On Sat, Jun 26, 2021 at 03:42:19PM -0600, Thomas Frohwein wrote:
> Hi,
> 
> This is a diff to update the ports for vulkan to the latest SDK release
> 1.2.176.1 from April. I've updated them and tested without issues with
> vkcube and vulkaninfo (both from the vulkan-tools port), as well as
> vkquake and the FNA game Cryptark with "fnaify /gldevice:Vulkan".
> 
> I tested vkquake with validation layers:
> 
> $ VK_INSTANCE_LAYERS=VK_LAYER_KHRONOS_validation vkquake
> 
> This seems to work and lists one error on start [1].

https://github.com/Novum/vkQuake/commit/db6d2a005a84cad0584a6b7c02af4d7ba56b1b0d

not yet in a non-beta release

> 
> A few notes on this update:
> 
> - For some reason, upstream named this "1.2.176.1-TAG". As versioning
>   has since progressed to 1.2.182 for non-SDK releases, I'm not
>   including the "-TAG".
> - All glslang tests succeed (this is the only one with working tests).
> - spirv-headers: use recent commit as no release tag since November
>   2020; version it as "1.5.4pl2" because there have been 2 patch tags
>   for 1.5.4 ("1.5.4.raytracing" and "1.5.4.raytracing.fixed").
> - spirv-tools: v2021.2 fails to build because of missing
>   SPV_OPERAND_TYPE_OPTIONAL_PACKED_VECTOR_FORMAT; therefore using
>   recent commit (2021.2pl20210623 from June 23rd).

Instead of 2021.2pl20210623 perhaps just 2021.2pl0?

> - vulkan-loader: remove BROKEN-i386 per the comment (was for clang
>   < 9.0.0)
> - vulkan-validation-layers: upstream defaults now to using "Robin Hood
>   Hashing." I haven't heard of this before; but per BUILD.md, this
>   would require a port of
>   https://github.com/martinus/robin-hood-hashing. Disable this for now;
>   we can figure out this out later and then enable.
> 
> Tested on Intel i7-10700 with the integrated UHD 630 GPU. Testing with
> amdgpu would be appreciated.

OpenBSD 6.9-current (GENERIC.MP) #93: Sat Jun 26 10:13:58 MDT 2021
    dera...@amd64.openbsd.org:/usr/src/sys/arch/amd64/compile/GENERIC.MP
..
amdgpu0: VEGA10 56 CU rev 0x01

vulkaninfo vkcube and vkquake are fine

> 
> ok? Comments?

Just the patch level comment above.  Everything else looks fine.
ok jsg@ either way

Reply via email to