68000 unaligned copy code performance - Page 2

Wepl · 16 February 2024, 08:55

If aligment of source and destination is different then I would recommend to align the destination to long. So if the cpu has a data cache and the source is cachable the source misaligment can be mitigated by the cache.

mc6809e · 22 February 2024, 08:49

Quote:

Originally Posted by Don_Adan

I dont remember exactly timings, but if this code is for 68000, then perhaps using movep.l can be the fastest option.

Movep.l ends up being a bit slower. 48 cycles for each pair of movep.l instructions. Same as four move.b (ao)+,(a1)+ instructions except using movep.l code still needs more time to get shifted bytes back into memory.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
68000 code optimisations	pmc	Coders. Asm / Hardware	248	17 September 2023 13:20
[C, RTG] My raycaster funtime.. sharing code, also looking for performance boost..	mateusz_s	Coders. General	26	23 April 2023 12:40
how to make this code 68000 compliant?	jotd	Coders. Asm / Hardware	2	30 January 2021 18:25
PErformance on 68000 machines	Amiga1992	project.WHDLoad	4	02 July 2014 11:16
68000 boot code	billt	Coders. General	15	05 May 2012 20:13

16 February 2024, 08:55	#21
Wepl Moderator Join Date: Nov 2001 Location: Germany Posts: 869	If aligment of source and destination is different then I would recommend to align the destination to long. So if the cpu has a data cache and the source is cachable the source misaligment can be mitigated by the cache.

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)