Crocus Gallium3D Nears Mainline Mesa For Gallium3D i965 Through Haswell Graphics

[smitty3268]

does this support "Resizable BAR" and "AMD Smart Access Memory" like ?

article about this : https://www.basnieuwenhuizen.nl/the-...ing-from-vram/

No, because that is all about memory access over PCI Express.

Meaning it doesn't apply to iGPUs which don't have any local gpu memory. They all just use regular system memory.

[Aryma]

No, because that is all about memory access over PCI Express.

Meaning it doesn't apply to iGPUs which don't have any local gpu memory. They all just use regular system memory.

APUs Are Affected Too


Even though APUs do not have VRAM they still are affected by the same issue. Typically the GPU gets a certain amount of memory pre-allocated at boot time as a carveout. There are some differences in how this is accessed from the GPU so from the perspective of the GPU this memory can be faster.

At the same time the Linux kernel only gives uncached access to that region from the CPU, so one could expect similar performance issues to crop up.

this problem didn't happen only with PCI Express

[smitty3268]

APUs Are Affected Too


Even though APUs do not have VRAM they still are affected by the same issue. Typically the GPU gets a certain amount of memory pre-allocated at boot time as a carveout. There are some differences in how this is accessed from the GPU so from the perspective of the GPU this memory can be faster.

At the same time the Linux kernel only gives uncached access to that region from the CPU, so one could expect similar performance issues to crop up.

this problem didn't happen only with PCI Express

Do you have a source with more information on this?

Resizable Bar is literally a term from the PCIE specifications. If AMD is doing something under the SAM name for APUs, that would be interesting to know about.

[Aryma]

Do you have a source with more information on this?

Resizable Bar is literally a term from the PCIE specifications. If AMD is doing something under the SAM name for APUs, that would be interesting to know about.

source in the same article I linked

Resizable Bar is for Nvidia and AMD made "smart access memory" is same thing

but look like only support :

• AMD 400 or 500 series motherboard with Ryzen 5000 CPU or a 10th- or 11th-generation Intel motherboard and CPU
• Nvidia GeForce RTX 30 graphics card

and this makes no sense the new hardware is fast enough only old hardware user who something like this

[smitty3268]

source in the same article I linked

Ok, I found your link to Baz's blog post.

That's not referring to Resizable Bar or SAM - or rather, the 1 paragraph at the very bottom of it is, under the "Other Limitations" section. I don't see anything there about it applying to APUs.

The main thrust of that blog post, which does talk about APUs, is talking about targeting the right type of memory. That's not got any kind of fancy marketing name associated with it, it's just good programming. Note you can directly control this from the app in Vulkan by picking the memory type you want to use, as the blog mentions. In OpenGL, the driver has to automatically pick one it thinks is correct, which as the blog mentions was sometimes incorrect and resulted in a game in 2015 doubling performance once it was fixed.

That kind of thing you can generally assume that every driver is at least attempting to do. How smart it's algorithm is, I guess we don't know, but it sounds like some fairly easy code to target in order to get some large optimizations, so I'd assume all the Intel drivers have probably been doing this for a long time.