Merge branch 'develop' into perk-kernel

Dockerfile

@@ -77,7 +77,7 @@ RUN cget -p $PREFIX install ccache@v4.1
 RUN cget -p /opt/cmake install kitware/cmake@v3.13.4
 
 ARG ONNXRUNTIME_REPO=https://github.com/Microsoft/onnxruntime
-ARG ONNXRUNTIME_BRANCH=master
+ARG ONNXRUNTIME_BRANCH=main
 ARG ONNXRUNTIME_COMMIT=24f1bd6156cf5968bbc76dfb0e801a9b9c56b9fc
 RUN git clone --single-branch --branch ${ONNXRUNTIME_BRANCH} --recursive ${ONNXRUNTIME_REPO} onnxruntime && \
     cd onnxruntime && \

src/targets/gpu/mlir.cpp

@@ -48,6 +48,10 @@
 #include <deque>
 #include <variant>
 
+#if defined(MLIR_MIGRAPHX_DIALECT_API_VERSION) && MLIR_MIGRAPHX_DIALECT_API_VERSION >= 2
+#define MIGRAPHX_MLIR_BARE_POINTER
+#endif
+
 namespace migraphx {
 inline namespace MIGRAPHX_INLINE_NS {
 namespace gpu {
@@ -606,9 +610,15 @@ instruction_ref insert_mlir(module& m,
                             code_object_op co,
                             const std::vector<instruction_ref>& inputs)
 {
+
     std::vector<instruction_ref> refs;
+    std::size_t last = 0;
+#ifdef MIGRAPHX_MLIR_BARE_POINTER
+    refs.reserve(inputs.size());
+    std::copy(inputs.begin(), inputs.end(), std::back_inserter(refs));
+    last = refs.size() - 1;
+#else
     refs.reserve(inputs.size() * 15);
-
     std::unordered_map<uint64_t, instruction_ref> literal_map{};
     auto get_literal = [&](uint64_t value) {
         auto fi = literal_map.find(value);
@@ -619,7 +629,6 @@ instruction_ref insert_mlir(module& m,
         return lit;
     };
 
-    std::size_t last = 0;
     for(auto input : inputs)
     {
         const size_t offset = 0;
@@ -643,6 +652,7 @@ instruction_ref insert_mlir(module& m,
                        [&](const auto& lval) { return get_literal(lval); });
         // refs.push_back(get_literal(1)); // G
     }
+#endif
     co.expected_inputs = to_shapes(refs);
     co.output_arg      = last;
     return m.insert_instruction(ins, co, refs);

test/onnx/onnx_test.cpp

@@ -954,13 +954,11 @@ TEST_CASE(conv_dynamic_img_same_upper)
 
 TEST_CASE(conv_dynamic_kernel_same_lower)
 {
-    std::cout << "here1\n";
     migraphx::program p;
     auto* mm = p.get_main_module();
     auto l0  = mm->add_parameter("0", {migraphx::shape::float_type, {1, 3, 5, 5}});
     auto l1  = mm->add_parameter(
         "1", {migraphx::shape::float_type, {{1, 1, 0}, {3, 3, 0}, {2, 4, 0}, {2, 4, 0}}});
-    std::cout << "here2\n";
     auto c0 = mm->add_instruction(
         migraphx::make_op("convolution",
                           {{"padding", {0, 0}},
@@ -970,12 +968,10 @@ TEST_CASE(conv_dynamic_kernel_same_lower)
                            {"use_dynamic_same_auto_pad", true}}),
         l0,
         l1);
-    std::cout << "here3\n";
     mm->add_return({c0});
 
     migraphx::onnx_options options;
     options.default_dyn_dim_value = {2, 4, 0};
-    std::cout << "here\n";
     auto prog = migraphx::parse_onnx("conv_dynamic_kernel_same_lower_test.onnx", options);
     EXPECT(p == prog);
 }