F distribution on the 41C

[quantalume]

(12-19-2015 09:18 AM)Dieter Wrote:  I often choose two-byte LBLs and the corresponding 3-byte GTOs because they are able to record their target address for longer jumps. You know that once a GTO has found the LBL it jumps to, this position is recorded within the GTO command ("compiled GTO") so that subsequent branches are executed much faster. Here the 3-byte-versions allow for longer jumps than their 2-byte counterparts, so just to be sure I prefer the former.

In this program two-byte GTOs (resp. single-byte LBLs) may be sufficient, at least in most cases. What makes me wonder is your remark that this speeded up the program slightly. I would have expected no change at best.

You obviously know a lot more about 41C programming than I do. I only measured the execution speed difference one time, and it was something like 0.25 seconds out of 17. It may well have come down to human error. Now that I have F distribution on my calc, perhaps I should repeat the experiment multiple times and determine the significance level of my measurement. ;-)

My only wish now would be trying to squeeze the program onto a single card (224 bytes). It currently stands at 252 bytes with the 1-byte LBLs, and the only other economy that immediately stands out is replacing the alpha LBLs, which wouldn't be enough. However, your beta function is valuable enough on its own, so perhaps it should become a permanent part of my calculator's memory.

David