X-Git-Url: https://git.sesse.net/?p=fjl;a=blobdiff_plain;f=dehuff.h;h=2669034554c25fb16362ace9879f5fa280be9592;hp=d93e7931ee17f081e734e7ff5397dd2f973f200d;hb=19a58db08afd149a862506936093423db756a2dc;hpb=618548d1f2e076a3da21368e708cf887dcbd20d2

diff --git a/dehuff.h b/dehuff.h
index d93e793..2669034 100644
--- a/dehuff.h
+++ b/dehuff.h
@@ -5,7 +5,8 @@
 #include <stdint.h>
 #include <sys/types.h>
 
-#include "input.h"
+#include "bytesource.h"
+#include "bitsource.h"
 
 // About 99% of all Huffman codes are <= 8 bits long (see codelen.txt),
 // and it's what libjpeg uses. Thus, it seems like a reasonable size.
@@ -13,11 +14,6 @@
 #define DEHUF_TABLE_SIZE (1 << DEHUF_TABLE_BITS)
 static const int DEHUF_SLOW_PATH = -1;
 
-// A function to read bytes from some input source. The bytes should be
-// already unstuffed (and thus without markers).
-// A return value of -1 indicates error, a return value of 0 indicates EOF.
-typedef ssize_t (raw_input_func_t)(void*, uint8_t*, size_t);
-
 struct huffman_table {
 	unsigned num_codes[17];     // BITS
 	unsigned char codes[256];   // HUFFVAL
@@ -47,19 +43,16 @@ enum coefficient_class {
 typedef struct huffman_table huffman_tables_t[NUM_COEFF_CLASSES][4];
 
 // Read Huffman tables from a stream, and compute the derived values.
-void read_huffman_tables(huffman_tables_t* dst, raw_input_func_t* input_func, void* userdata);
+void read_huffman_tables(huffman_tables_t* dst, input_func_t* input_func, void* userdata);
 
 unsigned read_huffman_symbol_slow_path(const struct huffman_table* table,
                                        struct bit_source* source);
 
-#include <stdio.h>
-
-static inline unsigned read_huffman_symbol(const struct huffman_table* table,
-                                           struct bit_source* source)
+static inline unsigned read_huffman_symbol_no_refill(
+	const struct huffman_table* table,
+	struct bit_source* source)
 {
-	// FIXME: We can read past the end of the stream here in some edge
-	// cases. We need to define some guarantees in the layers above.
-	possibly_refill(source, DEHUF_TABLE_BITS);
+	assert(source->bits_available >= DEHUF_TABLE_BITS);
 	unsigned lookup = peek_bits(source, DEHUF_TABLE_BITS);
 	int code = table->lookup_table_codes[lookup];
 	int length = table->lookup_table_length[lookup];
@@ -67,9 +60,45 @@ static inline unsigned read_huffman_symbol(const struct huffman_table* table,
 	if (code == DEHUF_SLOW_PATH) {
 		return read_huffman_symbol_slow_path(table, source);
 	}
-
+		
 	read_bits(source, length);
 	return code;
 }
 
+static inline unsigned read_huffman_symbol(const struct huffman_table* table,
+                                           struct bit_source* source)
+{
+	possibly_refill(source, DEHUF_TABLE_BITS);
+	return read_huffman_symbol_no_refill(table, source);
+}
+
+// procedure EXTEND (figure F.12)
+
+// Fast lookup table for (1 << (bits - 1)).
+// The table actually helps, since the load can go in parallel with the shift
+// operation below.
+static const int bit_thresholds[16] = {
+	0, 1 << 0, 1 << 1, 1 << 2, 1 << 3, 1 << 4, 1 << 5, 1 << 6, 1 << 7, 1 << 8, 1 << 9, 1 << 10, 1 << 11, 1 << 12, 1 << 13, 1 << 14
+};
+
+static inline unsigned extend(int val, unsigned bits)
+{
+#if defined(__GNUC__) && (defined(__i386__) || defined(__x86_64__))
+	// GCC should ideally be able to figure out that the conditional move is better, but
+	// it doesn't for various reasons, and this is pretty important for speed, so we hardcode.
+	asm("cmp %1, %0 ; cmovl %2, %0"
+		: "+r" (val)
+		: "g" (bit_thresholds[bits]),
+ 		  "r" (val + (-1 << bits) + 1)
+		: "cc");
+	return val;
+#else
+	if (val < bit_thresholds[bits]) {
+		return val + (-1 << bits) + 1;
+	} else {
+		return val;
+	}
+#endif
+}
+
 #endif /* !defined(_DEHUFF_H) */