Elena Demikhovsky 21706cbd24 AVX-512 Loop Vectorizer: Cost calculation for interleave load/store patterns. ...

X86 target does not provide any target specific cost calculation for interleave patterns.It uses the common target-independent calculation, which gives very high numbers. As a result, the scalar version is chosen in many cases. The situation on AVX-512 is even worse, since we have 3-src shuffles that significantly reduce the cost.

In this patch I calculate the cost on AVX-512. It will allow to compare interleave pattern with gather/scatter and choose a better solution (PR31426).

* Shiffle-broadcast cost will be changed in Simon's upcoming patch.

Differential Revision: https://reviews.llvm.org/D28118

llvm-svn: 290810

2017-01-02 10:37:52 +00:00

clang

DR1391: Check for implicit conversion sequences for non-dependent function

2017-01-02 02:42:17 +00:00

clang-tools-extra

[clang-tidy] Add delete null pointer check.

2016-12-31 12:45:59 +00:00

compiler-rt

Decrease kLargeMalloc block size in ASAN unit tests.

2016-12-28 21:57:51 +00:00

debuginfo-tests

New round of fixes for "Always compile debuginfo-tests for the host triple"

2014-10-18 23:47:59 +00:00

libclc

Fix build since r286752.

2016-11-14 16:06:33 +00:00

libcxx

chrono: address post commit comments from Howard

2017-01-01 22:04:38 +00:00

libcxxabi

clean up -Wmisleading-indentation warning

2016-12-31 18:09:51 +00:00

libunwind

EHABI: mark some functions as exported

2016-11-17 23:53:35 +00:00

lld

COFF: replace a magic number and assert more

2016-12-30 19:02:04 +00:00

lldb

Fix the variable view in the "gui" curses mode so that variables whose children change will update correctly. Previously the variable view would update the children once and not change. If you were stepping through code where the dynamic type of a variable would change the value and its children, or a synthetic type (like say for a std::vector<int>), the variable view wouldn't update. Now it caches the children and uses the process stop ID to tell when the children need to be updated.

2016-12-28 21:22:37 +00:00

llgo

[llgo] Remove support for LLVM attributes

2016-12-06 19:22:04 +00:00

llvm

AVX-512 Loop Vectorizer: Cost calculation for interleave load/store patterns.

2017-01-02 10:37:52 +00:00

openmp

Fix for the __kmpc_global_num_threads function to return the value of the __kmp_all_nth global var.

2016-12-21 21:20:20 +00:00

parallel-libs

[Axccel] Remove -Wno-missing-braces in build

2016-12-19 21:34:07 +00:00

polly

Update to isl-0.18-43-g0b4256f

2016-12-31 07:46:11 +00:00

Description

The LLVM Project is a collection of modular and reusable compiler and toolchain technologies.

5 GiB

Languages

LLVM 39.9%

C++ 32.5%

C 13.5%

Assembly 9.4%

MLIR 1.4%

Other 2.8%