From 69462b4c252064301c425e8e001d56eadae1d716 Mon Sep 17 00:00:00 2001
From: travisdowns-bot <60370964+travisdowns-bot@users.noreply.github.com>
Date: Wed, 3 Sep 2025 06:33:23 -0700
Subject: [PATCH] Comment from Rishi on kreg-facts

---
 _data/comments/kreg-facts/entry1756906402454.yml | 13 +++++++++++++
 1 file changed, 13 insertions(+)
 create mode 100644 _data/comments/kreg-facts/entry1756906402454.yml

diff --git a/_data/comments/kreg-facts/entry1756906402454.yml b/_data/comments/kreg-facts/entry1756906402454.yml
new file mode 100644
index 00000000..1fd042ec
--- /dev/null
+++ b/_data/comments/kreg-facts/entry1756906402454.yml
@@ -0,0 +1,13 @@
+_id: 8fa9b300-88ca-11f0-9585-874f1f8459af
+_parent: 'https://travisdowns.github.io/blog/2019/12/05/kreg-facts.html'
+replying_to_uid: ''
+message: >-
+  The reason CPU is hitting brakes at 95% fill of PRF could be because it
+  reserves few entries for a 2nd thread if it support hyperthreading/SMT. Some
+  out of order structures reserve few entries for forward progress as well, in
+  the event one op comes in non-speculatively. But PRF allocation might not be
+  out-of-order, so that might not apply here.
+name: Rishi
+email: 0a2adbcf9e25b9a649c8c302cbfa80bf
+hp: ''
+date: 1756906402