The repository is ready for deployment and release purposes.
Bi-weekly (twice weekly) builds have been setup by using a cron job.
All contributions are welcome! :)
https://github.com/mvaisakh/gcc-build
Bi-weekly (twice weekly) builds have been setup by using a cron job.
All contributions are welcome! :)
https://github.com/mvaisakh/gcc-build
GitHub
GitHub - mvaisakh/gcc-build: Bleeding Edge Bare Metal GCC build script.
Bleeding Edge Bare Metal GCC build script. Contribute to mvaisakh/gcc-build development by creating an account on GitHub.
Eva GCC now supports inline LLD.
Head over to my repo for the script.
ARM
ARM64
To use it with your kernel, make sure you apply this patch.
Head over to my repo for the script.
ARM
ARM64
To use it with your kernel, make sure you apply this patch.
Since GCC 10, there have been some inliner changes that led to a lot of regressions in kernel.
These are major regressions and can be noticed during actual usage.
According to @kerneltoast, increasing the instruction limit to 1000 can help overcome these regressions.
That fix didn't turn out to help me much. With @arter97's advice on restoring the GCC 9.x logic actually helped fix the regressions that I faced in my kernels.
Here are the patches that you'll need to overcome these regressions
These are major regressions and can be noticed during actual usage.
According to @kerneltoast, increasing the instruction limit to 1000 can help overcome these regressions.
That fix didn't turn out to help me much. With @arter97's advice on restoring the GCC 9.x logic actually helped fix the regressions that I faced in my kernels.
Here are the patches that you'll need to overcome these regressions
As pointed out by @CloudedQuartz, GCC+LLD depends on libxml2.so.2, which in some cases might spam messages about the same missing.
Install the libxml2 solves this issue.
So if you're using LLD, make sure you have
Install the libxml2 solves this issue.
So if you're using LLD, make sure you have
libxml2 installed.Thanks to extensive testing done by @mochi_wwww, conclusion can be made that GCC performs much better than clang.
Result comparison coming up shortly.
@EvaGCC
Result comparison coming up shortly.
@EvaGCC
Perftop on clang shows greater irq and cpu loads on kernel.
Compiled with Proton Clang
Courtesy of @mochi_wwww
@EvaGCC
Compiled with Proton Clang
Courtesy of @mochi_wwww
@EvaGCC
Hackbench scores stay persistent with GCC while the scores are very inconsistent with clang, going all around the place.
On @mochi_wwww's test cases,
GCC usually stood in single digits around 8,9,7,9,10..
Clang couldn't stay consistent with values around 12,14,16,13,18,11...
Big shoutout to @mochi_wwww for testing out my GCC-LLD.
@EvaGCC
On @mochi_wwww's test cases,
GCC usually stood in single digits around 8,9,7,9,10..
Clang couldn't stay consistent with values around 12,14,16,13,18,11...
Big shoutout to @mochi_wwww for testing out my GCC-LLD.
@EvaGCC
We are now live on XDA!
https://forum.xda-developers.com/t/gcc-toolchain-eva-gcc-calling-all-kernel-devs.4221409/
https://forum.xda-developers.com/t/gcc-toolchain-eva-gcc-calling-all-kernel-devs.4221409/
XDA Forums
[GCC][Toolchain] Eva GCC | Calling all kernel devs!
Introducing Eva GCC Toolchain
What is GCC?
The GNU Compiler Collection (GCC) is an optimizing compiler produced by the GNU Project supporting various programming languages, hardware...
What is GCC?
The GNU Compiler Collection (GCC) is an optimizing compiler produced by the GNU Project supporting various programming languages, hardware...
GCC+LLD has been setup as a cron job, and will be built twice every week (Sundays and Thursdays, approx 5:30am IST or 0:00 UTC) from today onwards.
If you face any issues, let me know in the comments below or join my support chat.
If you face any issues, let me know in the comments below or join my support chat.
Pretty absurd, but I wanna know..
Using my gcc, has it improved battery on your device?
Using my gcc, has it improved battery on your device?
Final Results
50%
Yes
50%
No
