Page 2 of 2
Re: Using the Analyser prim
Posted: Thu Jun 18, 2020 8:26 pm
by wlangfor@uoguelph.ca
You know, i was thinking the same thing the other day, that there's no need for an analyzer at to gauge things like db and such; is it not just a waste of time until you need to visually display it? Word to the wise making LU calc's.
Funny how most of us do it backwards.
Re: Using the Analyser prim
Posted: Fri Jun 19, 2020 9:20 am
by HughBanton
@trogluddite wrote:
The differences are only in what assumptions FS makes about their content, and what access it allows us...
A goldmine of info Trog, as ever. Thanks. The manual is very hazy about the difference between a float array and a mem, and you've made that much clearer.
Yes, I've seen examples of hanging things on to a mem and using its addresses directly. Never myself dared to attempt anything like that so far - I'd be sure to break something. Also .. memin has a dubious reputation in 3.0.6 as being crash prone, so I'd probably break at least TWO things ...
But I get the idea; I might have a go in FS4! MyCo seems to have got mem really stable now; I moved a few things to memrefin 6 months ago .. some worked better than others .. but I've now switched all memory reads to that method without further problems. Recommended.
Being able to read my array[32] without the 'slow' stage(2) counter, and without the Analyser, would be very satisfying. (I've got some much much bigger ones than this!)
Thanks
H
Re: Using the Analyser prim
Posted: Fri Jun 19, 2020 2:19 pm
by trogluddite
HughBanton wrote:memin has a dubious reputation in 3.0.6 as being crash prone
Unfortunately, when using assembly trickery, there's always the chance of a crash from the slightest error in code which uses pointers. Pointing to memory which hasn't been reserved by a proper request to the OS is an "access violation" - the sentence for which is instant death to the application with no right of appeal!
Re: Using the Analyser prim
Posted: Mon Mar 15, 2021 1:34 pm
by HughBanton
9 months later I finally got back to this. (DSP - Dreaded Serial Procrastination, hoho)
Yes Trog was correct, using a memrefin input you can indeed get the entire array result in the first cycle, making stage(2) redundant here and making these kind of calculations possible in a single pass.
However (even though you just can't contain your excitement!), obviously this is only suitable for the Slack forum so I'll post a demonstration over there.
H