Teach DAGCombine to fold fold (sra (trunc (sr x, c1)), c2) -> (trunc (sra x, c1+c2) when c1 equals the amount of bits that are truncated off.

This happens all the time when a smul is promoted to a larger type.

On x86-64 we now compile "int test(int x) { return x/10; }" into
  movslq  %edi, %rax
  imulq $1717986919, %rax, %rax
  movq  %rax, %rcx
  shrq  $63, %rcx
  sarq  $34, %rax <- used to be "shrq $32, %rax; sarl $2, %eax"
  addl  %ecx, %eax

This fires 96 times in gcc.c on x86-64.

llvm-svn: 124559

invocation. 

This mimics what gcc does and fixes libtool check for libraries.

llvm-svn: 124558

llvm-svn: 124557

llvm-svn: 124556

llvm-svn: 124555

llvm-svn: 124554

When emitting RTTI for a non-class type, compute the visibility of the RTTI data based on the explicit visibility of the type.

llvm-svn: 124553

llvm-svn: 124552

llvm-svn: 124551

llvm-svn: 124550

I'm still not sure if having the typenames be visible with -hidden-weak-vtables, but I think it makes sense.

llvm-svn: 124549

Set the visibility for typeinfo names.

llvm-svn: 124548

This is the one where one of the branches of the select is another select on
the same condition.

llvm-svn: 124547

Change CodeGenModule::setTypeVisibility to take a TypeVisibilityKind enum instead of an "IsForRTTI" flag.

llvm-svn: 124546

The DEBUG() call at line 606 demands to see raw_ostream's definition. I have no idea why this seems to only break MSVC.

llvm-svn: 124545

llvm-svn: 124544

llvm-svn: 124543

llvm-svn: 124542

llvm-svn: 124541

llvm-svn: 124540

Make emitting a VTT a two-step process, much like emitting a VTable. You first get the address of the VTT, and then pass it to EmitVTTDefinition.

llvm-svn: 124539

llvm-svn: 124538

Add a new function, to be used by CGRTTI, CGVTables and CGVTT (which each has their own copy of this code).

llvm-svn: 124537

llvm-svn: 124536

llvm-svn: 124535

llvm-svn: 124534

llvm-svn: 124533

This happens e.g. for code like "X - X%10" where we lower the modulo operation
to a series of multiplies and shifts that are then subtracted from X, leading to
this missed optimization.

llvm-svn: 124532

llvm-svn: 124531

by name or by pid (with or without waiting for a process to launch) and
catch the response asynchronously.

llvm-svn: 124530

llvm-svn: 124529

When trying to get the most derived class, don't assume that we can ignore all casts. We can only ignore derived-to-base and no-op casts. Fixes selfhost.

llvm-svn: 124528

llvm-svn: 124527

llvm-svn: 124526

llvm-svn: 124525

When calling a virtual member function on a base class and the most derived class is marked 'final', we can devirtualize the call.

llvm-svn: 124524

When doing a derived-to-base class through a virtual class, we don't have to get the vbase offset from the vtable if the derived class is marked final.

llvm-svn: 124523

llvm-svn: 124522

llvm-svn: 124521

Added a completion action class to the Process events so that we can make things like Attach and later Launch start their job, and then return to the event loop while waiting for the work to be done.

llvm-svn: 124520