Please re-enable nvapi (dxvk-nvapi) on aarch64 builds for NVIDIA DGX Spark features

[scaronni]

Please revert commit 2ee837e that disables building nvapi (through dxvk-nvapi) on aarch64.

This prevents these features in a setup like the NVIDIA Spark systems (aarch64 + FEX + Proton):

• NVIDIA Smooth Motion
• NVIDIA Reflex via the additional layer
• NVIDIA DLSS (Frame Generation / Super Resolution)

The hardware is now sold by many vendors:

• https://www.acer.com/us-en/desktops-and-all-in-ones/veriton-workstations/veriton-gn100-ai-mini-workstation
• https://www.asus.com/networking-iot-servers/desktop-ai-supercomputer/ultra-small-ai-supercomputers/asus-ascent-gx10/
• https://www.dell.com/en-us/dt/lp/dell-pro-max-nvidia-ai-dev
• https://www.lenovo.com/us/en/p/workstations/thinkstation-p-series/lenovo-thinkstation-pgx-sff/len102s0023
• https://www.hp.com/us-en/workstations/zgx-nano-ai-station.html
• https://www.gigabyte.com/Consumer/ai-top/AI-TOP-ATOM/
• https://us.msi.com/Landing/Next-Level-AI-Power-Right-at-Your-Desk

Thanks!

[ivyl]

It was originally disabled because it did not build with clang for ARM64. This may have changed since.

Have you tried to build it? Did it work at all? Does any of those features work?

[Saancreed]

I was able to successfully build nvapi.dll using llvm-mingw toolchain from @bylaws with both arm64ec-w64-mingw32-g++ and arm64ec-w64-mingw32-clang++ compilers, with a cross file like so in the latter case:

[binaries]
ar = 'arm64ec-w64-mingw32-ar'
c = 'arm64ec-w64-mingw32-clang'
cpp = 'arm64ec-w64-mingw32-clang++'
ld = 'arm64ec-w64-mingw32-ld'
widl = 'arm64ec-w64-mingw32-widl'
windres = 'arm64ec-w64-mingw32-windres'
strip = 'llvm-strip'

[host_machine]
system = 'windows'
cpu_family = 'aarch64'
cpu = 'aarch64'
endian = 'little'

But unfortunately, I don't have proper hardware to test this.

[rishibaghel25]

What's required spec ?? I can test :)