Add max-rss for gimli

Author: Mark-Simulacrum

triage/2020-08-03.md

@@ -25,7 +25,10 @@ Regressions
   A 1.5% regression on the ctfe stress test.
 * [std: Switch from libbacktrace to gimli (take 2) #74682](https://github.com/rust-lang/rust/pull/74682#issuecomment-668264270)
   ([instructions](https://perf.rust-lang.org/compare.html?start=cfc572cae2d1fc381cce476b5c787fd7221af98c&end=c058a8b8dc5dea0ed9b33e14da9e317e2749fcd7&stat=instructions%3Au))
-  An expected loss of up to 20% on smaller benchmarks. See discussion on
+  An expected loss of up to 20% on smaller benchmarks. This was also a
+  [max-rss](https://perf.rust-lang.org/compare.html?start=cfc572cae2d1fc381cce476b5c787fd7221af98c&end=c058a8b8dc5dea0ed9b33e14da9e317e2749fcd7&stat=max-rss)
+  regression, though how big is difficult to determine as this statistics is
+  noisy. See discussion on
   [Zulip](https://rust-lang.zulipchat.com/#narrow/stream/247081-t-compiler.2Fperformance/topic/gimli.20regressions).
   Optimistically, there is a follow-up PR, "metadata: track the simplified Self type for every trait impl
   #75008", which may [eliminate](https://github.com/rust-lang/rust/pull/75008#issuecomment-667577993) a good portion of these losses.