Hi,
> Another one that I'd need feedback of is the use of new dlist operations
> for this cached buffer list. I did not use in this patch the existing
> Postgres dlist architecture (ilist.h) because I want to save memory space
> as much as possible especially when NBuffers become large. Both dlist_node
> & dlist_head are 16 bytes. OTOH, two int pointers for this patch is 8 bytes.
In cb_dlist_combine(), the code block below can impact performance
especially for cases when the doubly linked list is long (IOW, many cached
buffers).
/* Point to the tail of main dlist */
while (curr_main->next != CACHEDBLOCK_END_OF_LIST)
curr_main = cb_dlist_next(curr_main);
Attached is an improved version of the previous patch, which adds a pointer
information of the TAIL field in order to speed up the abovementioned operation.
I stored the tail field in the prev pointer of the head entry (maybe not a
typical
approach). A more typical one is by adding a tail field (int tail) to
CachedBufferEnt,
but I didn’t do that because as I mentioned in previous email I want to avoid
using more memory as much as possible.
The patch worked as intended and passed the tests.
Any thoughts?
Regards,
Kirk Jamison
v2-Optimize-dropping-of-relation-buffers-using-dlist.patch
Description: v2-Optimize-dropping-of-relation-buffers-using-dlist.patch
