z13 added many SIMD (vector) instructions for both binary arithmetic and strings and most support from 8b to 128b wide operands. The compilers were updated to use SIMD for some functions, including binary integer and string; this was for PLI 4.5. You need PLI 5.3 (newest compiler) for ARCH(13) support. I've never investigated how extensive the SIMD exploitation was as retrofitted into PLI 4.5. I believe PLI 5.3 supports FIXED BIN(127,0) within the limits of the hardware -- for example there is 128 bit vector add/subtract with carry, and also 128b to 8b arithmetic, but not some of the other variants. This varies between z13, z14 and z15.
On Thu, Sep 10, 2020 at 7:54 AM Seymour J Metz <sme...@gmu.edu> wrote: > I was asking whether it supported vector instructions for FIXED, not for > FLOAT.; in particular, whether it supported FIXED BIN(127,0). > > > -- > Shmuel (Seymour J.) Metz > http://mason.gmu.edu/~smetz3 > > > ________________________________________ > From: IBM Mainframe Discussion List <IBM-MAIN@LISTSERV.UA.EDU> on behalf > of Attila Fogarasi <fogar...@gmail.com> > Sent: Wednesday, September 9, 2020 5:44 PM > To: IBM-MAIN@LISTSERV.UA.EDU > Subject: Re: PL/I support of vector instructions? > > There is no source change as the DFP option applies only to DEC FLOAT and > not to FIXED BIN. The data element storage sizes remain the same, but with > DFP the DEC FLOAT items become decimal floating point data and instructions > (and BIF) while with NODFP the DEC FLOAT is HEXADECIMAL float (as distinct > from the older BINARY float, which is not IEE754 compliant). Confusing > having 3 incompatible and numerically incongruent floating point formats on > the same machine, and only DFP should be used now as it is both faster. (on > all supported z hardware today), more accurate, and more storage dense. To > IBM's credit they have the best IEE754 implementation and went with the > expensive hardware option. > As for the vector facility, currently there are 32 vector registers, 128 > bit wide. The z14 extensions added DFP to vector and also quad-precision > capability. You do need to set multiple compiler options correctly to get > vector code generated :) > > On Thu, Sep 10, 2020 at 12:22 AM Seymour J Metz <sme...@gmu.edu> wrote: > > > No source change? Certainly the compiler could interpret FIXED BIN as > > FIXED BIN(127,0), but if I've coded, e.g., FIXED BIN(60,3), I would hope > > that the compiler would only use a doubleword unless and until I changed > > the source and recompiled. > > > > > > -- > > Shmuel (Seymour J.) Metz > > http://mason.gmu.edu/~smetz3 > > > > > > ________________________________________ > > From: IBM Mainframe Discussion List <IBM-MAIN@LISTSERV.UA.EDU> on behalf > > of Attila Fogarasi <fogar...@gmail.com> > > Sent: Wednesday, September 9, 2020 2:14 AM > > To: IBM-MAIN@LISTSERV.UA.EDU > > Subject: Re: PL/I support of vector instructions? > > > > Enterprise PL/I 5.2 supports ARCH(12) so has your desired vector > > instruction support. For completeness, so do Cobol 6.2, XL/C 2.3 and > Java > > 8.5. Again no source code change in PL/I is needed, just recompile with > > ARCH(12) option. Really sad that IBM doesn't publicize these features > > better.... vector usage can cut cpu and elapsed time dramatically (I've > > seen 80% reduction for intensive programs). Great reason to upgrade the > > hardware and compilers :) > > > > On Wed, Sep 9, 2020 at 3:59 PM Attila Fogarasi <fogar...@gmail.com> > wrote: > > > > > For greater precision Enterprise PLI supports the hardware Decimal > > > Floating Point which is enabled with the compiler option of FLOAT(DFP), > > > this then changes all the DECIMAL FLOAT data declarations to be DFP > > > allowing maximum 34 digits (and exact decimal precision, important for > > > financial applications). The PL/I language itself is unchanged, so > > > existing source code switches to using DFP with a recompile (though > > changes > > > in DCL might be desirable). Vector DFP is part of ARCH(12) which came > > with > > > z14. So Enterprise PLI fully supports it already, best I can tell, but > > > doesn't highlight such support in the doc. I do DFP in assembler as I > > > couldn't wait for the compilers to add support and also want finer > > control > > > (the devil is in the details). > > > > > > On Wed, Sep 9, 2020 at 2:53 PM Seymour J Metz <sme...@gmu.edu> wrote: > > > > > >> Has IBM announced an intention to support the vector instructions to > > >> allow more precision for FIXED DEC and FIXED BIN in PL/I? Are there > > other > > >> vendors that support greater precision? > > >> > > >> > > >> -- > > >> Shmuel (Seymour J.) Metz > > >> http://mason.gmu.edu/~smetz3 > > >> > > >> ---------------------------------------------------------------------- > > >> For IBM-MAIN subscribe / signoff / archive access instructions, > > >> send email to lists...@listserv.ua.edu with the message: INFO > IBM-MAIN > > >> > > > > > > > ---------------------------------------------------------------------- > > For IBM-MAIN subscribe / signoff / archive access instructions, > > send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN > > > > ---------------------------------------------------------------------- > > For IBM-MAIN subscribe / signoff / archive access instructions, > > send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN > > > > ---------------------------------------------------------------------- > For IBM-MAIN subscribe / signoff / archive access instructions, > send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN > > ---------------------------------------------------------------------- > For IBM-MAIN subscribe / signoff / archive access instructions, > send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN > ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
