Updates to Brotli compression format, decoder and encoder

This commit contains a batch of changes that were made to the Brotli
compression algorithm in the last three weeks. Most important changes:

  * Added UTF8 context model for good text compression.
  * Simplified context modeling by having only 4 context modes.
  * Per-block context mode selection.
  * Faster backward copying and bit reading functions.
  * More efficient histogram coding.
  * Streaming support for the decoder and encoder.
This commit is contained in:
Zoltan Szabadka
2013-11-15 19:02:17 +01:00
parent 79e99afe46
commit 1571db36a9
23 changed files with 1647 additions and 870 deletions
+5 -3
View File
@@ -22,9 +22,11 @@
namespace brotli {
// Input: length of data, and the bytes.
// Output: estimate of how many bits the literal will take entropy coded.
void EstimateBitCostsForLiterals(size_t len, const uint8_t *data, float *cost);
// Estimates how many bits the literals in the interval [pos, pos + len) in the
// ringbuffer (data, mask) will take entropy coded and writes these estimates
// to the ringbuffer (cost, mask).
void EstimateBitCostsForLiterals(size_t pos, size_t len, size_t mask,
const uint8_t *data, float *cost);
} // namespace brotli