Update EXC Only with master: Squashed commit of the following:

commit 62ac01fab0767f13a7424bcea8a940c92e3aa7e2 Author: David Williams-Young <dbwy@lbl.gov> Date: Fri May 31 13:36:46 2024 -0700 Fix EXC-only path for host integrators, UTs pass commit 57f305c39ca4c858d5050c5dfc32c3b1ffab2e82 Merge: 1c844b4 92bbbe2 Author: David Williams-Young <dbwy@lbl.gov> Date: Fri May 31 12:28:27 2024 -0700 Merge branch 'master' into merge_master commit 92bbbe2 Author: David Williams-Young <wavefunction91@gmail.com> Date: Fri May 31 10:50:04 2024 -0700 actions/checkout@v3 -> actions/checkout@v4 (#131) Bump actions/checkout version to quiet GHA warnings commit b61ee7a Author: mikovtun <91164556+mikovtun@users.noreply.github.com> Date: Fri May 31 10:21:21 2024 -0700 Spellcheck Error Messages (#129) * Fixed spelling in error messages and homogenized capitalization * Update include/gauxc/load_balancer.hpp Co-authored-by: David Williams-Young <wavefunction91@gmail.com> --------- Co-authored-by: David Williams-Young <wavefunction91@gmail.com> commit 905c36a Author: Ajay Panyala <ajay.panyala@gmail.com> Date: Fri May 31 09:07:08 2024 -0700 cutlass requires cuda CC >= 8.0 (#130) commit b9c2161 Author: David Williams-Young <dbwy@lbl.gov> Date: Tue May 21 15:13:43 2024 -0700 Refactor ShellBatched Integrators (#127) * Refactor of RKS/UKS/GKS Host drivers to reduce code replication + various QoL * ShellBatched Refactor - Host compiles and tests, device untested * Fix new ShellBatched for Device, add additional std::future::get to proagate exceptions * Update copyright year on old files, shellbatched -> shell_batched * document why the extra std::future::get is there commit 6a8f4bf Author: David Williams-Young <dbwy@lbl.gov> Date: Wed May 8 09:32:51 2024 -0700 Add Runtime Environment Query Functions (#126) * Add runtime environment query functions * Update gauxc_config.hpp.in commit cf6b85c Author: Ajay Panyala <ajay.panyala@gmail.com> Date: Tue May 7 09:48:17 2024 -0700 install nccl module file commit 3e44fcc Author: David Williams-Young <dbwy@lbl.gov> Date: Mon May 6 16:01:48 2024 -0700 Make CUTLASS a build-only dependency
wavefunction91 · May 31, 2024 · 50977ab · 50977ab
1 parent 1c844b4
commit 50977ab
Show file tree

Hide file tree

Showing 48 changed files with 1,263 additions and 977 deletions.
diff --git a/.github/workflows/build_and_test_compiler_zoo.yml b/.github/workflows/build_and_test_compiler_zoo.yml
@@ -22,7 +22,7 @@ jobs:
               openmp_flag: ON
 
     steps:
-    - uses: actions/checkout@v3
+    - uses: actions/checkout@v4
 
     - name: Setup Compiler 
       shell: bash
@@ -71,7 +71,7 @@ jobs:
             flags: [ {magma: OFF, cutlass: OFF}, {magma: ON, cutlass: OFF}, {magma: OFF, cutlass: ON} ]
 
     steps:
-    - uses: actions/checkout@v3
+    - uses: actions/checkout@v4
 
     - name: Setup Build Type
       shell: bash
@@ -109,7 +109,7 @@ jobs:
         image: dbwy/chemistry
 
     steps:
-    - uses: actions/checkout@v3
+    - uses: actions/checkout@v4
 
     - name: Setup Compiler 
       shell: bash
@@ -142,7 +142,7 @@ jobs:
         image: dbwy/chemistry
 
     steps:
-    - uses: actions/checkout@v3
+    - uses: actions/checkout@v4
 
     - name: Setup Compiler 
       shell: bash
@@ -170,7 +170,7 @@ jobs:
         image: dbwy/chemistry
 
     steps:
-    - uses: actions/checkout@v3
+    - uses: actions/checkout@v4
 
     - name: Setup Compiler 
       shell: bash

diff --git a/cmake/gauxc-cutlass.cmake b/cmake/gauxc-cutlass.cmake
@@ -1,7 +1,7 @@
 # Check that only CUDA CC 8.0+ is enabled
 foreach( cuda_arch ${CMAKE_CUDA_ARCHITECTURES} )
-  if( NOT cuda_arch EQUAL 80 )
-    message(FATAL_ERROR "GauXC Requires CUDA CC 8.0 For CUTLASS")
+  if( NOT cuda_arch GREATER_EQUAL 80 )
+    message(FATAL_ERROR "GauXC Requires CUDA CC >= 8.0 For CUTLASS")
   endif()
 endforeach()
 

diff --git a/include/gauxc/enums.hpp b/include/gauxc/enums.hpp
@@ -51,5 +51,11 @@ enum class ExecutionSpace {
   Device ///< Execute task on the device (e.g. GPU)
 };
 
+/// Supported Algorithms / Integrands
+enum class SupportedAlg {
+  XC,
+  DEN,
+  SNLINK
+};
 
 } // namespace GauXC
diff --git a/include/gauxc/gauxc_config.hpp.in b/include/gauxc/gauxc_config.hpp.in
@@ -18,7 +18,16 @@
 #cmakedefine GAUXC_HAS_HDF5
 #cmakedefine GAUXC_USE_FAST_RSQRT
 
+#ifdef GAUXC_HAS_HOST
+#cmakedefine GAUXC_CPU_XC_MAX_AM     @GAUXC_CPU_XC_MAX_AM@
+#cmakedefine GAUXC_CPU_SNLINK_MAX_AM @GAUXC_CPU_SNLINK_MAX_AM@
+#endif
+
 #cmakedefine GAUXC_HAS_DEVICE
+#ifdef GAUXC_HAS_DEVICE
+#cmakedefine GAUXC_GPU_XC_MAX_AM     @GAUXC_GPU_XC_MAX_AM@
+#cmakedefine GAUXC_GPU_SNLINK_MAX_AM @GAUXC_GPU_SNLINK_MAX_AM@
+#endif
 
 #if defined(__CUDACC__) || defined(__HIPCC__)
   #define HOST_DEVICE_ACCESSIBLE __host__ __device__

diff --git a/include/gauxc/load_balancer.hpp b/include/gauxc/load_balancer.hpp
@@ -26,7 +26,7 @@ namespace detail {
 /// State tracker for LoadBalancer instances 
 struct LoadBalancerState {
   bool modified_weights_are_stored = false; 
-    ///< Whether the load balancer currently sotred partitioned weights
+    ///< Whether the load balancer currently stores partitioned weights
 };
 
 
@@ -136,15 +136,15 @@ class LoadBalancerFactory {
    *    - "REPLICATED": Read as "REPLICATED-PETITE"
    *    - "REPLICATED-PETITE": Replicate the load balancer function, only keep
    *                           non negligible basis functions
-   *    - "REPLICATE-FILLIN": Same as "REPLICATED-PETITE" except if two 
-   *                          non-adjacent bfns are kept, the gaps are filled in.
-   *                          This gurantees contiguous memory access but leads
-   *                          to significantly more work. Not advised for general 
-   *                          usage
+   *    - "REPLICATED-FILLIN": Same as "REPLICATED-PETITE" except if two 
+   *                           non-adjacent bfns are kept, the gaps are filled in.
+   *                           This gurantees contiguous memory access but leads
+   *                           to significantly more work. Not advised for general 
+   *                           usage
    * 
    *    Currently accepted values for Device execution space:
    *      - "DEFAULT": Read as "REPLICATED"
-   *      - "REPLICATAED": Same as Host::REPLICATED-PETITE
+   *      - "REPLICATED": Same as Host::REPLICATED-PETITE
    */
   LoadBalancerFactory( ExecutionSpace ex, std::string kernel_name );
 

diff --git a/include/gauxc/util/environment.hpp b/include/gauxc/util/environment.hpp
@@ -0,0 +1,42 @@
+/**
+ * GauXC Copyright (c) 2020-2024, The Regents of the University of California,
+ * through Lawrence Berkeley National Laboratory (subject to receipt of
+ * any required approvals from the U.S. Dept. of Energy). All rights reserved.
+ *
+ * See LICENSE.txt for details
+ */
+#pragma once
+#include <gauxc/gauxc_config.hpp>
+#include <gauxc/enums.hpp>
+
+namespace GauXC {
+
+inline int gauxc_max_am(ExecutionSpace ex, SupportedAlg alg) {
+  switch(ex) {
+    #ifdef GAUXC_HAS_HOST
+    case ExecutionSpace::Host:
+      switch(alg) {
+        case SupportedAlg::XC: 
+        case SupportedAlg::DEN: 
+          return GAUXC_CPU_XC_MAX_AM;
+        case SupportedAlg::SNLINK:
+          return GAUXC_CPU_SNLINK_MAX_AM;
+        default: return -1;
+      }
+    #endif
+    #ifdef GAUXC_HAS_DEVICE
+    case ExecutionSpace::Device:
+      switch(alg) {
+        case SupportedAlg::XC: 
+        case SupportedAlg::DEN: 
+          return GAUXC_GPU_XC_MAX_AM;
+        case SupportedAlg::SNLINK:
+          return GAUXC_GPU_SNLINK_MAX_AM;
+        default: return -1;
+      }
+    #endif
+    default: return -1;
+  }
+}
+
+}
diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
@@ -100,6 +100,17 @@ add_subdirectory( external )
 
 add_library( gauxc::gauxc ALIAS gauxc )
 
+
+if(GAUXC_HAS_HOST)
+  set(GAUXC_CPU_XC_MAX_AM     6) 
+  set(GAUXC_CPU_SNLINK_MAX_AM 6)
+endif()
+
+if(GAUXC_HAS_DEVICE)
+  set(GAUXC_GPU_XC_MAX_AM     4) 
+  set(GAUXC_GPU_SNLINK_MAX_AM 2)
+endif()
+
 # Generate config file
 configure_file( 
   ${PROJECT_SOURCE_DIR}/include/gauxc/gauxc_config.hpp.in
@@ -151,10 +162,10 @@ set_target_properties(gauxc
     "GAUXC_HAS_MPI"        ${GAUXC_HAS_MPI}
     "GAUXC_HAS_OPENMP"     ${GAUXC_HAS_OPENMP}
     "GAUXC_HAS_HDF5"       ${GAUXC_HAS_HDF5}
-    "GAUXC_CPU_XC_MAX_AM"     6 
-    "GAUXC_CPU_SNLINK_MAX_AM" 6
-    "GAUXC_GPU_XC_MAX_AM"     4 
-    "GAUXC_GPU_SNLINK_MAX_AM" 2
+    "GAUXC_CPU_XC_MAX_AM"     ${GAUXC_CPU_XC_MAX_AM}     
+    "GAUXC_CPU_SNLINK_MAX_AM" ${GAUXC_CPU_SNLINK_MAX_AM}
+    "GAUXC_GPU_XC_MAX_AM"     ${GAUXC_GPU_XC_MAX_AM}     
+    "GAUXC_GPU_SNLINK_MAX_AM" ${GAUXC_GPU_SNLINK_MAX_AM}
 )
 set_property(TARGET gauxc APPEND PROPERTY EXPORT_PROPERTIES "${export_properties}")
 
@@ -204,6 +215,7 @@ configure_package_config_file(
 # Install CMake files
 install( FILES
   ${PROJECT_SOURCE_DIR}/cmake/modules/FindMAGMA.cmake
+  ${PROJECT_SOURCE_DIR}/cmake/modules/FindNCCL.cmake
   ${CMAKE_CURRENT_BINARY_DIR}/gauxc-config.cmake
   ${CMAKE_CURRENT_BINARY_DIR}/gauxc-config-version.cmake
   DESTINATION ${INSTALL_CONFIGDIR}

diff --git a/src/molgrid_defaults.cxx b/src/molgrid_defaults.cxx
@@ -119,7 +119,7 @@ std::tuple<RadialSize,AngularSize>
       }
 
     default:
-      GAUXC_GENERIC_EXCEPTION("Not A Resognized Standard Grid");
+      GAUXC_GENERIC_EXCEPTION("Not A Recognized Standard Grid");
       abort();
   }
 

diff --git a/src/xc_integrator/CMakeLists.txt b/src/xc_integrator/CMakeLists.txt
@@ -7,6 +7,7 @@
 #
 add_subdirectory(integrator_util)
 add_subdirectory(local_work_driver)
+add_subdirectory(shell_batched)
 add_subdirectory(replicated)
 add_subdirectory(xc_data)
 

diff --git a/src/xc_integrator/local_work_driver/device/cuda/CMakeLists.txt b/src/xc_integrator/local_work_driver/device/cuda/CMakeLists.txt
@@ -41,7 +41,7 @@ endforeach()
 add_subdirectory(obara_saika)
 if(GAUXC_ENABLE_CUTLASS)
     include( gauxc-cutlass )
-    target_link_libraries( gauxc PRIVATE gauxc_cutlass )
+    target_link_libraries( gauxc PRIVATE $<BUILD_INTERFACE:gauxc_cutlass> )
     target_sources( gauxc PRIVATE
       scheme1_cutlass_base.cxx 
       scheme1_cutlass_data_base.cxx   

diff --git a/src/xc_integrator/local_work_driver/device/scheme1_base.cxx b/src/xc_integrator/local_work_driver/device/scheme1_base.cxx
@@ -303,34 +303,34 @@ void AoSScheme1Base::eval_zmat_mgga_vxc_rks( XCDeviceData* _data, bool do_lapl){
 
 void AoSScheme1Base::eval_zmat_lda_vxc_uks( XCDeviceData* ){
 
-  GAUXC_GENERIC_EXCEPTION("UKS NOT YET IMPLEMENTED FOR DEVICE");
+  GAUXC_GENERIC_EXCEPTION("UKS Not Yet Implemented For Device");
 
 }
 
 void AoSScheme1Base::eval_zmat_gga_vxc_uks( XCDeviceData* ){
 
-  GAUXC_GENERIC_EXCEPTION("UKS NOT YET IMPLEMENTED FOR DEVICE");
+  GAUXC_GENERIC_EXCEPTION("UKS Not Yet Implemented For Device");
 
 }
 
 void AoSScheme1Base::eval_zmat_mgga_vxc_uks( XCDeviceData*, bool /*do_lapl*/){
-  GAUXC_GENERIC_EXCEPTION("UKS MGGA NOT YET IMPLEMENTED FOR DEVICE");
+  GAUXC_GENERIC_EXCEPTION("UKS MGGA Not Yet Implemented For Device");
 }
 
 void AoSScheme1Base::eval_zmat_lda_vxc_gks( XCDeviceData* ){
 
-  GAUXC_GENERIC_EXCEPTION("GKS NOT YET IMPLEMENTED FOR DEVICE");
+  GAUXC_GENERIC_EXCEPTION("GKS Not Yet Implemented For Device");
 
 }
 
 void AoSScheme1Base::eval_zmat_gga_vxc_gks( XCDeviceData* ){
 
-  GAUXC_GENERIC_EXCEPTION("GKS NOT YET IMPLEMENTED FOR DEVICE");
+  GAUXC_GENERIC_EXCEPTION("GKS Not Yet Implemented For Device");
 
 }
 
 void AoSScheme1Base::eval_zmat_mgga_vxc_gks( XCDeviceData*, bool /*do_lapl*/){
-  GAUXC_GENERIC_EXCEPTION("GKS MGGA NOT YET IMPLEMENTED FOR DEVICE");
+  GAUXC_GENERIC_EXCEPTION("GKS MGGA Not Yet Implemented For Device");
 }
 
 void AoSScheme1Base::eval_mmat_mgga_vxc_rks( XCDeviceData* _data, bool do_lapl){
@@ -356,10 +356,10 @@ void AoSScheme1Base::eval_mmat_mgga_vxc_rks( XCDeviceData* _data, bool do_lapl){
   data->device_backend_->check_error("mmat_mgga" __FILE__ ": " + std::to_string(__LINE__));
 }
 void AoSScheme1Base::eval_mmat_mgga_vxc_uks( XCDeviceData*, bool /*do_lapl*/){
-  GAUXC_GENERIC_EXCEPTION("UKS MGGA NOT YET IMPLEMENTED FOR DEVICE");
+  GAUXC_GENERIC_EXCEPTION("UKS MGGA Not Yet Implemented For Device");
 }
 void AoSScheme1Base::eval_mmat_mgga_vxc_gks( XCDeviceData*, bool /*do_lapl*/){
-  GAUXC_GENERIC_EXCEPTION("GKS MGGA NOT YET IMPLEMENTED FOR DEVICE");
+  GAUXC_GENERIC_EXCEPTION("GKS MGGA Not Yet Implemented For Device");
 }
 
 void AoSScheme1Base::eval_collocation( XCDeviceData* _data ) {
@@ -622,35 +622,35 @@ void AoSScheme1Base::eval_uvvar_mgga_rks( XCDeviceData* _data, bool do_lapl ){
 
 void AoSScheme1Base::eval_uvvar_lda_uks( XCDeviceData* ){
 
-  GAUXC_GENERIC_EXCEPTION("UKS NOT YET IMPLEMENTED FOR DEVICE"); 
+  GAUXC_GENERIC_EXCEPTION("UKS Not Yet Implemented For Device"); 
 
 }
 
 void AoSScheme1Base::eval_uvvar_gga_uks( XCDeviceData* ){
 
-  GAUXC_GENERIC_EXCEPTION("UKS NOT YET IMPLEMENTED FOR DEVICE");
+  GAUXC_GENERIC_EXCEPTION("UKS Not Yet Implemented For Device");
 
 }
 
 void AoSScheme1Base::eval_uvvar_mgga_uks( XCDeviceData*, bool /*do_lapl*/ ){
-  GAUXC_GENERIC_EXCEPTION("UKS MGGA NOT YET IMPLEMENTED FOR DEVICE"); 
+  GAUXC_GENERIC_EXCEPTION("UKS MGGA Not Yet Implemented For Device"); 
 }
 
 
 void AoSScheme1Base::eval_uvvar_lda_gks( XCDeviceData* ){
 
-    GAUXC_GENERIC_EXCEPTION("GKS NOT YET IMPLEMENTED FOR DEVICE");
+    GAUXC_GENERIC_EXCEPTION("GKS Not Yet Implemented For Device");
 
 }
 
 void AoSScheme1Base::eval_uvvar_gga_gks( XCDeviceData* ){
 
-    GAUXC_GENERIC_EXCEPTION("GKS NOT YET IMPLEMENTED FOR DEVICE");
+    GAUXC_GENERIC_EXCEPTION("GKS Not Yet Implemented For Device");
 
 }
 
 void AoSScheme1Base::eval_uvvar_mgga_gks( XCDeviceData*, bool /*do_lapl*/ ){
-  GAUXC_GENERIC_EXCEPTION("GKS MGGA NOT YET IMPLEMENTED FOR DEVICE"); 
+  GAUXC_GENERIC_EXCEPTION("GKS MGGA Not Yet Implemented For Device"); 
 }
 
 void AoSScheme1Base::eval_kern_exc_vxc_lda( const functional_type& func, 

diff --git a/src/xc_integrator/local_work_driver/factory.cxx b/src/xc_integrator/local_work_driver/factory.cxx
@@ -68,7 +68,7 @@ LocalWorkDriverFactory::ptr_return_t
 
 
   default:
-    GAUXC_GENERIC_EXCEPTION("Execution Space Not Regognized");
+    GAUXC_GENERIC_EXCEPTION("Execution Space Not Recognized");
 
   }