Mike Spertus [Mon, 6 Jun 2016 03:37:18 +0000 (03:37 +0000)]
Slightly improve Visual Studio visualization of clang::Expr
Now it gives the StmtClass of the Expr as well as the type. It's still
a long way from full visualization of expressions, but I have found
that having the class really helps when debugging, so definitely
worth submitting.
Craig Topper [Sat, 4 Jun 2016 05:43:41 +0000 (05:43 +0000)]
[X86] Use unsigned types for vector arithmetic in intrinsics to avoid undefined behavior for signed integer overflow.
This is really only needed for addition, subtraction, and multiplication, but I did the bitwise ops too for overall consistency. Clang currently doesn't set NSW for signed vector operations so the undefined behavior shouldn't happen today.
We would attempt to evaluate the sizeof a dependent type to check for an
integral overflow. However, because the dependent type is not yet resolved, we
cannot determine if the expression would overflow. Report a failure to perform
a symbolic evaluation of a constant involving the dependent type.
The assertion added earlier was overly strict. We need to strip the pointer
casts (as when constructing the GV). Correct the types (Function or Variable).
Pavel Labath [Thu, 2 Jun 2016 16:35:24 +0000 (16:35 +0000)]
[cmake] Fix builds with LLVM_ENABLE_PIC=0
Summary:
When this flag is specified, the target llvm-lto is not built, but is still
used as a dependency of the test targets. cmake 2.8 silently ignored this
situation, but with cmake_minimum_required(3.4) it becomes an error. Fix this
by avoiding the inclusion the target as a dependency.
Simon Pilgrim [Wed, 1 Jun 2016 21:46:51 +0000 (21:46 +0000)]
[X86][SSE] Replace (V)CVTTPS2DQ and VCVTTPD2DQ truncating (round to zero) f32/f64 to i32 with generic IR (clang)
The 'cvtt' truncation (round to zero) conversions can be safely represented as generic __builtin_convertvector (fptosi) calls instead of x86 intrinsics. We already do this (implicitly) for the scalar equivalents.
Note: I looked at updating _mm_cvttpd_epi32 as well but this still requires a lot more backend work to correctly lower (both for debug and optimized builds).
Chih-Hung Hsieh [Wed, 1 Jun 2016 20:48:46 +0000 (20:48 +0000)]
[driver][arm] add armv7 and thumb include and lib paths
Add a new test android-ndk-standalone.cpp
with new Android NDK release tree structure.
Detect armv7 sub architecture and thumb mode,
to add system include and link search paths.
Manman Ren [Wed, 1 Jun 2016 17:14:19 +0000 (17:14 +0000)]
ObjC lifetime: pull sugar off when the qualifiers conflict.
It's possible to have multiple local ObjCLifetime qualifiers. When there is
a conflict, we can't stop after we reach a type that is directly qualified.
We need to keep pulling sugar off and removing the ObjCLifetime qualifers.
Adding front-end support to several intrinsics (bit scanning, conversion and state reading intrinsics)
Adding LLVM front-end support to two intrinsics dealing with bit scan: _bit_scan_forward and _bit_scan_reverse.
Their functionality is as described in Intel intrinsics guide:
https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=_bit_scan_forward&expand=371,370
https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=_bit_scan_reverse&expand=371,370
Furthermore, adding clang front-end support to these conversion intrinsics: _mm256_cvtsd_f64, _mm256_cvtsi256_si32 and _mm256_cvtss_f32.
Finally, adding tests to all of the above, as well as to the state reading intrinsics _rdpmc and _rdtsc.
Their functionality is also specified in the Intel intrinsics guide.
The `isa' member was previously not given the correct DLL Storage. Ensure that
we give the `isa' constant `__CFConstantStringClassReference' the correct DLL
storage. Default to dllimport unless an explicit specification gives it a
dllexport storage.
[Parser] Fix look ahead after EOF while parsing objc message and lambdas
If a closing ')' isn't found for a macro instantiation inside a '[',
the next token is EOF, this leads to crashes if we try to look ahead of
that. This could be triggered whenever trying to parse lambdas or objs
message expressions.
Manman Ren [Tue, 31 May 2016 18:19:32 +0000 (18:19 +0000)]
PCH + module: make sure we write out macros associated with builtin identifiers.
When we import a module that defines a builtin identifier from prefix header and
precompile the prefix header, the macro information related to the identifier
is lost.
If we don't precompile the prefix header, the source file can still see the
macro information. The reason is that we write out the identifier in the pch
but not the macro information since the macro is not defined locally.
This is related to r251565. In that commit, if we read a builtin identifier from
a module that wasn't "interesting" to that module, we will still write it out to
a PCH that imports that module.
The fix is to write exported module macros for PCH as well.
Reid Kleckner [Tue, 31 May 2016 17:42:56 +0000 (17:42 +0000)]
Work around MinGW's macro definition of 'interface' to 'struct'
Previous attempts to rename the IBOutletCollection argument to something
other than "Interface" were undone (r127127 and r139620). Instead of
renaming it, work around this in tablegen, so the public facing getter
can have the usual name of 'getInterface'.
Eric Liu [Tue, 31 May 2016 13:34:20 +0000 (13:34 +0000)]
[clang-format] insert new #includes into correct blocks when cleaning up Replacement with cleanupAroundReplacements().
Summary:
When a replacement's offset is set to UINT_MAX or -1U, it is treated as
a header insertion replacement by cleanupAroundReplacements(). The new #include
directive is then inserted into the correct block.
Alexey Bataev [Tue, 31 May 2016 11:17:08 +0000 (11:17 +0000)]
[OPENMP] Update in ReleaseNotes for OpenMP support.
Added notes about full support of all non-offloading features of OpenMP
4.5 + info about option -fopenmp-version=[31|40|45] that allows to
control supported OpenMP version.
Simon Pilgrim [Mon, 30 May 2016 17:55:25 +0000 (17:55 +0000)]
[X86][SSE] _mm_store1_ps/_mm_store1_pd should require an aligned pointer
According to the gcc headers, intel intrinsics docs and msdn codegen the _mm_store1_pd (and its _mm_store_pd1 equivalent) should use an aligned pointer - the clang headers are the only implementation I can find that assume non-aligned stores (by storing with _mm_storeu_pd).
Additionally, according to the intel intrinsics docs and msdn codegen the _mm_store1_ps (_mm_store_ps1) requires a similarly aligned pointer.
This patch raises the alignment requirements to match the other implementations by calling _mm_store_ps/_mm_store_pd instead.
I've also added the missing _mm_store_pd1 intrinsic (which maps to _mm_store1_pd like _mm_store_ps1 does to _mm_store1_ps).
As a followup I'll update the llvm fast-isel tests to match this codegen.
Craig Topper [Mon, 30 May 2016 17:10:30 +0000 (17:10 +0000)]
[X86] Replace unaligned store builtins in SSE/AVX intrinsic files with code that will compile to a native unaligned store. Remove the builtins since they are no longer used.
Intrinsics will be removed from llvm in a future commit.
Adjust the constant CFString emission to emit into more appropriate sections on
ELF and COFF targets. It would previously try to use MachO section names
irrespective of the file format.
Alexey Bataev [Mon, 30 May 2016 13:05:14 +0000 (13:05 +0000)]
[OPENMP 4.5] Additional codegen for statically scheduled loops with
'simd' modifier.
Runtime library defines new schedule constant kmp_sch_static_balanced_chunked = 45 for static loop-based directives static with chunk adjustment (e.g., simd). Added codegen for this kind of schedule.
Alexey Bataev [Mon, 30 May 2016 09:06:50 +0000 (09:06 +0000)]
[OPENMP 4.5] Fixed codegen for 'priority' and destructors in task-based
directives.
'kmp_task_t' record type added a new field for 'priority' clause and
changed the representation of pointer to destructors for privates used
within loop-based directives.
Old representation:
typedef struct kmp_task { /* GEH: Shouldn't this be
aligned somehow? */
void *shareds; /**< pointer to block of
pointers to shared vars */
kmp_routine_entry_t routine; /**< pointer to routine
to call for executing task */
kmp_int32 part_id; /**< part id for the
task */
kmp_routine_entry_t destructors; /* pointer to function to
invoke deconstructors of firstprivate C++ objects */
/* private vars */
} kmp_task_t;
New representation:
typedef struct kmp_task { /* GEH: Shouldn't this be
aligned somehow? */
void *shareds; /**< pointer to block of
pointers to shared vars */
kmp_routine_entry_t routine; /**< pointer to routine
to call for executing task */
kmp_int32 part_id; /**< part id for the
task */
kmp_cmplrdata_t data1; /* Two known
optional additions: destructors and priority */
kmp_cmplrdata_t data2; /* Process
destructors first, priority second */
/* future data */
/* private vars */
} kmp_task_t;
Also excessive initialization of 'destructors' fields to 'null' was
removed from codegen if it is known that no destructors shal be used.
Currently a special bit is used in 'kmp_tasking_flags_t' bitfields
('destructors_thunk' bitfield).
Yaxun Liu [Mon, 30 May 2016 02:22:28 +0000 (02:22 +0000)]
[OpenCL] Add the default header file opencl-c.h for OpenCL C language
OpenCL has large number of "builtin" functions ("builtin" in the sense of OpenCL spec) which are defined in header files. To compile OpenCL kernels using these builtin functions, a header file is needed.
This header file is based on the Khronos implementation (https://github.com/KhronosGroup/SPIR/blob/spirv-1.0/lib/Headers/opencl.h) with heavy refactoring.
Re-commit after fixing failures on ppc64/systemz etc.
projects / clang / log