From 40b33cb6f8b8dc4d8551c3b522e7c5682a44540a Mon Sep 17 00:00:00 2001 From: Cliff Click Date: Sun, 20 Sep 2026 22:38:58 -0700 Subject: Final reg-alloc across chapters cleanup --- chapter22/README.md | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'chapter22/README.md') diff --git a/chapter22/README.md b/chapter22/README.md index bcb286d0..229c78d9 100644 --- a/chapter22/README.md +++ b/chapter22/README.md @@ -221,8 +221,8 @@ confused with comparing entire compiler chapters: lowering also changes. For example, the frozen Chapter 20 String input has no explicit return, and Chapter 22's default-return handling eliminates that workload at the default seed. -The full suite passes 416 tests plus the fuzzer; the statistics runner passes -55 selected tests. Routine backend comparisons use a fixed optimizer seed: +The full suite and statistics runner pass. Routine backend comparisons use a +fixed optimizer seed: shuffling optimizer worklists should normalize to essentially the same graph, so repeating allocation on those graphs adds little coverage. Use seed variation when investigating optimizer normalization or worklist-order failures. -- cgit v1.2.3