Einar-Lopez-Altamirano: CUDA Matrix Multiplication

.editorconfig

@@ -0,0 +1,61 @@
+[*]
+cpp_indent_braces=false
+cpp_indent_multi_line_relative_to=innermost_parenthesis
+cpp_indent_within_parentheses=indent
+cpp_indent_preserve_within_parentheses=false
+cpp_indent_case_labels=false
+cpp_indent_case_contents=true
+cpp_indent_case_contents_when_block=false
+cpp_indent_lambda_braces_when_parameter=true
+cpp_indent_goto_labels=one_left
+cpp_indent_preprocessor=leftmost_column
+cpp_indent_access_specifiers=false
+cpp_indent_namespace_contents=true
+cpp_indent_preserve_comments=false
+# Changed from ignore to remove
+cpp_new_line_before_open_brace_namespace=remove
+cpp_new_line_before_open_brace_type=remove
+cpp_new_line_before_open_brace_function=remove
+cpp_new_line_before_open_brace_block=remove
+cpp_new_line_before_open_brace_lambda=remove
+# ---
+cpp_new_line_scope_braces_on_separate_lines=false
+cpp_new_line_close_brace_same_line_empty_type=false
+cpp_new_line_close_brace_same_line_empty_function=false
+cpp_new_line_before_catch=true
+cpp_new_line_before_else=true
+cpp_new_line_before_while_in_do_while=false
+cpp_space_before_function_open_parenthesis=remove
+cpp_space_within_parameter_list_parentheses=false
+cpp_space_between_empty_parameter_list_parentheses=false
+cpp_space_after_keywords_in_control_flow_statements=true
+cpp_space_within_control_flow_statement_parentheses=false
+cpp_space_before_lambda_open_parenthesis=false
+cpp_space_within_cast_parentheses=false
+cpp_space_after_cast_close_parenthesis=false
+cpp_space_within_expression_parentheses=false
+cpp_space_before_block_open_brace=true
+cpp_space_between_empty_braces=false
+cpp_space_before_initializer_list_open_brace=false
+cpp_space_within_initializer_list_braces=true
+cpp_space_preserve_in_initializer_list=true
+cpp_space_before_open_square_bracket=false
+cpp_space_within_square_brackets=false
+cpp_space_before_empty_square_brackets=false
+cpp_space_between_empty_square_brackets=false
+cpp_space_group_square_brackets=true
+cpp_space_within_lambda_brackets=false
+cpp_space_between_empty_lambda_brackets=false
+cpp_space_before_comma=false
+cpp_space_after_comma=true
+cpp_space_remove_around_member_operators=true
+cpp_space_before_inheritance_colon=true
+cpp_space_before_constructor_colon=true
+cpp_space_remove_before_semicolon=true
+cpp_space_after_semicolon=false
+cpp_space_remove_around_unary_operator=true
+cpp_space_around_binary_operator=insert
+cpp_space_around_assignment_operator=insert
+cpp_space_pointer_reference_alignment=left
+cpp_space_around_ternary_operator=insert
+cpp_wrap_preserve_blocks=one_liners

CMakeLists.txt

@@ -1,4 +1,4 @@
-cmake_minimum_required(VERSION 3.18)
+cmake_minimum_required(VERSION 3.15)
 project(app LANGUAGES CXX CUDA)
 
 # Set C++ and CUDA standards

README.md

@@ -205,3 +205,34 @@ git push origin student-name
 
 Good luck, and enjoy accelerating matrix multiplication with CUDA!
 
+# Steps to run the code
+1. Copy files to Dione
+2. Access Dione
+3. Run the following commands
+
+```bash
+module load cuda
+module load GCC
+module load cmake
+nvcc -arch=sm_70 main.cu -o main -lm
+srun -p gpu --mem=10G -t 1:00:00 ./main <test_case>
+```
+
+For both naive and tiled matmul a threadsPerBlock of 16 by 16 was used.
+For tiled matmul a tile size of 16 was used.
+File transfer time was considered for both performance measurements. 
+
+# Results
+| Test Case | Dimensions ($m \times n \times p$) | Naive CPU (s) | Blocked CPU (s) | Parallel CPU (s) | Naive CUDA (s) | Tiled CUDA (s) | Tiled CUDA Speedup (vs. Naive CUDA) | Tiled CUDA Speedup (vs. Parallel CPU) |
+| :--- | :--- | :--- | :--- | :--- | :--- | :--- | :--- | :--- |
+| Case 0 | 64 × 64 × 64 | 0.001092 | 0.001221 | 0.000386 | 0.000104448 | 0.000069568 | 1.50x | 5.55x |
+| Case 1 | 128 × 64 × 128 | 0.003803 | 0.004292 | 0.001009 | 0.000164928 | 0.000141472 | 1.17x | 7.13x |
+| Case 2 | 100 × 128 × 56 | 0.002663 | 0.002945 | 0.000933999 | 0.000132320 | 0.000114848 | 1.15x | 8.13x |
+| Case 3 | 128 × 64 × 128 | 0.00387 | 0.00427 | 0.001542 | 0.000165760 | 0.000124928 | 1.33x | 12.34x |
+| Case 4 | 32 × 128 × 32 | 0.000464 | 0.000565 | 0.00024 | 0.000093344 | 0.000080416 | 1.16x | 2.98x |
+| Case 5 | 200 × 100 × 256 | 0.018409 | 0.021625 | 0.004873 | 0.000314528 | 0.000287104 | 1.10x | 16.97x |
+| Case 6 | 256 × 256 × 256 | 0.059081 | 0.068912 | 0.014825 | 0.000478688 | 0.000484800 | 0.99x | 30.58x |
+| Case 7 | 256 × 300 × 256 | 0.070174 | 0.082439 | 0.019428 | 0.000489696 | 0.000509376 | 0.96x | 38.14x |
+| Case 8 | 64 × 128 × 64 | 0.001954 | 0.002169 | 0.000576 | 0.000123552 | 0.000092864 | 1.33x | 6.20x |
+| Case 9 | 256 × 256 × 257 | 0.059597 | 0.06816 | 0.013344 | 0.000472896 | 0.000455456 | 1.04x | 29.30x |
+