11610 commits

Author	SHA1	Message	Date
Samuel Pitoiset	ad948a8ca6	radv/meta: fix missing sync for compute resolves ... Removed this by mistake during a rebase presumably. This fixes a regression with dEQP-VK.pipeline.monolithic.multisample.m10_resolve.* on <= GFX8. Fixes: 1746837a71 ("radv/meta: remove CB_RESOLVE") Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40449>	2026-03-17 11:36:53 +00:00
Rhys Perry	200555fd3b	radv: improve lower_array_layer_round_even condition ... If conformant_trunc_coord=true, then the hardware never truncates the layer and this lowering isn't needed. If conformant_trunc_coord=false, then this lowering is needed if TRUNC_COORD might be set, which is never the case if disable_trunc_coord=true. The previous version of this code always did the lowering if disable_trunc_coord=true, which shouldn't be necessary. fossil-db (navi31): Totals from 2244 (2.00% of 112041) affected shaders: MaxWaves: 50550 -> 50552 (+0.00%) Instrs: 4497677 -> 4495106 (-0.06%); split: -0.07%, +0.01% CodeSize: 23967696 -> 23958808 (-0.04%); split: -0.05%, +0.01% VGPRs: 165132 -> 165096 (-0.02%) Latency: 45174171 -> 45165056 (-0.02%); split: -0.02%, +0.00% InvThroughput: 7645548 -> 7639935 (-0.07%); split: -0.08%, +0.00% VClause: 66733 -> 66717 (-0.02%); split: -0.08%, +0.05% SClause: 83906 -> 83936 (+0.04%); split: -0.08%, +0.12% Copies: 297125 -> 297292 (+0.06%); split: -0.19%, +0.24% Branches: 69963 -> 69960 (-0.00%) PreSGPRs: 142985 -> 143002 (+0.01%) PreVGPRs: 138017 -> 137966 (-0.04%) VALU: 3022018 -> 3019855 (-0.07%); split: -0.07%, +0.00% SALU: 389844 -> 389663 (-0.05%); split: -0.12%, +0.07% VOPD: 809 -> 810 (+0.12%) Signed-off-by: Rhys Perry <pendingchaos02@gmail.com> Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Reviewed-by: Marek Olšák <marek.olsak@amd.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40409>	2026-03-17 10:27:07 +00:00
Zeyang Lyu	70a487adfb	radv: Fix incorrect misaligned_mask_invalid for VK_EXT_vertex_input_dynamic_state ... The cmd_buffer->vertex_bindings was not cleared when resetting the command buffer. The previous misaligned_mask_invalid bits were not preserved when calling radv_CmdBindVertexBuffers2 multiple times. Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40312>	2026-03-17 06:56:48 +00:00
Natalie Vock	b046eaf36d	radv/rt: Fix shared ray query stack on top of application LDS ... Since the stack pointer may wrap around the stack size in overflow cases, traversal logic calculates the real stack pointer with nir_umod_imm(b, stack, args->stack_entries * args->stack_stride). For ray queries, "stack" was initialized to "stack_base + local_invocation_idx * 4". This was completely broken, as the umod would later delete the stack base completely and overwrite the start of LDS, which belongs to the apps' shared memory. Instead, add the stack base as a constant offset in the load/store_stack callback. (This should also save 1 VALU per ray query) Also, delete radv_ray_traversal_args::stack_base since it's unused now. Cc: mesa-stable Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40420>	2026-03-16 12:43:59 +00:00
Samuel Pitoiset	e1aa011ce4	radv: stop passing radv_device for SPIR-V debug reports ... Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40379>	2026-03-16 11:55:45 +00:00
Samuel Pitoiset	639207701d	aco,radv,radeonsi: remove debug report support in ACO ... This doesn't seem very useful since ACO will abort and print the error messages to stderr. Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40379>	2026-03-16 11:55:45 +00:00
Samuel Pitoiset	d6fe5ee8a6	vulkan: do not pass vk_instance for debug report messages ... RADV wants to abstract the compiler from any instance/device/pdev objects. The previous NULL check for instance seems to be useless. Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40379>	2026-03-16 11:55:45 +00:00
David Rosca	25095cc393	radv/video: Fix coding pic_parameter_set_id in H264 slice header ... Cc: mesa-stable Reviewed-by: Ruijing Dong <ruijing.dong@amd.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40374>	2026-03-16 10:39:13 +00:00
David Rosca	0450e4ff65	radv/video: Fix AV1 encode min tile size ... Fixes: 37e71a5cb2 ("radv/video: add support for AV1 encoding") Reviewed-by: Ruijing Dong <ruijing.dong@amd.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40374>	2026-03-16 10:39:13 +00:00
Samuel Pitoiset	fd032536d9	radv: cleanup valid image layouts in radv_layout_is_htile_compressed() ... Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40154>	2026-03-16 09:16:09 +00:00
Samuel Pitoiset	205fe011ea	radv: always use separate depth/stencil layouts for rendering ... Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40154>	2026-03-16 09:16:09 +00:00
Samuel Pitoiset	7343aff8aa	radv: fix missing HTILE decompression with separate depth/stencil layouts ... These two layouts mean either the depth or the stencil aspect is compressed. Without TC-compat HTILE (mostly < GFX8), the driver must expand HTILE. This prevents regressions with Vulkan runtime code using separate depth/stencil layouts by default. Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40154>	2026-03-16 09:16:09 +00:00
Samuel Pitoiset	5192ee3c88	radv: handle separate depth/stencil layouts correctly for barriers ... Need to split the transitions, one for each aspect. Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40154>	2026-03-16 09:16:08 +00:00
Samuel Pitoiset	3c26a2fdb1	radv: handle separate depth/stencil layouts correctly for fbfetch decompressions ... Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40154>	2026-03-16 09:16:08 +00:00
Samuel Pitoiset	cb3cbf4303	radv/meta: stop setting the other depth/stencil attachments when unused ... It should be either both or depth-only or stencil-only. Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40154>	2026-03-16 09:16:07 +00:00
Samuel Pitoiset	feffa62a41	radv/meta: add separate ds layouts support to the HTILE expand pass ... This should work just fine. Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40154>	2026-03-16 09:16:07 +00:00
Samuel Pitoiset	58441d014e	radv: stop checking whether HTILE is compressed with the UNDEFINED layout ... This doesn't make sense. Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40154>	2026-03-16 09:16:07 +00:00
Samuel Pitoiset	a55b49569f	radv: only set the relevant image views for custom depth/stencil resolves ... So the rendering state has correct information, like the aspects. Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40154>	2026-03-16 09:16:07 +00:00
Samuel Pitoiset	1746837a71	radv/meta: remove CB_RESOLVE ... CB_RESOLVE isn't very fast and we already have two different paths, it's been removed in hw since GFX11. PAL and RadeonSI removed support for it too. Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39957>	2026-03-16 08:12:46 +00:00
Samuel Pitoiset	9c05ca4366	radv/meta: use the fragment resolve path by default ... CB_RESOLVE is going to be removed. The driver will fallback to the compute resolve path if necessary. Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39957>	2026-03-16 08:12:46 +00:00
Samuel Pitoiset	90a60c84b5	ac/rtld: remove radeon_info ... Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40375>	2026-03-16 07:50:02 +00:00
Samuel Pitoiset	f6b4acdf45	radv: remove radv_use_llvm_for_stage() ... This was useful few years ago to investigate ACO vs LLVM issues but now it seems unnecessary. Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40375>	2026-03-16 07:50:02 +00:00
Samuel Pitoiset	a7cc55c33b	radv: remove unnecessary radv_device parameter to few functions ... Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40375>	2026-03-16 07:50:01 +00:00
Samuel Pitoiset	1f20feacaa	radv: remove radv_nir_compiler_options::info ... Passing radeon_info is bad. Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40375>	2026-03-16 07:50:01 +00:00
Samuel Pitoiset	c8499885cb	radv: use radv_physical_device_cache_key::use_ngg_culling more ... Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40375>	2026-03-16 07:50:00 +00:00
Samuel Pitoiset	744cedf3a1	radv: remove unused radv_device parameter to few functions ... Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40375>	2026-03-16 07:50:00 +00:00
Samuel Pitoiset	8d1e148c37	radv: remove empty gather_shader_info_cs() ... Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40375>	2026-03-16 07:50:00 +00:00
Samuel Pitoiset	23dcabcb72	radv: tidy up radv_postprocess_nir() ... Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40375>	2026-03-16 07:49:59 +00:00
Natalie Vock	867d0b33b3	radv/rt: Bump ray query stack base limit for GFX12 ... GFX12 encoding added one bit to the stack offset, doubling the limit on the stack base offset that is possible to encode. In practice, this always allows using bvh_stack_push* instructions on GFX12 since LDS is still 64kB. Cc: mesa-stable Fixes: 59a39779 (radv/rt: Only use ds_bvh_stack_rtn if the stack base is possible to encode) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40405>	2026-03-14 15:27:49 +00:00
Benjamin Cheng	e76a5e69ff	radv/video_enc: Use variable slice mode when possible ... Reviewed-by: David Rosca <david.rosca@amd.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40329>	2026-03-12 20:22:30 +00:00
Samuel Pitoiset	1ed37ae5e6	radv/meta: remove redundant cache flushes when copying VRS rates to HTILE ... No need to invalidate the VCACHE again (applications are supposed to emit a barrier) and INV_SCACHE/INV_L2 are not necessary either. Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40342>	2026-03-12 18:38:21 +00:00
Samuel Pitoiset	2afbf363ca	radv: handle FRAGMENT_SHADING_RATE_ATTACHMENT_READ properly on GFX10.3 ... When VRS rates are copied from the VRS image to HTILE. Only the vector cache needs to be flushed becaue it's using VMEM. Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40342>	2026-03-12 18:38:21 +00:00
Samuel Pitoiset	f65290f6f9	radv: advertise VK_KHR_copy_memory_indirect on GFX8+ ... GFX6-7 don't support indirect unaligned dispatches. Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/37634>	2026-03-12 12:33:04 +00:00
Samuel Pitoiset	99b3f4c7ab	radv: implement VK_KHR_copy_memory_indirect ... Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/37634>	2026-03-12 12:33:02 +00:00
Samuel Pitoiset	6b4a04505f	radv/meta: make some functions non-static ... Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/37634>	2026-03-12 12:33:02 +00:00
Marek Olšák	796e1749c6	ac: replace some packet field definitions in sid.h by generated ones ... Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40183>	2026-03-11 18:54:20 +00:00
Georg Lehmann	07d5c2cbaa	radv: set no_signed_zero for FS store_output when format doesn't care ... Or when blending is enabled, because blending doesn't require IEEE math. Foz-DB Navi21: Totals from 367 (0.32% of 114627) affected shaders: MaxWaves: 7884 -> 7876 (-0.10%); split: +0.20%, -0.30% Instrs: 354948 -> 354386 (-0.16%); split: -0.16%, +0.00% CodeSize: 1905980 -> 1903172 (-0.15%); split: -0.15%, +0.00% VGPRs: 20208 -> 20216 (+0.04%); split: -0.08%, +0.12% Latency: 1855670 -> 1854973 (-0.04%); split: -0.06%, +0.02% InvThroughput: 540792 -> 539688 (-0.20%); split: -0.20%, +0.00% PreSGPRs: 18426 -> 18366 (-0.33%) PreVGPRs: 17213 -> 17249 (+0.21%); split: -0.05%, +0.26% VALU: 258793 -> 258237 (-0.21%); split: -0.22%, +0.00% SALU: 35168 -> 35166 (-0.01%); split: -0.01%, +0.01% Reviewed-by: Marek Olšák <marek.olsak@amd.com> Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40323>	2026-03-11 16:47:15 +00:00
Samuel Pitoiset	dfdaf6a277	radv: rewrite a comment explaining why PFP waits for ME with streamout ... Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40327>	2026-03-11 07:34:45 +00:00
Samuel Pitoiset	d9420eed9e	radv: fix missing L2 cache invalidation with streamout on GFX12 ... COPY_DATA emitted from the CP isn't coherent with L2, in case the buffer filled size needs to be copied. This fixes rare and random flickering with Mafia 3 Definitive Edition on RDNA4. Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/14697 Cc: mesa-stable Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40327>	2026-03-11 07:34:45 +00:00
Samuel Pitoiset	6c1d9612ef	radv: only emit FORCE_S_VALID(1) for MSAA depth/stencil images ... This affects GFX12 only. Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40303>	2026-03-10 17:38:55 +00:00
Samuel Pitoiset	7cd3d40f86	radv: set {color,ds}_samples for inherited rendering state ... There is no distinction for secondaries. Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40303>	2026-03-10 17:38:55 +00:00
Samuel Pitoiset	0da3714bd3	ac,radv,radeonsi: add has_db_force_stencil_valid_bug ... Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40303>	2026-03-10 17:38:54 +00:00
Samuel Pitoiset	cb5f2a0521	radv: initialize HiZ for UNDEFINED transitions on transfer queue ... This doesn't consider layers/mips because it doesn't seem possible, but it doesn't hurt correctness either, it just means HiZ is disabled. This fixes dEQP-VK.api.copy_and_blit.core.use_after_copy.*_tq on GFX12. Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40304>	2026-03-10 16:51:57 +00:00
Samuel Pitoiset	e293993fff	radv: optimize clipping performance with PA_SU_HARDWARE_SCREEN_OFFSET ... This optimization was missing in RADV for a very long time. Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/6492 Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40249>	2026-03-10 12:07:46 +00:00
Samuel Pitoiset	c7cfa5324d	radv: use common guardband computations ... That shouldn't change anything. Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40249>	2026-03-10 12:07:46 +00:00
Benjamin Cheng	8f1ace0571	radv: Disable video features for some DRM modifiers ... Filter out video decode/encode format features when the DRM modifier doesn't support video operations. Along with a CTS fix, this will fix dEQP-VK.video.formats.* on UVD/VCN1 (which do not support swizzled input). Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40203>	2026-03-09 23:27:46 +00:00
Samuel Pitoiset	8f10e2886c	radv: dump the PS epilog in the GPU hang report ... Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40219>	2026-03-09 09:53:27 +00:00
Samuel Pitoiset	4c7958ae6c	radv: apply the 1D workgroup optimization for mesh/task shaders too ... Suggested-by: Georg Lehmann <dadschoorse@gmail.com> Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40194>	2026-03-09 09:29:35 +00:00
Samuel Pitoiset	3c4cb16159	radv: fix local invocation index for mesh/task and quad derivatives on GFX12 ... It must be lowered. This fixes dEQP-VK.spirv_assembly.instruction.compute.compute_shader_derivatives.{mesh,task}.*. Cc: mesa-stable Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40194>	2026-03-09 09:29:34 +00:00
Samuel Pitoiset	4428541c54	radv/meta: fix HTILE fixup after copying depth/stencil image copies ... Typo, it should be false because it's after the copy. Fixes: 4f41818194 ("radv/meta: add a function to fixup HTILE metadata for copies on compute queue") Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40270>	2026-03-09 09:07:09 +00:00