Welcome to our forums
Are there perhaps any information available on the development (or history) on the Salford/Silverfrost Fortran compiler.
What I heard from tellings the compiler is somehow related to the University of Salford and then Silverfrost got involved. Or something like that.
Sadly I got no response to the history of the Salford/Silverfrost compiler. However, I do hope it is not a sad one or maybe it is so successful that it is not allowed to spread 😄 If I cannot get it in the Silverfrost forum I will have to keep to the telllings :!:
I would take me half an hour to dig out an acurate history and I don't see how it will help you. Here is a summary from memory.
Basically it began with a Fortran 77 compiler written by staff in the computing laboratory at Salford University for Prime computers. This was around the time when the 77 standard was first released. This compiler was later ported to 32 bit PCs. The current PC operating systems were 16 bit but FTN77 came with its own 32 bit OS add-on called DBOS. By this time the compiler team had formed a company (later called Salford Software Ltd) that was wholly owned by the University of Salford.
Salford Software continued writing and selling variious compilers over a period of more than 20 years. It was merged with another Salford University enterprise that initially marketed Novel Networks. In due course the compiler side of the company was passed on to Silverfrost whilst the networking side was extended to a much broader scope and continues to expand to this day.
Paul,
thanks for the info. At a first glance you might think that is does not help me. However, our IT manager (and I think in other companies as well) wants to invest in software that one can rely on. In the past we have seen Fortran compilers where for example the support was stopped.
At present we are using different compilers which is a result from employees coming from different company locations. One aspect in the decision of the choice of compilers is to have a look at the company history. I have little experience with other compilers, but the combination of FTN95 with SCC and more then 20 years in bussiness are good argument to continue using FTN95 😉 This might even be very useful for (new) programers searching for a Fortran compiler.
My shaky memory is...
It started in the late 1970s. I think David Vallance recruited David Bailey to write a Fortran compiler for an ICL machine - which the University had installed. This was a success and was ported to the Prime where FTN77 was a big success.
In the late 80s FTN77 was ported to the 80386 architecture. At the time PCs that had an 80386 ran it in 8086 mode and so to allow FTN77 to run in protected mode a DOS extender was written (DBOS). All this was written in a mixture of Fortran and assember (mostly Fortran 77).
At the same time FTN77/386 was being created the University formed a company 'Salford Software Marketing Ltd' to develop and sell the compiler products (which became Salford Software Ltd).
In 1991 SSM started development on their C++ compiler (which is bundled with FTN95). This C++ compiler was used to develop FTN95 which shares the same backend as the C++ compiler.
I have a vague recollection of there being a CDC version of a Salford Fortran back in the 1970's. I was a PhD student at Imperial College, and although the main Fortran compiler was 'Minnesota Fortran' (MNF), I'm sure that there was an alternative FTN compiler as well. My memory is decidedly shaky on this, as MNF had better support from the computer centre and it was what I used. I entertained the thought of using FTN because of the graphics support - I certainly read about Heshey fonts which I thought would be useful (and which the DBOS PC version also implemented).
When the PC version came out, it was 386/387 only, and it wouldn't work with all extended memory managers. It wiped the hard disk contents on one computer I used at my University. The DBOS (Dave Bailey Operating System?) DOS extender was powerful, and FTN77 was really fast compared to other compilers of the day, but DOS extenders were a blind alley technically.
I attended a one-day seminar at Salford shortly after ClearWin was launched - in the early 1990's. I think that DB himself was there, and presented part of it. I decided that the complexities of having Clearwin AND DBOS were too much, and I left specific programming for windows alone for a long time. I remember being told that FTN77 was mainly written in Fortran, simplifying setting it up on other computer types - I can't vouch for the truth of that, it may be an 'urban myth'.
The minute that you decide to use ANY extension to a compiler, you are locked in to that one and no other. In my view, Fortran-90 and -95 standards have simply attempted to duplicate (or supplant) existing features of Fortran rather than standardising graphics or useful interfaces (for example to Windows). Those guys on the Standards Committee will never standardise a Windows interface - they mainly use Unix ...
I don't think you will ever get a guaranteed indefinite support for anything. However, it looks likely to me that Win32 programs will work on the majority of PCs for the next decade, even if they are compiled today on XP. ClearWin is rather undervalued by Silverfrost. In my view it is possible to create a very solid commercial program interface with it. It is full of surprises. It could do with rather better instructions, however.
On the subject of speed, the Polyhedron benchmarks don't really show FTN95 in a good light relative to some other compilers. I think that in the main, they are twaddle: I've not the slightest problem with run speeds. FTN95 compiles faster than the others, and has more exhaustive diagnostics. That is more useful to the developer, provided that the run is 'fast enough'.
Eddie
jjgermis,
Over the past 30 years I have used many fortran compilers on a significant number of different platforms. I first used FTN77 from Salford in the 1980's and have stuck with it ever since in parallel with some of the other compilers. Now it is almost the only one I ever use, apart from gfortran on a Linux 64bit machine and a dabble with g95 on windows. For graphical applications (and windows interface) it is the only I now use, having long ago migrated from tektronix terminals under VAX/VMS and HPUX's starbase graphics.
I don't believe the Polyhedron benchmarks either, on a 64bit machine a 32bit silverfrost compiled program easily out performs a full 64bit gfortran version on a numerically intensive task.
You can write very professional looking programs with FTN95 as many of the contributors to this forum do.
As far as I remember, between FTN77 and FTN95 there was what was probably the first Fortran 90 compiler for PCs: the FTN90, which was developed by Salford in collaboration with NAG, and had a C++ backend. After some time, Salford launched FTN95 which was a pure Salford product, and PLATO, which was written in Clearwin+ and replaced something called SIDE (Salford Integrated Development Environment?) that used Microsoft Rich Edit control. I believe that Richard Putman was the developer of Plato. Maybe someone could correct me if I am wrong.
Agustin
Respect to Polyhedron comparisons. Of course main advantages are in capabilities of the compiler. it's the developer own time what is important, not a CPU one. But common perception of the average Joe is exactly opposite. Joe wants to see his compiler is the first, by 0.1 microsecond runtime or 1 frame per second graphics, by one more damn obsolete feature but still the first.
And the first 'black balls' in the FTN95 basket are all kinds of extensions. Here FTN95 is absolute leader, literally light years ahead, but Polyhedron compares what tightly packed bunch of other compilers can do. Polyhedron does not compare (though mentions somewhere of course, but nobody look at anything besides the nice looking Polyhedron spreadsheets) the graphics capabilities, fast compilation, OpenGL, Windows GUI builders or NET. It compares VAX style debug (D) lines and VAX style system intrinsics (SECNDS etc.). And adding all that together here FTN95 lately became 'last' from being first just several years ago. This could be fixed in one-two weekends. Really, is it hard to add VAX extensions ENCODE/DECODE or variable format expressions using <> and Q edit format, the C-style string constants or aliased dummy argument variable in debug and be the first here again? In debugging capabilities one of the most important parts of compiler it is and always was first.
Another weakness is run benchmarks. This is what makes average Joe happy and other compilers pay attention to his perceptions (which are often not a realities). Absoft for example abruptly in no time became a leader in execution speed sharing the first place with Intel. Here is the weakest part of FTN95. This could not be done in one weekend. But task is simplified now after Absoft & Intel have done that. Those at Silverfrost who are intimate with their beloved assembler and can compare how they optimize the code and how FTN95 is doing that will solve this quick. Right now I suspect FTN95 is not doing any optimization at all since even not a half-finished newborn g95 made on Joe's donation money is faster then FTN95.
I do not know if Posix & UNIX style libraries would be of much benefits but adding native 64bit OS support, interfaces to Matlab, Mathematica and OpenMP or better native parallel capabilities (single processors are dead), some elementary Fortran 2003 features would not be bad - but some of them are definitely years long tasks, though some just for several months. Dave Bailey with his knowledge of Mathematica could make a breakthrough here like he have done in 80th. Like making truly Visual Fortran as opposed to faked 'Visual' one of DEC/Compaq. People are tired of coding, they like clicking and making programs out of 'pre-chewed' 'Lego' pieces. If this would be done, users will excuse FTN95 ANYTHING, nobody cares about dead VAX extensions. But I've heard DB is not with Silverfrost and might be not interested in Fortran anymore.
And third weakness, is its general look. Just from first clicks of installation you become bored and confused. Nothing pops into your face and amazes you. As a result an average Joe will never reach compiler's best parts. 1 out of 1000 will find Clearwin or OpenGL. But the nice examples made with Clearwin, OpenGl, good graphics libraries could make from FTN95 a sweet candy. All do that - Intel, GINO, Winteracter. MATFOR. They make a good wrap for good impression. As to FTN95, using Clearwin for 10+ years, I was unable even to see how Visual Clearwin or NET works...EXEs or AVI videos are not supplied to save 3 MB on my harddrive 😦
Just a qucky to correct a couple of misconceptions in the above posts.
David Bailey is still part of our team.
Plato (versions 3 and 4) were written without using ClearWin+.
The bit about FTN77 being written in Fortran 77 being an urban myth:
It definitely was both for the Prime and 386 (both of which I helped support). DBOS was also written mostly in Fortran 77 with some 80386 assembler. In about 1990 DB and Tony Webster developed a macro substitution system and used it to convert things like:
CORE2(GRINFF+4)=OR(CORE2(GRINFF+4),VREG)
into
REGISTER_ALLOCATED(GRINFF)=1
There is some confusion about FTN90. Salford Software developed, with NAG a Fortran 90 compiler -- which was probably called FTN90. However, I don't think it ever got near release.
NAG released (and still may sell) a Fortran 90 compiler which translated Fortran 90 into C. That was entirely their own work.
I think the original Plato was written by Richard Putman in ClearWin+. The current version shares no code with this version.
Quoted from aebolzan As far as I remember, between FTN77 and FTN95 there was what was probably the first Fortran 90 compiler for PCs: the FTN90, which was developed by Salford in collaboration with NAG, and had a C++ backend. After some time, Salford launched FTN95 which was a pure Salford product, and PLATO, which was written in Clearwin+ and replaced something called SIDE (Salford Integrated Development Environment?) that used Microsoft Rich Edit control. I believe that Richard Putman was the developer of Plato. Maybe someone could correct me if I am wrong.
Agustin
I'm glad to hear that DB of DBOS fame is still with you. He impressed me enormously when I met him - just like I was impressed by FTN on my first encounter and every day since. Perhaps he could be encouraged to contribute openly to the forum?
I was also pleased to see that others share my view of the Polyhedron benchmarks. I'm not much of a believer in synthetic benchmarks. For instance, Intel cpus do much better at them than AMD, but if I ever want real work done, I leave my university's computers (all Intel - and I have a choice of hundreds) and go home to run things on a home-built AMD machine that appears to have inferior specs, but in practice, performs significantly better.
Is it worth suggesting that someone at Silverfrost finds out why the Polyhedron benchmark performance is so poor, in contrast to our real-world experience, and fixes it?
Like Dan I can't make head nor tail of .NET and visual clearwin - I'm obviously missing a manual somewhere. My understanding is that .NET sets out to make programs run on multiple machine architectures, but as there aren't any meaningful alternatives to x86, and .NET doesn't work with Linux or Mac, then it all seems rather pointless.
Plato 3 and 4 aren't written with ClearWin - now how did I guess that? Perhaps it is because they do things you can't do straightaway with ClearWin (but would like to - for example, dockable toolbars). Now how's that for a hint?
Eddie
Eddie,
Got you again!
and .NET doesn't work with Linux or Mac, then it all seems rather pointless.
visit:-
http://www.mono-project.com/Main_Page http://dev.mainsoft.com/Default.aspx?tabid=45
Thus a .net application produced with FTN95 should be able (in theory) to run on just about any platform (including 64bit?)
John
Quoted from Robert There is some confusion about FTN90. Salford Software developed, with NAG a Fortran 90 compiler -- which was probably called FTN90. However, I don't think it ever got near release.
NAG released (and still may sell) a Fortran 90 compiler which translated Fortran 90 into C. That was entirely their own work.
I think the original Plato was written by Richard Putman in ClearWin+. The current version shares no code with this version.
Well, may I say that logn time ago I got a couple of diskettes with probably the first public release of FTN90 (1.09?)? Salford sent them to me when we were still working with FTN77. FTN95 appeared likely one or two years later, when Salford finished the contract with NAGbut FTN90 was real.
About Plato: I thougth that Plato was still using Clearwin (Plato 2.09 was) as that was the limitation of Plato for printing the Fortran files in colour (at variance with SIDE). The point now is: why we cannot print our source code in colour using Plato?
Agustin
Colour printing is on the wish list for Plato. It is just something I have not got round to yet. It is a fairly straight forward addition but not trivial.
John,
I'll believe it when I see it for myself.
Eddie
FTN95 appeared likely one or two years later, when Salford finished the contract with NAGbut FTN90 was real.
Paul pointed out to me that Salford did sell it with our backend. I added the debug support so you would have thought I would remember...
We also had a Sheffield Pascal for DBOS. Didn't sell many though.
SHEFFIELD PASCAL FOR DBOS!!! Yes! I remember that! It was one of the products in the price list of Salford Software during the early 90' So you were at NAG during the development of FNT90? I am very sorry to tell you that last year I found that my original diskettes of FTN90 were lost, so i had no material prove about the real existence of FTN90. Thanks Paul for supporting my memories!.
Agustin
No, I was doing work for Salford Software. It was using the C++ debug information and it was 'filtered' to look like Fortran.