maxpan

11 ms·

1. 1.▲
   by maxpan 2y ago

   The quote above talks exclusively of instruction cache misses. In case you are really interested, the two kinds are related as L2 and L3 caches are shared my instructions and data. In terms of the execution profile, the kernel is very close
2. 2.▲
   by maxpan 2y ago

   Disclaimer: I'm an active developer of BOLT. It takes BOLT less than 10 second to optimize the Linux kernel once the profile is collected. There's no need for a distributed build system to take advantage of that. Overall, we have
3. 3.▲
   by maxpan 2y ago

   The performance loss due to cache misses in data-center applications far exceeds 5%. Combined data and instruction cache misses are contributing to more than half of stalled cycles. The following publication by Google from 2015 goes into de
4. 4.▲
   by maxpan 2y ago

   The gains mentioned are on top of compiler's PGO+LTO.
5. 5.▲
   by maxpan 8y ago

   Not yet, but the support is coming.
6. 6.▲
   by maxpan 8y ago

   BOLT can optimize the compiler itself. Either GCC or Clang.
7. 7.▲
   by maxpan 8y ago

   In many cases BOLT complements AutoFDO. AutoFDO affects several optimizations, and code layout is just one of them. Another critical optimization influenced by AutoFDO/PGO is function inlining. After inlining a callee code profile is o
8. 8.▲
   by maxpan 8y ago

   There's no need for another build as BOLT runs directly on a compiled binary, and could be integrated into an existing build system. Operating directly on a machine code allows BOLT to boost performance on top of AutoFDO/PGO and L