Jens Johansen
c812abd7bf
When compiling the cfe with an aot-compiled cfe this removes ~43 mb of context allocations in `BinaryPrinter._ensureCanonicalName` and has these changes when benchmarking: ``` msec task-clock:u: -5.0067% +/- 0.6053% (-288.75 +/- 34.91) (5767.31 -> 5478.56) page-faults:u: 2.7077% +/- 0.1426% (2942.48 +/- 155.01) (108670.96 -> 111613.44) cycles:u: -5.3274% +/- 0.6265% (-1278360098.44 +/- 150341087.36) (23995901283.80 -> 22717541185.36) instructions:u: -4.3319% +/- 0.0009% (-1195734804.64 +/- 241700.46) (27602908367.64 -> 26407173563.00) branch-misses:u: -3.4011% +/- 3.0065% (-3245774.08 +/- 2869267.25) (95434325.44 -> 92188551.36) seconds time elapsed: -5.0025% +/- 0.6041% (-0.29 +/- 0.03) (5.77 -> 5.48) seconds user: -5.3148% +/- 0.6838% (-0.29 +/- 0.04) (5.52 -> 5.22) Scavenge( new space) goes from 65 to 64 MarkSweep( promotion) goes from 10 to 9 Evacuate(store buffer) goes from 2 to 1 ``` although most of it is probably a happy accident with the GCs. Compiling the analysis server only gives: ``` page-faults:u: 0.6385% +/- 0.1266% (1058.68 +/- 209.95) (165813.00 -> 166871.68) instructions:u: -0.0158% +/- 0.0005% (-6794934.20 +/- 222091.71) (42904117069.28 -> 42897322135.08) Scavenge( new space) goes from 103 to 101 ``` Change-Id: I85ceece0df699bb2393f3ab72fe6130134b49132 Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/438681 Reviewed-by: Johnni Winther <johnniwinther@google.com> Commit-Queue: Jens Johansen <jensj@google.com>