Whether to develop a new programming language?
In fact, not once asking about redundancy
and excessive universalization of existing solutions...
The fact that the optimum solutions in the field of compilers
and of programming language today NO.
And here's why:
— optimization for specific processor architecture (CPU GPU Il) something quite abstract and decided it is not experienced by the system and not even a heuristic as simple DFAS (if...then) because of possible large loads. It would seem the solution is quite logical and optimal, but with the advent of clouds, clusters and other parallel buns it would be possible to find a "faster" comprehensive solution, and it is still there... but who need to come up with bikes with turbo?
— a man solves the problems of abstraction applications in most cases mandatory
a "black box" and a stack of templates (ideally 2-3), so there appear sablono, zombies, and other newproject. "0_o I discovered the world erlang'a" Yura said. "And we do not care." said team a.
Already happened to me more than once — in one case, the PLO in a different function... in the third brainfuck and website on fasm'e. who cares?.. all do not care — "just worked" no one thinks about speed and ephemeral "optimality". So I will say that people have too much freedom! Abstraction and the abstraction that would say this is a ball. but inside he decides to let the language itself, if the desired methane is methane, if we need silver that is silver.
— 0_o we all like the STL, it is so soft and fluffy... there are so many buns =)
why should we teach those fucking algorithms? why think about structure?
It may be redundant,- what are you talking about... works fine, even if it is slowly...
but it works. And so every day... people often try to solve the problem of the computation of optimality of algorithms for the final implementation, but almost all and always waiting for a complete epic fail.
In the words of uncle Knut: "low-level optimization is the root of all problems."
And here, too, people have too much freedom...
Actually what I?.. and the moral is quite simple:
existing solutions are partly integrated but because aparataj loads key points summarized and "universalservice" which leads to the loss of "optimality" of the output.
That is, a Universal solution may not be optimal, the only comprehensive...
A huge problem in our time is precisely the optimization, no matter under what multicore, multithreading, load sharing, low-level...
Let us for this is such a ephemeral torrent compiler Ala Skynet.
We do not care about any firmware, time, and memory.
We need the compiler "without compromise" and paradigms of the times і386.
IN fact never were the highlight performance — and uparatna platform already far ahead.
That's all, I think the reasons for "the decline of compilationunit" all clear.
Interested in the standard question: "but is the game worth the candle?"
I started development of such a "miracle" until (yet) for 1.5 years perfectly sculpted Mat. part.
Just don't want to make google wave.
Don't want to be someone purchased.
Want to find edinomyshlenikov.
PS in Advance I apologize for possible admittance errors — Russian not my native language.
Also I apologize for the unnecessary confusion of the presentation.
We hope for your understanding... Just "it's" already on order got.