donn@sdchema.UUCP (03/27/84)
The posted fix is for a bug which causes all the variables in a given COMMON block to be treated as one variable for the purpose of common subexpression elimination. This fix should solve many of the weird problems which have been reported recently. By the way, if you are running the mkargtemp() fix which Bob Corbett made to repair the bug that caused DO loop limits to get clobbered, I have some important bug fixes for that too. Bob's fix is running on at least some machines at Berkeley. I will send my fixes to Ralph Campbell at Berkeley; if you have mkargtemp() but aren't at Berkeley, send me mail. The bugs occur regardless of whether the optimizer is on and they are as nasty as the CSE bug mentioned above... Presently I will post a working version of mkargtemp(). I will also post the changes made by Ralph Campbell to speed up references to local variables, and (if I ever manage to get it done) my own fix to speed up references to COMMON variables. Revising my resume, Donn Seeley UCSD Chemistry Dept. ucbvax!sdcsvax!sdchema!donn 32 52' 30"N 117 14' 25"W (619) 452-4016 sdcsvax!sdchema!donn@nosc.ARPA