Re: Mi vs C (was calling RPG ILE with parameter list from MI)

Hi, Jon:

> Jon Paris wrote:
> This got me thinking. Has anyone here done any tests to compare the execution speed of MI programs vs comparable C programs? IBM doesn't write in MI any
> more and by sticking with MI you lose most of the optimization options that are available to ILE programs. The vast majority of MI instructions are surfaced in the
> C compiler and if I recall correctly are not called as functions (as they would be if called from RPG) but replaced by in-line code streams which should be a lot
> more efficient.

There are a large number of OPM MI instructions that are not available as ILE built-ins. As to performance, it depends on the algorithms used and how they are coded. In many cases, OPM MI can perform very well.

> C code has to be a lot more maintainable than MI and there are many more tools to work with it.

It is certainly possible to write very obtuse code in C, almost as cryptic and unreadable as some examples I have seen in APL.

> SO is there really any reason to use MI for new code? Does it really perform any better?

> Jon Paris
> www.Partner400.com
> www.SystemiDeveloper.com

There is still one issue that IBM has not yet resolved. If you make any syntax or semantic errors using any of the MI built-in functions in any ILE language, instead of getting a meaningful and helpful error message, you get a "dump" and an ugly message that says "compiler internal error - call support" and it opens up a "problem". For this reason, I use the OPM MI QPRCRTPG API to "debug" any MI code I write, before ever attempting to use equivalent embedded MI built-in functions, to make sure I have the arguments to the instructions coded correctly, etc. Until IBM "fixes" this, there will still be a need for the OPM MI compiler, IMHO.

Mark S. Waterbury