Is the GPUMEM option deprecated?

[erlend_eriksen]

I've been led to believe that putting GPUMEM=320M in the kernel command line options will lead to 320 megabytes being reserved for contiguous memory allocations from GPU and VPU. However, I can't see any reference to the GPUMEM option in uboot-imx: Search · gpumem · GitHub  .

The only reference I can find is in a diff from buildroot: [v2,4/7] board/boundarydevices: remove obsolete u-boot scripts - Patchwork . This diff seems to remove the option, and it seems like it only served as an alias for galcore.contiguousSize anyway.

Is the GPUMEM kernel command line option deprecated?

If not, it would be helpful if someone could point me to where it is defined in the relevant source code, like in uboot or the kernel or wherever.

[Bio_TICFSL]

Hello Erlend,

I believe you are right: "gpumem=" option specifies what amount of the CMA memory is reserved for the VPU and GPU and it would need to be lower than the "CONFIG_CMA_SIZE_MBYTES" .

Can you using set gpumem to higher value , default is 176MB, you can change to bigger. Such as 400MB to test.

"Gpumem pool is managed in GPU driver provided by verisilicon, So there is no documentation for it "

Regards

[erlend_eriksen]

Yes, thank you but this is an answer similar to what I have been given before. In practice I have found that it doesn't matter at all what value I assign to GPUMEM. Even if I set it to zero, contiguous memory allocations from the VPU still work fine.

So what I am wondering is if the option is indeed deprecated, or if it still matters in a way I haven't understood. Also, I don't understand why options given to the GPU driver would affect memory allocations made both by the VPU and GPU.