summaryrefslogtreecommitdiffstats
Commit message (Collapse)AuthorAgeFilesLines
* st/nine: Enforce LOD 0 for D3DUSAGE_AUTOGENMIPMAPAxel Davy2015-04-292-3/+10
| | | | | | | | | For D3DUSAGE_AUTOGENMIPMAP textures, applications can only lock/copy from/get surface descriptor for/etc the first level. Thus it makes sense to restrict the LOD to 0, and use only the first level to generate the sublevels. Signed-off-by: Axel Davy <[email protected]>
* st/nine: Some D3DUSAGE_AUTOGENMIPMAP fixesAxel Davy2015-04-293-2/+17
| | | | Signed-off-by: Axel Davy <[email protected]>
* st/nine: util_gen_mipmap doesn't need we reset states.Axel Davy2015-04-291-2/+0
| | | | | | | util_gen_mipmap uses pipe->blit, and thus doesn't need we restore all states after using it. Signed-off-by: Axel Davy <[email protected]>
* st/nine: D3DUSAGE_AUTOGENMIPMAP is forbidden for volumesAxel Davy2015-04-292-16/+1
| | | | Signed-off-by: Axel Davy <[email protected]>
* st/nine: Fix NineBaseTexture9_PreLoadAxel Davy2015-04-291-1/+1
| | | | | | | It wasn't uploading the texture when the lod had changed. Signed-off-by: Axel Davy <[email protected]>
* st/nine: Rewrite Managed texture uploadsAxel Davy2015-04-291-96/+125
| | | | | | | | That part of the code was quite obscure. This new implementation tries to make it clearer by separating the differents parts, and commenting more. Signed-off-by: Axel Davy <[email protected]>
* st/nine: Bound the dirty regions to resource sizeAxel Davy2015-04-293-0/+16
| | | | Signed-off-by: Axel Davy <[email protected]>
* st/nine: Simplify Surface9 Managed resources implementationAxel Davy2015-04-294-93/+53
| | | | | | | | | | | Remove the Surface9 code for dirty rects, used only for Managed resources. Instead convey the information to the parent texture. According to documentation, this seems to be the expected behaviour, and if documentation is wrong there, that's not a problem since it can only leads to more texture updates in corner cases. Signed-off-by: Axel Davy <[email protected]>
* st/nine: Remove impossible cases with Managed texturesAxel Davy2015-04-292-34/+8
| | | | | | | Copying to/from a Managed texture is forbidden. Rendering to a Managed texture is forbidden. Signed-off-by: Axel Davy <[email protected]>
* st/nine: Encapsulate variables for MANAGED resourceAxel Davy2015-04-2910-57/+60
| | | | Signed-off-by: Axel Davy <[email protected]>
* st/nine: Rework texture data allocationAxel Davy2015-04-297-37/+131
| | | | | | | | | | | | Some applications assume the memory for multilevel textures is allocated per continuous blocks. This patch implements that behaviour. v2: cache offsets Reviewed-by: Ilia Mirkin <[email protected]> Signed-off-by: Axel Davy <[email protected]>
* st/nine: Fix update_vertex_elements bad rebaseAxel Davy2015-04-291-3/+0
| | | | | | | | This code was supposed to be removed, but a rebase seems to have made it stay. Reviewed-by: Ilia Mirkin <[email protected]> Signed-off-by: Axel Davy <[email protected]>
* st/nine: Add debug warning when application uses sw processingAxel Davy2015-04-296-7/+27
| | | | | Reviewed-by: Ilia Mirkin <[email protected]> Signed-off-by: Axel Davy <[email protected]>
* st/nine: Rework update_vertex_buffersAxel Davy2015-04-291-11/+4
| | | | | | | | | | | | | | | | Previous code was trying to optimise to call set_vertex_buffers on big packets, and thus avoids as many calls as possible. However in practice doing so won't be faster (drivers implement set_vertex_buffers by a loop over the buffers we want to bind) When we want to unbind a buffer, we were calling set_vertex_buffers on a buffer with vtxbuf->buffer = NULL. It works on some drivers, but not on all of them, because it isn't in Gallium spec. This patch fixes that. Reviewed-by: Ilia Mirkin <[email protected]> Signed-off-by: Axel Davy <[email protected]>
* st/nine: Fix computation of const_used_sizeXavier Bouchoux2015-04-291-3/+6
| | | | | | | Was sometimes too large for PS. Reviewed-by: Axel Davy <[email protected]> Signed-off-by: Xavier Bouchoux <[email protected]>
* gallium/svga: Remove useless ARRAY_SIZE declarationAxel Davy2015-04-291-4/+0
| | | | | | | This is already declared in util/macros.h Reviewed-by: Brian Paul <[email protected]> Signed-off-by: Axel Davy <[email protected]>
* util/macros: Move DIV_ROUND_UP to util/macros.hAxel Davy2015-04-293-6/+3
| | | | | | | Move DIV_ROUND_UP to a shared location accessible everywhere Reviewed-by: Brian Paul <[email protected]> Signed-off-by: Axel Davy <[email protected]>
* st/nine: Fix behaviour of D3DUSAGE_QUERY_POSTPIXELSHADER_BLENDINGXavier Bouchoux2015-04-291-1/+2
| | | | | | | | | | Ignore D3DUSAGE_QUERY_POSTPIXELSHADER_BLENDING when D3DUSAGE_RENDERTARGET is not specified. This behaviour matches windows drivers. Reviewed-by: Axel Davy <[email protected]> Signed-off-by: Xavier Bouchoux <[email protected]>
* st/nine: Improve D3DQUERYTYPE_TIMESTAMPXavier Bouchoux2015-04-291-2/+7
| | | | | | | | | | | Avoid blocking when retrieving D3DQUERYTYPE_TIMESTAMP result with NineQuery9_GetData(), when D3DGETDATA_FLUSH is not specified. This mimics Win behaviour and gives slightly better performance for some games. Reviewed-by: Ilia Mirkin <[email protected]> Reviewed-by: Axel Davy <[email protected]> Signed-off-by: Xavier Bouchoux <[email protected]>
* st/nine: Fix D3DQUERYTYPE_TIMESTAMPFREQ queryXavier Bouchoux2015-04-291-1/+9
| | | | | | | | | | | | | | D3DQUERYTYPE_TIMESTAMPFREQ is supposed to give the frequency at which the clock of D3DQUERYTYPE_TIMESTAMP runs. PIPE_QUERY_TIMESTAMP returns a value in ns, thus the corresponding frequency is 1000000000. PIPE_QUERY_TIMESTAMP_DISJOINT returns the frequency at which PIPE_QUERY_TIMESTAMP value is updated. It isn't always 1000000000. Reviewed-by: Axel Davy <[email protected]> Signed-off-by: Xavier Bouchoux <[email protected]>
* st/nine: Change x86 FPU Control word on device creationTiziano Bacocco2015-04-291-0/+31
| | | | | | | | | As on wined3d and windows, when D3DCREATE_FPU_PRESERVE is not specified, change the fpu control word to all exceptions masked, single precision, round to nearest. Signed-off-by: Axel Davy <[email protected]> Signed-off-by: Tiziano Bacocco <[email protected]>
* st/nine: Do not advertise D3DDEVCAPS_TEXTURESYSTEMMEMORYAxel Davy2015-04-292-9/+3
| | | | | | | No major vendor advertises it, and we weren't supporting it. Reviewed-by: Ilia Mirkin <[email protected]> Signed-off-by: Axel Davy <[email protected]>
* st/nine: Fix comment in update_viewportAxel Davy2015-04-291-3/+2
| | | | | Reviewed-by: Ilia Mirkin <[email protected]> Signed-off-by: Axel Davy <[email protected]>
* st/nine: Workaround barycentrics issue on some cardsAxel Davy2015-04-293-1/+31
| | | | Signed-off-by: Axel Davy <[email protected]>
* st/nine: Clear struct pipe_blit_info before use.Xavier Bouchoux2015-04-292-0/+4
| | | | | | | | render_condition_enable was uninitialized. Reviewed-by: Ilia Mirkin <[email protected]> Reviewed-by: Axel Davy <[email protected]> Signed-off-by: Xavier Bouchoux <[email protected]>
* st/nine: NineDevice9_Clear skip fastpath for bigger depth-buffersPatrick Rudolph2015-04-291-4/+13
| | | | | | | | | | | | This adds an additional check to make sure the bound depth-buffer doesn't exceed the rendertarget size when clearing depth and color buffer at once. D3D9 clears only a rectangle with the same dimensions as the viewport, leaving other parts of the depth-buffer intact. This fixes failing WINE test visual.c:depth_buffer_test() Signed-off-by: Patrick Rudolph <[email protected]> Signed-off-by: Axel Davy <[email protected]>
* st/nine: Fix wrong assert in nine_shaderAxel Davy2015-04-291-4/+4
| | | | | | | The sampler src index was wrong for texldl and texldd Reviewed-by: Ilia Mirkin <[email protected]> Signed-off-by: Axel Davy <[email protected]>
* st/nine: Handle special LIT caseAxel Davy2015-04-291-1/+18
| | | | | Reviewed-by: Ilia Mirkin <[email protected]> Signed-off-by: Axel Davy <[email protected]>
* mesa: Fix glGetProgramiv(GL_ACTIVE_ATTRIBUTES).Jose Fonseca2015-04-291-2/+4
| | | | | | | | | | | | It's returning random values, because RESOURCE_VAR() is casting different objects into ir_variable pointers. This updates _mesa_count_active_attribs to filter the resources with the same logic used in _mesa_longest_attribute_name_length. https://bugs.freedesktop.org/show_bug.cgi?id=90207 Reviewed-by: Tapani Pälli <[email protected]>
* egl: misc fixes for EGL_MESA_image_dma_buf_exportMarc-André Lureau2015-04-292-2/+2
| | | | | | | Fix define and a function argument name introduced in commit 8f7338f284cdb1fef64c85e3293d2200d0cc6387 Signed-off-by: Dave Airlie <[email protected]>
* nvc0/ir: flush denorms to zero in non-compute shadersIlia Mirkin2015-04-282-1/+25
| | | | | | | | | | | | This will set the FTZ flag (flush denorms to zero) on all opcodes that can take it. This resolves issues in Unigine Heaven 4.0 where there were solid-filled boxes popping up. Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=89455 Cc: "10.4 10.5" <[email protected]> Signed-off-by: Ilia Mirkin <[email protected]>
* meta: remove unneeded #include colortab.hBrian Paul2015-04-281-1/+0
| | | | Reviewed-by: Anuj Phogat <[email protected]>
* mesa: remove unneeded #include colortab.hBrian Paul2015-04-281-1/+0
| | | | Reviewed-by: Anuj Phogat <[email protected]>
* mesa: remove unused options var in compile_shader()Brian Paul2015-04-281-3/+0
| | | | Reviewed-by: Anuj Phogat <[email protected]>
* docs: more details about Viewperf 12 medical-01 test issuesBrian Paul2015-04-281-0/+7
|
* nvc0: expose GLSL version 410Ilia Mirkin2015-04-281-1/+1
| | | | Signed-off-by: Ilia Mirkin <[email protected]>
* st/mesa: allow glsl version up to 410, enable ARB_shader_precisionIlia Mirkin2015-04-281-2/+4
| | | | | Signed-off-by: Ilia Mirkin <[email protected]> Reviewed-by: Marek Olšák <[email protected]>
* st/va: add h264 decoder level supportLeo Liu2015-04-281-0/+5
| | | | | Signed-off-by: Leo Liu <[email protected]> Reviewed-by: Christian König <[email protected]>
* st/omx/dec: add h264 decoder level supportLeo Liu2015-04-281-3/+7
| | | | | | | v2: use sps level idc as level to driver Signed-off-by: Leo Liu <[email protected]> Reviewed-by: Christian König <[email protected]>
* vl: add level idc in spsLeo Liu2015-04-281-0/+1
| | | | | Signed-off-by: Leo Liu <[email protected]> Reviewed-by: Christian König <[email protected]>
* st/omx/dec: separate create_video_codec to different codecsLeo Liu2015-04-284-18/+30
| | | | | | | v2: get frame size from port info Signed-off-by: Leo Liu <[email protected]> Reviewed-by: Christian König <[email protected]>
* st/vdpau: add h264 decoder level supportLeo Liu2015-04-281-0/+5
| | | | | Signed-off-by: Leo Liu <[email protected]> Reviewed-by: Christian König <[email protected]>
* gallium/util: get h264 level based on number of max references and resolutionLeo Liu2015-04-281-0/+36
| | | | | | | | v2: add commments for limitation of max references numbers, and what the caculation is based Signed-off-by: Leo Liu <[email protected]> Reviewed-by: Christian König <[email protected]>
* r600g,radeonsi: add a driver query returning GPU loadMarek Olšák2015-04-288-2/+194
| | | | Reviewed-by: Alex Deucher <[email protected]>
* r600g,radeonsi: add driver queries for GPU temperature and shader+memory clocksMarek Olšák2015-04-285-3/+58
| | | | Reviewed-by: Alex Deucher <[email protected]>
* gm107/ir: add lane/vertex count sysvalsIlia Mirkin2015-04-271-0/+2
| | | | Signed-off-by: Ilia Mirkin <[email protected]>
* gk110/ir: add support for writing per-patch and shader outputsIlia Mirkin2015-04-271-7/+3
| | | | Signed-off-by: Ilia Mirkin <[email protected]>
* freedreno/a3xx: color masking works like a blend for some formatsIlia Mirkin2015-04-271-0/+14
| | | | | | | | When there is a colormask active that does not cover all the channels, enable reading in the destination like with a combining blend operation. This fixes fbo-blending-formats on a3xx. Signed-off-by: Ilia Mirkin <[email protected]>
* freedreno/a3xx: add support for S8 and Z32F_S8Ilia Mirkin2015-04-2710-44/+236
| | | | | | | | | Enables ARB_depth_buffer_float. There is no sampling support for interleaved Z32F_S8, so we store the two textures separately, one as Z32F, the other as S8. As a result, we need a lot of additional logic for restores and transfers. Signed-off-by: Ilia Mirkin <[email protected]>
* freedreno/a3xx: add Z32F supportIlia Mirkin2015-04-275-10/+65
| | | | | | | | | 32-bit depth buffers are stored as unorm, and thus need special handling when moving to and from gmem. They are copied into gmem by writing depth, and resolved from gmem using a special resolve bit which apparently float-ifies the data. Signed-off-by: Ilia Mirkin <[email protected]>
generated by cgit v1.2.3 (git 2.39.1) at 2024-12-23 05:30:15 +0000