From d82558b5217645dbb847f47ee450f6364a6ddc66 Mon Sep 17 00:00:00 2001 From: Dmitriy Vasyura Date: Wed, 19 Nov 2025 11:22:46 -0800 Subject: [PATCH] Apply suggestion from @Copilot Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com> --- .github/prompts/micro-perf.prompt.md | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/.github/prompts/micro-perf.prompt.md b/.github/prompts/micro-perf.prompt.md index 17eb3d6800e..dbfad2dfc27 100644 --- a/.github/prompts/micro-perf.prompt.md +++ b/.github/prompts/micro-perf.prompt.md @@ -20,7 +20,7 @@ If the user provided explicit list of classes or functions to optimize, scope yo 3. Optimized function or class name should have the same name as original with '_new' suffix. 4. Create a file with '..perf.js' suffix with perf tests. For example if you are using model 'Foo' and optimizing file name utils.ts, you will create file named 'utils.foo.perf.js'. 5. **IMPORTANT**: You should use ESM format for the perf test files (i.e. use 'import' instead of 'require'). -5. The perf tests should contain comprehensive perf tests covering identified scenarios and common cases, and comparing old and new implementations. -6. The results of perf tests and your summary should be placed in another file with '..perf.md' suffix, for example 'utils.foo.perf.md'. -7. The results file must include section per optimized definition with a table with comparison of old vs new implementations with speedup ratios and analysis of results. -8. At the end ask the user if they want to apply the changes and if the answer is yes, replace original implementations with optimized versions but only in cases where there are significant perf gains and no serious regressions. Revert any other changes to the original code. +6. The perf tests should contain comprehensive perf tests covering identified scenarios and common cases, and comparing old and new implementations. +7. The results of perf tests and your summary should be placed in another file with '..perf.md' suffix, for example 'utils.foo.perf.md'. +8. The results file must include section per optimized definition with a table with comparison of old vs new implementations with speedup ratios and analysis of results. +9. At the end ask the user if they want to apply the changes and if the answer is yes, replace original implementations with optimized versions but only in cases where there are significant perf gains and no serious regressions. Revert any other changes to the original code.