=E2=80=94

For = way too many people, freedom=E2=80=99s just another word = for =E2=80=9CI should be able to do whatever I = want irrespective of how it affects others and not face = consequences in any = way.=E2=80=9D

<= /div>

= --Apple-Mail=_43C18022-AF78-47FD-9A70-2B6F9C372AEA-- From owner-chemistry@ccl.net Wed May 19 13:54:01 2021 From: "Chris Swain swain _ mac.com" To: CCL Subject: CCL: M1 port/fortran Message-Id: <-54331-210519135234-22755-0g5mx2hleO7IR4CVkNFxww[#]server.ccl.net> X-Original-From: Chris Swain Content-Transfer-Encoding: 7bit Content-Type: multipart/alternative; boundary=Apple-Mail-6EBE288F-B6AC-444A-A72B-52DA0C6AECAD Date: Wed, 19 May 2021 18:52:15 +0100 Mime-Version: 1.0 (1.0) Sent to CCL by: Chris Swain [swain{=}mac.com] --Apple-Mail-6EBE288F-B6AC-444A-A72B-52DA0C6AECAD Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: quoted-printable Might be useful? https://github.com/fxcoudert/gfortran-for-macOS/releases/tag/11-arm-alpha1 Sent from Chris's iPhone > On 19 May 2021, at 18:48, Joe Leonard jleonard42]-[gmail.com wrote: >=20 > =EF=BB=BFFolks, have any groups been successful getting Fortran codes (new= or old) ported to the new chipset? If so, which compiler did they use and h= ow did they handle any numerical instability they encountered? We are just s= tarting into this (NAG compiler) but are seeing instability. I was told the= re was a legacy compiler switch, and recall old Fortran code assumed variabl= es started at 0 (hence a -static flag back then)=E2=80=A6 Are there other p= ertinent switches that have been found? >=20 > Thanks in advance! >=20 > Joe Leonard >=20 > =E2=80=94 > For way too many people, freedom=E2=80=99s just another word for =E2=80=9C= I should be able to do whatever I want irrespective of how it affects others= and not face consequences in any way.=E2=80=9D >=20 --Apple-Mail-6EBE288F-B6AC-444A-A72B-52DA0C6AECAD Content-Type: text/html; charset=utf-8 Content-Transfer-Encoding: quoted-printable Might be useful?

https://github.com/fxcoudert/gfortran-for-macOS/releases/tag/11-arm-alph= a1

Sent from Chris's iPhone

On 19 May 2021, at 18:48, Joe Leonard jleona= rd42]-[gmail.com <owner-chemistry!^!ccl.net> wrote:

=EF=BB=BFFolks, have any groups= been successful getting Fortran codes (new or old) ported to the new chipse= t? If so, which compiler did they use and how did they handle any nume= rical instability they encountered? We are just starting into this (NA= G compiler) but are seeing instability. I was told there was a legacy c= ompiler switch, and recall old Fortran code assumed variables started at 0 (= hence a -static flag back then)=E2=80=A6 Are there other pertinent swi= tches that have been found?

Thanks in advance!

Joe Leonard

=
=
=
=
=E2=80=94
For way too many peopl= e, freedom=E2=80=99s just another word for =E2=80=9CI should b= e able to do whatever I want irrespective of how it affects others and&= nbsp;not face consequences in any way.=E2=80=9D
<= /div>
<= /div>
<= /div>

= --Apple-Mail-6EBE288F-B6AC-444A-A72B-52DA0C6AECAD-- From owner-chemistry@ccl.net Wed May 19 16:33:00 2021 From: "Ben Webb ben%a%salilab.org" To: CCL Subject: CCL: M1 port/fortran Message-Id: <-54332-210519153518-6880-FTT/jzTaM4V1Jhu/A8Neyw]|[server.ccl.net> X-Original-From: Ben Webb Content-Language: en-US Content-Transfer-Encoding: 7bit Content-Type: text/plain; charset=utf-8; format=flowed Date: Wed, 19 May 2021 12:35:07 -0700 MIME-Version: 1.0 Sent to CCL by: Ben Webb [ben/./salilab.org] On 5/19/21 9:16 AM, Joe Leonard jleonard42]-[gmail.com wrote: > Folks, have any groups been successful getting Fortran codes (new or > old) ported to the new chipset? No issues for us porting MODELLER (https://salilab.org/modeller/) to M1, although by Fortran standards it's not super-old (it's Fortran 90). We used the gfortran that's part of the gcc homebrew package (https://brew.sh/). We see about a 20% performance improvement with gfortran-10 on a 2020 Mac Mini (M1) compared to Intel Fortran on a 2018 Mac Mini (Intel). Ben -- ben{}salilab.org https://salilab.org/~ben/ "It is a capital mistake to theorize before one has data." - Sir Arthur Conan Doyle From owner-chemistry@ccl.net Wed May 19 17:08:00 2021 From: "Michel Petitjean petitjean.chiral%a%gmail.com" To: CCL Subject: CCL: M1 port/fortran Message-Id: <-54333-210519160449-12649-wzHwJ910l/xuGMiz83G3GA/./server.ccl.net> X-Original-From: Michel Petitjean Content-Transfer-Encoding: 8bit Content-Type: text/plain; charset="UTF-8" Date: Wed, 19 May 2021 22:04:32 +0200 MIME-Version: 1.0 Sent to CCL by: Michel Petitjean [petitjean.chiral]_[gmail.com] Dear Joe, Variables start at ...nobody can predict, except if you initialize them: never assume a default value of zero. The static flag is a linker option, it has nothing to deal with instabilities (see https://www.nag.com/nagware/np/r62_doc/manual/compiler_2_4.html) Instabilities are first a matter of numerical analysis and mathematical programming. The programmer must have skills in mathematics, but also a good understanding about how reals (float, double) are represented on computer, the operations performed on them, compiler library routines (sqrt, log, etc.), what mean underflow and overflow exceptions for reals, and so on. All that is machine and compiler dependent. Producing a code insensitive to numerical instabilities is a very difficult task. In the case of complex algorithms, it may even be an impossible task. In general, instabilities cannot be avoided: it depends on the input data. Consider a very large input real (this real can itself be transmitted by an other routine), take its square then the inverse of this square, or take its inverse then square it (well, the optimizer may change the order of these two operations): which output can be expected? What should be written in the user documentation? I am not experienced on Mac M1. May be you should give more information about your instabilities problem, so that somebody else could help you more. Best regards, Michel Petitjean University of Paris, BFA, CNRS UMR 8251, INSERM ERL U1133, F-75013 Paris, France Phone: +331 5727 8434; Fax: +331 5727 8372 E-mail: petitjean.chiral()gmail.com (preferred), michel.petitjean()univ-paris-diderot.fr http://petitjeanmichel.free.fr/itoweb.petitjean.freeware.html Le mer. 19 mai 2021 à 20:13, Joe Leonard jleonard42]-[gmail.com a écrit : > > Folks, have any groups been successful getting Fortran codes (new or old) ported to the new chipset? If so, which compiler did they use and how did they handle any numerical instability they encountered? We are just starting into this (NAG compiler) but are seeing instability. I was told there was a legacy compiler switch, and recall old Fortran code assumed variables started at 0 (hence a -static flag back then)… Are there other pertinent switches that have been found? > > Thanks in advance! > > Joe Leonard > > — > For way too many people, freedom’s just another word for “I should be able to do whatever I want irrespective of how it affects others and not face consequences in any way.” >