Can the compiled compiler be optimized by PGO?
pomelohan opened this issue · comments
Like ClangBuiltLinux, it divided two stages to optimize the compiler by PGO. Can we do this cool thing for GCC?
Like ClangBuiltLinux, it divided two stages to optimize the compiler by PGO. Can we do this cool thing for GCC?
Hi
It can be done but the workflow will be much longer and complex.
I'll have to chart out the stress tests, create profiles, store the profile, get them into workflow and then use these profiles for building gcc.
Might give it a try when I have time.
Like ClangBuiltLinux, it divided two stages to optimize the compiler by PGO. Can we do this cool thing for GCC?
Hi It can be done but the workflow will be much longer and complex.
I'll have to chart out the stress tests, create profiles, store the profile, get them into workflow and then use these profiles for building gcc.
Might give it a try when I have time.
😃 Thanks for your hard work, waiting for your good news!
Like ClangBuiltLinux, it divided two stages to optimize the compiler by PGO. Can we do this cool thing for GCC?
Hi It can be done but the workflow will be much longer and complex.
I'll have to chart out the stress tests, create profiles, store the profile, get them into workflow and then use these profiles for building gcc.
Might give it a try when I have time.
Are there any real time benefits from PGO in compilation of Android Kernel?
I mean it's pretty cool to get PGO but is it really sensible?
Like ClangBuiltLinux, it divided two stages to optimize the compiler by PGO. Can we do this cool thing for GCC?
Hi It can be done but the workflow will be much longer and complex.
I'll have to chart out the stress tests, create profiles, store the profile, get them into workflow and then use these profiles for building gcc.
Might give it a try when I have time.Are there any real time benefits from PGO in compilation of Android Kernel? I mean it's pretty cool to get PGO but is it really sensible?
In Android kernels, synthetic benchmarks yielded very little difference but the overall real life experience improved when I used PGO on my kernel.
But here the discussion is about compiler being compiled with PGO and frankly I'm not sure how the results are going to be.
Like ClangBuiltLinux, it divided two stages to optimize the compiler by PGO. Can we do this cool thing for GCC?
Hi It can be done but the workflow will be much longer and complex.
I'll have to chart out the stress tests, create profiles, store the profile, get them into workflow and then use these profiles for building gcc.
Might give it a try when I have time.Are there any real time benefits from PGO in compilation of Android Kernel? I mean it's pretty cool to get PGO but is it really sensible?
Practice leads to true knowledge, at least after having the PGO version, the one can compare the differences with the original version