From 0f8516f861de5951fd06d70da763721c8f5d5c02 Mon Sep 17 00:00:00 2001
From: ceriel <none@none>
Date: Thu, 26 Oct 1989 16:16:35 +0000
Subject: [PATCH] use more simple hashing algorithm, made namelist generation
 dependant on a #define

---
 lang/cem/cemcom.ansi/LLlex.c   |  2 +-
 lang/cem/cemcom.ansi/idf.c     | 17 +----------------
 lang/cem/cemcom.ansi/idf.str   | 19 +++----------------
 lang/cem/cemcom.ansi/main.c    |  9 ++++++++-
 lang/cem/cemcom.ansi/replace.c |  2 +-
 lang/cem/cemcom.ansi/stack.c   |  6 ++++++
 6 files changed, 20 insertions(+), 35 deletions(-)

diff --git a/lang/cem/cemcom.ansi/LLlex.c b/lang/cem/cemcom.ansi/LLlex.c
index 70b169340..b5ca0e0e6 100644
--- a/lang/cem/cemcom.ansi/LLlex.c
+++ b/lang/cem/cemcom.ansi/LLlex.c
@@ -316,7 +316,7 @@ garbage:
 		do	{			/* read the identifier	*/
 			if (++pos < idfsize) {
 				*tg++ = ch;
-				hash = ENHASH(hash, ch, pos);
+				hash = ENHASH(hash, ch);
 			}
 			ch = GetChar();
 		} while (in_idf(ch));
diff --git a/lang/cem/cemcom.ansi/idf.c b/lang/cem/cemcom.ansi/idf.c
index edfe1414e..880bc5f27 100644
--- a/lang/cem/cemcom.ansi/idf.c
+++ b/lang/cem/cemcom.ansi/idf.c
@@ -137,7 +137,7 @@ str2idf(tg)
 	hash = STARTHASH();
 	while (++pos < idfsize && (ch = *cp++))	{
 		*ncp++ = ch;
-		hash = ENHASH(hash, ch, pos);
+		hash = ENHASH(hash, ch);
 	}
 	hash = STOPHASH(hash);
 	*ncp++ = '\0';
@@ -757,18 +757,3 @@ free_formals(fm)
 		fm = tmp;
 	}
 }
-
-char hmask[IDFSIZE];
-
-init_hmask()
-{
-	/*	A simple congruence random number generator, as
-		described in Knuth, vol 2.
-	*/
-	register int h, rnd = HASH_X;
-	
-	for (h = 0; h < IDFSIZE; h++)	{
-		hmask[h] = rnd;
-		rnd = (HASH_A * rnd + HASH_C) & HASHMASK;
-	}
-}
diff --git a/lang/cem/cemcom.ansi/idf.str b/lang/cem/cemcom.ansi/idf.str
index 640d933cb..e5fcdcda0 100644
--- a/lang/cem/cemcom.ansi/idf.str
+++ b/lang/cem/cemcom.ansi/idf.str
@@ -7,24 +7,11 @@
 
 #include	"nopp.h"
 
-/*	Since the % operation in the calculation of the hash function
-	turns out to be expensive, it is replaced by the cheaper XOR (^).
-	Each character of the identifier is xored with an 8-bit mask which
-	depends on the position of the character; the sum of these results
-	is the hash value.  The random masks are obtained from a
-	congruence generator in idf.c.
-*/
+#define	HASHSIZE	307	/* must be odd */
 
-#define	HASHSIZE	256	/* must be a power of 2 */
-#define	HASH_X		0253	/* Knuth's X */
-#define	HASH_A		77	/* Knuth's a */
-#define	HASH_C		153	/* Knuth's c */
-
-extern char hmask[];		/* the random masks */
-#define	HASHMASK		(HASHSIZE-1)	/* since it is a power of 2 */
 #define	STARTHASH()		(0)
-#define	ENHASH(hs,ch,ps)	(hs + (ch ^ hmask[ps]))
-#define	STOPHASH(hs)		(hs & HASHMASK)
+#define	ENHASH(hs,ch)		((hs << 2) + ch)
+#define	STOPHASH(hs)		((unsigned)hs % HASHSIZE)
 
 struct idf	{
 	struct idf *next;
diff --git a/lang/cem/cemcom.ansi/main.c b/lang/cem/cemcom.ansi/main.c
index b25f2197b..d52d600a3 100644
--- a/lang/cem/cemcom.ansi/main.c
+++ b/lang/cem/cemcom.ansi/main.c
@@ -83,7 +83,6 @@ main(argc, argv)
 	/* parse and interpret the command line options	*/
 	prog_name = argv[0];
 
-	init_hmask();
 #ifndef NOPP
 	inctable = (char **) Malloc(10 * sizeof(char *));
 	inctable[0] = "";
@@ -125,7 +124,9 @@ main(argc, argv)
 
 char *source = 0;
 
+#ifdef GEN_NM_LIST
 char *nmlist = 0;
+#endif GEN_NM_LIST
 
 compile(argc, argv)
 	char *argv[];
@@ -157,15 +158,21 @@ compile(argc, argv)
 	case 2:
 		destination = argv[1];
 		break;
+#ifdef GEN_NM_LIST
 	case 3:
 		nmlist = argv[2];
 		destination = argv[1];
 		break;
+#endif GEN_NM_LIST
 #endif	LINT
 
 	default:
 #ifndef	LINT
+#ifdef GEN_NM_LIST
 		fatal("use: %s source destination [namelist]", prog_name);
+#else GEN_NM_LIST
+		fatal("use: %s source destination", prog_name);
+#endif GEN_NM_LIST
 #else	LINT
 		fatal("use: %s source", prog_name);
 #endif	LINT
diff --git a/lang/cem/cemcom.ansi/replace.c b/lang/cem/cemcom.ansi/replace.c
index c800ae108..7fcc8dcd8 100644
--- a/lang/cem/cemcom.ansi/replace.c
+++ b/lang/cem/cemcom.ansi/replace.c
@@ -313,7 +313,7 @@ actual(repl)
 			do {
 				if (++pos < idfsize) {
 					*p++ = ch;
-					hash = ENHASH(hash, ch, pos);
+					hash = ENHASH(hash, ch);
 				}
 				ch = GetChar();
 			} while (in_idf(ch));
diff --git a/lang/cem/cemcom.ansi/stack.c b/lang/cem/cemcom.ansi/stack.c
index b7d5b703f..94d766176 100644
--- a/lang/cem/cemcom.ansi/stack.c
+++ b/lang/cem/cemcom.ansi/stack.c
@@ -179,7 +179,9 @@ unstack_world()
 	lint_global_level(local_level);
 #endif	LINT
 
+#ifdef GEN_NM_LIST
 	open_name_list();
+#endif GEN_NM_LIST
 
 	while (se)	{
 		register struct idf *idf = se->se_idf;
@@ -227,8 +229,10 @@ unstack_world()
 		    && !def->df_initialized) {
 			/* space must be allocated */
 			bss(idf);
+#ifdef GEN_NM_LIST
 			if (def->df_sc != STATIC)
 				namelist(idf->id_text);	/* may be common */
+#endif GEN_NM_LIST
 			def->df_alloc = ALLOC_DONE;	/* see Note below */
 		}
 		se = se->next;
@@ -247,6 +251,7 @@ unstack_world()
 	*/
 }
 
+#ifdef GEN_NM_LIST
 /*	A list of potential common names is kept, to be fed to
 	an understanding loader.  The list is written to a file
 	the name of which is nmlist.  If nmlist == NULL, no name
@@ -269,3 +274,4 @@ namelist(nm)
 		sys_write(nfp, "\n", 1);
 	}
 }
+#endif GEN_NM_LIST