JSON support in MemoryAnalysis handler

[coleramos425]

This PR is an iteration on #21 that instead logs MemoryAnalysis output in the logduration message handler. The plugin behaves same as before in most instances, however now when user provides -t json JSON output will print to console. Additionally, the user can specify a log file (e.g. -l output.json) to save this JSON structure to an output file for further post-processing.

Example usage:

root@51a53ca4a7c7:/workspace# logdur_install/bin/logDuration/omniprobe -i -a MemoryAnalysis -t json -- audacious/maestro/examples/build/bank_conflict/b2b_matrix_transpose 
  ___                  _                 _          
 / _ \ _ __ ___  _ __ (_)_ __  _ __ ___ | |__   ___ 
| | | | '_ ` _ \| '_ \| | '_ \| '__/ _ \| '_ \ / _ \
| |_| | | | | | | | | | | |_) | | | (_) | |_) |  __/
 \___/|_| |_| |_|_| |_|_| .__/|_|  \___/|_.__/ \___|
                        |_|                         

Found config file at /workspace/logdur_install/bin/logDuration/runtime_config.txt

Omniprobe is developed by Advanced Micro Devices, Research and Advanced Development
Copyright (c) 2025 Advanced Micro Devices. All rights reserved.

No Triton cache location provided; assuming HIP run.
HANDLER: libMemAnalysis64.so
Memory Analysis Wrapper loaded.
handlerManager: OpenedlibMemAnalysis64.so
Adding /workspace/audacious/maestro/examples/build/bank_conflict/b2b_matrix_transpose
Found 12 kernels
Adding linux-vdso.so.1
Adding /opt/rocm/lib/libamdhip64.so.6
Adding /lib/x86_64-linux-gnu/libstdc++.so.6
Adding /lib/x86_64-linux-gnu/libm.so.6
Adding /lib/x86_64-linux-gnu/libgcc_s.so.1
Adding /lib/x86_64-linux-gnu/libc.so.6
Adding /opt/rocm/lib/librocprofiler-register.so.0
Adding /opt/rocm/lib/libamd_comgr.so.2
Adding /opt/rocm/lib/libhsa-runtime64.so.1
Adding /lib/x86_64-linux-gnu/libnuma.so.1
Adding /lib64/ld-linux-x86-64.so.2
Adding /lib/x86_64-linux-gnu/libz.so.1
Adding /lib/x86_64-linux-gnu/libzstd.so.1
Adding /lib/x86_64-linux-gnu/libelf.so.1
Adding /opt/amdgpu/lib/x86_64-linux-gnu/libdrm.so.2
Adding /opt/amdgpu/lib/x86_64-linux-gnu/libdrm_amdgpu.so.1
Adding /workspace/logdur_install/lib/logDuration/liblogDuration64.so
Adding /workspace/logdur_install/lib/libdh_comms.so
Adding /workspace/logdur_install/lib/libkernelDB64.so.1
Adding /lib/x86_64-linux-gnu/libdwarf.so.1
Adding /workspace/logdur_install/lib/libMemAnalysis64.so
Adding /workspace/audacious/maestro/examples/build/bank_conflict/b2b_matrix_transpose
[...Removed for brevity...]
>>>>>>>> HSA intercept registered.
{
  "kernel_analyses": [
    {
      "kernel_info": {
        "name": "matrixTransposeShared_0(float*, float const*, int, int) [clone .kd]",
        "dispatch_id": 1
      },
      "cache_analysis": {
        "accesses": [
          {
            "source_location": {
              "file": "/workspace/audacious/maestro/examples/bank_conflict/b2b_matrix_transpose/b2b_matrix_transpose.hip",
              "line": 40,
              "column": 38
            },
            "code_context": "tile[threadIdx.y][threadIdx.x] = in[y * width + x];",
            "access_info": {
              "type": "read",
              "execution_count": 16384,
              "ir_bytes": 4,
              "isa_bytes": 4,
              "isa_instruction": "global_load_dword",
              "cache_lines": {
                "needed": 65536,
                "used": 65536
              }
            }
          },
          {
            "source_location": {
              "file": "/workspace/audacious/maestro/examples/bank_conflict/b2b_matrix_transpose/b2b_matrix_transpose.hip",
              "line": 49,
              "column": 25
            },
            "code_context": "out[y * height + x] = tile[threadIdx.x][threadIdx.y];",
            "access_info": {
              "type": "write",
              "execution_count": 16384,
              "ir_bytes": 4,
              "isa_bytes": 4,
              "isa_instruction": "global_store_dword",
              "cache_lines": {
                "needed": 65536,
                "used": 65536
              }
            }
          }
        ]
      },
      "bank_conflicts": {
        "accesses": [
          {
            "source_location": {
              "file": "/workspace/audacious/maestro/examples/bank_conflict/b2b_matrix_transpose/b2b_matrix_transpose.hip",
              "line": 40,
              "column": 36
            },
            "code_context": "tile[threadIdx.y][threadIdx.x] = in[y * width + x];",
            "access_info": {
              "type": "write",
              "execution_count": 16384,
              "ir_bytes": 4,
              "total_conflicts": 0
            }
          },
          {
            "source_location": {
              "file": "/workspace/audacious/maestro/examples/bank_conflict/b2b_matrix_transpose/b2b_matrix_transpose.hip",
              "line": 49,
              "column": 27
            },
            "code_context": "out[y * height + x] = tile[threadIdx.x][threadIdx.y];",
            "access_info": {
              "type": "read",
              "execution_count": 16384,
              "ir_bytes": 4,
              "total_conflicts": 229376
            }
          }
        ]
      }
    }
  ],
  "metadata": {
    "version": "null",
    "kernels_found": 1,
    "timestamp": "2025-06-30 17:00:30",
    "gpu_info": {
      "architecture": "gfx90a",
      "cache_line_size": 128
    }
  }
},
54525952 bytes processed in 1.878942 seconds (29.0 MiB/s)
{
  "kernel_analyses": [
    {
      "kernel_info": {
        "name": "matrixTransposeShared_1(float*, float const*, int, int) [clone .kd]",
        "dispatch_id": 2
      },
      "cache_analysis": {
        "accesses": [
          {
            "source_location": {
              "file": "/workspace/audacious/maestro/examples/bank_conflict/b2b_matrix_transpose/b2b_matrix_transpose.hip",
              "line": 63,
              "column": 38
            },
            "code_context": "tile[threadIdx.y][threadIdx.x] = in[y * width + x];",
            "access_info": {
              "type": "read",
              "execution_count": 16384,
              "ir_bytes": 4,
              "isa_bytes": 4,
              "isa_instruction": "global_load_dword",
              "cache_lines": {
                "needed": 65536,
                "used": 65536
              }
            }
          },
          {
            "source_location": {
              "file": "/workspace/audacious/maestro/examples/bank_conflict/b2b_matrix_transpose/b2b_matrix_transpose.hip",
              "line": 72,
              "column": 25
            },
            "code_context": "out[y * height + x] = tile[threadIdx.x][threadIdx.y];",
            "access_info": {
              "type": "write",
              "execution_count": 16384,
              "ir_bytes": 4,
              "isa_bytes": 4,
              "isa_instruction": "global_store_dword",
              "cache_lines": {
                "needed": 65536,
                "used": 65536
              }
            }
          }
        ]
      },
      "bank_conflicts": {
        "accesses": [
          {
            "source_location": {
              "file": "/workspace/audacious/maestro/examples/bank_conflict/b2b_matrix_transpose/b2b_matrix_transpose.hip",
              "line": 63,
              "column": 36
            },
            "code_context": "tile[threadIdx.y][threadIdx.x] = in[y * width + x];",
            "access_info": {
              "type": "write",
              "execution_count": 16384,
              "ir_bytes": 4,
              "total_conflicts": 0
            }
          },
          {
            "source_location": {
              "file": "/workspace/audacious/maestro/examples/bank_conflict/b2b_matrix_transpose/b2b_matrix_transpose.hip",
              "line": 72,
              "column": 27
            },
            "code_context": "out[y * height + x] = tile[threadIdx.x][threadIdx.y];",
            "access_info": {
              "type": "read",
              "execution_count": 16384,
              "ir_bytes": 4,
              "total_conflicts": 229376
            }
          }
        ]
      }
    }
  ],
  "metadata": {
    "version": "null",
    "kernels_found": 1,
    "timestamp": "2025-06-30 17:00:32",
    "gpu_info": {
      "architecture": "gfx90a",
      "cache_line_size": 128
    }
  }
}
54525952 bytes processed in 3.588229 seconds (15.2 MiB/s)
Memory Analysis Wrapper unloaded.
Comms Runner shutting down
Cache Watcher shutting down

[coleramos425]

Most recent commit finalizes this PR by ensuring valid JSON syntax when producing a MemoryAnalysis report for an application with several dispatches
71a9d32

The core challenge is creating a mechanism that can reliably signal when all processing is complete, allowing for final JSON output refinement without disrupting the existing plugin architecture. Implementing this logic in the omniprobe module seems to be the simplest solution