summaryrefslogtreecommitdiffstats
path: root/src/gallium/drivers/radeonsi
Commit message (Collapse)AuthorAgeFilesLines
* radeonsi: move code for setting one shader image into separate functionMarek Olšák2016-06-011-71/+82
| | | | | | | v2: fix set_shader_images(..., NULL). Found by Christoph Haag. Reviewed-by: Nicolai Hähnle <[email protected]> Reviewed-by: Bas Nieuwenhuizen <[email protected]>
* radeonsi: set some image descriptor fields at bind timeMarek Olšák2016-06-014-71/+111
| | | | | | | | mainly the fields that can change by reallocating a texture and changing the tile mode Reviewed-by: Nicolai Hähnle <[email protected]> Reviewed-by: Bas Nieuwenhuizen <[email protected]>
* radeonsi: fix the raster config setup for 1 RB iceland chipsAlex Deucher2016-06-011-1/+4
| | | | | | | | | I didn't realize there were 1 and 2 RB variants when this code was originally added. Reviewed-by: Marek Olšák <[email protected]> Signed-off-by: Alex Deucher <[email protected]> Cc: 11.1 11.2 12.0 <[email protected]>
* radeonsi: Decompress DCC textures in a render feedback loop.Bas Nieuwenhuizen2016-05-314-0/+113
| | | | | | | | | By using a counter to quickly reject textures that are not bound to a framebuffer, the performance impact when binding sampler_views/images is not too large. Signed-off-by: Bas Nieuwenhuizen <[email protected]> Reviewed-by: Marek Olšák <[email protected]>
* radeonsi: Add counter to check if a texture is bound to a framebuffer.Bas Nieuwenhuizen2016-05-313-0/+21
| | | | | Signed-off-by: Bas Nieuwenhuizen <[email protected]> Reviewed-by: Marek Olšák <[email protected]>
* radeonsi: fix CP DMA hazard with index buffer fetchesMarek Olšák2016-05-311-2/+20
| | | | Reviewed-by: Alex Deucher <[email protected]>
* radeonsi: do GL-compliant integer resolvesMarek Olšák2016-05-311-37/+1
| | | | | | | The GL spec has been clarified and the new rule says we should just copy 1 sample. u_blitter does the right thing. Reviewed-by: Nicolai Hähnle <[email protected]>
* radeonsi: don't flush TC at the end of IBs on DRM >= 3.2.0Marek Olšák2016-05-311-2/+3
| | | | | | It's not needed since it was fixed in the kernel. Reviewed-by: Alex Deucher <[email protected]>
* radeonsi: Don't offset OFFCHIP_BUFFERING on pre-VI cards.Bas Nieuwenhuizen2016-05-301-2/+6
| | | | | | Signed-off-by: Bas Nieuwenhuizen <[email protected]> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=96239 Reviewed-by: Marek Olšák <[email protected]>
* radeonsi: enable OpenGL 4.3Bas Nieuwenhuizen2016-05-271-0/+4
| | | | | Signed-off-by: Bas Nieuwenhuizen <[email protected]> Reviewed-by: Marek Olšák <[email protected]>
* radeonsi: always reserve output space for tess factorsMarek Olšák2016-05-271-1/+6
| | | | | Reviewed-by: Bas Nieuwenhuizen <[email protected]> Tested-by: Dave Airlie <[email protected]>
* radeonsi: Allow TES distribution between shader engines.Bas Nieuwenhuizen2016-05-264-15/+40
| | | | | | | | | | | | | The R_028B50_VGT_TESS_DISTRIBUTION value is copied from amdgpu-pro. Smaller values in the ACCUM fields seem to decrease the performance advantage from this patch, higher values don't seem to matter. v2: Add distribution mode field enums. Signed-off-by: Bas Nieuwenhuizen <[email protected]> Reviewed-by: Nicolai Hähnle <[email protected]> Reviewed-by: Marek Olšák <[email protected]>
* radeonsi: Process multiple patches per threadgroup.Bas Nieuwenhuizen2016-05-261-15/+35
| | | | | | | | | | | | | | | | | | | | | | | Using more than 1 wave per threadgroup does increase performance generally. Not using too many patches per threadgroup also increases performance. Both catalyst and amdgpu-pro seem to use 40 patches as their maximum, but I haven't really seen any performance increase from limiting the number of patches to 40 instead of 64. Note that the trick where we overlap the input and output LDS does not work anymore as the insertion of the tess factors changes the patch stride. v2: - Add comment about LDS assumptions. - Add constant for buffer size. - Fix code style. v3: - Correct limits for not splitting patches between waves. - Set max num_patches to 40 as in the proprietary driver. Signed-off-by: Bas Nieuwenhuizen <[email protected]> Reviewed-by: Marek Olšák <[email protected]>
* radeonsi: Add barrier before writing the tess factors.Bas Nieuwenhuizen2016-05-261-0/+6
| | | | | | | | The factors may be stored to LDs by another invocation than the invocation for vertex 0. Signed-off-by: Bas Nieuwenhuizen <[email protected]> Reviewed-by: Marek Olšák <[email protected]>
* radeonsi: Enable dynamic HS.Bas Nieuwenhuizen2016-05-262-5/+16
| | | | | | | | | | This allows running the TES on different CU's than the TCS which results in performance improvements. v2: Only write the control word from one invocation. Signed-off-by: Bas Nieuwenhuizen <[email protected]> Reviewed-by: Marek Olšák <[email protected]>
* radeonsi: Remove LDS layout user SGPR's from TES.Bas Nieuwenhuizen2016-05-263-13/+10
| | | | | | | | They are unused. Signed-off-by: Bas Nieuwenhuizen <[email protected]> Reviewed-by: Nicolai Hähnle <[email protected]> Reviewed-by: Marek Olšák <[email protected]>
* radeonsi: Use buffer loads and stores for passing data from TCS to TES.Bas Nieuwenhuizen2016-05-261-16/+50
| | | | | | | | | | | | | | | | We always try to use 4-component loads, as LLVM does not combine loads and they bypass the L1 cache. We can't use a similar strategy for stores and this is especially notable with the tess factors, as they are often set with separate MOV's per component in the TGSI. We keep storing to LDS and the LDS space, so we can load the outputs later, either due to the shader, of for wrting the tess factors. Signed-off-by: Bas Nieuwenhuizen <[email protected]> Reviewed-by: Nicolai Hähnle <[email protected]> Reviewed-by: Marek Olšák <[email protected]>
* radeonsi: Store inputs to memory when not using a TCS.Bas Nieuwenhuizen2016-05-263-0/+49
| | | | | | | | | | | | | | | | | We need to copy the VS outputs to memory. I decided to do this using a shader key, as the value depends on other shaders. I also switch the fixed function TCS over to monolithic, as otherwisze many of the user SGPR's need to be passed to the epilog, which increases register pressure, or complexity to avoid that. The main body of the fixed function TCS is not that interesting to precompile anyway, since we do it on demand and it is very small. v2: Use u_bit_scan64. Signed-off-by: Bas Nieuwenhuizen <[email protected]> Reviewed-by: Marek Olšák <[email protected]>
* radeonsi: Add offchip buffer address calculation.Bas Nieuwenhuizen2016-05-261-0/+124
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Instead of creating a memory area per patch and per vertex, we put the same attribute of every vertex & patch together. Most loads and stores access the same attribute across all lanes, only for different patches and vertices. For the TCS this results in tightly packed data for 4-component stores. For the TES this is not the case as within a patch the loads often also access the same vertex. However if there are < 4 vertices/patch, this still results in a reduction of the number of cache lines. In the LDS situation we only do better than worst case if the data per patch < 64 bytes, which due to the tessellation factors is pretty much never. We do not use hardware swizzling for this. It would slightly reduce the number of executed VALU instructions, but I had issues with increased wait times that I haven't been able to solve yet. Furthermore, the tbuffer_store intrinsic does not support both VGPR offset and an index, so we have a problem storing indirectly indexed outputs. This can be solved by temporarily storing arrays in LDS and then copying them, but I don't think that is worth the effort. The difference in VALU cycles hardware swizzling gives is about 0.2% of total busy cycles. That is without handling the array case. I chose for attributes instead of components as they are often accessed together, and the software swizzling takes VALU cycles for calculating offsets. v2: - Rename functions to get_tcs_tes_buffer_address. - multiply by 16 as late as possible. - Use tgsi_full_src_register_from_dst. - Remove some bad comments. Signed-off-by: Bas Nieuwenhuizen <[email protected]> Reviewed-by: Marek Olšák <[email protected]>
* radeonsi: Add user SGPR for the layout of the offchip buffer.Bas Nieuwenhuizen2016-05-263-4/+20
| | | | | | Signed-off-by: Bas Nieuwenhuizen <[email protected]> Reviewed-by: Nicolai Hähnle <[email protected]> Reviewed-by: Marek Olšák <[email protected]>
* radeonsi: Use correct parameter index for LS_OUT_LAYOUT.Bas Nieuwenhuizen2016-05-261-3/+4
| | | | | | | | | This happens to be in the right position, but that changes when TCS/TES get new parameters. Signed-off-by: Bas Nieuwenhuizen <[email protected]> Reviewed-by: Nicolai Hähnle <[email protected]> Reviewed-by: Marek Olšák <[email protected]>
* radeonsi: Add buffer load functions.Bas Nieuwenhuizen2016-05-261-0/+114
| | | | | | | | | | v2: - Use llvm.admgcn.buffer.load instrinsics for new LLVM. - Code style fixes. v3: - Code style fix. Signed-off-by: Bas Nieuwenhuizen <[email protected]> Reviewed-by: Marek Olšák <[email protected]>
* radeonsi: Define build_tbuffer_store_dwords earlier to support new users.Bas Nieuwenhuizen2016-05-261-69/+69
| | | | | | Signed-off-by: Bas Nieuwenhuizen <[email protected]> Reviewed-by: Nicolai Hähnle <[email protected]> Reviewed-by: Marek Olšák <[email protected]>
* radeonsi: Add offchip tessellation parameters.Bas Nieuwenhuizen2016-05-263-6/+34
| | | | | | Signed-off-by: Bas Nieuwenhuizen <[email protected]> Reviewed-by: Nicolai Hähnle <[email protected]> Reviewed-by: Marek Olšák <[email protected]>
* radeonsi: Add buffer for offchip storage between TCS and TES.Bas Nieuwenhuizen2016-05-264-0/+23
| | | | | | | | | | | The buffer is quite large, but should only be allocated if the application uses tessellation. Most non-games don't. v2: - Use the correct register for SI. - Add define for block size. Signed-off-by: Bas Nieuwenhuizen <[email protected]> Reviewed-by: Marek Olšák <[email protected]>
* gallium: Add a pipe cap for whether primitive restart works for patches.Kenneth Graunke2016-05-231-0/+1
| | | | | | | | | | | | | | | Some hardware supports primitive restart on patch primitives, and other hardware does not. Modern GL and ES include a query for this feature; adding a capability bit will allow us to answer it. As far as I know, AMD hardware does not support this feature, while NVIDIA and Intel hardware does. However, most Gallium drivers do not appear to support tessellation shaders yet. So, I've enabled it for nvc0 and disabled it everywhere else. Signed-off-by: Kenneth Graunke <[email protected]> Reviewed-by: Ilia Mirkin <[email protected]> Reviewed-by: Marek Olšák <[email protected]>
* radeonsi: add a workaround for a bug in LLVM <= 3.8Marek Olšák2016-05-191-0/+7
| | | | | | This is not directly applicable to stable and needs to be backported. Reviewed-by: Nicolai Hähnle <[email protected]>
* radeonsi: Mixed colorbuffer formats are unsupportedAxel Davy2016-05-181-0/+10
| | | | | | | | | | Besides depth/stencil, the hardware doesn't support mixed formats. The GL state tracker doesn't make use of them. Signed-off-by: Axel Davy <[email protected]> Reviewed-by: Marek Olšák <[email protected]>
* radeonsi: Change default behaviour for undefined COLOR0Axel Davy2016-05-181-0/+3
| | | | | | | | | d3d 9 needs COLOR0 to be 1.0 on all channels when undefined. 0.0 for the others is fine. GL behaviour is undefined. Signed-off-by: Axel Davy <[email protected]> Reviewed-by: Marek Olšák <[email protected]>
* radeonsi: force level zero on image instructions in non-fragment shaders (v2)Nicolai Hähnle2016-05-171-0/+5
| | | | | | | | | | | | | | | | | | | | | | | Section 8.9 (Texture Functions) of the OpenGL Shading Language 4.5 specification: However, automatic level of detail is computed only for fragment shaders. Other shaders operate as though the base level of detail were computed as zero. and Section 8.9.3 (Texture Gather Functions): When performing a texture gather operation, the minification and magnification filters are ignored, and the rules for LINEAR filtering in the OpenGL Specification are applied to the base level of the texture image to identify the four texels i_0 j_1, i_1 j_1, i_1 j_0, and i_0 j_0. Of course, explicit LOD or derivative variants work in all shader types. This fixes several GL4x-CTS.texture_gather.* tests. v2: TG4 is always level zero (thanks, Ilia) Reviewed-by: Marek Olšák <[email protected]>
* radeonsi: emit TXQ in separate functionsNicolai Hähnle2016-05-171-52/+78
| | | | | | | | | TXQ is sufficiently different that having in it in the same code path as texture sampling/fetching opcodes doesn't make much sense. v2: guard against NULL pointer dereferences Reviewed-by: Marek Olšák <[email protected]> (v1)
* gallium/radeon: add radeon_emitted to check for non-trivial IBsNicolai Hähnle2016-05-171-2/+2
| | | | Reviewed-by: Marek Olšák <[email protected]>
* gallium/radeon: use radeon_emitNicolai Hähnle2016-05-172-25/+25
| | | | | | | Mostly generated using a sed-script, with manual fix-up for multi-line statements. Reviewed-by: Marek Olšák <[email protected]>
* Treewide: Remove Elements() macroJan Vesely2016-05-175-11/+11
| | | | | Signed-off-by: Jan Vesely <[email protected]> Reviewed-by: Brian Paul <[email protected]>
* gallium: Add a pipe cap for arb_cull_distanceTobias Klausmann2016-05-141-0/+1
| | | | | | | | | This lets us safely enable or disable the extension as needed Signed-off-by: Tobias Klausmann <[email protected]> Reviewed-by: Edward O'Callaghan <[email protected]> Reviewed-by: Marek Olšák <[email protected]> Signed-off-by: Dave Airlie <[email protected]>
* radeonsi/sid_tables: rename reg_table to sid_reg_tableNicolai Hähnle2016-05-132-3/+3
| | | | | | | | | This is purely cosmetic, making it easier to assign blame for space used in the binary in case somebody else makes a similar cleanup effort in the future. Reviewed-by: Marek Olšák <[email protected]> Reviewed-by: Bas Nieuwenhuizen <[email protected]>
* radeonsi/sid_tables: store offset into global fields table instead of pointerNicolai Hähnle2016-05-132-9/+16
| | | | | | | This avoids relocations in the final binary. Reviewed-by: Marek Olšák <[email protected]> Reviewed-by: Bas Nieuwenhuizen <[email protected]>
* radeonsi/sid_tables: store strings by offset instead of by pointerNicolai Hähnle2016-05-132-28/+141
| | | | | | | This saves some space and avoids the need for relocations. Acked-by: Marek Olšák <[email protected]> Reviewed-by: Bas Nieuwenhuizen <[email protected]>
* radeonsi: Set declared tessellation LDS size to hardware size.Bas Nieuwenhuizen2016-05-101-16/+2
| | | | | | | | | | | | | | | The calculated limit gave problems on SI as it was > 32 KiB and the hardware LDS size on SI is only 32 KiB. It isn't correct anyway when processing multiple patches in a threadgroup. As we potentially have any number of patches such that the used LDS is at most the hardware LDS size, and exact size per patch is not known at compile time, this seems like the only valid bound. Signed-off-by: Bas Nieuwenhuizen <[email protected]> Reviewed-by: Nicolai Hähnle <[email protected]> Reviewed-by: Marek Olšák <[email protected]>
* radeonsi: consolidate radeon_add_to_buffer_list calls for DMAMarek Olšák2016-05-102-33/+0
| | | | | Reviewed-by: Alex Deucher <[email protected]> Reviewed-by: Nicolai Hähnle <[email protected]>
* gallium/radeon: flush if DMA IB memory usage is too highMarek Olšák2016-05-102-6/+6
| | | | | | | | This prevents IB rejections due to insane memory usage from many concecutive texture uploads. Reviewed-by: Alex Deucher <[email protected]> Reviewed-by: Nicolai Hähnle <[email protected]>
* radeonsi: add new SDMA texture copy codeMarek Olšák2016-05-101-0/+438
| | | | | | | | | | | | | | | This implements: - Linear-to-linear partial copies. (unaligned) - Tiled-to-linear and linear-to-tiled partial copies. (unaligned except 1-2 Bpp) - Tiled-to-tiled partial copies aligned to 8x8. v2: Extend the SDMA L2T VM fault workaround to T2L. - Same algorithm, just applied to T2L. (and using a 0-based address and surface.bo_size instead of buf->size) Reviewed-by: Alex Deucher <[email protected]> (v1) Reviewed-by: Nicolai Hähnle <[email protected]>
* gallium/radeon: fix (S)DMA read-after-write hazardsMarek Olšák2016-05-102-0/+3
| | | | | Reviewed-by: Alex Deucher <[email protected]> Reviewed-by: Nicolai Hähnle <[email protected]>
* radeonsi: raise the max size for SDMA buffer copiesMarek Olšák2016-05-102-3/+3
| | | | | Reviewed-by: Alex Deucher <[email protected]> Reviewed-by: Nicolai Hähnle <[email protected]>
* radeonsi: remove SDMA texture copy codeMarek Olšák2016-05-101-215/+2
| | | | | | | | | Most of this has never worked according to the new test. The new code will be radically different. Reviewed-by: Alex Deucher <[email protected]> Reviewed-by: Nicolai Hähnle <[email protected]>
* radeonsi: only expose *_init_*dma_functions from (S)DMA filesMarek Olšák2016-05-105-34/+31
| | | | | | | just normalizing the interfaces Reviewed-by: Alex Deucher <[email protected]> Reviewed-by: Nicolai Hähnle <[email protected]>
* gallium/radeon: implement randomized SDMA texture copy testing (v2)Marek Olšák2016-05-101-0/+3
| | | | | | | | | | | v2: - adjustments for exercising all important SDMA code paths - decrease the probability of getting huge sizes (faster testing) - increase the probability of getting power-of-two dimensions - change the memory cap to 128MB (faster testing) - better detect which engine has been used Reviewed-by: Alex Deucher <[email protected]> Reviewed-by: Nicolai Hähnle <[email protected]>
* gallium/radeon: use a common function for DMA blit preparationMarek Olšák2016-05-102-22/+5
| | | | | | | this is more robust and probably fixes some bugs already Reviewed-by: Alex Deucher <[email protected]> Reviewed-by: Nicolai Hähnle <[email protected]>
* gallium/radeon: use gart_page_size instead of hardcoded 4096Marek Olšák2016-05-101-1/+1
| | | | | Reviewed-by: Bas Nieuwenhuizen <[email protected]> Reviewed-by: Nicolai Hähnle <[email protected]>
* radeonsi: workaround for tesselation on SINicolai Hähnle2016-05-091-0/+8
| | | | | | | | | | | | | | | | | | We request more than 32KB of LDS here, which SI doesn't have. Since LLVM recently started checking the size of declared LDS allocations, all shaders involved in tesselation fail to compile on SI. Note that the entire calculation here seems wrong, given how we calculate indices for generic attributes, so the number ends up wrong on CI+ as well. A proper solution is clearly needed, but this patch should serve as a band-aid for SI in the meantime. Also note that the real size of the LDS allocation in hardware is independent from what we tell LLVM, so this is really more of a "cosmetic" change. Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=95198 Cc: "11.2" <[email protected]> Reviewed-by: Marek Olšák <[email protected]>
generated by cgit v1.2.3 (git 2.39.1) at 2024-12-28 18:54:38 +0000