Comments on: one malloc to rule them all http://blog.reverberate.org/2009/02/20/one-malloc-to-rule-them-all/ parsing, performance, minimalism with C99 Mon, 06 Feb 2012 23:44:51 +0000 hourly 1 http://wordpress.org/?v=3.3.1 By: Benjamin A. Shelton | Blog » Blog Archive » Links: May 13th http://blog.reverberate.org/2009/02/20/one-malloc-to-rule-them-all/comment-page-1/#comment-1320 Benjamin A. Shelton | Blog » Blog Archive » Links: May 13th Thu, 14 May 2009 00:19:41 +0000 http://blog.reverberate.org/?p=124#comment-1320 [...] a fascinating discussion regarding different types of malloc()s—and how they all effectively promise the same [...] [...] a fascinating discussion regarding different types of malloc()s—and how they all effectively promise the same [...]

]]> By: Josef "Jeff" Sipek http://blog.reverberate.org/2009/02/20/one-malloc-to-rule-them-all/comment-page-1/#comment-1319 Josef "Jeff" Sipek Mon, 11 May 2009 19:20:18 +0000 http://blog.reverberate.org/?p=124#comment-1319 You are assuming that the faults are not in Firefox. Firefox has lots of problems, including managing memory, unnecessary IO, etc., etc. You are assuming that the faults are not in Firefox. Firefox has lots of problems, including managing memory, unnecessary IO, etc., etc.

]]> By: josh http://blog.reverberate.org/2009/02/20/one-malloc-to-rule-them-all/comment-page-1/#comment-1318 josh Mon, 11 May 2009 18:38:57 +0000 http://blog.reverberate.org/?p=124#comment-1318 Good point Matt, though you make me realize I also forgot up-front overhead (6MB for tcmalloc) and how fast memory is returned to the OS (tcmalloc is not very good about this). Good point Matt, though you make me realize I also forgot up-front overhead (6MB for tcmalloc) and how fast memory is returned to the OS (tcmalloc is not very good about this).

]]> By: Matt H http://blog.reverberate.org/2009/02/20/one-malloc-to-rule-them-all/comment-page-1/#comment-1317 Matt H Mon, 11 May 2009 17:03:32 +0000 http://blog.reverberate.org/?p=124#comment-1317 You forgot about overhead per allocation. To quote google's page: Another benefit of TCMalloc is space-efficient representation of small objects. For example, N 8-byte objects can be allocated while using space approximately 8N * 1.01 bytes. I.e., a one-percent space overhead. ptmalloc2 uses a four-byte header for each object and (I think) rounds up the size to a multiple of 8 bytes and ends up using 16N bytes. I didn't find any info on TSLF's page on overhead, but I think its important, especially for dynamic languages where the VM makes lots of small allocations You forgot about overhead per allocation. To quote google’s page:

Another benefit of TCMalloc is space-efficient representation of small objects. For example, N 8-byte objects can be allocated while using space approximately 8N * 1.01 bytes. I.e., a one-percent space overhead. ptmalloc2 uses a four-byte header for each object and (I think) rounds up the size to a multiple of 8 bytes and ends up using 16N bytes.

I didn’t find any info on TSLF’s page on overhead, but I think its important, especially for dynamic languages where the VM makes lots of small allocations

]]>