Upgrade to hcc 2.0

.clang-tidy

@@ -5,6 +5,10 @@ CheckOptions:
     value:           risky   
   - key:             modernize-loop-convert.NamingStyle
     value:           lower_case
+  - key:             performance-unnecessary-copy-initialization.AllowedTypes
+    value:           'shape'
+  - key:             performance-unnecessary-value-param.AllowedTypes
+    value:           'shape'
   - key:             readability-function-size.BranchThreshold
     value:           '15'
   - key:             readability-function-size.LineThreshold

CMakeLists.txt

@@ -52,7 +52,9 @@ rocm_enable_clang_tidy(
         -clang-analyzer-optin.performance.Padding
         -clang-diagnostic-deprecated-declarations
         -clang-diagnostic-extern-c-compat
+        -clang-diagnostic-disabled-macro-expansion
         -clang-diagnostic-unused-command-line-argument
+        -cppcoreguidelines-macro-usage
         -cppcoreguidelines-pro-bounds-array-to-pointer-decay
         -cppcoreguidelines-pro-bounds-constant-array-index
         -cppcoreguidelines-pro-bounds-pointer-arithmetic
@@ -70,13 +72,12 @@ rocm_enable_clang_tidy(
         -hicpp-explicit-conversions
         -hicpp-no-array-decay
         -hicpp-special-member-functions
+        -hicpp-uppercase-literal-suffix
         -hicpp-use-override
         # This check is broken
-        -hicpp-use-auto
         -llvm-header-guard
         -llvm-include-order
         -misc-macro-parentheses
-        -modernize-use-auto
         -modernize-use-override
         -modernize-pass-by-value
         -modernize-use-default-member-init
@@ -84,7 +85,12 @@ rocm_enable_clang_tidy(
         -readability-braces-around-statements
         -readability-else-after-return
         -readability-named-parameter
+        -readability-uppercase-literal-suffix,
+        -*-avoid-c-arrays
         -*-explicit-constructor
+        -*-magic-numbers
+        -*-non-private-member-variables-in-classes
+        -*-use-auto
         -*-use-emplace
         -*-use-equals-default
     ERRORS
@@ -94,6 +100,7 @@ rocm_enable_clang_tidy(
         ".*hpp"
     EXTRA_ARGS
         -DMIGRAPHX_USE_CLANG_TIDY
+        "-Dmain\\\\(...\\\\)=main\\\\(__VA_ARGS__\\\\) // NOLINT"
 
 )
 include(ROCMCppCheck)

src/constant_propagate.cpp

@@ -13,7 +13,7 @@ struct match_const_add
         return match::name("add")(match::args(match::name("@literal"), match::name("@literal")));
     }
 
-    void apply(program& p, match::matcher_result r) const
+    void apply(program& p, const match::matcher_result& r) const
     {
         auto ins  = r.result;
         auto arg1 = ins->inputs().at(0)->get_literal();

src/include/migraphx/make_shared_array.hpp

@@ -10,7 +10,7 @@ inline namespace MIGRAPHX_INLINE_NS {
 template <typename T>
 std::shared_ptr<T> make_shared_array(size_t size)
 {
-    return std::shared_ptr<T>(new T[size], std::default_delete<T[]>());
+    return std::shared_ptr<T>(new T[size], std::default_delete<T[]>()); // NOLINT
 }
 
 } // namespace MIGRAPHX_INLINE_NS

src/include/migraphx/operators.hpp

@@ -289,7 +289,7 @@ struct transpose
         }
         std::vector<size_t> output_lens(input_lens.size());
         std::vector<size_t> output_strides(input_lens.size());
-        for(int i = 0; i < output_lens.size(); i++)
+        for(std::size_t i = 0; i < output_lens.size(); i++)
         {
             output_lens[i]    = input_lens[dims[i]];
             output_strides[i] = input_strides[dims[i]];
@@ -326,7 +326,7 @@ struct concat
     std::size_t axis = 0;
     std::string name() const { return "concat"; }
     std::vector<std::size_t> compute_offsets(const shape& output_shape,
-                                             const std::vector<argument> args) const
+                                             const std::vector<argument>& args) const
     {
         std::vector<std::size_t> offsets;
         std::vector<std::size_t> offset(args[0].get_shape().lens().size(), 0);

src/include/migraphx/tensor_view.hpp

@@ -164,7 +164,7 @@ bool operator!=(const tensor_view<T>& x, const tensor_view<U>& y)
 }
 
 template <class T>
-tensor_view<T> make_view(shape s, T* data)
+tensor_view<T> make_view(const shape& s, T* data)
 {
     return {s, data};
 }

src/include/migraphx/type_name.hpp

@@ -18,7 +18,7 @@ const std::string& get_type_name()
         name = typeid(PrivateMigraphTypeNameProbe).name();
         name = name.substr(7);
 #else
-        const char parameter_name[] = "PrivateMigraphTypeNameProbe =";
+        const char parameter_name[] = "PrivateMigraphTypeNameProbe =";// NOLINT
 
         name = __PRETTY_FUNCTION__;
 

src/onnx/mnist.cpp

@@ -14,7 +14,10 @@
 
 auto reverse_int(unsigned int i)
 {
-    unsigned char c1, c2, c3, c4;
+    unsigned char c1;
+    unsigned char c2;
+    unsigned char c3;
+    unsigned char c4;
     c1 = i & 255u;
     c2 = (i >> 8u) & 255u;
     c3 = (i >> 16u) & 255u;
@@ -32,7 +35,9 @@ read_mnist_images(const std::string& full_path, int& number_of_images, int& imag
 
     if(file.is_open())
     {
-        int magic_number = 0, n_rows = 0, n_cols = 0;
+        int magic_number = 0;
+        int n_rows = 0;
+        int n_cols = 0;
 
         file.read(reinterpret_cast<char*>(&magic_number), sizeof(magic_number));
         magic_number = reverse_int(magic_number);

src/onnx/verify_onnx.cpp

@@ -116,7 +116,7 @@ void verify_reduced_program(F f, double tolerance = 80)
 {
     migraphx::program p = f();
     auto n              = std::distance(p.begin(), p.end());
-    for(int i = 0; i < n; i++)
+    for(std::size_t i = 0; i < n; i++)
     {
         verify_reduced(f, i, tolerance);
     }

src/targets/cpu/lowering.cpp

@@ -166,7 +166,8 @@ struct cpu_im2col
             const std::size_t& stride_h = op.stride[0];
             const std::size_t& stride_w = op.stride[1];
 
-            int kdiv2_h, kdiv2_w;
+            int kdiv2_h;
+            int kdiv2_w;
             kdiv2_h = kernel_h / 2;
             kdiv2_w = kernel_w / 2;
             // calculate output sizes

src/targets/gpu/abs.cpp

@@ -18,7 +18,8 @@ argument miopen_abs::compute(context& ctx,
                              const shape& output_shape,
                              const std::vector<argument>& args) const
 {
-    float alpha = 1, beta = 0;
+    float alpha = 1;
+    float beta = 0;
     auto x_desc = make_tensor(args[0].get_shape());
     auto y_desc = make_tensor(output_shape);
     miopenActivationForward(ctx.get_stream().get_miopen(),

src/targets/gpu/batchnorm.cpp

@@ -22,7 +22,8 @@ argument miopen_batch_norm_inference::compute(context& ctx,
     auto y_desc  = make_tensor(output_shape);
     auto bn_desc = make_tensor(args[3].get_shape());
 
-    float alpha = 1.0, beta = 0.0f;
+    float alpha = 1.0;
+    float beta = 0.0f;
 
     miopenBatchNormalizationForwardInference(ctx.get_stream().get_miopen(),
                                              miopenBatchNormMode_t(op.bn_mode),

src/targets/gpu/convolution.cpp

@@ -21,7 +21,8 @@ argument miopen_convolution::compute(context& ctx,
     auto w_desc = make_tensor(args[1].get_shape());
     auto y_desc = make_tensor(output_shape);
 
-    float alpha = 1, beta = 0;
+    float alpha = 1;
+    float beta = 0;
     miopenConvolutionForward(ctx.get_stream().get_miopen(),
                              &alpha,
                              x_desc.get(),

src/targets/gpu/elu.cpp

@@ -18,7 +18,8 @@ argument miopen_elu::compute(context& ctx,
                              const shape& output_shape,
                              const std::vector<argument>& args) const
 {
-    float alpha = 1, beta = 0;
+    float alpha = 1;
+    float beta = 0;
     auto x_desc = make_tensor(args[0].get_shape());
     auto y_desc = make_tensor(output_shape);
     miopenActivationForward(ctx.get_stream().get_miopen(),

src/targets/gpu/fuse_ops.cpp

@@ -265,7 +265,8 @@ struct miopen_conv_bias
     argument compute(context& ctx, const shape&, const std::vector<argument>& args) const
     {
         auto fargs  = make_fused_args();
-        float alpha = 1, beta = 0;
+        float alpha = 1;
+        float beta = 0;
         miopenSetOpArgsConvForward(fargs.get(), conv, &alpha, &beta, args[1].implicit());
         miopenSetOpArgsBiasForward(fargs.get(), bias, &alpha, &beta, args[3].implicit());
         return f.execute(ctx, fargs, args[0], args[4]);
@@ -308,7 +309,8 @@ struct miopen_conv_bias_relu
     argument compute(context& ctx, const shape&, const std::vector<argument>& args) const
     {
         auto fargs  = make_fused_args();
-        float alpha = 1, beta = 0;
+        float alpha = 1;
+        float beta = 0;
         miopenSetOpArgsConvForward(fargs.get(), conv, &alpha, &beta, args[1].implicit());
         miopenSetOpArgsBiasForward(fargs.get(), bias, &alpha, &beta, args[3].implicit());
         miopenSetOpArgsActivForward(fargs.get(), relu, &alpha, &beta, 0, 0, 0);

src/targets/gpu/hip.cpp

@@ -16,7 +16,8 @@ std::string hip_error(int error) { return hipGetErrorString(static_cast<hipError
 
 std::size_t get_available_gpu_memory()
 {
-    size_t free, total;
+    size_t free;
+    size_t total;
     auto status = hipMemGetInfo(&free, &total);
     if(status != hipSuccess)
         MIGRAPHX_THROW("Failed getting available memory: " + hip_error(status));
@@ -72,13 +73,13 @@ argument allocate_gpu(const shape& s, bool host)
     return {s, [p]() mutable { return reinterpret_cast<char*>(p.get()); }};
 }
 
-argument to_gpu(argument arg, bool host)
+argument to_gpu(const argument& arg, bool host)
 {
     auto p = share(write_to_gpu(arg.data(), arg.get_shape().bytes(), host));
     return {arg.get_shape(), [p]() mutable { return reinterpret_cast<char*>(p.get()); }};
 }
 
-argument from_gpu(argument arg)
+argument from_gpu(const argument& arg)
 {
     argument result;
     arg.visit([&](auto x) {
@@ -98,7 +99,7 @@ void set_device(std::size_t id)
 
 void gpu_sync() { hipDeviceSynchronize(); }
 
-void copy_to_gpu(argument src, argument dst)
+void copy_to_gpu(const argument& src, const argument& dst)
 {
     std::size_t src_size = src.get_shape().bytes();
     std::size_t dst_size = dst.get_shape().bytes();

src/targets/gpu/include/migraphx/gpu/hip.hpp

@@ -9,17 +9,17 @@ namespace migraphx {
 inline namespace MIGRAPHX_INLINE_NS {
 namespace gpu {
 
-migraphx::argument allocate_gpu(const migraphx::shape& s, bool host = false);
+argument allocate_gpu(const shape& s, bool host = false);
 
-migraphx::argument to_gpu(migraphx::argument arg, bool host = false);
+argument to_gpu(const argument& arg, bool host = false);
 
-migraphx::argument from_gpu(migraphx::argument arg);
+argument from_gpu(const argument& arg);
 
 void set_device(std::size_t id);
 
 void gpu_sync();
 
-void copy_to_gpu(argument src, argument dst);
+void copy_to_gpu(const argument& src, const argument& dst);
 
 struct hip_allocate
 {

src/targets/gpu/leaky_relu.cpp

@@ -18,7 +18,8 @@ argument miopen_leaky_relu::compute(context& ctx,
                                     const shape& output_shape,
                                     const std::vector<argument>& args) const
 {
-    float alpha = 1, beta = 0;
+    float alpha = 1;
+    float beta = 0;
     auto x_desc = make_tensor(args[0].get_shape());
     auto y_desc = make_tensor(output_shape);
     miopenActivationForward(ctx.get_stream().get_miopen(),

src/targets/gpu/pooling.cpp

@@ -20,7 +20,8 @@ argument miopen_pooling::compute(context& ctx,
     auto x_desc = make_tensor(args[0].get_shape());
     auto y_desc = make_tensor(output_shape);
 
-    float alpha = 1, beta = 0;
+    float alpha = 1;
+    float beta = 0;
 
     miopenPoolingForward(ctx.get_stream().get_miopen(),
                          pd.get(),

src/targets/gpu/relu.cpp

@@ -18,7 +18,8 @@ argument miopen_relu::compute(context& ctx,
                               const shape& output_shape,
                               const std::vector<argument>& args) const
 {
-    float alpha = 1, beta = 0;
+    float alpha = 1;
+    float beta = 0;
     auto x_desc = make_tensor(args[0].get_shape());
     auto y_desc = make_tensor(output_shape);
     miopenActivationForward(ctx.get_stream().get_miopen(),