Heikki Linnakangas <heikki.linnakangas@enterprisedb.com> writes:
> Alvaro Herrera wrote:
>> I thought about that too. I admit I am not sure if this really works
>> portably; however I don't want to add a palloc() to that routine.
> It should work, AFAIK, but unaligned memcpy()s and write()s can be a
> significantly slower. There can be only one write() happening at any
> time, so you could just palloc() a single 8k buffer in TopMemoryContext
> in backend startup, and always use that.
Some time ago, we arranged for shared buffers to be aligned on *more*
than maxalign boundaries (cf BUFFERALIGN) because on at least some
platforms this makes a very significant difference in the speed of
copying to/from kernel space. If you are going to double-buffer it
is going to be important to have the intermediate buffer just as well
aligned. A local char array won't be acceptable, and even for a
palloc'd one you'll need to take some extra steps (like wasting
ALIGNOF_BUFFER extra bytes so you can align the pointer palloc gives).
regards, tom lane