3 ms·

Yes, it's correct modern C. I'm not sure what that paragraph is supposed ot mean. AFAICT there is no difference between old and modern C in how the cases can ap

Yes, it's correct modern C. I'm not sure what that paragraph is supposed ot mean. AFAICT there is no difference between old and modern C in how the cases can appear in the body of the switch. Here is the first edition of The C Programming Language on switch: > switch (expression) statement >The usual arithmetic conversion is performed on the expression, but the result must be int. The statement is typically compound. Any statement within the statement may be labeled with one or more case prefixes as follows > case constant_expression : https://archive.org/details/TheCProgrammingLanguageFirstEdition/page/n209/mode/2up https://archive.org/details/TheCProgrammingLanguageFirstEdit...

SV_BubbleTime 6y ago
I can’t at all support this with a source, but I had THOUGHT the register keyword needed to make this fast was largely up to how the compiler decides to use it. That register can’t be required to be register vs ram, but it’s encouraged to be IF the compiler can do it. So I thought the downside was you can declare “register” but it might not obey but then definitely can not use the & to get the memory address by using they keyword. Which on portable code register isn’t recommended. So I suppose it’s not entirely fair to say all modern or all past support this. Right?
- marvy 6y ago
  You're tangling a bunch of stuff together in a confusing mess. Let's take one thing at a time. 1. You're right that you can't use the & operator on a variable that was declared using the register keyword. 2. However, if you look very carefully at the example code, it actually follows this rule. In fact, it doesn't use the & operator at all. There's a lot of pointers all over the place, but no usage of the & operator. 3. You're also right that the register keyword is just a hint that the compiler is free to ignore. 4. Expanding on the previous point: modern compilers usually will ignore it, but in the olden days, compilers were not so good at register allocation, so this keyword was needed if you wanted speeds competitive with assembly. Since this is code from the 80s, no surprise that it uses this keyword. 5. Now that we've got all that out of the way: the register keyword is a complete distraction here, and has nothing at all to do Duff's device. Since modern compilers ignore it anyway, you can take the original example and just delete all uses of the word "register". The point of Duff's device is how to do manual loop unrolling without then having a special "pre-loop" or "post-loop" which takes care of "leftovers" because your array size was not divisible by 8, or 4, or whatever unroll factor you happened to use.

SV_BubbleTime 6y ago

I can’t at all support this with a source, but I had THOUGHT the register keyword needed to make this fast was largely up to how the compiler decides to use it. That register can’t be required to be register vs ram, but it’s encouraged to be IF the compiler can do it. So I thought the downside was you can declare “register” but it might not obey but then definitely can not use the & to get the memory address by using they keyword. Which on portable code register isn’t recommended. So I suppose it’s not entirely fair to say all modern or all past support this. Right?

marvy 6y ago
You're tangling a bunch of stuff together in a confusing mess. Let's take one thing at a time. 1. You're right that you can't use the & operator on a variable that was declared using the register keyword. 2. However, if you look very carefully at the example code, it actually follows this rule. In fact, it doesn't use the & operator at all. There's a lot of pointers all over the place, but no usage of the & operator. 3. You're also right that the register keyword is just a hint that the compiler is free to ignore. 4. Expanding on the previous point: modern compilers usually will ignore it, but in the olden days, compilers were not so good at register allocation, so this keyword was needed if you wanted speeds competitive with assembly. Since this is code from the 80s, no surprise that it uses this keyword. 5. Now that we've got all that out of the way: the register keyword is a complete distraction here, and has nothing at all to do Duff's device. Since modern compilers ignore it anyway, you can take the original example and just delete all uses of the word "register". The point of Duff's device is how to do manual loop unrolling without then having a special "pre-loop" or "post-loop" which takes care of "leftovers" because your array size was not divisible by 8, or 4, or whatever unroll factor you happened to use.