[Git][xorg-team/lib/mesa][upstream-experimental] 69 commits: .pick_status.json: Update to 85c082ddbed727940535911e6bf87f7d274525bf
"Timo Aaltonen (@tjaalton)" <[email protected]>
| Newsgroups | gmane.linux.debian.devel.x |
|---|---|
| Message-ID | <[email protected]> |
Timo Aaltonen pushed to branch upstream-experimental at X Strike Force / lib / mesa Commits: 147283da by Eric Engestrom at 2026-08-05T21:36:21+02:00 .pick_status.json: Update to 85c082ddbed727940535911e6bf87f7d274525bf Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 2bf73d21 by Mary Guillemard at 2026-08-05T21:36:21+02:00 nvk: Reenable compression support on Turing with nouveau 1.4.3 Dave landed a proper fix for it that landed, let's reenable it again as nouveau got another version bump for NVDEC support. Signed-off-by: Mary Guillemard <[email protected]> Fixes: e81a292165a ("nvk: Disable compression on Turing") Reviewed-by: Mel Henning <[email protected]> (cherry picked from commit 68ee90c22ecf30fa8135b37cd8abbee8b33b73aa) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 60405449 by Karol Herbst at 2026-08-05T21:36:22+02:00 nouveau: Fix return of dangling pointer in nouveau_fence_new Backport-to: * Fixes: b3aae9c5568 ("nouveau: eliminate busy waiting on fences") Closes: https://gitlab.freedesktop.org/mesa/mesa/-/work_items/15670 Reviewed-by: Mel Henning <[email protected]> (cherry picked from commit 6f19fc972f4d3004a6719ccbcca221e33098a481) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 964e92bf by Derek Lesho at 2026-08-05T21:36:22+02:00 zink: Fix zink_bo_unmap synchronization for client pointer support. MR #41127 was incomplete: When checking if we can unmap the cpu_ptr, zink_bo_map can't use the current cpu_ptr. Setting to NULL triggers the sync path and prevents the race, even if there are more than two threads. Fixes: ce45069c ("zink: Guard bo map/unmap on map_count.") (cherry picked from commit 64894956d17709ded2fc6a68e463dd78ce8b3402) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - a8058d78 by Frank Binns at 2026-08-05T21:36:22+02:00 zink: gate some EXT_descriptor_indexing related code This isn't a required extension and similar code is already conditional in init_layouts(). Backport-to: 26.1 Signed-off-by: Frank Binns <[email protected]> Reviewed-by: Mike Blumenkrantz <[email protected]> Reviewed-by: Erik Faye-Lund <[email protected]> (cherry picked from commit c2dd50cb516ab42d8c880f9114bd046e51bc20a2) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 2b49ede1 by Lionel Landwerlin at 2026-08-05T21:36:22+02:00 anv: fix barrier for Wa_1508744258 / Wa_14024015672 We need to flush the RCC writes because the register programming for Wa_1508744258 and 3DSTATE_3D_MODE Wa_14024015672 do not. Otherwise we might trigger the HW mode while some RCC operations are still going. Fixes some hard hangs on BMG with applications using MSAA. Signed-off-by: Lionel Landwerlin <[email protected]> Fixes: ba0336ab3f ("anv: Reduce RHWO optimization (Wa_1508744258)") Reviewed-by: Tapani Pälli <[email protected]> (cherry picked from commit 7c73cd054a3d70cb5cf12fa4eeb9561d0849abc4) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 52b617c7 by Lionel Landwerlin at 2026-08-05T21:36:22+02:00 iris: fix barrier for Wa_1508744258 / Wa_14024015672 We need to flush the RCC writes because the register programming for Wa_1508744258 and 3DSTATE_3D_MODE Wa_14024015672 do not. Otherwise we might trigger the HW mode while some RCC operations are still going. Signed-off-by: Lionel Landwerlin <[email protected]> Cc: mesa-stable Reviewed-by: Tapani Pälli <[email protected]> (cherry picked from commit e36f7e6e092c485a72415b4fd54efe80b17b9b49) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - fdf45a96 by Georg Lehmann at 2026-08-05T21:36:22+02:00 nir/unsigned_upper_bound: fix float to int conversions AC:Shadows multiplies fsat(x) by 2**32 and converts that to unsigned. The float upper bound is larger than UINT32_MAX, so we get an UB analysis result, which happens to be zero. nir_opt_uub then replaces it with 0. Fixes: 72ac3f60261 ("nir: add nir_unsigned_upper_bound and nir_addition_might_overflow") Closes: https://gitlab.freedesktop.org/mesa/mesa/-/work_items/15899 Reviewed-by: Rhys Perry <[email protected]> Reviewed-by: Timur Kristóf <[email protected]> (cherry picked from commit c439d52c946c0e0f0b471410d75f83c69355b5ec) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - addd4f93 by Janne Grunau at 2026-08-05T21:36:22+02:00 hk: xfb: Avoid assertions in nir_slot_num_components Avoids assertions that the shader stage is MESA_SHADER_TESS_CTRL for VARYING_SLOT_TESS_LEVEL_OUTER, VARYING_SLOT_TESS_LEVEL_INNER and VARYING_SLOT_BOUNDING_BOX1 when calling `nir_slot_num_components(i, MESA_SHADER_VERTEX)` in hk_handle_passthrough_gs(). Fixes: 8bb4f24ef1bd ("poly: allow specifying component count in passthrough GS key") Signed-off-by: Janne Grunau <[email protected]> (cherry picked from commit d4d8b05477ab399db283d28838de2228bf17effd) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 3dbf3967 by Janne Grunau at 2026-08-05T21:36:22+02:00 poly: Fix comment after moving passthrough_gs Replace outdated reference to hk_passthrough_gs_key_size after moving passthrough_gs from hk to poly in https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/41192 (noticed as comment in the merge request). Fixes: 12e59536ab6c ("poly: move hk passthrough GS code to libpoly") Signed-off-by: Janne Grunau <[email protected]> (cherry picked from commit 395de865be7bc1cf07538340753942884e547b21) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 40287468 by Lorenzo Rossi at 2026-08-05T21:36:22+02:00 pan/nir: Fix header static inline function This would've caused compile warnings if a file imported it without using the right functions Signed-off-by: Lorenzo Rossi <[email protected]> Fixes: bf9fa6e6191 ("pan/nir: Load texel buffer conversion descriptors in NIR") Reviewed-by: Ashley Smith <[email protected]› Reviewed-by: Christian Gmeiner <[email protected]> (cherry picked from commit fc8dc63cc6de5bcc82f8b8542cff0caa27387dcc) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 1d10b996 by Nick Hamilton at 2026-08-05T21:36:22+02:00 glthread: fix check for unroll draws using user VBOs when the ctx supports GLES This was originally fixed by 56b4d19 but when the gl ctx direct accesses were replaced in b5f388a the condition was mistakenly inverted. Fixes: b5f388a2220 ("mesa: replace gl ctx direct access") Signed-off-by: Nick Hamilton <[email protected]> Reviewed-by: Marek Olšák <[email protected]> (cherry picked from commit e79dfca9f9f2eba934d35dd77206503c5f9e8def) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - a1649478 by Danylo Piliaiev at 2026-08-05T21:36:22+02:00 tu: Fix tu_event not being reset on creation It didn't blow up probably because DXVK always resets events and vkd3d-proton doesn't use them. Fixes: 513184fa443 ("tu: Suballoc VkEvent BOs") Signed-off-by: Danylo Piliaiev <[email protected]> (cherry picked from commit ae84646e30cb2a4e2665b19b5d41b0b66827f116) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 0ea20600 by Danylo Piliaiev at 2026-08-05T21:36:22+02:00 tu: Merge disable_write_for_rp from secondary to primary LRZ write being disabled in one secondary wasn't propagated further. Fixes: 534cf4feeb5 ("tu/lrz: Improve LRZ around stencil tests and reads_dest cases") Signed-off-by: Danylo Piliaiev <[email protected]> (cherry picked from commit 201cbc4880398c19d06bf8da4d333084ef1d7f8c) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 4ff74a5a by Valentine Burley at 2026-08-05T21:36:22+02:00 zink: Use ShaderLayer capability for gl_Layer when available For gl_Layer, the SPIR-V specification allows the use of ShaderLayer or ShaderViewportIndexLayerEXT capabilities instead of Geometry. Fixes VUID-vkCmdPipelineBarrier-dstStageMask-04090 on devices without geometry shader support. Cc: mesa-stable Signed-off-by: Valentine Burley <[email protected]> (cherry picked from commit cb3bdc1bfd86b4caa0f90e9d8f807b8035cbd839) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 67537f66 by Christian Gmeiner at 2026-08-05T21:36:22+02:00 etnaviv: Drop stale stream output relocs in etna_update_hwxfb(..) A TFB_BUFFER_ADDR slot the current vertex shader does not write keeps the reloc of an earlier one, while etna_emit_state(..) always emits all four slots. Once the state tracker destroys that buffer, the buffer object lands in the reuse cache, and the next draw has append_bo(..) ref it straight out of its bucket, without the list_del and device ref that a real allocation does. The flush drops that reference back to zero, so the object gets parked a second time and hits the guard in etna_bo_cache_free(..): etna_bo_cache_free: Assertion `!list_is_linked(&bo->list)' failed. Clear all four slots up front instead. Deleting a transform feedback buffer while a vertex shader with different xfb outputs is bound and then drawing was enough to abort a dEQP run in the dEQP-GLES3.functional.draw_buffers_indexed.* group. Fixes: 0c6c1fa484a ("etnaviv: Implement hardware based streamout support") Signed-off-by: Christian Gmeiner <[email protected]> Reviewed-by: Lucas Stach <[email protected]> (cherry picked from commit 7b8ff5692ba6e9b4499ecabdc8e8b16445d12a60) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 26a6389c by Georg Lehmann at 2026-08-05T21:36:22+02:00 nir: mark some AMD specific shuffles as subgroup ops Cc: mesa-stable Reviewed-by: Timur Kristóf <[email protected]> (cherry picked from commit d06485a14654d95d2592f4c537334e744afc8218) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 5869c0a2 by Lucas Stach at 2026-08-05T21:36:22+02:00 etnaviv: clean up index buffer handling code a bit Reduce the scope of the index_offset variable, which isn't needed outside the index buffer handling and move the comment to the appropriate location. Cc: mesa-stable Signed-off-by: Lucas Stach <[email protected]> Reviewed-by: Christian Gmeiner <[email protected]> (cherry picked from commit fb4cad4c05ee40ab1fcaf6bac0cc38596151df4e) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 78142bb7 by Lucas Stach at 2026-08-05T21:36:22+02:00 etnaviv: move index buffer handling in draw_vbo after derived state handling After the derived states have been updated we know if we need to flush the commandstream to make room for the states we are about to emit. As the flush will invalidate the cached index BO pointer, we need to make sure that the index buffer handling happens after the potential flush. Cc: mesa-stable Signed-off-by: Lucas Stach <[email protected]> Reviewed-by: Christian Gmeiner <[email protected]> (cherry picked from commit 0833edefbea9cc00026d82f84e943193259a0149) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 1875ea57 by Lucas Stach at 2026-08-05T21:36:22+02:00 etnaviv: reserve state emission space early in draw_vbo Currently the space needed for the state emission is reserved right before we want to emit those states. However, this is too late, as we already did quite a bit of work in draw_vbo to attach the current draw state to the context. If the reservation causes a flush, most of this state is cleared and not properly accounted to the new commandbuffer. This leads to states not being properly emitted for the next draw, causing all kinds of flaky behavior including GPU hangs. As the GPU init states might be emitted together with this draw command we need to account for their size when reserving space in the cmdstream. Cc: mesa-stable Signed-off-by: Lucas Stach <[email protected]> Reviewed-by: Christian Gmeiner <[email protected]> (cherry picked from commit 120476415f416c075f5aa6bbcc3e4b5cf57d4571) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - dd228ece by Lucas Stach at 2026-08-05T21:36:22+02:00 etnaviv: move sampler source update before draw space reservation Sampler source updates will possibly insert blits for TS resolve or texture compatible sibling updates. As those take up command buffer space, they need to be done before reserving the space for the draw state emission. Cc: mesa-stable Signed-off-by: Lucas Stach <[email protected]> Reviewed-by: Christian Gmeiner <[email protected]> (cherry picked from commit 6ab174a3bc9e4ee72258d315243d392747875384) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - f977ea4e by Lionel Landwerlin at 2026-08-05T21:36:23+02:00 jay: copy resource_intel surface handle value If some debug modification adds a printf on the surface handle value, then jay currently segfault because it'll look for a def that was never set. Signed-off-by: Lionel Landwerlin <[email protected]> Fixes: e42e3193137d ("intel: add Jay") (cherry picked from commit a62bd97754eb52d2ba875528f832c5acbcf9d2db) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 55a3efba by Jeremy Huddleston at 2026-08-05T21:36:23+02:00 llvmpipe: don't create a screen when the process is not allowed to JIT On macOS, a process with library validation enabled but without the com.apple.security.cs.allow-jit entitlement cannot create writable and executable mappings, so LLVM aborts the first time it emits code. Because llvmpipe is preferred over softpipe whenever both are built, such a process crashes instead of falling back. Probe the policy with a MAP_JIT mapping, since there is no API to query it, and fail screen creation when it is denied so the existing software driver fallback selects softpipe. Cc: mesa-stable Assisted-by: Claude Opus 5 Reviewed-by: [email protected] Signed-off-by: Jeremy Huddleston Sequoia <[email protected]> (cherry picked from commit ea60532e25126eea862d6bd09f9d2e79cb1f339b) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 867cc407 by Rahul Mahantappa Bhadrashette at 2026-08-05T21:36:23+02:00 radeonsi: emit MESHLET registers for mesh shaders in gfx10_emit_shader_ngg GFX11 hardware that uses the gfx10_emit_shader_ngg() code path (has_set_context_pairs_packed=false) was missing emission of SPI_SHADER_GS_MESHLET_DIM and SPI_SHADER_GS_MESHLET_EXP_ALLOC registers for mesh shaders. This caused immediate GPU hang when executing mesh shader draw calls, as the GPU attempted mesh shader execution with uninitialized workgroup dimension and export allocation registers. The gfx11_dgpu_emit_shader_ngg() path already had proper MESHLET emission, but gfx10_emit_shader_ngg() did not, affecting GFX11 APUs that take this code path. Fixes: 74894150f18 ("radeonsi: init pm4 state for mesh shader") Reviewed-by: Qiang Yu <[email protected]> Reviewed-by: Pierre-Eric Pelloux-Prayer <[email protected]> Signed-off-by: Rahul Mahantappa Bhadrashette <[email protected]> (cherry picked from commit 271ad6c5cdc4cf5c062c82a5972720a449ae06df) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - f4472471 by Lorenzo Rossi at 2026-08-05T21:36:23+02:00 pan/compiler/stats: Fix ALU not being used in instruction bounds A last minute typo made the MAX bound read zeroed memory instead of the previously written ALU cycles. Signed-off-by: Lorenzo Rossi <[email protected]> Fixes: 320667060c5 ("panfrost/compiler: Properly compute Valhall ALU bound") (cherry picked from commit cc82e157111d488c5c7540f4042884803293ebf7) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - faf2c729 by Jesse Natalie at 2026-08-05T21:36:23+02:00 d3d12: Disable vao fast path for AMD Closes: https://gitlab.freedesktop.org/mesa/mesa/-/work_items/15924 Cc: mesa-stable (cherry picked from commit 6c306fd75a9a3bad31ef779f8607d557524e2277) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 24c06f9e by Neha Bhende at 2026-08-05T21:36:23+02:00 svga: fix shared memory index for svga driver Now TGSI shader is using correct index for shared memory This fixes shared spec@arb_compute_shader@execution@shared-atomic* on svga driver Fixes: d55b2cafb3a6 ("nir_to_tgsi: fix shared memory index") (cherry picked from commit 707cc8f5ab97485a0c4cf1a2ab13dbdf5c9469c4) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 3ef3829e by Ian Romanick at 2026-08-05T21:36:23+02:00 brw/validate: Gfx11 can't have accumulator src0 in 3-src instructions There are a lot of contradictions, incorrect information, and copy pasta in the various documentation. :( v2: Update after Caio's fixes to accumulator source encoding for 3-source instructions. Fixes: c3a5b62c081 ("brw/validate: Perform more 3-src validation in brw_validate instead of brw_eu_emit") Reviewed-by: Caio Oliveira <[email protected]> (cherry picked from commit ed4bb05db770ac44b8a542747a9942e5a4a6ac1e) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 7dd9ca6f by Jeremy Huddleston at 2026-08-05T21:36:23+02:00 glx/apple: silence OpenGL deprecation warnings in libglx 68bef96fa70b added -DGL_SILENCE_DEPRECATION to the libappleglx target, but applegl_glx.c and glxcmds.c are compiled into libglx target. Apply the flag there as well. Fixes: 68bef96fa70b ("glx/apple: silence OpenGL deprecation warnings") Assisted-by: Claude Opus 4.8 Signed-off-by: Jeremy Huddleston Sequoia <[email protected]> (cherry picked from commit 0baeca38c72ae4af7c464644189853b66af153e7) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 685edb55 by Faith Ekstrand at 2026-08-05T21:36:23+02:00 compiler/rust/cfg: Remap node edges in remove_unreachable() It's not enough to just delete the unused nodes. We also need to remap edges. We weren't remapping edges before, so if we had an unreachable node, any edges to or from nodes after that node got screwed up. This means we need the same algorithm for remove_unreachable() as we use for rev_post_order_sort(). Fixes: 3467bdbb352e ("compiler/rust/cfg: Make sorting optional in CFGBuilder::as_cfg()") Reviewed-by: Lorenzo Rossi <[email protected]> (cherry picked from commit a8259a857aa9c90eef3ffecca4ccea35a3492b55) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 0fe92173 by Danylo Piliaiev at 2026-08-05T21:36:23+02:00 tu: Fix memory leak of FDM patch-points When resuming RP in a separate command buffer patchpoints_ctx is NULL but `resuming == true`. Fixes: d4ad50752ff ("tu: Fix memory leak of patchpoints_ctx in dynamic rendering") Signed-off-by: Danylo Piliaiev <[email protected]> (cherry picked from commit 7e34817cdc0406ab452762fc97726dfd2d2c51e3) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 75614ef4 by Marek Olšák at 2026-08-05T21:36:23+02:00 nir/opt_varyings: fix incorrect counting of emit_vertex within a block It incorrectly counted emit_vertex within a block, which broke deduplication. We need to reset the counter at the first gathered instruction of a block, not at the first emit_vertex of a block. Fixes: ca137e545c10a9f4bf0f2 - nir/opt_varyings: rewrite elimination of duplicated outputs Closes: https://gitlab.freedesktop.org/mesa/mesa/-/work_items/15872 Reviewed-by: Aitor Camacho <[email protected]> (cherry picked from commit cf41ce6ac45c97fb8565e6e4326380955f922f50) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 1419cb1b by Rhys Perry at 2026-08-05T21:36:23+02:00 nir/load_store_vectorize: rework barriers to use entries We can't compare entry keys from before vectorization with those created after: a = load_ssbo(offset=4) //key=4 b = load_ssbo(offset=0) //key=0 c = load_ubo(a) //key=0+a.x barrier d = load_ubo(b) -> vectorize a/b at barrier and continue a = load_ssbo(offset=0) //key=0 c = load_ubo(a.y) //key=0+a.x d = load_ubo(a.x) //key=0+a.x -> incorrectly vectorizes c/d Rework how barriers are handled so that all entry keys for the block are created before any vectorization happens. This also fixes an issue where a nir_var_mem_global barrier didn't block global access vectorization. Signed-off-by: Rhys Perry <[email protected]> Backport-to: * Reviewed-by: Georg Lehmann <[email protected]> (cherry picked from commit 2da4c70d036b2a3115be5a990943def759fcc0d9) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 639e6edb by Rhys Perry at 2026-08-05T21:36:23+02:00 nir/load_store_vectorize: recreate entry key when adding from predecessor fossil-db (gfx1201): Totals from 12 (0.01% of 210263) affected shaders: Instrs: 31512 -> 31450 (-0.20%) CodeSize: 166652 -> 166292 (-0.22%) Latency: 85478 -> 83168 (-2.70%) InvThroughput: 13092 -> 13072 (-0.15%); split: -0.19%, +0.04% VClause: 389 -> 377 (-3.08%) Copies: 2004 -> 1979 (-1.25%) Branches: 1704 -> 1696 (-0.47%) PreSGPRs: 584 -> 575 (-1.54%) PreVGPRs: 487 -> 493 (+1.23%) VALU: 14114 -> 14106 (-0.06%); split: -0.09%, +0.04% SALU: 6889 -> 6874 (-0.22%) VMEM: 577 -> 562 (-2.60%) We can't compare entry keys from before vectorization with those created after: block1: a = load(offset=4) //key=4 b = load(offset=0) //key=0 c = load(a) //key=0+a.x block2: d = load(b) -> vectorize a/b a = load(offset=0) //key=0 c = load(a.y) //key=0+a.x block2: d = load(a.x) -> enter block2: a = load(offset=0) //key=0 c = load(a.y) //key=0+a.x block2: d = load(a.x) //key=0+a.x -> incorrectly vectorizes c/d Signed-off-by: Rhys Perry <[email protected]> Closes: https://gitlab.freedesktop.org/mesa/mesa/-/work_items/15831 Fixes: 4ca7ee7bd78 ("nir/opt_load_store_vectorize: Allow to vectorize at most one entry of each type across blocks") Reviewed-by: Georg Lehmann <[email protected]> (cherry picked from commit 728b81fd3cae3235e91a4aacc19a1b7fc30211e8) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 44981f97 by Rhys Perry at 2026-08-05T21:36:23+02:00 vtn: don't fail at uniformity decorations on variables and OpBufferPointer The results of OpVariable and OpBufferPointerEXT are objects, and the SPIR-V spec says these can be applied only to objects. Signed-off-by: Rhys Perry <[email protected]> Fixes: f35839a6bc5 ("spirv: Handle OpBufferPointerKHR") Closes: https://gitlab.freedesktop.org/mesa/mesa/-/work_items/15868 Reviewed-by: Georg Lehmann <[email protected]> Reviewed-by: Samuel Pitoiset <[email protected]> (cherry picked from commit 4fa60540d36fefbf5a346986c56c99d33a64bcdd) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 57626da6 by Faith Ekstrand at 2026-08-05T21:36:23+02:00 compiler/rust/nir: Implement Send+Sync for nir_shader_compiler_options Fixes: a84cd8930c89 ("kraid: Expose our own NIR compiler options") Acked-by: Ahmed Hesham <[email protected]> Reviewed-by: Lorenzo Rossi <[email protected]> (cherry picked from commit 8be93459a68edb8dd04a593627b7da7878bf9d8e) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 1c7b2c9f by Faith Ekstrand at 2026-08-05T21:36:23+02:00 kraid: Use nir_shader_compiler_options directly Now that it's Send+Sync, we can just use it directly instead of messing about with the array type. This fixes the alignment which was tripping up rusticl. Fixes: a84cd8930c89 ("kraid: Expose our own NIR compiler options") Acked-by: Ahmed Hesham <[email protected]> Reviewed-by: Lorenzo Rossi <[email protected]> (cherry picked from commit 0815c20329e3f639cb1a6a11a5858cef32441cd6) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 9cf23e43 by Dylan Baker at 2026-08-05T21:36:23+02:00 intel/gen: decode type of src1 in basic 2 source after setting IMM Otherwise we don't correctly translate IMM sources (such as TYPE_V) from their non-IMM types (such as TYPE_B) when decoding. Fixes: 03c32e90a98 ("intel/gen: Add gen encoding module") Closes: https://gitlab.freedesktop.org/mesa/mesa/-/work_items/15914 (cherry picked from commit a2feaa67a07ed02450d17c020e0cf3704964ffd5) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 83fc310f by Jesse Natalie at 2026-08-05T21:36:23+02:00 mesa: Fix shared state bookkeeping for dynamic share list changes (wglShareLists) Fixes: b3133e250e1 ("gallium: add pipe_context::resource_release to eliminate buffer refcounting") Reviewed-by: Mike Blumenkrantz <[email protected]> (cherry picked from commit 2eee75cfc7ad77e853a84529601dcb0b9c6e2ffb) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 57669541 by Lionel Landwerlin at 2026-08-05T21:36:23+02:00 anv: fixup the logic dealing with STATE_BYTE_STRIDE First simplify the logic with just a boolean tracking whether a special stride was programmed. Second always unset STATE_BYTE_STRIDE as the end of primary command buffers. Signed-off-by: Lionel Landwerlin <[email protected]> Fixes: 778cb59086 ("anv: optimize STATE_BYTE_STRIDE emission") Closes: https://gitlab.freedesktop.org/mesa/mesa/-/work_items/15803 Reviewed-by: Ivan Briano <[email protected]> (cherry picked from commit b683b5a33e7a9fba20bb872c0772c886dcfc8ace) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 3f227fda by Georg Lehmann at 2026-08-05T21:36:23+02:00 aco/optimizer: fix skip_smem_offset_align We actually have to update the temporary. Foz-DB Navi48: Totals from 318 (0.15% of 209101) affected shaders: Instrs: 334072 -> 333210 (-0.26%); split: -0.27%, +0.01% CodeSize: 1743652 -> 1740388 (-0.19%); split: -0.20%, +0.01% Latency: 3467824 -> 3467554 (-0.01%); split: -0.01%, +0.00% InvThroughput: 517308 -> 517239 (-0.01%) SClause: 10408 -> 10383 (-0.24%); split: -0.25%, +0.01% Copies: 29332 -> 29325 (-0.02%); split: -0.03%, +0.00% Branches: 9747 -> 9745 (-0.02%) PreSGPRs: 17170 -> 17166 (-0.02%) SALU: 66476 -> 65933 (-0.82%) Fixes: b798ace4434 ("aco/optimizer: fix skip_smem_offset_align with non temp register operands") Reviewed-by: Timur Kristóf <[email protected]> Reviewed-by: Marek Olšák <[email protected]> (cherry picked from commit b47e255b21fa382759c9c68ccbc53da037452670) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - c1cf5f0d by Rhys Perry at 2026-08-05T21:36:23+02:00 radv: drop support for cooperativeMatrixRobustBufferAccess Including this in the key breaks Fossilize-based shader caching, and I'm not aware of anything which needs it. Signed-off-by: Rhys Perry <[email protected]> Fixes: 0cdb7594d77 ("radv: track cooperative matrix robustness") Reviewed-by: Samuel Pitoiset <[email protected]> (cherry picked from commit 7a3a8748c95cc61b73d9b8c0aa93622f554396c4) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - f33b91a3 by Rob Clark at 2026-08-05T21:36:23+02:00 nir/convert_address_format: Split convert_def into two passes If we need to convert some use's back to the original address format, we cannot create the new instructions as part of iterating the uses, as this itself would create a new use which should not be re-written. Split this loop into two passes, with the first one simply collecting the uses that need to be rewritten, and a second one which actually does the re-writing. Fixes: e71449798191 ("nir: add nir_convert_address_format pass") Signed-off-by: Rob Clark <[email protected]> (cherry picked from commit b759fa487f944aac50ac7c8857f7ff6a870e8b21) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 72a890d3 by Rob Clark at 2026-08-05T21:36:23+02:00 nir/convert_address_format: Handle non-deref sources Sometimes a deref can show up as a "non-deref" src, for example storing a pointer that results from a deref chain. These need to be converted back to the original address format. Fixes CL CTS test_printf address_space. Fixes: e71449798191 ("nir: add nir_convert_address_format pass") Signed-off-by: Rob Clark <[email protected]> (cherry picked from commit 679a10c08d16ef2ac9d4e18820a2b42318733ac1) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 672a0af2 by Aitor Camacho at 2026-08-05T21:36:23+02:00 kk: Fix icd json api version Fixes: ae820e24269 ("kk: expose Vulkan 1.4") Signed-off-by: Aitor Camacho <[email protected]> (cherry picked from commit b11bdf7d3197f3c63c522c94985c5eee90ce4c60) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 4c143e5d by Georg Lehmann at 2026-08-05T21:36:23+02:00 nir: support phi sources in nir_rematerialize_deref_in_use_blocks We can't simply skip them, otherwise lowering a loop to lcssa doesn't ensure lcssa for derefs. Cc: mesa-stable Reviewed-by: Rhys Perry <[email protected]> (cherry picked from commit 461ff319cdd595df38f7e1738820914c6e253786) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 70eecce1 by Georg Lehmann at 2026-08-05T21:36:23+02:00 nir/to_lcssa: fix progress for derefs We don't insert a phi, but we still modify the shader. Cc: mesa-stable Reviewed-by: Rhys Perry <[email protected]> (cherry picked from commit 97afff90a1dd4fd7d9edf1b388486c777b03db9e) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - c0b4b0f4 by Georg Lehmann at 2026-08-05T21:36:23+02:00 nir/to_lcssa: move constants before the loop instead of creating a phi Various NIR instructions really need constant sources, so replacing them with phis breaks all kinds of assumptions. In case of loop unrolling, this creates phis that will be cleaned up later, so often the harmful effect is hidden. In case of nir_simplify_loop(loop, nir_jump_break), the new phis can't be removed later, so everything breaks as soon as there is a nir_op_extract that uses a constant from the loop. Backport-to: 26.2 Closes: https://gitlab.freedesktop.org/mesa/mesa/-/work_items/15947 Closes: https://gitlab.freedesktop.org/mesa/mesa/-/work_items/15278 Reviewed-by: Rhys Perry <[email protected]> (cherry picked from commit 50c2c924f8e4a1c0f2ef2caa42fa9710591a3bbd) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - a37f690e by Lionel Landwerlin at 2026-08-05T21:36:23+02:00 anv: only consider active view-capable queues for image views Transfer queues don't operate on views. Signed-off-by: Lionel Landwerlin <[email protected]> Cc: mesa-stable Reviewed-by: José Roberto de Souza <[email protected]> (cherry picked from commit 51735259659644ccec13a3cb1b2f22d37c3d635c) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 2cb3ed47 by Caio Oliveira at 2026-08-05T21:36:23+02:00 nir, spirv: Match debug printf argument alignment with u_printf The u_printf consumer expects each argument to start at a 4-byte-aligned offset. Use the same alignment when laying out the argument data and calculating its buffer size. Fixes: 768f4782e391 ("spirv: Fix debugPrintfEXT not working with multiple arguments") Assisted-by: Claude Code (Opus 5) Reviewed-by: Lionel Landwerlin <[email protected]> (cherry picked from commit ef80cf0f1e57eb3bbce0dfea5f396a87c35f8b30) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 65550c46 by Caio Oliveira at 2026-08-05T21:36:23+02:00 nir, spirv: Pad 3-component debug printf arguments to 4 components u_printf decodes vec3 arguments using a four-component stride. Reserve the padding component in both debug printf paths to match it. Fixes: 768f4782e391 ("spirv: Fix debugPrintfEXT not working with multiple arguments") Assisted-by: Claude Code (Opus 4.8) Reviewed-by: Lionel Landwerlin <[email protected]> (cherry picked from commit db70cf645dc92e300b18975e3d6d4be0f3799e49) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 11e46dfc by Lorenzo Rossi at 2026-08-05T21:36:23+02:00 pan/bi: Propagate swizzle in bi_optimizer_result_type bi_optimizer_result_type fuses a FCMP_V2F16 + MUX.v2i16 pair into a single FCMP_V2F16.f1/.i1 instruction. The MUX's condition input (src[2]) is the FCMP result; when that source carries a non-identity swizzle (e.g. H10 to swap lanes), the fusion was silently discarding it, causing the fused FCMP to compare the wrong lane pairs. Fix this by composing the MUX condition swizzle into both FCMP source operands before completing the fusion, mirroring what bi_fuse_discard_fcmp already does. A non-16-bit comparison cannot absorb a 16-bit lane swizzle, so we bail out in that case and leave the instructions unfused. This bug would be more present when we enable mediump io on panfrost with: dEQP-GLES31.functional.shaders.arrays_of_arrays.es31.parameter.out.bvec4_4x2_vertex Fixes: 501a66cb5ce ("pan/bi: Fuse result types") Signed-off-by: Lorenzo Rossi <[email protected]> Reviewed-by: Faith Ekstrand <[email protected]> Acked-by: Eric R. Smith <[email protected]> (cherry picked from commit 3111d15c9352fa3fc1fb4c59d5278b533a4d8826) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 4faec877 by Tanner Van De Walle at 2026-08-05T21:36:23+02:00 nir/nir_builder: inline dst_bit_size calculation in assert Move the dst_bit_size calculation inline so that we don't trigger C4189 warnings. Fixes: 067d786dd5e ("nir: disallow converting to sized booleans from nir_type_convert()") Reviewed-by: Jesse Natalie <[email protected]> (cherry picked from commit 64f9ff1d26fe56acedc50f484bc0e2bb2ed07942) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 00d3ef4b by Jose Maria Casanova Crespo at 2026-08-05T21:36:23+02:00 vc4: save the fragment constant buffer around the YUV blit vc4_yuv_blit() binds its own fragment constant buffer and then calls util_blitter_restore_constant_buffer_state(), but VC4_BLIT does not save that state, so the restore installs a zeroed constant buffer. The next fragment shader that reads a uniform then dereferences a NULL user buffer in vc4_write_uniforms(). Reproduced on a Pi3 by sampling an imported linear NV12 dmabuf with a fragment shader that reads a uniform: the shadow-texture blit runs at the top of the same draw, before the uniforms are written. Fixes: 99fab42da98 ("vc4: add blitter operations") Backport-to: * Assisted-by: Claude Opus 5 Reviewed-by: Maíra Canal <[email protected]> (cherry picked from commit 4eaf48e221d938165289e948b6b93730a12cda3e) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - e7b8abac by Jose Maria Casanova Crespo at 2026-08-05T21:36:23+02:00 vc4: unbind the textures around the blitter clears The quad the blitter draws for a partial Z/S clear, or for a surface clear, is a draw like any other, so vc4_predraw_check_textures() refreshes the shadow texture of a bound raster-order texture from inside util_blitter. That re-enters the blitter, which is not re-entrant: the inner blit invalidates the saved state and the outer restore then binds INVALID_PTR. Unbind the textures for the duration of the clear, the way vc4_yuv_blit() already does, and let u_blitter put them back. Reproduced on a Pi3 with an imported NV12 dmabuf bound (its BO is never private, so the shadow is refreshed on every draw) followed by a depth-only clear: u_blitter:575: Caught recursion. This is a driver bug. vc4_rasterizer_state_bind (hwcso=0xffffffffffffffff) util_blitter_restore_vertex_states util_blitter_clear_custom vc4_clear Backport-to: * Assisted-by: Claude Opus 5 Reviewed-by: Maíra Canal <[email protected]> (cherry picked from commit e3aab667bcc0f1f4eaa531ecb2986244d96bb1e9) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 86f80f66 by Valentine Burley at 2026-08-05T21:36:23+02:00 freedreno: Increase reg_size_vec4 for A702 A702 was inheriting reg_size_vec4 = 48 from a6xx_gen1_low, which was sized for A610's 64-thread wave. A702 uses a 16-thread wave, so the inherited value underestimated the capacity the compiler should assume per wave. Fixes: 639d7946c71 ("freedreno: Add initial A702 support") Signed-off-by: Valentine Burley <[email protected]> (cherry picked from commit 6a282e8f18a16c3445ce3498ecd5c09667e97555) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 398382cf by Valentine Burley at 2026-08-05T21:36:24+02:00 freedreno: Fix VPC_RAST_STREAM_CNTL register layouts It turns out that offset 0x9107 is actually VPC_RAST_STREAM_CNTL for both A6XX and A7XX, and the register at 0x9980 (previously marked as VPC_RAST_STREAM_CNTL for A6XX) is actually PC_RAST_STREAM_CNTL. VPC_UNKNOWN_9107 is removed entirely as it was just an alias for VPC_RAST_STREAM_CNTL at 0x9107. Fixes dEQP-GLES3.functional.rasterizer_discard failures with zink-on-tu on A702. Cc: mesa-stable Signed-off-by: Valentine Burley <[email protected]> (cherry picked from commit 418f2963a1500481b97afeecd94959588434ad4a) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 4bb89e49 by Samuel Pitoiset at 2026-08-05T21:36:24+02:00 spirv: fix handling OpAbortKHR This should be considered a terminator block instruction as explained in the SPIR-V spec. This fixes parsing SPIR-V for dEQP-VK-experimental.postmortem.shader_abort.* since the tests have been updated. Fixes: 88fb73c8838 ("spirv: implement SPV_KHR_abort") Signed-off-by: Samuel Pitoiset <[email protected]> (cherry picked from commit f321d7ebf427b503d28c8345deae6ad070d478cf) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 11db4dbf by Samuel Pitoiset at 2026-08-05T21:36:24+02:00 radv: fix invalid assertions in DGC when queues aren't enabled This fixes a SIGFPE issues on BC250 because the compute queue is disabled. Cc: mesa-stable Signed-off-by: Samuel Pitoiset <[email protected]> (cherry picked from commit 075f0eb7abc59a7f2d4f2178ea95899b20e8b98b) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - fd9b0bba by Kenneth Graunke at 2026-08-05T21:36:24+02:00 intel: Temporarily disable madvise on iris on xe.ko Enabling madvise for iris on xe.ko led to a large performance regression (see discussion in !42952). It turns out that this was primarily caused by a bug in xe.ko, which has been fixed and was marked for stable branches. iris's usage of madvise is also poorly tuned, and we may want some fixes for that (see !43016 for a proposal). This patch temporarily reverts the feature for Mesa 26.2, matching the previous behavior. This gives us more time for the xe.ko fix to propagate to people's systems, and also for us to tune the iris code. Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - df5f30f7 by Eric Engestrom at 2026-08-05T21:36:24+02:00 [26.2 only] docs/new_features: mention that VK_EXT_host_image_copy was exposed on RADV/GFX10.3+ This was forgotten in 87e95c5e50ead830a1c8 ("radv: advertise VK_EXT_host_image_copy by default on GFX10.3+"); see https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40996#note_3597624 Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 2e29ea10 by Vlad Zahorodnii at 2026-08-05T21:36:24+02:00 wsi/wayland: Add support for wl_fixes.ack_global_remove The wl_fixes.ack_global_remove request signals the compositor that the client will not bind the removed global. It can be used by the compositor to decide when it is safe to actually destroy the corresponding global. If a global is destroyed too soon, some clients may get disconnected. Signed-off-by: Vlad Zahorodnii <[email protected]> (cherry picked from commit c541f5e93061f98bbe83c98bea2336184966142b) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 4d672e8c by Mike Blumenkrantz at 2026-08-05T21:36:24+02:00 zink/ci: switch zink/anv jobs to surfaceless+i915 this makes the trace jobs >=99% stable (cherry picked from commit 97d586cc66406277f0035ec25a1fa847b5307151) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 0b9aaa6c by Valentine Burley at 2026-08-05T21:36:24+02:00 ci/piglit: Switch all trace jobs to surfaceless+gbm Closes: https://gitlab.freedesktop.org/mesa/mesa/-/work_items/15969 Signed-off-by: Valentine Burley <[email protected]> (cherry picked from commit 5f6b357015170067b6ff01fa008ce49c2df71b9f) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - 20b40644 by Karol Herbst at 2026-08-05T21:36:24+02:00 clc: make libclc optional for configs not needing it Backport-to: * Acked-by: Erik Faye-Lund <[email protected]> Reviewed-by: Eric Engestrom <[email protected]> (cherry picked from commit 019669c465ab2aee294af7047f354943c5b901ee) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - f1ab1828 by Karol Herbst at 2026-08-05T21:36:24+02:00 clc: use our downstream fork of libclc There are a couple of issues with the upstream versions and release schedules and overall experience makes it dificult for us to rely on users using a proper functional implementation. And we want to ship a conformant OpenCL implementation here and that's difficult if users run untested libclc versions or libclc with known bugs that aren't going to get fixed upstream because the release is already EoL. Backport-to: * Closes: https://gitlab.freedesktop.org/mesa/mesa/-/work_items/15617 Acked-by: Erik Faye-Lund <[email protected]> Reviewed-by: Eric Engestrom <[email protected]> (cherry picked from commit 6f2f3ceab8c2424236d98e4bb2cc55ec12492dc3) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - b8f6be5a by Karol Herbst at 2026-08-05T21:36:24+02:00 rusticl: warn if we load not our own libclc fork Backport-to: * Acked-by: Erik Faye-Lund <[email protected]> Reviewed-by: Eric Engestrom <[email protected]> (cherry picked from commit ba022bc4cede739190a8112b81ed04cab80ef4fe) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/43510> - - - - - f8e86a04 by Eric Engestrom at 2026-08-05T23:14:26+02:00 docs: add release notes for 26.2.0 - - - - - 9f0a7610 by Eric Engestrom at 2026-08-05T23:14:26+02:00 VERSION: bump for 26.2.0 - - - - - 103 changed files: - .gitlab-ci/test/gitlab-ci-inc.yml - .pick_status.json - VERSION - docs/relnotes.rst - + docs/relnotes/26.2.0.rst - − docs/relnotes/new_features.txt - meson.build - src/amd/compiler/aco_optimizer.cpp - src/amd/vulkan/radv_device.c - src/amd/vulkan/radv_dgc.c - src/amd/vulkan/radv_physical_device.c - src/amd/vulkan/radv_physical_device.h - src/asahi/vulkan/hk_cmd_draw.c - src/compiler/clc/meson.build - src/compiler/nir/nir_builder.c - src/compiler/nir/nir_convert_address_format.c - src/compiler/nir/nir_deref.c - src/compiler/nir/nir_intrinsics.py - src/compiler/nir/nir_lower_printf.c - src/compiler/nir/nir_opt_load_store_vectorize.c - src/compiler/nir/nir_opt_varyings.c - src/compiler/nir/nir_range_analysis.c - src/compiler/nir/nir_to_lcssa.c - src/compiler/nir/tests/control_flow_tests.cpp - src/compiler/rust/cfg.rs - src/compiler/rust/nir.rs - src/compiler/spirv/spirv_to_nir.c - src/compiler/spirv/vtn_cfg.c - src/compiler/spirv/vtn_private.h - src/compiler/spirv/vtn_structured_cfg.c - src/compiler/spirv/vtn_variables.c - src/egl/drivers/dri2/platform_wayland.c - src/freedreno/common/freedreno_devices.py - src/freedreno/registers/adreno/a6xx.xml - src/freedreno/tests/reference/crash.log - src/freedreno/tests/reference/crash_prefetch.log - src/freedreno/tests/reference/dEQP-VK.draw.indirect_draw.indexed.indirect_draw_count.triangle_list.log - src/freedreno/tests/reference/fd-clouds.log - src/freedreno/vulkan/tu_clear_blit.cc - src/freedreno/vulkan/tu_cmd_buffer.cc - src/freedreno/vulkan/tu_event.cc - src/freedreno/vulkan/tu_pipeline.cc - src/gallium/drivers/d3d12/d3d12_screen.cpp - src/gallium/drivers/etnaviv/etnaviv_context.c - src/gallium/drivers/etnaviv/etnaviv_emit.c - src/gallium/drivers/etnaviv/etnaviv_emit.h - src/gallium/drivers/etnaviv/etnaviv_state.c - src/gallium/drivers/freedreno/a6xx/fd6_emit.cc - src/gallium/drivers/iris/iris_state.c - src/gallium/drivers/llvmpipe/lp_screen.c - src/gallium/drivers/nouveau/nouveau_context.h - src/gallium/drivers/nouveau/nouveau_fence.c - src/gallium/drivers/nouveau/nouveau_fence.h - src/gallium/drivers/nouveau/nouveau_screen.c - src/gallium/drivers/nouveau/nv30/nv30_context.c - src/gallium/drivers/nouveau/nv50/nv50_context.c - src/gallium/drivers/nouveau/nv50/nv50_context.h - src/gallium/drivers/nouveau/nv50/nv50_vbo.c - src/gallium/drivers/nouveau/nvc0/nvc0_context.c - src/gallium/drivers/nouveau/nvc0/nvc0_context.h - src/gallium/drivers/nouveau/nvc0/nvc0_query_hw.c - src/gallium/drivers/nouveau/nvc0/nvc0_vbo.c - src/gallium/drivers/radeonsi/gfx/si_state_shaders.cpp - src/gallium/drivers/svga/ci/svga-fails.txt - src/gallium/drivers/svga/svga_tgsi_vgpu10.c - src/gallium/drivers/vc4/vc4_blit.c - src/gallium/drivers/vc4/vc4_draw.c - src/gallium/drivers/zink/ci/gitlab-ci-inc.yml - src/gallium/drivers/zink/ci/traces-zink.yml - src/gallium/drivers/zink/nir_to_spirv/nir_to_spirv.c - src/gallium/drivers/zink/nir_to_spirv/nir_to_spirv.h - src/gallium/drivers/zink/zink_bo.c - src/gallium/drivers/zink/zink_compiler.c - src/gallium/drivers/zink/zink_descriptors.c - src/gallium/frontends/rusticl/core/device.rs - src/gallium/frontends/rusticl/mesa/compiler/nir.rs - src/gallium/winsys/nouveau/drm/nouveau.c - src/gallium/winsys/nouveau/drm/nouveau.h - src/glx/meson.build - src/intel/compiler/brw/brw_validate.cpp - src/intel/compiler/gen/gen_encoding.cpp - src/intel/compiler/gen/tests/basic.txt - src/intel/compiler/jay/jay_from_nir.c - src/intel/dev/xe/intel_device_info.c - src/intel/vulkan/anv_device.c - src/intel/vulkan/anv_image_view.c - src/intel/vulkan/anv_private.h - src/intel/vulkan/genX_cmd_buffer.c - src/intel/vulkan/genX_cmd_draw.c - src/kosmickrisp/vulkan/meson.build - src/mesa/main/context.c - src/mesa/main/glthread_draw.c - src/mesa/main/shared.c - src/nouveau/vulkan/nvkmd/nouveau/nvkmd_nouveau_pdev.c - src/panfrost/compiler/bifrost/bi_opt_mod_props.c - src/panfrost/compiler/bifrost/bifrost_compile.c - src/panfrost/compiler/kraid/compile.rs - src/panfrost/compiler/pan_nir.h - src/poly/nir/poly_nir.h - src/util/os_misc.c - src/util/os_misc.h - src/vulkan/device-select-layer/device_select_wayland.c - src/vulkan/wsi/wsi_common_wayland.c The diff was not included because it is too large. View it on GitLab: https://salsa.debian.org/xorg-team/lib/mesa/-/compare/5db7111ffb9f932815baf7b299cd4db772e6d301...9f0a761020bca92f2b07156a0621e5360cb8eca5 -- View it on GitLab: https://salsa.debian.org/xorg-team/lib/mesa/-/compare/5db7111ffb9f932815baf7b299cd4db772e6d301...9f0a761020bca92f2b07156a0621e5360cb8eca5 You're receiving this email because of your account on salsa.debian.org. Manage all notifications: https://salsa.debian.org/-/profile/notifications | Help: https://salsa.debian.org/help