Skip to content

Conversation

@OCHyams
Copy link
Contributor

@OCHyams OCHyams commented Jul 18, 2025

Rather than visit all instructions each iteration of the analysis, we can cache the interesting ones ahead of time while performing other setup related tasks.

Marginal performance improvement. Note in this compile-tim-tracker comparison I've enabled assignment tracking support with lto, which isn't the same as on main:
http://llvm-compile-time-tracker.com/compare.php?from=80d92fc648d0160e0bbd191ba67b0f05218d4b69&to=1a38acc6e758919bbce5a451b7fe39c89a134c01&stat=instructions%3Au

@OCHyams OCHyams requested review from SLTozer and jmorse July 18, 2025 13:43
// Process the instructions in "frames". A "frame" includes a single
// non-debug instruction followed any debug instructions before the
// next non-debug instruction. A pointer-int int value of 1 on an
// instruction means non-debug instructions follow (end of frame).
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

err, note to self, this comment has the bit value the wrong way around

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

2 participants