7f15f3e7ad
With concurrent inlining, the inlining phase happens on the background thread and the data needed for the inlining phase is serialized on the main thread. The serialization phase tries to gather data about functions called which is sometimes more expensive than inlining phase itself. So it's better not to use concurrent inlining for TurboFan compilations when tiering up from Turboprop to TurboFan. Turboprop compilations don't inline and hence it is OK to continue using concurrent inlining for Turboprop compilations. Bug: v8:9684 Change-Id: Ib529905213fa7f0df84ee52218adc27f7c219f60 Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2557504 Commit-Queue: Mythri Alle <mythria@chromium.org> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> Cr-Commit-Position: refs/heads/master@{#71405} |
||
---|---|---|
.. | ||
local | ||
objects | ||
outproc | ||
testproc | ||
utils | ||
__init__.py | ||
base_runner.py | ||
num_fuzzer.py | ||
OWNERS | ||
PRESUBMIT.py | ||
standard_runner.py | ||
test_config.py | ||
trycatch_loader.js |