I have used M4000/M6000 GPU before where GPU memory page size was 64KB. On TX1 module it seems that the memory page size is 4KB only. I first thought just changing the kernel configuration by enabling the flag CONFIG_ARM64_64K_PAGES=y will do the trick but with this flag kernel would not build. So I want to first ask if there is any way to make the page size to 64KB.
The reason why I want 64KB page is that we are doing huge amount to DMA transfers from FPGA to GPU memory and FPGA has limited entries for translation table. Also I see that there is no rDMA support on TX1. Could someone tell me why is that?