4 ms·
Amusing. Over at Berkeley, they developed a new, open CPU design that is intended to be useful for pretty much everything from education to industry, and to ke
by hga 12y ago
Amusing. Over at Berkeley, they developed a new, open CPU design that is intended to be useful for pretty much everything from education to industry, and to keep it simple it has no support for integer errors beyond making it cheap to detect divide by zero: http://riscv.org/ http://riscv.org/
- renox 12y agoGiven that the MIPS (introduced in 1981) had integer operations with traps on overflow, I find very strange this reasoning.
- hga 12y agoIn part, perhaps this is Worse is Better AKA the New Jersey way. One part is that this makes it simpler for education, but I'm not convinced, especially by their rationalizations in the published ISA, that this is acceptable for "industry", which they claim to also be aiming for. The long integer cryptography people certainly are distressed.
- renox 12y agoI think that we're talking about different things: you're talking cryptography people who need carry bit and integer instructions with carry to efficiently implement long integers I'm talking about programmers who likes when their programs core dump instead of returning them garbage (could also be useful for security maybe) and 'trap on integer overflow' instructions help here.
- hga 12y agoActually I'm meaning to talk about both; I wanted to bring up the long integer folks because that's a less obvious need to embed integer overflow machinery in your CPU.