On 20 September 2013 09:28, Farley, Peter x23353 <peter.far...@broadridge.com> wrote: > Tim, I have to disagree in part with the statement you made below that "... > it couldn't be avoided". It most certainly *could* have been avoided. > > It may have been a "... reasonable, responsible technical choice" from IBM's > more-than-occasionally myopic point of view, but not necessarily the best > choice, either technically or politically. A better choice, IMHO, would have > been to use the XL C/C++ back end which supports not only load-module output > but also supports architectures back to before the G1 series (the XL C/C++ > documentation for ARCH(0) says: "Produces code that is executable on all > models."). All the benefits of newer-architecture instructions and advanced > optimization algorithms would have been made available without *requiring* > program-object output and thus PDSE storage.
At risk of going off on a bit of a tangent, I've been wondering quite a bit about this choice. And more fundamentally, why there are two current code generator back-ends in IBM with obviously overlapping function. And I think they even come from the same lab in Toronto. I had imagined the key feature of a Java back end to be Just In Time compiling, because in general JVM classes cannot be pre compiled, but need to be translated on the fly, often in small chunks yet with good performance, so that a tiny function doesn't have a big startup cost when called, yet its performance can improve with "experience". A COBOL back-end - like one for C or PL/I or FORTRAN or any traditional language - would presumably want to generate the best possible code with the luxuries of lots of available compile-time resources, cross-procedure/module optimization, and so on. Now these goals are obviously not incompatible, indeed they overlap hugely, but I wonder if this is just the latest in a decades-long series of IBM internal competitions where both products are allowed to get out the door before one is eventually killed off. And if the Java one is winning the new business, as it were, is the C one going to be kicked out of C and PL/I in due time? Tony H. ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
