diff options
author | Matt A. Tobin <email@mattatobin.com> | 2019-11-14 21:08:43 -0500 |
---|---|---|
committer | Matt A. Tobin <email@mattatobin.com> | 2019-11-14 21:08:43 -0500 |
commit | 1d30f6fa8413746ddc408f93710d701493af273d (patch) | |
tree | abe84e83d704e13c60c90db7ac4b9e363d8a81fc /modules/brotli/enc/hash_rolling_inc.h | |
parent | 9308ec68e863e4c6e650680370a5d7baa9f0d1f3 (diff) | |
parent | 00573571a226a0c59dd744da67483864a22911aa (diff) | |
download | UXP-1d30f6fa8413746ddc408f93710d701493af273d.tar UXP-1d30f6fa8413746ddc408f93710d701493af273d.tar.gz UXP-1d30f6fa8413746ddc408f93710d701493af273d.tar.lz UXP-1d30f6fa8413746ddc408f93710d701493af273d.tar.xz UXP-1d30f6fa8413746ddc408f93710d701493af273d.zip |
Merge branch 'master' into mailnews-work
Diffstat (limited to 'modules/brotli/enc/hash_rolling_inc.h')
-rw-r--r-- | modules/brotli/enc/hash_rolling_inc.h | 216 |
1 files changed, 216 insertions, 0 deletions
diff --git a/modules/brotli/enc/hash_rolling_inc.h b/modules/brotli/enc/hash_rolling_inc.h new file mode 100644 index 000000000..17f8a408e --- /dev/null +++ b/modules/brotli/enc/hash_rolling_inc.h @@ -0,0 +1,216 @@ +/* NOLINT(build/header_guard) */ +/* Copyright 2018 Google Inc. All Rights Reserved. + + Distributed under MIT license. + See file LICENSE for detail or copy at https://opensource.org/licenses/MIT +*/ + +/* template parameters: FN, JUMP, NUMBUCKETS, MASK, CHUNKLEN */ +/* NUMBUCKETS / (MASK + 1) = probability of storing and using hash code. */ +/* JUMP = skip bytes for speedup */ + +/* Rolling hash for long distance long string matches. Stores one position + per bucket, bucket key is computed over a long region. */ + +#define HashRolling HASHER() + +static const uint32_t FN(kRollingHashMul32) = 69069; +static const uint32_t FN(kInvalidPos) = 0xffffffff; + +/* This hasher uses a longer forward length, but returning a higher value here + will hurt compression by the main hasher when combined with a composite + hasher. The hasher tests for forward itself instead. */ +static BROTLI_INLINE size_t FN(HashTypeLength)(void) { return 4; } +static BROTLI_INLINE size_t FN(StoreLookahead)(void) { return 4; } + +/* Computes a code from a single byte. A lookup table of 256 values could be + used, but simply adding 1 works about as good. */ +static uint32_t FN(HashByte)(uint8_t byte) { + return (uint32_t)byte + 1u; +} + +static uint32_t FN(HashRollingFunctionInitial)(uint32_t state, uint8_t add, + uint32_t factor) { + return (uint32_t)(factor * state + FN(HashByte)(add)); +} + +static uint32_t FN(HashRollingFunction)(uint32_t state, uint8_t add, + uint8_t rem, uint32_t factor, + uint32_t factor_remove) { + return (uint32_t)(factor * state + + FN(HashByte)(add) - factor_remove * FN(HashByte)(rem)); +} + +typedef struct HashRolling { + uint32_t state; + uint32_t* table; + size_t next_ix; + + uint32_t chunk_len; + uint32_t factor; + uint32_t factor_remove; +} HashRolling; + +static BROTLI_INLINE HashRolling* FN(Self)(HasherHandle handle) { + return (HashRolling*)&(GetHasherCommon(handle)[1]); +} + +static void FN(Initialize)( + HasherHandle handle, const BrotliEncoderParams* params) { + HashRolling* self = FN(Self)(handle); + size_t i; + self->state = 0; + self->next_ix = 0; + + self->factor = FN(kRollingHashMul32); + + /* Compute the factor of the oldest byte to remove: factor**steps modulo + 0xffffffff (the multiplications rely on 32-bit overflow) */ + self->factor_remove = 1; + for (i = 0; i < CHUNKLEN; i += JUMP) { + self->factor_remove *= self->factor; + } + + self->table = (uint32_t*)((HasherHandle)self + sizeof(HashRolling)); + for (i = 0; i < NUMBUCKETS; i++) { + self->table[i] = FN(kInvalidPos); + } + + BROTLI_UNUSED(params); +} + +static void FN(Prepare)(HasherHandle handle, BROTLI_BOOL one_shot, + size_t input_size, const uint8_t* data) { + HashRolling* self = FN(Self)(handle); + size_t i; + /* Too small size, cannot use this hasher. */ + if (input_size < CHUNKLEN) return; + self->state = 0; + for (i = 0; i < CHUNKLEN; i += JUMP) { + self->state = FN(HashRollingFunctionInitial)( + self->state, data[i], self->factor); + } + BROTLI_UNUSED(one_shot); +} + +static BROTLI_INLINE size_t FN(HashMemAllocInBytes)( + const BrotliEncoderParams* params, BROTLI_BOOL one_shot, + size_t input_size) { + return sizeof(HashRolling) + NUMBUCKETS * sizeof(uint32_t); + BROTLI_UNUSED(params); + BROTLI_UNUSED(one_shot); + BROTLI_UNUSED(input_size); +} + +static BROTLI_INLINE void FN(Store)(HasherHandle BROTLI_RESTRICT handle, + const uint8_t* BROTLI_RESTRICT data, const size_t mask, const size_t ix) { + BROTLI_UNUSED(handle); + BROTLI_UNUSED(data); + BROTLI_UNUSED(mask); + BROTLI_UNUSED(ix); +} + +static BROTLI_INLINE void FN(StoreRange)(HasherHandle handle, + const uint8_t* data, const size_t mask, const size_t ix_start, + const size_t ix_end) { + BROTLI_UNUSED(handle); + BROTLI_UNUSED(data); + BROTLI_UNUSED(mask); + BROTLI_UNUSED(ix_start); + BROTLI_UNUSED(ix_end); +} + +static BROTLI_INLINE void FN(StitchToPreviousBlock)(HasherHandle handle, + size_t num_bytes, size_t position, const uint8_t* ringbuffer, + size_t ring_buffer_mask) { + /* In this case we must re-initialize the hasher from scratch from the + current position. */ + HashRolling* self = FN(Self)(handle); + size_t position_masked; + size_t available = num_bytes; + if ((position & (JUMP - 1)) != 0) { + size_t diff = JUMP - (position & (JUMP - 1)); + available = (diff > available) ? 0 : (available - diff); + position += diff; + } + position_masked = position & ring_buffer_mask; + /* wrapping around ringbuffer not handled. */ + if (available > ring_buffer_mask - position_masked) { + available = ring_buffer_mask - position_masked; + } + + FN(Prepare)(handle, BROTLI_FALSE, available, + ringbuffer + (position & ring_buffer_mask)); + self->next_ix = position; + BROTLI_UNUSED(num_bytes); +} + +static BROTLI_INLINE void FN(PrepareDistanceCache)( + HasherHandle handle, int* BROTLI_RESTRICT distance_cache) { + BROTLI_UNUSED(handle); + BROTLI_UNUSED(distance_cache); +} + +static BROTLI_INLINE void FN(FindLongestMatch)(HasherHandle handle, + const BrotliEncoderDictionary* dictionary, + const uint8_t* BROTLI_RESTRICT data, const size_t ring_buffer_mask, + const int* BROTLI_RESTRICT distance_cache, const size_t cur_ix, + const size_t max_length, const size_t max_backward, + const size_t gap, const size_t max_distance, + HasherSearchResult* BROTLI_RESTRICT out) { + HashRolling* self = FN(Self)(handle); + const size_t cur_ix_masked = cur_ix & ring_buffer_mask; + size_t pos = self->next_ix; + + if ((cur_ix & (JUMP - 1)) != 0) return; + + /* Not enough lookahead */ + if (max_length < CHUNKLEN) return; + + for (pos = self->next_ix; pos <= cur_ix; pos += JUMP) { + uint32_t code = self->state & MASK; + + uint8_t rem = data[pos & ring_buffer_mask]; + uint8_t add = data[(pos + CHUNKLEN) & ring_buffer_mask]; + size_t found_ix = FN(kInvalidPos); + + self->state = FN(HashRollingFunction)( + self->state, add, rem, self->factor, self->factor_remove); + + if (code < NUMBUCKETS) { + found_ix = self->table[code]; + self->table[code] = (uint32_t)pos; + if (pos == cur_ix && found_ix != FN(kInvalidPos)) { + /* The cast to 32-bit makes backward distances up to 4GB work even + if cur_ix is above 4GB, despite using 32-bit values in the table. */ + size_t backward = (uint32_t)(cur_ix - found_ix); + if (backward <= max_backward) { + const size_t found_ix_masked = found_ix & ring_buffer_mask; + const size_t len = FindMatchLengthWithLimit(&data[found_ix_masked], + &data[cur_ix_masked], + max_length); + if (len >= 4 && len > out->len) { + score_t score = BackwardReferenceScore(len, backward); + if (score > out->score) { + out->len = len; + out->distance = backward; + out->score = score; + out->len_code_delta = 0; + } + } + } + } + } + } + + self->next_ix = cur_ix + JUMP; + + /* NOTE: this hasher does not search in the dictionary. It is used as + backup-hasher, the main hasher already searches in it. */ + BROTLI_UNUSED(dictionary); + BROTLI_UNUSED(distance_cache); + BROTLI_UNUSED(gap); + BROTLI_UNUSED(max_distance); +} + +#undef HashRolling |