Hi Dieter, I just finished the first prototype for the two-threads-only file server dpi. I've had this in mind for sometime because the old one, despite solving a deadlock race, used too much threads when the page requested too much different files (one per file). This can cause some problems on restricted environments as PDAs or uclibc for instance. I hope this solves the issues you told me on your LP64 machine. Ah, the CVS also has a lot more fixes of -Wconversion warnings. Urbi et orbi: please test these new dpis because the current CVS is very close to becoming dillo 0.8.4! How: dpidc stop make make install-strip (this avoids the upgrading problems with running dpis/dpid). ============ Have great hollydays and enjoy as much as you can! -- Cheers Jorge.- PS: Dieter: are you subscribed to dillo-dev?
On Tue, Dec 28, 2004 at 06:59:07PM -0300, Jorge Arellano Cid wrote:
Hi Dieter,
I just finished the first prototype for the two-threads-only file server dpi. I've had this in mind for sometime because the old one, despite solving a deadlock race, used too much threads when the page requested too much different files (one per file).
i haven't looked at it yet, and i appologize for commenting before doing do, but is this something that could be made configurable? i ask only since some machines (a lot that i own anyway *G*) do better IO than CPU stuff, so doing lots of simultaneous IO would make dillo faster there. -brian -- "Now you know why I got the everliving hell OUT of Windows administration. Knowing it doesn't make it any easier. It's just broken-as-designed."
On Thu, Dec 30, 2004 at 10:49:14PM -0500, Brian Hechinger wrote:
On Tue, Dec 28, 2004 at 06:59:07PM -0300, Jorge Arellano Cid wrote:
Hi Dieter,
I just finished the first prototype for the two-threads-only file server dpi. I've had this in mind for sometime because the old one, despite solving a deadlock race, used too much threads when the page requested too much different files (one per file).
i haven't looked at it yet, and i appologize for commenting before doing do, but is this something that could be made configurable?
i ask only since some machines (a lot that i own anyway *G*) do better IO than CPU stuff, so doing lots of simultaneous IO would make dillo faster there.
On single-CPU sytems this server will be faster in most cases. On multi-CPU systems I doubt a difference because the threaded tasks are almost pure IO and not CPU-intensive. OTOH, computer architecture magic can surprise anyone! ;-) -- Cheers Jorge.-
participants (2)
-
Brian Hechinger -
Jorge Arellano Cid