martin@CITI.UMICH.EDU (12/14/88)
I have completed and tested some enchancements to the PURDUE patches posted earlier by Gene Spafford. They are packaged in the same format as the purdue patches and are diffed against the purdue patches. (e.g. You need to have purdue installed to apply the purdue+ patches) I will put this shar file onto the expo machine at: /contrib/PurduePlus-speedups.mfb.Z Here is something describing the nature of these changes. About Purdue+ ------------- These are some simple yet blinding modifications to the Purdue speedups for X.V11R3. The speedups effect only machines using the mc68020 or VAX processor and even then only if gcc is used (the GNU C Compiler is available from prep.ai.mit.edu). - The main idea behind these speedups is to use a single bitfield (asm) instruction for all getbits and putbits calls. This is an idea implemented first by tmdonahu@athena.mit.edu on the VAX, I have extended this idea to the mc68020 processor bitfield instructions. - Also included in purdue+ are some of the more complex enchancements made by Donahue to the mfb routines. These were changes made to mfbtegblt.c, mfbbitblt.c and mfbbres.c . The changes to mfbtegblt.c (Terminal Emulator Glyphblt) are the most dramatic of these changes. In a nut shell, if the font is <= 8 pixels wide, then four of them are or'ed together into a single word and they are (all four) put onto the screen with one call. This helps xterm and xemacs when using the smaller fonts. (6x13, 8x13, etc..) - Also included is an obvious bug fix to the miarc.c file which only seems to surface badly when the code is compiled with GCC. I have sent this bug to the xbugs@expo list, but have not seen a (blessed) fix for it yet. Martin Friedmann Martin@citi.umich.edu P.s. The network is acting funny, so I will put these things on the net later tonight.