Quasi-random numbers - results for SOBOL & NIEDERREITER methods identical?

[davidspurr]

Using MKL, I get identical QRN's whether I use either the SOBOL or the NIEDERREITER methods. Is this correct, or am I doing something wrong?

Code extract used to generate the QRN's is as below:

USE mkl_vsl

integer status !Error/status
integer brng !Index of generator to initialize stream
integer method !Method = 0 in case of uniform distribution
integer dimen
TYPE (VSL_STREAM_STATE) stream !Random stream
TYPE (VSL_STREAM_STATE) stream2 !Random stream 2

integer, PARAMETER :: nc = 10
integer, PARAMETER :: nr = 1000
real sobel_num(nc,nr)
real niederr_num(nc,nr)

fmt = '(10(:,",",f8.6))'
OPEN(11, file='TestRandomOut.csv')

!SOBEL
write(11,'(///,a,/)') 'SOBOL'
method = 0
sobel_num = 0.0
brng = VSL_BRNG_SOBOL
dimen = 1
status = vslnewstream( stream, brng, dimen ) !Create SOBOL stream
status = vsrnguniform( method, stream, nc*nr, sobel_num, 0.0, 1.0 )
status = vsldeletestream( stream ) !Delete stream
write(11,fmt) sobel_num

!NIEDERREITER
write(11,'(///,a,/)') 'NIEDERREITER'
method = 0
niederr_num = 0.0
brng = VSL_BRNG_NIEDERR
dimen = 1
status = vslnewstream( stream2, brng, dimen ) !Create NIEDERREITER stream
status = vsrnguniform( method, stream2, nc*nr, niederr_num, 0.0, 1.0 )
status = vsldeletestream( stream2 ) !Delete stream
write(11,fmt) niederr_num

[Andrey_N_Intel]

VSL implementations of Sobol and Niederreiter QRNGs are both Gray code based (see Vslnotes.pdf for implementation details). The generators are different in their initialization schemes. However, for particular dimension d=1 the generators are initialized similarly, and you have identical outputs.

[davidspurr]

Thanks for the explanation - yes the results were for d=1.