Fall back to the selection dag isel to select tail calls.
authorChad Rosier <mcrosier@apple.com>
Tue, 11 Dec 2012 00:18:02 +0000 (00:18 +0000)
committerChad Rosier <mcrosier@apple.com>
Tue, 11 Dec 2012 00:18:02 +0000 (00:18 +0000)
commit425e951734c3a0615e22ec94ffa51cc16ce6e483
tree498b09ecb92c4701f5626d07a50ac1c0d5401830
parent0e3e9b79f6e471b46438251249f18121408e5189
Fall back to the selection dag isel to select tail calls.

This shouldn't affect codegen for -O0 compiles as tail call markers are not
emitted in unoptimized compiles.  Testing with the external/internal nightly
test suite reveals no change in compile time performance.  Testing with -O1,
-O2 and -O3 with fast-isel enabled did not cause any compile-time or
execution-time failures.  All tests were performed on my x86 machine.
I'll monitor our arm testers to ensure no regressions occur there.

In an upcoming clang patch I will be marking the objc_autoreleaseReturnValue
and objc_retainAutoreleaseReturnValue as tail calls unconditionally.  While
it's theoretically true that this is just an optimization, it's an
optimization that we very much want to happen even at -O0, or else ARC
applications become substantially harder to debug.

Part of rdar://12553082

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@169796 91177308-0d34-0410-b5e6-96231b3b80d8
include/llvm/CodeGen/FastISel.h
lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp
lib/Target/ARM/ARMFastISel.cpp
lib/Target/X86/X86FastISel.cpp
test/CodeGen/X86/tailcall-fastisel.ll