Full text: XVth ISPRS Congress (Part A2)

92 
  
  
493 
STEPS 
This step is performed by FEP. The subscene images, which are processed 
in STEP2, are combined into the scene. 
The software is written in FORTRAN77 which allows automatical 
vectorization by the FORTRAN77/VP. Though the high speed processing is 
expected by this vector compiler, the algorithmic consideration is important 
(3) 
to extract higher performance of VP-200. 
The largest portion of SAR data processing time is consumed by FFT, which 
is often used to evaluate the performance of supercomputers. There are some 
algorithms for implementing FFT and Fig.4 shows two of 5, Though the in- 
place FFT algorithm does not require temporary storages, its transfer pattern is 
considerably complicated. Since the vector length decreases by each stage, it 
is difficult to get a high enough performance by vector processing. This 
algorithm is appropriate for scalar machines with small main storage. On the 
other hand, though the isogeometric FFT algorithm requires temporary storages, 
the vector length is retained constant because its transfer pattern is fixed 
over all stages. So, it is easy to extract a very high performance of a vector 
computer. The isogeometric FFT is employed in the SAR data processing. It 
takes 213 microseconds to perform binary-radix, 1024 point complex FFT using 
FACOM VP-200. 
 
	        
Waiting...

Note to user

Dear user,

In response to current developments in the web technology used by the Goobi viewer, the software no longer supports your browser.

Please use one of the following browsers to display this page correctly.

Thank you.