llvm-svn: 39979

llvm-svn: 39978

llvm-svn: 39977

Missed the case where alloca is used but the stack size (not including callee-saved portion) is zero. Thanks Dan.

llvm-svn: 39974

llvm-svn: 39971

llvm-svn: 39967

llvm-svn: 39962

having that implementation file #include <cstdlib>.

llvm-svn: 39952

beyond the end of an alloca to make FastDSE faster and more aggressive.

llvm-svn: 39945

dead stores on 400.perlbench.

llvm-svn: 39929

llvm-svn: 39926

llvm-gcc build to succeed. Without this change it fails in libstdc++
compilation. This causes no regressions in dejagnu tests. However, 
someone who knows this code better might want to review it.

llvm-svn: 39924

have an error, and refector out the code for binary operators into
ConstantFoldBinaryFP and use it for all binary floating-point operations
which may have an error. These functions still rely exclusively on errno
to detect errors though.

llvm-svn: 39923

llvm-svn: 39921

llvm-svn: 39896

llvm-svn: 39894

Always pass the constant as the second parameter to HowManyLessThans.

Remove obsolete "isSigned" parameter.

llvm-svn: 39893

First teach instcombine that sign bit checks only demand the 
sign bit, this allows simplify demanded bits to hack on 
expressions better.

Second, teach instcombine that ashr is useless if only the
sign bit is demanded.

llvm-svn: 39880

(X << 31) <s 0  --> (X&1) != 0

This happens dozens of times in the CFE.

llvm-svn: 39879

llvm-svn: 39857

llvm-svn: 39856

This patch fills the last necessary bits to enable exceptions
handling in LLVM. Currently only on x86-32/linux.

In fact, this patch adds necessary intrinsics (and their lowering) which
represent really weird target-specific gcc builtins used inside unwinder.

After corresponding llvm-gcc patch will land (easy) exceptions should be
more or less workable. However, exceptions handling support should not be 
thought as 'finished': I expect many small and not so small glitches
everywhere.

llvm-svn: 39855

llvm-svn: 39852

llvm-svn: 39851

llvm-svn: 39844

llvm-svn: 39843

Handle GEPs with all-zero indices in the same way we handle pointer-pointer bitcasts. Also, fix a potentia infinite loop.

This brings FastDSE to parity with old DSE on 175.vpr.

llvm-svn: 39839

llvm-svn: 39834

llvm-svn: 39827

feedback. This theoretically makes the common (scalar) case more efficient.

llvm-svn: 39823

This drastically improves the effect of FastDSE on kimwitu++.

llvm-svn: 39819

llvm-svn: 39817

Change name of -sched option and DEBUG_TYPE to
pre-RA-sched; adjust testcases.

llvm-svn: 39816

llvm-svn: 39813

llvm-svn: 39790

llvm-svn: 39789

Remove the end-block handling code.  It was unsafe, and making it safe would have resulted in falling back to the slow DSE case.  I need to think some more about the right way to handle this.

llvm-svn: 39788

following jump tables that it earlier inserted.  This
would be OK on other targets but is needed for correctness
only on ARM (constant islands needs to find jump tables).

llvm-svn: 39782

should be output as unsigned values.  Checked against
gcc.

llvm-svn: 39775

thanks Benoit Boissinot!

llvm-svn: 39774