Mesa

Mesa is an open-source 3D graphics library that provides the commonly-used DRI and Gallium3D open-source Linux graphics drivers.

broadcom/compiler: sort constant UBO loads by index and offset
Iago Toral Quiroga: This implements a NIR pass that groups together constant UBO loads for the same UBO index in order of increasing offset when the distance between them is small enough that it enables the "skip unifa write" optimization.

broadcom/compiler: add a definition for the unifa skip distance
Iago Toral Quiroga: We will be using this distance to setup another optimization in a follow-up patch.

r600: Enable sb for nir only on specific request
Gert Wollny: SB si known to be buggy and the ultimate aim is to make it go away.

egl: support no error attribute set to false with ES 1.1
Tapani Pälli: We advertise the extension so it should support case where context has attribute EGL_CONTEXT_OPENGL_NO_ERROR_KHR set to EGL_FALSE indepdendent of ES version used.

microsoft/compiler: Add some more float16 support
Jesse Natalie: We can support float16 constants, b2f16, and casts to float16.

microsoft/compiler: Implement new double pack/unpack alu ops
Jesse Natalie: MakeDouble is pretty straightforward, but SplitDouble is interesting since it returns a unique 2-element struct.

microsoft/compiler: Add a lowering pass to emit double [un]pack instructions
Jesse Natalie: Whenever we have an ALU op that's operating on a double, we'll unpack it as an integer, then repack it as a float.

nir: Add a new opcode for [un]packing doubles
Jesse Natalie: HLSL doesn't support bitcasting a 64bit integer to a double.

Other Activity This Week

radv: Support DCC modifiers fully
Bas Nieuwenhuizen

radv: Add retiling for foreign queues
Bas Nieuwenhuizen: This way modifier images that don't go through the present layout get the retile executed properly.

radv: Support DCC without DCC/FCE predicates
Bas Nieuwenhuizen: Imported images can't have this as part of the associated memory.

radv: Support DCC without a fast clear value
Bas Nieuwenhuizen: For imported images we can't have one in the associated memory.

lavapipe: add vk1.1 image swapchain support
Dave Airlie: Adding support to create images from memory in the swapchain.

gallium/threaded_context: add another rule for buffer mapping
Mike Blumenkrantz: a synchronous driver can use PIPE_MAP_ONCE to infer that a buffer is guaranteed to not be mapped multiple times, as this is only used when doing map -> memcpy -> unmap directly

venus: bring up Android support
Yiwei Zhang

venus: properly enable WSI for different platforms
Yiwei Zhang

venus: initial support for WSI
Chia-I Wu

venus: initial support for transform feedback
Chia-I Wu

venus: initial support for command buffers
Chia-I Wu

venus: initial support for module and pipelines
Chia-I Wu

venus: initial support for events and queries
Chia-I Wu

venus: initial support for render pass and fb
Chia-I Wu

venus: initial support for descriptor sets
Chia-I Wu

venus: initial support for buffers/images/samplers
Chia-I Wu

venus: initial support for VkDeviceMemory commands
Chia-I Wu

venus: initial support for queue/fence/semaphore
Chia-I Wu

venus: initial support for VkDevice commands
Chia-I Wu

venus: initial support for VkPhysicalDevice commands
Chia-I Wu

venus: initial support for vkCreateInstance
Chia-I Wu: Connect to the renderer.

venus: add a ring buffer
Chia-I Wu: vn_ring is a fast way to submit commands, skipping vn_renderer entirely.

venus: add a CS encoder/decoder
Chia-I Wu

venus: add experimental renderers
Chia-I Wu: There are a virtio-gpu renderer and a vtest renderer.

venus: add generated venus-protocol headers
Chia-I Wu

anv: Remove vkCreateDmaBufINTEL (v4)
Chad Versace: Superceded by VK_EXT_image_drm_format_modifier.

anv/image: Add 'offset' param to add_surface()
Chad Versace: Will be used by VkImageDrmFormatModifierExplicitCreateInfoEXT.

anv: Implement image acquire/release of modifier images
Chad Versace: If VkImageMemoryBarrier acquires ownership of the image on VK_QUEUE_FAMILY_FOREIGN_EXT or VK_QUEUE_FAMILY_EXTERNAL, and if the image has a DRM format modifier with an aux surface, then we must ensure that the image's aux state is consistent with the modifier.

zink: add env var to force uniform inlining
Mike Blumenkrantz

zink: implement uniform inlining
Mike Blumenkrantz

aux/trace: add GALLIUM_TRACE_TRIGGER mode
Mike Blumenkrantz: similar to amd/radv driver debug modes for sqtt, this specifies a filename which is checked on every flush(PIPE_FLUSH_END_OF_FRAME); when it exists, the next frame (and only that frame) is captured into the trace

util/hash_table: add macro for destructively iterating entries
Mike Blumenkrantz: a common usage for hash tables is for tracking exactly one instance of a pointer for a given period of time, after which the table's entries are purged and it is reused

util/set: add macro for destructively iterating set entries
Mike Blumenkrantz: a common usage for sets is for tracking exactly one instance of a pointer for a given period of time, after which the set's entries are purged and it is reused

aux/trace: add screen deduplication for zink+lavapipe tracing
Mike Blumenkrantz

lavapipe: enable Vulkan 1.1 support
Dave Airlie: This is not fully conformant but the failures are the same as the Vulkan 1.

lavapipe: enable correct workgroup sizing
Dave Airlie

gallivm: add compute shader subgroup system values support
Dave Airlie

gallivm: add subgroup lowering support
Dave Airlie

gallivm: add subgroup read invocation support
Dave Airlie

gallivm: add subgroup ballot support
Dave Airlie

gallivm: add subgroup reduction + in/ex scan support
Dave Airlie

gallivm: add subgroup elect intrinsic support
Dave Airlie

gallivm: add subgroup system values support
Dave Airlie: This just adds support for retrieving subgroup system values.

gallivm: add subgroup vote 64-bit and feq support
Dave Airlie: These are both required for vulkan subgroup support.

lavapipe: add vulkan 1.1 properties/features apis
Dave Airlie: Still have to add subgroups before advertising 1.

radv: enable DCC for concurrent images on GFX10
Samuel Pitoiset: The driver now supports DCC stores on GFX10 that means that we can keep DCC compressed on all layouts/queues.

radv: enable DCC stores with MSAA 4x/8x on GFX10+
Samuel Pitoiset: Should work now.

nir/lower_idiv: Add 8-bit and 16-bit lowering path
Alyssa Rosenzweig: Roundtrip to a larger float and divide there.

ci: Use a single template for LAVA jobs
Tomeu Vizoso: It's a pain to keep both templates in sync and there aren't really that many differences.

gallium/u_threaded: split draws that don't fit in a batch
Pierre-Eric Pelloux-Prayer: Multidraws with large draw count will try to call tc_add_slot_based_call with a payload larger than TC_CALLS_PER_BATCH.

pan/bi: Remove conversion lowering
Alyssa Rosenzweig: Inherited from Midgard, it's easier to just do at NIR->BIR time now that we have a builder that doesn't suck.

radv: clean up fence syncobj code
Samuel Pitoiset: Since RADV requires DRM 3.

broadcom/compiler: optimize branch emission for uniform break/continue
Iago Toral Quiroga: A break/continue in a loop is typically emitted like this: if (cond) { break/continue; } else { } If cond is uniform, we'll emit code for a uniform if statement and that will emit a branch right before the if to jump directly to the else (or the block after the else in this case, since the else is empty) in case cond evaluates to false.

broadcom/compiler: implement restriction for branch after setmsf
Iago Toral Quiroga

lavapipe: add support for KHR_imageless_framebuffer
Dave Airlie

lavapipe: implement CmdBindVertexBuffers with CmdBindVertexBuffers2EXT
Joshua Ashton: Easier than fixing the broken size logic here.

vulkan/util: Fix implicit declaration of ffs for Android build
Mauro Rossi: Similar to 641320ce02ff and 68bb26af63ac to avoid Android build errors

Mesa Release Tags

intel: Rename GEN_VERSIONx10 macro to GFX_VERx10
Anuj Phogat: Commands used to do the changes: export SEARCH_PATH="src/intel src/gallium/drivers/iris src/mesa/drivers/dri/i965" grep -E "GEN_VERSION" -rIl $SEARCH_PATH | xargs sed -ie "s/GEN_VERSION/GFX_VER/g"

VERSION: bump for 21.1.0 cycle
Dylan Baker

Bump version for 21.0 devel
Dylan Baker

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

Most Popular In The Past Month

llvmpipe: add reduction mode support
Dave Airlie

zink: add caching for descriptor sets
Mike Blumenkrantz

zink: split descriptor sets based on usage
Mike Blumenkrantz

lima: implement shader disk cache
Vasily Khoruzhick: Wire up disk cache routines and change fs and vs keys to use nir_sha1 instead of pointer to uncompiled shader to be able to reuse them for disk cache.

radv: Enable DCC for image stores on GFX10.
Bas Nieuwenhuizen: Since image stores can now compress and we can't track image stores this also stops using predication for DCC decompression. In GFX10 this was benchmarked to be faster.

anv: Add a pass for lowering A64 UBO access
Jason Ekstrand: Instead of load_global_constant_offset/bounded, we want to use the Intel-specific block load intrinsic whenever we can.

etnaviv: add support for NTE
Christian Gmeiner

meson, util: Make zlib optional again
Jesse Natalie: This adds a HAVE_COMPRESSION macro, which is undefined if neither zlib nor zstd are present, and is used to no-op compress.h/c.

radv: Dedupe winsyses per device
Bas Nieuwenhuizen: The heap usage/budget needs to be tracked per process instead of per device.

v3dv: Add support for the on-disk shader cache
Alejandro Piñeiro: Quoting Jason's commit message (afa8f5892), that also applies here: "The Vulkan API provides a mechanism for applications to cache their own shaders and manage on-disk pipeline caching themselves.

Mesa

Today's Activity

Yesterday's Activity

Other Activity This Week

Mesa Release Tags

Most Popular In The Past Month