Return-Path: Delivered-To: apmail-apr-dev-archive@apr.apache.org Received: (qmail 13404 invoked by uid 500); 12 Jul 2001 18:39:47 -0000 Mailing-List: contact dev-help@apr.apache.org; run by ezmlm Precedence: bulk List-Post: List-Help: List-Unsubscribe: List-Subscribe: Delivered-To: mailing list dev@apr.apache.org Received: (qmail 9820 invoked from network); 12 Jul 2001 18:33:29 -0000 Message-ID: <3B4DED5F.8040101@xbc.nu> Date: Thu, 12 Jul 2001 20:33:03 +0200 From: Branko =?ISO-8859-2?Q?=C8ibej?= User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:0.9.2) Gecko/20010628 X-Accept-Language: sl, en-gb, en MIME-Version: 1.0 To: dev@apr.apache.org Subject: Re: SMS usage patterns, hierarchies References: <5.1.0.14.2.20010711152506.00aa6020@mail.charter.net> Content-Type: text/plain; charset=ISO-8859-2; format=flowed Content-Transfer-Encoding: 8bit X-Spam-Rating: h31.sny.collab.net 1.6.2 0/1000/N Greg Marr wrote: > At 01:01 PM 07/11/2001, dean gaudet wrote: > >> i know nothing about the quality of windows malloc on multi-cpu boxes. > > > As of VC++ 6.0 SP4: > There is a small block freelist for blocks 0x3F8 or smaller. It uses > a critical section around the allocator. Anything larger falls > through to the HeapAlloc Win32 API function. Except that on Win2k, the CRT library configures itself (dynamically, at program startup) to pass every single malloc off to HeapAlloc, thereby slowing small block allocation tremendously. In the project I work on, our app got slowdowns of up to 10x on certain test cases when we switched to SP4. We switched back to SP3 at once, of course. Grrrrr. -- Brane �ibej home: http://www.xbc.nu/brane/ work: http://www.hermes-softlab.com/ ACM : http://www.acm.org/