[llvm-commits] [llvm] r45141 - /llvm/trunk/lib/Target/README.txt

Chris Lattner sabre at nondot.org
Mon Dec 17 17:19:18 PST 2007


Author: lattner
Date: Mon Dec 17 19:19:18 2007
New Revision: 45141

URL: http://llvm.org/viewvc/llvm-project?rev=45141&view=rev
Log:
add a missed case.

Modified:
    llvm/trunk/lib/Target/README.txt

Modified: llvm/trunk/lib/Target/README.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/README.txt?rev=45141&r1=45140&r2=45141&view=diff

==============================================================================
--- llvm/trunk/lib/Target/README.txt (original)
+++ llvm/trunk/lib/Target/README.txt Mon Dec 17 19:19:18 2007
@@ -464,3 +464,21 @@
 }
 
 //===---------------------------------------------------------------------===//
+
+This code is often produced by the SMAX expansion in SCEV:
+
+define i32 @foo(i32 %a) {
+entry:
+        %tmp15 = sub i32 99, %a         ; <i32> [#uses=2]
+        %tmp16 = icmp slt i32 %tmp15, 0         ; <i1> [#uses=1]
+        %smax = select i1 %tmp16, i32 0, i32 %tmp15             ; <i32> [#uses=1]
+        %tmp12 = add i32 %smax, %a              ; <i32> [#uses=1]
+        %tmp13 = add i32 %tmp12, 1              ; <i32> [#uses=1]
+        ret i32 %tmp13
+}
+
+Note that the tmp12 add can be pushed through the select operands, turning
+it into a "select %tmp16, %a, 99".  We apparently already do this in dag 
+combine because it isn't present in X86 output.
+
+//===---------------------------------------------------------------------===//





More information about the llvm-commits mailing list