Fix make-array transforms.

[sbcl.git] / OPTIMIZATIONS

diff --git a/OPTIMIZATIONS b/OPTIMIZATIONS
index de7e3c3..42119ac 100644 (file)
--- a/OPTIMIZATIONS
+++ b/OPTIMIZATIONS
@@ -157,11 +157,6 @@ sbcl-0.8.12.30, this affects at least DUMP-OBJECT through
 COMPOUND-OBJECT-P, and (LABELS MAYBE-EMIT-MAKE-LOAD-FORMS GROVEL)
 through TYPEP UNBOXED-ARRAY, within the compiler itself.
 --------------------------------------------------------------------------------
-#18
-(lambda (x) (declare (null x)) (sxhash x)) goes through SYMBOL-HASH
-rather than either constant-folding or manipulating NIL-VALUE or
-NULL-TN directly.
---------------------------------------------------------------------------------
 #20
 (defun-with-dx foo (x)
   (flet ((make (x)
@@ -173,15 +168,13 @@ NULL-TN directly.
       (declare (dynamic-extent l))
       (mapc #'print l))))
 
-Result of MAKE is not stack allocated, which means that
-stack-allocation of structures is impossible.
+Result of MAKE is not stack allocated.
 --------------------------------------------------------------------------------
 #22
 IR2 does not perform unused code flushing.
 --------------------------------------------------------------------------------
 #24
-a. Iterations on &REST lists, returning them as VALUES could be
-   rewritten with &MORE vectors.
+a. Iterations on &REST lists could be rewritten with &MORE vectors.
 b. Implement local unknown-values mv-call (useful for fast type checking).
 --------------------------------------------------------------------------------
 #26
@@ -210,6 +203,10 @@ enough to eliminate something like ':initial-element 0'.  Such an optimization
 is valid if the vector is being allocated in the heap, but not if it is being
 allocated on the stack.  You could remove this optimization, but that makes
 the heap-allocated case somewhat slower...)
+
+To do this, extend ALLOCATE-VECTOR with ALLOW-JUNK argument, and when
+stack allocating don't zero if it is true -- and probably ALLOW-JUNK iff
+the vector is a specialized one (cannot have pointers.)
 --------------------------------------------------------------------------------
 #28
 a. Accessing raw slots in structure instances is more inefficient than
@@ -244,21 +241,6 @@ vm-support-routine COMBINATION-IMPLEMENTATION-STYLE and its use in
 src/compiler/ir1opt.lisp, IR1-OPTIMIZE-COMBINATION.  The above
 optimizations are left as an exercise for the reader.)
 --------------------------------------------------------------------------------
-#30
-(defun foo (x y)
-  (< x y))
-
-FOO's IR1 representation is roughly:
-
-(defun foo (x y)
-  (if (< x y)
-      T
-      NIL))
-
-However, if a full call is generated for < (and similarly for other
-predicate functions), then the IF is unnecessary, since the return value
-of (< x y) is already T or NIL.
---------------------------------------------------------------------------------
 #31
 The typecheck generated for a declaration like (integer 0 45) on x86 looks
 like:
@@ -402,3 +384,23 @@ can be transformed into
 which allows compiler-macro-expansion for FOO. (Only constant
 arguments can be moved inside the new lambda -- otherwise evaluation
 order is altered.)
+
+--------------------------------------------------------------------------------
+#41
+
+The unibyte external formats are written in a very generic way.  Three
+optimizations immediately applicable that could be automatically
+generated:
+
+(a) if the external format merely permutes the first 256 characters, a
+    constant-time lookup (rather than a binary search) could be
+    performed on output.  This applies at least to EBCDIC, which
+    currently has a hand-rolled mapper instead.
+
+(b) if there are no undefined characters corresponding to the 256
+    codes, then no error checking need be done on input.
+
+(c) if there is a way to use particular bits of the exceptional
+    characters, constant-time output (rather than binary search) can
+    still be achieved as used to be done by the latin-9 external
+    format before 1.0.31.