/MERGE option is a bit complicated for many reasons. Firstly, it takes both
positive and negative arguments. That means we have to have one of three
distinctive values (set, clear or unchange) for each permission bit. In this
patch we represent the three values using two bitmasks.

Secondly, the permissions specified by the parameter is bitwise or-ed with the
default permissions of a section. There is an exception for that rule; if one
of READ, WRITE or EXECUTE bit is specified, unspecified bits need to be
cleared. (So if you specify only WRITE for example, the resulting section will
not have WRITE nor EXECUTE bits.)

Lastly, multiple /merge options are allowed.

llvm-svn: 195882

NOTE: This is a candidate for the 3.4 branch.
llvm-svn: 195881

SGPRs are spilled into VGPRs using the {READ,WRITE}LANE_B32 instructions.

v2:
  - Fix encoding of Lane Mask
  - Use correct register flags, so we don't overwrite the low dword
    when restoring multi-dword registers.

v3:
  - Register spilling seems to hang the GPU, so replace all shaders
    that need spilling with a dummy shader.

v4:
  - Fix *LANE definitions
  - Change destination reg class for 32-bit SMRD instructions

v5:
  - Remove small optimization that was crashing Serious Sam 3.

https://bugs.freedesktop.org/show_bug.cgi?id=68224
https://bugs.freedesktop.org/show_bug.cgi?id=71285

NOTE: This is a candidate for the 3.4 branch.
llvm-svn: 195880

Writing to the M0 register from an SMRD instruction hangs the GPU, so
we need to use the SGPR_32 register class, which does not include M0.

NOTE: This is a candidate for the 3.4 branch.
llvm-svn: 195879

NOTE: This is a candidate for the 3.4 branch.
llvm-svn: 195878

llvm-svn: 195877

llvm-svn: 195876

Renaming Struct to GenericRecord for clarity; could not use Record as the enumerant because that's already taken by the Record class.

llvm-svn: 195875

cross-reference debug output with encoded stack-maps, and to create stackmap
test-cases. 

llvm-svn: 195874

This was removed in the great unused local cleanup of r195814.

Adding cast-to-void to suppress unused variable warnings.

Should fix certain builds like ARCMT.

llvm-svn: 195872

Converts the lockable attributes to use the simple attribute handler because the semantic checking was already automated.

llvm-svn: 195866

In particular, check the name of the symbol we are putting in the constant pool.

llvm-svn: 195865

Automated checking for C++ when determining what argument to send to the diagnostic for attribute subjects. In turn, this allows the Subjects to be enabled for some more attributes and improves diagnostics. Updated a test case based on the improved diagnostic.

llvm-svn: 195864

This patch is to improve the readability of the tests before making a change
to /merge option.

llvm-svn: 195863

Adding the Subjects entry back for the noreturn attributes. This caused some test cases to be updated because the original diagnostic was about applying to methods as well as functions, but the semantic checking disallowed methods.

llvm-svn: 195862

MO_ExternalSymbol and MO_JumpTableIndex don't show up in inline asm.

llvm-svn: 195861

llvm-svn: 195859

llvm-svn: 195858

llvm-svn: 195857

/MERGE:foo=bar command line option merges section foo to section bar. If
section bar does not exist, foo is just renamed as bar.

llvm-svn: 195856

llvm-svn: 195855

llvm-svn: 195854

llvm-svn: 195853

Atom ordinals are the indeces in a file. Currently the PECOFF reader assigns
ordinals for each section, so it's (incorrectly) assigning duplicate ordinals.

llvm-svn: 195852

Enabling the subject list for the warn_unused attribute, and adding a test case. Previously, would issue a "warning ignored" diagnostic instead of the more specific "only applies to."

llvm-svn: 195851

llvm-svn: 195850

We currently error in clang with:
"error: thread-local storage is unsupported for the current target", but we
can start to get the llvm level ready.

When compiling

template<typename T>
struct foo {
  static __declspec(thread) int bar;
};
template<typename T>
__declspec(therad) int foo<T>::bar;
template struct foo<int>;

msvc produces

SECTION HEADER #3
   .tls$ name
       0 physical address
       0 virtual address
       4 size of raw data
     12F file pointer to raw data (0000012F to 00000132)
       0 file pointer to relocation table
       0 file pointer to line numbers
       0 number of relocations
       0 number of line numbers
C0301040 flags
         Initialized Data
         COMDAT; sym= "public: static int foo<int>::bar" (?bar@?$foo@H@@2HA)
         4 byte align
         Read Write

gcc produces a ".data$__emutls_v.<symbol>" for the testcase with
__declspec(thread) replaced with thread_local.

llvm-svn: 195849

Refactored semantic handlers that would unilaterally apply the attribute to a Decl so they all go through a common interface. No functional change intended.

llvm-svn: 195848

MO_ConstantPoolIndex is handled in printLeaMemReference.
MO_JumpTableIndex and MO_ExternalSymbol don't show up in inline asm.

llvm-svn: 195847

llvm-svn: 195846

llvm-svn: 195845

llvm-svn: 195844

llvm-svn: 195843

Removing the description part of a subset subject. It is unused, and wouldn't be future-proof regarding translation issues. It is also covered by SubjectList's CustomDiag member.

llvm-svn: 195842

Laying the basic groundwork for table generating the diagnostics for attribute subjects. This makes some modifications to the way subjects are listed in Attr.td, and updates the attr emitter to handle the new constructs.

I have disabled some attribute subject lines on purpose in Attr.td;
this part is a WIP with the goal being to restore those subjects
incrementally. By commenting them out, it leaves the original behavior
the same as before for those attributes and so those are not
functionality changes.

llvm-svn: 195841

llvm-svn: 195840

llvm-svn: 195839

llvm-svn: 195838

More steps are needed to actually make it usable:
* sanitizer runtimes should be compiled with -DSANITIZER_LIBBACKTRACE.
* libbacktrace headers should be installed.
* user has to manually link in libbacktrace.a into the executable.

We can easily solve the first two problems in the build system, but
detecting/linking libbacktrace to all the tests we have and end-user programs
is more challenging (and will unlikely work w/o Driver support).

Based on the patch by Jakub Jelinek!

llvm-svn: 195837

print thread creation stack
and stacks where ignores were enabled.

llvm-svn: 195836