FPC and FLOAT [[Re: [CLOSED] I can't debug a program that uses PasLibVLC in the IDE.]]

[Fred vS]

I noted that it is extremely difficult to debug a application that deals with audio (or video + audio).
This because debugging "eats" latency with his analysis and so disturbs the flow of samples to sound card.

OTOH I think OP's said (although he's not explicitly confirmed) that he was able to run without breakpoints set, and even though Martin makes the point that the debugger silently intercepts various events I'd not have thought it would make that much difference on a modern CPU.

MarkMLl

I dont know what is the point now with fpc + float calculation?
I do have problems with fpc 3.2.0 with -O2 and -O3 on the ARM Rpi when "Direct DSP" (live effect to samples that is done by a float calculation).

Do you know if there is (or already is) plan to make fpc float calculation faster?
Maybe using fpc LLVM enabled?

Did somebody try the result of float calculation using native-fpc vs LLVM-fpc?

Notre that is only by curiosity, fpc float calculation is not that bad  .

[MarkMLl]

I dont know what is the point now with fpc + float calculation?
I do have problems with fpc 3.2.0 with -O2 and -O3 on the ARM Rpi when "Direct DSP" (live effect to samples that is done by a float calculation).

Fred, are you sure you haven't misthreaded that? OP's determined that the issue is whether the program's run under the debugger, not the compilation options.

MarkMLl

[Fred vS]

I dont know what is the point now with fpc + float calculation?
I do have problems with fpc 3.2.0 with -O2 and -O3 on the ARM Rpi when "Direct DSP" (live effect to samples that is done by a float calculation).

Fred, are you sure you haven't misthreaded that? OP's determined that the issue is whether the program's run under the debugger, not the compilation options.

MarkMLl

OK, OK, sorry, like always I branch off and take the opportunity to ask questions that deviate from the subject.

I apologize.

 

( But about the speed of float calculation, do you have some news?  )

Fre;D

[MarkMLl]

( But about the speed of float calculation, do you have some news?  )

Vectorize! Vectorize!
Let nothing evade your eyes!

(With apologies to Tom Lehrer)

MarkMLl

[Fred vS]

( But about the speed of float calculation, do you have some news?  )

Vectorize! Vectorize!
Let nothing evade your eyes!

(With apologies to Tom Lehrer)

MarkMLl

Hello Mark.

The topic was closed by the author, so maybe he will not be too angry if we discuss off-road.

About vector and Tom Lehrer.

Sorry but I dont have (yet) the skill of Tom.

Anyway, even if I could vectorize some DSP (not easy), it will still need to do float calculation.
Or I miss something.

Fre;D