This patch adds support for LZ4 compression in the Linux Kernel.
LZ4 Compression APIs for kernel are based on LZ4 implementation
by Yann Collet and changed with kernel coding style.
LZ4 homepage : http://fastcompression.blogspot.com/p/lz4.html
LZ4 source repository : http://code.google.com/p/lz4/
svn revision : r90
Two APIs are added:
lz4_compress() support basic lz4 compression whereas lz4hc_compress() support
high compression or CPU performance get lower but compression ratio get higher.
Also, we require the pre-allocated working memory with the defined size and
destination buffer must be allocated with the size of lz4_compressbound.
Signed-off-by: Chanho Min <[email protected]>
---
include/linux/lz4.h | 36 ++++
lib/Kconfig | 6 +
lib/Makefile | 2 +
lib/lz4/Makefile | 2 +
lib/lz4/lz4_compress.c | 443 +++++++++++++++++++++++++++++++++++++
lib/lz4/lz4defs.h | 66 +++++-
lib/lz4/lz4hc_compress.c | 539 ++++++++++++++++++++++++++++++++++++++++++++++
7 files changed, 1092 insertions(+), 2 deletions(-)
create mode 100644 lib/lz4/lz4_compress.c
create mode 100644 lib/lz4/lz4hc_compress.c
diff --git a/include/linux/lz4.h b/include/linux/lz4.h
index 7f6c75a..d21c13f 100644
--- a/include/linux/lz4.h
+++ b/include/linux/lz4.h
@@ -9,6 +9,8 @@
* it under the terms of the GNU General Public License version 2 as
* published by the Free Software Foundation.
*/
+#define LZ4_MEM_COMPRESS (4096 * sizeof(unsigned char *))
+#define LZ4HC_MEM_COMPRESS (65538 * sizeof(unsigned char *))
/*
* lz4_compressbound()
@@ -21,6 +23,40 @@ static inline size_t lz4_compressbound(size_t isize)
}
/*
+ * lz4_compress()
+ * src : source address of the original data
+ * src_len : size of the original data
+ * dst : output buffer address of the compressed data
+ * This requires 'dst' of size LZ4_COMPRESSBOUND.
+ * dst_len : is the output size, which is returned after compress done
+ * workmem : address of the working memory.
+ * This requires 'workmem' of size LZ4_MEM_COMPRESS.
+ * return : Success if return 0
+ * Error if return (< 0)
+ * note : Destination buffer and workmem must be already allocated with
+ * the defined size.
+ */
+int lz4_compress(const unsigned char *src, size_t src_len,
+ unsigned char *dst, size_t *dst_len, void *wrkmem);
+
+ /*
+ * lz4hc_compress()
+ * src : source address of the original data
+ * src_len : size of the original data
+ * dst : output buffer address of the compressed data
+ * This requires 'dst' of size LZ4_COMPRESSBOUND.
+ * dst_len : is the output size, which is returned after compress done
+ * workmem : address of the working memory.
+ * This requires 'workmem' of size LZ4HC_MEM_COMPRESS.
+ * return : Success if return 0
+ * Error if return (< 0)
+ * note : Destination buffer and workmem must be already allocated with
+ * the defined size.
+ */
+int lz4hc_compress(const unsigned char *src, size_t src_len,
+ unsigned char *dst, size_t *dst_len, void *wrkmem);
+
+/*
* lz4_decompress()
* src : source address of the compressed data
* src_len : is the input size, whcih is returned after decompress done
diff --git a/lib/Kconfig b/lib/Kconfig
index d086415..7bfdbbc 100644
--- a/lib/Kconfig
+++ b/lib/Kconfig
@@ -189,6 +189,12 @@ config LZO_COMPRESS
config LZO_DECOMPRESS
tristate
+config LZ4_COMPRESS
+ tristate
+
+config LZ4HC_COMPRESS
+ tristate
+
config LZ4_DECOMPRESS
tristate
diff --git a/lib/Makefile b/lib/Makefile
index 0feeb19..378d1c4 100644
--- a/lib/Makefile
+++ b/lib/Makefile
@@ -73,6 +73,8 @@ obj-$(CONFIG_REED_SOLOMON) += reed_solomon/
obj-$(CONFIG_BCH) += bch.o
obj-$(CONFIG_LZO_COMPRESS) += lzo/
obj-$(CONFIG_LZO_DECOMPRESS) += lzo/
+obj-$(CONFIG_LZ4_COMPRESS) += lz4/
+obj-$(CONFIG_LZ4HC_COMPRESS) += lz4/
obj-$(CONFIG_LZ4_DECOMPRESS) += lz4/
obj-$(CONFIG_XZ_DEC) += xz/
obj-$(CONFIG_RAID6_PQ) += raid6/
diff --git a/lib/lz4/Makefile b/lib/lz4/Makefile
index 7f548c6..8085d04 100644
--- a/lib/lz4/Makefile
+++ b/lib/lz4/Makefile
@@ -1 +1,3 @@
+obj-$(CONFIG_LZ4_COMPRESS) += lz4_compress.o
+obj-$(CONFIG_LZ4HC_COMPRESS) += lz4hc_compress.o
obj-$(CONFIG_LZ4_DECOMPRESS) += lz4_decompress.o
diff --git a/lib/lz4/lz4_compress.c b/lib/lz4/lz4_compress.c
new file mode 100644
index 0000000..fd94058
--- /dev/null
+++ b/lib/lz4/lz4_compress.c
@@ -0,0 +1,443 @@
+/*
+ * LZ4 - Fast LZ compression algorithm
+ * Copyright (C) 2011-2012, Yann Collet.
+ * BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
+
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are
+ * met:
+ *
+ * * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * * Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following disclaimer
+ * in the documentation and/or other materials provided with the
+ * distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ * You can contact the author at :
+ * - LZ4 homepage : http://fastcompression.blogspot.com/p/lz4.html
+ * - LZ4 source repository : http://code.google.com/p/lz4/
+ *
+ * Changed for kernel use by:
+ * Chanho Min <[email protected]>
+ */
+
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/lz4.h>
+#include <asm/unaligned.h>
+#include "lz4defs.h"
+
+/*
+ * LZ4_compressCtx :
+ * -----------------
+ * Compress 'isize' bytes from 'source' into an output buffer 'dest' of
+ * maximum size 'maxOutputSize'. * If it cannot achieve it, compression
+ * will stop, and result of the function will be zero.
+ * return : the number of bytes written in buffer 'dest', or 0 if the
+ * compression fails
+ */
+static inline int lz4_compressctx(void *ctx,
+ const char *source,
+ char *dest,
+ int isize,
+ int maxoutputsize)
+{
+ HTYPE *hashtable = (HTYPE *)ctx;
+ const u8 *ip = (u8 *)source;
+#if LZ4_ARCH64
+ const BYTE * const base = ip;
+#else
+ const int base = 0;
+#endif
+ const u8 *anchor = ip;
+ const u8 *const iend = ip + isize;
+ const u8 *const mflimit = iend - MFLIMIT;
+ #define MATCHLIMIT (iend - LASTLITERALS)
+
+ u8 *op = (u8 *) dest;
+ u8 *const oend = op + maxoutputsize;
+ int length;
+ const int skipstrength = SKIPSTRENGTH;
+ u32 forwardh;
+ int lastrun;
+
+ /* Init */
+ if (isize < MINLENGTH)
+ goto _last_literals;
+
+ memset((void *)hashtable, 0, LZ4_MEM_COMPRESS);
+
+ /* First Byte */
+ hashtable[LZ4_HASH_VALUE(ip)] = ip - base;
+ ip++;
+ forwardh = LZ4_HASH_VALUE(ip);
+
+ /* Main Loop */
+ for (;;) {
+ int findmatchattempts = (1U << skipstrength) + 3;
+ const u8 *forwardip = ip;
+ const u8 *ref;
+ u8 *token;
+
+ /* Find a match */
+ do {
+ u32 h = forwardh;
+ int step = findmatchattempts++ >> skipstrength;
+ ip = forwardip;
+ forwardip = ip + step;
+
+ if (unlikely(forwardip > mflimit))
+ goto _last_literals;
+
+ forwardh = LZ4_HASH_VALUE(forwardip);
+ ref = base + hashtable[h];
+ hashtable[h] = ip - base;
+ } while ((ref < ip - MAX_DISTANCE) || (A32(ref) != A32(ip)));
+
+ /* Catch up */
+ while ((ip > anchor) && (ref > (u8 *)source) &&
+ unlikely(ip[-1] == ref[-1])) {
+ ip--;
+ ref--;
+ }
+
+ /* Encode Literal length */
+ length = (int)(ip - anchor);
+ token = op++;
+ /* check output limit */
+ if (unlikely(op + length + (2 + 1 + LASTLITERALS) +
+ (length >> 8) > oend))
+ return 0;
+
+ if (length >= (int)RUN_MASK) {
+ int len;
+ *token = (RUN_MASK << ML_BITS);
+ len = length - RUN_MASK;
+ for (; len > 254 ; len -= 255)
+ *op++ = 255;
+ *op++ = (u8)len;
+ } else
+ *token = (length << ML_BITS);
+
+ /* Copy Literals */
+ LZ4_BLINDCOPY(anchor, op, length);
+_next_match:
+ /* Encode Offset */
+ LZ4_WRITE_LITTLEENDIAN_16(op, (u16)(ip - ref));
+
+ /* Start Counting */
+ ip += MINMATCH;
+ /* MinMatch verified */
+ ref += MINMATCH;
+ anchor = ip;
+ while (likely(ip < MATCHLIMIT - (STEPSIZE - 1))) {
+ #if LZ4_ARCH64
+ u64 diff = A64(ref) ^ A64(ip);
+ #else
+ u32 diff = A32(ref) ^ A32(ip);
+ #endif
+ if (!diff) {
+ ip += STEPSIZE;
+ ref += STEPSIZE;
+ continue;
+ }
+ ip += LZ4_NBCOMMONBYTES(diff);
+ goto _endcount;
+ }
+ #if LZ4_ARCH64
+ if ((ip < (MATCHLIMIT - 3)) && (A32(ref) == A32(ip))) {
+ ip += 4;
+ ref += 4;
+ }
+ #endif
+ if ((ip < (MATCHLIMIT - 1)) && (A16(ref) == A16(ip))) {
+ ip += 2;
+ ref += 2;
+ }
+ if ((ip < MATCHLIMIT) && (*ref == *ip))
+ ip++;
+_endcount:
+ /* Encode MatchLength */
+ length = (int)(ip - anchor);
+ /* Check output limit */
+ if (unlikely(op + (1 + LASTLITERALS) + (length >> 8) > oend))
+ return 0;
+ if (length >= (int)ML_MASK) {
+ *token += ML_MASK;
+ length -= ML_MASK;
+ for (; length > 509 ; length -= 510) {
+ *op++ = 255;
+ *op++ = 255;
+ }
+ if (length > 254) {
+ length -= 255;
+ *op++ = 255;
+ }
+ *op++ = (u8)length;
+ } else
+ *token += length;
+
+ /* Test end of chunk */
+ if (ip > mflimit) {
+ anchor = ip;
+ break;
+ }
+
+ /* Fill table */
+ hashtable[LZ4_HASH_VALUE(ip-2)] = ip - 2 - base;
+
+ /* Test next position */
+ ref = base + hashtable[LZ4_HASH_VALUE(ip)];
+ hashtable[LZ4_HASH_VALUE(ip)] = ip - base;
+ if ((ref > ip - (MAX_DISTANCE + 1)) && (A32(ref) == A32(ip))) {
+ token = op++;
+ *token = 0;
+ goto _next_match;
+ }
+
+ /* Prepare next loop */
+ anchor = ip++;
+ forwardh = LZ4_HASH_VALUE(ip);
+ }
+
+_last_literals:
+ /* Encode Last Literals */
+ lastrun = (int)(iend - anchor);
+ if (((char *)op - dest) + lastrun + 1
+ + ((lastrun + 255 - RUN_MASK) / 255) > (u32)maxoutputsize)
+ return 0;
+
+ if (lastrun >= (int)RUN_MASK) {
+ *op++ = (RUN_MASK << ML_BITS);
+ lastrun -= RUN_MASK;
+ for (; lastrun > 254 ; lastrun -= 255)
+ *op++ = 255;
+ *op++ = (u8)lastrun;
+ } else
+ *op++ = (lastrun << ML_BITS);
+ memcpy(op, anchor, iend - anchor);
+ op += iend - anchor;
+
+ /* End */
+ return (int)(((char *)op) - dest);
+}
+
+static inline int lz4_compress64kctx(void *ctx,
+ const char *source,
+ char *dest,
+ int isize,
+ int maxoutputsize)
+{
+ u16 *hashtable = (u16 *)ctx;
+ const u8 *ip = (u8 *) source;
+ const u8 *anchor = ip;
+ const u8 *const base = ip;
+ const u8 *const iend = ip + isize;
+ const u8 *const mflimit = iend - MFLIMIT;
+ #define MATCHLIMIT (iend - LASTLITERALS)
+
+ u8 *op = (u8 *) dest;
+ u8 *const oend = op + maxoutputsize;
+ int len, length;
+ const int skipstrength = SKIPSTRENGTH;
+ u32 forwardh;
+ int lastrun;
+
+ /* Init */
+ if (isize < MINLENGTH)
+ goto _last_literals;
+
+ memset((void *)hashtable, 0, LZ4_MEM_COMPRESS);
+
+ /* First Byte */
+ ip++;
+ forwardh = LZ4_HASH64K_VALUE(ip);
+
+ /* Main Loop */
+ for (;;) {
+ int findmatchattempts = (1U << skipstrength) + 3;
+ const u8 *forwardip = ip;
+ const u8 *ref;
+ u8 *token;
+
+ /* Find a match */
+ do {
+ u32 h = forwardh;
+ int step = findmatchattempts++ >> skipstrength;
+ ip = forwardip;
+ forwardip = ip + step;
+
+ if (forwardip > mflimit)
+ goto _last_literals;
+
+ forwardh = LZ4_HASH64K_VALUE(forwardip);
+ ref = base + hashtable[h];
+ hashtable[h] = (u16)(ip - base);
+ } while (A32(ref) != A32(ip));
+
+ /* Catch up */
+ while ((ip > anchor) && (ref > (u8 *)source)
+ && (ip[-1] == ref[-1])) {
+ ip--;
+ ref--;
+ }
+
+ /* Encode Literal length */
+ length = (int)(ip - anchor);
+ token = op++;
+ /* Check output limit */
+ if (unlikely(op + length + (2 + 1 + LASTLITERALS)
+ + (length >> 8) > oend))
+ return 0;
+ if (length >= (int)RUN_MASK) {
+ *token = (RUN_MASK << ML_BITS);
+ len = length - RUN_MASK;
+ for (; len > 254 ; len -= 255)
+ *op++ = 255;
+ *op++ = (u8)len;
+ } else
+ *token = (length << ML_BITS);
+
+ /* Copy Literals */
+ LZ4_BLINDCOPY(anchor, op, length);
+
+_next_match:
+ /* Encode Offset */
+ LZ4_WRITE_LITTLEENDIAN_16(op, (u16)(ip - ref));
+
+ /* Start Counting */
+ ip += MINMATCH;
+ /* MinMatch verified */
+ ref += MINMATCH;
+ anchor = ip;
+
+ while (ip < MATCHLIMIT - (STEPSIZE - 1)) {
+ #if LZ4_ARCH64
+ u64 diff = A64(ref) ^ A64(ip);
+ #else
+ u32 diff = A32(ref) ^ A32(ip);
+ #endif
+
+ if (!diff) {
+ ip += STEPSIZE;
+ ref += STEPSIZE;
+ continue;
+ }
+ ip += LZ4_NBCOMMONBYTES(diff);
+ goto _endcount;
+ }
+ #if LZ4_ARCH64
+ if ((ip < (MATCHLIMIT - 3)) && (A32(ref) == A32(ip))) {
+ ip += 4;
+ ref += 4;
+ }
+ #endif
+ if ((ip < (MATCHLIMIT - 1)) && (A16(ref) == A16(ip))) {
+ ip += 2;
+ ref += 2;
+ }
+ if ((ip < MATCHLIMIT) && (*ref == *ip))
+ ip++;
+_endcount:
+
+ /* Encode MatchLength */
+ len = (int)(ip - anchor);
+ /* Check output limit */
+ if (unlikely(op + (1 + LASTLITERALS) + (len >> 8) > oend))
+ return 0;
+ if (len >= (int)ML_MASK) {
+ *token += ML_MASK;
+ len -= ML_MASK;
+ for (; len > 509 ; len -= 510) {
+ *op++ = 255;
+ *op++ = 255;
+ }
+ if (len > 254) {
+ len -= 255;
+ *op++ = 255;
+ }
+ *op++ = (u8)len;
+ } else
+ *token += len;
+
+ /* Test end of chunk */
+ if (ip > mflimit) {
+ anchor = ip;
+ break;
+ }
+
+ /* Fill table */
+ hashtable[LZ4_HASH64K_VALUE(ip-2)] = (u16)(ip - 2 - base);
+
+ /* Test next position */
+ ref = base + hashtable[LZ4_HASH64K_VALUE(ip)];
+ hashtable[LZ4_HASH64K_VALUE(ip)] = (u16)(ip - base);
+ if (A32(ref) == A32(ip)) {
+ token = op++;
+ *token = 0;
+ goto _next_match;
+ }
+
+ /* Prepare next loop */
+ anchor = ip++;
+ forwardh = LZ4_HASH64K_VALUE(ip);
+ }
+
+_last_literals:
+ /* Encode Last Literals */
+ lastrun = (int)(iend - anchor);
+ if (op + lastrun + 1 + (lastrun - RUN_MASK + 255) / 255 > oend)
+ return 0;
+ if (lastrun >= (int)RUN_MASK) {
+ *op++ = (RUN_MASK << ML_BITS);
+ lastrun -= RUN_MASK;
+ for (; lastrun > 254 ; lastrun -= 255)
+ *op++ = 255;
+ *op++ = (u8)lastrun;
+ } else
+ *op++ = (lastrun << ML_BITS);
+ memcpy(op, anchor, iend - anchor);
+ op += iend - anchor;
+ /* End */
+ return (int)(((char *)op) - dest);
+}
+
+int lz4_compress(const unsigned char *src, size_t src_len,
+ unsigned char *dst, size_t *dst_len, void *wrkmem)
+{
+ int ret = -1;
+ int out_len = 0;
+
+ if (src_len < LZ4_64KLIMIT)
+ out_len = lz4_compress64kctx(wrkmem, src, dst, src_len,
+ lz4_compressbound(src_len));
+ else
+ out_len = lz4_compressctx(wrkmem, src, dst, src_len,
+ lz4_compressbound(src_len));
+
+ if (out_len < 0)
+ goto exit;
+
+ *dst_len = out_len;
+
+ return 0;
+exit:
+ return ret;
+}
+EXPORT_SYMBOL_GPL(lz4_compress);
+
+MODULE_LICENSE("GPL");
+MODULE_DESCRIPTION("LZ4 compressor");
diff --git a/lib/lz4/lz4defs.h b/lib/lz4/lz4defs.h
index 43ac31d..abcecdc 100644
--- a/lib/lz4/lz4defs.h
+++ b/lib/lz4/lz4defs.h
@@ -22,23 +22,40 @@
* Architecture-specific macros
*/
#define BYTE u8
+typedef struct _U16_S { u16 v; } U16_S;
+typedef struct _U32_S { u32 v; } U32_S;
+typedef struct _U64_S { u64 v; } U64_S;
#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) \
|| defined(CONFIG_ARM) && __LINUX_ARM_ARCH__ >= 6 \
&& defined(ARM_EFFICIENT_UNALIGNED_ACCESS)
-typedef struct _U32_S { u32 v; } U32_S;
-typedef struct _U64_S { u64 v; } U64_S;
+#define A16(x) (((U16_S *)(x))->v)
#define A32(x) (((U32_S *)(x))->v)
#define A64(x) (((U64_S *)(x))->v)
#define PUT4(s, d) (A32(d) = A32(s))
#define PUT8(s, d) (A64(d) = A64(s))
+#define LZ4_WRITE_LITTLEENDIAN_16(p, v) \
+ do { \
+ A16(p) = v; \
+ p += 2; \
+ } while (0)
#else /* CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS */
+#define A64(x) get_unaligned((u64 *)&(((U16_S *)(x))->v))
+#define A32(x) get_unaligned((u32 *)&(((U16_S *)(x))->v))
+#define A16(x) get_unaligned((u16 *)&(((U16_S *)(x))->v))
+
#define PUT4(s, d) \
put_unaligned(get_unaligned((const u32 *) s), (u32 *) d)
#define PUT8(s, d) \
put_unaligned(get_unaligned((const u64 *) s), (u64 *) d)
+
+#define LZ4_WRITE_LITTLEENDIAN_16(p, v) \
+ do { \
+ put_unaligned(v, (u16 *)(p)); \
+ p += 2; \
+ } while (0)
#endif
#define COPYLENGTH 8
@@ -46,6 +63,29 @@ typedef struct _U64_S { u64 v; } U64_S;
#define ML_MASK ((1U << ML_BITS) - 1)
#define RUN_BITS (8 - ML_BITS)
#define RUN_MASK ((1U << RUN_BITS) - 1)
+#define MEMORY_USAGE 14
+#define MINMATCH 4
+#define SKIPSTRENGTH 6
+#define LASTLITERALS 5
+#define MFLIMIT (COPYLENGTH + MINMATCH)
+#define MINLENGTH (MFLIMIT + 1)
+#define MAXD_LOG 16
+#define MAXD (1 << MAXD_LOG)
+#define MAXD_MASK (u32)(MAXD - 1)
+#define MAX_DISTANCE (MAXD - 1)
+#define HASH_LOG (MAXD_LOG - 1)
+#define HASHTABLESIZE (1 << HASH_LOG)
+#define MAX_NB_ATTEMPTS 256
+#define OPTIMAL_ML (int)((ML_MASK-1)+MINMATCH)
+#define LZ4_64KLIMIT ((1<<16) + (MFLIMIT - 1))
+#define HASHLOG64K ((MEMORY_USAGE - 2) + 1)
+#define HASH64KTABLESIZE (1U << HASHLOG64K)
+#define LZ4_HASH_VALUE(p) (((A32(p)) * 2654435761U) >> \
+ ((MINMATCH * 8) - (MEMORY_USAGE-2)))
+#define LZ4_HASH64K_VALUE(p) (((A32(p)) * 2654435761U) >> \
+ ((MINMATCH * 8) - HASHLOG64K))
+#define HASH_VALUE(p) (((A32(p)) * 2654435761U) >> \
+ ((MINMATCH * 8) - HASH_LOG))
#if LZ4_ARCH64/* 64-bit */
#define STEPSIZE 8
@@ -65,6 +105,13 @@ typedef struct _U64_S { u64 v; } U64_S;
LZ4_WILDCOPY(s, d, e); \
} \
} while (0)
+#define HTYPE u32
+
+#ifdef __BIG_ENDIAN
+#define LZ4_NBCOMMONBYTES(val) (__builtin_clzll(val) >> 3)
+#else
+#define LZ4_NBCOMMONBYTES(val) (__builtin_ctzll(val) >> 3)
+#endif
#else /* 32-bit */
#define STEPSIZE 4
@@ -83,6 +130,14 @@ typedef struct _U64_S { u64 v; } U64_S;
} while (0)
#define LZ4_SECURECOPY LZ4_WILDCOPY
+#define HTYPE const u8*
+
+#ifdef __BIG_ENDIAN
+#define LZ4_NBCOMMONBYTES(val) (__builtin_clz(val) >> 3)
+#else
+#define LZ4_NBCOMMONBYTES(val) (__builtin_ctz(val) >> 3)
+#endif
+
#endif
#define LZ4_READ_LITTLEENDIAN_16(d, s, p) \
@@ -92,3 +147,10 @@ typedef struct _U64_S { u64 v; } U64_S;
do { \
LZ4_COPYPACKET(s, d); \
} while (d < e)
+
+#define LZ4_BLINDCOPY(s, d, l) \
+ do { \
+ u8 *e = (d) + l; \
+ LZ4_WILDCOPY(s, d, e); \
+ d = e; \
+ } while (0)
diff --git a/lib/lz4/lz4hc_compress.c b/lib/lz4/lz4hc_compress.c
new file mode 100644
index 0000000..a9a9c2a
--- /dev/null
+++ b/lib/lz4/lz4hc_compress.c
@@ -0,0 +1,539 @@
+/*
+ * LZ4 HC - High Compression Mode of LZ4
+ * Copyright (C) 2011-2012, Yann Collet.
+ * BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are
+ * met:
+ *
+ * * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * * Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following disclaimer
+ * in the documentation and/or other materials provided with the
+ * distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ * You can contact the author at :
+ * - LZ4 homepage : http://fastcompression.blogspot.com/p/lz4.html
+ * - LZ4 source repository : http://code.google.com/p/lz4/
+ *
+ * Changed for kernel use by:
+ * Chanho Min <[email protected]>
+ */
+
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/lz4.h>
+#include <asm/unaligned.h>
+#include "lz4defs.h"
+
+struct lz4hc_data {
+ const u8 *base;
+ HTYPE hashtable[HASHTABLESIZE];
+ u16 chaintable[MAXD];
+ const u8 *nexttoupdate;
+} __attribute__((__packed__));
+
+static inline int lz4hc_init(struct lz4hc_data *hc4, const u8 *base)
+{
+ memset((void *)hc4->hashtable, 0, sizeof(hc4->hashtable));
+ memset(hc4->chaintable, 0xFF, sizeof(hc4->chaintable));
+
+#if LZ4_ARCH64
+ hc4->nexttoupdate = base + 1;
+#else
+ hc4->nexttoupdate = base;
+#endif
+ hc4->base = base;
+ return 1;
+}
+
+/* Update chains up to ip (excluded) */
+static inline void lz4hc_insert(struct lz4hc_data *hc4, const u8 *ip)
+{
+ u16 *chaintable = hc4->chaintable;
+ HTYPE *hashtable = hc4->hashtable;
+#if LZ4_ARCH64
+ const BYTE * const base = hc4->base;
+#else
+ const int base = 0;
+#endif
+
+ while (hc4->nexttoupdate < ip) {
+ const u8 *p = hc4->nexttoupdate;
+ size_t delta = p - (hashtable[HASH_VALUE(p)] + base);
+ if (delta > MAX_DISTANCE)
+ delta = MAX_DISTANCE;
+ chaintable[(size_t)(p) & MAXD_MASK] = (u16)delta;
+ hashtable[HASH_VALUE(p)] = (p) - base;
+ hc4->nexttoupdate++;
+ }
+}
+
+static inline size_t lz4hc_commonlength(const u8 *p1, const u8 *p2,
+ const u8 *const matchlimit)
+{
+ const u8 *p1t = p1;
+
+ while (p1t < matchlimit - (STEPSIZE - 1)) {
+#if LZ4_ARCH64
+ u64 diff = A64(p2) ^ A64(p1t);
+#else
+ u32 diff = A32(p2) ^ A32(p1t);
+#endif
+ if (!diff) {
+ p1t += STEPSIZE;
+ p2 += STEPSIZE;
+ continue;
+ }
+ p1t += LZ4_NBCOMMONBYTES(diff);
+ return p1t - p1;
+ }
+#if LZ4_ARCH64
+ if ((p1t < (matchlimit-3)) && (A32(p2) == A32(p1t))) {
+ p1t += 4;
+ p2 += 4;
+ }
+#endif
+
+ if ((p1t < (matchlimit - 1)) && (A16(p2) == A16(p1t))) {
+ p1t += 2;
+ p2 += 2;
+ }
+ if ((p1t < matchlimit) && (*p2 == *p1t))
+ p1t++;
+ return p1t - p1;
+}
+
+static inline int lz4hc_insertandfindbestmatch(struct lz4hc_data *hc4,
+ const u8 *ip, const u8 *const matchlimit, const u8 **matchpos)
+{
+ u16 *const chaintable = hc4->chaintable;
+ HTYPE *const hashtable = hc4->hashtable;
+ const u8 *ref;
+#if LZ4_ARCH64
+ const BYTE * const base = hc4->base;
+#else
+ const int base = 0;
+#endif
+ int nbattempts = MAX_NB_ATTEMPTS;
+ size_t repl = 0, ml = 0;
+ u16 delta;
+
+ /* HC4 match finder */
+ lz4hc_insert(hc4, ip);
+ ref = hashtable[HASH_VALUE(ip)] + base;
+
+ /* potential repetition */
+ if (ref >= ip-4) {
+ /* confirmed */
+ if (A32(ref) == A32(ip)) {
+ delta = (u16)(ip-ref);
+ repl = ml = lz4hc_commonlength(ip + MINMATCH,
+ ref + MINMATCH, matchlimit) + MINMATCH;
+ *matchpos = ref;
+ }
+ ref -= (size_t)chaintable[(size_t)(ref) & MAXD_MASK];
+ }
+
+ while ((ref >= ip - MAX_DISTANCE) && nbattempts) {
+ nbattempts--;
+ if (*(ref + ml) == *(ip + ml)) {
+ if (A32(ref) == A32(ip)) {
+ size_t mlt =
+ lz4hc_commonlength(ip + MINMATCH,
+ ref + MINMATCH, matchlimit) + MINMATCH;
+ if (mlt > ml) {
+ ml = mlt;
+ *matchpos = ref;
+ }
+ }
+ }
+ ref -= (size_t)chaintable[(size_t)(ref) & MAXD_MASK];
+ }
+
+ /* Complete table */
+ if (repl) {
+ const BYTE *ptr = ip;
+ const BYTE *end;
+ end = ip + repl - (MINMATCH-1);
+ /* Pre-Load */
+ while (ptr < end - delta) {
+ chaintable[(size_t)(ptr) & MAXD_MASK] = delta;
+ ptr++;
+ }
+ do {
+ chaintable[(size_t)(ptr) & MAXD_MASK] = delta;
+ /* Head of chain */
+ hashtable[HASH_VALUE(ptr)] = (ptr) - base;
+ ptr++;
+ } while (ptr < end);
+ hc4->nexttoupdate = end;
+ }
+
+ return (int)ml;
+}
+
+static inline int lz4hc_insertandgetwidermatch(struct lz4hc_data *hc4,
+ const u8 *ip, const u8 *startlimit, const u8 *matchlimit, int longest,
+ const u8 **matchpos, const u8 **startpos)
+{
+ u16 *const chaintable = hc4->chaintable;
+ HTYPE *const hashtable = hc4->hashtable;
+#if LZ4_ARCH64
+ const BYTE * const base = hc4->base;
+#else
+ const int base = 0;
+#endif
+ const u8 *ref;
+ int nbattempts = MAX_NB_ATTEMPTS;
+ int delta = (int)(ip - startlimit);
+
+ /* First Match */
+ lz4hc_insert(hc4, ip);
+ ref = hashtable[HASH_VALUE(ip)] + base;
+
+ while ((ref >= ip - MAX_DISTANCE) && (ref >= hc4->base)
+ && (nbattempts)) {
+ nbattempts--;
+ if (*(startlimit + longest) == *(ref - delta + longest)) {
+ if (A32(ref) == A32(ip)) {
+ const u8 *reft = ref + MINMATCH;
+ const u8 *ipt = ip + MINMATCH;
+ const u8 *startt = ip;
+
+ while (ipt < matchlimit-(STEPSIZE - 1)) {
+ #if LZ4_ARCH64
+ u64 diff = A64(reft) ^ A64(ipt);
+ #else
+ u32 diff = A32(reft) ^ A32(ipt);
+ #endif
+
+ if (!diff) {
+ ipt += STEPSIZE;
+ reft += STEPSIZE;
+ continue;
+ }
+ ipt += LZ4_NBCOMMONBYTES(diff);
+ goto _endcount;
+ }
+ #if LZ4_ARCH64
+ if ((ipt < (matchlimit - 3))
+ && (A32(reft) == A32(ipt))) {
+ ipt += 4;
+ reft += 4;
+ }
+ ipt += 2;
+ #endif
+ if ((ipt < (matchlimit - 1))
+ && (A16(reft) == A16(ipt))) {
+ reft += 2;
+ }
+ if ((ipt < matchlimit) && (*reft == *ipt))
+ ipt++;
+_endcount:
+ reft = ref;
+
+ while ((startt > startlimit)
+ && (reft > hc4->base)
+ && (startt[-1] == reft[-1])) {
+ startt--;
+ reft--;
+ }
+
+ if ((ipt - startt) > longest) {
+ longest = (int)(ipt - startt);
+ *matchpos = reft;
+ *startpos = startt;
+ }
+ }
+ }
+ ref -= (size_t)chaintable[(size_t)(ref) & MAXD_MASK];
+ }
+ return longest;
+}
+
+static inline int lz4_encodesequence(const u8 **ip, u8 **op, const u8 **anchor,
+ int ml, const u8 *ref)
+{
+ int length, len;
+ u8 *token;
+
+ /* Encode Literal length */
+ length = (int)(*ip - *anchor);
+ token = (*op)++;
+ if (length >= (int)RUN_MASK) {
+ *token = (RUN_MASK << ML_BITS);
+ len = length - RUN_MASK;
+ for (; len > 254 ; len -= 255)
+ *(*op)++ = 255;
+ *(*op)++ = (u8)len;
+ } else
+ *token = (length << ML_BITS);
+
+ /* Copy Literals */
+ LZ4_BLINDCOPY(*anchor, *op, length);
+
+ /* Encode Offset */
+ LZ4_WRITE_LITTLEENDIAN_16(*op, (u16)(*ip - ref));
+
+ /* Encode MatchLength */
+ len = (int)(ml - MINMATCH);
+ if (len >= (int)ML_MASK) {
+ *token += ML_MASK;
+ len -= ML_MASK;
+ for (; len > 509 ; len -= 510) {
+ *(*op)++ = 255;
+ *(*op)++ = 255;
+ }
+ if (len > 254) {
+ len -= 255;
+ *(*op)++ = 255;
+ }
+ *(*op)++ = (u8)len;
+ } else
+ *token += len;
+
+ /* Prepare next loop */
+ *ip += ml;
+ *anchor = *ip;
+
+ return 0;
+}
+
+int lz4_compresshcctx(struct lz4hc_data *ctx,
+ const char *source,
+ char *dest,
+ int isize)
+{
+ const u8 *ip = (const u8 *)source;
+ const u8 *anchor = ip;
+ const u8 *const iend = ip + isize;
+ const u8 *const mflimit = iend - MFLIMIT;
+ const u8 *const matchlimit = (iend - LASTLITERALS);
+
+ u8 *op = (u8 *)dest;
+
+ int ml, ml2, ml3, ml0;
+ const u8 *ref = NULL;
+ const u8 *start2 = NULL;
+ const u8 *ref2 = NULL;
+ const u8 *start3 = NULL;
+ const u8 *ref3 = NULL;
+ const u8 *start0;
+ const u8 *ref0;
+ int lastrun;
+
+ ip++;
+
+ /* Main Loop */
+ while (ip < mflimit) {
+ ml = lz4hc_insertandfindbestmatch(ctx, ip, matchlimit, (&ref));
+ if (!ml) {
+ ip++;
+ continue;
+ }
+
+ /* saved, in case we would skip too much */
+ start0 = ip;
+ ref0 = ref;
+ ml0 = ml;
+_search2:
+ if (ip+ml < mflimit)
+ ml2 = lz4hc_insertandgetwidermatch(ctx, ip + ml - 2,
+ ip + 1, matchlimit, ml, &ref2, &start2);
+ else
+ ml2 = ml;
+ /* No better match */
+ if (ml2 == ml) {
+ lz4_encodesequence(&ip, &op, &anchor, ml, ref);
+ continue;
+ }
+
+ if (start0 < ip) {
+ /* empirical */
+ if (start2 < ip + ml0) {
+ ip = start0;
+ ref = ref0;
+ ml = ml0;
+ }
+ }
+ /*
+ * Here, start0==ip
+ * First Match too small : removed
+ */
+ if ((start2 - ip) < 3) {
+ ml = ml2;
+ ip = start2;
+ ref = ref2;
+ goto _search2;
+ }
+
+_search3:
+ /*
+ * Currently we have :
+ * ml2 > ml1, and
+ * ip1+3 <= ip2 (usually < ip1+ml1)
+ */
+ if ((start2 - ip) < OPTIMAL_ML) {
+ int correction;
+ int new_ml = ml;
+ if (new_ml > OPTIMAL_ML)
+ new_ml = OPTIMAL_ML;
+ if (ip + new_ml > start2 + ml2 - MINMATCH)
+ new_ml = (int)(start2 - ip) + ml2 - MINMATCH;
+ correction = new_ml - (int)(start2 - ip);
+ if (correction > 0) {
+ start2 += correction;
+ ref2 += correction;
+ ml2 -= correction;
+ }
+ }
+ /*
+ * Now, we have start2 = ip+new_ml,
+ * with new_ml=min(ml, OPTIMAL_ML=18)
+ */
+ if (start2 + ml2 < mflimit)
+ ml3 = lz4hc_insertandgetwidermatch(ctx,
+ start2 + ml2 - 3, start2, matchlimit,
+ ml2, &ref3, &start3);
+ else
+ ml3 = ml2;
+
+ /* No better match : 2 sequences to encode */
+ if (ml3 == ml2) {
+ /* ip & ref are known; Now for ml */
+ if (start2 < ip+ml)
+ ml = (int)(start2 - ip);
+
+ /* Now, encode 2 sequences */
+ lz4_encodesequence(&ip, &op, &anchor, ml, ref);
+ ip = start2;
+ lz4_encodesequence(&ip, &op, &anchor, ml2, ref2);
+ continue;
+ }
+
+ /* Not enough space for match 2 : remove it */
+ if (start3 < ip + ml + 3) {
+ /*
+ * can write Seq1 immediately ==> Seq2 is removed,
+ * so Seq3 becomes Seq1
+ */
+ if (start3 >= (ip + ml)) {
+ if (start2 < ip + ml) {
+ int correction =
+ (int)(ip + ml - start2);
+ start2 += correction;
+ ref2 += correction;
+ ml2 -= correction;
+ if (ml2 < MINMATCH) {
+ start2 = start3;
+ ref2 = ref3;
+ ml2 = ml3;
+ }
+ }
+
+ lz4_encodesequence(&ip, &op, &anchor, ml, ref);
+ ip = start3;
+ ref = ref3;
+ ml = ml3;
+
+ start0 = start2;
+ ref0 = ref2;
+ ml0 = ml2;
+ goto _search2;
+ }
+
+ start2 = start3;
+ ref2 = ref3;
+ ml2 = ml3;
+ goto _search3;
+ }
+
+ /*
+ * OK, now we have 3 ascending matches; let's write at least
+ * the first one ip & ref are known; Now for ml
+ */
+ if (start2 < ip + ml) {
+ if ((start2 - ip) < (int)ML_MASK) {
+ int correction;
+ if (ml > OPTIMAL_ML)
+ ml = OPTIMAL_ML;
+ if (ip + ml > start2 + ml2 - MINMATCH)
+ ml = (int)(start2 - ip) + ml2
+ - MINMATCH;
+ correction = ml - (int)(start2 - ip);
+ if (correction > 0) {
+ start2 += correction;
+ ref2 += correction;
+ ml2 -= correction;
+ }
+ } else
+ ml = (int)(start2 - ip);
+ }
+ lz4_encodesequence(&ip, &op, &anchor, ml, ref);
+
+ ip = start2;
+ ref = ref2;
+ ml = ml2;
+
+ start2 = start3;
+ ref2 = ref3;
+ ml2 = ml3;
+
+ goto _search3;
+ }
+
+ /* Encode Last Literals */
+ lastrun = (int)(iend - anchor);
+ if (lastrun >= (int)RUN_MASK) {
+ *op++ = (RUN_MASK << ML_BITS);
+ lastrun -= RUN_MASK;
+ for (; lastrun > 254 ; lastrun -= 255)
+ *op++ = 255;
+ *op++ = (u8) lastrun;
+ } else
+ *op++ = (lastrun << ML_BITS);
+ memcpy(op, anchor, iend - anchor);
+ op += iend - anchor;
+ /* End */
+ return (int) (((char *)op) - dest);
+}
+
+int lz4hc_compress(const unsigned char *src, size_t src_len,
+ unsigned char *dst, size_t *dst_len, void *wrkmem)
+{
+ int ret = -1;
+ int out_len = 0;
+
+ struct lz4hc_data *hc4 = (struct lz4hc_data *)wrkmem;
+ lz4hc_init(hc4, (const u8 *)src);
+ out_len = lz4_compresshcctx((struct lz4hc_data *)hc4, (const u8 *)src,
+ (char *)dst, (int)src_len);
+
+ if (out_len < 0)
+ goto exit;
+
+ *dst_len = out_len;
+ return 0;
+
+exit:
+ return ret;
+}
+EXPORT_SYMBOL_GPL(lz4hc_compress);
+
+MODULE_LICENSE("GPL");
+MODULE_DESCRIPTION("LZ4HC compressor");
--
1.7.9.5
On Thu, Mar 14, 2013 at 10:48 AM, Chanho Min <[email protected]> wrote:
> +#ifdef __BIG_ENDIAN
> +#define LZ4_NBCOMMONBYTES(val) (__builtin_clzll(val) >> 3)
> +#else
> +#define LZ4_NBCOMMONBYTES(val) (__builtin_ctzll(val) >> 3)
> +#endif
>
> #else /* 32-bit */
> #define STEPSIZE 4
> @@ -83,6 +130,14 @@ typedef struct _U64_S { u64 v; } U64_S;
> } while (0)
>
> #define LZ4_SECURECOPY LZ4_WILDCOPY
> +#define HTYPE const u8*
> +
> +#ifdef __BIG_ENDIAN
> +#define LZ4_NBCOMMONBYTES(val) (__builtin_clz(val) >> 3)
> +#else
> +#define LZ4_NBCOMMONBYTES(val) (__builtin_ctz(val) >> 3)
> +#endif
It seems at least m68k and sparc don't have the __builtin_clz() functions:
m68k-allmodconfig (http://kisskb.ellerman.id.au/kisskb/buildresult/8572593/):
ERROR: "__clzsi2" [lib/lz4/lz4hc_compress.ko] undefined!
ERROR: "__clzsi2" [lib/lz4/lz4_compress.ko] undefined!
sparc64-allmodconfig (http://kisskb.ellerman.id.au/kisskb/buildresult/8572790/):
ERROR: "__clzdi2" [lib/lz4/lz4hc_compress.ko] undefined!
ERROR: "__clzdi2" [lib/lz4/lz4_compress.ko] undefined!
sparc-allmodconfig (http://kisskb.ellerman.id.au/kisskb/buildresult/8572795/):
ERROR: "__clzdi2" [lib/lz4/lz4hc_compress.ko] undefined!
ERROR: "__clzdi2" [lib/lz4/lz4_compress.ko] undefined!
There may be more, hidden by build errors before reaching the link or modpost
phases.
The gmail-whitespace-damaged patch below fixes the m68k and sparc builds
for me, does this look acceptable?
diff --git a/lib/lz4/lz4defs.h b/lib/lz4/lz4defs.h
index abcecdc..81d3251 100644
--- a/lib/lz4/lz4defs.h
+++ b/lib/lz4/lz4defs.h
@@ -107,12 +107,6 @@ typedef struct _U64_S { u64 v; } U64_S;
} while (0)
#define HTYPE u32
-#ifdef __BIG_ENDIAN
-#define LZ4_NBCOMMONBYTES(val) (__builtin_clzll(val) >> 3)
-#else
-#define LZ4_NBCOMMONBYTES(val) (__builtin_ctzll(val) >> 3)
-#endif
-
#else /* 32-bit */
#define STEPSIZE 4
@@ -132,12 +126,12 @@ typedef struct _U64_S { u64 v; } U64_S;
#define LZ4_SECURECOPY LZ4_WILDCOPY
#define HTYPE const u8*
-#ifdef __BIG_ENDIAN
-#define LZ4_NBCOMMONBYTES(val) (__builtin_clz(val) >> 3)
-#else
-#define LZ4_NBCOMMONBYTES(val) (__builtin_ctz(val) >> 3)
#endif
+#ifdef __BIG_ENDIAN
+#define LZ4_NBCOMMONBYTES(val) (((STEPSIZE * 8) - 1 - __fls(val)) >> 3)
+#else
+#define LZ4_NBCOMMONBYTES(val) (__ffs(val) >> 3)
#endif
#define LZ4_READ_LITTLEENDIAN_16(d, s, p) \
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]
In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
On Mon, Apr 22, 2013 at 11:22 AM, Chanho Min <[email protected]> wrote:
>>> +#define HTYPE const u8*
>>> +
>>> +#ifdef __BIG_ENDIAN
>>> +#define LZ4_NBCOMMONBYTES(val) (__builtin_clz(val) >> 3)
>>> +#else
>>> +#define LZ4_NBCOMMONBYTES(val) (__builtin_ctz(val) >> 3)
>>> +#endif
>>
>>It seems at least m68k and sparc don't have the __builtin_clz() functions:
>>
>>m68k-allmodconfig (http://kisskb.ellerman.id.au/kisskb/buildresult/8572593/):
>>
>>ERROR: "__clzsi2" [lib/lz4/lz4hc_compress.ko] undefined!
>>ERROR: "__clzsi2" [lib/lz4/lz4_compress.ko] undefined!
>
> gcc seems to define __builtin_clz as __clzsi2 in some architecture.
> But, kernel doesn't link libgcc.a.
> If kernel should use gcc's built-in function without libgcc.a,
> do we need to port __clzsi2 to 'arch/*/lib/*'?
That's another option.
So far no one used __builtin_c[lt]z in generic code. We always used
__fls and __ffs.
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]
In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
On Mon, 22 Apr 2013 11:24:21 +0200 Geert Uytterhoeven <[email protected]> wrote:
> On Mon, Apr 22, 2013 at 11:22 AM, Chanho Min <[email protected]> wrote:
> >>> +#define HTYPE const u8*
> >>> +
> >>> +#ifdef __BIG_ENDIAN
> >>> +#define LZ4_NBCOMMONBYTES(val) (__builtin_clz(val) >> 3)
> >>> +#else
> >>> +#define LZ4_NBCOMMONBYTES(val) (__builtin_ctz(val) >> 3)
> >>> +#endif
> >>
> >>It seems at least m68k and sparc don't have the __builtin_clz() functions:
> >>
> >>m68k-allmodconfig (http://kisskb.ellerman.id.au/kisskb/buildresult/8572593/):
> >>
> >>ERROR: "__clzsi2" [lib/lz4/lz4hc_compress.ko] undefined!
> >>ERROR: "__clzsi2" [lib/lz4/lz4_compress.ko] undefined!
> >
> > gcc seems to define __builtin_clz as __clzsi2 in some architecture.
> > But, kernel doesn't link libgcc.a.
> > If kernel should use gcc's built-in function without libgcc.a,
> > do we need to port __clzsi2 to 'arch/*/lib/*'?
>
> That's another option.
Without having seen the patch .... Yes, if we fix it this way then we
also fix it for future callers of __builtin_clz().
On Mon, 22 Apr 2013 18:22:18 +0900 "Chanho Min" <[email protected]> wrote:
> >> +#define HTYPE const u8*
> >> +
> >> +#ifdef __BIG_ENDIAN
> >> +#define LZ4_NBCOMMONBYTES(val) (__builtin_clz(val) >> 3)
> >> +#else
> >> +#define LZ4_NBCOMMONBYTES(val) (__builtin_ctz(val) >> 3)
> >> +#endif
> >
> >It seems at least m68k and sparc don't have the __builtin_clz() functions:
> >
> >m68k-allmodconfig (http://kisskb.ellerman.id.au/kisskb/buildresult/8572593/):
> >
> >ERROR: "__clzsi2" [lib/lz4/lz4hc_compress.ko] undefined!
> >ERROR: "__clzsi2" [lib/lz4/lz4_compress.ko] undefined!
>
> gcc seems to define __builtin_clz as __clzsi2 in some architecture.
> But, kernel doesn't link libgcc.a.
> If kernel should use gcc's built-in function without libgcc.a,
> do we need to port __clzsi2 to 'arch/*/lib/*'?
This breaks alpha (gcc-4.4.4) as well. Can we please get this fixed
promptly?
>> gcc seems to define __builtin_clz as __clzsi2 in some architecture.
>> But, kernel doesn't link libgcc.a.
>> If kernel should use gcc's built-in function without libgcc.a,
>> do we need to port __clzsi2 to 'arch/*/lib/*'?
>
>This breaks alpha (gcc-4.4.4) as well. Can we please get this fixed
>promptly?
__clzsi2 can be implemented by using generic functions.
It can be overridden by linking arch-specific versions may not be implemented.
does this way look acceptable?
diff --git a/lib/Makefile b/lib/Makefile
index af79e8c..e17b3ee 100644
--- a/lib/Makefile
+++ b/lib/Makefile
@@ -23,7 +23,7 @@ lib-y += kobject.o klist.o
obj-y += bcd.o div64.o sort.o parser.o halfmd4.o debug_locks.o random32.o \
bust_spinlocks.o hexdump.o kasprintf.o bitmap.o scatterlist.o \
- gcd.o lcm.o list_sort.o uuid.o flex_array.o \
+ gcd.o lcm.o list_sort.o uuid.o flex_array.o clz.o\
bsearch.o find_last_bit.o find_next_bit.o llist.o memweight.o kfifo.o
obj-y += string_helpers.o
obj-$(CONFIG_TEST_STRING_HELPERS) += test-string_helpers.o
diff --git a/lib/clz.c b/lib/clz.c
index e69de29..6794b83 100644
--- a/lib/clz.c
+++ b/lib/clz.c
@@ -0,0 +1,23 @@
+#include <linux/kernel.h>
+
+int __attribute__((weak)) __clzsi2(int val)
+{
+ return BITS_PER_LONG - fls(val);
+}
+EXPORT_SYMBOL(__clzsi2);
+
+#if BITS_PER_LONG == 32
+int __attribute__((weak)) __clzdi2(long val)
+{
+ return BITS_PER_LONG - fls((int)val);
+}
+EXPORT_SYMBOL(__clzdi2);
+#elif BITS_PER_LONG == 64
+int __attribute__((weak)) __clzdi2(long val)
+{
+ return BITS_PER_LONG - fls64((u64)val);
+}
+EXPORT_SYMBOL(__clzdi2);
+#else
+#error BITS_PER_LONG not 32 or 64
+#endif
Hi,
On Fri, 26 Apr 2013 14:02:01 +0900 "Chanho Min" <[email protected]> wrote:
>
>
> @@ -0,0 +1,23 @@
> +#include <linux/kernel.h>
> +
> +int __attribute__((weak)) __clzsi2(int val)
We have __weak in <linux/compiler.h>
--
Cheers,
Stephen Rothwell [email protected]
On Fri, 26 Apr 2013 15:51:05 +1000 Stephen Rothwell <[email protected]> wrote:
> Hi,
>
> On Fri, 26 Apr 2013 14:02:01 +0900 "Chanho Min" <[email protected]> wrote:
> >
> >
> > @@ -0,0 +1,23 @@
> > +#include <linux/kernel.h>
> > +
> > +int __attribute__((weak)) __clzsi2(int val)
>
> We have __weak in <linux/compiler.h>
And lib/clz.c needs a few more includes, for EXPORT_SYMBOL,
BITS_PER_LONG maybe.
Plus a changelog and a signed-off-by, please.