Discussion in 'Videocards - NVIDIA GeForce Drivers Section' started by klunka, Jul 31, 2020.
Lol, you are right, but you get the point : P
i doubt it, the best Devs in the World working at nvidia and they do not understand it?
is not like "i know some Basics" - "Welcome to the nvidia dev team"
the nvidia code is very complex and only Aliens are allowed to touch it
There can be undocumented adverse effects with older complex hardware that strange code deals with. Its not always obvious what problems occurred and how they could be handled.
Unless you know everything the code was created to cope with, the ability to understand it is limited.
This makes it difficult to change it or write effective new code without the wealth of experience that went into writing the original.
A lot is not as straight forward as you might think, its not about ability, its about how much time needs to be invested to make it worthwhile to make the changes.
And whether the chance of producing a good enough result is possible within a time frame.
The mention of "Aliens" makes me think GiveMe was not serious.
I think he is.
His end statement, while amusing, takes a direct dig at NVidia that follows the earlier comments.
ie he doesnt feel they have the knowledge to deal with this or are barred from touching sacred code.
Probably about as hard as it is for you to comprehend theres no evidence of its use on Windows
There is no use for the command or no use removing it? And is it hard to remove because I'm dumb, or easy to remove because I'm not? Please be more clear in your insults.
He means that its use is documented only in the Linux driver.
Currently there is no source mentioning that it is being used in the Windows driver.
"most Linux 2.6 kernels' implementations of the global_flush_tlb() interface fail to flush caches on all or some CPUs, for a variety of reasons."
"Due to the above, the NVIDIA Linux graphics driver is forced to perform heavy-weight flush/invalidation operations to avoid problems due to stale cache lines and/or TLB entries."
I think than in 561.45 MB of drivers there is more than a few line of bloatware old or useless code XD
If you want to know the size of the actual drivers, look at the kernel- and user space modules size in file system.
Pretty dull discussion, tbqh. It would be much more interesting if somebody professionally analyzed why especially the Nvidia D3D12 driver sometimes has relatively slow CPU performance. E.g. I noticed that CPU bound performance in SotTRS regressed even further with Turing vs. Pascal, Polaris is ~15-20% faster. The Linux Vulkan port runs fast on Turing too when CPU bound. Though it has GPU bound performance penalty in return, which is again not the case on AMD GPUs...
half the size is GFE
the other half excluding all other software drivers that aren't necessarily installed, is the shader recompilers that the driver includes.
I'm aware of this. My sentence wasn't so serious
But the core it's true... there is surely some bloatware or old code inside the package