From 043c3514ed9a7ddf3bd8e680405759953fd4b9c3 Mon Sep 17 00:00:00 2001 From: Huanyi Chen Date: Sun, 4 Feb 2024 20:27:41 -0500 Subject: [PATCH] Add after-action report --- lectures/flipped/L13.md | 13 ++++++++----- lectures/flipped/L14.md | 8 ++++++++ 2 files changed, 16 insertions(+), 5 deletions(-) diff --git a/lectures/flipped/L13.md b/lectures/flipped/L13.md index ac4a65d5..4b0fc511 100644 --- a/lectures/flipped/L13.md +++ b/lectures/flipped/L13.md @@ -25,11 +25,6 @@ Compare the runtimes Talk about the **fast inverse square root** in Quoke III (very interesting!) and **loop perforation** if time permits. -# After-action report, huanyi, 08Feb23 - -Went through all the n-body sample codes. However, it turns out the bin version -does not speed up much. Need to take a look sometime later. - # After-action report, plam, 10Feb23 Gave the example of Lab 2 and not requiring exact answers. @@ -46,3 +41,11 @@ Loop perforation: talked about an example of summing a bunch of numbers. I gave a very short introduction of the N-body problem. I'll cover the code in the next class. + +# After-action report, huanyi, 05Feb24 + +The performance needs further investigation. The bin version is slower than the +parallel one. + +I talked about the fast inverse square root, although briefly. Better to refer +to the referenced post to explain it. diff --git a/lectures/flipped/L14.md b/lectures/flipped/L14.md index ad4548a3..ca384a9a 100644 --- a/lectures/flipped/L14.md +++ b/lectures/flipped/L14.md @@ -153,3 +153,11 @@ inject "fence" to ensure other students see the same order as themselves see it. I did a bit of reorg in the above notes. It's still not very flipped. But I kind of did the whole thing, and did not talk about cache coherence. Did not check the runtime between SeqCst and Relaxed. + +# After-action report, huanyi, 05Feb24 + +I went through everything. It appears changing `SeqCst` to `Relaxed` had no +effect on RustExplorer, but it worked on my laptop. I also talked about memory +consistency model vs. cache coherence. I think the definition from the paper is +reasonable. There was an interesting question, why compiler reording is included +in the memory consistency topic?