The third Radeon drm-next-3.8 pull by Alex Deucher provides CS ioctl support for a-synchrnous DMA ring access while the other changes come down to bug-fixes.
- Add 6xx/7xx CS parser for async DMA (v2)
- Add evergreen/cayman CS parser for async DMA (v2)
- Add VM CS parser support for async DMA on cayman/TN/SI
- Enable the async DMA rings in the CS ioctl
- Bump version for CS ioctl support for async DMA
- Add more pedantic checks in the CP DMA checker
- Add W|RREG32_IDX for MM_INDEX|DATA based mmio accesss
- Make indirect register access concurrency-safe
- Fix fence locking in the pageflip callback
- Fix regression with eviction since evict caching changes
- Fix htile buffer size computation for command stream checker
The CS ioctl support for a-synchrnous DMA ring access is most important from this pull since it allows using this performance-enhancing feature from user-space with the Radeon Gallium3D driver in Mesa. The improvements made within the Radeon driver for the Linux 3.8 kernel can lead to serious performance improvements.
The "htile buffer size computation for command stream checker" fix is also important for the work being done by Jerome Glisse on finally having working R600 Gallium3D HyperZ support.
This newest pull request can be found on dri-devel.