From 93177be0b8b6548e27b54302c7e62a7e26750a20 Mon Sep 17 00:00:00 2001
From: Kaz Kylheku <kaz@kylheku.com>
Date: Fri, 7 Apr 2023 19:21:56 -0700
Subject: compiler: iterate on level 4-5 optimizations.

* stdlib/optimize.tl (basic-blocks num-blocks): New
method.

* stdlib/compiler.tl (compiler optimize): At optimization
level 6, instead of performing one extra pass of
jump threading, dead-code elimintation and peephole
optimizations, keep iterating on these until the number
of basic blocks stays the same.

* txr.1: Documented.
---
 stdlib/compiler.tl | 17 ++++++++---------
 stdlib/optimize.tl |  3 +++
 txr.1              |  4 +++-
 3 files changed, 14 insertions(+), 10 deletions(-)

diff --git a/stdlib/compiler.tl b/stdlib/compiler.tl
index 2a8f97cd..bf6496bb 100644
--- a/stdlib/compiler.tl
+++ b/stdlib/compiler.tl
@@ -1723,15 +1723,14 @@
           bb.(thread-jumps)
           bb.(elim-dead-code))
         (when (>= olev 5)
-          bb.(calc-liveness)
-          bb.(peephole)
-          (when (>= olev 6)
-             bb.(link-graph)
-             bb.(thread-jumps))
-          bb.(elim-dead-code)
-          (when (>= olev 6)
-            bb.(calc-liveness)
-            bb.(peephole)))
+          (let ((nblocks nil))
+            (while* (and (>= olev 6)
+                         (neql nblocks (set nblocks bb.(num-blocks))))
+              bb.(calc-liveness)
+              bb.(peephole)
+              bb.(link-graph)
+              bb.(thread-jumps)
+              bb.(elim-dead-code))))
         (cond
           ((>= olev 7)
            bb.(merge-jump-thunks)
diff --git a/stdlib/optimize.tl b/stdlib/optimize.tl
index ff256db7..8bf8b8c9 100644
--- a/stdlib/optimize.tl
+++ b/stdlib/optimize.tl
@@ -85,6 +85,9 @@
         (mapdo (do set [bb.hash @1.label] @1) bb.list))
       bb.(link-graph))
 
+    (:method num-blocks (bb)
+      (len bb.list))
+
     (:method get-insns (bb)
       [mappend .insns bb.list])
 
diff --git a/txr.1 b/txr.1
index 2a5819a5..e4044901 100644
--- a/txr.1
+++ b/txr.1
@@ -87774,7 +87774,9 @@ Data flow optimizations are applied, such as elimination of dead register
 moves, or useless propagations of values from one register to another.
 More peephole optimizations are applied.
 .IP 6
-Another pass of the levels 4 and 5 optimizations is performed.
+Additional iterations of the levels 4 and 5 optimizations are performed,
+if the previous iterations have coalesced some basic blocks of the program
+graph.
 .IP 7
 Certain more rarely applicable optimizations are applied which reduce code size
 by merging some identical code blocks, or improving some more rarely
-- 
cgit v1.2.3