Siarhei Siamashka

I think that it is a good idea to test problems in isolation. Which means one test for DRAM and another test for DVFS. The whole purpose of having dedicated...

@kika123 Yes, it is always possible to do the 3D rendering entirely on the client side and then use the standard X11 API to move it to the screen. No...

How can we confirm that these optimizations are improving something? Some of them are not on a critical code path. The others seem to be redundant when using an optimizing...

Thanks, looks good. However the Signed-off-by header seems to be missing in the commit.

Thanks for submitting it as a github tracker issue. I'll try to find some time for this.

@mittorn Thanks for investigating the problem and providing a fix! Just to make it complete and ready to be applied, please also add a description for this new option to...

The memory subsystem can't optimize two consecutive "simple" reads because the address used by the second read is calculated from the value obtained from the first read. So the second...

I'm not sure about the exynos-drm, but I have plans to try improving the generic modesetting driver. Since it got glamor support now, also improving the software rendering backend in...

Well, but your patch does not fix any real problem. Or does it? NEON is already used by the 32-bit build of tinymembench, see https://github.com/ssvb/tinymembench/wiki/PINE64-%28Allwinner-A64%29 The tricky thing is that...

Hmm, can you dump the /proc/cpuinfo from a 32-bit application on S905? Maybe its kernel is missing this patch - http://lists.infradead.org/pipermail/linux-arm-kernel/2014-October/296878.html