4 ms·

I thought ENIAC and Turing aren't really connected and independent to each other? And then again for development of physical computer, seems like Neumann have m

by mcmoor 1y ago

I thought ENIAC and Turing aren't really connected and independent to each other? And then again for development of physical computer, seems like Neumann have more claim to him because of his architecture.

vlovich123 1y ago
Turing literally defined the mathematical underpinning of what it meant to build a computer in the first place and what operations it needed to support (the language and the computer instruction itself being indistinguishable at the time). Without this work we don’t have any kind of meaningful understanding of how to translate algorithms into something that can run on a machine. Neumann’s contributions were valuable and notable as they proposed to make machines more flexible than how the ENIAC worked, but these are largely just “applied engineering” improvements not foundational theoretical CS concepts; the Harvard architecture is equally valid and flexible as Neumann’s design, even preceding Neumann’s idea of not differentiating code and data. It just that the simplicity at the time of Neumann machines won out even though today we pay for their penalty in other ways. But without Neumann or Harvard architectures you still end up building machines and likely these ideas would eventually have been proposed and tried by others without any meaningful delay in development. Without Turing’s work it’s possible that computers would have been delayed by decades until someone else figured this stuff out. Without the transistor computers would just be an odd curio that’s in use in some places and basically the entire field wouldn’t exist (cryptography wouldn’t be a really thing, AI, databases, programming languages etc).
Animats 1y ago
I think that's right. The ENIAC people were working on ballistics for large guns, and needed to crunch numbers and create printed tables. The Turing team was working on key-testing, like a Bitcoin miner. Colossus was a key tester, trying patterns zipping by on endless loops of tape, and recording matches. Any general purpose machine today can do both, but the early machines were specialized. What general purpose architecture should roughly look like was known all the way back to Babbage. Some kind of long list of instructions, with unconditional and conditional branches. Zuse and Atanasoff both got there before WWII. The big hangup was memory. All early computing struggled with memory limitations, as I mention occasionally. CPUs and instruction decoding came long before enough memory to store programs. Memory was still a million dollars a megabyte in the mid-1970s.
- vlovich123 1y ago
  No general purpose compute was not known during Babbage’s time. He happened to have been in the process of building a Turing complete machine, but that’s largely accidental - he didn’t understand why it could do so many things and it wasn’t until the work of Hilbert, Gödel, Turing and Church that the picture really came into focus and these machines really started to get built consistently and reliably with a key understanding of what features the machine needed to have. Z3 wasn’t Turing complete and z4 can’t start Turing’s work

by mcmoor 1y ago

vlovich123 1y ago

Turing literally defined the mathematical underpinning of what it meant to build a computer in the first place and what operations it needed to support (the language and the computer instruction itself being indistinguishable at the time). Without this work we don’t have any kind of meaningful understanding of how to translate algorithms into something that can run on a machine. Neumann’s contributions were valuable and notable as they proposed to make machines more flexible than how the ENIAC worked, but these are largely just “applied engineering” improvements not foundational theoretical CS concepts; the Harvard architecture is equally valid and flexible as Neumann’s design, even preceding Neumann’s idea of not differentiating code and data. It just that the simplicity at the time of Neumann machines won out even though today we pay for their penalty in other ways. But without Neumann or Harvard architectures you still end up building machines and likely these ideas would eventually have been proposed and tried by others without any meaningful delay in development. Without Turing’s work it’s possible that computers would have been delayed by decades until someone else figured this stuff out. Without the transistor computers would just be an odd curio that’s in use in some places and basically the entire field wouldn’t exist (cryptography wouldn’t be a really thing, AI, databases, programming languages etc).

Animats 1y ago

I think that's right. The ENIAC people were working on ballistics for large guns, and needed to crunch numbers and create printed tables. The Turing team was working on key-testing, like a Bitcoin miner. Colossus was a key tester, trying patterns zipping by on endless loops of tape, and recording matches. Any general purpose machine today can do both, but the early machines were specialized. What general purpose architecture should roughly look like was known all the way back to Babbage. Some kind of long list of instructions, with unconditional and conditional branches. Zuse and Atanasoff both got there before WWII. The big hangup was memory. All early computing struggled with memory limitations, as I mention occasionally. CPUs and instruction decoding came long before enough memory to store programs. Memory was still a million dollars a megabyte in the mid-1970s.

vlovich123 1y ago
No general purpose compute was not known during Babbage’s time. He happened to have been in the process of building a Turing complete machine, but that’s largely accidental - he didn’t understand why it could do so many things and it wasn’t until the work of Hilbert, Gödel, Turing and Church that the picture really came into focus and these machines really started to get built consistently and reliably with a key understanding of what features the machine needed to have. Z3 wasn’t Turing complete and z4 can’t start Turing’s work