bs@sbsvax.UUCP (Bernard Sieloff) (08/08/88)
Subject: [n]eqn(1) speedup Description: This is no *real* bug, but it chews unnecessarily a lot of CPU time, if you have some BIG [n]eqn input files. Input to eqn with no effect to eqn itself will be passed unaltered to stdout, but by calling printf()! Eqn behavior is heavily sped up, if the "passing"-printf() calls are changed to fputs() calls. Fix: Apply the following diff with Larry Wall's "patch" program. The concerned file is io.c in eqn's source directory. *** io.c.orig Sun Sep 4 00:19:07 1983 --- io.c Sun Aug 7 20:39:03 1988 *************** *** 32,38 **** eqline = linect; if (in[0]=='.' && in[1]=='E' && in[2]=='Q') { for (i=11; i<100; used[i++]=0); ! printf("%s",in); printf(".nr 99 \\n(.s\n.nr 98 \\n(.f\n"); markline = 0; init(); --- 32,38 ---- eqline = linect; if (in[0]=='.' && in[1]=='E' && in[2]=='Q') { for (i=11; i<100; used[i++]=0); ! fputs(in, stdout); /* [bs] */ printf(".nr 99 \\n(.s\n.nr 98 \\n(.f\n"); markline = 0; init(); *************** *** 58,64 **** else if (type == lefteq) inline(); else ! printf("%s",in); } return(0); } --- 58,64 ---- else if (type == lefteq) inline(); else ! fputs(in, stdout); /* [bs] */ } return(0); } -- Bernard Sieloff | UUCP: ...!uunet!unido!sbsvax!bs Universitaet des Saarlandes | or bs@sbsvax.UUCP FB 10 - Informatik (Dept. of CS) | CSNET: bs%sbsvax.uucp@Germany.CSnet Bau 36, Im Stadtwald 15 | ARPA: bs%sbsvax.uucp@uunet.UU.NET D-6600 Saarbruecken 11, W. Germany | Phone: +49 681 302 2434