3 ms·

The blog post sadly doesn't cover what alternatives are out there, and why they settled on their choice. Besides tcmalloc, there are jmalloc, hoard, the more r

• curlypaul924 6y ago
  Do you think any of these stand a chance at replacing Doug Lea's malloc in glibc, or does dlmalloc provide more desirable behavior for the general case?
  • alexhutcheson 6y ago
    The FSF requires copyright assignment for code added to glibc, which creates an impediment to including code from other projects.
• rurban 6y ago
  The best allocator for this kind of problem would of course be a compacting GC. First it would be much faster, it would be safer, and easier to use. For my bigger libraries glibc malloc and free has insane runtimes. Eg a final free could last over 1 minute. If you see this with a GC you would throw it out immediately. But apparently people don't measure and spread false rumors about GC's. Even a very poor and primitive GC as in emacs is not that slow as glibc malloc/free. A good one is miles faster. My GC has an upper limit of 10ms, not minutes.