[Heap] Don't try and find SFIs FunctionLiteralId when flushing it's bytecode.
For compiled SFIs, finding the FunctionLiteralId is a slow operation that requires doing a binary search in the script's SFI table. Given the SFI has been flushed, it is unlikely we will need the FunctionLiteralId again, so just mark the UncompiledData has having an invalid FunctionLiteralID such that we only do the search if it is required from the UncompiledData. This addresses a significant regression on gc_latency when bytecode flushing was enabled. BUG=chromium:927038,v8:8395 Change-Id: I7123c3fe05034ff96e84f9d1d36d5f312a2a56e0 Reviewed-on: https://chromium-review.googlesource.com/c/1450118Reviewed-by: Ulan Degenbaev <ulan@chromium.org> Commit-Queue: Ross McIlroy <rmcilroy@chromium.org> Cr-Commit-Position: refs/heads/master@{#59319}
Showing
Please
register
or
sign in
to comment