PCIe with Asynchronous DMA support

Hi,
We are working on a project using AlphaData board connected to Xavier’s PCIe slot. But PCIe DMA read and write rates are only 200+MBps. The Xavier’s L4T version is 3.5.1.
I also noticed there is an issue that the earlier version of kernel (32.2.1?) doesn’t support “PCIe Asynchronous DMA”. I’m not sure if it is the problem.
I was wondering which version of kernel start to support Asynchronous DMA? and should I configure to enable it?

Thanks!

more information on our problem:
The AlphaData board has a Xilinx FPGA with DMA engine. We use ChipScope to monitor activities of PCIe bus.
We see shorts bursts of data over the PCIe that seem at speed for ~ 2 GBps but long periods in between.
The same FPGA board on a PC performs over 2 GBps, where we do not see the gaps in between the bursts.

any suggestions are greatly appreciated,
Thanks!

We added an option ‘nvidia,update_fc_fixup;’ in file: ( is it correct file for Xavier?)
hardware/nvidia/soc/t19x/kernel-dts/tegra194-soc/tegra194-soc-pcie.dtsi
and rebuild dtb and flash to Xavier. The PCIe DMA speed has no improvement.

What are the PCIe’s credit limit of data and posted write header in Xavier?
How can we read them? and update them?

Thanks,