Fcalva/cyclone
1
0
Fork 0
mirror of https://github.com/justinethier/cyclone.git synced 2025-07-16 17:27:33 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

Added notes

Browse source
This commit is contained in:
Justin Ethier 2016-08-02 21:49:57 -04:00
parent 918e78cb21
commit d9c27b7756
1 changed files with 14 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

14
include/cyclone/types.h
View file

@ -153,6 +153,20 @@ struct gc_thread_data_t {
/** /**
* Group heap pages by type, to attempt to limit fragmentation * Group heap pages by type, to attempt to limit fragmentation
* and improve performance. * and improve performance.
TODO: starting to run into problems when adding additional "sizes" of heap page,
possibly due to increasing amounts of page faults due to non-locality???
Basically for X86_64 everything works great when a 96 byte heap is added, but slows way down when additional
heaps (128, 160) are also added.
32 bit x86 is starting to have trouble with just a 96 byte heap added.
In the future, a better solution might be to allocate arrays (closureN's, vectors, bytevectors, and strings)
as fixed-size chunks to prevent heap fragmentation. The advantage is then we have no fragmentation directly.
But, an array will no longer be contiguous so they may cause other problems, and the runtime has to change
to work with non-contiguous arrays. This would also cause a lot of problems for strings since the built-in
functions would no longer work (EG: strlen, etc).
*/ */
typedef enum { typedef enum {
HEAP_SM = 0 // 32 byte objects (min gc_heap_align) HEAP_SM = 0 // 32 byte objects (min gc_heap_align)