4 ms·

One wonders whether this is a - cost cutting measure. Maintain the zOS and AIX/POWER backends, kick the rest of the XL compilers teams. or - Advance the stat

by jabl 7y ago

One wonders whether this is a - cost cutting measure. Maintain the zOS and AIX/POWER backends, kick the rest of the XL compilers teams. or - Advance the state of the art in compilers, by avoiding as much wheel reinvention as before.

apaprocki 7y ago
A mix of both? It seemed obvious by C++11 completion in GCC/Clang that achieving later standards in the vendor compiler front-end implementations would be a huge investment of time with diminishing returns. Since the C++ committee adopted the planned 3 year release cycle, that kind of sealed their fate (IMO). It was much easier to coast when there was an 8-year hiatus between standards (03 to 11) in the “modern” C++/Linux age.
- chungus_khan 7y ago
  It's a lot like web tech and browsers. It used to be a much more manageable effort to create an entirely new browser engine, but now that the standard is larger and faster moving, smaller entrants haven't been able to keep up. It's an interesting consequence of these faster moving targets that they trend towards consolidation down to two competing entrants in both cases.
  - PaulHoule 7y ago
    That was 1995 when a "new browser engine" was easy. Not long, however, open source web browsers developed a bad reputation. Back then one reason was that 40% or so of all web pages had broken HTML (e.g. <tag) but that Netscape rendered them anyway. To make a usable web browser you needed to render broken HTML the same way Netscape did, which was undocumented.
    - erichocean 7y ago
      Sure, but that specific issue has been resolved 100% AFAIK with the HTML 5 parsing algorithm being fully specified now.
      - PaulHoule 7y ago
        That's right. Overall the HTML 5 specification represents a huge step in specificity over other specs in the industry. On the other hand if you want to make a modern web browser now you need a very fast Javascript engine and a highly complex rendering engine...
whatsmyusername 7y ago
It's IBM soooooooo...
nitrud 7y ago
IBM's investment in LLVM and Clang open source technology is not a cost cutting measure, it is the opposite; IBM is increasing overall investment in compiler technology and improving LLVM in general, as evidenced by recent contributions to AIX enablement and loop optimizations. By leveraging this very robust and successful open source compiler infrastructure, IBM can prevent duplication of effort as much as possible.

by jabl 7y ago

apaprocki 7y ago

A mix of both? It seemed obvious by C++11 completion in GCC/Clang that achieving later standards in the vendor compiler front-end implementations would be a huge investment of time with diminishing returns. Since the C++ committee adopted the planned 3 year release cycle, that kind of sealed their fate (IMO). It was much easier to coast when there was an 8-year hiatus between standards (03 to 11) in the “modern” C++/Linux age.

chungus_khan 7y ago
It's a lot like web tech and browsers. It used to be a much more manageable effort to create an entirely new browser engine, but now that the standard is larger and faster moving, smaller entrants haven't been able to keep up. It's an interesting consequence of these faster moving targets that they trend towards consolidation down to two competing entrants in both cases.
- PaulHoule 7y ago
  That was 1995 when a "new browser engine" was easy. Not long, however, open source web browsers developed a bad reputation. Back then one reason was that 40% or so of all web pages had broken HTML (e.g. <tag) but that Netscape rendered them anyway. To make a usable web browser you needed to render broken HTML the same way Netscape did, which was undocumented.
  - erichocean 7y ago
    Sure, but that specific issue has been resolved 100% AFAIK with the HTML 5 parsing algorithm being fully specified now.
    - PaulHoule 7y ago
      That's right. Overall the HTML 5 specification represents a huge step in specificity over other specs in the industry. On the other hand if you want to make a modern web browser now you need a very fast Javascript engine and a highly complex rendering engine...

whatsmyusername 7y ago

It's IBM soooooooo...

nitrud 7y ago

IBM's investment in LLVM and Clang open source technology is not a cost cutting measure, it is the opposite; IBM is increasing overall investment in compiler technology and improving LLVM in general, as evidenced by recent contributions to AIX enablement and loop optimizations. By leveraging this very robust and successful open source compiler infrastructure, IBM can prevent duplication of effort as much as possible.