Path: utzoo!attcan!uunet!lll-winken!csd4.milw.wisc.edu!nic.MR.NET!shamash!com50!pai!erc From: erc@pai.UUCP (Eric Johnson) Newsgroups: comp.sys.mac Subject: Re: New Mac Rumours (really language-based chips) Summary: TI Lisp chip, SOAR, Lilith, etc. Message-ID: <409@pai.UUCP> Date: 24 Feb 89 14:16:23 GMT References: <70755@ti-csl.csc.ti.com> Organization: Prime Automation, Inc., Burnsville, MN Lines: 57 In article <70755@ti-csl.csc.ti.com>, holland@m2.csc.ti.com (Fred Hollander) writes: > > This is very interesting. I don't have the strongest background in > hardware architecture, but, could you please explain how a processor > could be optimized for a specific high level language? Certain high-level languages make certain assumptions about an underlying machine, or more usually, an underlying "virtual" machine. Most attempts to optimize a processor to a specific language seem to focus at making the processor instruction set the "virtual" machine instruction set. Some examples (I may make mistakes so please correct any inaccurate info): TI's LISP chip which focuses on the tagged data values in LISP and the use of CDR and CAR. (There are also competitors, such as the MacIvory one, but I couldn't resist mentioning TI.) Lilith Herr Docktor Professor N. Wirth's (inventor of Pascal, etc.) Modula-based workstation. The processor was aimed at the Modula language. SOAR Smalltalk On A Risc -- one of many attempts to implement Smalltalk byte-codes as a machine instruction set. Forth chip (From Novix?) A chip designed around the lanuage Forth (which seems rather strange as I always thought Forth was one of the few languages heavily optimized to the traditional stack-based architecture). Intel 80x86 I have read arguments stating that the Intel architecture was based around the concept of Pascal, especially with the four segments: Data, Code, Stack and Extra (personally, I think the Intel architecture is a crime against humanity :-). Other places to look would include the Warren Absract Machine for Prolog (I'm sure someone is working on putting this on a chip) and in object-oriented languages. Sometimes, chip-makers go the opposite route and create a new language optimized for the hardware, and Inmos did with Occam for their Transputer. Anyway, I hope this sheds some light, -Eric > > Fred Hollander > Computer Science Center > Texas Instruments, Inc. > hollander@ti.com > > The above statements are my own and not representative of Texas Instruments. -- Eric F. Johnson | Phone +1 612-894-0313 | Are we Prime Automation,Inc | UUCP: bungia!pai!erc | having 12201 Wood Lake Drive | UUCP: sun!tundra!pai!erc | fun Burnsville, MN 55337 USA | DOMAIN: erc@pai.mn.org | yet?