From owner-freebsd-hackers Mon Jul 24 04:47:27 1995 Return-Path: hackers-owner Received: (from majordom@localhost) by freefall.cdrom.com (8.6.11/8.6.6) id EAA16985 for hackers-outgoing; Mon, 24 Jul 1995 04:47:27 -0700 Received: from dns.netvision.net.il (dns.NetVision.net.il [194.90.1.5]) by freefall.cdrom.com (8.6.11/8.6.6) with ESMTP id EAA16979 ; Mon, 24 Jul 1995 04:47:22 -0700 Received: from gena@NetVision.net.il (gena@burka.NetVision.net.il [194.90.6.15]) by dns.netvision.net.il (8.6.12/8.6.9) with SMTP id OAA11397; Mon, 24 Jul 1995 14:48:04 +0300 Date: Mon, 24 Jul 1995 14:48:04 +0300 In-Reply-To: <199507241032.DAA14618@freefall.cdrom.com> Reply-To: gena@NetVision.net.il Message-ID: X-Face: #v>4HN>#D_"[olq9y`HqTYkLVB89Xy|3')Vs9v58JQ*u-xEJVKY`xa.}E?z0RkLI/P&;BJmi0#u=W0).-Y'J4(dw{"54NhSG|YYZG@[)(`e! >jN#L!~qI5fE-JHS+< Organization: NetVision Ltd. X-Mailer: XFMail 0.2-Beta on FreeBSD From: Gennady Sorokopud To: Poul-Henning Kamp Subject: RE: new malloc.c please test Cc: Sender: hackers-owner@freebsd.org Precedence: bulk Hello! I just tried it and after installing new libc my system started to misbehave. Many programs (like vi) complained about memory allocation and my Xserver just crashed. I'm running -current (supped yesterday) and i compiled malloc exactly like you said. ( i defined j in free_pages and also i got 2 warnings during compilation: FreeBSD i386 constants hardcoded. i386 inline assembly used. but i guess it's ok) In message <199507241032.DAA14618@freefall.cdrom.com> Poul-Henning Kamp writes: >OK, here is the next version of my "new" malloc.c This one is running >in my libc right now, and giving no problems so far. > >Please give it a whirl: > cp malloc.c /usr/src/lib/libc/stdlib > cd /usr/src/lib/libc > make all install > >In particular I am very interested in data regarding performance, time and >resource wise compared to the regular malloc in libc, -lgnumalloc and so >on. If some of you could roll it into a Xserver I would be very interested >in the results, and likewise what impact it has on "make world" kind of >things. > >If you compile with -DSANITY it will core-dump the program if it detects >bad pointers or metadata. > >If you have time, you can try to fiddle with the "NFP" definition, and >see what it buys you. > >Please email all results to me. > >Poul-Henning >