Eric Anholt wrote:

Hi Eric,

thanks for the patch! I've checked it in with one small change: I've left out the hunk in Float.c below.

However, I'm still seeing insanely long times to run simple tests like DoublePrint.java. Not sure what's going on here.

you could try kaffe -prof DoublePrint with an installed version, and then look into the profiling output. I'm not sure whether that works on amd64, but it may be worth trying out.


Index: libraries/clib/native/Float.c
===================================================================
RCS file: /cvs/kaffe/kaffe/libraries/clib/native/Float.c,v
retrieving revision 1.14
diff -u -r1.14 Float.c
--- libraries/clib/native/Float.c       21 May 2004 15:19:24 -0000      1.14
+++ libraries/clib/native/Float.c       21 Oct 2004 07:56:25 -0000
@@ -144,7 +145,7 @@
        }
 #endif
 /* Parse value; check for empty parse */
-       value = (float) strtod(buf, &endbuf);
+       value = strtof(buf, &endbuf);
        if (endbuf == buf) {
                msg = startbuf;         /* this is what JDK 1.1.6 does */
                goto bail;

If one uses strtof here, the resulting strings are not correct, afaict, for exponent-less floats.


cheers,
dalibor topic

_______________________________________________
kaffe mailing list
[EMAIL PROTECTED]
http://kaffe.org/cgi-bin/mailman/listinfo/kaffe

Reply via email to