This is my first foray into SISAL (Streams and Iteration in a Single Assignment Language), an old forgotten scientific computing language out of the Lawrence Livermore National Laboratory circa 1985. It feels like a precursor to something like Julia.
For more info and further reading, refer to https://kestrel.nmt.edu/~raymond/software/sisal/sisal.xhtml
The below assumes a UNIX environment. Documentation is scarce, YMMV
- Download the tarball for SISAL from Sourceforge: https://sourceforge.net/projects/sisal/
- Extract the tarball and enter the directory:
tar xvfz sisal-14.1.0.tgz
cd sisal-14.1.0- Run
./configure. If neccessary, include the flags below to get your C compiler into the GNU89 standard:
./configure CFLAGS="-g -O2 -fcommon -std=gnu89 -Wno-int-conversion -Wno-implicit-function-declaration -Wno-implicit-int" LIBS="-lm"makeit happen (may needsudoat the end to copy files where they need to be):
make
sudo make install- You can safely exit the directory and set up camp elsewhere.
Given how old the project is, the following CFLAGS need to be passed:
-std=gnu89: the code uses pre-C99 constructs; function declarations without return types, implicit int, K&R style in places-fcommon: relies on tentative definitions for global variables, which GCC 10+ made an error by default-Wno-int-conversion: casting between pointers and integers without explicit casts, common in 1990s C-Wno-implicit-function-declaration: calling functions before declaring them, also pre-C99 practice-Wno-implicit-int: functions with no declared return type, defaulting to intLIBS="-lm": theconfigurescript doesn't automatically detect the math library dependency on your system
Assuming we have a simple SISAL programme that takes in an integer value N and prints to screen the squares from 1 to N, it first needs to be compiled and then run:
% Generate an array of squares
% Tells the compiler this is the only exposed function
define Main
% Define a custom type
type IntArray = array [integer]; % use real for floats
function square(x: integer returns integer)
x * x
end function
function Main(N: integer returns IntArray)
for i in 1, N
returns array of square(i)
end for
end function
Save this as squares.sis and run:
sisalc -o squares squares.sis
You will see a long output from the C compiler. This is harmless and is just modern GCC complaining about 32-bit pointers because we are not in 1989 any more:
LL Parse, using binary files
* Reading file: squares.sis...
version 1.8 (Mar 28, 1989)
accepted
16 lines in program
0 errors ( calls to corrector)
0 tokens inserted; 0 tokens deleted.
0 semantic errors
In file included from squares.c:11:
/usr/local/include/sisal/sisal.h: In function ‘ArrayDump’:
/usr/local/include/sisal/sisal.h:193:31: warning: format ‘%x’ expects argument of type ‘unsigned int’, but argument 3 has type ‘struct Array *’ [-Wformat=]
193 | fprintf( stderr, "Array: 0x%x (%d), sizeof: %d\n",Array,Array,sizeof(*Array));
| ~^ ~~~~~
| | |
| unsigned int struct Array *
/usr/local/include/sisal/sisal.h:193:35: warning: format ‘%d’ expects argument of type ‘int’, but argument 4 has type ‘struct Array *’ [-Wformat=]
193 | fprintf( stderr, "Array: 0x%x (%d), sizeof: %d\n",Array,Array,sizeof(*Array));
| ~^ ~~~~~
| | |
| int struct Array *
/usr/local/include/sisal/sisal.h:193:48: warning: format ‘%d’ expects argument of type ‘int’, but argument 5 has type ‘long unsigned int’ [-Wformat=]
193 | fprintf( stderr, "Array: 0x%x (%d), sizeof: %d\n",Array,Array,sizeof(*Array));
| ~^ ~~~~~~~~~~~~~~
| | |
| int long unsigned int
| %ld
/usr/local/include/sisal/sisal.h:196:32: warning: format ‘%x’ expects argument of type ‘unsigned int’, but argument 3 has type ‘struct ArrayPhys *’ [-Wformat=]
196 | fprintf( stderr, " Phys: 0x%x (%d)\n",Array->Phys,Array->Phys);
| ~^ ~~~~~~~~~~~
| | |
| unsigned int struct ArrayPhys *
/usr/local/include/sisal/sisal.h:196:36: warning: format ‘%d’ expects argument of type ‘int’, but argument 4 has type ‘struct ArrayPhys *’ [-Wformat=]
196 | fprintf( stderr, " Phys: 0x%x (%d)\n",Array->Phys,Array->Phys);
| ~^ ~~~~~~~~~~~
| | |
| int struct ArrayPhys *
/usr/local/include/sisal/sisal.h:197:31: warning: format ‘%d’ expects argument of type ‘int’, but argument 3 has type ‘LOCK_TYPE’ {aka ‘pthread_mutex_t’} [-Wformat=]
197 | fprintf( stderr, " Mutex: %d\n",Array->Mutex);
| ~^ ~~~~~~~~~~~~
| | |
| int LOCK_TYPE {aka pthread_mutex_t}
While it is tempting to run ./squares 10, it will not accept standard command line arguments like a modern C or Python script. Because SISAL was built for continuous dataflow processing on mainframes, it explicitly listens for stdin.
$ ./squares 10
x86_64-unknown-linux-gnu SISAL 1.2 (PThreads) ?.?
You will see it hangs until you provide a value followed by Enter. A much cleaner way to automate this is by using echo
$ echo 10 | ./squares
x86_64-unknown-linux-gnu SISAL 1.2 (PThreads) ?.?
[ 1,10: 1 4 9 16 25 36 49 64 81 100 ]
In terms of run time and binary size, I reckon it's near-C speed and small! For instance, the compiled binary for the above code is only 248Kb! As far as scientific computing goes, it performs parallel programming by design and was running on supercomputers back in the day.
This was an afternoon's work exploring something new so it's difficult to say. I like the syntax so far and the compilation speed, binary size, and run time. It doesn't have most of your modern things like networking capabilities and struggles to parse files (to my knowledge) without relying on C or Fortran, but it's so far been decent to build mathy things with.
Of all the 'dead' languages I've seen (mainly Wikipedia code snippets), this one looks pretty cool. The goal is a bigger project and perhaps testing it as a drop-in bit of math inside Julia (via ccall).