[lld] Benchmark of the last few (800ish) commits

[Rafael Espíndola via llvm-commits]

I left my desktop running lld benchmarks over the weekend and got the
result for that last 800 revisions or so.

A good news is that the relocation processing rewrite was an
improvement in most tests. The chrome graph is fairly representative.
The other good news is that the regression that I first noticed in the
linux kernel seems to have been a fluke.

The bad news is that there seems to be quite a bit of variation from
revision to revision. The graph of branch misses in particular for
chrome is typical: large variations when code that is not executed is
changed.

I will try to remember to implement -falign-functions in clang and
experiment with -falign-functions=32 and -falign-functions=64 to see
if that improves the stability.

Cheers,
Rafael
-------------- next part --------------
A non-text attachment was scrubbed...
Name: chrome.png
Type: image/png
Size: 570533 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20180115/02974fd1/attachment-0003.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: linux.png
Type: image/png
Size: 629904 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20180115/02974fd1/attachment-0004.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: chrome-branch-miss.png
Type: image/png
Size: 438836 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20180115/02974fd1/attachment-0005.png>