| commit | b4cf5e32a94024c1bfdc6ca432677c31306a3fb5 | [log] [tgz] |
|---|---|---|
| author | Googler <noreply@google.com> | Tue Sep 13 22:43:37 2016 +0000 |
| committer | Dmitry Lomov <dslomov@google.com> | Wed Sep 14 09:37:13 2016 +0000 |
| tree | 39b4daeff52f3a8d12ac3c477c73a43e8b3501bd | |
| parent | f92ecb3d12a8a966524cda067029dc0aaae409b3 [diff] |
Change C++ singlejar to use stdio for all output, instead of a mix of sendfile() and write() without buffering. sendfile() is not a speedup for typical jar contents; the average compressed class file is less than 2KB, so memcpy()ing the data into a large buffer is cheaper than the extra system calls. The existing non-sendfile code was making four calls to lseek() per archive member. I removed all of those by using pread() or caching the output position. Configure stdio to use a 128KB buffer to make fewer write() calls to the output file. This is a noticeable speedup over the default when writing to a fuse filesystem. (I think this wouldn't be necessary if our fuse filesystems would set st_blksize appropriately in stat() results.) Remove needless thread-hostile calls to umask(). -- MOS_MIGRATED_REVID=133057985
{Fast, Correct} - Choose two
Bazel is a build tool that builds code quickly and reliably. It is used to build the majority of Google‘s software, and thus it has been designed to handle build problems present in Google’s development environment, including:
A massive, shared code repository, in which all software is built from source. Bazel has been built for speed, using both caching and parallelism to achieve this. Bazel is critical to Google's ability to continue to scale its software development practices as the company grows.
An emphasis on automated testing and releases. Bazel has been built for correctness and reproducibility, meaning that a build performed on a continuous build machine or in a release pipeline will generate bitwise-identical outputs to those generated on a developer's machine.
Language and platform diversity. Bazel's architecture is general enough to support many different programming languages within Google, and can be used to build both client and server software targeting multiple architectures from the same underlying codebase.
Find more background about Bazel in our FAQ.