Create new accessors to get arguments for call/invoke instructions. It breaks

[oota-llvm.git] / lib / Target / README.txt

diff --git a/lib/Target/README.txt b/lib/Target/README.txt
index 052a5756da9faafdd4b97653c36fadb00cc76048..7fa73edaefee2509128c39beb0e591949c7f4c87 100644 (file)
--- a/lib/Target/README.txt
+++ b/lib/Target/README.txt
@@ -263,19 +263,6 @@ if anyone cared enough about sincos.
 
 //===---------------------------------------------------------------------===//
 
-Turn this into a single byte store with no load (the other 3 bytes are
-unmodified):
-
-define void @test(i32* %P) {
-       %tmp = load i32* %P
-        %tmp14 = or i32 %tmp, 3305111552
-        %tmp15 = and i32 %tmp14, 3321888767
-        store i32 %tmp15, i32* %P
-        ret void
-}
-
-//===---------------------------------------------------------------------===//
-
 quantum_sigma_x in 462.libquantum contains the following loop:
 
       for(i=0; i<reg->size; i++)
@@ -1843,3 +1830,36 @@ entry:
 
 //===---------------------------------------------------------------------===//
 
+We should use DSE + llvm.lifetime.end to delete dead vtable pointer updates.
+See GCC PR34949
+
+Another interesting case is that something related could be used for variables
+that go const after their ctor has finished.  In these cases, globalopt (which
+can statically run the constructor) could mark the global const (so it gets put
+in the readonly section).  A testcase would be:
+
+#include <complex>
+using namespace std;
+const complex<char> should_be_in_rodata (42,-42);
+complex<char> should_be_in_data (42,-42);
+complex<char> should_be_in_bss;
+
+Where we currently evaluate the ctors but the globals don't become const because
+the optimizer doesn't know they "become const" after the ctor is done.  See
+GCC PR4131 for more examples.
+
+//===---------------------------------------------------------------------===//
+
+In this code:
+
+long foo(long x) {
+  return x > 1 ? x : 1;
+}
+
+LLVM emits a comparison with 1 instead of 0. 0 would be equivalent
+and cheaper on most targets.
+
+LLVM prefers comparisons with zero over non-zero in general, but in this
+case it choses instead to keep the max operation obvious.
+
+//===---------------------------------------------------------------------===//