Skip to content
Snippets Groups Projects
Select Git revision
  • master default
  • player-without-audio
  • release-1.2.0-ptalarbvorm
  • theora-1.0
  • theora-1.1
  • theora-gumboot
  • theora-monty-post-1-1
  • theora-multithread
  • theora-oil
  • theora-playtime
  • theora_multithread_decode_omp
  • theora_multithread_decode_pthread
  • theorarm-merge-branch
  • windows-ci
  • v1.2.0
  • v1.2.0beta1
  • v1.0
  • v1.2.0alpha1
  • v1.1.0
  • v1.1.1
20 results
You can move around the graph by using the arrow keys.
Created with Raphaël 2.2.01Aug23Jul18171614111097129Jun282726252422211715827May2211864329Apr2524221917161514131211108543228Mar27262524201914131211729Feb271513321Jan191716109865431Dec30272622181165330Nov292827262221201615731Oct30292826252318131211109643228Sep2221201816151492128Aug272524231026Jul232213109519Jun28May1410930Apr171615131229Jun1211987432130May282726107654327Apr2620161510529Mar282221725Feb24232220181320Jan1917121023Dec181776543128Nov23154227Oct2621206529Sep287525Aug242120196529Jul262524231211728Jun27222018146422May96125Apr7Mar43227Feb26122Jan19116529Dec151423Nov191727Oct22130Sep26241627AugCheck for an uninitialized state before dereferencing in propagating Commit some minor motion compensation search tweaks.Minor chroma-counting bugfix (affects bit estimation, not correctness)Add a dead zone to DC quant as a hysteresis measure as we can'tThis version of the pipelined multithread decoder with openMP uses a generic value for MCU sizeA pipelined implementationSome bug fixesA faster version with greater MCU sizeCommit first cut of per-token optimization. Like SKIP blocks, it does not yet automatically choose its lambda.correct an idiot bug in token log trackingActivate tokenize/rollback code. Not yet used for intended purpose, but it is working.Add a first cut of [only partially tested] tokenization rollbackIn-progress code cleanup in tokenization; make it a little easier to grok before diving into per-token opt which will render it nicely opaque again.The opemMP parallel implementationThe Ralph pthread implementation merged to new versiontheora_multithr…theora_multithread_decode_pthread remote_mirror_3c0a8dd8d474c1a086b0cebe7b503994/theora_multithread_decode_pthreadfull-frame dct coeff storage array eliminated. Tokenization is Rearrange tokenization to be inline with quantizationRefactoring of tokenization process to allow coding DC last, such that Correct an mv choice correctness bug in 4mv->1mv mode demotionLoop filter fixAlso update 32 bit MMX with loop filter fixCorrect an introduced bug that cam in the with the new loop filter asm;Many bugs fixed, still busted DON'T USE ITCommit for distributed debugging. THIS VERSION IS BROKEN: DON'T USE ITMore tokenization reflow to move toward T/Q/T in one passComplete rewrite of the DFarray code; it's now tracked and coded as part of the main encode loop so that the bit usage information of the block runs is available to block coding decisions.Move some more of the skip cleanup details out of the main encoding loop and into the luma MB coding specific calls.Rearrange the master encode loop to consider chroma in mode determination, but to perform quant/encode in strict sequential-plane superblock order. Preparation for transform/quant/code in one step.Eliminate use of SSE instruction in loop filter MMX code by replacing Staging a fix for mainline Theora: refactor code slightly to use the MMX add export symbols flag for mingw build fixes #1383Include <stddef.h> in theora[enc|dec].h directly in order to get size_t.Take MB coding overhead into account when evaluating blocks/MBs for Code rearrangement for SB frarray encoding.Get initial RD-based skip determination into SVN. Lambda currently Merge x86 assembly for forward DCT from Thusnelda branch; an attempt to fix Set eos properly in png2theora.Remove trailing whitespace.Actually need the round-toward-zero behavior for reencode.Update Xcode build files.