alf: Cleanup and commentary.

4 files changed, 26 insertions, 13 deletions

diff --git a/src/alf.1 b/src/alf.1
index 5a3df8f..46a1ef2 100644
--- a/src/alf.1
+++ b/src/alf.1
@@ -27,7 +27,7 @@ level margin: \\n[rst2man-indent\\n[rst2man-indent-level]]
 .\" new: \\n[rst2man-indent\\n[rst2man-indent-level]]
 .in \\n[rst2man-indent\\n[rst2man-indent-level]]u
 ..
-.TH "ALF" 1 "2025-12-11" "0.4.0" "Urchlay's Atari 8-bit Tools"
+.TH "ALF" 1 "2025-12-12" "0.4.0" "Urchlay's Atari 8-bit Tools"
 .SH NAME
 alf \- create Atari 8-bit ALF archives
 .\" RST source for alf(1) man page. Convert with:
@@ -271,8 +271,8 @@ a DOS capable of handling multi\-megabyte files...
 Performance is pretty good, as of alftools\-0.4.0. For small files
 like you\(aqd use on an Atari (up to 50KB), it\(aqs basically instantaneous
 (0.008 seconds) on the author\(aqs modest i7 workstation. For a 1MB text
-file, it takes 0.026 sec (faster than \fBarc\fP!). For 1MB of random
-garbage, it\(aqs 0.043 sec (and the resulting ALF file is 36% larger than
+file, it takes 0.022 sec (faster than \fBarc\fP!). For 1MB of random
+garbage, it\(aqs 0.03 sec (and the resulting ALF file is 36% larger than
 the garbage).
 .sp
 By comparison, \fBzip\fP takes 0.06 seconds to compress the 1MB text file,
diff --git a/src/alf.rst b/src/alf.rst
index dbee22b..0c72aa2 100644
--- a/src/alf.rst
+++ b/src/alf.rst
@@ -238,8 +238,8 @@ Performance
 Performance is pretty good, as of alftools-0.4.0. For small files
 like you'd use on an Atari (up to 50KB), it's basically instantaneous
 (0.008 seconds) on the author's modest i7 workstation. For a 1MB text
-file, it takes 0.026 sec (faster than **arc**\!). For 1MB of random
-garbage, it's 0.043 sec (and the resulting ALF file is 36% larger than
+file, it takes 0.022 sec (faster than **arc**\!). For 1MB of random
+garbage, it's 0.03 sec (and the resulting ALF file is 36% larger than
 the garbage).
 
 By comparison, **zip** takes 0.06 seconds to compress the 1MB text file,
diff --git a/src/bytorder.h b/src/bytorder.h
index cb8231e..2551950 100644
--- a/src/bytorder.h
+++ b/src/bytorder.h
@@ -2,13 +2,14 @@
 #  error Cannot define both ALF_LSB_FIRST and ALF_MSB_FIRST
 #endif
 
+/* user defined one or the other in COPT, use it. */
 #if defined(ALF_LSB_FIRST) || defined(ALF_MSB_FIRST)
 #  define ALF_ENDIAN_OK
 #endif
 
 /* try to get the byte order on this platform.
    if we succeed, either ALF_LSB_FIRST or ALF_MSB_FIRST will
-   defined, and so will ALF_ENDIAN_OK. */
+   be defined, and so will ALF_ENDIAN_OK. */
 
 #ifndef ALF_ENDIAN_OK
 #  if defined(__BYTE_ORDER__)
@@ -24,12 +25,13 @@
 #  endif
 #endif
 
+/* if the __BYTE_ORDER__ macro wasn't defined... */
 #ifndef ALF_ENDIAN_OK
-#  if defined(__LITTLE_ENDIAN__)
+#  if defined(__LITTLE_ENDIAN__ || defined __LITTLE_ENDIAN || defined LITTLE_ENDIAN)
 #    define ALF_LSB_FIRST
 #    define ALF_ENDIAN_OK
 #  else
-#    if defined(__BIG_ENDIAN__)
+#    if defined(__BIG_ENDIAN__ || defined __BIG_ENDIAN || defined BIG_ENDIAN)
 #      define ALF_MSB_FIRST
 #      define ALF_ENDIAN_OK
 #    endif
diff --git a/src/crunch.c b/src/crunch.c
index f572192..2755248 100644
--- a/src/crunch.c
+++ b/src/crunch.c
@@ -30,6 +30,7 @@ unsigned int input_len, output_len, out_bitpos;
 short tokens[MAX_TOKENS][256];
 
 int token_bits;
+int shiftamt; /* precalculated MAX_BITS - token_bits */
 int max_token;
 int curr_token = INIT_TOKEN;
 int in_pos;
@@ -121,6 +122,9 @@ void init_table(void) {
 	token_bits = INITIAL_BITS;
 	max_token = 1 << INITIAL_BITS;
 	curr_token = INIT_TOKEN;
+
+	/* precalculated for store_token() */
+	shiftamt = MAX_BITS - token_bits;
 }
 
 void check_output_len(void) {
@@ -136,9 +140,9 @@ void inc_output_len(void) {
 	output_buf[output_len] = 0;
 }
 
-#ifdef ALF_ENDIAN_OK
+#if !defined(APPEND_BIT) && defined ALF_ENDIAN_OK
 
-/* This is 20% faster, but it requires knowing the endianness of
+/* This is 25% faster, but it requires knowing the endianness of
    the platform at compile time. See bytorder.h for gory details. */
 union { unsigned int ui; u8 bytes[4]; } ui2bytes;
 
@@ -154,11 +158,17 @@ void store_token(int tok) {
 		putchar('\n');
 	}
 
-	tok <<= (MAX_BITS - token_bits);
-	ui2bytes.ui = tok << (12 - out_bitpos);
+	/* align token so, no matter what its size (9 thru 12),
+	   its top bit is at bit 23 of a 32-bit int. */
+	ui2bytes.ui = tok << ((12 - out_bitpos) + shiftamt);
+
+	/* always append 12 bits (it's quicker to do that than have
+	   conditionals to decide whether the last byte is needed) */
 	output_buf[output_len] |= ui2bytes.bytes[HIBYTE];
 	output_buf[output_len + 1] = ui2bytes.bytes[MIDBYTE];
 	output_buf[output_len + 2] = ui2bytes.bytes[LOBYTE];
+
+	/* update based on actual token size */
 	out_bitpos += token_bits;
 	output_len += out_bitpos / 8;
 	out_bitpos %= 8;
@@ -227,11 +237,12 @@ void make_token(short tok, u8 chr) {
 				dump_tokens();
 			}
 			store_token(TOK_RESET); /* stored at the *old* token size! */
-			token_bits = INITIAL_BITS;
 			init_table();
 			return; /* since we're starting over, *don't* make a token */
 		} else {
 			token_bits++;
+			/* precalculated for store_token() */
+			shiftamt--;
 			if(opt_verbose > 1) {
 				printf("token_bits increased to %d\n", token_bits);
 			}