scott@CANNON.CS.USU.EDU (Scott Cannon) (08/01/90)
>We are very interested in using the transputer in a multiprocessor >embedded design. I am looking for information on other's experience >in using multiple transputers in an embedded system design. I am also >looking for information or pointers to information regarding C++ >compilers, (etc.).. I don't know what your platform is, but here is a brief comment on using Turbo C++ on a PC host. An embedded system consists of course, of two software systems; the transputer code and the code running on the host to handle the I/O interface and provide user routines which can be called from the transputer system (cioext() on LSC). Turbo C++ works fine for the host code if the large/huge model is used, but compilation, linking, and editing interaction are so much slower than plain Turbo C, that we've gone back to that. The advantages of OOPS just didn't seem worth the frustration, even on 286/386 systems. By the way, Turbo C++ requires more than 6MB of disk. While I have the floor... We found that by a) using the Alta Technology Superlink XL card (with a 16-bit PC interface) to host our trams, and b) doing block disk I/O in cioext(), we were able to speed up disk reads by a factor of about 2 over plain-vanilla fread()'s on a transputer. S. Cannon, Utah State Univ. scott@cannon.cs.usu.edu