• Dan <murpsoft hotmail.com> Nov 24 2007
• BCS <ao pathlink.com> Nov 24 2007
• Dan <murpsoft hotmail.com> Nov 25 2007

Dan <murpsoft hotmail.com> writes:

Stewart Gordon Wrote:
 Meanwhile, Phobos has std.bitarray.  I also have a bit array implementation 
 capable of arbitrary slicing:
 http://pr.stewartsplace.org.uk/d/sutil/


Yeah, more or less.

 4-bit type?  I guess some of the bit array code could be modified to give 
 nibble arrays, if that's what you mean....
 
 Stewart.


Yeah okay.  A nibble[] then.  Or I might opt to go a little smaller.

See, while almost no complex operations are done on DNA, pattern-matches are
done often, and trying to do that on the human genome is to pattern-match
against 3 billion base-pairs (6GB if you're tight-ship, 24GB if you use char)

My guess is that the process would be an exercise in DMA and cache prefetch
timing.

So back on subject, how does one implement a nibble[] without a nibble type? 
Would I be using a struct containing byte[] and overloading opIndex and such to
reflect half the "nibble" index into the "byte" index?

: p

BCS <ao pathlink.com> writes:

Reply to Dan,

 Stewart Gordon Wrote:
 
 Meanwhile, Phobos has std.bitarray.  I also have a bit array
 implementation
 capable of arbitrary slicing:
 http://pr.stewartsplace.org.uk/d/sutil/


 
 4-bit type?  I guess some of the bit array code could be modified to
 give nibble arrays, if that's what you mean....
 
 Stewart.
 


 
 See, while almost no complex operations are done on DNA,
 pattern-matches are done often, and trying to do that on the human
 genome is to pattern-match against 3 billion base-pairs (6GB if you're
 tight-ship, 24GB if you use char)


your off by a factor of 8 (or swaped B and b)

3 * 1024^3 * 2 bits =   6Gb = 0.75GB 
3 * 1024^3 * 8 bits = 24Gb = 3.00GB

but 2.25 GB is still a lot of RAM and IO. 

Dan <murpsoft hotmail.com> writes:

BCS Wrote:

 Reply to Dan,
 
 Stewart Gordon Wrote:
 
 Meanwhile, Phobos has std.bitarray.  I also have a bit array
 implementation
 capable of arbitrary slicing:
 http://pr.stewartsplace.org.uk/d/sutil/


 
 4-bit type?  I guess some of the bit array code could be modified to
 give nibble arrays, if that's what you mean....
 
 Stewart.
 


 
 See, while almost no complex operations are done on DNA,
 pattern-matches are done often, and trying to do that on the human
 genome is to pattern-match against 3 billion base-pairs (6GB if you're
 tight-ship, 24GB if you use char)


 your off by a factor of 8 (or swaped B and b)
 
 3 * 1024^3 * 2 bits =   6Gb = 0.75GB 
 3 * 1024^3 * 8 bits = 24Gb = 3.00GB
 
 but 2.25 GB is still a lot of RAM and IO. 
 
 


b

: p