.9
------
382.48 4381.74
41.57 115.61
Table 6.6.1 Comparison of CPU time for PEC rough surface h
O.IA o .
Surface size (~ )
Number of unknowns
rms height ("-0) 0.05 0.10 0.15 0.20 0.25 0.30 0.05 0.10 0.05 0.10 Sequential code 49,368 77,643 114,385 161,313 207,095 275,485 270,204 469,831
-------
CPU time (Sec.) Speedup factor 9.77 11.57 11.55 11.79 11.92 12.23 13.20 13.43 -------------
Parallel code 5,051 6,712 9,906 13,682 17,371 22,517 20,477 34,981 133,798 249,460
98,340
16 x 16 32 x 32
393,216 1,572,864
-------
Table 6.6.2 Comparison of CPU time for lossy dielectric rough surface
45 + i30.
when the number of unknowns increases. For the surfaces with the same size, the CPU time increases with h because of the slow convergence of the CGM for rougher surfaces. For example, it only needs 2 iterations (n = 1) with 40 CG terms in the first iteration and 1 CG term in the second iteration for the 8 x 8>'~ surface with h = 0.05>'0' In contrast, it needs 3 iterations (n = 2) with 100 CG terms in the first iteration, 41 CG terms in the second iteration and 4 CG terms in the last iteration when h increases to 0.3>'0' The speedup factors become large when the CPU time increases. It is shown that the best speedup factor is around 13.7 for the PEC surface with 32,768 surface unknowns and h = 0.1>.0' and 13.4 for the lossy dielectric surface with 0.4 million surface unknowns for the same h. For the PEC surfaces, CPU times of the parallel code are obtained when the strong matrix is computed once
6.4 Numerical Results
Total time Communication time Near interaction calculation time Far interaction calculation time Emissivity calculation time
4584.8 s 120.8 s 2918.6 s 289.4 s 1225.7 s
Table 6.6.3 CPU time distribution for the case of 8 x 8'\~ surface with h
0.05'\0'
and stored. When the number of unknowns exceeds 131,000, we recalculate the strong matrix in the sequential code when necessary. In the last two rows of Table 6.6.1, we give the comparison of CPU time for one CG term when the surface sizes are 32 x 32 and 64 x 64>'~, respectively. The strong matrix is recalculated in the sequential code. One can see that the speedup factor can be up to 115.6 for 1 CG term. To complete the conjugate gradient solution of sequential code will be prohibitive long. For dielectric surfaces, when the number of unknowns exceeds 1.5 millions, the code on a single processor cannot be run and therefore parallel computation is the only alternative. In the 1.5 million unknown example, the computer code requires about 78% of the 4.0 GBytes of RAM. We should emphasize that the CPU time comparisons shown in Table 6.6.1 and 6.6.2 are based on the collocation method. For the lossy dielectric surface, it is found that near-field integration is necessary. Although this near-field integration requires additional CPU time, the resulting matrix is better conditioned so that the number of CG terms may be reduced. For examples, with near-field integration, the CPU time for the 8 x 8 and 16 x 16>'~ surface with h = 0.05>'0 reduces from 5,051 to 4,585 and 20,474 to 19,085 seconds, respectively. It is expected that similar speed-up factors are obtained when replacing the collocation method with the nearfield integration. Table 6.6.3 shows the CPU time distribution for a 8 x 8>'~ surface with near-field integration, where h = 0.05>'0' The communication time is only a small part of the total CPU time. Most of the CPU time is spent on calculating the sparse matrix and sparse-matrix-vector multiplication. Communication time only attributes to 2.6% of the total CPU time while that of the far interaction calculation based on FFT is 6.3%. Table 6.6.4 shows the comparison of the computed brightness temperature versus surface size and rms height, and the comparison between the results obtained with and without the near-field integration. The incident angle for the 8 x 8>'~ surfaces is 30 while that of the remaining ones is 50 . The result of an infinite flat surface with the same permittivity is also given for reference. It should be pointed out that the brightness temperatures of the rough surfaces are obtained from one realization. Different realizations