Fix a buffer overflow in the LZMA encoder. It was due to my

misunderstanding of the code. There's no tiny fix for this
problem, so I also cleaned up the code in general.

This reduces the speed of the encoder 2-5 % in the fastest
compression mode ("lzma -1"). High compression modes should
have no noticeable performance difference.

This commit breaks things (especially LZMA_SYNC_FLUSH) but I
will fix them once the new format and LZMA2 has been roughly
implemented. Plain LZMA won't support LZMA_SYNC_FLUSH at all
and won't be supported in the new .lzma format. This may
change still but this is what it looks like now.

Support for known uncompressed size (that is, LZMA or LZMA2
without EOPM) is likely to go away. This means there will
be API changes.

8 files changed, 532 insertions, 620 deletions

diff --git a/src/liblzma/lz/lz_encoder.c b/src/liblzma/lz/lz_encoder.c
index 03f8fa94..82b9103f 100644
--- a/src/liblzma/lz/lz_encoder.c
+++ b/src/liblzma/lz/lz_encoder.c
@@ -134,16 +134,13 @@ extern lzma_ret
 lzma_lz_encoder_reset(lzma_lz_encoder *lz, lzma_allocator *allocator,
 		bool (*process)(lzma_coder *coder, uint8_t *restrict out,
 			size_t *restrict out_pos, size_t out_size),
-		lzma_vli uncompressed_size,
 		size_t history_size, size_t additional_buffer_before,
 		size_t match_max_len, size_t additional_buffer_after,
 		lzma_match_finder match_finder, uint32_t match_finder_cycles,
 		const uint8_t *preset_dictionary,
 		size_t preset_dictionary_size)
 {
-	lz->sequence = SEQ_START;
-	lz->uncompressed_size = uncompressed_size;
-	lz->temp_size = 0;
+	lz->sequence = SEQ_RUN;
 
 	///////////////
 	// In Window //
@@ -395,10 +392,6 @@ fill_window(lzma_coder *coder, lzma_allocator *allocator, const uint8_t *in,
 		in_used = *in_pos - in_start;
 	}
 
-	assert(coder->lz.uncompressed_size >= in_used);
-	if (coder->lz.uncompressed_size != LZMA_VLI_VALUE_UNKNOWN)
-		coder->lz.uncompressed_size -= in_used;
-
 	// If end of stream has been reached or flushing completed, we allow
 	// the encoder to process all the input (that is, read_pos is allowed
 	// to reach write_pos). Otherwise we keep keep_size_after bytes
@@ -431,24 +424,6 @@ fill_window(lzma_coder *coder, lzma_allocator *allocator, const uint8_t *in,
 				- coder->lz.keep_size_after;
 	}
 
-	// Switch to finishing mode if we have got all the input data.
-	// lzma_lz_encode() won't return LZMA_STREAM_END until LZMA_FINISH
-	// is used.
-	//
-	// NOTE: When LZMA is used together with other filters, it is possible
-	// that coder->lz.sequence gets set to SEQ_FINISH before the next
-	// encoder has returned LZMA_STREAM_END. This is somewhat ugly, but
-	// works correctly, because the next encoder cannot have any more
-	// output left to be produced. If it had, then our known Uncompressed
-	// Size would be invalid, which would mean that we have a bad bug.
-// 	if (ret == LZMA_OK && coder->lz.uncompressed_size == 0)
-// 		coder->lz.sequence = SEQ_FINISH;
-	// The above breaks normal encoding with known uncompressed size
-	// if input chunk size is a multiple of uncompressed size. Commenting
-	// the above out breaks LZMA_SYNC_FLUSH at end of stream whose
-	// uncompressed size is known. Support for encoding with known
-	// uncompressed may get dropped completely so I won't fix this now.
-
 	// Restart the match finder after finished LZMA_SYNC_FLUSH.
 	if (coder->lz.pending > 0
 			&& coder->lz.read_pos < coder->lz.read_limit) {
@@ -475,67 +450,6 @@ lzma_lz_encode(lzma_coder *coder, lzma_allocator *allocator,
 		uint8_t *restrict out, size_t *restrict out_pos,
 		size_t out_size, lzma_action action)
 {
-	// Flush the temporary output buffer, which may be used when the
-	// encoder runs of out of space in primary output buffer (the out,
-	// *out_pos, and out_size variables).
-	if (coder->lz.temp_size > 0) {
-		const size_t out_avail = out_size - *out_pos;
-		if (out_avail < coder->lz.temp_size) {
-			// Cannot copy everything. Copy as much as possible
-			// and move the data in lz.temp to the beginning of
-			// that buffer.
-			memcpy(out + *out_pos, coder->lz.temp, out_avail);
-			*out_pos += out_avail;
-			memmove(coder->lz.temp, coder->lz.temp + out_avail,
-					coder->lz.temp_size - out_avail);
-			coder->lz.temp_size -= out_avail;
-			return LZMA_OK;
-		}
-
-		// We can copy everything from coder->lz.temp to out.
-		memcpy(out + *out_pos, coder->lz.temp, coder->lz.temp_size);
-		*out_pos += coder->lz.temp_size;
-		coder->lz.temp_size = 0;
-	}
-
-	switch (coder->lz.sequence) {
-	case SEQ_START:
-		assert(coder->lz.read_pos == coder->lz.write_pos);
-
-		// If there is no new input data and LZMA_SYNC_FLUSH is used
-		// immediatelly after previous LZMA_SYNC_FLUSH finished or
-		// at the very beginning of the input stream, we return
-		// LZMA_STREAM_END immediatelly. Writing a flush marker
-		// to the very beginning of the stream or right after previous
-		// flush marker is not allowed by the LZMA stream format.
-		if (*in_pos == in_size && action == LZMA_SYNC_FLUSH)
-			return LZMA_STREAM_END;
-
-		coder->lz.sequence = SEQ_RUN;
-		break;
-
-	case SEQ_FLUSH_END:
-		// During an earlier call to this function, flushing was
-		// otherwise finished except some data was left pending
-		// in coder->lz.buffer. Now we have copied all that data
-		// to the output buffer and can return LZMA_STREAM_END.
-		coder->lz.sequence = SEQ_START;
-		assert(action == LZMA_SYNC_FLUSH);
-		return LZMA_STREAM_END;
-
-	case SEQ_END:
-		// This is like the above flushing case, but for finishing
-		// the encoding.
-		//
-		// NOTE: action is not necesarily LZMA_FINISH; it can
-		// be LZMA_RUN or LZMA_SYNC_FLUSH too in case it is used
-		// at the end of the stream with known Uncompressed Size.
-		return action != LZMA_RUN ? LZMA_STREAM_END : LZMA_OK;
-
-	default:
-		break;
-	}
-
 	while (*out_pos < out_size
 			&& (*in_pos < in_size || action != LZMA_RUN)) {
 		// Read more data to coder->lz.buffer if needed.
@@ -546,27 +460,10 @@ lzma_lz_encode(lzma_coder *coder, lzma_allocator *allocator,
 
 		// Encode
 		if (coder->lz.process(coder, out, out_pos, out_size)) {
-			if (coder->lz.sequence == SEQ_FLUSH) {
-				assert(action == LZMA_SYNC_FLUSH);
-				if (coder->lz.temp_size == 0) {
-					// Flushing was finished successfully.
-					coder->lz.sequence = SEQ_START;
-				} else {
-					// Flushing was otherwise finished,
-					// except that some data was left
-					// into coder->lz.buffer.
-					coder->lz.sequence = SEQ_FLUSH_END;
-				}
-			} else {
-				// NOTE: action may be LZMA_RUN here in case
-				// Uncompressed Size is known and we have
-				// processed all the data already.
-				assert(coder->lz.sequence == SEQ_FINISH);
-				coder->lz.sequence = SEQ_END;
-			}
-
-			return action != LZMA_RUN && coder->lz.temp_size == 0
-					? LZMA_STREAM_END : LZMA_OK;
+			// Setting this to SEQ_RUN for cases when we are
+			// flushing. It doesn't matter when finishing.
+			coder->lz.sequence = SEQ_RUN;
+			return action != LZMA_RUN ? LZMA_STREAM_END : LZMA_OK;
 		}
 	}
 
diff --git a/src/liblzma/lz/lz_encoder.h b/src/liblzma/lz/lz_encoder.h
index b13e4b83..da0e0804 100644
--- a/src/liblzma/lz/lz_encoder.h
+++ b/src/liblzma/lz/lz_encoder.h
@@ -24,32 +24,19 @@
 #include "common.h"
 
 
-#define LZMA_LZ_TEMP_SIZE 64
-
-
 typedef struct lzma_lz_encoder_s lzma_lz_encoder;
 struct lzma_lz_encoder_s {
 	enum {
-		SEQ_START,
 		SEQ_RUN,
 		SEQ_FLUSH,
-		SEQ_FLUSH_END,
 		SEQ_FINISH,
-		SEQ_END
 	} sequence;
 
+	/// Function to do the actual encoding from the sliding input window
+	/// to the output stream.
 	bool (*process)(lzma_coder *coder, uint8_t *restrict out,
 			size_t *restrict out_pos, size_t out_size);
 
-	/// Uncompressed Size or LZMA_VLI_VALUE_UNKNOWN if using EOPM. We need
-	/// to track Uncompressed Size to prevent writing flush marker to the
-	/// very end of stream that doesn't use EOPM.
-	lzma_vli uncompressed_size;
-
-	/// Temporary buffer for range encoder.
-	uint8_t temp[LZMA_LZ_TEMP_SIZE];
-	size_t temp_size;
-
 	///////////////
 	// In Window //
 	///////////////
@@ -145,7 +132,6 @@ extern lzma_ret lzma_lz_encoder_reset(lzma_lz_encoder *lz,
 		lzma_allocator *allocator,
 		bool (*process)(lzma_coder *coder, uint8_t *restrict out,
 			size_t *restrict out_pos, size_t out_size),
-		lzma_vli uncompressed_size,
 		size_t history_size, size_t additional_buffer_before,
 		size_t match_max_len, size_t additional_buffer_after,
 		lzma_match_finder match_finder, uint32_t match_finder_cycles,
diff --git a/src/liblzma/lzma/lzma_encoder.c b/src/liblzma/lzma/lzma_encoder.c
index 4eee84f3..b0052182 100644
--- a/src/liblzma/lzma/lzma_encoder.c
+++ b/src/liblzma/lzma/lzma_encoder.c
@@ -26,113 +26,248 @@
 #include "fastpos.h"
 
 
-////////////
-// Macros //
-////////////
-
-// These are as macros mostly because they use local range encoder variables.
-
-#define literal_encode(subcoder, symbol) \
-do { \
-	uint32_t context = 1; \
-	int i = 8; \
-	do { \
-		--i; \
-		const uint32_t bit = ((symbol) >> i) & 1; \
-		bit_encode(subcoder[context], bit); \
-		context = (context << 1) | bit; \
-	} while (i != 0); \
-} while (0)
-
-
-#define literal_encode_matched(subcoder, match_byte, symbol) \
-do { \
-	uint32_t context = 1; \
-	int i = 8; \
-	do { \
-		--i; \
-		uint32_t bit = ((symbol) >> i) & 1; \
-		const uint32_t match_bit = ((match_byte) >> i) & 1; \
-		const uint32_t subcoder_index = 0x100 + (match_bit << 8) + context; \
-		bit_encode(subcoder[subcoder_index], bit); \
-		context = (context << 1) | bit; \
-		if (match_bit != bit) { \
-			while (i != 0) { \
-				--i; \
-				bit = ((symbol) >> i) & 1; \
-				bit_encode(subcoder[context], bit); \
-				context = (context << 1) | bit; \
-			} \
-			break; \
-		} \
-	} while (i != 0); \
-} while (0)
-
-
-#define length_encode(length_encoder, symbol, pos_state, update_price) \
-do { \
-	assert((symbol) <= MATCH_MAX_LEN); \
-	if ((symbol) < LEN_LOW_SYMBOLS) { \
-		bit_encode_0((length_encoder).choice); \
-		bittree_encode((length_encoder).low[pos_state], \
-				LEN_LOW_BITS, symbol); \
-	} else { \
-		bit_encode_1((length_encoder).choice); \
-		if ((symbol) < LEN_LOW_SYMBOLS + LEN_MID_SYMBOLS) { \
-			bit_encode_0((length_encoder).choice2); \
-			bittree_encode((length_encoder).mid[pos_state], \
-					LEN_MID_BITS, \
-					(symbol) - LEN_LOW_SYMBOLS); \
-		} else { \
-			bit_encode_1((length_encoder).choice2); \
-			bittree_encode((length_encoder).high, LEN_HIGH_BITS, \
-					(symbol) - LEN_LOW_SYMBOLS \
-					- LEN_MID_SYMBOLS); \
-		} \
-	} \
-	if (update_price) \
-		if (--(length_encoder).counters[pos_state] == 0) \
-			lzma_length_encoder_update_table(&(length_encoder), pos_state); \
-} while (0)
-
-
-///////////////
-// Functions //
-///////////////
-
-/// \brief      Updates price table of the length encoder
-///
-/// Like all the other prices in LZMA, these are used by lzma_get_optimum().
-///
-extern void
-lzma_length_encoder_update_table(lzma_length_encoder *lencoder,
-		const uint32_t pos_state)
+/////////////
+// Literal //
+/////////////
+
+static inline void
+literal_normal(lzma_range_encoder *rc, probability *subcoder, uint32_t symbol)
+{
+	uint32_t context = 1;
+	uint32_t bit_count = 8; // Bits per byte
+	do {
+		const uint32_t bit = (symbol >> --bit_count) & 1;
+		rc_bit(rc, &subcoder[context], bit);
+		context = (context << 1) | bit;
+	} while (bit_count != 0);
+}
+
+
+static inline void
+literal_matched(lzma_range_encoder *rc, probability *subcoder,
+		uint32_t match_byte, uint32_t symbol)
+{
+	uint32_t context = 1;
+	uint32_t bit_count = 8;
+
+	do {
+		uint32_t bit = (symbol >> --bit_count) & 1;
+		const uint32_t match_bit = (match_byte >> bit_count) & 1;
+		rc_bit(rc, &subcoder[(0x100 << match_bit) + context], bit);
+		context = (context << 1) | bit;
+
+		if (match_bit != bit) {
+			// The bit from the literal being encoded and the bit
+			// from the previous match differ. Finish encoding
+			// as a normal literal.
+			while (bit_count != 0) {
+				bit = (symbol >> --bit_count) & 1;
+				rc_bit(rc, &subcoder[context], bit);
+				context = (context << 1) | bit;
+			}
+
+			break;
+		}
+
+	} while (bit_count != 0);
+}
+
+
+static inline void
+literal(lzma_coder *coder)
+{
+	// Locate the literal byte to be encoded and the subcoder.
+	const uint8_t cur_byte = coder->lz.buffer[
+			coder->lz.read_pos - coder->additional_offset];
+	probability *subcoder = literal_get_subcoder(coder->literal_coder,
+			coder->now_pos, coder->previous_byte);
+
+	if (is_literal_state(coder->state)) {
+		// Previous LZMA-symbol was a literal. Encode a normal
+		// literal without a match byte.
+		literal_normal(&coder->rc, subcoder, cur_byte);
+	} else {
+		// Previous LZMA-symbol was a match. Use the last byte of
+		// the match as a "match byte". That is, compare the bits
+		// of the current literal and the match byte.
+		const uint8_t match_byte = coder->lz.buffer[
+				coder->lz.read_pos - coder->reps[0] - 1
+				- coder->additional_offset];
+		literal_matched(&coder->rc, subcoder, match_byte, cur_byte);
+	}
+
+	update_literal(coder->state);
+	coder->previous_byte = cur_byte;
+}
+
+
+//////////////////
+// Match length //
+//////////////////
+
+static inline void
+length(lzma_range_encoder *rc, lzma_length_encoder *lc,
+		const uint32_t pos_state, uint32_t len)
+{
+	assert(len <= MATCH_MAX_LEN);
+	len -= MATCH_MIN_LEN;
+
+	if (len < LEN_LOW_SYMBOLS) {
+		rc_bit(rc, &lc->choice, 0);
+		rc_bittree(rc, lc->low[pos_state], LEN_LOW_BITS, len);
+	} else {
+		rc_bit(rc, &lc->choice, 1);
+		len -= LEN_LOW_SYMBOLS;
+
+		if (len < LEN_MID_SYMBOLS) {
+			rc_bit(rc, &lc->choice2, 0);
+			rc_bittree(rc, lc->mid[pos_state], LEN_MID_BITS, len);
+		} else {
+			rc_bit(rc, &lc->choice2, 1);
+			len -= LEN_MID_SYMBOLS;
+			rc_bittree(rc, lc->high, LEN_HIGH_BITS, len);
+		}
+	}
+}
+
+
+///////////
+// Match //
+///////////
+
+static inline void
+match(lzma_coder *coder, const uint32_t pos_state,
+		const uint32_t distance, const uint32_t len)
+{
+	update_match(coder->state);
+
+	length(&coder->rc, &coder->match_len_encoder, pos_state, len);
+	coder->prev_len_encoder = &coder->match_len_encoder;
+
+	const uint32_t pos_slot = get_pos_slot(distance);
+	const uint32_t len_to_pos_state = get_len_to_pos_state(len);
+	rc_bittree(&coder->rc, coder->pos_slot_encoder[len_to_pos_state],
+			POS_SLOT_BITS, pos_slot);
+
+	if (pos_slot >= START_POS_MODEL_INDEX) {
+		const uint32_t footer_bits = (pos_slot >> 1) - 1;
+		const uint32_t base = (2 | (pos_slot & 1)) << footer_bits;
+		const uint32_t pos_reduced = distance - base;
+
+		if (pos_slot < END_POS_MODEL_INDEX) {
+			rc_bittree_reverse(&coder->rc,
+				&coder->pos_encoders[base - pos_slot - 1],
+				footer_bits, pos_reduced);
+		} else {
+			rc_direct(&coder->rc, pos_reduced >> ALIGN_BITS,
+					footer_bits - ALIGN_BITS);
+			rc_bittree_reverse(
+					&coder->rc, coder->pos_align_encoder,
+					ALIGN_BITS, pos_reduced & ALIGN_MASK);
+			++coder->align_price_count;
+		}
+	}
+
+	coder->reps[3] = coder->reps[2];
+	coder->reps[2] = coder->reps[1];
+	coder->reps[1] = coder->reps[0];
+	coder->reps[0] = distance;
+	++coder->match_price_count;
+}
+
+
+////////////////////
+// Repeated match //
+////////////////////
+
+static inline void
+rep_match(lzma_coder *coder, const uint32_t pos_state,
+		const uint32_t rep, const uint32_t len)
 {
-	const uint32_t num_symbols = lencoder->table_size;
-	const uint32_t a0 = bit_get_price_0(lencoder->choice);
-	const uint32_t a1 = bit_get_price_1(lencoder->choice);
-	const uint32_t b0 = a1 + bit_get_price_0(lencoder->choice2);
-	const uint32_t b1 = a1 + bit_get_price_1(lencoder->choice2);
+	if (rep == 0) {
+		rc_bit(&coder->rc, &coder->is_rep0[coder->state], 0);
+		rc_bit(&coder->rc,
+				&coder->is_rep0_long[coder->state][pos_state],
+				len != 1);
+	} else {
+		const uint32_t distance = coder->reps[rep];
+		rc_bit(&coder->rc, &coder->is_rep0[coder->state], 1);
+
+		if (rep == 1) {
+			rc_bit(&coder->rc, &coder->is_rep1[coder->state], 0);
+		} else {
+			rc_bit(&coder->rc, &coder->is_rep1[coder->state], 1);
+			rc_bit(&coder->rc, &coder->is_rep2[coder->state],
+					rep - 2);
+
+			if (rep == 3)
+				coder->reps[3] = coder->reps[2];
+
+			coder->reps[2] = coder->reps[1];
+		}
+
+		coder->reps[1] = coder->reps[0];
+		coder->reps[0] = distance;
+	}
+
+	if (len == 1) {
+		update_short_rep(coder->state);
+	} else {
+		length(&coder->rc, &coder->rep_len_encoder, pos_state, len);
+		coder->prev_len_encoder = &coder->rep_len_encoder;
+		update_long_rep(coder->state);
+	}
+}
+
 
-	uint32_t *prices = lencoder->prices[pos_state];
-	uint32_t i = 0;
+//////////
+// Main //
+//////////
 
-	for (i = 0; i < num_symbols && i < LEN_LOW_SYMBOLS; ++i)
-		prices[i] = a0 + bittree_get_price(lencoder->low[pos_state],
-				LEN_LOW_BITS, i);
+static void
+encode_symbol(lzma_coder *coder, uint32_t pos, uint32_t len)
+{
+	const uint32_t pos_state = coder->now_pos & coder->pos_mask;
+
+	if (len == 1 && pos == UINT32_MAX) {
+		// Literal i.e. eight-bit byte
+		rc_bit(&coder->rc,
+				&coder->is_match[coder->state][pos_state], 0);
+		literal(coder);
+	} else {
+		// Some type of match
+		rc_bit(&coder->rc,
+			&coder->is_match[coder->state][pos_state], 1);
+
+		if (pos < REP_DISTANCES) {
+			// It's a repeated match i.e. the same distance
+			// has been used earlier.
+			rc_bit(&coder->rc, &coder->is_rep[coder->state], 1);
+			rep_match(coder, pos_state, pos, len);
+		} else {
+			// Normal match
+			rc_bit(&coder->rc, &coder->is_rep[coder->state], 0);
+			match(coder, pos_state, pos - REP_DISTANCES, len);
+		}
 
-	for (; i < num_symbols && i < LEN_LOW_SYMBOLS + LEN_MID_SYMBOLS; ++i)
-		prices[i] = b0 + bittree_get_price(lencoder->mid[pos_state],
-				LEN_MID_BITS, i - LEN_LOW_SYMBOLS);
+		coder->previous_byte = coder->lz.buffer[
+				coder->lz.read_pos + len - 1
+				- coder->additional_offset];
+	}
 
-	for (; i < num_symbols; ++i)
-		prices[i] = b1 + bittree_get_price(
-				lencoder->high, LEN_HIGH_BITS,
-				i - LEN_LOW_SYMBOLS - LEN_MID_SYMBOLS);
+	assert(coder->additional_offset >= len);
+	coder->additional_offset -= len;
+	coder->now_pos += len;
+}
 
-	lencoder->counters[pos_state] = num_symbols;
 
-	return;
+static void
+encode_eopm(lzma_coder *coder)
+{
+	const uint32_t pos_state = coder->now_pos & coder->pos_mask;
+	rc_bit(&coder->rc, &coder->is_match[coder->state][pos_state], 1);
+	rc_bit(&coder->rc, &coder->is_rep[coder->state], 0);
+	match(coder, pos_state, UINT32_MAX, MATCH_MIN_LEN);
 }
 
 
@@ -145,15 +280,6 @@ extern bool
 lzma_lzma_encode(lzma_coder *coder, uint8_t *restrict out,
 		size_t *restrict out_pos, size_t out_size)
 {
-#define rc_buffer coder->lz.temp
-#define rc_buffer_size coder->lz.temp_size
-
-	// Local copies
-	lzma_range_encoder rc = coder->rc;
-	size_t out_pos_local = *out_pos;
-	const uint32_t pos_mask = coder->pos_mask;
-	const bool best_compression = coder->best_compression;
-
 	// Initialize the stream if no data has been encoded yet.
 	if (!coder->is_initialized) {
 		if (coder->lz.read_pos == coder->lz.read_limit) {
@@ -167,20 +293,10 @@ lzma_lzma_encode(lzma_coder *coder, uint8_t *restrict out,
 			// Do the actual initialization.
 			uint32_t len;
 			uint32_t num_distance_pairs;
-			lzma_read_match_distances(coder, &len, &num_distance_pairs);
-
-			bit_encode_0(coder->is_match[coder->state][0]);
-			update_literal(coder->state);
-
-			const uint8_t cur_byte = coder->lz.buffer[
-					coder->lz.read_pos - coder->additional_offset];
-			probability *subcoder = literal_get_subcoder(coder->literal_coder,
-					coder->now_pos, coder->previous_byte);
-			literal_encode(subcoder, cur_byte);
+			lzma_read_match_distances(coder,
+					&len, &num_distance_pairs);
 
-			coder->previous_byte = cur_byte;
-			--coder->additional_offset;
-			++coder->now_pos;
+			encode_symbol(coder, UINT32_MAX, 1);
 
 			assert(coder->additional_offset == 0);
 		}
@@ -191,19 +307,19 @@ lzma_lzma_encode(lzma_coder *coder, uint8_t *restrict out,
 
 	// Encoding loop
 	while (true) {
-		// Check that there is free output space.
-		if (out_pos_local == out_size)
-			break;
-
-		assert(rc_buffer_size == 0);
+		// Encode pending bits, if any.
+		if (rc_encode(&coder->rc, out, out_pos, out_size))
+			return false;
 
 		// Check that there is some input to process.
 		if (coder->lz.read_pos >= coder->lz.read_limit) {
 			// If flushing or finishing, we must keep encoding
 			// until additional_offset becomes zero to make
 			// all the input available at output.
-			if (coder->lz.sequence == SEQ_RUN
-					|| coder->additional_offset == 0)
+			if (coder->lz.sequence == SEQ_RUN)
+				return false;
+
+			if (coder->additional_offset == 0)
 				break;
 		}
 
@@ -218,8 +334,6 @@ lzma_lzma_encode(lzma_coder *coder, uint8_t *restrict out,
 		}
 #endif
 
-		const uint32_t pos_state = coder->now_pos & pos_mask;
-
 		uint32_t pos;
 		uint32_t len;
 
@@ -230,175 +344,32 @@ lzma_lzma_encode(lzma_coder *coder, uint8_t *restrict out,
 		//   - UINT32_MAX: not a match but a literal
 		// Value ranges for len:
 		//   - [MATCH_MIN_LEN, MATCH_MAX_LEN]
-		if (best_compression)
+		if (coder->best_compression)
 			lzma_get_optimum(coder, &pos, &len);
 		else
 			lzma_get_optimum_fast(coder, &pos, &len);
 
-		if (len == 1 && pos == UINT32_MAX) {
-			// It's a literal.
-			bit_encode_0(coder->is_match[coder->state][pos_state]);
-
-			const uint8_t cur_byte = coder->lz.buffer[
-					coder->lz.read_pos - coder->additional_offset];
-			probability *subcoder = literal_get_subcoder(coder->literal_coder,
-					coder->now_pos, coder->previous_byte);
-
-			if (is_literal_state(coder->state)) {
-				literal_encode(subcoder, cur_byte);
-			} else {
-				const uint8_t match_byte = coder->lz.buffer[
-						coder->lz.read_pos
-						- coder->rep_distances[0] - 1
-						- coder->additional_offset];
-				literal_encode_matched(subcoder, match_byte, cur_byte);
-			}
-
-			update_literal(coder->state);
-			coder->previous_byte = cur_byte;
-
-		} else {
-			// It's a match.
-			bit_encode_1(coder->is_match[coder->state][pos_state]);
-
-			if (pos < REP_DISTANCES) {
-				// It's a repeated match i.e. the same distance
-				// has been used earlier.
-				bit_encode_1(coder->is_rep[coder->state]);
-
-				if (pos == 0) {
-					bit_encode_0(coder->is_rep0[coder->state]);
-					const uint32_t symbol = (len == 1) ? 0 : 1;
-					bit_encode(coder->is_rep0_long[coder->state][pos_state],
-							symbol);
-				} else {
-					const uint32_t distance = coder->rep_distances[pos];
-					bit_encode_1(coder->is_rep0[coder->state]);
-
-					if (pos == 1) {
-						bit_encode_0(coder->is_rep1[coder->state]);
-					} else {
-						bit_encode_1(coder->is_rep1[coder->state]);
-						bit_encode(coder->is_rep2[coder->state], pos - 2);
-
-						if (pos == 3)
-							coder->rep_distances[3] = coder->rep_distances[2];
-
-						coder->rep_distances[2] = coder->rep_distances[1];
-					}
-
-					coder->rep_distances[1] = coder->rep_distances[0];
-					coder->rep_distances[0] = distance;
-				}
-
-				if (len == 1) {
-					update_short_rep(coder->state);
-				} else {
-					length_encode(coder->rep_len_encoder,
-							len - MATCH_MIN_LEN, pos_state,
-							best_compression);
-					update_long_rep(coder->state);
-				}
-
-			} else {
-				bit_encode_0(coder->is_rep[coder->state]);
-				update_match(coder->state);
-				length_encode(coder->match_len_encoder, len - MATCH_MIN_LEN,
-						pos_state, best_compression);
-				pos -= REP_DISTANCES;
-
-				const uint32_t pos_slot = get_pos_slot(pos);
-				const uint32_t len_to_pos_state = get_len_to_pos_state(len);
-				bittree_encode(coder->pos_slot_encoder[len_to_pos_state],
-						POS_SLOT_BITS, pos_slot);
-
-				if (pos_slot >= START_POS_MODEL_INDEX) {
-					const uint32_t footer_bits = (pos_slot >> 1) - 1;
-					const uint32_t base = (2 | (pos_slot & 1)) << footer_bits;
-					const uint32_t pos_reduced = pos - base;
-
-					if (pos_slot < END_POS_MODEL_INDEX) {
-						bittree_reverse_encode(
-								coder->pos_encoders + base - pos_slot - 1,
-								footer_bits, pos_reduced);
-					} else {
-						rc_encode_direct_bits(pos_reduced >> ALIGN_BITS,
-								footer_bits - ALIGN_BITS);
-						bittree_reverse_encode(coder->pos_align_encoder,
-								ALIGN_BITS, pos_reduced & ALIGN_MASK);
-						++coder->align_price_count;
-					}
-				}
-
-				coder->rep_distances[3] = coder->rep_distances[2];
-				coder->rep_distances[2] = coder->rep_distances[1];
-				coder->rep_distances[1] = coder->rep_distances[0];
-				coder->rep_distances[0] = pos;
-				++coder->match_price_count;
-			}
-
-			coder->previous_byte = coder->lz.buffer[
-					coder->lz.read_pos + len - 1
-					- coder->additional_offset];
-		}
-
-		assert(coder->additional_offset >= len);
-		coder->additional_offset -= len;
-		coder->now_pos += len;
+		encode_symbol(coder, pos, len);
 	}
 
-	// Check if everything is done.
-	bool all_done = false;
-	if (coder->lz.sequence != SEQ_RUN
-			&& coder->lz.read_pos == coder->lz.write_pos
-			&& coder->additional_offset == 0) {
-		assert(coder->longest_match_was_found == false);
-
-		if (coder->lz.uncompressed_size == LZMA_VLI_VALUE_UNKNOWN
-				|| coder->lz.sequence == SEQ_FLUSH) {
-			// Write special marker: flush marker or end of payload
-			// marker. Both are encoded as a match with distance of
-			// UINT32_MAX. The match length codes the type of the marker.
-			const uint32_t pos_state = coder->now_pos & pos_mask;
-			bit_encode_1(coder->is_match[coder->state][pos_state]);
-			bit_encode_0(coder->is_rep[coder->state]);
-			update_match(coder->state);
-
-			const uint32_t len = coder->lz.sequence == SEQ_FLUSH
-					? LEN_SPECIAL_FLUSH : LEN_SPECIAL_EOPM;
-			length_encode(coder->match_len_encoder, len - MATCH_MIN_LEN,
-					pos_state, best_compression);
-
-			const uint32_t pos_slot = (1 << POS_SLOT_BITS) - 1;
-			const uint32_t len_to_pos_state = get_len_to_pos_state(len);
-			bittree_encode(coder->pos_slot_encoder[len_to_pos_state],
-					POS_SLOT_BITS, pos_slot);
-
-			const uint32_t footer_bits = 30;
-			const uint32_t pos_reduced
-					= (UINT32_C(1) << footer_bits) - 1;
-			rc_encode_direct_bits(pos_reduced >> ALIGN_BITS,
-					footer_bits - ALIGN_BITS);
+	assert(!coder->longest_match_was_found);
 
-			bittree_reverse_encode(coder->pos_align_encoder, ALIGN_BITS,
-					pos_reduced & ALIGN_MASK);
-		}
+	if (coder->is_flushed) {
+		coder->is_flushed = false;
+		return true;
+	}
 
-		// Flush the last bytes of compressed data from
-		// the range coder to the output buffer.
-		rc_flush();
+	// We don't support encoding old LZMA streams without EOPM, and LZMA2
+	// doesn't use EOPM at LZMA level.
+	if (coder->write_eopm)
+		encode_eopm(coder);
 
-		rc_reset(rc);
+	rc_flush(&coder->rc);
 
-		// All done. Note that some output bytes might be
-		// pending in coder->lz.temp. lzma_lz_encode() will
-		// take care of those bytes.
-		all_done = true;
+	if (rc_encode(&coder->rc, out, out_pos, out_size)) {
+		coder->is_flushed = true;
+		return false;
 	}
 
-	// Store local variables back to *coder.
-	coder->rc = rc;
-	*out_pos = out_pos_local;
-
-	return all_done;
+	return true;
 }
diff --git a/src/liblzma/lzma/lzma_encoder_getoptimum.c b/src/liblzma/lzma/lzma_encoder_getoptimum.c
index 535508ee..b175e4cb 100644
--- a/src/liblzma/lzma/lzma_encoder_getoptimum.c
+++ b/src/liblzma/lzma/lzma_encoder_getoptimum.c
@@ -104,6 +104,36 @@ do { \
 
 
 static void
+fill_length_prices(lzma_length_encoder *lc, uint32_t pos_state)
+{
+	const uint32_t num_symbols = lc->table_size;
+	const uint32_t a0 = bit_get_price_0(lc->choice);
+	const uint32_t a1 = bit_get_price_1(lc->choice);
+	const uint32_t b0 = a1 + bit_get_price_0(lc->choice2);
+	const uint32_t b1 = a1 + bit_get_price_1(lc->choice2);
+
+	uint32_t *prices = lc->prices[pos_state];
+	uint32_t i = 0;
+
+	for (i = 0; i < num_symbols && i < LEN_LOW_SYMBOLS; ++i)
+		prices[i] = a0 + bittree_get_price(lc->low[pos_state],
+				LEN_LOW_BITS, i);
+
+	for (; i < num_symbols && i < LEN_LOW_SYMBOLS + LEN_MID_SYMBOLS; ++i)
+		prices[i] = b0 + bittree_get_price(lc->mid[pos_state],
+				LEN_MID_BITS, i - LEN_LOW_SYMBOLS);
+
+	for (; i < num_symbols; ++i)
+		prices[i] = b1 + bittree_get_price(lc->high, LEN_HIGH_BITS,
+				i - LEN_LOW_SYMBOLS - LEN_MID_SYMBOLS);
+
+	lc->counters[pos_state] = num_symbols;
+
+	return;
+}
+
+
+static void
 fill_distances_prices(lzma_coder *coder)
 {
 	uint32_t temp_prices[FULL_DISTANCES];
@@ -254,6 +284,9 @@ extern void
 lzma_get_optimum(lzma_coder *restrict coder,
 		uint32_t *restrict back_res, uint32_t *restrict len_res)
 {
+	uint32_t position = coder->now_pos;
+	uint32_t pos_state = position & coder->pos_mask;
+
 	// Update the price tables. In the C++ LZMA SDK 4.42 this was done in both
 	// initialization function and in the main loop. In liblzma they were
 	// moved into this single place.
@@ -265,6 +298,13 @@ lzma_get_optimum(lzma_coder *restrict coder,
 			fill_align_prices(coder);
 	}
 
+	if (coder->prev_len_encoder != NULL) {
+		if (--coder->prev_len_encoder->counters[pos_state] == 0)
+			fill_length_prices(coder->prev_len_encoder, pos_state);
+
+		coder->prev_len_encoder = NULL;
+	}
+
 
 	if (coder->optimum_end_index != coder->optimum_current_index) {
 		*len_res = coder->optimum[coder->optimum_current_index].pos_prev
@@ -312,7 +352,7 @@ lzma_get_optimum(lzma_coder *restrict coder,
 	uint32_t rep_max_index = 0;
 
 	for (uint32_t i = 0; i < REP_DISTANCES; ++i) {
-		reps[i] = coder->rep_distances[i];
+		reps[i] = coder->reps[i];
 		const uint32_t back_offset = reps[i] + 1;
 
 		if (buf[0] != *(buf - back_offset)
@@ -356,13 +396,8 @@ lzma_get_optimum(lzma_coder *restrict coder,
 		return;
 	}
 
-	const uint32_t pos_mask = coder->pos_mask;
-
 	coder->optimum[0].state = coder->state;
 
-	uint32_t position = coder->now_pos;
-	uint32_t pos_state = (position & pos_mask);
-
 	coder->optimum[1].price = bit_get_price_0(
 				coder->is_match[coder->state][pos_state])
 			+ literal_get_price(
@@ -575,7 +610,7 @@ lzma_get_optimum(lzma_coder *restrict coder,
 	current_byte = *buf;
 	match_byte = *(buf - reps[0] - 1);
 
-	pos_state = position & pos_mask;
+	pos_state = position & coder->pos_mask;
 
 	const uint32_t cur_and_1_price = cur_price
 			+ bit_get_price_0(coder->is_match[state][pos_state])
@@ -640,7 +675,7 @@ lzma_get_optimum(lzma_coder *restrict coder,
 			uint32_t state_2 = state;
 			update_literal(state_2);
 
-			const uint32_t pos_state_next = (position + 1) & pos_mask;
+			const uint32_t pos_state_next = (position + 1) & coder->pos_mask;
 			const uint32_t next_rep_match_price = cur_and_1_price
 					+ bit_get_price_1(coder->is_match[state_2][pos_state_next])
 					+ bit_get_price_1(coder->is_rep[state_2]);
@@ -719,7 +754,7 @@ lzma_get_optimum(lzma_coder *restrict coder,
 			uint32_t state_2 = state;
 			update_long_rep(state_2);
 
-			uint32_t pos_state_next = (position + len_test) & pos_mask;
+			uint32_t pos_state_next = (position + len_test) & coder->pos_mask;
 
 			const uint32_t cur_and_len_char_price = price
 					+ length_get_price(coder->rep_len_encoder,
@@ -732,7 +767,7 @@ lzma_get_optimum(lzma_coder *restrict coder,
 
 			update_literal(state_2);
 
-			pos_state_next = (position + len_test + 1) & pos_mask;
+			pos_state_next = (position + len_test + 1) & coder->pos_mask;
 
 			const uint32_t next_rep_match_price = cur_and_len_char_price
 					+ bit_get_price_1(coder->is_match[state_2][pos_state_next])
@@ -829,7 +864,7 @@ lzma_get_optimum(lzma_coder *restrict coder,
 					uint32_t state_2 = state;
 					update_match(state_2);
 					uint32_t pos_state_next
-							= (position + len_test) & pos_mask;
+							= (position + len_test) & coder->pos_mask;
 
 					const uint32_t cur_and_len_char_price = cur_and_len_price
 							+ bit_get_price_0(
@@ -844,7 +879,7 @@ lzma_get_optimum(lzma_coder *restrict coder,
 								buf[len_test]);
 
 					update_literal(state_2);
-					pos_state_next = (pos_state_next + 1) & pos_mask;
+					pos_state_next = (pos_state_next + 1) & coder->pos_mask;
 
 					const uint32_t next_rep_match_price
 							= cur_and_len_char_price
diff --git a/src/liblzma/lzma/lzma_encoder_getoptimumfast.c b/src/liblzma/lzma/lzma_encoder_getoptimumfast.c
index e6cee19d..fa06be21 100644
--- a/src/liblzma/lzma/lzma_encoder_getoptimumfast.c
+++ b/src/liblzma/lzma/lzma_encoder_getoptimumfast.c
@@ -65,7 +65,7 @@ lzma_get_optimum_fast(lzma_coder *restrict coder,
 	uint32_t rep_max_index = 0;
 
 	for (uint32_t i = 0; i < REP_DISTANCES; ++i) {
-		const uint32_t back_offset = coder->rep_distances[i] + 1;
+		const uint32_t back_offset = coder->reps[i] + 1;
 
 		// If the first two bytes (2 == MATCH_MIN_LEN) do not match,
 		// this rep_distance[i] is not useful. This is indicated
@@ -168,7 +168,7 @@ lzma_get_optimum_fast(lzma_coder *restrict coder,
 		--num_available_bytes;
 
 		for (uint32_t i = 0; i < REP_DISTANCES; ++i) {
-			const uint32_t back_offset = coder->rep_distances[i] + 1;
+			const uint32_t back_offset = coder->reps[i] + 1;
 
 			if (buf[1] != *(buf + 1 - back_offset)
 					|| buf[2] != *(buf + 2 - back_offset)) {
diff --git a/src/liblzma/lzma/lzma_encoder_init.c b/src/liblzma/lzma/lzma_encoder_init.c
index 0e3fb769..306cea8c 100644
--- a/src/liblzma/lzma/lzma_encoder_init.c
+++ b/src/liblzma/lzma/lzma_encoder_init.c
@@ -42,7 +42,7 @@ length_encoder_reset(lzma_length_encoder *lencoder,
 
 	// NLength::CPriceTableEncoder::UpdateTables()
 	for (size_t pos_state = 0; pos_state < num_pos_states; ++pos_state)
-		lzma_length_encoder_update_table(lencoder, pos_state);
+		lencoder->counters[pos_state] = 1;
 
 	return;
 }
@@ -112,7 +112,6 @@ lzma_lzma_encoder_init(lzma_next_coder *next, lzma_allocator *allocator,
 	{
 		const lzma_ret ret = lzma_lz_encoder_reset(
 				&next->coder->lz, allocator, &lzma_lzma_encode,
-				filters[0].uncompressed_size,
 				options->dictionary_size, OPTS,
 				options->fast_bytes, MATCH_MAX_LEN + 1 + OPTS,
 				options->match_finder,
@@ -143,13 +142,13 @@ lzma_lzma_encoder_init(lzma_next_coder *next, lzma_allocator *allocator,
 	next->coder->fast_bytes = options->fast_bytes;
 
 	// Range coder
-	rc_reset(next->coder->rc);
+	rc_reset(&next->coder->rc);
 
 	// State
 	next->coder->state = 0;
 	next->coder->previous_byte = 0;
 	for (size_t i = 0; i < REP_DISTANCES; ++i)
-		next->coder->rep_distances[i] = 0;
+		next->coder->reps[i] = 0;
 
 	// Bit encoders
 	for (size_t i = 0; i < STATES; ++i) {
@@ -190,6 +189,8 @@ lzma_lzma_encoder_init(lzma_next_coder *next, lzma_allocator *allocator,
 
 	next->coder->now_pos = 0;
 	next->coder->is_initialized = false;
+	next->coder->is_flushed = false,
+	next->coder->write_eopm = true;
 
 	// Initialize the next decoder in the chain, if any.
 	{
diff --git a/src/liblzma/lzma/lzma_encoder_private.h b/src/liblzma/lzma/lzma_encoder_private.h
index 0feaf26a..4159b468 100644
--- a/src/liblzma/lzma/lzma_encoder_private.h
+++ b/src/liblzma/lzma/lzma_encoder_private.h
@@ -26,14 +26,6 @@
 #include "lz_encoder.h"
 #include "range_encoder.h"
 
-// We need space for about two encoding loops, because there is no check
-// for available buffer space before end of payload marker gets written.
-// 2*26 bytes should be enough for this... but Lasse isn't very sure about
-// the exact value. 64 bytes certainly is enough. :-)
-#if LZMA_LZ_TEMP_SIZE < 64
-#	error LZMA_LZ_TEMP_SIZE is too small.
-#endif
-
 
 #define move_pos(num) \
 do { \
@@ -72,7 +64,7 @@ typedef struct {
 	uint32_t pos_prev;  // pos_next;
 	uint32_t back_prev;
 
-	uint32_t backs[4];
+	uint32_t backs[REP_DISTANCES];
 
 } lzma_optimal;
 
@@ -90,7 +82,7 @@ struct lzma_coder_s {
 	// State
 	lzma_lzma_state state;
 	uint8_t previous_byte;
-	uint32_t rep_distances[REP_DISTANCES];
+	uint32_t reps[REP_DISTANCES];
 
 	// Misc
 	uint32_t match_distances[MATCH_MAX_LEN * 2 + 2 + 1];
@@ -99,6 +91,8 @@ struct lzma_coder_s {
 	uint32_t now_pos; // Lowest 32 bits are enough here.
 	bool best_compression;           ///< True when LZMA_MODE_BEST is used
 	bool is_initialized;
+	bool is_flushed;
+	bool write_eopm;
 
 	// Literal encoder
 	lzma_literal_coder *literal_coder;
@@ -119,6 +113,7 @@ struct lzma_coder_s {
 	// Length encoders
 	lzma_length_encoder match_len_encoder;
 	lzma_length_encoder rep_len_encoder;
+	lzma_length_encoder *prev_len_encoder;
 
 	// Optimal
 	lzma_optimal optimum[OPTS];
diff --git a/src/liblzma/rangecoder/range_encoder.h b/src/liblzma/rangecoder/range_encoder.h
index b216e648..b156ee7f 100644
--- a/src/liblzma/rangecoder/range_encoder.h
+++ b/src/liblzma/rangecoder/range_encoder.h
@@ -4,7 +4,7 @@
 /// \brief      Range Encoder
 //
 //  Copyright (C) 1999-2006 Igor Pavlov
-//  Copyright (C) 2007 Lasse Collin
+//  Copyright (C) 2007-2008 Lasse Collin
 //
 //  This library is free software; you can redistribute it and/or
 //  modify it under the terms of the GNU Lesser General Public
@@ -24,14 +24,218 @@
 #include "range_common.h"
 
 
+/// Maximum number of symbols that can be put pending into lzma_range_encoder
+/// structure between calls to lzma_rc_encode(). For LZMA, 52+5 is enough
+/// (match with big distance and length followed by range encoder flush).
+#define RC_SYMBOLS_MAX 58
+
+
 typedef struct {
 	uint64_t low;
 	uint64_t cache_size;
 	uint32_t range;
 	uint8_t cache;
+
+	/// Number of symbols in the tables
+	size_t count;
+
+	/// rc_encode()'s position in the tables
+	size_t pos;
+
+	/// Symbols to encode
+	enum {
+		RC_BIT_0,
+		RC_BIT_1,
+		RC_DIRECT_0,
+		RC_DIRECT_1,
+		RC_FLUSH,
+	} symbols[RC_SYMBOLS_MAX];
+
+	/// Probabilities associated with RC_BIT_0 or RC_BIT_1
+	probability *probs[RC_SYMBOLS_MAX];
+
 } lzma_range_encoder;
 
 
+static inline void
+rc_reset(lzma_range_encoder *rc)
+{
+	rc->low = 0;
+	rc->cache_size = 1;
+	rc->range = UINT32_MAX;
+	rc->cache = 0;
+	rc->count = 0;
+	rc->pos = 0;
+}
+
+
+static inline void
+rc_bit(lzma_range_encoder *rc, probability *prob, uint32_t bit)
+{
+	rc->symbols[rc->count] = bit;
+	rc->probs[rc->count] = prob;
+	++rc->count;
+}
+
+
+static inline void
+rc_bittree(lzma_range_encoder *rc, probability *probs,
+		uint32_t bit_count, uint32_t symbol)
+{
+	uint32_t model_index = 1;
+
+	do {
+		const uint32_t bit = (symbol >> --bit_count) & 1;
+		rc_bit(rc, &probs[model_index], bit);
+		model_index = (model_index << 1) | bit;
+	} while (bit_count != 0);
+}
+
+
+static inline void
+rc_bittree_reverse(lzma_range_encoder *rc, probability *probs,
+		uint32_t bit_count, uint32_t symbol)
+{
+	uint32_t model_index = 1;
+
+	do {
+		const uint32_t bit = symbol & 1;
+		symbol >>= 1;
+		rc_bit(rc, &probs[model_index], bit);
+		model_index = (model_index << 1) | bit;
+	} while (--bit_count != 0);
+}
+
+
+static inline void
+rc_direct(lzma_range_encoder *rc,
+		uint32_t value, uint32_t bit_count)
+{
+	do {
+		rc->symbols[rc->count++]
+				= RC_DIRECT_0 + ((value >> --bit_count) & 1);
+	} while (bit_count != 0);
+}
+
+
+static inline void
+rc_flush(lzma_range_encoder *rc)
+{
+	for (size_t i = 0; i < 5; ++i)
+		rc->symbols[rc->count++] = RC_FLUSH;
+}
+
+
+static inline bool
+rc_shift_low(lzma_range_encoder *rc,
+		uint8_t *out, size_t *out_pos, size_t out_size)
+{
+	if ((uint32_t)(rc->low) < (uint32_t)(0xFF000000)
+			|| (uint32_t)(rc->low >> 32) != 0) {
+		do {
+			if (*out_pos == out_size)
+				return true;
+
+			out[*out_pos] = rc->cache + (uint8_t)(rc->low >> 32);
+			++*out_pos;
+			rc->cache = 0xFF;
+
+		} while (--rc->cache_size != 0);
+
+		rc->cache = (rc->low >> 24) & 0xFF;
+	}
+
+	++rc->cache_size;
+	rc->low = (rc->low & 0x00FFFFFF) << SHIFT_BITS;
+
+	return false;
+}
+
+
+static inline bool
+rc_encode(lzma_range_encoder *rc,
+		uint8_t *out, size_t *out_pos, size_t out_size)
+{
+	while (rc->pos < rc->count) {
+		// Normalize
+		if (rc->range < TOP_VALUE) {
+			if (rc_shift_low(rc, out, out_pos, out_size))
+				return true;
+
+			rc->range <<= SHIFT_BITS;
+		}
+
+		// Encode a bit
+		switch (rc->symbols[rc->pos]) {
+		case RC_BIT_0: {
+			probability prob = *rc->probs[rc->pos];
+			rc->range = (rc->range >> BIT_MODEL_TOTAL_BITS) * prob;
+			prob += (BIT_MODEL_TOTAL - prob) >> MOVE_BITS;
+			*rc->probs[rc->pos] = prob;
+			break;
+		}
+
+		case RC_BIT_1: {
+			probability prob = *rc->probs[rc->pos];
+			const uint32_t bound = prob
+					* (rc->range >> BIT_MODEL_TOTAL_BITS);
+			rc->low += bound;
+			rc->range -= bound;
+			prob -= prob >> MOVE_BITS;
+			*rc->probs[rc->pos] = prob;
+			break;
+		}
+
+		case RC_DIRECT_0:
+			rc->range >>= 1;
+			break;
+
+		case RC_DIRECT_1:
+			rc->range >>= 1;
+			rc->low += rc->range;
+			break;
+
+		case RC_FLUSH:
+			// Prevent further normalizations.
+			rc->range = UINT32_MAX;
+
+			// Flush the last five bytes (see rc_flush()).
+			do {
+				if (rc_shift_low(rc, out, out_pos, out_size))
+					return true;
+			} while (++rc->pos < rc->count);
+
+			// Reset the range encoder so we are ready to continue
+			// encoding if we weren't finishing the stream.
+			rc_reset(rc);
+			return false;
+
+		default:
+			assert(0);
+			break;
+		}
+
+		++rc->pos;
+	}
+
+	rc->count = 0;
+	rc->pos = 0;
+
+	return false;
+}
+
+
+static inline uint64_t
+rc_pending(const lzma_range_encoder *rc)
+{
+	return rc->cache_size + 5 - 1;
+}
+
+
+////////////
+// Prices //
+////////////
+
 #ifdef HAVE_SMALL
 /// Probability prices used by *_get_price() macros. This is initialized
 /// by lzma_rc_init() and is not modified later.
@@ -49,183 +253,6 @@ lzma_rc_prob_prices[BIT_MODEL_TOTAL >> MOVE_REDUCING_BITS];
 #endif
 
 
-/// Resets the range encoder structure.
-#define rc_reset(rc) \
-do { \
-	(rc).low = 0; \
-	(rc).range = UINT32_MAX; \
-	(rc).cache_size = 1; \
-	(rc).cache = 0; \
-} while (0)
-
-
-//////////////////
-// Bit encoding //
-//////////////////
-
-// These macros expect that the following variables are defined:
-//  - lzma_range_encoder rc;
-//  - uint8_t *out;
-//  - size_t out_pos_local;  // Local copy of *out_pos
-//  - size_t size_out;
-//
-// Macros pointing to these variables are also needed:
-//  - uint8_t rc_buffer[]; // Don't use a pointer, must be real array!
-//  - size_t rc_buffer_size;
-
-
-// Combined from NRangeCoder::CEncoder::Encode()
-// and NRangeCoder::CEncoder::UpdateModel().
-#define bit_encode(prob, symbol) \
-do { \
-	probability rc_prob = prob; \
-	const uint32_t rc_bound \
-			= (rc.range >> BIT_MODEL_TOTAL_BITS) * rc_prob; \
-	if ((symbol) == 0) { \
-		rc.range = rc_bound; \
-		rc_prob += (BIT_MODEL_TOTAL - rc_prob) >> MOVE_BITS; \
-	} else { \
-		rc.low += rc_bound; \
-		rc.range -= rc_bound; \
-		rc_prob -= rc_prob >> MOVE_BITS; \
-	} \
-	prob = rc_prob; \
-	rc_normalize(); \
-} while (0)
-
-
-// Optimized version of bit_encode(prob, 0)
-#define bit_encode_0(prob) \
-do { \
-	probability rc_prob = prob; \
-	rc.range = (rc.range >> BIT_MODEL_TOTAL_BITS) * rc_prob; \
-	rc_prob += (BIT_MODEL_TOTAL - rc_prob) >> MOVE_BITS; \
-	prob = rc_prob; \
-	rc_normalize(); \
-} while (0)
-
-
-// Optimized version of bit_encode(prob, 1)
-#define bit_encode_1(prob) \
-do { \
-	probability rc_prob = prob; \
-	const uint32_t rc_bound = (rc.range >> BIT_MODEL_TOTAL_BITS) \
-			* rc_prob; \
-	rc.low += rc_bound; \
-	rc.range -= rc_bound; \
-	rc_prob -= rc_prob >> MOVE_BITS; \
-	prob = rc_prob; \
-	rc_normalize(); \
-} while (0)
-
-
-///////////////////////
-// Bit tree encoding //
-///////////////////////
-
-#define bittree_encode(probs, bit_levels, symbol) \
-do { \
-	uint32_t model_index = 1; \
-	for (int32_t bit_index = bit_levels - 1; \
-			bit_index >= 0; --bit_index) { \
-		const uint32_t bit = ((symbol) >> bit_index) & 1; \
-		bit_encode((probs)[model_index], bit); \
-		model_index = (model_index << 1) | bit; \
-	} \
-} while (0)
-
-
-#define bittree_reverse_encode(probs, bit_levels, symbol) \
-do { \
-	uint32_t model_index = 1; \
-	for (uint32_t bit_index = 0; bit_index < bit_levels; ++bit_index) { \
-		const uint32_t bit = ((symbol) >> bit_index) & 1; \
-		bit_encode((probs)[model_index], bit); \
-		model_index = (model_index << 1) | bit; \
-	} \
-} while (0)
-
-
-/////////////////
-// Direct bits //
-/////////////////
-
-#define rc_encode_direct_bits(value, num_total_bits) \
-do { \
-	for (int32_t rc_i = (num_total_bits) - 1; rc_i >= 0; --rc_i) { \
-		rc.range >>= 1; \
-		if ((((value) >> rc_i) & 1) == 1) \
-			rc.low += rc.range; \
-		rc_normalize(); \
-	} \
-} while (0)
-
-
-//////////////////
-// Buffer "I/O" //
-//////////////////
-
-// Calls rc_shift_low() to write out a byte if needed.
-#define rc_normalize() \
-do { \
-	if (rc.range < TOP_VALUE) { \
-		rc.range <<= SHIFT_BITS; \
-		rc_shift_low(); \
-	} \
-} while (0)
-
-
-// Flushes all the pending output.
-#define rc_flush() \
-	for (int32_t rc_i = 0; rc_i < 5; ++rc_i) \
-		rc_shift_low()
-
-
-// Writes the compressed data to next_out.
-// TODO: Notation change?
-//       (uint32_t)(0xFF000000)  =>  ((uint32_t)(0xFF) << TOP_BITS)
-// TODO: Another notation change?
-//       rc.low = (uint32_t)(rc.low) << SHIFT_BITS;
-//       =>
-//       rc.low &= TOP_VALUE - 1;
-//       rc.low <<= SHIFT_BITS;
-#define rc_shift_low() \
-do { \
-	if ((uint32_t)(rc.low) < (uint32_t)(0xFF000000) \
-			|| (uint32_t)(rc.low >> 32) != 0) { \
-		uint8_t rc_temp = rc.cache; \
-		do { \
-			rc_write_byte(rc_temp + (uint8_t)(rc.low >> 32)); \
-			rc_temp = 0xFF; \
-		} while(--rc.cache_size != 0); \
-		rc.cache = (uint8_t)((uint32_t)(rc.low) >> 24); \
-	} \
-	++rc.cache_size; \
-	rc.low = (uint32_t)(rc.low) << SHIFT_BITS; \
-} while (0)
-
-
-// Write one byte of compressed data to *next_out. Updates out_pos_local.
-// If out_pos_local == out_size, the byte is appended to rc_buffer.
-#define rc_write_byte(b) \
-do { \
-	if (out_pos_local == out_size) { \
-		rc_buffer[rc_buffer_size++] = (uint8_t)(b); \
-		assert(rc_buffer_size < sizeof(rc_buffer)); \
-	} else { \
-		assert(rc_buffer_size == 0); \
-		out[out_pos_local++] = (uint8_t)(b); \
-	} \
-} while (0)
-
-
-//////////////////
-// Price macros //
-//////////////////
-
-// These macros expect that the following variables are defined:
-//  - uint32_t lzma_rc_prob_prices;
-
 #define bit_get_price(prob, symbol) \
 	lzma_rc_prob_prices[((((prob) - (symbol)) ^ (-(symbol))) \
 			& (BIT_MODEL_TOTAL - 1)) >> MOVE_REDUCING_BITS]