I have implemented new version of the Small C plus by Cain, Van Zandt, Hendrix, Yorston version that emits code that will run on a TMS9900/TMS99000 Single Board Computer with at least 64k Bytes of memory. This new version of Small-C is a major improvement over other versions of Small C and includes unsigned integers and chars, floating point support, as well as structures and unions and other standard C language support that are absent from earlier versions. Indeed it compiles and runs the VT100 (https://github.com/TeraTermProject/teraterm) screen editor ED2 https://github.com/mdlougheed/ed2 without difficulty.
A screen capture of ED2 running on a SBC (https://github.com/AlexanderAdelAU/TMS9900-SBC) that was compiled under this version of Small C is shown here:
The source for the BDOS, and IO libraries that are required to run ED2 are included on this site and is built using the following commands:
smallcp -C src/ED2
R99 src/ED2 SCHCLC
link99 -M -S src/ED2 iolib99.LIB clib99.LIB.
The code produced by the Small C compiler can be either compiled to produce standalone code (using the -M) flag and assembled using the A99 assembler to run at an absolute location in memory. If it needs to be linked to other runtime libraries then it will produce relocatable assembly code (using the -C flag) that can be assembled and linked with other modules using the R99 (Relocatable assembler) assembler in conjuction with the linker loader (Link99). The relocatable applications use the standard REL format.
It is compiled using MingW GCC and therefore operates as a Cross Compiler however the C code is standard C and should be able to self compile with the appropriate OS support on the target computer. The Eclipse project files are included as a zip file that may help building and testing the compiler in your own environment.
A test programme for the Byte Sieve.c benchmark is included and runs on a 20MHz TMS99105A in 18 seconds.
The floating point libraray implementation is a retargetting of Anders Hejlsberg's Z80 Floating Point library (FloatM48.z80)
An IOLIB is included that supports CP/M compatible (BDOS in this respository) operating system calls and a core set of 40bit floating point library routines. The core set includes, FPADD, FPMUL,FPSUB, FPDIV, SIN, COS, TAN, POW,LN,EXP, SQRT. Other functions such as trancedentals will be added later. The TMS9900 Floating Point Library filename is CFLOAT48.A99 and was ported from a Z80 Floating Point library.
A sample test programme that demonstrates how to calculate a Double Integral using Simpson's Rule is shown below and is compiled using the following syntax:
copy DoubleIntegralSimpsonsRule.c, cDSR99.c
smallcp -C cDSR99
R99 cDSR99 SCHCLC
link99 -M -S cDSR99 clib99.LIB iolib99.LIB strlib99.LIB fplib99.LIB
The programme performs a double integration on the following mathemtical function"
zd = cos(pi * cos(x) / 2.0) * cos(pi * (1.0 - sin(x) * cos(y)) / 4.0);
zd = pow(zd, 2.0)/sin(x);
This is a very good test of the Floating Point library.
/*
============================================================================
Name : DoubleIntegralSimpsonsRule.c
Author : Alex Cameron
Version :
Copyright : None
Description : Simpson integration technique for
evaluating double integrals
============================================================================
*/
#include "float.h"
double fxy[100], fy[12];
double pi;
double f();
main() {
double llx, lly, ulx, uly, x, y;
double hx, hy, ef, of, simpson;
int nosx, nosy, i, j;
/*
*
============================================================================
*
* Simpson integration constants.
* nos -> number of strips
* ul -> upper limit of integration
* ll -> lower limit of integration
* h -> incremental value per strip
=============================================================================
*/
pi = 3.1415926;
nosx = 10;
nosy = 8;
llx = 0.1e-8;
lly = 0.1e-8;
ulx = pi;
uly = 2.0 * pi;
hx = (ulx - llx) / nosx;
hy = (uly - lly) / nosy;
printf("\n\nDouble integration parameters: \n");
printf(" Step size hx and hy: %10.6f,%10.6f\n", hx, hy);
printf(" Number of steps (x,y): %d,%d\n\n", nosx, nosy);
/*
Calculate all the points within the integration domain
*/
for (j = 0; j <= nosy; j++) {
y = j * hy + lly;
for (i = 0; i <= nosx; i++) {
x = i * hx + llx;
fxy[i+j*nosx] = f(x, y);
}
}
/*
========================================================
Now perform a Simpson integration along
the x axis and accumulate results using
the y axis variable as an index
===========================================================
*/
for (j = 0; j <= nosy; j++) {
of = fxy[1 + j*nosx];
ef = 0.0;
for (i = 2; i <= nosx - 2; i += 2) {
ef += fxy[i + j*nosx];
of += fxy[i + 1 + j*nosx];
}
fy[j] = fxy[0 + j*nosy] + fxy[nosx + j*nosx] + 2.0 * ef + 4.0 * of;
}
/*
* ========================================================
* Lastly perform Simpson integration
* along the y axis.
* ========================================================
*/
of = fy[1];
ef = 0.0;
for (j = 2; j <= nosy - 2; j += 2) {
ef += fy[j];
of += fy[j + 1];
}
simpson = (hx * hy / 9.0) * (fy[0] + fy[nosy] + 2.0 * ef + 4.0 * of);
printf("Result = %10.6f\n\n", simpson);
}
/*
* ==============================================================
* Enter the function to be integrated here.
* plot abs(((cos(pi*cos(x)/2.0)*cos(pi*(1.0-sin(x)*cos(y))/4.0))^2)/sin(x))
* ===============================================================
*/
double f(x, y)
double x, y; {
double zd;
zd = cos(pi * cos(x) / 2.0) * cos(pi * (1.0 - sin(x) * cos(y)) / 4.0);
zd = pow(zd, 2.0)/sin(x);
return (fabs(zd));
}
A output of running the above is shown below.
%cDSR99
Double integration parameters:
Step size hx and hy: 0.314159, 0.785398
Number of steps (x,y): 10,8
Result = 3.827845
%