[Fwd: Re: Bugzilla issue 230 float_pkg]

Since David's opinion agrees with the current committee consensus, 
unless there are objections,
I will mark this issue closed.

Chuck Swart


-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.

attached mail follows:

Chuck Swart - MTI wrote:
> Hello David:
> 
> Jim Lewis submitted Bugzilla Issue Number 230 which suggests that we 
> change the float_pkg to use
> values which require less hardware to implement.  We are probably not 
> going to follow his suggestion
> for the upcoming version because there is considerable merit in the 
> existing package, and there is not
> yet consensus on the best parameter choices for an "FPGA" package. 
> (There might be an agreed package
> for the next version.) However, we were considering renaming the package 
> from float_pkg to float_754_pkg.
> We ask your opinion on this change and we wonder how many people are 
> using the package under the
> name float_pkg and how strongly that name change would affect them.

This comes from a conversation that Jim and I had a few months ago.  The 
"float_pkg" as it is, is IEEE 754/854 compatible, which is what is 
considered (from a software sense) floating point.   In some opinions 
nothing less will do.

Having said that, I would never actually put that hardware into anything 
other than a CPU.   I would eliminate several functions which (in most 
cases) add very little to the algorithm.   My "fpga" package, which was 
designed around an 18 bit ALU is much more efficient in those parts.

Having said that, at this point I would not change anything.   I could 
give you an "fpga" package, but I am concerned about FPGA vendors saying 
that it favors their parts and not others.

-- 
NAME:     David W. Bishop           INTERNET: dbishop@vhdl.org