Skip to content

Commit 8906ad5

Browse files
authored
Merge pull request #50 from Nullpoint56/release-notes-template
1.0.0-beta.2 notes in the release template
2 parents 8c51440 + bd7e978 commit 8906ad5

1 file changed

Lines changed: 1 addition & 4 deletions

File tree

‎docs/releases/v1.0.0-beta.2.md‎

Lines changed: 1 addition & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -11,7 +11,7 @@
1111
### Fixed
1212
- Vulkan applications presenting while RivaTuner Statistics Server runs (the overlay of MSI Afterburner, CapFrameX and HWiNFO) ended at the first present with 0xC0000409. NRI now lists physical devices with `vkEnumeratePhysicalDevices`, as the layer expects, without the extra call beta.1 made after device creation.
1313
- D3D12 micromaps: the debug layer refused every micromap creation (`[1162] ... InputBuffer and PerOmmDescs can't be null`); the size query now passes the dummy addresses the DXR specification asks for.
14-
- Vulkan AV1 encode: session parameters from a sequence without a colour description were refused (`VK_ERROR_INVALID_VIDEO_STD_PARAMETERS_KHR`); the colours are now passed as "unspecified", as AV1 defines them.
14+
- Vulkan AV1 encode: session parameters from a sequence without a colour description were refused (`VK_ERROR_INVALID_VIDEO_STD_PARAMETERS_KHR`); the colours are now passed as "unspecified", as AV1 defines them. NVIDIA's Vulkan driver (617.14 and 621.14 measured) then still fails the encode silently into a bitstream buffer without `HOST_VISIBLE` memory, a driver defect: put the AV1 bitstream in `HOST_READBACK` memory there (guide: Video).
1515
- `CreateStreamer` with the default (zero) constant buffer size was refused by NRI validation for its constant buffer location, which is not used then.
1616
- Lavapipe on Windows: every pipeline without a logic op crashed the driver; NRI now passes a valid value for the disabled logic op (a driver workaround, harmless everywhere).
1717
- `WriteShaderGroupIdentifiers` with a group range past the pipeline's groups ended the process on D3D12 and Vulkan, also with NRI validation on; validation now refuses the range with `INVALID_ARGUMENT`.
@@ -21,9 +21,6 @@
2121
- Command-buffer, queue and thread annotations allocated native memory for their name on every call; names up to 255 UTF-8 bytes are now encoded on the stack.
2222
- Documentation: `SwapChainBits.LATEST_READY` with `VSYNC` does not cap rendering at the refresh rate; the display shows the newest ready image at each vblank.
2323

24-
### Notes
25-
- NVIDIA's Vulkan driver (617.14 and 621.14 measured) fails an AV1 encode into a bitstream buffer without `HOST_VISIBLE` memory, silently: put the AV1 bitstream in `HOST_READBACK` memory there (guide: Video).
26-
2724
## Breaking changes
2825
- `NriDeviceOptions.VkEnumeratePhysicalDevicesAfterCreation` is gone: NRI enumerates physical devices itself, so the option has nothing left to do. Remove it from device options.
2926
- `NriApi.HAS_BUFFER` is `static readonly`; it could be assigned, process-wide. Use `NriBuffer.HasBuffer`.

0 commit comments

Comments
 (0)