rab@mimsy.UUCP (06/17/87)
I recently read a paper on a `very_long_instruction_word' (VLIW) computer, specifically the Multiflow TRACE computer. Each instruction is up to 1024 bits long. Different bit fields directly control different functional units in the cpu, so there is no need for microcode or decoding logic. This results in very fast execution, but requires a lot of memory, and very smart compilers. They have C and Fortran compilers, and recently got UNIX up and running. As memory prices continue to decline, this kind of computer may become quite common. If anyone has experience with VLIW computers, what do you think of them? How do they really perform and what types of problems do you have with them? rab@mimsy.umd.edu seismo!mimsy!rab
bpendlet@esunix.UUCP (06/18/87)
in article <7082@mimsy.UUCP>, rab@mimsy.UUCP (Bob Bruce) says: > > > I recently read a paper on a `very_long_instruction_word' > (VLIW) computer, specifically the Multiflow TRACE computer. Read "BullDog: A Compiler for VLIW Architectures" by John R. Ellis, The MIT Press. This was the 1985 ACM Doctoral Dissertation for 1985. Or read the proceedings of the SIGMICRO annual workshop on microprogramming for the last 19 years. > > As memory prices continue to decline, this kind of computer may If you look inside of most CISC computers you will find a custom designed LIW or VLIW computer. As far as I can tell VLIW is another way of saying horizontal microcode. But everyone "knows" that no compiler can generate horizontal microcode! So maybe thats why they decided to call it VLIW. Of course most microcoded machines are optimized to emulate a specific instruction set rather than executing application programs. > > rab@mimsy.umd.edu > seismo!mimsy!rab Bob Pendleton -- Bob Pendleton @ Evans & Sutherland UUCP Address: {decvax,ucbvax,ihnp4,allegra}!decwrl!esunix!bpendlet Alternate: {ihnp4,seismo}!utah-cs!utah-gr!uplherc!esunix!bpendlet I am solely responsible for what I say.