This repository was archived by the owner on Mar 21, 2024. It is now read-only.

Draft: Revert/barrier parity#192

Closed

wmaxey wants to merge 44 commits intoNVIDIA:mainfrom

wmaxey:revert/barrier-parity

Commits on Aug 4, 2021

Added parity waiting

ogiroux
authored and
wmaxey
committed
Added try_wait options

ogiroux
authored and
wmaxey
committed
Add a small parity wait test
wmaxey
committed
Extend the test to measure both phases, make barrier_(try_)wait_parity member functions
wmaxey
committed
Re-add the __try_wait API as pipeline uses internal symbols of barrier for some reason
wmaxey
committed
Remove all uses of non-compliant __CUDA_ARCH__/preprocessor macros from libcudacxx
wmaxey
committed
Fix an issue in NVRTC tests
wmaxey
committed
Move files and implementation for atomic refactor
wmaxey
committed
WIP: single interface wrap/unwrap
wmaxey
committed
Finish atomic refactor, bones of atomic_ref are in place
wmaxey
committed
Refactor a few #ifdef __CUDA__ things and fix statics/shared memory usage in tests
wmaxey
committed
Implement CUDA atomic ref, implement MSVC atomics layer, fix several test utilities
wmaxey
committed
Fix atomic_mscv header, fix macro processing in __threading_support on MSVC, fix a test that used preprocessor directives within NV_TARGET
wmaxey
committed
Fix some local errors and warnings, put back the SHARED macro, fix pipeline tests
wmaxey
committed
Add symlink to nv/target within libcxx
wmaxey
committed
Fix CV handling of atomics on the libcxx/std layer
wmaxey
committed
Disable C11 atomics in __config
wmaxey
committed
fixups for Clang specific issues in atomic, make sure <nv/target> is within a C++ context
wmaxey
committed
Fix CUDA and MSVC atomic layers
wmaxey
committed
uncomment a couple tests from pipeline
wmaxey
committed
Revert tests, will <nv/target>-ify later
wmaxey
committed
Rebuild atomic_cuda_generated
wmaxey
committed
Dedup MSVC by splitting the atomic base class into a seperate header and creating back-end intrinsics in the MSVC header
wmaxey
committed
Missed grabbing important parts of the nvcxx-compatibility branch when cherry-picking
wmaxey
committed
Pickup more nv/target specializations from nvcxx_compatibility branch
wmaxey
committed
Uglify the atomic detail:: and host:: namespaces
wmaxey
committed
Rename __skip_amt to __atomic_ptr_inc
wmaxey
committed
Refactor and dedup some code in the __cxx_atomic cuda layer, fix runtime lock free check
wmaxey
committed
Fix set-but-not-used warnings for atomic intrinsics in atomic_base.h
wmaxey
committed
Fix static_assert in bad_atomic_alignment test.
wmaxey
committed
Suppress pointless comparison warnings where tests are impacted by the introduction of <nv/target>
wmaxey
committed
Add a missing license header to the atomic_c11.h file
wmaxey
committed
Fix pointless comparison warnings on two other pipeline tests
wmaxey
committed
Wrap/Unwrap store, exchange, and load to make sure the 'written to' parts of atomics are aligned
wmaxey
committed
Change method of ensuring that atomic types match
wmaxey
committed
Fix spurious warnings in atomic_base.h
wmaxey
committed
Rename __to_gcc_order to __cxx_atomic_order_to_int
wmaxey
committed
Reset barrier/latch/semaphore to head, as those will be nv/target'd later
wmaxey
committed
Fix a few includes occuring within internal namespaces
wmaxey
committed
Create changelog for release 1.6.0
wmaxey
committed
Bump libcudacxx API version to 1.6.0
wmaxey
committed
Add clearer notification about ABI break. Fix capitalization.
wmaxey
committed
Revert "Re-add the __try_wait API as pipeline uses internal symbols of barrier for some reason"
wmaxey
committed
Update Release 1.6.0?
wmaxey
committed