Thomas M. Turner's Tech Ramblings  

I recently updated my RegIf.exe program. Turns out I couldn't test the '(Default)' value of a key; how I never had tested that, I don't know.

Program: regif.exe

Type: Win32 console (UPX packed)

Purpose: Designed for use in batch files, to conditionally take actions based on values found, or not found, in the registry.

Current version: 1.3

We've been using SPSS for Windows, since late 1994, and some version on the VAX system before that. We've always used SPSS syntax files, almost entirely. While I personally have had nothing but grief getting SPSS 11.5, 11.51, and 12.0 to install properly in client/server mode, we've never had serious problems actually getting the product to run our syntax files and give us output... until recently.

Up until a few weeks ago, we had been running SPSS on some older Pentium 4s and Pentium IIIs. Then, in April, we finally purchased several new Dell OptiPlex GX270 computers and I first had problems imaging them with PowerQuest ImageCenter.

Here's what led me to the conclusion that Hyper-Threading enabled on our new Dell OptiPlex GX270 computers was causing syntax files in SPSS for Windows 12.0 to crash.

First, I thought it was SPSS version 12.0.1, since I had upgraded those having the problems, from version 11.51, along with their new computers. However, my office Dell Dimension 8200 ran the most problematic syntax file we had (and all others I've ever run on it) just fine, in version 12.0.1. Not long after getting the new computers, the SPSS 12.0.2 patch came out and so of course I tried it, but our problems continued.

One brief thought I had was somehow SPSS running on Windows XP was doing something strange, since they also went from using Windows 2000, when the new computers arrived. However, I subsequently ruled that out, by getting the most problematic syntax file to run fine on two non-GX270 systems that had Windows XP installed.

Most of my problem solving was in assuming the problem was in running our syntax and data off our Novell NetWare 6.0 file server. Now, we've been running our SPSS files off some NetWare file server since 1994, but it just seemed like there was some sort of server caching and/or file locking problem. Copying the syntax and data files to the local hard drive, off of the server, however, did not solve the problem.

Here is a screen shot of what happened when running a doomed syntax file. The SPSS syntax file crashes also weren't consistent between the various computers and users, which is the most frustrating thing for a tech support person trying to systematically reproduce and solve a problem. One of my earlier suggestions was to put more EXECUTE commands in the syntax files, to try to run code in sections. This helped somewhat, at least we think it did, but some syntax files would crash anyway, if not as often?
The apparent solution came when I reasoned out what critical difference there was between the computers that could run syntax files without crashing and these new Dell computers that were crash happy in SPSS. The only difference my tests hadn't seemed to rule out was that the new Dell OptiPlex GX270 computers had Hyper-Threading enabled, while all the older computers that ran the SPSS syntax properly did not have HT-capable CPUs.

Still not thinking this would help, but out of any other ideas, I disabled Hyper-Threading in the GX270 computer's BIOS and we ran the most problematic, crash-happy SPSS syntax file... and it didn't crash! We ran a couple of other smaller syntax files, still no crash. Ok, so far, so good, but what if I re-enable HT and tried it, would it go back to crashing? Sure enough, the SPSS syntax file crashed with Hyper-Threading enabled. Finally, I disabled Hyper-Threading, once more, and many syntax runs later, no more SPSS crashes on the Dell OptiPlex GX270 systems! Two days later and still no crashing, on the GX270 computer with HT disabled.