summaryrefslogtreecommitdiffstats
path: root/src/gallium/drivers/virgl
Commit message (Collapse)AuthorAgeFilesLines
* virgl: use virgl_transfer_inline_write even lessGurchetan Singh2019-02-151-1/+1
| | | | | | | | | | | | | | We've noticed the Team Fortress 2 engine seems to do many small calls to glSubData(..). Let's pick our heuristic based on the resource base width, not the size of a particular upload. This will cause transfers to be batched together in the transfer queue. Revelant glbench microbenchmark -- Before: buffer_upload_dynamic_element_array_131072 = 131.17 mbytes_sec After: buffer_upload_dynamic_element_array_131072 = 6828.24 mbytes_sec Reviewed-by: Gert Wollny <[email protected]>
* virgl: use transfer queueGurchetan Singh2019-02-155-18/+36
| | | | | | | | | | This improves Unigine Valley benchmark by 3 to 10 fps (depending on the scene). It also improves the Team Fortress 2 benchmark from 6 fps to 13 fps (host: 20 fps). Reviewed-by: Gert Wollny <[email protected]>
* virgl: introduce transfer queueGurchetan Singh2019-02-155-0/+390
| | | | | | | | | | | | | | | | | | | | Transfers will be placed here at unmap time instead of incurring a VM exit. There's an attempt to deduplicate intersecting 1D transfers, which are surprisingly common. This can also help with mipmapped texture upload and smaller textures, where the majority of the time is spent in the guest kernel / QEMU -- not virglrenderer. This is shown by the GLbench texture upload benchmark: Before: texture_upload_rgba_teximage2d_32 = 64.23 mtexel_sec After: texture_upload_rgba_teximage2d_32 = 367.44 mtexel_sec v2: Split up list iteration functions (@gerddie) v3: Support for optimizing glBufferSubData Reviewed-by: Gert Wollny <[email protected]>
* virgl: add encoder functions for new protocolGurchetan Singh2019-02-152-0/+28
| | | | | | Let's encode the new protocol with new helper functions. Reviewed-by: Gert Wollny <[email protected]>
* virgl: make winsys modifications for encoded transfersGurchetan Singh2019-02-152-3/+5
| | | | | | | | | | | | | The idea is to have two command buffers: 1) One for transfers 2) One for commands, which can include transfers At flush time, (2) will be filled. Otherwise, (1) will be used to submit transfers if there are enough of them. v2: Pass size directly to cmd_buf_create (@gerddie) Reviewed-by: Gert Wollny <[email protected]>
* virgl: add extra checks in virgl_res_needs_flush_waitGurchetan Singh2019-02-151-4/+9
| | | | | | | | | | | | | | | | | | | | | | This is motivated by the following scenario: glSubBufferData(GL_ARRAY_BUFFER, ...) glFlush(..) glSubBufferData(GL_ARRAY_BUFFER, ...) glSubBufferData(GL_ARRAY_BUFFER, ...) glSubBufferData(GL_ARRAY_BUFFER, ...) This increases @davidriley's Team Fortress 2 apitrace from 1 fps to 6 fps and helps with the Chromium glbench microbenchmarks: Before: texture_update_rgba_texsubimage2d_2048 = 554.96 mtexel_sec buffer_upload_dynamic_array_12 = 0.02 mbytes_sec buffer_upload_dynamic_array_576 = 1.07 mbytes_sec After: texture_update_rgba_texsubimage2d_2048 = 612.29 mtexel_sec buffer_upload_dynamic_array_12 = 2.22 mbytes_sec buffer_upload_dynamic_array_576 = 164.89 mbytes_sec Reviewed-by: Gert Wollny <[email protected]>
* virgl: pass virgl transfer to virgl_res_needs_flush_waitGurchetan Singh2019-02-155-14/+22
| | | | Reviewed-by: Gert Wollny <[email protected]>
* virgl: keep track of number of computationsGurchetan Singh2019-02-152-3/+3
| | | | | | It's good to keep track of these things. Reviewed-by: Gert Wollny <[email protected]>
* virgl: limit command length to 16 bitsGurchetan Singh2019-02-152-5/+8
| | | | | | | | | | | Much of our logic is based around the idea the upper 16 bits of a command dword can encode the length of the command. Now that the command buffer >= 2^16 - 1, we should check for this. v2: alignment, and only check VIRGL_ENCODE_MAX_DWORDS Reviewed-by: Gert Wollny <[email protected]>
* virgl: use virgl_transfer in inline writeGurchetan Singh2019-02-151-26/+40
| | | | | | | | | | | | | Let's define a helper function and use it. This commit also allows resources to be emitted into different command buffers. Like the ioctls, send 0 for layer_stride and stride. If we actually send the real values, there are various assumptions in virglrenderer for non-1D buffers that may need to be modified. Reviewed-by: Gert Wollny <[email protected]>
* virgl: add protocol for resource transfersGurchetan Singh2019-02-152-0/+12
| | | | | | | | | Mostly similar to VIRGL_CCMD_RESOURCE_INLINE_WRITE. However, this uses the resource's already attached iovecs rather than the command buffer to transfer the data. v2: Used (1 << 16) not (1 << 15) [@gerddie] Reviewed-by: Gert Wollny <[email protected]>
* virgl: when creating / freeing transfers, pass slab pool directlyGurchetan Singh2019-02-154-14/+14
| | | | | | | This will allow us to destroy transfers w/o having a pointer to the context. Reviewed-by: Gert Wollny <[email protected]>
* virgl: unmap uploader at flush timeGurchetan Singh2019-02-151-2/+3
| | | | | | This should save some memory when allocating and freeing transfers. Reviewed-by: Gert Wollny <[email protected]>
* virgl: make alignment smaller when uploading index user buffersGurchetan Singh2019-02-151-1/+1
| | | | | | | | Since we're just uploading to guest memory, let's just align to dword size. Fixes: e0f932 ("u_upload_mgr: pass alignment to u_upload_data manually") Reviewed-by: Gert Wollny <[email protected]>
* virgl: track level cleanliness rather than resource cleanlinessGurchetan Singh2019-02-156-14/+20
| | | | | | This allows a minor optimization for texture upload. Reviewed-by: Gert Wollny <[email protected]>
* virgl: don't mark unclean after a flushGurchetan Singh2019-02-151-1/+0
| | | | | | | The guest memory is still clean until host GL touches it, which we should track elsewhere. Reviewed-by: Gert Wollny <[email protected]>
* virgl: use virgl_resource_dirty helperGurchetan Singh2019-02-156-16/+19
| | | | Reviewed-by: Gert Wollny <[email protected]>
* virgl: add ability to do finer grain dirty trackingGurchetan Singh2019-02-158-13/+15
| | | | | | There are levels to cleanliness. Reviewed-by: Gert Wollny <[email protected]>
* gallium: add PIPE_CAP_MAX_VARYINGSKarol Herbst2019-02-071-0/+4
| | | | | | | | | | | | | | | | | Some NVIDIA hardware can accept 128 fragment shader input components, but only have up to 124 varying-interpolated input components. We add a new cap to express this cleanly. For most drivers, this will have the same value as PIPE_SHADER_CAP_MAX_INPUTS for the fragment shader. Fixes KHR-GL45.limits.max_fragment_input_components Signed-off-by: Karol Herbst <[email protected]> [imirkin: rebased, improved docs/commit message] Signed-off-by: Ilia Mirkin <[email protected]> Acked-by: Rob Clark <[email protected]> Acked-by: Eric Anholt <[email protected]> Reviewed-by: Marek Olšák <[email protected]> Cc: 19.0 <[email protected]>
* virgl: ARB_query_buffer_object supportDave Airlie2019-01-317-1/+58
| | | | | | v1.1: fix size define. Reviewed-by: Gurchetan Singh <[email protected]>
* virgl: enable elapsed time queriesDave Airlie2019-01-311-1/+1
| | | | | | GL underneath always has GL_TIME_ELAPSED so always enable these. Reviewed-by: Gurchetan Singh <[email protected]>
* virgl: Set sRGB write control CAP based on host capabilitiesGert Wollny2019-01-283-0/+13
| | | | | | | | | v2: - Use the renamed CAPS - add assetions to make sure that mesa doesn't try to switch destination surface formats when it is not supported. (Ilia Mirkin) Signed-off-by: Gert Wollny <[email protected]> Reviewed-by: Gurchetan Singh <[email protected]>
* virgl: use primconvert provoking vertex properlyDave Airlie2019-01-082-8/+24
| | | | | | | | This stores the raster state and calls the correct primconvert interface using the currently bound raster state. Reviewed-By: Gert Wollny <[email protected]> Signed-off-by: Dave Airlie <[email protected]>
* virgl: remove empty fileGurchetan Singh2019-01-031-0/+0
| | | | | Fixes: 174f53 ("virgl: consolidate transfer code") Reviewed-by: Erik Faye-Lund <[email protected]>
* virgl: don't flush an empty rangeGurchetan Singh2019-01-031-0/+4
| | | | | | | | | | | | | | Otherwise, the gl-1.0-long-dlist Piglit test crashes. Fixes: db7757 ("virgl: modify how we handle GL_MAP_FLUSH_EXPLICIT_BIT") Reported by airlied@ v2: Exit on any invalid range (Erik) Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=109190 Reviewed-by: Dave Airlie <[email protected]> Reviewed-by: Erik Faye-Lund <[email protected]> Tested-by: Jakob Bornecrantz <[email protected]>
* virgl: move resource creation / import / destruction to common codeGurchetan Singh2018-12-194-114/+89
| | | | | | We can remove some duplicated code. Reviewed-by: Elie Tournier <[email protected]>
* virgl: move resource metadata into base resourceGurchetan Singh2018-12-194-91/+71
| | | | | | A resource is just a buffer with some metadata. Reviewed-by: Elie Tournier <[email protected]>
* virgl: modify how we handle GL_MAP_FLUSH_EXPLICIT_BITGurchetan Singh2018-12-194-69/+25
| | | | | | | | | | | | | | | | | | Previously, we ignored the the glUnmap(..) operation and flushed before we flush the cbuf. Now, let's just flush the data when we unmap. Neither method is optimal, for example: glMapBufferRange(.., 0, 100, GL_MAP_FLUSH_EXPLICIT_BIT) glFlushMappedBufferRange(.., 25, 30) glFlushMappedBufferRange(.., 65, 70) We'll end up flushing 25 --> 70. Maybe we can fix this later. v2: Add fixme comment in the code (Elie) Reviewed-by: Elie Tournier <[email protected]>
* virgl: make virgl_buffers use resource helpersGurchetan Singh2018-12-192-20/+11
| | | | | | We can reuse the helpers we created. Reviewed-by: Elie Tournier <[email protected]>
* virgl: make transfer code with PIPE_BUFFER targetsGurchetan Singh2018-12-191-2/+4
| | | | | | | util_format_get_blocksize returns 1 for R8 formats (all PIPE_BUFFERs are R8). Reviewed-by: Elie Tournier <[email protected]>
* virgl: consolidate transfer codeGurchetan Singh2018-12-195-59/+73
| | | | | | | | We could allocate and destroy transfers in one place. v2: Keep l_stride around. Reviewed-by: Elie Tournier <[email protected]>
* virgl: store layer_stride in metadataGurchetan Singh2018-12-192-6/+6
| | | | Reviewed-by: Elie Tournier <[email protected]>
* virgl: move vrend_get_tex_image_offset to common codeGurchetan Singh2018-12-193-26/+28
| | | | | | Will be reused. Reviewed-by: Elie Tournier <[email protected]>
* virgl: move virgl_resource_layout to common codeGurchetan Singh2018-12-193-42/+51
| | | | | | Will be reused. Reviewed-by: Elie Tournier <[email protected]>
* virgl: move texture metadata to common codeGurchetan Singh2018-12-192-12/+18
| | | | | | Will be reused. Reviewed-by: Elie Tournier <[email protected]>
* virgl: remove unnessecary codeGurchetan Singh2018-12-191-3/+0
| | | | | | | | | With commit 89b479, we moved to tracking buffer cleanliness when binding. TEST=dEQP-GLES31.functional.image_load_store.buffer.load_store.r32ui Reviewed-by: Elie Tournier <[email protected]>
* virgl: texture_transfer_pool --> transfer_poolGurchetan Singh2018-12-196-11/+11
| | | | | | It's used for all types of resources. Reviewed-by: Elie Tournier <[email protected]>
* virgl: work around bad assumptions in virglrendererErik Faye-Lund2018-12-131-1/+32
| | | | | | | | | | | | | | | | | | | | Virglrenderer does the wrong thing when given an instance divisor; it tries to use the element-index rather than the binding-index as the argument to glVertexBindingDivisor(). This worked fine as long as there was a 1:1 relationship between elements and bindings, which was the case util 19a91841c34 "st/mesa: Use Array._DrawVAO in st_atom_array.c.". So let's detect instance divisors, and restore a 1:1 relationship in that case. This will make old versions of virglrenderer behave correctly. For newer versions, we can consider making a better interface, where the instance divisor isn't specified per element, but rather per binding. But let's save that for another day. Signed-off-by: Erik Faye-Lund <[email protected]> Fixes: 19a91841c34 "st/mesa: Use Array._DrawVAO in st_atom_array.c." Reviewed-by: Mathias Fröhlich <[email protected]> Tested-By: Gert Wollny <[email protected]>
* virgl: wrap vertex element state in a structErik Faye-Lund2018-12-132-9/+21
| | | | | | | | | This just has one member for now; the handle. But this is about to change. Signed-off-by: Erik Faye-Lund <[email protected]> Reviewed-by: Mathias Fröhlich <[email protected]> Tested-By: Gert Wollny <[email protected]>
* virgl: simplify virgl_hw_set_index_bufferErik Faye-Lund2018-12-131-3/+2
| | | | | | Signed-off-by: Erik Faye-Lund <[email protected]> Reviewed-by: Mathias Fröhlich <[email protected]> Tested-By: Gert Wollny <[email protected]>
* virgl: simplify virgl_hw_set_vertex_buffersErik Faye-Lund2018-12-131-4/+2
| | | | | | Signed-off-by: Erik Faye-Lund <[email protected]> Reviewed-by: Mathias Fröhlich <[email protected]> Tested-By: Gert Wollny <[email protected]>
* virgl: force linear texturing supportErik Faye-Lund2018-12-121-2/+3
| | | | | | | | | | | | | | | When I made sure that half-float texture-filtering was required for ES3, I didn't realize that virgl doesn't report support for this correctly. This regressed the GLES version available on top of several drivers, including i965 from 3.2 to 2.0. This is going to need protocol changes to fix properly, so let's just restore the previous behavior by enabling floating-point filtering unconditionally for now. Signed-off-by: Erik Faye-Lund <[email protected]> Fixes: fcf9fcee3c8 "mesa/main: do not require float-texture filtering for es3" Reviewed-by: Gurchetan Singh <[email protected]>
* virgl: fix const warning on debug flags.Dave Airlie2018-12-042-3/+3
| | | | Fixes: 8d4bb6e5c (virgl: Add command and flags to initiate debugging on the host (v2))
* virgl: don't mark buffers as unclean after a writeGurchetan Singh2018-11-302-1/+10
| | | | | | | | | | | | | | | | | We can mark the buffer unclean if it's ever bound as a TBO, SSBO, ABO, or image. This improves dEQP-GLES3.performance.buffer.data_upload.function_call.map_buffer_range.new_specified_buffer.flag_write_full.stream_draw from 9.58 MB/s to 451.17 MB/s. v2: Track buffer cleanliness as a function of bindings (Ilia). v3: virgl_modify_clean --> virgl_dirty_res (Erik) Tested-By: Gert Wollny <[email protected]> Reviewed-by: Erik Faye-Lund <[email protected]>
* virgl: avoid large inline transfersGurchetan Singh2018-11-301-1/+5
| | | | | | | | | | | | | | | | | | | | We flush everytime the command buffer (16 kB) is full, which is quite costly. This improves dEQP-GLES3.performance.buffer.data_upload.function_call.buffer_data.new_buffer.usage_stream_draw from 111.16 MB/s to 1930.36 MB/s. In addition, I made the benchmark produce buffers from 0 --> VIRGL_MAX_CMDBUF_DWORDS * 4, and tried ((VIRGL_MAX_CMDBUF_DWORDS * 4) / 2), ((VIRGL_MAX_CMDBUF_DWORDS * 4) / 4), etc. I didn't notice any clear differences, so let's just go with the most obvious heuristic. Tested-By: Gert Wollny <[email protected]> Reviewed-by: Erik Faye-Lund <[email protected]>
* virgl: quadruple command buffer sizeGurchetan Singh2018-11-301-1/+1
| | | | | | | | | | | | Tested running WebGL aquarium on Nvidia host (10,000 fishes) This moves us from 7 fps to 9 fps. After quadrupling, performance gains diminish. v2: Remove change ID (Erik) Tested-By: Gert Wollny <[email protected]> Reviewed-by: Erik Faye-Lund <[email protected]>
* virgl: fix undefined shift to use unsigned.Dave Airlie2018-11-291-1/+1
| | | | | | Ported from virglrenderer. Signed-off-by: Dave Airlie <[email protected]>
* virgl: Don't try handling server fences when they are not supportedGert Wollny2018-11-281-2/+4
| | | | | | | | | | | | | | | | | | | | | | | | vtest doesn't implement the according API and would segfault: Program received signal SIGSEGV, Segmentation fault. #0 0x0000000000000000 in ?? () #1 in virgl_fence_server_sync at src/gallium/drivers/virgl/virgl_context.c:1049 #2 in st_server_wait_sync at src/mesa/state_tracker/st_cb_syncobj.c:155 so just don't do the call when the function pointers are not set. Fixes dEQP: dEQP-GLES3.functional.fence_sync.wait_sync_smalldraw dEQP-GLES3.functional.fence_sync.wait_sync_largedraw Fixes: d1a1c21e7621b5177febf191fcd3d3b8ef69dc96 virgl: native fence fd support Signed-off-by: Gert Wollny <[email protected]> Reviewed-by: Dave Airlie <[email protected]> Reviewed-by: Robert Foss <[email protected]>
* virgl: add assert and missing function parameterRobert Foss2018-11-211-1/+4
| | | | | | | | | | Verify the pipe_fd_type to be of PIPE_FD_TYPE_NATIVE_SYNC. Fixes: d1a1c21e7621b5177feb "virgl: native fence fd support" Suggested-by: Eric Engestrom <[email protected]> Signed-off-by: Robert Foss <[email protected]> Reviewed-by: Emil Velikov <[email protected]>
* virgl: Clean up fences commitRobert Foss2018-11-181-1/+1
| | | | | | | | Remove a dead variable, a int->bool conversion and some whitespace changes. Signed-off-by: Robert Foss <[email protected]> Reviewed-by: Emil Velikov <[email protected]>
generated by cgit v1.2.3 (git 2.39.1) at 2024-12-28 04:36:36 +0000