Improvements with the new fermi hardware

How can I improve the CEV predictor with the new NVIDIA 480 ?

One possibility is to increase the size of the table I use in the compare session where now in the 8800 GT I use a table of 24 bit index of boolean with a matching probability of about 1 – ((1 – (80/3)/(2^(8*3)))^((2^16)/8)) (see String Matching ) .

With the new card I can use a 2^32 booleans table using 2^(32-3) =536870912 bytes of memory ( 3 bits to manage 1 byte )  available on this card so the matching probability become 1 – ((1 – (80/4)/(2^(8*4)))^((2^16)/8)) and to obtain the same matching probability I can increase the size of the block from 2^16 to 2^25 so I can exam 2^(25-16) =512 times data for each cycle .

This does not means that I can improve by a factor of 512 but that I can reduce slow operations by a factor of 512 .

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s