Merge pull request lammps#4413 from stanmoore1/kk_update_4.5.0

Update Kokkos library in LAMMPS to v4.5.1

cmake/Modules/Packages/KOKKOS.cmake

@@ -7,26 +7,13 @@ endif()
 
 ########################################################################
 # consistency checks and Kokkos options/settings required by LAMMPS
-if(Kokkos_ENABLE_CUDA)
-  option(Kokkos_ENABLE_IMPL_CUDA_MALLOC_ASYNC "CUDA asynchronous malloc support" OFF)
-  mark_as_advanced(Kokkos_ENABLE_IMPL_CUDA_MALLOC_ASYNC)
-  if(Kokkos_ENABLE_IMPL_CUDA_MALLOC_ASYNC)
-    message(STATUS "KOKKOS: CUDA malloc async support enabled")
-  else()
-    message(STATUS "KOKKOS: CUDA malloc async support disabled")
-  endif()
-endif()
 if(Kokkos_ENABLE_HIP)
   option(Kokkos_ENABLE_HIP_MULTIPLE_KERNEL_INSTANTIATIONS "Enable multiple kernel instantiations with HIP" ON)
   mark_as_advanced(Kokkos_ENABLE_HIP_MULTIPLE_KERNEL_INSTANTIATIONS)
   option(Kokkos_ENABLE_ROCTHRUST "Use RoCThrust library" ON)
   mark_as_advanced(Kokkos_ENABLE_ROCTHRUST)
-
-  if(Kokkos_ARCH_AMD_GFX942 OR Kokkos_ARCH_AMD_GFX940)
-    option(Kokkos_ENABLE_IMPL_HIP_UNIFIED_MEMORY "Enable unified memory with HIP" ON)
-    mark_as_advanced(Kokkos_ENABLE_IMPL_HIP_UNIFIED_MEMORY)
-  endif()
 endif()
+
 # Adding OpenMP compiler flags without the checks done for
 # BUILD_OMP can result in compile failures. Enforce consistency.
 if(Kokkos_ENABLE_OPENMP)
@@ -70,8 +57,8 @@ if(DOWNLOAD_KOKKOS)
   list(APPEND KOKKOS_LIB_BUILD_ARGS "-DCMAKE_CXX_EXTENSIONS=${CMAKE_CXX_EXTENSIONS}")
   list(APPEND KOKKOS_LIB_BUILD_ARGS "-DCMAKE_TOOLCHAIN_FILE=${CMAKE_TOOLCHAIN_FILE}")
   include(ExternalProject)
-  set(KOKKOS_URL "https://github.com/kokkos/kokkos/archive/4.4.01.tar.gz" CACHE STRING "URL for KOKKOS tarball")
-  set(KOKKOS_MD5 "de6ee80d00b6212b02bfb7f1e71a8392" CACHE STRING "MD5 checksum of KOKKOS tarball")
+  set(KOKKOS_URL "https://github.com/kokkos/kokkos/archive/4.5.01.tar.gz" CACHE STRING "URL for KOKKOS tarball")
+  set(KOKKOS_MD5 "4d832aa0284169d9e3fbae3165286bc6" CACHE STRING "MD5 checksum of KOKKOS tarball")
   mark_as_advanced(KOKKOS_URL)
   mark_as_advanced(KOKKOS_MD5)
   GetFallbackURL(KOKKOS_URL KOKKOS_FALLBACK)
@@ -96,7 +83,7 @@ if(DOWNLOAD_KOKKOS)
   add_dependencies(LAMMPS::KOKKOSCORE kokkos_build)
   add_dependencies(LAMMPS::KOKKOSCONTAINERS kokkos_build)
 elseif(EXTERNAL_KOKKOS)
-  find_package(Kokkos 4.4.01 REQUIRED CONFIG)
+  find_package(Kokkos 4.5.01 REQUIRED CONFIG)
   target_link_libraries(lammps PRIVATE Kokkos::kokkos)
 else()
   set(LAMMPS_LIB_KOKKOS_SRC_DIR ${LAMMPS_LIB_SOURCE_DIR}/kokkos)

doc/src/Build_extras.rst

@@ -547,16 +547,7 @@ They must be specified in uppercase.
       - Local machine
    *  - AMDAVX
       - HOST
-      - AMD 64-bit x86 CPU (AVX 1)
-   *  - ZEN
-      - HOST
-      - AMD Zen class CPU (AVX 2)
-   *  - ZEN2
-      - HOST
-      - AMD Zen2 class CPU (AVX 2)
-   *  - ZEN3
-      - HOST
-      - AMD Zen3 class CPU (AVX 2)
+      - AMD chip
    *  - ARMV80
       - HOST
       - ARMv8.0 Compatible CPU
@@ -572,105 +563,126 @@ They must be specified in uppercase.
    *  - A64FX
       - HOST
       - ARMv8.2 with SVE Support
+   *  - ARMV9_GRACE
+      - HOST
+      - ARMv9 NVIDIA Grace CPU
    *  - SNB
       - HOST
-      - Intel Sandy/Ivy Bridge CPU (AVX 1)
+      - Intel Sandy/Ivy Bridge CPUs
    *  - HSW
       - HOST
-      - Intel Haswell CPU (AVX 2)
+      - Intel Haswell CPUs
    *  - BDW
       - HOST
-      - Intel Broadwell Xeon E-class CPU (AVX 2 + transactional mem)
-   *  - SKL
-      - HOST
-      - Intel Skylake Client CPU
-   *  - SKX
-      - HOST
-      - Intel Skylake Xeon Server CPU (AVX512)
+      - Intel Broadwell Xeon E-class CPUs
    *  - ICL
       - HOST
-      - Intel Ice Lake Client CPU (AVX512)
+      - Intel Ice Lake Client CPUs (AVX512)
    *  - ICX
       - HOST
-      - Intel Ice Lake Xeon Server CPU (AVX512)
-   *  - SPR
+      - Intel Ice Lake Xeon Server CPUs (AVX512)
+   *  - SKL
       - HOST
-      - Intel Sapphire Rapids Xeon Server CPU (AVX512)
+      - Intel Skylake Client CPUs
+   *  - SKX
+      - HOST
+      - Intel Skylake Xeon Server CPUs (AVX512)
    *  - KNC
       - HOST
       - Intel Knights Corner Xeon Phi
    *  - KNL
       - HOST
       - Intel Knights Landing Xeon Phi
+   *  - SPR
+      - HOST
+      - Intel Sapphire Rapids Xeon Server CPUs (AVX512)
    *  - POWER8
       - HOST
-      - IBM POWER8 CPU
+      - IBM POWER8 CPUs
    *  - POWER9
       - HOST
-      - IBM POWER9 CPU
+      - IBM POWER9 CPUs
+   *  - ZEN
+      - HOST
+      - AMD Zen architecture
+   *  - ZEN2
+      - HOST
+      - AMD Zen2 architecture
+   *  - ZEN3
+      - HOST
+      - AMD Zen3 architecture
    *  - RISCV_SG2042
       - HOST
-      - SG2042 (RISC-V) CPU
+      - SG2042 (RISC-V) CPUs
+   *  - RISCV_RVA22V
+      - HOST
+      - RVA22V (RISC-V) CPUs
    *  - KEPLER30
       - GPU
-      - NVIDIA Kepler generation CC 3.0 GPU
+      - NVIDIA Kepler generation CC 3.0
    *  - KEPLER32
       - GPU
-      - NVIDIA Kepler generation CC 3.2 GPU
+      - NVIDIA Kepler generation CC 3.2
    *  - KEPLER35
       - GPU
-      - NVIDIA Kepler generation CC 3.5 GPU
+      - NVIDIA Kepler generation CC 3.5
    *  - KEPLER37
       - GPU
-      - NVIDIA Kepler generation CC 3.7 GPU
+      - NVIDIA Kepler generation CC 3.7
    *  - MAXWELL50
       - GPU
-      - NVIDIA Maxwell generation CC 5.0 GPU
+      - NVIDIA Maxwell generation CC 5.0
    *  - MAXWELL52
       - GPU
-      - NVIDIA Maxwell generation CC 5.2 GPU
+      - NVIDIA Maxwell generation CC 5.2
    *  - MAXWELL53
       - GPU
-      - NVIDIA Maxwell generation CC 5.3 GPU
+      - NVIDIA Maxwell generation CC 5.3
    *  - PASCAL60
       - GPU
-      - NVIDIA Pascal generation CC 6.0 GPU
+      - NVIDIA Pascal generation CC 6.0
    *  - PASCAL61
       - GPU
-      - NVIDIA Pascal generation CC 6.1 GPU
+      - NVIDIA Pascal generation CC 6.1
    *  - VOLTA70
       - GPU
-      - NVIDIA Volta generation CC 7.0 GPU
+      - NVIDIA Volta generation CC 7.0
    *  - VOLTA72
       - GPU
-      - NVIDIA Volta generation CC 7.2 GPU
+      - NVIDIA Volta generation CC 7.2
    *  - TURING75
       - GPU
-      - NVIDIA Turing generation CC 7.5 GPU
+      - NVIDIA Turing generation CC 7.5
    *  - AMPERE80
       - GPU
-      - NVIDIA Ampere generation CC 8.0 GPU
+      - NVIDIA Ampere generation CC 8.0
    *  - AMPERE86
       - GPU
-      - NVIDIA Ampere generation CC 8.6 GPU
+      - NVIDIA Ampere generation CC 8.6
    *  - ADA89
       - GPU
-      - NVIDIA Ada Lovelace generation CC 8.9 GPU
+      - NVIDIA Ada generation CC 8.9
    *  - HOPPER90
       - GPU
-      - NVIDIA Hopper generation CC 9.0 GPU
+      - NVIDIA Hopper generation CC 9.0
    *  - AMD_GFX906
       - GPU
-      - AMD GPU MI50/MI60
+      - AMD GPU MI50/60
    *  - AMD_GFX908
       - GPU
       - AMD GPU MI100
    *  - AMD_GFX90A
       - GPU
       - AMD GPU MI200
+   *  - AMD_GFX940
+      - GPU
+      - AMD GPU MI300
    *  - AMD_GFX942
       - GPU
       - AMD GPU MI300
+   *  - AMD_GFX942_APU
+      - GPU
+      - AMD APU MI300A
    *  - AMD_GFX1030
       - GPU
       - AMD GPU V620/W6800
@@ -679,7 +691,7 @@ They must be specified in uppercase.
       - AMD GPU RX7900XTX
    *  - AMD_GFX1103
       - GPU
-      - AMD Phoenix APU with Radeon 740M/760M/780M/880M/890M
+      - AMD APU Phoenix
    *  - INTEL_GEN
       - GPU
       - SPIR64-based devices, e.g. Intel GPUs, using JIT
@@ -702,7 +714,7 @@ They must be specified in uppercase.
       - GPU
       - Intel GPU Ponte Vecchio
 
-This list was last updated for version 4.3.0 of the Kokkos library.
+This list was last updated for version 4.5.1 of the Kokkos library.
 
 .. tabs::
 

lib/kokkos/CHANGELOG.md

@@ -1,7 +1,112 @@
 # CHANGELOG
 
+## 4.5.01
+
+[Full Changelog](https://github.com/kokkos/kokkos/compare/4.5.00...4.5.01)
+
+### Bug Fixes
+
+* Fix re-builds after cleaning the binary tree when doing `add_subdirectory` on the Kokkos source [\#7557](https://github.com/kokkos/kokkos/pull/7557)
+* Update mdspan to include fix for submdspan and bracket operator with clang 15&16 [\#7559](https://github.com/kokkos/kokkos/pull/7559)
+* Fix DynRankView performance regression by re-introducing shortcut operator() impls [\#7606](https://github.com/kokkos/kokkos/pull/7606)
+* Add missing MI300A (`GFX942_APU`) option to Makefile build-system
+
+## 4.5.00
+
+[Full Changelog](https://github.com/kokkos/kokkos/compare/4.4.01...4.5.00)
+
+### Features
+
+* SYCL backend graduated to production ready
+* Introduce new `SequentialHostInit` view allocation property [\#7229](https://github.com/kokkos/kokkos/pull/7229) (backported in 4.4.01)
+* Support building with Run-Time Type Information (RTTI) disabled
+* Add new `KOKKOS_RELOCATABLE_FUNCTION` function annotation macro [\#5993](https://github.com/kokkos/kokkos/pull/5993)
+
+### Backend and Architecture Enhancements
+
+#### CUDA
+
+* Adding occupancy tuning for CUDA architectures [\#6788](https://github.com/kokkos/kokkos/pull/6788)
+* By default disable `cudaMallocAsync` (i.e., revert the change made in version 4.2) [\#7353](https://github.com/kokkos/kokkos/pull/7353)
+
+#### HIP
+
+* Add support for AMD Phoenix APUs with Radeon 740M/760M/780M/880M/890M [\#7162](https://github.com/kokkos/kokkos/pull/7162)
+* Update maximum waves per CU values for consumer card [\#7347](https://github.com/kokkos/kokkos/pull/7347)
+* Check that Kokkos is running on the architecture it was compiled for [\#7379](https://github.com/kokkos/kokkos/pull/7379)
+* Add opt-in option to use `hipMallocAsync` instead of `hipMalloc` [\#7324](https://github.com/kokkos/kokkos/pull/7324)
+* Introduce new architecture option `AMD_GFX942_APU` for MI300A [\#7462](https://github.com/kokkos/kokkos/pull/7462)
+
+#### SYCL
+
+* Move the `SYCL` backend out of the `Experimental` namespace [\#7171](https://github.com/kokkos/kokkos/pull/7171)
+* Introduce `KOKKOS_ENABLE_SYCL_RELOCATABLE_DEVICE_CODE` as CMake option [\#5993](https://github.com/kokkos/kokkos/pull/5993)
+
+#### OpenACC
+
+* Add support for building with the Clacc compiler [\#7198](https://github.com/kokkos/kokkos/pull/7198)
+* Workaround NVHPC collapse clause bug for `MDRangePolicy` [\#7425](https://github.com/kokkos/kokkos/pull/7425)
+
+#### HPX
+
+* Implement `Experimental::partition_space` to produce truly independent execution spaces [\#7287](https://github.com/kokkos/kokkos/pull/7287)
+
+#### Threads
+
+* Fix compilation for `parallel_reduce` `MDRange` with `Dynamic` scheduling [\#7478](https://github.com/kokkos/kokkos/pull/7478)
+* Fix race conditions on ARM architectures [\#7498](https://github.com/kokkos/kokkos/pull/7498)
+
+#### OpenMP
+
+* Fix run time behavior when compiling with `-fvisibility-hidden` [\#7284](https://github.com/kokkos/kokkos/pull/7284) (backported in 4.4.01)
+* Fix linking with Cray Clang compiler [\#7341](https://github.com/kokkos/kokkos/pull/7341)
+
+#### Serial
+
+* Allow `Kokkos_ENABLE_ATOMICS_BYPASS` to skip mutexes to remediate performance regression in 4.4 [\#7369](https://github.com/kokkos/kokkos/pull/7369)
+
+### General Enhancements
+
+* Improve `View` initialization/destruction for non-scalar trivial and trivially-destructible types [\#7219](https://github.com/kokkos/kokkos/pull/7219) [\#7225](https://github.com/kokkos/kokkos/pull/7225)
+* Add getters for default tile sizes used in `MDRangePolicy` [\#6839](https://github.com/kokkos/kokkos/pull/6839)
+* Improve performance of `Kokkos::sort` when `std::sort` is used [\#7264](https://github.com/kokkos/kokkos/pull/7264)
+* Add range-based for loop support for `Array<T, N>` [\#7293](https://github.com/kokkos/kokkos/pull/7293)
+* Allow functors as reducers for nested team parallel reduce [\#6921](https://github.com/kokkos/kokkos/pull/6921)
+* Avoid making copies of string rvalue reference arguments to `view_alloc()` [\#7364](https://github.com/kokkos/kokkos/pull/7364)
+* Add `atomic_{mod,xor,nand,lshift,rshift}` [\#7458](https://github.com/kokkos/kokkos/pull/7458)
+* Allow using `SequentialHostInit` with `Kokkos::DualView` [\#7456](https://github.com/kokkos/kokkos/pull/7456)
+* Add `Graph::instantiate()` [\#7240](https://github.com/kokkos/kokkos/pull/7240)
+* Allow an arbitrary execution space instance to be used in `Kokkos::Graph::submit()` [\#7249](https://github.com/kokkos/kokkos/pull/7249)
+* Enable compile-time diagnostic of illegal reduction target for graphs [\#7460](https://github.com/kokkos/kokkos/pull/7460)
+
+### Build System Changes
+
+* Make sure backend-specific options such as `IMPL_CUDA_MALLOC_ASYNC` only show when that backend is actually enabled [\#7228](https://github.com/kokkos/kokkos/pull/7228)
+* Major refactoring removing `TriBITS` paths [\#6164](https://github.com/kokkos/kokkos/pull/6164)
+* Add support for SpacemiT K60 (RISC-V) [\#7160](https://github.com/kokkos/kokkos/pull/7160)
+
+### Deprecations
+
+* Deprecate Tasking interface [\#7393](https://github.com/kokkos/kokkos/pull/7393)
+* Deprecate `atomic_query_version`, `atomic_assign`, `atomic_compare_exchange_strong`, `atomic_{inc, dec}rement` [\#7458](https://github.com/kokkos/kokkos/pull/7458)
+* Deprecate `{OpenMP,HPX}::is_asynchronous()` [\#7322](https://github.com/kokkos/kokkos/pull/7322)
+
+### Bug Fixes
+
+* Fix undefined behavior in `BinSort` when sorting within bins on host [\#7223](https://github.com/kokkos/kokkos/pull/7223)
+* Using CUDA limits to set extents for blocks, grids [\#7235](https://github.com/kokkos/kokkos/pull/7235)
+* Fix `deep_copy (serial_exec, dst, src)` with multiple host backends [\#7245](https://github.com/kokkos/kokkos/pull/7245)
+* Skip `RangePolicy` bounds conversion checks if roundtrip convertibility is not provided [\#7172](https://github.com/kokkos/kokkos/pull/7172)
+* Allow extracting host and device views from `DualView` with `const` value type [\#7242](https://github.com/kokkos/kokkos/pull/7242)
+* Fix `TeamPolicy` array reduction for CUDA and HIP [\#6296](https://github.com/kokkos/kokkos/pull/6296)
+* Fix implicit copy assignment operators in few AVX2 masks being deleted [\#7296](https://github.com/kokkos/kokkos/pull/7296)
+* Fix configuring without architecture flags for SYCL [\#7303](https://github.com/kokkos/kokkos/pull/7303)
+* Set an initial value index during join of `MinLoc`, `MaxLoc` or `MinMaxLoc` [\#7330](https://github.com/kokkos/kokkos/pull/7330)
+* Fix storage lifetime of driver for global launch of graph nodes for CUDA and HIP [\#7365](https://github.com/kokkos/kokkos/pull/7365)
+* Make `value_type` for `RandomAccessIterator` non-`const` [\#7485](https://github.com/kokkos/kokkos/pull/7485)
+
 ## [4.4.01](https://github.com/kokkos/kokkos/tree/4.4.01)
-[Full Changelog](https://github.com/kokkos/kokkos/compare/4.0.00...4.4.01)
+[Full Changelog](https://github.com/kokkos/kokkos/compare/4.4.00...4.4.01)
 
 ### Features:
 * Introduce new SequentialHostInit view allocation property [\#7229](https://github.com/kokkos/kokkos/pull/7229)
@@ -13,7 +118,7 @@
 
 ### Bug Fixes
 * OpenMP: Fix issue related to the visibility of an internal symbol with shared libraries that affected `ScatterView` in particular [\#7284](https://github.com/kokkos/kokkos/pull/7284)
-* Fix implicit copy assignment operators in few AVX2 masks being deleted [#7296](https://github.com/kokkos/kokkos/pull/7296)
+* Fix implicit copy assignment operators in few AVX2 masks being deleted [\#7296](https://github.com/kokkos/kokkos/pull/7296)
 
 ## [4.4.00](https://github.com/kokkos/kokkos/tree/4.4.00)
 [Full Changelog](https://github.com/kokkos/kokkos/compare/4.3.01...4.4.00)
@@ -57,6 +162,7 @@
 * SIMD: Allow flexible vector width for 32 bit types [\#6802](https://github.com/kokkos/kokkos/pull/6802)
 * Updates for `Kokkos::Array`: add `kokkos_swap(Array<T, N>)` specialization [\#6943](https://github.com/kokkos/kokkos/pull/6943), add `Kokkos::to_array` [\#6375](https://github.com/kokkos/kokkos/pull/6375),  make `Kokkos::Array` equality-comparable [\#7148](https://github.com/kokkos/kokkos/pull/7148)
 * Structured binding support for `Kokkos::complex` [\#7040](https://github.com/kokkos/kokkos/pull/7040)
+* Introduce `KOKKOS_DEDUCTION_GUIDE` macro to allow for portable user-defined deduction guides [\#6954](https://github.com/kokkos/kokkos/pull/6954)
 
 ### Build System Changes
 * Do not require OpenMP support for languages other than CXX [\#6965](https://github.com/kokkos/kokkos/pull/6965)
@@ -1388,7 +1494,7 @@
 **Closed issues:**
 
 - Silent error (Validate storage level arg to set_scratch_size) [\#3097](https://github.com/kokkos/kokkos/issues/3097)
-- Remove KOKKKOS\_ENABLE\_PROFILING Option [\#3095](https://github.com/kokkos/kokkos/issues/3095)
+- Remove KOKKOS\_ENABLE\_PROFILING Option [\#3095](https://github.com/kokkos/kokkos/issues/3095)
 - Cuda 11 -\> allow C++17 [\#3083](https://github.com/kokkos/kokkos/issues/3083)
 - In source build failure not explained [\#3081](https://github.com/kokkos/kokkos/issues/3081)
 - Allow naming of Views for initialization kernel [\#3070](https://github.com/kokkos/kokkos/issues/3070)