I agree with you, but I would phrase it in another way. Adele Goldberg is one of many bright people in this industry that did not get as much in the spotlight for their contributions as their peers; but, in many ways, she accomplished more than most of these “unfortunates” ever did, having received ACM Software Systems Award along with Alan Kay and Dan Ingalls for her contributions to Smalltalk, and served as president of the Association—all of this a testament to her unique brilliance.
“Our goal was to teach hardware designers and their software counterparts what we thought might be the need for something special in the hardware. It turned out we were wrong. There was a wonderful project at Berkeley that was part of this, which was the SOAR Project, which later became the Sun Microsystems SPARC. SOAR is “Smalltalk-On-A-RISC,” a project by Dave Patterson and David Ungar. It turned out that you really didn’t necessarily need special hardware, and the Motorola 68000 was handling Smalltalk quite well.”
It definitely was the case in the 1980s that it was better to figure out how to use general purpose machines and especially those designed to run C/Pascal effectively (MIPS, SPARC, ARM … the 68k predated that) than to make limited production custom hardware because the general purpose stuff had the funding to improve quickly.
It will be interesting to see if that’s true now that hardware progress has slowed down. Maybe a few helpful instructions can be sneaked into “conventional” ISAs. The RISC-V “J” extension group is looking at that. The performance of Javascript is very important now, has a lot of money spent on it, and has a lot in common with the needs of Smalltalk or Lisp.
Maybe a few helpful instructions can be sneaked into “conventional” ISAs. The RISC-V “J” extension group is looking at that
I know that Jecel Assumpcao of the Squeak community – who has designed Smalltalk based processors over the years – is a part of the J extension working group.
Unlike earlier microprocessors, like the Intel ones, the 68k was very much made with running C/Pascal in mind. Note the large number of general purpose registers or the LINK/UNLK opcodes for handling stack frames.
The real difference compared to later RISC designs was that it wasn’t solely designed for compilers.
When the 68000 design was started in 1976 (or even when it was introduced in 1979 – basically exactly in parallel with the 8086 development, but Motorola was late getting working chips out) the use of high level languages was not yet common.
Like the DEC VAX, the 68000 was primarily designed to make the human assembly-language programmer more efficient.
Yes, those crazy people in the technical documentation department of AT&T had written a toy OS for the PDP 11 in C, but they were out of step with the industry and not yet well known. The first version in C was done (started?) in 1973, but there was not a non-PDP 11 version until 1978.
The following article in BYTE from the time repeatedly talks about how the 68000 makes it easier for the programmer to write code. Not for a C or Pascal compiler to write code, but for the programmer to write code. That’s what things such as the LINK/UNLK instructions are for. Assembly language programmers want a frame pointer so that variables stay at the same offset from it as the stack is pushed and popped. Compilers don’t care – they can keep track.
The more I learn of Adele Goldberg, the more I think she doesn’t get nearly enough credit for Smalltalk.
Her book, Smalltalk 80, is a literal masterpiece. My copy is among my most prized possessions.
I agree with you, but I would phrase it in another way. Adele Goldberg is one of many bright people in this industry that did not get as much in the spotlight for their contributions as their peers; but, in many ways, she accomplished more than most of these “unfortunates” ever did, having received ACM Software Systems Award along with Alan Kay and Dan Ingalls for her contributions to Smalltalk, and served as president of the Association—all of this a testament to her unique brilliance.
I will tell my grandchildren that this was the Adele everyone talked about in my generation.
Thanks! Interestingly it is not the same as the Oral History of Adele Goldberg interview I knew about.
“Our goal was to teach hardware designers and their software counterparts what we thought might be the need for something special in the hardware. It turned out we were wrong. There was a wonderful project at Berkeley that was part of this, which was the SOAR Project, which later became the Sun Microsystems SPARC. SOAR is “Smalltalk-On-A-RISC,” a project by Dave Patterson and David Ungar. It turned out that you really didn’t necessarily need special hardware, and the Motorola 68000 was handling Smalltalk quite well.”
It definitely was the case in the 1980s that it was better to figure out how to use general purpose machines and especially those designed to run C/Pascal effectively (MIPS, SPARC, ARM … the 68k predated that) than to make limited production custom hardware because the general purpose stuff had the funding to improve quickly.
It will be interesting to see if that’s true now that hardware progress has slowed down. Maybe a few helpful instructions can be sneaked into “conventional” ISAs. The RISC-V “J” extension group is looking at that. The performance of Javascript is very important now, has a lot of money spent on it, and has a lot in common with the needs of Smalltalk or Lisp.
I know that Jecel Assumpcao of the Squeak community – who has designed Smalltalk based processors over the years – is a part of the J extension working group.
Unlike earlier microprocessors, like the Intel ones, the 68k was very much made with running C/Pascal in mind. Note the large number of general purpose registers or the LINK/UNLK opcodes for handling stack frames. The real difference compared to later RISC designs was that it wasn’t solely designed for compilers.
When the 68000 design was started in 1976 (or even when it was introduced in 1979 – basically exactly in parallel with the 8086 development, but Motorola was late getting working chips out) the use of high level languages was not yet common.
Like the DEC VAX, the 68000 was primarily designed to make the human assembly-language programmer more efficient.
Yes, those crazy people in the technical documentation department of AT&T had written a toy OS for the PDP 11 in C, but they were out of step with the industry and not yet well known. The first version in C was done (started?) in 1973, but there was not a non-PDP 11 version until 1978.
The following article in BYTE from the time repeatedly talks about how the 68000 makes it easier for the programmer to write code. Not for a C or Pascal compiler to write code, but for the programmer to write code. That’s what things such as the LINK/UNLK instructions are for. Assembly language programmers want a frame pointer so that variables stay at the same offset from it as the stack is pushed and popped. Compilers don’t care – they can keep track.
http://www.easy68k.com/paulrsm/doc/dpbm68k1.htm