Hello

Hello

[Angel Dimitrov]

  Hello,

 My name is Angel. I am from Bulgaria. I am compiling with your compiler my
codes. Happy to say "hi" to you.

 The Intel Xeon Phi x100 is a fast coprocessor. The question is: Is it
possible to compile Fortran 90 code on Xeon. Any documentation on compilers
flags will be very useful. The Intel fortran compiler has flag option for
many integrated cores. I do not know how to compile code on Xeon
coprocessors with gfortran.

 Thank you for your ideas.

 Best,

       Angel

Re: Hello

[Paul Richard Thomas]

Hi Angel,

The off-load facility of the Intel compilers was, of course, designed
by the company that produced the Xeon Phi. The gcc collection has no
such facility to my knowledge and is unlikely to since the Xeon Phi
series has been discontinued (summer 2018).

That said, you can run the gcc collection native on the Phi's. I
bought a Knight's Corner (3120) when they were selling for ~$300 and
tinkered with it for a while. While I could get programmes to compile
and run, the wierd sharing of the SMD units between threads meant that
custom code had to be written to benefit and I was unable to get
anywhere near to the advertised ~1Tflop. Essentially, I found that
each core functioned as a pentium, although a few hundred Gflop is not
to be sneezed at for a desktop.

Use the Intel documentation for information on how to run native. It's
so long ago that I have been through two workstation since then and
have parted company with all my notes and test code.

The Knight's Landing devices, on the other hand, have AVX-512
available on every thread and should run gcc/gfortran with
vecorization on without any problem. Again. without an off-loader,
codes will have to be run native.

I hope that this helps.

Regards

Paul

On Tue, 28 May 2019 at 20:13, Angel Dimitrov <stormlaboratory@gmail.com> wrote:
>
>   Hello,
>
>  My name is Angel. I am from Bulgaria. I am compiling with your compiler my
> codes. Happy to say "hi" to you.
>
>  The Intel Xeon Phi x100 is a fast coprocessor. The question is: Is it
> possible to compile Fortran 90 code on Xeon. Any documentation on compilers
> flags will be very useful. The Intel fortran compiler has flag option for
> many integrated cores. I do not know how to compile code on Xeon
> coprocessors with gfortran.
>
>  Thank you for your ideas.
>
>  Best,
>
>        Angel



-- 
"If you can't explain it simply, you don't understand it well enough"
- Albert Einstein

Hello

[Geek Moses]

I've read that I need to sign releases to work on any ideas here, and I do
have an idea for a great compiler.  It would be a front end for Fortran,
that allows for building GUIs.

    There are several things I keep trying to say.  First, a Compiler,
should not be designed or built without a Decompiler.  If some-one wants to
build proprietary software, let them use another compiler.  Very simply,
with a minor modification of the operating system, you can decompile any
program from the list of compilers your machine uses.  If it fails to
recognize the simple Assembly Language/Machine Code Macros, there is this
incredible problem we call a virus, another labeled spyware, and still
another we know as adware.  When ever you download a binary, unzip a binary,
move a binary form a removable drive to the computer, or attempt to run it,
it is simply decompiled first.  Disabling Assembly Language Code, would save
us by maintaining the high level computer languages.  We are not trying to
recognize strings or variables, just the base set of assembly language,
machine code macros.

    Until we find a heuristic search that allows us to determine
automatically if a program is a virus, at least we can still read the source
code.

    That's why I want to use Simulated Logic Gates, in a higher level
language and Dedicated Processes in a program that is similar to digital
electronics and digital simulators.  Assume that a Assembly Language Macro
for Open File, and another for Save File, always works faster and better
than a list of simulated Boolean gates.  Then we tamp out a Chip with Pins,
and the functions are listed each pin.  In the part number, and with the
part description and pinout, comes the list of what type variables and or
string/encoding to use.  Because, there is a way to decipher if it is or is
not a virus.  The source code we use is not readable by a machine, and
whatever language we use has to look more like flow chart or a program flow
diagram to track information.  Logic Gates make that Easy.

Look at this web page:
http://en.wikipedia.org/wiki/Logic_gate

There are a total of 7 commands in the language that are not Fortran.

I've seen people use If Statements to represent Logic Gates.  When I look,
EAX = Number of inputs, EBX + 1 for each input.  The If Statements are all
associated to producing ones from variables or strings, in true false
statements.  An if statement represents an array of logic gates and is more
simply defined as a Decoder.  Sometimes the output is a pointer, to a
graphic file, or another set of instructions/Fortran, or a Soundfile.  The
inputs set by the location of the pointer/mouse.  The graphic symbols that
are missing, are a mouse, keyboard, speaker, screen, GUI.  The whole machine
should run faster because, you would be consistently removing long string
names, and long variable names.  I want all of those variable names moved
into the headers for the simulators to be able to share other programs.
But, I don't want the processor handling them, they take up too much time.
The only time a long variable name should be there is during the Alpha
Testing, and the Beta versions should be compiled for optimization.

Windows, uses long string and variable names everywhere, and after all of
these years, when once upon a time short variable names we X and short
string names were Y$.  Each character is 1 byte  32 is 4 and 64 is 8.  The
motherboard is moving all of this much slower from the upper memory than the
CPU, and the CPU is using up clock cycles just handling data.  It's only
important during debug and development.

    If you can trace data from any hardware to the screen, and from any
hardware to the sound card, things you can't see are limited to drivers.
Everything that is not a virus, has user controls, and user controls control
all of the flow into and out of them if they are linked programs.  The use
of implied wires, and nodal logic, as with logic gates, allows us to track
information, and an automated system say, oh, that is a virus.  But, not
with proprietary software that cannot be decompiled.

    Yea, I want to make Linux the absolute best software ever.  But, I ain't
kidding.  If you can get the programmers together with the digital
electronics gurus who actually build circuits, you can actually bring viral
circus to an end.  Especailly, when the list of macros your compiler spits
out is far less than the list of viruses out there.  No need for virus
scanning in the old school sense, looking for a list of over 60,000
different possibilities of code combinations, it's just a list of how ever
many commands and functions your high level compiler language spits out.