News for hackers. Stuff that matters.

A comment on A Simple Graph-Based Intermediate Representation (1995) [pdf]. Parent comment

Re:A Simple Graph-Based Intermediate Representation (1995) [... by pizlonator on Monday February 03, 2020 @01:58PM (#22227599)
Meh. Seems over engineered and too general. Maybe it’ll be cool for ML. Probably best left ignored for anything else.

The whole thing about using it for C++ optimizations... I’ll believe it when I see a statistically significant speedup. Until then it’s vapor.

[ Parent | on HN ]
Re:A Simple Graph-Based Intermediate Representation (1995) [... by halayli on Monday February 03, 2020 @03:39PM (#22228749)
you don't need to limit yourself to c/c++. It can be used for various other reasons.

As for c/c++ specifically, you can optimize much better and safer if you have 1 or more intermediate dialects prior to llvm's IR. You don't need to believe it. This is compiler 101.

[ Parent | on HN ]
Re:A Simple Graph-Based Intermediate Representation (1995) [... by pizlonator on Monday February 03, 2020 @04:46PM (#22229539)
Of course you can optimize better if you have the right intermediate dialects.

I’m saying that MLIR is unproven for that purpose outside of the ML space.

[ Parent | on HN ]