/*
- * lzx-common.c - Common data for LZX compression and decompression.
+ * lzx-common.c - Common code for LZX compression and decompression.
*/
/*
- * Copyright (C) 2012, 2013 Eric Biggers
+ * Copyright (C) 2012, 2013, 2014 Eric Biggers
*
- * This file is part of wimlib, a library for working with WIM files.
+ * This file is free software; you can redistribute it and/or modify it under
+ * the terms of the GNU Lesser General Public License as published by the Free
+ * Software Foundation; either version 3 of the License, or (at your option) any
+ * later version.
*
- * wimlib is free software; you can redistribute it and/or modify it under the
- * terms of the GNU General Public License as published by the Free
- * Software Foundation; either version 3 of the License, or (at your option)
- * any later version.
- *
- * wimlib is distributed in the hope that it will be useful, but WITHOUT ANY
- * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
- * A PARTICULAR PURPOSE. See the GNU General Public License for more
+ * This file is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+ * FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more
* details.
*
- * You should have received a copy of the GNU General Public License
- * along with wimlib; if not, see http://www.gnu.org/licenses/.
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this file; if not, see http://www.gnu.org/licenses/.
*/
#ifdef HAVE_CONFIG_H
# include "config.h"
#endif
+#include <string.h>
+
+#include "wimlib/bitops.h"
#include "wimlib/endianness.h"
#include "wimlib/lzx.h"
+#include "wimlib/unaligned.h"
#include "wimlib/util.h"
#ifdef __SSE2__
# include <emmintrin.h>
#endif
-/* Mapping: position slot => first match offset that uses that position slot.
+/* Mapping: offset slot => first match offset that uses that offset slot.
*/
-const u32 lzx_position_base[LZX_MAX_POSITION_SLOTS] = {
+const u32 lzx_offset_slot_base[LZX_MAX_OFFSET_SLOTS] = {
0 , 1 , 2 , 3 , 4 , /* 0 --- 4 */
6 , 8 , 12 , 16 , 24 , /* 5 --- 9 */
32 , 48 , 64 , 96 , 128 , /* 10 --- 14 */
2097152 /* 50 */
};
-/* Mapping: position slot => how many extra bits must be read and added to the
- * corresponding position base to decode the match offset. */
-#ifdef USE_LZX_EXTRA_BITS_ARRAY
-const u8 lzx_extra_bits[LZX_MAX_POSITION_SLOTS] = {
+/* Mapping: offset slot => how many extra bits must be read and added to the
+ * corresponding offset slot base to decode the match offset. */
+const u8 lzx_extra_offset_bits[LZX_MAX_OFFSET_SLOTS] = {
0 , 0 , 0 , 0 , 1 ,
1 , 2 , 2 , 3 , 3 ,
4 , 4 , 5 , 5 , 6 ,
17, 17, 17, 17, 17,
17
};
-#endif
/* Round the specified compression block size (not LZX block size) up to the
* next valid LZX window size, and return its order (log2). Or, if the block
{
unsigned order;
- if (max_block_size == 0 || max_block_size > (1 << LZX_MAX_WINDOW_ORDER))
+ if (max_block_size == 0 || max_block_size > LZX_MAX_WINDOW_SIZE)
return 0;
- order = bsr32(max_block_size);
+ order = fls32(max_block_size);
- if ((1 << order) != max_block_size)
+ if (((u32)1 << order) != max_block_size)
order++;
return max(order, LZX_MIN_WINDOW_ORDER);
unsigned
lzx_get_num_main_syms(unsigned window_order)
{
- u32 window_size = 1 << window_order;
+ u32 window_size = (u32)1 << window_order;
/* NOTE: the calculation *should* be as follows:
*
* u32 max_offset = window_size - LZX_MIN_MATCH_LEN;
- * u32 max_formatted_offset = max_offset + LZX_OFFSET_OFFSET;
- * u32 num_position_slots = 1 + lzx_get_position_slot_raw(max_formatted_offset);
+ * u32 max_adjusted_offset = max_offset + LZX_OFFSET_OFFSET;
+ * u32 num_offset_slots = 1 + lzx_get_offset_slot_raw(max_adjusted_offset);
*
* However since LZX_MIN_MATCH_LEN == LZX_OFFSET_OFFSET, we would get
- * max_formatted_offset == window_size, which would bump the number of
- * position slots up by 1 since every valid LZX window size is equal to
- * a position base value. The format doesn't do this, and instead
+ * max_adjusted_offset == window_size, which would bump the number of
+ * offset slots up by 1 since every valid LZX window size is equal to a
+ * offset slot base value. The format doesn't do this, and instead
* disallows matches with minimum length and maximum offset. This sets
- * max_formatted_offset = window_size - 1, so instead we must calculate:
+ * max_adjusted_offset = window_size - 1, so instead we must calculate:
*
- * num_position_slots = 1 + lzx_get_position_slot_raw(window_size - 1);
+ * num_offset_slots = 1 + lzx_get_offset_slot_raw(window_size - 1);
*
* ... which is the same as
*
- * num_position_slots = lzx_get_position_slot_raw(window_size);
+ * num_offset_slots = lzx_get_offset_slot_raw(window_size);
*
- * ... since every valid window size is equal to a position base value.
+ * ... since every valid window size is equal to an offset base value.
*/
- unsigned num_position_slots = lzx_get_position_slot_raw(window_size);
+ unsigned num_offset_slots = lzx_get_offset_slot_raw(window_size);
/* Now calculate the number of main symbols as LZX_NUM_CHARS literal
- * symbols, plus 8 symbols per position slot (since there are 8 possible
- * length headers, and we need all (position slot, length header)
+ * symbols, plus 8 symbols per offset slot (since there are 8 possible
+ * length headers, and we need all (offset slot, length header)
* combinations). */
- return LZX_NUM_CHARS + (num_position_slots << 3);
+ return LZX_NUM_CHARS + (num_offset_slots << 3);
}
static void
-do_translate_target(sle32 *target, s32 input_pos)
+do_translate_target(void *target, s32 input_pos)
{
s32 abs_offset, rel_offset;
- /* XXX: This assumes unaligned memory accesses are okay. */
- rel_offset = le32_to_cpu(*target);
+ rel_offset = get_unaligned_u32_le(target);
if (rel_offset >= -input_pos && rel_offset < LZX_WIM_MAGIC_FILESIZE) {
if (rel_offset < LZX_WIM_MAGIC_FILESIZE - input_pos) {
/* "good translation" */
/* "compensating translation" */
abs_offset = rel_offset - LZX_WIM_MAGIC_FILESIZE;
}
- *target = cpu_to_le32(abs_offset);
+ put_unaligned_u32_le(abs_offset, target);
}
}
static void
-undo_translate_target(sle32 *target, s32 input_pos)
+undo_translate_target(void *target, s32 input_pos)
{
s32 abs_offset, rel_offset;
- /* XXX: This assumes unaligned memory accesses are okay. */
- abs_offset = le32_to_cpu(*target);
+ abs_offset = get_unaligned_u32_le(target);
if (abs_offset >= 0) {
if (abs_offset < LZX_WIM_MAGIC_FILESIZE) {
/* "good translation" */
rel_offset = abs_offset - input_pos;
-
- *target = cpu_to_le32(rel_offset);
+ put_unaligned_u32_le(rel_offset, target);
}
} else {
if (abs_offset >= -input_pos) {
/* "compensating translation" */
rel_offset = abs_offset + LZX_WIM_MAGIC_FILESIZE;
-
- *target = cpu_to_le32(rel_offset);
+ put_unaligned_u32_le(rel_offset, target);
}
}
}
* in calculating the translated jump targets. But in WIM files, this file size
* is always the same (LZX_WIM_MAGIC_FILESIZE == 12000000).
*/
-static
-#ifndef __SSE2__
-inline /* Although inlining the 'process_target' function still speeds up the
- SSE2 case, it bloats the binary more. */
-#endif
-void
-lzx_e8_filter(u8 *data, u32 size, void (*process_target)(sle32 *, s32))
+static void
+lzx_e8_filter(u8 *data, u32 size, void (*process_target)(void *, s32))
{
-#ifdef __SSE2__
- /* SSE2 vectorized implementation for x86_64. This speeds up LZX
- * decompression by about 5-8% overall. (Usually --- the performance
- * actually regresses slightly in the degenerate case that the data
- * consists entirely of 0xe8 bytes. Also, this optimization affects
- * compression as well, but the percentage improvement is less because
- * LZX compression is much slower than LZX decompression. ) */
- __m128i *p128 = (__m128i *)data;
- u32 valid_mask = 0xFFFFFFFF;
-
- if (size >= 32 && (uintptr_t)data % 16 == 0) {
- __m128i * const end128 = p128 + size / 16 - 1;
-
- /* Create a vector of all 0xe8 bytes */
- const __m128i e8_bytes = _mm_set1_epi8(0xe8);
-
- /* Iterate through the 16-byte vectors in the input. */
- do {
- /* Compare the current 16-byte vector with the vector of
- * all 0xe8 bytes. This produces 0xff where the byte is
- * 0xe8 and 0x00 where it is not. */
- __m128i cmpresult = _mm_cmpeq_epi8(*p128, e8_bytes);
-
- /* Map the comparison results into a single 16-bit
- * number. It will contain a 1 bit when the
- * corresponding byte in the current 16-byte vector is
- * an e8 byte. Note: the low-order bit corresponds to
- * the first (lowest address) byte. */
- u32 e8_mask = _mm_movemask_epi8(cmpresult);
-
- if (!e8_mask) {
- /* If e8_mask is 0, then none of these 16 bytes
- * have value 0xe8. No e8 translation is
- * needed, and there is no restriction that
- * carries over to the next 16 bytes. */
- valid_mask = 0xFFFFFFFF;
- } else {
- /* At least one byte has value 0xe8.
- *
- * The AND with valid_mask accounts for the fact
- * that we can't start an e8 translation that
- * overlaps the previous one. */
- while ((e8_mask &= valid_mask)) {
-
- /* Count the number of trailing zeroes
- * in e8_mask. This will produce the
- * index of the byte, within the 16, at
- * which the next e8 translation should
- * be done. */
- u32 bit = __builtin_ctz(e8_mask);
-
- /* Do (or undo) the e8 translation. */
- u8 *p8 = (u8 *)p128 + bit;
- (*process_target)((sle32 *)(p8 + 1),
- p8 - data);
-
- /* Don't start an e8 translation in the
- * next 4 bytes. */
- valid_mask &= ~((u32)0x1F << bit);
- }
- /* Moving on to the next vector. Shift and set
- * valid_mask accordingly. */
- valid_mask >>= 16;
- valid_mask |= 0xFFFF0000;
- }
- } while (++p128 < end128);
+
+#if !defined(__SSE2__) && !defined(__AVX2__)
+ /*
+ * A worthwhile optimization is to push the end-of-buffer check into the
+ * relatively rare E8 case. This is possible if we replace the last six
+ * bytes of data with E8 bytes; then we are guaranteed to hit an E8 byte
+ * before reaching end-of-buffer. In addition, this scheme guarantees
+ * that no translation can begin following an E8 byte in the last 10
+ * bytes because a 4-byte offset containing E8 as its high byte is a
+ * large negative number that is not valid for translation. That is
+ * exactly what we need.
+ */
+ u8 *tail;
+ u8 saved_bytes[6];
+ u8 *p;
+
+ if (size <= 10)
+ return;
+
+ tail = &data[size - 6];
+ memcpy(saved_bytes, tail, 6);
+ memset(tail, 0xE8, 6);
+ p = data;
+ for (;;) {
+ while (*p != 0xE8)
+ p++;
+ if (p >= tail)
+ break;
+ (*process_target)(p + 1, p - data);
+ p += 5;
}
+ memcpy(tail, saved_bytes, 6);
+#else
+ /* SSE2 or AVX-2 optimized version for x86_64 */
+
+ u8 *p = data;
+ u64 valid_mask = ~0;
+
+ if (size <= 10)
+ return;
+#ifdef __AVX2__
+# define ALIGNMENT_REQUIRED 32
+#else
+# define ALIGNMENT_REQUIRED 16
+#endif
- u8 *p8 = (u8 *)p128;
- while (!(valid_mask & 1)) {
- p8++;
+ /* Process one byte at a time until the pointer is properly aligned. */
+ while ((uintptr_t)p % ALIGNMENT_REQUIRED != 0) {
+ if (p >= data + size - 10)
+ return;
+ if (*p == 0xE8 && (valid_mask & 1)) {
+ (*process_target)(p + 1, p - data);
+ valid_mask &= ~0x1F;
+ }
+ p++;
valid_mask >>= 1;
+ valid_mask |= (u64)1 << 63;
}
-#else /* __SSE2__ */
- u8 *p8 = data;
-#endif /* !__SSE2__ */
-
- if (size > 10) {
- /* Finish any bytes that weren't processed by the vectorized
- * implementation. */
- u8 *p8_end = data + size - 10;
- do {
- if (*p8 == 0xe8) {
- (*process_target)((sle32 *)(p8 + 1), p8 - data);
- p8 += 5;
- } else {
- p8++;
+
+ if (data + size - p >= 64) {
+
+ /* Vectorized processing */
+
+ /* Note: we use a "trap" E8 byte to eliminate the need to check
+ * for end-of-buffer in the inner loop. This byte is carefully
+ * positioned so that it will never be changed by a previous
+ * translation before it is detected. */
+
+ u8 *trap = p + ((data + size - p) & ~31) - 32 + 4;
+ u8 saved_byte = *trap;
+ *trap = 0xE8;
+
+ for (;;) {
+ u32 e8_mask;
+ u8 *orig_p = p;
+ #ifdef __SSE2__
+ const __m128i e8_bytes = _mm_set1_epi8(0xE8);
+ for (;;) {
+ /* Read the next 32 bytes of data and test them
+ * for E8 bytes. */
+ __m128i bytes1 = *(const __m128i *)p;
+ __m128i bytes2 = *(const __m128i *)(p + 16);
+ __m128i cmpresult1 = _mm_cmpeq_epi8(bytes1, e8_bytes);
+ __m128i cmpresult2 = _mm_cmpeq_epi8(bytes2, e8_bytes);
+ u32 mask1 = _mm_movemask_epi8(cmpresult1);
+ u32 mask2 = _mm_movemask_epi8(cmpresult2);
+ /* The masks have a bit set for each E8 byte.
+ * We stay in this fast inner loop as long as
+ * there are no E8 bytes. */
+ if (mask1 | mask2) {
+ e8_mask = mask1 | (mask2 << 16);
+ break;
+ }
+ p += 32;
}
- } while (p8 < p8_end);
+ #else
+ /* AVX-2 */
+ const __m256i e8_bytes = _mm256_set1_epi8(0xE8);
+ for (;;) {
+ __m256i bytes = *(const __m256i *)p;
+ __m256i cmpresult = _mm256_cmpeq_epi8(bytes, e8_bytes);
+ e8_mask = _mm256_movemask_epi8(cmpresult);
+ if (e8_mask)
+ break;
+ p += 32;
+ }
+ #endif
+
+ /* Did we pass over data with no E8 bytes? */
+ if (p != orig_p)
+ valid_mask = ~0;
+
+ /* Are we nearing end-of-buffer? */
+ if (p == trap - 4)
+ break;
+
+ /* Process the E8 bytes. However, the AND with
+ * 'valid_mask' ensures we never process an E8 byte that
+ * was itself part of a translation target. */
+ while ((e8_mask &= valid_mask)) {
+ unsigned bit = ffs32(e8_mask);
+ (*process_target)(p + bit + 1, p + bit - data);
+ valid_mask &= ~((u64)0x1F << bit);
+ }
+
+ valid_mask >>= 32;
+ valid_mask |= 0xFFFFFFFF00000000;
+ p += 32;
+ }
+
+ *trap = saved_byte;
+ }
+
+ /* Approaching the end of the buffer; process one byte a time. */
+ while (p < data + size - 10) {
+ if (*p == 0xE8 && (valid_mask & 1)) {
+ (*process_target)(p + 1, p - data);
+ valid_mask &= ~0x1F;
+ }
+ p++;
+ valid_mask >>= 1;
+ valid_mask |= (u64)1 << 63;
}
+#endif /* __SSE2__ || __AVX2__ */
}
void