On Thu, Sep 6, 2012 at 10:50 PM, Rugxulo <rugx...@gmail.com> wrote:
> Anyways, here's a concrete example:   building p7zip 9.20.1 (7za.exe
> only) via G++ 4.7.1 (DJGPP 2.04 host and target) on this particular
> machine:
> 1). native FreeDOS, no LFNs, XMGR + UIDE, built atop RDISK, HDPMI32 -r
> 2). native FreeDOS, same as above but with DOSLFN loaded
> 3). DOSEMU + FreeDOS (with its own LFNs, XMS, DPMI, etc.) atop Lucid
> PuppyLinux 5.18
> 1). (SFN) Elapsed time: 354.050 seconds (0:05:54.050)
> 2). (LFN) Elapsed time: 430.840 seconds (0:07:10.840)
> 3). (EMU) Elapsed time: 303.190 seconds (0:05:03.190)
> So, for the record, fastest to slowest is DOSEMU, native FreeDOS, and
> LFN-enabled FreeDOS. But we're only talking 5 vs 6 vs 7 mins. here,
> indeed quite noticeable here but nothing too horribly different.

This isn't the whole story. Just so no one gets confused or gets the
wrong idea, here's the deal:

a). Without cache loaded at all, in native FreeDOS, with compiler
installed to HD but compiling atop RAM disk, this build takes 25
b). With cache loaded it's about 5 or 6 minutes (again, under native FreeDOS).
c). Optimal is apparently using both software cache (e.g. UIDE or
LBACACHE) and RAM disk (e.g. SHSURDRV or RDISK). When I put the
compiler proper (build tools, etc.) onto RAM disk and have suitable
cache enabled, the build time (also atop RAM disk) drops to 2 minutes!

It's apparently a mistake to think that the software cache alone would
suffice to be "close enough" to optimal ("too many XMS moves"??).

(Just FYI in case anyone wants to do any heavy compiling ....)

Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and 
threat landscape has changed and how IT managers can respond. Discussions 
will include endpoint security, mobile security and the latest in malware 
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
Freedos-user mailing list

Reply via email to