[FEA] Atomics codegen refactor (#1993)

* Initial draft of new atomics backend

* Change atomic fetch ops back to tag dispatch

* Save wip

* Add load/store and support for MMIO

* Begin working on exch

* Enable formatting exchange

* Several signed-ness fixes

* Make atomics ptx tests build. Lit tests are a WIP.

* Fix load/store, some volatileness, and min/max

* Formatting and enabled codegen in all builds

* Make integral.pass.cpp pass

* Make the rest of the atomics tests pass

* Use 128b ld/st instead of vector load as it is not atomic across the whole atom

* Fix copy-paste mistake in load/store

* Whitespace fixup

* Fix 128b .exch using .cas operands

* Make codegen link fmt as PRIVATE

Co-authored-by: Allison Piper <[email protected]>

* Simplify MMIO down to a static array.

Co-authored-by: Bernhard Manfred Gruber <[email protected]>

* Static -> Inline for codegen functions. Replace endl with '\n'.

* Supply the output stream directly to `fmt::format`

* Update fmtlib.

* Revert `fmt::format(out...)` changes. They don't work on MSVC.

* Fixup libcudacxx codegen CMake stuff

* Remove sneaky cstdef include that was auto-added

* [pre-commit.ci] auto code formatting

---------

Co-authored-by: Allison Piper <[email protected]>
Co-authored-by: Bernhard Manfred Gruber <[email protected]>
Co-authored-by: Michael Schellenberger Costa <[email protected]>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

CMakePresets.json

@@ -478,7 +478,7 @@
       ],
       "filter": {
         "exclude": {
-          "name": "^libcudacxx\\.test\\.lit$"
+          "name": "^libcudacxx\\.test\\.(lit|atomics\\.codegen\\.diff)$"
         }
       }
     },
@@ -487,7 +487,7 @@
       "configurePreset": "libcudacxx-codegen",
       "filter": {
         "include": {
-          "name": "^libcudacxx\\.atomics\\.codegen.*$"
+          "name": "^libcudacxx\\.test\\.atomics\\.codegen.*$"
         }
       }
     },

libcudacxx/CMakeLists.txt

@@ -33,7 +33,7 @@ include(CTest)
 enable_testing()
 
 # Add codegen module
-option(libcudacxx_ENABLE_CODEGEN "Enable ctest-based testing." OFF)
+option(libcudacxx_ENABLE_CODEGEN "Enable libcudacxx's atomics backend codegen and tests." OFF)
 if (libcudacxx_ENABLE_CODEGEN)
   add_subdirectory(codegen)
 endif()

libcudacxx/codegen/CMakeLists.txt

@@ -1,44 +1,44 @@
 ## Codegen adds the following build targets
 #    libcudacxx.atomics.codegen
-#    libcudacxx.atomics.codegen.execute
 #    libcudacxx.atomics.codegen.install
 ## Test targets:
-#    libcudacxx.atomics.codegen.diff
+#    libcudacxx.test.atomics.codegen.diff
 
-add_custom_target(libcudacxx.atomics.codegen)
+include(${CMAKE_SOURCE_DIR}/cub/cmake/CPM.cmake)
+CPMAddPackage("gh:fmtlib/fmt#11.0.1")
 
 add_executable(
     codegen
     EXCLUDE_FROM_ALL
     codegen.cpp
 )
 
-target_compile_features(
-    codegen PRIVATE cxx_std_14
-)
+target_link_libraries(codegen PRIVATE fmt)
 
-add_dependencies(libcudacxx.atomics.codegen codegen)
+set_property(TARGET codegen PROPERTY CXX_STANDARD 17)
 
 set(atomic_generated_output "${libcudacxx_BINARY_DIR}/codegen/cuda_ptx_generated.h")
 set(atomic_install_location "${libcudacxx_SOURCE_DIR}/include/cuda/std/__atomic/functions")
 
 add_custom_target(
-    libcudacxx.atomics.codegen.execute
-    COMMAND codegen
+    libcudacxx.atomics.codegen
+    COMMAND codegen "${atomic_generated_output}"
     BYPRODUCTS "${atomic_generated_output}"
 )
 
-add_dependencies(libcudacxx.atomics.codegen libcudacxx.atomics.codegen.execute)
-
 add_custom_target(
     libcudacxx.atomics.codegen.install
     COMMAND ${CMAKE_COMMAND} -E copy "${atomic_generated_output}" "${atomic_install_location}/cuda_ptx_generated.h"
+    DEPENDS libcudacxx.atomics.codegen
     BYPRODUCTS "${atomic_install_location}/cuda_ptx_generated.h"
 )
 
-add_dependencies(libcudacxx.atomics.codegen.install libcudacxx.atomics.codegen.execute)
-
 add_test(
-    NAME libcudacxx.atomics.codegen.diff
+    NAME libcudacxx.test.atomics.codegen.diff
     COMMAND ${CMAKE_COMMAND} -E compare_files "${atomic_install_location}/cuda_ptx_generated.h" "${atomic_generated_output}"
 )
+
+set_tests_properties(
+    libcudacxx.test.atomics.codegen.diff
+    PROPERTIES REQUIRED_FILES "${atomic_generated_output}"
+)