summaryrefslogtreecommitdiff
path: root/src
AgeCommit message (Collapse)Author
2006-11-01disable debug printfBrian Paul
2006-11-01disable debug printfsBrian Paul
2006-11-01Correct a little bug, radeon->glCtx is a pointer.Jerome Glisse
2006-11-01Remove deprecated GetBufferSize interface.Jerome Glisse
2006-11-01fix compiler warningsKeith Whitwell
2006-11-01Patch from <haihao.xiang@intel.com> -- pass program stringKeith Whitwell
notification callback through to tnl/ module, fixes glean crash.
2006-11-01merge the (rest of) texmem branchKeith Whitwell
2006-11-01Merge texmem-0-3-branch.Keith Whitwell
2006-11-01Add LIT instruction to fragment program.Jerome Glisse
2006-11-01Import texmem i915 driver to its new location as i915tex.Keith Whitwell
2006-11-01Add CMPH instruction to fragprogJerome Glisse
CMPH a0, a1, a2 -> if a2 > 0.5 return a1 else return a0 Guessed by examinating LIT instruction handling of FGLRX.
2006-11-01Patch from <haihao.xiang@intel.com> -- pass program stringKeith Whitwell
notification callback through to tnl/ module, fixes glean crash.
2006-11-01change DEBUG_FRAG conditionsBrian Paul
2006-11-01fix some bugs in interpolate_specular()Brian Paul
2006-11-01Added a few new ctx->Const. fields for shader-related limits.Brian Paul
2006-10-31make use of Parameters->StateFlags in _swrast_update_fragment_program()Brian Paul
2006-10-31minor clean-ups, simplifications and new commentsBrian Paul
2006-10-31fix MINGW32 problemsBrian Paul
2006-10-31We need to call _mesa_load_state_parameters() whenever a fragment programBrian Paul
is enabled and we've changed any GL state that might be referenced by a fragment program (such as state.fog.params). Do this during swrast validation, not in _swrast_exec_fragment_program().
2006-10-31signal _NEW_LIGHT in _tnl_copy_to_current()Brian Paul
2006-10-31In TEX instructions, use precomputed lambda only if using fragment.texcoord[k]Brian Paul
to sample texture[k]. Use zero otherwise. Not foolproof, but a good compromise.
2006-10-31For TEX instructions use lambda=0. When sampling from texture unit K weBrian Paul
were using the partial derivatives of texcoord[K] but the coordinate used for texture sampling may be something totally different (and texcoord[K] might not be a real texture coord at all). Net result was a bogus LOD is sometimes used, often resulting in using the smallest mipmap level (a constant color). Just use zero for now (undef LAMBDA_ZERO to override). Plus, some additional debug code.
2006-10-30Fix bug #8799.Ian Romanick
Properly resize bith the drawable and the readable in mgaUpdateRects. Eliminate the use of the deprecated GetBufferSize interface. Bump driver date.
2006-10-30Refactor mgaXMesaSetFrontClipRects and mgaXMesaSetBackClipRects.Ian Romanick
Combine mgaXMesaSetFrontClipRects and mgaXMesaSetBackClipRects into a single new function called mga_set_cliprects. This enables a small refactor in mgaDDDrawBuffer.
2006-10-30Get DRI drawable from framebuffer object.Ian Romanick
In the software rasterization fallbacks, get the __DRIdrawablePrivate pointer from the driRenderbuffer that is passed in. This eliminates the need for the mesa_drawable pointer in MGA context structure. This partially fixes bug #8799. wincopy no longer produces a black window for the destination window, but it does produce incorrect results if the destination window is resized.
2006-10-30code tweaks, remove old commentsBrian Paul
2006-10-30fix _mesa_problem() textBrian Paul
2006-10-30Move gl_vertex_program_machine struct out of mtypes.h and put into nvvertexec.h.Brian Paul
Massage nvvertexec.c code to work more like s_nvfragprog.c - another step toward unifying vertex/fragment program execution.
2006-10-30remove debug codeBrian Paul
2006-10-29Rather than always allocating 100 instructions, put instructions into aBrian Paul
temporary buffer, then copy into a malloc'd array of the correct size. According to Glean texCombine, we never need more than 21 instructions.
2006-10-29New _mesa_exec_vertex_state_program() function. Start of some re-org.Brian Paul
2006-10-29fix wrong inequalityBrian Paul
2006-10-29Change _mesa_init_instruction() to initialize an array of instructions.Brian Paul
2006-10-29Program instructions are initialized before parsing so no need to setBrian Paul
Abs, NegateAbs, CondMask and CondSwizzle fields to defaults here. CondMask was getting mis-set. Removing the incorrect assignment improves fragment program speed by 15%.
2006-10-28Added PROGRAM_CONSTANT to get_register_pointer().Brian Paul
More DEBUG_FRAG code.
2006-10-28special case END in _mesa_print_instruction()Brian Paul
2006-10-28Implement _mesa_append_fog_code() for fragment program fog options.Brian Paul
2006-10-26If DEBUG, check that all array indices really do fall in [start,end] inBrian Paul
the DrawRangeElements() call. Warn the user if that's not true.
2006-10-26remove code for inserting mvp transform into position-invariant vertex progs ↵Roland Scheidegger
and instead use _mesa_insert_mvp_code().
2006-10-25Previously, if a fragment program referenced fragment.fogcoord but theBrian Paul
program didn't use the ARB_fog_linear option, the fragment program's fragment.fogcoord register wasn't loaded properly.
2006-10-25change some bogus comments regarding the base e exponential function used ↵Roland Scheidegger
for fog in vertex programs.
2006-10-24fix (per-vertex) fog when using ARB_vp by incorporating fog factor ↵Roland Scheidegger
computation into the vertex program (not yet fixed for swtnl). Simplify (and correct) the VTX_TCL_OUTPUT_VTXFMT handling when using vertex programs, turns out it's solely driven by the needs of the past-vertex stage of the pipeline, this should fix lockups with ill-specified applications using vertex programs (for instance applications enabling fog but not writing to fog coord output will now get (conformant) undefined results instead of lockups).
2006-10-24s/GLchan/GLubyte/Brian Paul
2006-10-24whitespace clean-upBrian Paul
2006-10-23Emit cliprects in the userspace driver as required, rather thanKeith Whitwell
passing them to the kernel. This works because all drawing commands in the 965 driver are emitted with the lock held and the batchbuffer is always flushed prior to releasing the lock. This allows multiple cliprects to be dealt with, without replaying entire batchbuffers and redundantly re-emitting state.
2006-10-22assorted clean-upsBrian Paul
2006-10-22don't save color/z if fragment is killedBrian Paul
2006-10-22minor tweaks good for about 12% faster executionBrian Paul
2006-10-22silence used var warningsBrian Paul
2006-10-22Color clamping fixes.Brian Paul
generated by cgit v1.2.3 (git 2.39.1) at 2024-11-09 05:10:17 +0000