Compiler related FM3 reduction in performance?

O2b60Again

O2b60Again

Member
Hi all ...my first post
re: Cliff's statement
The firmware is approaching release quality. We finally got to the bottom of some data alignment issues. Seems there's a bug in the compiler but we found a workaround. The wait will be worth it.

It's a C compiler issue. We use SIMD processing which requires that vectors are aligned at even-word addresses. The compiler is failing to honor the alignment #pragma in some cases.

It was stated by Fractal that there would/could be a 5% - 15% CPU performance hit (I think)
due to the Compiler work around.
Has the Compiler problem been fixed and is that 5% - 15% CPU performance hit been recovered ?

Thanks
 
I guess it depends upon whether Analog Devices ever fixed their buggy compiler. I'd love to have that 5-15% back ;)
 
O2b60Again said:
It was stated by Fractal that there would/could be a 5% - 15% CPU performance hit (I think)
due to the Compiler work around.
Click to expand...

I don't recall that 5%-15% statement at all. It (not the statement but the compiler error) would be unacceptable for FAS.
 
Well, the compiler problem is definitely real. Cliff did talk about that. Apparently, analog devices sent him a compiler with a bug in it. He had to then create a workaround solution to get things to work correctly. I never heard about the 5% to 15% hit however.
 
O2b60Again said:
Hi all ...my first post
re: Cliff's statement
The firmware is approaching release quality. We finally got to the bottom of some data alignment issues. Seems there's a bug in the compiler but we found a workaround. The wait will be worth it.

It's a C compiler issue. We use SIMD processing which requires that vectors are aligned at even-word addresses. The compiler is failing to honor the alignment #pragma in some cases.

It was stated by Fractal that there would/could be a 5% - 15% CPU performance hit (I think)
due to the Compiler work around.
Has the Compiler problem been fixed and is that 5% - 15% CPU performance hit been recovered ?

Thanks
Click to expand...

Please provide a reference for the CPU performance hit - I never heard about it.
 
No worries then, eh? For us end users, anyway. Something that will get eventually get sorted out with a fix to the compiler, with stability and performance benefits in the release firmware. And its not like I'm experiencing any down time: my FM3 is working its mojo as long as I keep my fingers moving. Extra work for the code-meisters at Fractal, though. I'll honor them with performance enhanced arpeggios.
 
Last edited:
hippietim said:
Please provide a reference for the CPU performance hit - I never heard about it.
Click to expand...
I read it in one of the FM3 threads (pretty sure) ... however I can't find it now.
Maybe it was wiped for some reason. I could be wrong however. Nobody seems
to have it read it but me. Oh well ... glad I'm wrong though.
 
O2b60Again said:
I read it in one of the FM3 threads (pretty sure) ... however I can't find it now.
Maybe it was wiped for some reason. I could be wrong however. Nobody seems
to have it read it but me. Oh well ... glad I'm wrong though.
Click to expand...
You’re probably interleaving some other thread where someone was talking about taking a cpu hit due to a preset issue or something along those lines.
 
You must log in or register to reply here.
Back
Top Bottom