That's pretty cool. I've worked on projects where the link stage would take minutes, and therefore constitute most of the incremental build time, and it's not like you can avoid it. Visual C++ has a special incremental linker, it always surprised me that nothing similar ever hit the mainstream GNU toolchain. It'll be interesting to see if this gains traction.

I'm just waiting for Google to start making their own motherboards, chipsets, and CPU architectures. That seems to be the logical conclusion of their tendency to respond to the uniquely demanding requirements of their mission by creating new, improved infrastructure.

Nope, far from it, they're using the most commodity hardware infrastructure they can find. Their whole system is built around cheap, widely-available hardware. Even with the volumes they are using, they would not get production high enough to compete with the commodity PC market on price.

They are trying to get more efficient power supplies.. http://tech.slashdot.org/article.pl?sid=06/09/26/2039213

I guess it wasn't obvious that I wasn't really seriously waiting for Google to start slinging wafers.

how they monetize this? it ads adsense for every new executable? :)

Some intern probably wrote a MapReduce job which discovered that Google engineers collectively spend thousands of hours per year linking, so writing a new linker can be justified as a cost savings. :-)

OTOH, http://xkcd.com/303/

doesnt link linux. FAIL.

jk.. this is one of the better projects ive seen come out of google. 5 times speed increase is huge for big apps. does anyone know if it was a 20% project? hopefully this gets picked up and ported around so that it becomes a viable alternative to the gcc linker.

He doesn't fully realize that he is the only one (with the other writing code for this project) that is really giving back something, not "Google". This corporate thinking is sad.

Although you may be partially right in this case, I think that people take on this corporate-hating mentality too often (against Google, against Sun, against Microsoft even!). Give credit where it's due - it's a pretty impressive thing for a company to let its employee's spend 20% of their time on their own interests, and then pay them for it.

Credit to google for the openmindness, but 1) Google is doing this for its own interests of course. 2) this does not change the fact that the main credit is due to the guy that wrote the code that could have used his time to produce something not useful for the external community.

Google paid him for doing this.

FINALLY! Now can someone please rewrite GCC? That is the ugliest piece of code I've ever seen.

Is it hindering innovation to that extent to justify the rewrite? Remember, GCC is the one of the most fundamental pieces of software running the entire OSS eco-system. It's internal beauty isn't so much important as it's utility to this system.

Apparently GCC is crimping Apple's style enough to justify a rewrite, although that may be due to licensing and lack of modularity more than ugly code.

Spoken like someone who's never tried to port GCC to a new CPU. It's seriously unfun.

Check out clang, it's looking pretty promising.

It's being done. It's called LLVM. Apple have been sponsoring it.

I wish I knew someone still running gentoo so I could see their weeping for joy.

Another open source release by google which no one really needs. Is is it really that important that the linking takes a few minutes less. There are probably tons of interesting projects at google which they could release, but they are too evil to do that, because they fear their competitors. Not like yahoo and microsoft, which give much more back to the community then google.