- Desktop
- Enterprise
- Gaming
- Graphics
- Hardware
- Linux Kernel
- Multimedia
- Operating Systems
- Programming
- System Internals
- All Projects
- Latest News
- gallium: Add a nir-to-TGSI pass
- add MenuToggleButton for split toggle/menu button
- genxml: drop gen10
- tdf#134121 pie chart: add leader lines
- zink: add handling for gs in ntv
- aco: implement 16-bit literals
- pan/bi: Add load_output support
- v3dv: add a fast path for vkCmdClearAttachments
- Latest News (cont.)
- v3dv/pipeline_cache: cache nir shaders
- v3dv: add a CPU path for buffer to image copies
- v3dv/bo: adding a BO cache
- v3dv/bo: add dump stats info
- v3dv: implement compute dispatch
- v3dv: implement indirect compute dispatch
- v3dv: implement fallback for partial image copies
- v3dv: implement partial image to buffer copies
- v3dv: implement depth bias
- v3dv: make the driver more robust against OOM
- v3dv/descriptor_set: added support for samplers
Featured content is also available via:
Mesa
Mesa is an open-source 3D graphics library that provides the commonly-used DRI and Gallium3D open-source Linux graphics drivers.
Today's Activity
- meson: Don't enable libunwind by in 'auto' mode on Android
Eric Anholt: On Android we're expected to use their backtrace library. - softpipe: Switch to using NIR as the shader format from mesa/st
Eric Anholt: This causes our TGSI to use far more temps, since NTT is currently not releasing temps from registers. - gallium: Add a nir-to-TGSI pass
Eric Anholt: The goal is to replace glsl_to_tgsi.cpp and its supporting code (~10k LOC). - nir: Add an option to not lower source mods for f64/u64/i64.
Eric Anholt: TGSI can't handle them, but we want to use this pass for nir-to-tgsi. - nir: Add a block start/end ip to live instr index metadata
Eric Anholt: I wanted it for the per-instruction live intervals metadata, and it's not much to store in general. - ci: Enable NIR_VALIDATE everywhere
Eric Anholt: I wasted a bunch of time today tracking down a spurious test results change due to a driver invoking UB by running tests where NIR validation had failed (instruction reading from components beyond vector size). - genxml: drop gen10
Lionel Landwerlin: Finishing off the job started in !6899
Yesterday's Activity
- clover/nir: add a constant folding pass before lowering mem const
Dave Airlie: If we lower mem constants first, then direct array accesses to constants never get lowered, so do a constant fold pass first to remove direct const array accesses. - isl: Enable Tigerlake HDC:L1 caches via MOCS in various cases
Kenneth Graunke: Thanks to Felix Degrood for discovering that we missed enabling this additional caching on Tigerlake! Felix also benchmarked the changes. We now use MOCS 48 (HDC:L1 + L3 + LLC) for render targets, textures, and pull constant buffers. - anv: Add a descriptor_count to descriptor sets
Jason Ekstrand: This is useful for asserting in-bounds descriptor set access. - anv: Bump the number of update-after-bind descriptors to 1M
Jason Ekstrand: It's a bit hard to exactly map our implementation to the limits described by Vulkan.
Other Activity This Week
- radeonsi: enable NGG culling by default on Navi1x PRO cards
Marek Olšák - radeonsi: enable NGG on Navi14 PRO cards
Marek Olšák: These are the only cards where it works. - radeonsi: enable NGG culling by default on gfx10.3 dGPUs
Marek Olšák - zink: add ntv handling for ARB_sample_shading
Mike Blumenkrantz - zink: add gallium handling for geometry shaders
Mike Blumenkrantz - zink: add handling for gs in ntv
Mike Blumenkrantz - zink: add ntv handling for geometry shader variables
Mike Blumenkrantz - vc4: Enable nir_lower_io for uniforms
Jose Maria Casanova Crespo: Altough the driver isn't expected to receive nir_var_uniform types from GLSL this happens currently for one of the internal driver shaders. - aco: implement 16-bit literals
Rhys Perry: We can copy any value into a 16-bit subregister with a 3 dword - panfrost: Add blend shader support to bifrost
Boris Brezillon - pan/bi: Add load_output support
Boris Brezillon: This is mapped to the LD_TILE instruction. - panfrost: Add a "Bifrost Internal Blend" descriptor
Boris Brezillon: This descriptor can be passed directly as a constant to the bifrost BLEND instruction and we'll need to pass this information to blend shaders. - pan/bi: Support indirect jumps
Boris Brezillon: We need that for blend shaders which are passed the return address through r48. - panfrost: Fix fixed-function blend on bifrost
Boris Brezillon: The conversion from a 32b float to a 16b fixed-point number was wrong. - v3d/compiler: implement load interpolated input intrinsics
Iago Toral Quiroga: We will lower GLSL interpolateAt functions to these. - broadcom/compiler: track partially interpolated fragment inputs
Iago Toral Quiroga: We will need these to implement GLSL's interpolateAt*() functions where we are required to perform interpolation in the shader at arbitrary offsets. - isl: Allow CCS for 8bpp surfaces with 3+ miplevels
Nanley Chery: I can't find a restriction for enabling CCS on these surfaces in recent versions of the Bspec. - iris: Add fast-clear restriction for 8bpp surfaces
Nanley Chery: For 8bpp surfaces on TGL, prevent LOD1+ from being fast-cleared. - zink: add more explicit fencing for transfer maps
Mike Blumenkrantz: we're using our (primitive) buffer r/w tracking here to ensure that our src buffers are synchronized before we do any kind of read operation on them - zink: add a mechanism to track current resource usage in batches
Mike Blumenkrantz - radv/aco: disable NGG GS support because it randomly hangs the GPU
Samuel Pitoiset: Disable ACO NGG GS until the random GPU hangs are fixed (one CTS run == one GPU hang here).
Mesa Release Tags
- radv: remove useless assignment of MAX_API_VERSION
Samuel Pitoiset: It's overwritten few lines below. - Bump development version and clear new_features
Dylan Baker - anv: Advertise shaderIntegerFunctions2
Jason Ekstrand: We advertised the extension string but never the feature bit. - cut 20.1 branch
Eric Engestrom - VERSION: bump after 20.0 branch point
Eric Engestrom: Closes: #2457 - Bump VERSION to 20.0.0-devel
Dylan Baker - ac: replace HAVE_LLVM with LLVM_VERSION_MAJOR for atomic-optimizations
Marek Olšák: trivial - scons: define MESA_LLVM_VERSION_STRING like the other build systems do
Eric Engestrom - lima/ppir: use ra_get_best_spill_node to select spill node
Erico Nunes: ra_get_best_spill_node is what other users of the mesa register allocator use. - radv: consider MESA_VK_VERSION_OVERRIDE when setting the api version
Eleni Maria Stea: Before setting the physical device API version, we should check if the MESA_VK_VERSION_OVERRIDE environment variable is set and take it into account.
Most Popular In The Past Month
- panfrost: gen_pack: Add the aggregate concept
Boris Brezillon: Panfrost descriptors are big and are usually built from a combination of sub-descriptors. - turnip: Add kgsl backend
Kristian H. Kristensen: Lacking a bit around fences and wsi integration, but there's enough here to actually drive the GPU. - panfrost: Add support for rbg16 formats
Boris Brezillon - radeonsi: implement 16-bit FS color outputs
Marek Olšák: This removes type conversions from 16 bits to 32 bits in the main function and then back to 16 bits in the epilog. - intel: Add support for i945g to intel_stub_gpu
Eric Anholt: I wanted to check the impact of nir-to-tgsi on the i915g driver, but I don't have an i915 set up. - v3dv: add a fast path for vkCmdClearAttachments
Iago Toral Quiroga: Since vkCmdClearAttachments executes inside a render pass, we would benefit from converting it to a draw within the current subpass job to improve batching and avoid expensive tile load/store operations. - radv: remove useless assignment of MAX_API_VERSION
Samuel Pitoiset: It's overwritten few lines below. - nir: Add nir_lower_multiview pass
Connor Abbott: Taken mostly directly from the anv pass. - nir: Add a conversion and rounding intrinsic
Jason Ekstrand: This new intrinsic is capable of handling the full range of conversions from OpenCL including rounding modes and possible saturation. - v3dv: implement partial image to buffer copies
Iago Toral Quiroga: We implement this by blitting the requested region to a linear image setup to use the buffer memory store at the requested offset. Because we can't store linear depth/stencil images, we implement copies of depth/stencil aspects by using a compatible color blit.