Update reduction

Author: jamesnulliu

.clangd

@@ -1,24 +1,3 @@
----
-If:
-  PathMatch: [.*\.c, .*\.h]
-CompileFlags:
-  Add: [-xc, -std=c11]
-  Remove: [-xc++, -std=c++*]
-Diagnostics:
-  ClangTidy:
-    Remove:
-      - modernize-*
-      - cppcoreguidelines-*
-      - performance-*
-      - google-*
-      - cert-dcl58-cpp
-      - cert-err58-cpp
-      - misc-new-delete-overloads
-      - misc-non-private-member-variables-in-classes
-      - misc-uniqueptr-reset-release
----
-If:
-  PathMatch: [.*\.cpp, .*\.cu, .*\.hpp, .*\.cuh]
 CompileFlags:
   Add: 
     - -std=c++20
@@ -36,6 +15,10 @@ CompileFlags:
     - -Xcompiler*
     - -arch=*
 
+Index:
+  Background: Build
+  StandardLibrary: Yes
+
 Diagnostics:
   UnusedIncludes: None
   MissingIncludes: None

.vscode/settings.json

@@ -28,7 +28,7 @@
         "--header-insertion=never",
         "--header-insertion-decorators",
         "--background-index",
-        // "-j=8",
+        "-j=16",
         "--pch-storage=memory",
         "--function-arg-placeholders=false",
     ],

configs/lib-tests.yml

@@ -7,4 +7,6 @@ OpTest:
     - nInputs: 256
       divider: 4
     - nInputs: 2048
-      divider: 4
+      divider: 4
+  MulReduction:
+    - nInputs: 32

csrc/include/pmpp/pch.hpp

@@ -0,0 +1,12 @@
+#pragma once
+
+#include <cuda_runtime.h>
+#include <torch/torch.h>
+#include <torch/python.h>
+#include <type_traits>
+#include <algorithm>
+
+#include "pmpp/system.hpp"
+#include "pmpp/types/cu_types.cuh"
+#include "pmpp/types/cxx_types.hpp"
+#include "pmpp/types/torch_types.hpp"

csrc/include/pmpp/system.h csrc/include/pmpp/system.hppcsrc/include/pmpp/system.h renamed to csrc/include/pmpp/system.hpp

@@ -2,6 +2,7 @@
 
 #include <iterator>
 #include <ranges>
+#include <type_traits>
 
 namespace pmpp
 {

csrc/include/pmpp/types/c_types.h

@@ -1,3 +1,5 @@
+#pragma once
+
 #include <cuda_bf16.h>
 #include <cuda_fp16.h>
 #include <cuda_fp8.h>

csrc/include/pmpp/types/concepts.hpp

@@ -0,0 +1,6 @@
+#pragma once
+
+#include "pmpp/types/concepts.hpp"
+#include "pmpp/types/cu_types.cuh"
+#include "pmpp/types/cxx_types.hpp"
+#include "pmpp/types/torch_types.hpp"

csrc/include/pmpp/types/cu_types.cuh

@@ -13,9 +13,16 @@ target_include_directories(
     $<INSTALL_INTERFACE:include>
 )
 
+target_precompile_headers(
+    ${LIB_NAME} PUBLIC
+    $<BUILD_INTERFACE:${PROJECT_SOURCE_DIR}/include/pmpp/pch.hpp>
+    $<INSTALL_INTERFACE:include/pmpp/pch.hpp>
+)
+
 target_link_libraries(
     ${LIB_NAME} 
     PUBLIC 
+    Python::Python
     ${TORCH_LIBRARIES}
     fmt::fmt
     msft_proxy