[PATCH] D94739: ADT: Fix reference invalidation in SmallVector::emplace_back and the size+value version of SmallVector::assign

[Duncan P. N. Exon Smith via Phabricator via llvm-commits]

dexonsmith added a reviewer: nikic.
dexonsmith added a subscriber: nikic.
dexonsmith added a comment.

I did a spot check of code size / compile-time, capturing clang's binary size in KiB (78460KiB => 78484KiB) and 5 runs of compiling VirtualFileSystem.cpp.o (looks like it's in the noise):

  78460	baseline/build/bin/clang
          3.84 real         3.74 user         0.09 sys
          3.85 real         3.75 user         0.09 sys
          3.83 real         3.73 user         0.09 sys
          3.84 real         3.73 user         0.10 sys
          3.84 real         3.74 user         0.09 sys
  78484	invalidation/build/bin/clang
          3.83 real         3.73 user         0.09 sys
          3.87 real         3.77 user         0.09 sys
          3.83 real         3.73 user         0.09 sys
          3.83 real         3.72 user         0.10 sys
          3.90 real         3.80 user         0.09 sys

"baseline" is ceaf0110ff5e0c2de1f03d65d13703d34d0d5737 <https://reviews.llvm.org/rGceaf0110ff5e0c2de1f03d65d13703d34d0d5737>, and "invalidation" has this patch applied on top.

I also checked a big file, X86ISelLowering.cpp.o:

  78460	baseline/build/bin/clang
         31.05 real        30.70 user         0.34 sys
         30.83 real        30.46 user         0.36 sys
         30.94 real        30.59 user         0.34 sys
         31.15 real        30.76 user         0.38 sys
         30.80 real        30.43 user         0.36 sys
  78484	invalidation/build/bin/clang
         31.67 real        30.71 user         0.37 sys
         30.90 real        30.53 user         0.36 sys
         31.53 real        31.16 user         0.35 sys
         31.30 real        30.93 user         0.36 sys
         31.01 real        30.65 user         0.34 sys

Best of 5 gives 30.43 vs. 30.53, which is less than 0.4%; not sure if that's significant (certainly my environment isn't stabilized, so it's not a high quality result).

@nikic, maybe I can trigger your compile-time tracker myself (I just sent you an email), but if that doesn't work out, any chance you could kick off a perf run for this patch as well?


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D94739/new/

https://reviews.llvm.org/D94739