I wonder if large lookup tables/table-driven state machines are still as good as they used to be. After all, even with all the on-chip caches, the additional memory accesses today seem to be slower than doing some multi-instruction SIMD voodoo.
At least the GNU version of wc [0] uses AVX2 for line counting, if available. Though it falls back to a simple character-by-character loop if you ask for a character count [not to be confused with a byte count!] or a word count.
There are only 7 bits in ASCII. An 8th can be used for parity when transmitting data but a regular program will never see it. Anything above 0x7F is simply not a character.
But this is not what's the code doing, is it? It's not doing (ch & 0x7F), it's doing ch <= 0x7F. And the parity checking/filtering is done in the tape drive/serial port driver anyhow, it would never reach wc in the first place.
I am going about the parity bit. 0x46 has odd number of bits set (three, to be precise) so for the parity to check out (that is, the number of bits set has to be even), a parity bit needs to be set and the resulting encoding has to be 0xC6, with four bits set.
Assuming parity is enabled, the parity check is done at a lower level (serial port, TTY driver, etc.) and you'll never see it from the application. I used to mess around with serial ports and terminals a ton in my youth.
The parity bit is not part of the character. It's external, an error detecting device. To read ASCII you always look at bits 6..0, seven bits. You don't filter away the character because it has the parity bit set, you filter off the parity bit (whether it's set or not).
The [z] and [ð] are phonemically different in English, just as [i] and [i:] are, so it'd actually be "Ðe riform is komplijt". American rhotacism prevents us from spelling it "rifoom" as would be proper, unfortunately.
A fun read on word count optimization can be found in Abrash's Black Book:
https://www.jagregory.com/abrash-black-book/#lessons-learned...
You can gloss over the asm if you wish, the tricks that are explained around it are worth it imho.
I wonder if large lookup tables/table-driven state machines are still as good as they used to be. After all, even with all the on-chip caches, the additional memory accesses today seem to be slower than doing some multi-instruction SIMD voodoo.
At least the GNU version of wc [0] uses AVX2 for line counting, if available. Though it falls back to a simple character-by-character loop if you ask for a character count [not to be confused with a byte count!] or a word count.
[0] https://git.savannah.gnu.org/cgit/coreutils.git/tree/src/wc_...
Those `goto`s between two different for loops is crazy.
Assembly / machine code thinking.
More like a relic of (actual) "spaghetti code", it was relatively common in really old Lisp code.
Not that crazy given that it closely mirrors it's state machine structure.
> A word is a maximal string of characters delimited by spaces, tabs or newlines.
And then the actual code explicitly filters out and ignores every character larger than 0x7F. Just why.
Probably because they're not characters. They're just bytes undefined by ASCII.
ASCII is 7 bits (the eight bit would be parity), so that makes perfect sense, in an ASCII world.
So the character e.g. "B" would have this parity bit set and therefore should be filtered out and not count as a letter, in the ASCII world?
There are only 7 bits in ASCII. An 8th can be used for parity when transmitting data but a regular program will never see it. Anything above 0x7F is simply not a character.
Parity bits are not part of the character. They are for detecting transmission errors. You filter off the parity bit before looking at the byte.
But this is not what's the code doing, is it? It's not doing (ch & 0x7F), it's doing ch <= 0x7F. And the parity checking/filtering is done in the tape drive/serial port driver anyhow, it would never reach wc in the first place.
What in the hell are you going on about? B is 0x46 which is < 0x7F.
I am going about the parity bit. 0x46 has odd number of bits set (three, to be precise) so for the parity to check out (that is, the number of bits set has to be even), a parity bit needs to be set and the resulting encoding has to be 0xC6, with four bits set.
Assuming parity is enabled, the parity check is done at a lower level (serial port, TTY driver, etc.) and you'll never see it from the application. I used to mess around with serial ports and terminals a ton in my youth.
The parity bit is not part of the character. It's external, an error detecting device. To read ASCII you always look at bits 6..0, seven bits. You don't filter away the character because it has the parity bit set, you filter off the parity bit (whether it's set or not).
Because they thought that a word is something said in a human language that they can understand.
Mi ne pensas ke lingvoj kiuj usas ekskluzive la basan latinan alfabeton estas komprepeneblaj per si mem.
Cool how my native language is Spanish and I can almost-understand 80% of Esperanto.
Ze riform iz komplit.
The [z] and [ð] are phonemically different in English, just as [i] and [i:] are, so it'd actually be "Ðe riform is komplijt". American rhotacism prevents us from spelling it "rifoom" as would be proper, unfortunately.
The brevity carried over to Plan 9. Re-posting my older comment (https://news.ycombinator.com/item?id=4023385):
http://en.wikipedia.org/wiki/Plan_9_from_Bell_Labs follows the Unix philosophy. A lot of legacy has been shed. I can count 13 options to ls, 11 options to sed and just 5 to sed.
The standard Plan 9 shell, Rc, is described in mere ~500 lines of manpage, while Bash takes whooping ~5400 lines.
Oh, and there is no `dll hell' in P9 :-)