llc -enable-value-prop, propagation of value info (sign/zero ext info) from
one MBB to another
debug info for optimized code
-
+ interpreter + libffi
-->
<!-- for announcement email:
<!--
Random stuff:
+Pure project: http://code.google.com/p/pure-lang/
+
+
xcore backend!
fortran on darwin!
-.ll parser rewrite.
+.ll parser rewrite, caret diags, better errors, less fragile (less likely to
+ crash on strange things). No longer depends on flex/bison.
GCC inliner off, llvm handles always-inline.
cmake mature?
-x86 backend FS/GS segment address spaces?
+x86 backend GS segment -> addr space 256 (r62980)
nocapture
memdep (used by GVN and memcpyopt) is faster / more aggressive.
how to write a backend doc docs/WritingAnLLVMBackend.html
input vectors.
loop optimizer improves floating point induction variables
llvm/Analysis/DebugInfo.h classes, llvm-gcc and clang and codegen use them.
+ DebugInfoBuilder gone.
asmprinters seperate from targets for jits
PBQP register allocator now supports register coalescing.
JIT supports exceptions on linux/x86-64.
X86 backend now supports -disable-mmx.
noalias attribute on return value indicates that function returns new memory
(e.g. malloc).
-postalloc scheduler: anti dependence breaking?
+postalloc scheduler: anti dependence breaking, hazard recognizer?
llvmc2 renamed to llvmc
Jump threading more powerful: it is iterative, handles threading based on values
with fully redundant and partially redundant loads.
unit test framework based on Google Test.
vector shift support + X86 backend.
+x86 JIT now detects core i7 and atom, autoconfiguring itself appropriately.
+SROA is more aggressive about promoting unions.
+non-zero __builtin_return_address values on X86.
+x86-64 now uses red zone (unless -mno-red-zone option is specified).
+private linkage.
+
+llvm-gcc defaults to -fno-math-errno on all x86 targets.
+initial support for debug line numbers when optimization enabled, not useful in
+ 2.5 but will be for 2.6.
-->
</ul>