ECPoint_mul/ec_points_mul take a lot of time on Arm32
Categories
(NSS :: Libraries, defect, P3)
Tracking
(Not tracked)
People
(Reporter: jesup, Unassigned)
References
(Blocks 1 open bug)
Details
(Keywords: perf, perf:pageload, Whiteboard: [geckoview:p2])
Attachments
(1 obsolete file)
This was from loading allrecipes.com. for a 720ms block of time on socket thread, most of it was spent in ECPoint_mul (and below) or ec_points_mul (and below); some was spent in ec25519 (bug 1550579).
Comment 1•5 years ago
|
||
The current logic is that if the compiler somehow uses gold rather than
BFD ld by default, we let it, but if it uses BFD ld, we default to lld.
When doing Android builds, the compiler finds the linker in the NDK, and
the default ld
binary is gold. So we currently end up using fold for
Android builds.
Here, we change the logic such that we use lld when the default linker
the compiler uses is either BFD ld or gold. We can't go with "is not
lld" because the other possible kind, ld64, is what we actually want to
use on mac, since lld doesn't support mach-o fully just yet.
Comment 2•5 years ago
|
||
Comment on attachment 9070398 [details]
Bug 1557505 - Default to lld linker for local builds in more cases.
Revision D34040 was moved to bug 1557507. Setting attachment 9070398 [details] to obsolete.
Updated•5 years ago
|
Updated•5 years ago
|
Comment 3•4 years ago
|
||
We spend 364ms total on a S7 arm64 device, https://share.firefox.dev/3aIYTjj. Although it's not in one big block so I'm inclined to say this is a qf:p3.
Updated•3 years ago
|
Updated•2 years ago
|
Description
•