libFLAC: Support 64bit brword/bwword

This patch allows FLAC__BYTES_PER_WORD to be set to 8, but is disabled by
default.

Patch-from: lvqcl <lvqcl.mail@gmail.com>

5 files changed, 222 insertions, 55 deletions

diff --git a/include/share/endswap.h b/include/share/endswap.h
index 4fde4c15..b189f3ec 100644
--- a/include/share/endswap.h
+++ b/include/share/endswap.h
@@ -43,11 +43,13 @@ static inline unsigned short __builtin_bswap16(unsigned short a)
 
 #define	ENDSWAP_16(x)		(__builtin_bswap16 (x))
 #define	ENDSWAP_32(x)		(__builtin_bswap32 (x))
+#define	ENDSWAP_64(x)		(__builtin_bswap64 (x))
 
 #elif defined _MSC_VER		/* Windows. Apparently in <stdlib.h>. */
 
 #define	ENDSWAP_16(x)		(_byteswap_ushort (x))
 #define	ENDSWAP_32(x)		(_byteswap_ulong (x))
+#define	ENDSWAP_64(x)		(_byteswap_uint64 (x))
 
 #elif defined HAVE_BYTESWAP_H		/* Linux */
 
@@ -55,16 +57,18 @@ static inline unsigned short __builtin_bswap16(unsigned short a)
 
 #define	ENDSWAP_16(x)		(bswap_16 (x))
 #define	ENDSWAP_32(x)		(bswap_32 (x))
+#define	ENDSWAP_64(x)		(bswap_64 (x))
 
 #else
 
 #define	ENDSWAP_16(x)		((((x) >> 8) & 0xFF) | (((x) & 0xFF) << 8))
 #define	ENDSWAP_32(x)		((((x) >> 24) & 0xFF) | (((x) >> 8) & 0xFF00) | (((x) & 0xFF00) << 8) | (((x) & 0xFF) << 24))
+#define	ENDSWAP_64(x)		((ENDSWAP_32(((x) >> 32) & 0xFFFFFFFF)) | (ENDSWAP_32((x) & 0xFFFFFFFF) << 32))
 
 #endif
 
 
-/* Host to little-endian byte swapping. */
+/* Host to little-endian byte swapping (for MD5 calculation) */
 #if CPU_IS_BIG_ENDIAN
 
 #define H2LE_16(x)		ENDSWAP_16 (x)
diff --git a/src/libFLAC/bitreader.c b/src/libFLAC/bitreader.c
index 67d0f371..3559bf12 100644
--- a/src/libFLAC/bitreader.c
+++ b/src/libFLAC/bitreader.c
@@ -45,18 +45,43 @@
 #include "share/endswap.h"
 
 /* Things should be fastest when this matches the machine word size */
-/* WATCHOUT: if you change this you must also change the following #defines down to FLAC__clz_uint32 below to match */
-/* WATCHOUT: there are a few places where the code will not work unless uint32_t is >= 32 bits wide */
+/* WATCHOUT: if you change this you must also change the following #defines down to COUNT_ZERO_MSBS2 below to match */
+/* WATCHOUT: there are a few places where the code will not work unless brword is >= 32 bits wide */
 /*           also, some sections currently only have fast versions for 4 or 8 bytes per word */
-#define FLAC__BYTES_PER_WORD 4		/* sizeof uint32_t */
-#define FLAC__BITS_PER_WORD (8 * FLAC__BYTES_PER_WORD)
+
+#if 1
+
+typedef FLAC__uint32 brword;
+#define FLAC__BYTES_PER_WORD 4		/* sizeof brword */
+#define FLAC__BITS_PER_WORD 32
 #define FLAC__WORD_ALL_ONES ((FLAC__uint32)0xffffffff)
-/* SWAP_BE_WORD_TO_HOST swaps bytes in a uint32_t (which is always big-endian) if necessary to match host byte order */
+/* SWAP_BE_WORD_TO_HOST swaps bytes in a brword (which is always big-endian) if necessary to match host byte order */
 #if WORDS_BIGENDIAN
 #define SWAP_BE_WORD_TO_HOST(x) (x)
 #else
 #define SWAP_BE_WORD_TO_HOST(x) ENDSWAP_32(x)
 #endif
+/* counts the # of zero MSBs in a word */
+#define COUNT_ZERO_MSBS(word) FLAC__clz_uint32(word)
+#define COUNT_ZERO_MSBS2(word) FLAC__clz2_uint32(word)
+
+#else
+
+typedef FLAC__uint64 brword;
+#define FLAC__BYTES_PER_WORD 8		/* sizeof brword */
+#define FLAC__BITS_PER_WORD 64
+#define FLAC__WORD_ALL_ONES ((FLAC__uint64)FLAC__U64L(0xffffffffffffffff))
+/* SWAP_BE_WORD_TO_HOST swaps bytes in a brword (which is always big-endian) if necessary to match host byte order */
+#if WORDS_BIGENDIAN
+#define SWAP_BE_WORD_TO_HOST(x) (x)
+#else
+#define SWAP_BE_WORD_TO_HOST(x) ENDSWAP_64(x)
+#endif
+/* counts the # of zero MSBs in a word */
+#define COUNT_ZERO_MSBS(word) FLAC__clz_uint64(word)
+#define COUNT_ZERO_MSBS2(word) FLAC__clz2_uint64(word)
+
+#endif
 
 /*
  * This should be at least twice as large as the largest number of words
@@ -77,7 +102,7 @@ static const unsigned FLAC__BITREADER_DEFAULT_CAPACITY = 65536u / FLAC__BITS_PER
 struct FLAC__BitReader {
 	/* any partially-consumed word at the head will stay right-justified as bits are consumed from the left */
 	/* any incomplete word at the tail will be left-justified, and bytes from the read callback are added on the right */
-	uint32_t *buffer;
+	brword *buffer;
 	unsigned capacity; /* in words */
 	unsigned words; /* # of completed words in buffer */
 	unsigned bytes; /* # of bytes in incomplete word at buffer[words] */
@@ -89,7 +114,7 @@ struct FLAC__BitReader {
 	void *client_data;
 };
 
-static inline void crc16_update_word_(FLAC__BitReader *br, uint32_t word)
+static inline void crc16_update_word_(FLAC__BitReader *br, brword word)
 {
 	register unsigned crc = br->read_crc16;
 #if FLAC__BYTES_PER_WORD == 4
@@ -142,7 +167,7 @@ static FLAC__bool bitreader_read_from_client_(FLAC__BitReader *br)
 		return false; /* no space left, buffer is too small; see note for FLAC__BITREADER_DEFAULT_CAPACITY  */
 	target = ((FLAC__byte*)(br->buffer+br->words)) + br->bytes;
 
-	/* before reading, if the existing reader looks like this (say uint32_t is 32 bits wide)
+	/* before reading, if the existing reader looks like this (say brword is 32 bits wide)
 	 *   bitstream :  11 22 33 44 55            br->words=1 br->bytes=1 (partial tail word is left-justified)
 	 *   buffer[BE]:  11 22 33 44 55 ?? ?? ??   (shown layed out as bytes sequentially in memory)
 	 *   buffer[LE]:  44 33 22 11 ?? ?? ?? 55   (?? being don't-care)
@@ -175,7 +200,7 @@ static FLAC__bool bitreader_read_from_client_(FLAC__BitReader *br)
 	 */
 #if WORDS_BIGENDIAN
 #else
-	end = (br->words*FLAC__BYTES_PER_WORD + br->bytes + bytes + (FLAC__BYTES_PER_WORD-1)) / FLAC__BYTES_PER_WORD;
+	end = (br->words*FLAC__BYTES_PER_WORD + br->bytes + (unsigned)bytes + (FLAC__BYTES_PER_WORD-1)) / FLAC__BYTES_PER_WORD;
 	for(start = br->words; start < end; start++)
 		br->buffer[start] = SWAP_BE_WORD_TO_HOST(br->buffer[start]);
 #endif
@@ -186,7 +211,7 @@ static FLAC__bool bitreader_read_from_client_(FLAC__BitReader *br)
 	 *   buffer[LE]:  44 33 22 11 88 77 66 55 CC BB AA 99 ?? FF EE DD
 	 * finally we'll update the reader values:
 	 */
-	end = br->words*FLAC__BYTES_PER_WORD + br->bytes + bytes;
+	end = br->words*FLAC__BYTES_PER_WORD + br->bytes + (unsigned)bytes;
 	br->words = end / FLAC__BYTES_PER_WORD;
 	br->bytes = end % FLAC__BYTES_PER_WORD;
 
@@ -236,7 +261,7 @@ FLAC__bool FLAC__bitreader_init(FLAC__BitReader *br, FLAC__BitReaderReadCallback
 	br->words = br->bytes = 0;
 	br->consumed_words = br->consumed_bits = 0;
 	br->capacity = FLAC__BITREADER_DEFAULT_CAPACITY;
-	br->buffer = malloc(sizeof(uint32_t) * br->capacity);
+	br->buffer = malloc(sizeof(brword) * br->capacity);
 	if(br->buffer == 0)
 		return false;
 	br->read_callback = rcb;
@@ -281,7 +306,7 @@ void FLAC__bitreader_dump(const FLAC__BitReader *br, FILE *out)
 				if(i < br->consumed_words || (i == br->consumed_words && j < br->consumed_bits))
 					fprintf(out, ".");
 				else
-					fprintf(out, "%01u", br->buffer[i] & (1 << (FLAC__BITS_PER_WORD-j-1)) ? 1:0);
+					fprintf(out, "%01u", br->buffer[i] & ((brword)1 << (FLAC__BITS_PER_WORD-j-1)) ? 1:0);
 			fprintf(out, "\n");
 		}
 		if(br->bytes > 0) {
@@ -290,7 +315,7 @@ void FLAC__bitreader_dump(const FLAC__BitReader *br, FILE *out)
 				if(i < br->consumed_words || (i == br->consumed_words && j < br->consumed_bits))
 					fprintf(out, ".");
 				else
-					fprintf(out, "%01u", br->buffer[i] & (1 << (br->bytes*8-j-1)) ? 1:0);
+					fprintf(out, "%01u", br->buffer[i] & ((brword)1 << (br->bytes*8-j-1)) ? 1:0);
 			fprintf(out, "\n");
 		}
 	}
@@ -315,7 +340,7 @@ FLAC__uint16 FLAC__bitreader_get_read_crc16(FLAC__BitReader *br)
 
 	/* CRC any tail bytes in a partially-consumed word */
 	if(br->consumed_bits) {
-		const uint32_t tail = br->buffer[br->consumed_words];
+		const brword tail = br->buffer[br->consumed_words];
 		for( ; br->crc16_align < br->consumed_bits; br->crc16_align += 8)
 			br->read_crc16 = FLAC__CRC16_UPDATE((unsigned)((tail >> (FLAC__BITS_PER_WORD-8-br->crc16_align)) & 0xff), br->read_crc16);
 	}
@@ -363,33 +388,34 @@ FLAC__bool FLAC__bitreader_read_raw_uint32(FLAC__BitReader *br, FLAC__uint32 *va
 		if(br->consumed_bits) {
 			/* this also works when consumed_bits==0, it's just a little slower than necessary for that case */
 			const unsigned n = FLAC__BITS_PER_WORD - br->consumed_bits;
-			const uint32_t word = br->buffer[br->consumed_words];
+			const brword word = br->buffer[br->consumed_words];
 			if(bits < n) {
-				*val = (word & (FLAC__WORD_ALL_ONES >> br->consumed_bits)) >> (n-bits);
+				*val = (FLAC__uint32)((word & (FLAC__WORD_ALL_ONES >> br->consumed_bits)) >> (n-bits)); /* The result has <= 32 non-zero bits */
 				br->consumed_bits += bits;
 				return true;
 			}
-			*val = word & (FLAC__WORD_ALL_ONES >> br->consumed_bits);
+			/* (FLAC__BITS_PER_WORD - br->consumed_bits <= bits) ==> (FLAC__WORD_ALL_ONES >> br->consumed_bits) has no more than 'bits' non-zero bits */
+			*val = (FLAC__uint32)(word & (FLAC__WORD_ALL_ONES >> br->consumed_bits));
 			bits -= n;
 			crc16_update_word_(br, word);
 			br->consumed_words++;
 			br->consumed_bits = 0;
 			if(bits) { /* if there are still bits left to read, there have to be less than 32 so they will all be in the next word */
 				*val <<= bits;
-				*val |= (br->buffer[br->consumed_words] >> (FLAC__BITS_PER_WORD-bits));
+				*val |= (FLAC__uint32)(br->buffer[br->consumed_words] >> (FLAC__BITS_PER_WORD-bits));
 				br->consumed_bits = bits;
 			}
 			return true;
 		}
-		else {
-			const uint32_t word = br->buffer[br->consumed_words];
+		else { /* br->consumed_bits == 0 */
+			const brword word = br->buffer[br->consumed_words];
 			if(bits < FLAC__BITS_PER_WORD) {
-				*val = word >> (FLAC__BITS_PER_WORD-bits);
+				*val = (FLAC__uint32)(word >> (FLAC__BITS_PER_WORD-bits));
 				br->consumed_bits = bits;
 				return true;
 			}
-			/* at this point 'bits' must be == FLAC__BITS_PER_WORD; because of previous assertions, it can't be larger */
-			*val = word;
+			/* at this point bits == FLAC__BITS_PER_WORD == 32; because of previous assertions, it can't be larger */
+			*val = (FLAC__uint32)word;
 			crc16_update_word_(br, word);
 			br->consumed_words++;
 			return true;
@@ -404,12 +430,12 @@ FLAC__bool FLAC__bitreader_read_raw_uint32(FLAC__BitReader *br, FLAC__uint32 *va
 		if(br->consumed_bits) {
 			/* this also works when consumed_bits==0, it's just a little slower than necessary for that case */
 			FLAC__ASSERT(br->consumed_bits + bits <= br->bytes*8);
-			*val = (br->buffer[br->consumed_words] & (FLAC__WORD_ALL_ONES >> br->consumed_bits)) >> (FLAC__BITS_PER_WORD-br->consumed_bits-bits);
+			*val = (FLAC__uint32)((br->buffer[br->consumed_words] & (FLAC__WORD_ALL_ONES >> br->consumed_bits)) >> (FLAC__BITS_PER_WORD-br->consumed_bits-bits));
 			br->consumed_bits += bits;
 			return true;
 		}
 		else {
-			*val = br->buffer[br->consumed_words] >> (FLAC__BITS_PER_WORD-bits);
+			*val = (FLAC__uint32)(br->buffer[br->consumed_words] >> (FLAC__BITS_PER_WORD-bits));
 			br->consumed_bits += bits;
 			return true;
 		}
@@ -565,7 +591,7 @@ FLAC__bool FLAC__bitreader_read_byte_block_aligned_no_crc(FLAC__BitReader *br, F
 	/* step 2: read whole words in chunks */
 	while(nvals >= FLAC__BYTES_PER_WORD) {
 		if(br->consumed_words < br->words) {
-			const uint32_t word = br->buffer[br->consumed_words++];
+			const brword word = br->buffer[br->consumed_words++];
 #if FLAC__BYTES_PER_WORD == 4
 			val[0] = (FLAC__byte)(word >> 24);
 			val[1] = (FLAC__byte)(word >> 16);
@@ -630,9 +656,9 @@ FLAC__bool FLAC__bitreader_read_unary_unsigned(FLAC__BitReader *br, unsigned *va
 	*val = 0;
 	while(1) {
 		while(br->consumed_words < br->words) { /* if we've not consumed up to a partial tail word... */
-			uint32_t b = br->buffer[br->consumed_words] << br->consumed_bits;
+			brword b = br->buffer[br->consumed_words] << br->consumed_bits;
 			if(b) {
-				i = FLAC__clz_uint32(b);
+				i = COUNT_ZERO_MSBS(b);
 				*val += i;
 				i++;
 				br->consumed_bits += i;
@@ -660,9 +686,9 @@ FLAC__bool FLAC__bitreader_read_unary_unsigned(FLAC__BitReader *br, unsigned *va
 		 */
 		if(br->bytes*8 > br->consumed_bits) {
 			const unsigned end = br->bytes * 8;
-			uint32_t b = (br->buffer[br->consumed_words] & (FLAC__WORD_ALL_ONES << (FLAC__BITS_PER_WORD-end))) << br->consumed_bits;
+			brword b = (br->buffer[br->consumed_words] & (FLAC__WORD_ALL_ONES << (FLAC__BITS_PER_WORD-end))) << br->consumed_bits;
 			if(b) {
-				i = FLAC__clz_uint32(b);
+				i = COUNT_ZERO_MSBS(b);
 				*val += i;
 				i++;
 				br->consumed_bits += i;
@@ -717,7 +743,7 @@ FLAC__bool FLAC__bitreader_read_rice_signed_block(FLAC__BitReader *br, int vals[
 	 * bitreader functions that use them, and before returning */
 	unsigned cwords, words, lsbs, msbs, x, y;
 	unsigned ucbits; /* keep track of the number of unconsumed bits in word */
-	uint32_t b;
+	brword b;
 	int *val, *end;
 
 	FLAC__ASSERT(0 != br);
@@ -758,7 +784,7 @@ FLAC__bool FLAC__bitreader_read_rice_signed_block(FLAC__BitReader *br, int vals[
 
 	while(val < end) {
 		/* read the unary MSBs and end bit */
-		x = y = FLAC__clz2_uint32(b);
+		x = y = COUNT_ZERO_MSBS2(b);
 		if(x == FLAC__BITS_PER_WORD) {
 			x = ucbits;
 			do {
@@ -767,7 +793,7 @@ FLAC__bool FLAC__bitreader_read_rice_signed_block(FLAC__BitReader *br, int vals[
 				if (cwords >= words)
 					goto incomplete_msbs;
 				b = br->buffer[cwords];
-				y = FLAC__clz2_uint32(b);
+				y = COUNT_ZERO_MSBS2(b);
 				x += y;
 			} while(y == FLAC__BITS_PER_WORD);
 		}
@@ -777,7 +803,7 @@ FLAC__bool FLAC__bitreader_read_rice_signed_block(FLAC__BitReader *br, int vals[
 		msbs = x;
 
 		/* read the binary LSBs */
-		x = b >> (FLAC__BITS_PER_WORD - parameter);
+		x = (FLAC__uint32)(b >> (FLAC__BITS_PER_WORD - parameter)); /* parameter < 32, so we can cast to 32-bit unsigned */
 		if(parameter <= ucbits) {
 			ucbits -= parameter;
 			b <<= parameter;
@@ -788,7 +814,7 @@ FLAC__bool FLAC__bitreader_read_rice_signed_block(FLAC__BitReader *br, int vals[
 				goto incomplete_lsbs;
 			b = br->buffer[cwords];
 			ucbits += FLAC__BITS_PER_WORD - parameter;
-			x |= b >> ucbits;
+			x |= (FLAC__uint32)(b >> ucbits);
 			b <<= FLAC__BITS_PER_WORD - ucbits;
 		}
 		lsbs = x;
diff --git a/src/libFLAC/bitwriter.c b/src/libFLAC/bitwriter.c
index 5582efc8..742fba36 100644
--- a/src/libFLAC/bitwriter.c
+++ b/src/libFLAC/bitwriter.c
@@ -46,33 +46,52 @@
 
 /* Things should be fastest when this matches the machine word size */
 /* WATCHOUT: if you change this you must also change the following #defines down to SWAP_BE_WORD_TO_HOST below to match */
-/* WATCHOUT: there are a few places where the code will not work unless uint32_t is >= 32 bits wide */
-#define FLAC__BYTES_PER_WORD 4
+/* WATCHOUT: there are a few places where the code will not work unless bwword is >= 32 bits wide */
+
+#if 1
+
+typedef FLAC__uint32 bwword;
+#define FLAC__BYTES_PER_WORD 4		/* sizeof bwword */
 #define FLAC__BITS_PER_WORD 32
 #define FLAC__WORD_ALL_ONES ((FLAC__uint32)0xffffffff)
-/* SWAP_BE_WORD_TO_HOST swaps bytes in a uint32_t (which is always big-endian) if necessary to match host byte order */
+/* SWAP_BE_WORD_TO_HOST swaps bytes in a bwword (which is always big-endian) if necessary to match host byte order */
 #if WORDS_BIGENDIAN
 #define SWAP_BE_WORD_TO_HOST(x) (x)
 #else
 #define SWAP_BE_WORD_TO_HOST(x) ENDSWAP_32(x)
 #endif
 
+#else
+
+typedef FLAC__uint64 bwword;
+#define FLAC__BYTES_PER_WORD 8		/* sizeof bwword */
+#define FLAC__BITS_PER_WORD 64
+#define FLAC__WORD_ALL_ONES ((FLAC__uint64)FLAC__U64L(0xffffffffffffffff))
+/* SWAP_BE_WORD_TO_HOST swaps bytes in a bwword (which is always big-endian) if necessary to match host byte order */
+#if WORDS_BIGENDIAN
+#define SWAP_BE_WORD_TO_HOST(x) (x)
+#else
+#define SWAP_BE_WORD_TO_HOST(x) ENDSWAP_64(x)
+#endif
+
+#endif
+
 /*
  * The default capacity here doesn't matter too much.  The buffer always grows
  * to hold whatever is written to it.  Usually the encoder will stop adding at
  * a frame or metadata block, then write that out and clear the buffer for the
  * next one.
  */
-static const unsigned FLAC__BITWRITER_DEFAULT_CAPACITY = 32768u / sizeof(uint32_t); /* size in words */
+static const unsigned FLAC__BITWRITER_DEFAULT_CAPACITY = 32768u / sizeof(bwword); /* size in words */
 /* When growing, increment 4K at a time */
-static const unsigned FLAC__BITWRITER_DEFAULT_INCREMENT = 4096u / sizeof(uint32_t); /* size in words */
+static const unsigned FLAC__BITWRITER_DEFAULT_INCREMENT = 4096u / sizeof(bwword); /* size in words */
 
 #define FLAC__WORDS_TO_BITS(words) ((words) * FLAC__BITS_PER_WORD)
 #define FLAC__TOTAL_BITS(bw) (FLAC__WORDS_TO_BITS((bw)->words) + (bw)->bits)
 
 struct FLAC__BitWriter {
-	uint32_t *buffer;
-	uint32_t accum; /* accumulator; bits are right-justified; when full, accum is appended to buffer */
+	bwword *buffer;
+	bwword accum; /* accumulator; bits are right-justified; when full, accum is appended to buffer */
 	unsigned capacity; /* capacity of buffer in words */
 	unsigned words; /* # of complete words in buffer */
 	unsigned bits; /* # of used bits in accum */
@@ -85,7 +104,7 @@ static
 FLAC__bool bitwriter_grow_(FLAC__BitWriter *bw, unsigned bits_to_add)
 {
 	unsigned new_capacity;
-	uint32_t *new_buffer;
+	bwword *new_buffer;
 
 	FLAC__ASSERT(0 != bw);
 	FLAC__ASSERT(0 != bw->buffer);
@@ -107,7 +126,7 @@ FLAC__bool bitwriter_grow_(FLAC__BitWriter *bw, unsigned bits_to_add)
 	FLAC__ASSERT(new_capacity > bw->capacity);
 	FLAC__ASSERT(new_capacity >= bw->words + ((bw->bits + bits_to_add + FLAC__BITS_PER_WORD - 1) / FLAC__BITS_PER_WORD));
 
-	new_buffer = safe_realloc_mul_2op_(bw->buffer, sizeof(uint32_t), /*times*/new_capacity);
+	new_buffer = safe_realloc_mul_2op_(bw->buffer, sizeof(bwword), /*times*/new_capacity);
 	if(new_buffer == 0)
 		return false;
 	bw->buffer = new_buffer;
@@ -149,7 +168,7 @@ FLAC__bool FLAC__bitwriter_init(FLAC__BitWriter *bw)
 
 	bw->words = bw->bits = 0;
 	bw->capacity = FLAC__BITWRITER_DEFAULT_CAPACITY;
-	bw->buffer = malloc(sizeof(uint32_t) * bw->capacity);
+	bw->buffer = malloc(sizeof(bwword) * bw->capacity);
 	if(bw->buffer == 0)
 		return false;
 
@@ -184,13 +203,13 @@ void FLAC__bitwriter_dump(const FLAC__BitWriter *bw, FILE *out)
 		for(i = 0; i < bw->words; i++) {
 			fprintf(out, "%08X: ", i);
 			for(j = 0; j < FLAC__BITS_PER_WORD; j++)
-				fprintf(out, "%01u", bw->buffer[i] & (1u << (FLAC__BITS_PER_WORD-j-1)) ? 1:0);
+				fprintf(out, "%01u", bw->buffer[i] & ((bwword)1 << (FLAC__BITS_PER_WORD-j-1)) ? 1:0);
 			fprintf(out, "\n");
 		}
 		if(bw->bits > 0) {
 			fprintf(out, "%08X: ", i);
 			for(j = 0; j < bw->bits; j++)
-				fprintf(out, "%01u", bw->accum & (1 << (bw->bits-j-1)) ? 1:0);
+				fprintf(out, "%01u", bw->accum & ((bwword)1 << (bw->bits-j-1)) ? 1:0);
 			fprintf(out, "\n");
 		}
 	}
@@ -337,7 +356,7 @@ inline FLAC__bool FLAC__bitwriter_write_raw_uint32(FLAC__BitWriter *bw, FLAC__ui
 		bw->accum = val;
 	}
 	else { /* at this point bits == FLAC__BITS_PER_WORD == 32  and  bw->bits == 0 */
-		bw->buffer[bw->words++] = SWAP_BE_WORD_TO_HOST(val);
+		bw->buffer[bw->words++] = SWAP_BE_WORD_TO_HOST((bwword)val);
 	}
 
 	return true;
@@ -533,8 +552,8 @@ FLAC__bool FLAC__bitwriter_write_rice_signed_block(FLAC__BitWriter *bw, const FL
 		msbits = uval >> parameter;
 		total_bits = lsbits + msbits;
 
-		if(bw->bits && bw->bits + total_bits < FLAC__BITS_PER_WORD) { /* i.e. if the whole thing fits in the current uint32_t */
-			/* ^^^ if bw->bits is 0 then we may have filled the buffer and have no free uint32_t to work in */
+		if(bw->bits && bw->bits + total_bits < FLAC__BITS_PER_WORD) { /* i.e. if the whole thing fits in the current bwword */
+			/* ^^^ if bw->bits is 0 then we may have filled the buffer and have no free bwword to work in */
 			bw->bits += total_bits;
 			uval |= mask1; /* set stop bit */
 			uval &= mask2; /* mask off unused top bits */
@@ -544,7 +563,7 @@ FLAC__bool FLAC__bitwriter_write_rice_signed_block(FLAC__BitWriter *bw, const FL
 		else {
 			/* slightly pessimistic size check but faster than "<= bw->words + (bw->bits+msbits+lsbits+FLAC__BITS_PER_WORD-1)/FLAC__BITS_PER_WORD" */
 			/* OPT: pessimism may cause flurry of false calls to grow_ which eat up all savings before it */
-			if(bw->capacity <= bw->words + bw->bits + msbits + 1 /* lsbits always fit in 1 uint32_t */ && !bitwriter_grow_(bw, total_bits))
+			if(bw->capacity <= bw->words + bw->bits + msbits + 1 /* lsbits always fit in 1 bwword */ && !bitwriter_grow_(bw, total_bits))
 				return false;
 
 			if(msbits) {
diff --git a/src/libFLAC/include/private/bitmath.h b/src/libFLAC/include/private/bitmath.h
index 220b8dcc..bdd154a9 100644
--- a/src/libFLAC/include/private/bitmath.h
+++ b/src/libFLAC/include/private/bitmath.h
@@ -91,7 +91,31 @@ static inline unsigned int FLAC__clz_uint32(FLAC__uint32 v)
 #endif
 }
 
-/* This one works with input 0 */
+/* Used when 64-bit bsr/clz is unavailable; can use 32-bit bsr/clz when possible */
+static inline unsigned int FLAC__clz_soft_uint64(FLAC__uint64 word)
+{
+	return (FLAC__uint32)(word>>32) ? FLAC__clz_uint32((FLAC__uint32)(word>>32)) :
+		FLAC__clz_uint32((FLAC__uint32)word) + 32;
+}
+
+static inline unsigned int FLAC__clz_uint64(FLAC__uint64 v)
+{
+	/* Never used with input 0 */
+	FLAC__ASSERT(v > 0);
+#if defined(__GNUC__) && (__GNUC__ >= 4 || (__GNUC__ == 3 && __GNUC_MINOR__ >= 4))
+	return __builtin_clzll(v);
+#elif (defined(__INTEL_COMPILER) || defined(_MSC_VER)) && (defined(_M_IA64) || defined(_M_X64))
+	{
+		unsigned long idx;
+		_BitScanReverse64(&idx, v);
+		return idx ^ 63U;
+	}
+#else
+	return FLAC__clz_soft_uint64(v);
+#endif
+}
+
+/* These two functions work with input 0 */
 static inline unsigned int FLAC__clz2_uint32(FLAC__uint32 v)
 {
 	if (!v)
@@ -99,6 +123,13 @@ static inline unsigned int FLAC__clz2_uint32(FLAC__uint32 v)
 	return FLAC__clz_uint32(v);
 }
 
+static inline unsigned int FLAC__clz2_uint64(FLAC__uint64 v)
+{
+	if (!v)
+		return 64;
+	return FLAC__clz_uint64(v);
+}
+
 /* An example of what FLAC__bitmath_ilog2() computes:
  *
  * ilog2( 0) = assertion failure
diff --git a/src/test_libFLAC/bitwriter.c b/src/test_libFLAC/bitwriter.c
index a652b301..d8f744c9 100644
--- a/src/test_libFLAC/bitwriter.c
+++ b/src/test_libFLAC/bitwriter.c
@@ -33,8 +33,21 @@
  * the definition here to get at the internals.  Make sure this is kept up
  * to date with what is in ../libFLAC/bitwriter.c
  */
+#if 1
+
 typedef FLAC__uint32 bwword;
+#define FLAC__BYTES_PER_WORD 4
 #define FLAC__BITS_PER_WORD 32
+#define PRI_BWWORD "08x"
+
+#else
+
+typedef FLAC__uint64 bwword;
+#define FLAC__BYTES_PER_WORD 8
+#define FLAC__BITS_PER_WORD 64
+#define PRI_BWWORD "016" PRIx64
+
+#endif
 
 struct FLAC__BitWriter {
 	bwword *buffer;
@@ -53,11 +66,21 @@ FLAC__bool test_bitwriter(void)
 	FLAC__BitWriter *bw;
 	FLAC__bool ok;
 	unsigned i, j;
+#if FLAC__BYTES_PER_WORD == 4
 #if WORDS_BIGENDIAN
 	static bwword test_pattern1[5] = { 0xaaf0aabe, 0xaaaaaaa8, 0x300aaaaa, 0xaaadeadb, 0x00eeface };
 #else
 	static bwword test_pattern1[5] = { 0xbeaaf0aa, 0xa8aaaaaa, 0xaaaa0a30, 0xdbeaadaa, 0x00eeface };
 #endif
+#elif FLAC__BYTES_PER_WORD == 8
+#if WORDS_BIGENDIAN
+	static bwword test_pattern1[3] = { FLAC__U64L(0xaaf0aabeaaaaaaa8), FLAC__U64L(0x300aaaaaaaadeadb), FLAC__U64L(0x0000000000eeface) };
+#else
+	static bwword test_pattern1[3] = { FLAC__U64L(0xa8aaaaaabeaaf0aa), FLAC__U64L(0xdbeaadaaaaaa0a30), FLAC__U64L(0x0000000000eeface) };
+#endif
+#else
+#error FLAC__BYTES_PER_WORD is neither 4 nor 8 -- not implemented
+#endif
 	unsigned words, bits; /* what we think bw->words and bw->bits should be */
 
 	printf("\n+++ libFLAC unit test: bitwriter\n\n");
@@ -196,7 +219,7 @@ FLAC__bool test_bitwriter(void)
 		return false;
 	}
 	if((bw->accum & 0x00ffffff) != test_pattern1[words]) {
-		printf("FAILED pattern match (bw->accum=%08X != %08X)\n", bw->accum&0x00ffffff, test_pattern1[words]);
+		printf("FAILED pattern match (bw->accum=%" PRI_BWWORD " != %" PRI_BWWORD ")\n", bw->accum&0x00ffffff, test_pattern1[words]);
 		FLAC__bitwriter_dump(bw, stdout);
 		return false;
 	}
@@ -229,7 +252,7 @@ FLAC__bool test_bitwriter(void)
 		return false;
 	}
 	if((bw->accum & 0x3fffffff) != test_pattern1[words]) {
-		printf("FAILED pattern match (bw->accum=%08X != %08X)\n", bw->accum&0x3fffffff, test_pattern1[words]);
+		printf("FAILED pattern match (bw->accum=%" PRI_BWWORD " != %" PRI_BWWORD ")\n", bw->accum&0x3fffffff, test_pattern1[words]);
 		FLAC__bitwriter_dump(bw, stdout);
 		return false;
 	}
@@ -299,11 +322,15 @@ FLAC__bool test_bitwriter(void)
 	printf("testing utf8_uint32(0x00010000)... ");
 	FLAC__bitwriter_clear(bw);
 	FLAC__bitwriter_write_utf8_uint32(bw, 0x00010000);
+#if FLAC__BYTES_PER_WORD == 4
 #if WORDS_BIGENDIAN
 	ok = TOTAL_BITS(bw) == 32 && bw->buffer[0] == 0xF0908080;
 #else
 	ok = TOTAL_BITS(bw) == 32 && bw->buffer[0] == 0x808090F0;
 #endif
+#elif FLAC__BYTES_PER_WORD == 8
+	ok = TOTAL_BITS(bw) == 32 && (bw->accum & 0xffffffff) == 0xF0908080;
+#endif
 	printf("%s\n", ok?"OK":"FAILED");
 	if(!ok) {
 		FLAC__bitwriter_dump(bw, stdout);
@@ -313,11 +340,15 @@ FLAC__bool test_bitwriter(void)
 	printf("testing utf8_uint32(0x001FFFFF)... ");
 	FLAC__bitwriter_clear(bw);
 	FLAC__bitwriter_write_utf8_uint32(bw, 0x001FFFFF);
+#if FLAC__BYTES_PER_WORD == 4
 #if WORDS_BIGENDIAN
 	ok = TOTAL_BITS(bw) == 32 && bw->buffer[0] == 0xF7BFBFBF;
 #else
 	ok = TOTAL_BITS(bw) == 32 && bw->buffer[0] == 0xBFBFBFF7;
 #endif
+#elif FLAC__BYTES_PER_WORD == 8
+	ok = TOTAL_BITS(bw) == 32 && (bw->accum & 0xffffffff) == 0xF7BFBFBF;
+#endif
 	printf("%s\n", ok?"OK":"FAILED");
 	if(!ok) {
 		FLAC__bitwriter_dump(bw, stdout);
@@ -327,11 +358,15 @@ FLAC__bool test_bitwriter(void)
 	printf("testing utf8_uint32(0x00200000)... ");
 	FLAC__bitwriter_clear(bw);
 	FLAC__bitwriter_write_utf8_uint32(bw, 0x00200000);
+#if FLAC__BYTES_PER_WORD == 4
 #if WORDS_BIGENDIAN
 	ok = TOTAL_BITS(bw) == 40 && bw->buffer[0] == 0xF8888080 && (bw->accum & 0xff) == 0x80;
 #else
 	ok = TOTAL_BITS(bw) == 40 && bw->buffer[0] == 0x808088F8 && (bw->accum & 0xff) == 0x80;
 #endif
+#elif FLAC__BYTES_PER_WORD == 8
+	ok = TOTAL_BITS(bw) == 40 && (bw->accum & FLAC__U64L(0xffffffffff)) == FLAC__U64L(0xF888808080);
+#endif
 	printf("%s\n", ok?"OK":"FAILED");
 	if(!ok) {
 		FLAC__bitwriter_dump(bw, stdout);
@@ -341,11 +376,15 @@ FLAC__bool test_bitwriter(void)
 	printf("testing utf8_uint32(0x03FFFFFF)... ");
 	FLAC__bitwriter_clear(bw);
 	FLAC__bitwriter_write_utf8_uint32(bw, 0x03FFFFFF);
+#if FLAC__BYTES_PER_WORD == 4
 #if WORDS_BIGENDIAN
 	ok = TOTAL_BITS(bw) == 40 && bw->buffer[0] == 0xFBBFBFBF && (bw->accum & 0xff) == 0xBF;
 #else
 	ok = TOTAL_BITS(bw) == 40 && bw->buffer[0] == 0xBFBFBFFB && (bw->accum & 0xff) == 0xBF;
 #endif
+#elif FLAC__BYTES_PER_WORD == 8
+	ok = TOTAL_BITS(bw) == 40 && (bw->accum & FLAC__U64L(0xffffffffff)) == FLAC__U64L(0xFBBFBFBFBF);
+#endif
 	printf("%s\n", ok?"OK":"FAILED");
 	if(!ok) {
 		FLAC__bitwriter_dump(bw, stdout);
@@ -355,11 +394,15 @@ FLAC__bool test_bitwriter(void)
 	printf("testing utf8_uint32(0x04000000)... ");
 	FLAC__bitwriter_clear(bw);
 	FLAC__bitwriter_write_utf8_uint32(bw, 0x04000000);
+#if FLAC__BYTES_PER_WORD == 4
 #if WORDS_BIGENDIAN
 	ok = TOTAL_BITS(bw) == 48 && bw->buffer[0] == 0xFC848080 && (bw->accum & 0xffff) == 0x8080;
 #else
 	ok = TOTAL_BITS(bw) == 48 && bw->buffer[0] == 0x808084FC && (bw->accum & 0xffff) == 0x8080;
 #endif
+#elif FLAC__BYTES_PER_WORD == 8
+	ok = TOTAL_BITS(bw) == 48 && (bw->accum & FLAC__U64L(0xffffffffffff)) == FLAC__U64L(0xFC8480808080);
+#endif
 	printf("%s\n", ok?"OK":"FAILED");
 	if(!ok) {
 		FLAC__bitwriter_dump(bw, stdout);
@@ -369,11 +412,15 @@ FLAC__bool test_bitwriter(void)
 	printf("testing utf8_uint32(0x7FFFFFFF)... ");
 	FLAC__bitwriter_clear(bw);
 	FLAC__bitwriter_write_utf8_uint32(bw, 0x7FFFFFFF);
+#if FLAC__BYTES_PER_WORD == 4
 #if WORDS_BIGENDIAN
 	ok = TOTAL_BITS(bw) == 48 && bw->buffer[0] == 0xFDBFBFBF && (bw->accum & 0xffff) == 0xBFBF;
 #else
 	ok = TOTAL_BITS(bw) == 48 && bw->buffer[0] == 0xBFBFBFFD && (bw->accum & 0xffff) == 0xBFBF;
 #endif
+#elif FLAC__BYTES_PER_WORD == 8
+	ok = TOTAL_BITS(bw) == 48 && (bw->accum & FLAC__U64L(0xffffffffffff)) == FLAC__U64L(0xFDBFBFBFBFBF);
+#endif
 	printf("%s\n", ok?"OK":"FAILED");
 	if(!ok) {
 		FLAC__bitwriter_dump(bw, stdout);
@@ -443,11 +490,15 @@ FLAC__bool test_bitwriter(void)
 	printf("testing utf8_uint64(0x0000000000010000)... ");
 	FLAC__bitwriter_clear(bw);
 	FLAC__bitwriter_write_utf8_uint64(bw, FLAC__U64L(0x0000000000010000));
+#if FLAC__BYTES_PER_WORD == 4
 #if WORDS_BIGENDIAN
 	ok = TOTAL_BITS(bw) == 32 && bw->buffer[0] == 0xF0908080;
 #else
 	ok = TOTAL_BITS(bw) == 32 && bw->buffer[0] == 0x808090F0;
 #endif
+#elif FLAC__BYTES_PER_WORD == 8
+	ok = TOTAL_BITS(bw) == 32 && (bw->accum & 0xffffffff) == 0xF0908080;
+#endif
 	printf("%s\n", ok?"OK":"FAILED");
 	if(!ok) {
 		FLAC__bitwriter_dump(bw, stdout);
@@ -457,11 +508,15 @@ FLAC__bool test_bitwriter(void)
 	printf("testing utf8_uint64(0x00000000001FFFFF)... ");
 	FLAC__bitwriter_clear(bw);
 	FLAC__bitwriter_write_utf8_uint64(bw, FLAC__U64L(0x00000000001FFFFF));
+#if FLAC__BYTES_PER_WORD == 4
 #if WORDS_BIGENDIAN
 	ok = TOTAL_BITS(bw) == 32 && bw->buffer[0] == 0xF7BFBFBF;
 #else
 	ok = TOTAL_BITS(bw) == 32 && bw->buffer[0] == 0xBFBFBFF7;
 #endif
+#elif FLAC__BYTES_PER_WORD == 8
+	ok = TOTAL_BITS(bw) == 32 && (bw->accum & 0xffffffff) == 0xF7BFBFBF;
+#endif
 	printf("%s\n", ok?"OK":"FAILED");
 	if(!ok) {
 		FLAC__bitwriter_dump(bw, stdout);
@@ -471,11 +526,15 @@ FLAC__bool test_bitwriter(void)
 	printf("testing utf8_uint64(0x0000000000200000)... ");
 	FLAC__bitwriter_clear(bw);
 	FLAC__bitwriter_write_utf8_uint64(bw, FLAC__U64L(0x0000000000200000));
+#if FLAC__BYTES_PER_WORD == 4
 #if WORDS_BIGENDIAN
 	ok = TOTAL_BITS(bw) == 40 && bw->buffer[0] == 0xF8888080 && (bw->accum & 0xff) == 0x80;
 #else
 	ok = TOTAL_BITS(bw) == 40 && bw->buffer[0] == 0x808088F8 && (bw->accum & 0xff) == 0x80;
 #endif
+#elif FLAC__BYTES_PER_WORD == 8
+	ok = TOTAL_BITS(bw) == 40 && (bw->accum & FLAC__U64L(0xffffffffff)) == FLAC__U64L(0xF888808080);
+#endif
 	printf("%s\n", ok?"OK":"FAILED");
 	if(!ok) {
 		FLAC__bitwriter_dump(bw, stdout);
@@ -485,11 +544,15 @@ FLAC__bool test_bitwriter(void)
 	printf("testing utf8_uint64(0x0000000003FFFFFF)... ");
 	FLAC__bitwriter_clear(bw);
 	FLAC__bitwriter_write_utf8_uint64(bw, FLAC__U64L(0x0000000003FFFFFF));
+#if FLAC__BYTES_PER_WORD == 4
 #if WORDS_BIGENDIAN
 	ok = TOTAL_BITS(bw) == 40 && bw->buffer[0] == 0xFBBFBFBF && (bw->accum & 0xff) == 0xBF;
 #else
 	ok = TOTAL_BITS(bw) == 40 && bw->buffer[0] == 0xBFBFBFFB && (bw->accum & 0xff) == 0xBF;
 #endif
+#elif FLAC__BYTES_PER_WORD == 8
+	ok = TOTAL_BITS(bw) == 40 && (bw->accum & FLAC__U64L(0xffffffffff)) == FLAC__U64L(0xFBBFBFBFBF);
+#endif
 	printf("%s\n", ok?"OK":"FAILED");
 	if(!ok) {
 		FLAC__bitwriter_dump(bw, stdout);
@@ -499,11 +562,15 @@ FLAC__bool test_bitwriter(void)
 	printf("testing utf8_uint64(0x0000000004000000)... ");
 	FLAC__bitwriter_clear(bw);
 	FLAC__bitwriter_write_utf8_uint64(bw, FLAC__U64L(0x0000000004000000));
+#if FLAC__BYTES_PER_WORD == 4
 #if WORDS_BIGENDIAN
 	ok = TOTAL_BITS(bw) == 48 && bw->buffer[0] == 0xFC848080 && (bw->accum & 0xffff) == 0x8080;
 #else
 	ok = TOTAL_BITS(bw) == 48 && bw->buffer[0] == 0x808084FC && (bw->accum & 0xffff) == 0x8080;
 #endif
+#elif FLAC__BYTES_PER_WORD == 8
+	ok = TOTAL_BITS(bw) == 48 && (bw->accum & FLAC__U64L(0xffffffffffff)) == FLAC__U64L(0xFC8480808080);
+#endif
 	printf("%s\n", ok?"OK":"FAILED");
 	if(!ok) {
 		FLAC__bitwriter_dump(bw, stdout);
@@ -513,11 +580,15 @@ FLAC__bool test_bitwriter(void)
 	printf("testing utf8_uint64(0x000000007FFFFFFF)... ");
 	FLAC__bitwriter_clear(bw);
 	FLAC__bitwriter_write_utf8_uint64(bw, FLAC__U64L(0x000000007FFFFFFF));
+#if FLAC__BYTES_PER_WORD == 4
 #if WORDS_BIGENDIAN
 	ok = TOTAL_BITS(bw) == 48 && bw->buffer[0] == 0xFDBFBFBF && (bw->accum & 0xffff) == 0xBFBF;
 #else
 	ok = TOTAL_BITS(bw) == 48 && bw->buffer[0] == 0xBFBFBFFD && (bw->accum & 0xffff) == 0xBFBF;
 #endif
+#elif FLAC__BYTES_PER_WORD == 8
+	ok = TOTAL_BITS(bw) == 48 && (bw->accum & FLAC__U64L(0xffffffffffff)) == FLAC__U64L(0xFDBFBFBFBFBF);
+#endif
 	printf("%s\n", ok?"OK":"FAILED");
 	if(!ok) {
 		FLAC__bitwriter_dump(bw, stdout);
@@ -527,11 +598,15 @@ FLAC__bool test_bitwriter(void)
 	printf("testing utf8_uint64(0x0000000080000000)... ");
 	FLAC__bitwriter_clear(bw);
 	FLAC__bitwriter_write_utf8_uint64(bw, FLAC__U64L(0x0000000080000000));
+#if FLAC__BYTES_PER_WORD == 4
 #if WORDS_BIGENDIAN
 	ok = TOTAL_BITS(bw) == 56 && bw->buffer[0] == 0xFE828080 && (bw->accum & 0xffffff) == 0x808080;
 #else
 	ok = TOTAL_BITS(bw) == 56 && bw->buffer[0] == 0x808082FE && (bw->accum & 0xffffff) == 0x808080;
 #endif
+#elif FLAC__BYTES_PER_WORD == 8
+	ok = TOTAL_BITS(bw) == 56 && (bw->accum & FLAC__U64L(0xffffffffffffff)) == FLAC__U64L(0xFE828080808080);
+#endif
 	printf("%s\n", ok?"OK":"FAILED");
 	if(!ok) {
 		FLAC__bitwriter_dump(bw, stdout);
@@ -541,11 +616,15 @@ FLAC__bool test_bitwriter(void)
 	printf("testing utf8_uint64(0x0000000FFFFFFFFF)... ");
 	FLAC__bitwriter_clear(bw);
 	FLAC__bitwriter_write_utf8_uint64(bw, FLAC__U64L(0x0000000FFFFFFFFF));
+#if FLAC__BYTES_PER_WORD == 4
 #if WORDS_BIGENDIAN
 	ok = TOTAL_BITS(bw) == 56 && bw->buffer[0] == 0xFEBFBFBF && (bw->accum & 0xffffff) == 0xBFBFBF;
 #else
 	ok = TOTAL_BITS(bw) == 56 && bw->buffer[0] == 0xBFBFBFFE && (bw->accum & 0xffffff) == 0xBFBFBF;
 #endif
+#elif FLAC__BYTES_PER_WORD == 8
+	ok = TOTAL_BITS(bw) == 56 && (bw->accum & FLAC__U64L(0xffffffffffffff)) == FLAC__U64L(0xFEBFBFBFBFBFBF);
+#endif
 	printf("%s\n", ok?"OK":"FAILED");
 	if(!ok) {
 		FLAC__bitwriter_dump(bw, stdout);
@@ -558,11 +637,19 @@ FLAC__bool test_bitwriter(void)
 	j = bw->capacity;
 	for(i = 0; i < j; i++)
 		FLAC__bitwriter_write_raw_uint32(bw, 0xaaaaaaaa, 32);
+#if FLAC__BYTES_PER_WORD == 4
 #if WORDS_BIGENDIAN
 	ok = TOTAL_BITS(bw) == i*32+4 && bw->buffer[0] == 0x5aaaaaaa && (bw->accum & 0xf) == 0xa;
 #else
 	ok = TOTAL_BITS(bw) == i*32+4 && bw->buffer[0] == 0xaaaaaa5a && (bw->accum & 0xf) == 0xa;
 #endif
+#elif FLAC__BYTES_PER_WORD == 8
+#if WORDS_BIGENDIAN
+	ok = TOTAL_BITS(bw) == i*32+4 && bw->buffer[0] == FLAC__U64L(0x5aaaaaaaaaaaaaaa) && (bw->accum & 0xf) == 0xa;
+#else
+	ok = TOTAL_BITS(bw) == i*32+4 && bw->buffer[0] == FLAC__U64L(0xaaaaaaaaaaaaaa5a) && (bw->accum & 0xf) == 0xa;
+#endif
+#endif
 	printf("%s\n", ok?"OK":"FAILED");
 	if(!ok) {
 		FLAC__bitwriter_dump(bw, stdout);