adjust readme and add clang tidy/format

.clang-format

@@ -0,0 +1,3 @@
+BasedOnStyle: LLVM
+IndentWidth: 4
+AccessModifierOffset: -4

.clang-tidy

@@ -0,0 +1,11 @@
+Checks: '-*,clang-diagnostic-*,misc-*,performance-*,clang-analyzer-*,bugprone-*'
+CheckOptions:
+  - { key: readability-identifier-naming.NamespaceCase,       value: lower_case }
+  - { key: readability-identifier-naming.ClassCase,           value: CamelCase  }
+  - { key: readability-identifier-naming.PrivateMemberPrefix, value: m_         }
+  - { key: readability-identifier-naming.StructCase,          value: CamelCase  }
+  - { key: readability-identifier-naming.FunctionCase,        value: lower_case }
+  - { key: readability-identifier-naming.VariableCase,        value: lower_case }
+  - { key: readability-identifier-naming.MemberCase,          value: lower_case }
+  - { key: readability-identifier-naming.GlobalConstantCase,  value: UPPER_CASE }
+  - { key: readability-simplify-boolean-expr,                 value: true       }

README.md

@@ -24,25 +24,22 @@
 
 # Backends
 
-- [ ] Generic (already implemented but mixed up with AVX2 backend)
-- [X] AVX2 (going to be replaced by ISPC)
-- [X] CUDA (wip, some kernels are missing)
-- [ ] SYCL ? HIP ? Vulkan ?
+- [X] ISPC
+- [X] CUDA
 
 # Maintainability
 
 - [X] Clangd
-- [ ] Clang-Tidy
+- [X] Clang-Format
+- [X] Clang-Tidy
 - [X] Testing framework (via `xmake test`)
 - [X] Compilation CI
-- [ ] Performance regression CI
 - [ ] Documented code (doxygen style)
 
 ## Build and run
 
 Requirements: 
 - Windows (MSVC compiler) or Linux (GCC or Clang)
-- CPU that supports AVX2
 - [xmake](https://xmake.io/#/) build system
     - For Windows, the [installer](https://github.com/xmake-io/xmake/releases) is recommended
     - For Linux, run `curl -fsSL https://xmake.io/shget.text | bash`

vlm/backends/cuda/src/vlm_backend_cuda.cu

@@ -97,12 +97,11 @@ public:
     cudaStream_t stream() {return(m_stream);}
     cublasHandle_t cublas() {return(m_cublas);}
 
-protected:
+private:
     cusolverDnHandle_t m_cusolver = nullptr;
     cudaStream_t m_stream = nullptr;
     cublasHandle_t m_cublas = nullptr;
-
-private:
+
     CtxManager() = default;
     ~CtxManager() = default;
 };