forums.silverfrost.com

StamK · Joined: 12 Oct 2016 Posts: 159

Since the early days of 64bit up to now, we have noticed that debugging is much slower in 64bit than in 32bit. This is not only when starting up but also when stepping inside the program. Have other users noticed that? For developers who use the debugger all the time, this is a big issue and it is forcing the actual development to be done in 32bit and 64bit to be used only to test compatibility.
It would be good for this issue to be investigated and some kind of solution to be found.

PaulLaidler · Posted: Fri Dec 22, 2023 11:35 am Post subject:

StamK

Are you running SDBG64 directly or are you using the integrated access from Plato?

StamK · Joined: 12 Oct 2016 Posts: 159

SDBG64 directly

DanRRight · Posted: Mon Dec 25, 2023 1:36 am Post subject:

Interesting. I had more complaints/improvement suggestions about SDBG than all people here combined but never complained about slowness to the extent that i thought to go back to 32bit. What version and what processor+memory do you have?

Have you used new feature Trace allowing to trace steps back? This could be slowing debugging according to Robert but i have not noticed that too.

Last week i had one case in last years of slowness when SDBG was loading slow but my computer was overloaded of its all cores plus all threads with other running codes by 100%.

Also typically i do not run the code under SDBG64 for a long time to catch the error, this could be why i have not noticed potential slowness if it really slows things down to the extent that it becomes unbearable. I do not use 32bit, because the code just can not scale down to 32bits anymore

As a potential reason for slowness could also be special cases where FTN95 really lags in performance (see Polyhedron tests)

PaulLaidler · Posted: Tue Dec 26, 2023 9:41 am Post subject:

Version 9.0 of FTN95 includes significant improvements in runtime performance. The current published results for the Polyhedron benchmarks are therefore out of date.