sankar@Neon.Stanford.EDU (Sriram Sankar) (03/09/90)
I am trying to evaluate different PC compilers and intend to eventually buy one. For some reason, I get the feeling that Meridian has a good compiler, albeit with a few problems. I'd like feedback from you on the Meridian compiler and also how it relates to other PC compilers. I'll summarize the responses that I receive. I have had some experience with Meridian's 2.? basic compiler and a beta version of their 4.0 extended compiler. I have conducted some tests which seem to indicate that code generated for an AT 386 25Mhz machine by the Meridian compiler for a program with a *lot* of tasks and task communication runs *much* faster than code generated by the VADS compiler for a Sequent with the same kind of processor. I can attribute a part of this to the UNIX overhead, but not all. On the down side, the Meridian compilers I have used seem to have many bugs because of which it is nearly impossible to get it to compile a large Ada program and get it running without a whole lot of workarounds. Some programs even refuse to run after all this. Then there are other problems such as the limit on the number of tasks due to limits on the size of the total stack space, and that I have had very little success in compiling programs using the debugger option (to load the debugger). One of my benchmarks is an approx. 10000 line (+ 5000 lines of assembly) windowing system for the AT's EGA and VGA displays. I have also tried to get UC Irvine's alex program and most of our many lines of Ada code at Stanford to run with little success. The windowing system (which is one of my smaller benchmarks) runs. I am sure that Meridian's response to all this is that their latest compiler does not have these bugs anymore, and maybe they are correct to a large extent. But I do not wish to buy this compiler and then find that it still has a lot of these bugs. So, does anyone have experience with the Meridian compiler with large programs? Is there anyone in the San Francisco area who would be willing to let me run some benchmarks on their Meridian compiler? I will post a summary of all responses and also what I find out about this compiler and other compilers for the PC-AT. Thanks in advance. Sriram Sankar Research Associate ERL449, Computer Systems Laboratory Stanford University Stanford, CA 94305 (415)723-1835 Email: sankar@neon.stanford.edu.