Re: How to implement an non-uniform lookup table in hardware?
- From: glen herrmannsfeldt <gah@xxxxxxxxxxxxxxxx>
- Date: Fri, 11 Dec 2009 18:23:07 +0000 (UTC)
MBALOVER <mbalover9@xxxxxxxxx> wrote:
I am curious about how a non-uniform look-up table is implemented in
hardware?
As far as I know, for an input X to a uniform LUT with 2^M entries, we
just take M most significant bits of X to determine its corresponding
address in the LUT.
For a non-uniform LUT, how do people do to determine the index of an
input?
I thought of another possibility. You can have a series of look-up
tables, each indicating where in the next on to look up the next
set of input bits.
That is, the first table gets, say, 10 most significant bits.
Then latch the output (for pipelineing) and feed it along with,
some of the same bits and, say, the next five bits to another table.
Again latch and index with more bits. Three levels of table should
allow for a large number of input bits. If you use the BRAMs in
most currrent FPGAs it shouldn't be hard at all.
Otherwise, how would you do it in software?
-- glen
.
- Follow-Ups:
- Re: How to implement an non-uniform lookup table in hardware?
- From: Frnak McKenney
- Re: How to implement an non-uniform lookup table in hardware?
- References:
- How to implement an non-uniform lookup table in hardware?
- From: MBALOVER
- How to implement an non-uniform lookup table in hardware?
- Prev by Date: Re: what is happening to my old friend, comp.dsp??
- Next by Date: squared mean vs. mean squared
- Previous by thread: Re: How to implement an non-uniform lookup table in hardware?
- Next by thread: Re: How to implement an non-uniform lookup table in hardware?
- Index(es):
Relevant Pages
|
