Editing APL (programming language) (section)

== Interpretation and compilation of APL ==
APL is traditionally an [[interpreted language]], having language characteristics such as [[Strong and weak typing|weak variable typing]] not well suited to [[compiler|compilation]].<ref>{{cite book |last1=Budd |first1=Timothy |title=An APL Compiler |publisher=Springer-Verlag |year=1988 |isbn=978-0-387-96643-4 |url=https://books.google.com/books?id=rTb2BwAAQBAJ&pg=PA1}}</ref> However, with arrays as its core data structure<ref>{{cite web |title=What is APL? |website=www.sigapl.org |publisher=SIGAPL |url=http://www.sigapl.org/about.php |access-date=January 20, 2015}}</ref> it provides opportunities for performance gains through [[Data parallelism|parallelism]],<ref>{{cite journal |last1=Ju |first1=Dz-Ching |last2=Ching |first2=Wai-Mee |title=Exploitation of APL data parallelism on a shared-memory MIMD machine |journal=Newsletter ACM SIGPLAN Notices |date=1991 |volume=26 |issue=7 |pages=61–72 |doi=10.1145/109625.109633 |s2cid=8584353}}</ref> [[parallel computing]],<ref>{{cite web |last1=Hsu |first1=Aaron W. |last2=Bowman |first2=William J. |title=Revisiting APL in the Modern Era |url=http://www.cs.princeton.edu/~dpw/obt/abstracts/obt12_submission_11.pdf |access-date=January 20, 2015 |archive-url=https://ghostarchive.org/archive/20221009/http://www.cs.princeton.edu/~dpw/obt/abstracts/obt12_submission_11.pdf |archive-date=2022-10-09 |url-status=live |website=www.cs.princeton.edu}}</ref><ref>{{cite journal |last1=Ching |first1=W.-M. |last2=Ju |first2=D. |title=Execution of automatically parallelized APL programs on RP3 |journal=IBM Journal of Research & Development |date=1991 |volume=35 |issue=5/6 |pages=767–777 |url=http://domino.research.ibm.com/tchjr/journalindex.nsf/c469af92ea9eceac85256bd50048567c/f892e104dfc4d0fd85256bfa0067fb42!OpenDocument |access-date=January 20, 2015 |doi=10.1147/rd.355.0767 |archive-date=March 3, 2016 |archive-url=https://web.archive.org/web/20160303231727/http://domino.research.ibm.com/tchjr/journalindex.nsf/c469af92ea9eceac85256bd50048567c/f892e104dfc4d0fd85256bfa0067fb42!OpenDocument |url-status=dead|url-access=subscription }}</ref> [[Massively parallel (computing)|massively parallel]] applications,<ref>{{cite journal |last1=Blelloch |first1=Guy E. |last2=Sabot |first2=Gary W. |title=Compiling Collection-Oriented Languages onto Massively Parallel Computers |journal=Journal of Parallel and Distributed Computing |volume=8 |issue=2 |citeseerx=10.1.1.51.5088 |pages=119–134 |quote=Collection oriented languages include APL, APL2 |year=1990 |doi=10.1016/0743-7315(90)90087-6}}</ref><ref>{{cite web |last1=Jendrsczok |first1=Johannes |last2=Hoffmann |first2=Rolf |last3=Ediger |first3=Patrick |last4=Keller |first4=Jörg |title=Implementing APL-like data parallel functions on a GCA machine |pages=1–6 |url=https://www.fernuni-hagen.de/imperia/md/content/fakultaetfuermathematikundinformatik/pv/97-08/papergca_09_1_.pdf |website=www.fernuni-hagen.de |access-date=January 22, 2015 |quote=GCA – Global Cellular Automation. Inherently massively parallel. 'APL has been chosen because of the ability to express matrix and vector' structures. |archive-url=https://web.archive.org/web/20150122211834/https://www.fernuni-hagen.de/imperia/md/content/fakultaetfuermathematikundinformatik/pv/97-08/papergca_09_1_.pdf |archive-date=January 22, 2015 |url-status=dead}}</ref> and [[very-large-scale integration]] (VLSI),<ref>{{cite book |last1=Brenner |first1=Norman |chapter=VLSI circuit design using APL with fortran subroutines |title=Proceedings of the international conference on APL - APL '84 |date=1984 |volume=14 |issue=4 |pages=77–79 |doi=10.1145/800058.801079 |isbn=0-89791-137-7 |quote=APL for interactiveness and ease of coding |s2cid=30863491}}<br/>{{*}}{{cite book |last1=Brenner |first1=Norman |date=1984 |title=Proceedings of the international conference on APL – APL '84 |chapter=VLSI circuit design using APL with fortran subroutines |publisher=ACM SIGAPL |isbn=978-0897911375}}</ref><ref>{{cite book |doi=10.1109/PACRIM.1989.48437 |quote=VLSI module generators are described. APL and C, as examples of interpreted and compiled languages, can be interfaced to an advanced graphics display. |chapter=Towards a graphics/Procedural environment for constructing VLSI module generators |title=Conference Proceeding IEEE Pacific Rim Conference on Communications, Computers and Signal Processing |pages=606–611 |year=1989 |last1=Gamble |first1=D.J. |last2=Hobson |first2=R.F. |s2cid=7921438}}</ref> and from the outset APL has been regarded as a high-performance language<ref>{{cite journal |last1=Lee |first1=Robert S. |title=Two Implementations of APL |journal=PC Magazine |date=1983 |volume=2 |issue=5 |page=379 |url=https://books.google.com/books?id=qURs4j9vKn4C&q=%22IBM%20APL%27s%20fast%20execution%22&pg=PA379 |access-date=January 20, 2015}}</ref> – for example, it was noted for the speed with which it could perform complicated matrix operations "because it operates on arrays and performs operations like matrix inversion internally".<ref>{{cite web |website=MARTHA and LLAMA |title=The APL Computer Language |url=http://marthallama.org/apl/ |access-date=January 20, 2015 |archive-date=February 13, 2015 |archive-url=https://web.archive.org/web/20150213004709/http://marthallama.org/apl/ |url-status=dead}}</ref>

Nevertheless, APL is rarely purely interpreted and compilation or partial compilation techniques that are, or have been, used include the following:

=== Idiom recognition ===
Most APL interpreters support [[Programming idiom|idiom]] recognition<ref>{{cite book |last1=Metzger |first1=Robert |last2=Wen |first2=Zhaofang |title=Automatic Algorithm Recognition and Replacement: A New Approach to Program Optimization |publisher=The MIT press |year=2000 |isbn=9780262133685 |url=https://books.google.com/books?id=u38h_fV3UqgC&pg=PA12 |access-date=May 6, 2018}}</ref> and evaluate common idioms as single operations.<ref>{{cite journal |last1=Snyder |first1=Lawrence |date=1982 |title=Recognition and Selection of Idioms for Code Optimization |journal=Acta Informatica |volume=17 |issue=3 |doi=10.1007/BF00264357 |s2cid=8369972}}</ref><ref>{{cite thesis |last1=Cheng |first1=Feng Sheng |date=1981 |title=Idiom matching: an optimization technique for an APL compiler |url=https://lib.dr.iastate.edu/cgi/viewcontent.cgi?article=7896&context=rtd |publisher=Iowa State University |access-date=May 6, 2018}}</ref> For example, by evaluating the idiom <code>BV/⍳⍴A</code> as a single operation (where <code>BV</code> is a Boolean vector and <code>A</code> is an array), the creation of two intermediate arrays is avoided.<ref>{{cite web |url=http://help.dyalog.com/16.0/Content/Language/Defined%20Functions%20and%20Operators/Idiom%20Recognition/Idiom%20Recognition.htm |title=Idiom Recognition |publisher=dyalog.com |access-date=May 6, 2018}}</ref>

=== Optimised bytecode ===
Weak typing in APL means that a name may reference an array (of any datatype), a function or an operator. In general, the interpreter cannot know in advance which form it will be and must therefore perform analysis, syntax checking etc. at run-time.<ref>{{cite journal |last1=Strawn |first1=George O. |date=March 1977 |title=Does APL really need run-time parsing? |journal=Software: Practice and Experience |volume=7 |issue=2 |pages=193–200 |doi=10.1002/spe.4380070207 |s2cid=1463012}}</ref> However, in certain circumstances, it is possible to deduce in advance what type a name is expected to reference and then generate [[bytecode]] which can be executed with reduced run-time overhead. This bytecode can also be optimised using compilation techniques such as [[constant folding]] or [[common subexpression elimination]].<ref name="DyalogCompiler">{{cite web|title=Compiler User Guide|url=http://docs.dyalog.com/16.0/Compiler%20User%20Guide.pdf |archive-url=https://ghostarchive.org/archive/20221009/http://docs.dyalog.com/16.0/Compiler%20User%20Guide.pdf |archive-date=2022-10-09 |url-status=live |website=www.dyalog.com|publisher=Dyalog Ltd.|access-date=May 7, 2018}}</ref> The interpreter will execute the bytecode when present and when any assumptions which have been made are met. Dyalog APL includes support for optimised bytecode.<ref name="DyalogCompiler"/>

=== Compilation ===
[[Compiler|Compilation]] of APL has been the subject of research and experiment since the language first became available; the first compiler is considered to be the Burroughs APL-700<ref name="DriscollOrth">
{{cite journal |last1=Driscoll |first1=Graham C. Jr. |last2=Orth |first2=Donald L. |s2cid=2299699 |date=November 1986 |title=Compiling APL: The Yorktown APL Translator |journal=IBM Journal of Research and Development |volume=30 |issue=6 |pages=583–593 |doi=10.1147/rd.306.0583}}</ref> which was released around 1971.<ref>{{cite web |url=http://www.sigapl.org/APLChronology.php |title=Chronology of APL |website=www.sigapl.org| publisher=ACM |access-date=May 7, 2018}}</ref> In order to be able to compile APL, language limitations have to be imposed.<ref name="DriscollOrth"/><ref>
{{cite journal |last1=Wai-Mee |first1=Ching |s2cid=17306407 |date=November 1986 |title=Program Analysis and Code Generation in an APL/370 Compiler|journal=IBM Journal of Research and Development |volume=30 |issue=6 |pages=594–602 |doi=10.1147/rd.306.0594}}</ref> APEX is a research APL compiler which was written by [[Robert Bernecky]] and is available under the [[GNU General Public License]].<ref>{{cite web |title= The APEX Project |url= http://www.snakeisland.com/apexup.htm}}</ref>

The [[STSC]] APL Compiler is a hybrid of a bytecode optimiser and a compiler – it enables compilation of functions to [[machine code]] provided that its sub-functions and globals are [[Declaration (computer programming)|declared]], but the interpreter is still used as a [[runtime library]] and to execute functions which do not meet the compilation requirements.<ref>{{cite web |title= APL Compiler (message from Jim Weigang to the comp.lang.apl Newsgroup) |url= http://www.chilton.com/~jimw/aplcomp.html |date= Apr 5, 1994}}</ref>