Cannot really add anything valuable to this discussion since my experience is too little, just wanted to say thank you for your tremendous effort because you are, and i think by far, the best teacher I have had on this and your course about performance aware programming i have recently started is just overwhelmingly amazing and I enjoy every second of it. It sparks a joy in programming that i have seemed to bury in the years of web development i have lost myself in haha
Original Spectre paper explicitly mentions ARM and RISC-V as ISA's that can be vulnerable to speculative attacks: https://spectreattack.com/spectre.pdf
Of course actual hw implementation matters, it is not a property of ISA.
Yes. The original article seemed to be confused about a lot of things like this, and seemed to imply that somehow x64 was unique in being pipelined and having speculative execution, among other things. It didn't make a lot of sense.
As someone who is passionate about software development and continuously striving to improve, I would greatly appreciate your perspective on the "best" approach to learning programming effectively. With the vast array of resources, programming languages, and learning methodologies available, it can be overwhelming to determine the optimal path.
He really has good energy in this video! Usually people who are too excited and call things they don't fully understand/remember super cool turn me off but for some reason, he seems genuine to me.
Maybe the article is bad, but I too recently started thinking that Arm is probably going to take over. As I understand, it is easier to evolve RISC architecture than CISC. If that is not true, then I don't understand the benefit of having RISC at all.
As I mentioned, and as the Chips and Cheese response article says as well, RISC vs. CISC isn't a particularly meaningful distinction anymore. Desktop and mobile ISAs and microarchitectures are too complex nowadays for anything to really be considered RISC in my opinion. RISC-V, for example, has very complex instructions in its instruction set, even if you only consider reasonable extensions that people need (like RISC-V V). I defy anyone to argue that LMUL and VGATHER, for example, are "RISC-like" instructions in the sense that the term RISC originally meant.
Cannot really add anything valuable to this discussion since my experience is too little, just wanted to say thank you for your tremendous effort because you are, and i think by far, the best teacher I have had on this and your course about performance aware programming i have recently started is just overwhelmingly amazing and I enjoy every second of it. It sparks a joy in programming that i have seemed to bury in the years of web development i have lost myself in haha
Incredible Stuff Casey!
Original Spectre paper explicitly mentions ARM and RISC-V as ISA's that can be vulnerable to speculative attacks: https://spectreattack.com/spectre.pdf
Of course actual hw implementation matters, it is not a property of ISA.
Yes. The original article seemed to be confused about a lot of things like this, and seemed to imply that somehow x64 was unique in being pipelined and having speculative execution, among other things. It didn't make a lot of sense.
- Casey
Dear Casey Muratori,
As someone who is passionate about software development and continuously striving to improve, I would greatly appreciate your perspective on the "best" approach to learning programming effectively. With the vast array of resources, programming languages, and learning methodologies available, it can be overwhelming to determine the optimal path.
Thank you for your time and consideration.
He really has good energy in this video! Usually people who are too excited and call things they don't fully understand/remember super cool turn me off but for some reason, he seems genuine to me.
Maybe the article is bad, but I too recently started thinking that Arm is probably going to take over. As I understand, it is easier to evolve RISC architecture than CISC. If that is not true, then I don't understand the benefit of having RISC at all.
As I mentioned, and as the Chips and Cheese response article says as well, RISC vs. CISC isn't a particularly meaningful distinction anymore. Desktop and mobile ISAs and microarchitectures are too complex nowadays for anything to really be considered RISC in my opinion. RISC-V, for example, has very complex instructions in its instruction set, even if you only consider reasonable extensions that people need (like RISC-V V). I defy anyone to argue that LMUL and VGATHER, for example, are "RISC-like" instructions in the sense that the term RISC originally meant.
- Casey