Execute HQ9+: Difference between revisions

So, there is a careful factorisation of the text phrases into FORMAT and WRITE statements. Note that "free-format" output (as with <code>WRITE (6,*)</code>) starts in the second column, whereas formatted output starts in the first column. Inspection of the code file HQ9.exe shows that the compiler has recognised that the multiple appearances of the text literals "bottles" (three) and "bottle" (two) are the same and there is only one value of each constant in the code file. However, it has not noticed that the text "bottle" can be extracted from "bottles", which could in turn be found within a larger text literal "No bottles of beer on the wall" which also contains the subsequence " on the wall" - perhaps the code to do this would consume more space than would be saved by having a single multiple-use text constant in the code for those, or perhaps the problem is just too difficult in general to be worth the effort of devising and executing a worthwhile analysis, given that justonly a few bytes might be saved in a code file of 480Kb. This of course must contain the format interpretation subsystem and so forth, not just the code for the Fortran source. Even so, this program (with minor changes to the syntax) could be written in Fortran IV for an IBM1130, and would run in a computer with a total memory size of 8Kb. On such systems, much thought would go in to minimising space lost to verbose texts and good exposition as well as such reuse opportunities: gaining access to 32Kb or even 64Kb systems would be a great relief. But these days, memory space is not at a premium, and we are told that modern compilers produce excellent code.