bayes@hpislx.HP.COM (Scott Bayes) (04/06/90)
I wanted to thank you for the effort you obviously put into this post. I had known most of the techniques you explained, but many of them have grown rusty with misuse. Sometimes they don't even pop out of the bag of tricks when they'd be appropriate, due to nneglect. A couple of paragraphs near the end caught my notice: > THINK Pascal 2.0 won't let you take the address of a nested routine. > I don't know if this restriction has been lifted in 3.0. Do they give a reason for the restriction? Is it perhaps the dangling static link problem? Just too hard too explain? > The Bind routine is generating code at run-time. This *may* cause > problems with the 68020/030(/040?) instruction cache, if you're making > really frequent calls to Bind (which I've never done). Personally, > I think the '020/030 cache is brain-damaged, and should be fixed > so that it invalidates properly on data writes. Do you mean use the f0-f2 lines as a tag, invalidating any cache cells that show up with 2 different f-tags? Sounds like it would obviate the need for flush-after-store for instructions. Anyway, thank you for a fine posting. Scott Bayes Hewlett-Packard Company MSO Tech Support