mph@behemoth.phx.mcd.mot.com (Mark Huth) (05/31/89)
Okay, I need some help from this wonderful resource we call the 'Net. I have the following configuration: 2000, 2620 (2 Meg), 2052, 2090, 40 Meg Mini-scribe SCSI, 20 Meg Seagate Winchester. 20 Meg has small initial partition, and the rest FFS. 40 Meg has floppy-size initial partition, and 2 20 Meg FFS partitions. Running released 1.3 Workbench with 1.2 ROMs. Also, Dmouse, RSLClock, and CA Shell. Startup does FastMemFirst. While running the '020 I decided to do some benchmarks with the Lattice C 5.02 compiler. The benchmark was simply to compile Stevie that was recently posted to the net. The makefile works great under LMK. Whis would seem to be a simple matter, but I ran into a number on non-reproducible crashes. The QUAD: device is assigned to RAM: #1) Run the make with includes and compiler on one disk and the source on another, only QUAD: in ramdisk. This usually works (takes 10::52 to complete) but not always - at least twice, this guru'ed with type 3 errors (address error - on '020 this can only occur if the destination of a branch or jmp instruction is odd). After reboot, re-running lmk from where it left off worked just fine and the make completed. #2) Run the make with the includes in RAM:include - this also completed about 50% of the time - time was 9::36. Guru'ed once. #3) Add the source to the RAM:stevie directory and cd to it. Went real fast, but never completed benchmark - noticed a number of symptoms - compiler would lock up in the Optimizer, but other tasks would be okay - twice the screen was blank (by dmouse) and the system was totally locked - may have been in an alert with the screen turned off) two other times it guru'ed with the usual type 3 or type 4 error. Once I noticed that the source files in RAM;stevie had been corrupted. #4) Remove dmouse and RSLClock from system. Retry RAM:include test - completes. Okay, now I go to display the timer files (using and more) and - crash direct to reset - reboot not successful. Needs three fingers. #5) Reboot on 68000 - run disk (24::57), ram includes (24::23) and ram source (22::48) benchmarks without a hitch. Do I have a hardware problem??? Has anyone else seen these types of non-reproducible or complex failures?? Helpful hints? I should mention that the 2620 has spent several nights running mandelbrot traces without crashing. Also, it stays up for days at a time doing nothing when I'm not home and I have never found it crashed. Is Lattice C 5.02 doing something subtle to the memory? Thanks for your input, Mark Huth