id	summary	reporter	owner	description	type	status	priority	milestone	component	version	severity	resolution	keywords	cc	mlocked
1433	pespacket.cpp Allocator optimization	Janne <janne-mythtv@…>	danielk	"With [9223] 2/3 of the time of !ParseTable was spent in get_{188|4096}_block, since ~80% of the calls actually do memory allocation.

The attached patch halves the size of the blocks, unrolls a expensive vector method from the inner loop and frees the allocators only if more than one block was allocated.

The last point is probably the simplest compromise between saving memory allocations and smallest memory footprint.

With the patch get_{188|4096}_block drops to 3%, and the ratio of calls with memory allocation drops  under 1%."	enhancement	closed	minor	unknown	dvb		medium	fixed			0