Added Proper Include What You Use (CNugteren#627)

* Made all utilities and backend stuff available through the utilities/utilities.hpp header to avoid needless dozens of includes.

* Removed timing.hpp from include because it used by tuners and diagnostics.

* Added clblast_half.h into the utilities.hpp to be exported because it is a utility to interface with half.

* Made the backend files not actually export themselves to prevent suggestions for IWYU.

* Removed transistive includes from utility because there isn't too many headers that it is necessary.

* Updated src/tuning/* with IWYU.

* Updated src/routines/ and src/routines/level1 with IWYU.

* Did all the files in src/routines/level2/

* Did IWYU on some of the files in src/routines/level3

* Finished IWYU on level 3 routines.

* Added IWYU to the levelx routines.

* Fixed the general purpose files in src/, the CUDA API should also be fixed but it might not be.

* Exported 2 utilities in utilities/utilities.hpp because they ensure the library works with MSVC and android.

* Applied IWYU to the database files that needed to be changed.

* Applied IWYU to clblast_c.h in the include folder.

* Removed android.hpp from include because it isn't supposed to be included explicitly.

* Did minimalist changes to the tests to make sure they compile. Tested with cblas.

* Added the .clangd used to apply IWYU for future contributors.

* Removed unwanted includes of CL/cl.h

* Fixed stddef include for the C API.

* Removed open blas specific include that should not be there.

* Fixed compilation error in test/routines/levelx/xhad.hpp

* Update changelog.

* Fixed clang format

* Added missing include in a routine test to fix appveyor compilation.

* Removed unused file.

* Fixed HEADERS var.

* Added Include What You Use for src/tuning/tuning.hpp

* Added routines.hpp again to avoid duplication in api_common, clblast_cuda, and clblast.cpp

* Improved readability and decreased the amount of pragmas.

* Updated CMakeLists.txt; added routines.hpp into the HEADERS var.

Author: EJainDev

.clangd

@@ -0,0 +1,3 @@
+Diagnostics:
+  MissingIncludes: Strict
+  UnusedIncludes: Strict

CHANGELOG

@@ -5,6 +5,7 @@ Development version (next version)
 - Enabled parallel kernel compilation for faster kernel tuning (see doc/tuning.md)
 - Removed all generator scripts for the API
 - Added 3 new error codes for vector Z in routine XHAD
+- Added include what you use (IWYU) with Clangd Include Cleaner
 
 Version 1.6.3
 - Fixed a bug in the GEMMK=1 kernel (with 2D register tiling) when MWG!=NWG

include/clblast_c.h

@@ -11,6 +11,8 @@
 #ifndef CLBLAST_CLBLAST_C_H_
 #define CLBLAST_CLBLAST_C_H_
 
+#include <stddef.h>
+
 // Includes the normal OpenCL C header
 #ifndef CL_TARGET_OPENCL_VERSION
 #define CL_TARGET_OPENCL_VERSION 110

src/api_common.cpp

@@ -6,10 +6,17 @@
 //
 // =================================================================================================
 
+#include <cstddef>
 #include <string>
+#include <unordered_map>
+#include <vector>
 
 #include "cache.hpp"
+#include "clblast.h"
+#include "database/database_structure.hpp"
 #include "routines/routines.hpp"
+#include "utilities/backend.hpp"
+#include "utilities/clblast_exceptions.hpp"
 #include "utilities/utilities.hpp"
 
 namespace clblast {

src/cache.cpp

@@ -9,11 +9,13 @@
 
 #include "cache.hpp"
 
+#include <memory>
 #include <mutex>
 #include <string>
-#include <vector>
+#include <utility>
 
 #include "database/database.hpp"
+#include "utilities/backend.hpp"
 
 namespace clblast {
 // =================================================================================================

src/cache.hpp

@@ -10,11 +10,15 @@
 #ifndef CLBLAST_CACHE_H_
 #define CLBLAST_CACHE_H_
 
+#include <functional>
 #include <map>
+#include <memory>
 #include <mutex>
 #include <string>
+#include <tuple>
 
-#include "utilities/utilities.hpp"
+#include "clblast.h"
+#include "utilities/backend.hpp"
 
 namespace clblast {
 // =================================================================================================

src/clblast.cpp

@@ -11,9 +11,16 @@
 
 #include "clblast.h"
 
+#include <complex>
+#include <cstddef>
 #include <string>
+#include <vector>
 
+#include "database/database.hpp"
 #include "routines/routines.hpp"
+#include "utilities/backend.hpp"
+#include "utilities/clblast_exceptions.hpp"
+#include "utilities/utilities.hpp"
 
 namespace clblast {
 

src/clblast_c.cpp

@@ -9,10 +9,11 @@
 
 #include "clblast_c.h"
 
+#include <cstddef>
 #include <string>
-#include <unordered_map>
 
 #include "clblast.h"
+#include "utilities/clblast_exceptions.hpp"
 #include "utilities/utilities.hpp"
 
 // Shortcuts to the clblast namespace

src/clblast_cuda.cpp

@@ -11,9 +11,15 @@
 
 #include "clblast_cuda.h"
 
+#include <complex>
+#include <cstddef>
 #include <string>
 
+#include "clblast.h"
+#include "database/database.hpp"
 #include "routines/routines.hpp"
+#include "utilities/backend.hpp"
+#include "utilities/clblast_exceptions.hpp"
 
 namespace clblast {
 

src/clblast_netlib_c.cpp

@@ -12,8 +12,10 @@
 #include "clblast_netlib_c.h"
 
 #include <cstdlib>
+#include <stdexcept>
 
 #include "clblast.h"
+#include "utilities/backend.hpp"
 #include "utilities/utilities.hpp"
 
 // Shortcuts to the clblast namespace