A pet peeve I have is people say computers see things as "0s and 1s" but the thing is the 0s and 1s are not addressable, bytes are and are usually the smallest unit you can do math on, it would be more accurate to say computers see everything in terms of numbers between 0 and 255
-
Show this thread
-
It's not like we have obscure British microcomputers with 5 bit words floating around these days we have to be careful to be inclusive of, the byte won, that's just what a computer is now, we can acknowledge it
5 replies 5 retweets 72 likesShow this thread -
Replying to @mcclure111
there is a concentrated effort in LLVM right now to add support for computers with non-8-bit bytes, including non-multiple-of-8 bytes. it is led by multiple very large companies
3 replies 0 retweets 19 likes -
-
Replying to @mcclure111
nope. one of these companies is Synopsys which should tell you everything
2 replies 0 retweets 4 likes -
Replying to @whitequark
*scratches head* wait. do they want to use LLVM with *FPGAs*?
1 reply 0 retweets 7 likes -
Replying to @mcclure111
i assume it means they have some specialized softcores they want to use LLVM with that have like 18-bit bytes or something. i know there are many DSPs with 16-bit bytes but they're going further still
3 replies 0 retweets 6 likes
I would assume it's for high level synthesis (HLS). Most C/C++/OpenCL HLS tools are based on clang and llvm. And in HLS you want to have C datatypes of any arbitrary bit width.
-
-
Their ASIP tool is indeed a HLS tool. What they do is let you specify a custom processor that is tailored for your specific code. You can tweak the processor and see how it affects the design size and performance. I would guess non 8 bit comes from the instruction size.
0 replies 0 retweets 0 likesThanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.