Author: Heiner Marxen
Date: 09:38:43 11/02/99
Go up one level in this thread
I agree completely! Once upon a time ... well, 20 years ago I wrote a chess problem program (precursor of CHEST) completely in assembler (IBM 370), and I did optimize it really thoroughly. I performed my own nanoseconds timing of all the important instructions, etc. I used even self modifying code <g>. When I later compared to the speed of compiled Fortran, I found a factor of roughly 2, which I still consider to be comparatively large (i.e. my hand optimizations were good). Algorithmic changes quickly saved much more than just a factor of 2, once I changed to C, while the compiled (and optimized) code were still rather good. For years I had the habit of manually inspecting the assembler code generated by the C compiler, and most of the time I didn´t find anything dramatic. Nowadays I rarely inspect the generated assembler. Of course, it helps to know a bit about the way the compiler produces code. Overall, I will never ever fall back to writing non-trivial code in assembler. It is far too hard to maintain, a factor of 2 is not worth the hassle. Also, portability is a great issue for me: what would I do with highly optimized IBM/370 assembler, today? [I still have those paper listings.] IMO, YMMV. Heiner
This page took 0.01 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.