«

Hi all,

Can anyone provide me inputs and tips on how the merge-sort (i.e.
merging two pre-sorted arrays) can be implemented in VHDL. I am looking
for a parallel implementation (or otherwise also) wherein I want the
merged-sorted array in the next clock cycle itself.

Thanx in advance.

What you want is only the last step of a merge sort.
You still need to need to implement it in VHDL (or
whatever HDL you choose), but here is a pointer
that may help you.

http://www-unix.mcs.anl.gov/dbpp/text/node127.html

--
SystemVerilog DPI tutorial on Project VeriPage:
http://www.project-veripage.com/dpi_tutorial_1.php
For subscribing to Project VeriPage mailing list:
<URL: http://www.project-veripage.com/list/?p=subscribe&id=1>

Hi Swapnajit,

Thank you so much for the inputs.

You got it correctly i.e. I want only the last step of merge-sort. For
sorting I am using parallel VHDL implementation of Insertion Sort for
my application (I know it consumes lot of LEs on my FPGA, but I found
it quite suitable for my application).

In each clock cycle, I am required to sort two sequences (for which
insertion sort suits best). I get the sorted sequences in the next
clock cycle. I would like to merge these two sorted sequences now. And
I do have a latency of one more clock cycle. So I want a similar
implementation of merge-sort also, wherein single sorted sequence
should be obtained in the next clock cyle itself.

However I doubt if the hypercube implementation will help here. Ya, the
compare-exchange method seems to be of some use in this case.

vizziee.