X-Git-Url: http://repo.macrolet.net/gitweb/?a=blobdiff_plain;f=OPTIMIZATIONS;h=3c3edea60ba46a6f485fa4828c717722c1be7831;hb=22c1de0a40df83bb5628974010a879cb2c17ff53;hp=7f45bee38b0cde660fe092d60e25efa307a9fb07;hpb=852a4504e8a9040adf52d374dafa4813bc661173;p=sbcl.git

diff --git a/OPTIMIZATIONS b/OPTIMIZATIONS
index 7f45bee..3c3edea 100644
--- a/OPTIMIZATIONS
+++ b/OPTIMIZATIONS
@@ -1,3 +1,4 @@
+#1
 (defun mysl (s)
     (declare (simple-string s))
     (declare (optimize (speed 3) (safety 0) (debug 0)))
@@ -10,14 +11,12 @@
 
 * On X86 I is represented as a tagged integer.
 
-* EQL uses "CMP reg,reg" instead of "CMP reg,im". This causes
-  allocation of an extra register and an extra move.
-
 * Unnecessary move:
   3: SLOT S!11[EDX] {SB-C::VECTOR-LENGTH 1 7} => t23[EAX]
   4: MOVE t23[EAX] => t24[EBX]
 
 --------------------------------------------------------------------------------
+#2
 (defun quux (v)
   (declare (optimize (speed 3) (safety 0) (space 2) (debug 0)))
   (declare (type (simple-array double-float 1) v))
@@ -34,6 +33,7 @@
   and emits costy MOVE ... => FR1.
 
 --------------------------------------------------------------------------------
+#3
 (defun bar (n)
   (declare (optimize (speed 3) (safety 0) (space 2))
            (type fixnum n))
@@ -42,10 +42,8 @@
     (length v)))
 
 * IR1 does not optimize away (MAKE-LIST N).
-
-* IR1 thinks that the type of V in (LENGTH V) is (OR LIST SIMPLE-VECTOR), not
-  SIMPLE-VECTOR.
 --------------------------------------------------------------------------------
+#4
 (defun bar (v1 v2)
   (declare (optimize (speed 3) (safety 0) (space 2))
            (type (simple-array base-char 1) v1 v2))
@@ -64,10 +62,7 @@ VOP DATA-VECTOR-SET/SIMPLE-STRING V2!14[EDI] t32[EAX] t30[S2]>t33[CL]
 
 * And why two moves?
 --------------------------------------------------------------------------------
-(loop repeat 1.5)
-
-uses generic arithmetic
---------------------------------------------------------------------------------
+#6
 09:49:05 <jtra> I have found a case in those where suboptimal code is
   generate with nested loops, it might be moderately easy to fix that
 09:49:28 <jtra> see
@@ -77,7 +72,6 @@ uses generic arithmetic
   memory location for iteration variable
 
 ;;; -*- mode: lisp -*-
-;;; $Id$
 ;;; http://www.bagley.org/~doug/shootout/
 ;;; from Friedrich Dominicus
 
@@ -96,3 +90,71 @@ uses generic arithmetic
                 (incf x)))))))
    (format t "~A~%" x)))
 --------------------------------------------------------------------------------
+#7
+(defun foo (x)
+  (declare (optimize speed (debug 0)))
+  (if (< x 0) x (foo (1- x))))
+
+SBCL generates a full call of FOO (but CMUCL does not).
+--------------------------------------------------------------------------------
+#8
+(defun foo (d)
+  (declare (optimize (speed 3) (safety 0) (debug 0)))
+  (declare (type (double-float 0d0 1d0) d))
+  (loop for i fixnum from 1 to 5
+     for x1 double-float = (sin d) ;;; !!!
+     do (loop for j fixnum from 1 to 4
+             sum x1 double-float)))
+
+Without the marked declaration Python will use boxed representation for X1.
+
+This is equivalent to
+
+(let ((x nil))
+  (setq x 0d0)
+  ;; use of X as DOUBLE-FLOAT
+)
+
+The initial binding is effectless, and without it X is of type
+DOUBLE-FLOAT. Unhopefully, IR1 does not optimize away effectless
+SETs/bindings, and IR2 does not perform type inference.
+--------------------------------------------------------------------------------
+#9 "Multi-path constant folding"
+(defun foo (x)
+  (if (= (cond ((irgh x) 0)
+               ((buh x) 1)
+               (t 2))
+         0)
+      :yes
+      :no))
+
+This code could be optimized to
+
+(defun foo (x)
+  (cond ((irgh x) :yes)
+        ((buh x) :no)
+        (t :no)))
+--------------------------------------------------------------------------------
+#11
+(inverted variant of #9)
+
+(lambda (x)
+  (let ((y (sap-alien x c-string)))
+    (list (alien-sap y)
+          (alien-sap y))))
+
+It could be optimized to
+
+(lambda (x) (list x x))
+
+(if Y were used only once, the current compiler would optimize it)
+--------------------------------------------------------------------------------
+#12
+(typep (truly-the (simple-array * (*)) x) 'simple-vector)
+
+tests lowtag.
+--------------------------------------------------------------------------------
+#13
+FAST-+/FIXNUM and similar should accept unboxed arguments in interests
+of representation selection. Problem: inter-TN dependencies.
+--------------------------------------------------------------------------------