carm-roofline 1.0.0.dev2__tar.gz → 1.0.0.dev3__tar.gz

{carm_roofline-1.0.0.dev2 → carm_roofline-1.0.0.dev3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: carm-roofline
-Version: 1.0.0.dev2
+Version: 1.0.0.dev3
 Summary: CARM: Cache-Aware Roofline Model benchmarking and visualization toolkit
 Author: CHaMP Team
 License:                   GNU LESSER GENERAL PUBLIC LICENSE
@@ -525,9 +525,10 @@ Description-Content-Type: text/markdown
 License-File: LICENSE
 Requires-Dist: rich>=14.3.2
 Requires-Dist: rich-argparse>=1.7.0
-Requires-Dist: numpy>=1.24.4
+Requires-Dist: numpy>=2.0.0
 Requires-Dist: matplotlib>=3.7.2
 Requires-Dist: tomli>=2.0.0
+Requires-Dist: platformdirs>=4.0.0
 Provides-Extra: gui
 Requires-Dist: dash[diskcache]; extra == "gui"
 Requires-Dist: dash-bootstrap-components; extra == "gui"
@@ -545,42 +546,45 @@ Requires-Dist: carm-roofline[dev,gui]; extra == "all"
 Dynamic: license-file
 
 
-# CARM Roofline
+# The CARM Tool
 
-CARM is a Cache-Aware Roofline Model toolkit for CPU benchmarking and visualization.
+
+Cache-Aware Roofline Model (CARM) toolkit - benchmark, visualize and profile Intel, AMD, ARM, RISC-V CPUs and NVIDIA, AMD GPUs.
 
 The package provides:
 
-- automatic microbenchmark generation for supported ISAs
+- automatic benchmarking for supported ISAs
 - arithmetic, memory, and roofline benchmark modes
-- CLI output in table, JSON, CSV, and plot formats
 - optional web GUI for interactive result visualization
 
+> [!WARNING]
+> The development branch and associated python package are still experimental, and do not include all features yet. For a stable release, please refer to the main branch.
+
 ## Installation
 
 Install from PyPI:
 
 ```bash
+# command-line interface only:
 pip install carm-roofline
+# with GUI extras:
+pip install carm-roofline[gui]
+# with all extras (GUI + development tools):
+pip install carm-roofline[all]
 ```
 
-Install with optional GUI dependencies:
-
-```bash
-pip install "carm-roofline[gui]"
-```
-
-Install all extras (GUI + development tools):
+To avoid conflicts with other packages, we recommend installing in a virtual environment:
 
 ```bash
-pip install "carm-roofline[all]"
+python -m venv carm-env
+source carm-env/bin/activate
+pip install carm-roofline[all]
 ```
 
 ## Requirements
 
-- Python 3.10+
-- Linux (POSIX)
-- GCC toolchain (for compiling benchmark binaries)
+- Python 3.9+
+- GCC (for compiling benchmark binaries)
 
 ## Command-Line Interface
 
@@ -594,7 +598,7 @@ Available subcommands:
 
 - `benchmark` - run benchmark suites
 - `gui` - launch the results dashboard
-- `profile` - reserved (not implemented yet)
+- `profile` - coming soon
 
 Show help:
 
@@ -627,12 +631,12 @@ carm benchmark --dry-run --test arithmetic --test-time 1 --verbose 4
 
 ## Output and Results
 
-By default, results are written under `carm_results`.
+By default, results are written under the platform user data directory for the `carm` app (via platformdirs). On linux, this is typically `~/.local/share/carm/`, on Windows, `%APPDATA%\carm\`, and on MacOS, `~/Library/Application Support/carm/`.
 
 Control output directory and format(s):
 
 ```bash
-carm benchmark --output-dir carm_results --output-fmt table json csv plot
+carm benchmark --output-dir ./carm_results --output-fmt table json csv plot
 ```
 
 Use `--name` to label result files:
@@ -652,12 +656,12 @@ carm gui
 Custom results location and port:
 
 ```bash
-carm gui --results-dir carm_results --gui-port 8050
+carm gui --results-dir ./carm_results --gui-port 8050
 ```
 
 ## Publications and Citation
 
-If you use CARM in papers or reports, please cite:
+If you use the CARM and the CARM Tool in papers or reports, please cite:
 
 <p>
   <a href="https://doi.org/10.1109/L-CA.2013.6" alt="Publication">

{carm_roofline-1.0.0.dev2 → carm_roofline-1.0.0.dev3}/README.md

@@ -1,40 +1,43 @@
 
-# CARM Roofline
+# The CARM Tool
 
-CARM is a Cache-Aware Roofline Model toolkit for CPU benchmarking and visualization.
+
+Cache-Aware Roofline Model (CARM) toolkit - benchmark, visualize and profile Intel, AMD, ARM, RISC-V CPUs and NVIDIA, AMD GPUs.
 
 The package provides:
 
-- automatic microbenchmark generation for supported ISAs
+- automatic benchmarking for supported ISAs
 - arithmetic, memory, and roofline benchmark modes
-- CLI output in table, JSON, CSV, and plot formats
 - optional web GUI for interactive result visualization
 
+> [!WARNING]
+> The development branch and associated python package are still experimental, and do not include all features yet. For a stable release, please refer to the main branch.
+
 ## Installation
 
 Install from PyPI:
 
 ```bash
+# command-line interface only:
 pip install carm-roofline
+# with GUI extras:
+pip install carm-roofline[gui]
+# with all extras (GUI + development tools):
+pip install carm-roofline[all]
 ```
 
-Install with optional GUI dependencies:
-
-```bash
-pip install "carm-roofline[gui]"
-```
-
-Install all extras (GUI + development tools):
+To avoid conflicts with other packages, we recommend installing in a virtual environment:
 
 ```bash
-pip install "carm-roofline[all]"
+python -m venv carm-env
+source carm-env/bin/activate
+pip install carm-roofline[all]
 ```
 
 ## Requirements
 
-- Python 3.10+
-- Linux (POSIX)
-- GCC toolchain (for compiling benchmark binaries)
+- Python 3.9+
+- GCC (for compiling benchmark binaries)
 
 ## Command-Line Interface
 
@@ -48,7 +51,7 @@ Available subcommands:
 
 - `benchmark` - run benchmark suites
 - `gui` - launch the results dashboard
-- `profile` - reserved (not implemented yet)
+- `profile` - coming soon
 
 Show help:
 
@@ -81,12 +84,12 @@ carm benchmark --dry-run --test arithmetic --test-time 1 --verbose 4
 
 ## Output and Results
 
-By default, results are written under `carm_results`.
+By default, results are written under the platform user data directory for the `carm` app (via platformdirs). On linux, this is typically `~/.local/share/carm/`, on Windows, `%APPDATA%\carm\`, and on MacOS, `~/Library/Application Support/carm/`.
 
 Control output directory and format(s):
 
 ```bash
-carm benchmark --output-dir carm_results --output-fmt table json csv plot
+carm benchmark --output-dir ./carm_results --output-fmt table json csv plot
 ```
 
 Use `--name` to label result files:
@@ -106,12 +109,12 @@ carm gui
 Custom results location and port:
 
 ```bash
-carm gui --results-dir carm_results --gui-port 8050
+carm gui --results-dir ./carm_results --gui-port 8050
 ```
 
 ## Publications and Citation
 
-If you use CARM in papers or reports, please cite:
+If you use the CARM and the CARM Tool in papers or reports, please cite:
 
 <p>
   <a href="https://doi.org/10.1109/L-CA.2013.6" alt="Publication">

{carm_roofline-1.0.0.dev2 → carm_roofline-1.0.0.dev3}/benchmark/suites/memory_sweep.py

@@ -16,7 +16,7 @@ from output_utils import debug
 from units import Bandwidth, Bytes
 
 from ..benchmark import MemoryBenchmark
-from ..generation import MemoryBenchmarkParams
+from ..generation import MemoryBenchmarkParams, MemoryLayoutMode
 from .base import ISABenchmarkSuite
 
 if TYPE_CHECKING:
@@ -92,8 +92,7 @@ class MemorySweepBenchmarkSuite(ISABenchmarkSuite):
             (level_num, level_info.name)
             for level_num, level_info in zip(
                 topology.available_cache_levels(),
-                topology,
-                strict=True,
+                iter(topology),
             )
         ]
 
@@ -159,13 +158,15 @@ class MemorySweepBenchmarkSuite(ISABenchmarkSuite):
             size_per_thread = Bytes(int(size_bytes))
 
             # Classify: find the smallest cache level whose per-thread capacity
-            # is ≥ the sweep size.  Fall back to DRAM label if none matches.
+            # is >= the sweep size. Fall back to DRAM label if none matches.
             assigned_level_name: str = dram_level_name
             assigned_affinity: CacheAwareThreadAffinity = dram_affinity
             for _lvl_num, lvl_name, avail, aff in cache_levels_for_classification:
                 if size_per_thread <= avail:
                     assigned_level_name = lvl_name
                     assigned_affinity = aff
+                    # L1 can suffer from 4k aliasing - use single-array layout to prevent speculative forwarding issues
+                    mem_layout = MemoryLayoutMode.single if _lvl_num == 1 else MemoryLayoutMode.split
                     break
 
             # Use sweep{i:02d} as memory_level_name to guarantee unique C
@@ -178,6 +179,7 @@ class MemorySweepBenchmarkSuite(ISABenchmarkSuite):
                 load_store_ratio=benchmark.ld_st_ratio,
                 size_per_thread=size_per_thread,
                 memory_level_name=sweep_label,
+                layout_mode=mem_layout,
             )
 
             bench_spec = isa.generate_memory(params, context)

{carm_roofline-1.0.0.dev2 → carm_roofline-1.0.0.dev3}/carm_roofline.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: carm-roofline
-Version: 1.0.0.dev2
+Version: 1.0.0.dev3
 Summary: CARM: Cache-Aware Roofline Model benchmarking and visualization toolkit
 Author: CHaMP Team
 License:                   GNU LESSER GENERAL PUBLIC LICENSE
@@ -525,9 +525,10 @@ Description-Content-Type: text/markdown
 License-File: LICENSE
 Requires-Dist: rich>=14.3.2
 Requires-Dist: rich-argparse>=1.7.0
-Requires-Dist: numpy>=1.24.4
+Requires-Dist: numpy>=2.0.0
 Requires-Dist: matplotlib>=3.7.2
 Requires-Dist: tomli>=2.0.0
+Requires-Dist: platformdirs>=4.0.0
 Provides-Extra: gui
 Requires-Dist: dash[diskcache]; extra == "gui"
 Requires-Dist: dash-bootstrap-components; extra == "gui"
@@ -545,42 +546,45 @@ Requires-Dist: carm-roofline[dev,gui]; extra == "all"
 Dynamic: license-file
 
 
-# CARM Roofline
+# The CARM Tool
 
-CARM is a Cache-Aware Roofline Model toolkit for CPU benchmarking and visualization.
+
+Cache-Aware Roofline Model (CARM) toolkit - benchmark, visualize and profile Intel, AMD, ARM, RISC-V CPUs and NVIDIA, AMD GPUs.
 
 The package provides:
 
-- automatic microbenchmark generation for supported ISAs
+- automatic benchmarking for supported ISAs
 - arithmetic, memory, and roofline benchmark modes
-- CLI output in table, JSON, CSV, and plot formats
 - optional web GUI for interactive result visualization
 
+> [!WARNING]
+> The development branch and associated python package are still experimental, and do not include all features yet. For a stable release, please refer to the main branch.
+
 ## Installation
 
 Install from PyPI:
 
 ```bash
+# command-line interface only:
 pip install carm-roofline
+# with GUI extras:
+pip install carm-roofline[gui]
+# with all extras (GUI + development tools):
+pip install carm-roofline[all]
 ```
 
-Install with optional GUI dependencies:
-
-```bash
-pip install "carm-roofline[gui]"
-```
-
-Install all extras (GUI + development tools):
+To avoid conflicts with other packages, we recommend installing in a virtual environment:
 
 ```bash
-pip install "carm-roofline[all]"
+python -m venv carm-env
+source carm-env/bin/activate
+pip install carm-roofline[all]
 ```
 
 ## Requirements
 
-- Python 3.10+
-- Linux (POSIX)
-- GCC toolchain (for compiling benchmark binaries)
+- Python 3.9+
+- GCC (for compiling benchmark binaries)
 
 ## Command-Line Interface
 
@@ -594,7 +598,7 @@ Available subcommands:
 
 - `benchmark` - run benchmark suites
 - `gui` - launch the results dashboard
-- `profile` - reserved (not implemented yet)
+- `profile` - coming soon
 
 Show help:
 
@@ -627,12 +631,12 @@ carm benchmark --dry-run --test arithmetic --test-time 1 --verbose 4
 
 ## Output and Results
 
-By default, results are written under `carm_results`.
+By default, results are written under the platform user data directory for the `carm` app (via platformdirs). On linux, this is typically `~/.local/share/carm/`, on Windows, `%APPDATA%\carm\`, and on MacOS, `~/Library/Application Support/carm/`.
 
 Control output directory and format(s):
 
 ```bash
-carm benchmark --output-dir carm_results --output-fmt table json csv plot
+carm benchmark --output-dir ./carm_results --output-fmt table json csv plot
 ```
 
 Use `--name` to label result files:
@@ -652,12 +656,12 @@ carm gui
 Custom results location and port:
 
 ```bash
-carm gui --results-dir carm_results --gui-port 8050
+carm gui --results-dir ./carm_results --gui-port 8050
 ```
 
 ## Publications and Citation
 
-If you use CARM in papers or reports, please cite:
+If you use the CARM and the CARM Tool in papers or reports, please cite:
 
 <p>
   <a href="https://doi.org/10.1109/L-CA.2013.6" alt="Publication">

{carm_roofline-1.0.0.dev2 → carm_roofline-1.0.0.dev3}/carm_roofline.egg-info/SOURCES.txt

@@ -1,8 +1,5 @@
-DBI_AI_Calculator.py
 LICENSE
-PMU_AI_Calculator.py
 README.md
-SDE_AI_Calculator.py
 arguments.py
 carm.py
 context.py
@@ -10,6 +7,7 @@ error.py
 exec_interface.py
 output_utils.py
 pyproject.toml
+results_paths.py
 run_config.py
 units.py
 utils.py
@@ -22,8 +20,6 @@ architecture/frequency.py
 architecture/memory.py
 architecture/riscv.py
 architecture/x86.py
-architecture/tests/REFACTOR.md
-architecture/tests/frequency
 architecture/tests/frequency.c
 architecture/tests/arm/features.c
 architecture/tests/arm/frequency.h
@@ -31,9 +27,7 @@ architecture/tests/arm/vlen.c
 architecture/tests/riscv/frequency.h
 architecture/tests/riscv/rvv_version.c
 architecture/tests/riscv/vlen.c
-architecture/tests/x86/cache
 architecture/tests/x86/cache.c
-architecture/tests/x86/features
 architecture/tests/x86/features.c
 architecture/tests/x86/frequency.h
 architecture/tests/x86/x86_avx512/frequency.h

{carm_roofline-1.0.0.dev2 → carm_roofline-1.0.0.dev3}/carm_roofline.egg-info/requires.txt

@@ -1,8 +1,9 @@
 rich>=14.3.2
 rich-argparse>=1.7.0
-numpy>=1.24.4
+numpy>=2.0.0
 matplotlib>=3.7.2
 tomli>=2.0.0
+platformdirs>=4.0.0
 
 [all]
 carm-roofline[dev,gui]

{carm_roofline-1.0.0.dev2 → carm_roofline-1.0.0.dev3}/carm_roofline.egg-info/top_level.txt

@@ -7,6 +7,7 @@ error
 exec_interface
 gui
 output_utils
+results_paths
 run_config
 test_bench
 units

{carm_roofline-1.0.0.dev2 → carm_roofline-1.0.0.dev3}/gui/config.py

@@ -4,6 +4,7 @@ import argparse
 from pathlib import Path
 
 from arguments import InsertsArguments
+from results_paths import default_results_root
 
 
 class GUIConfig(InsertsArguments):
@@ -21,8 +22,8 @@ class GUIConfig(InsertsArguments):
         parser.add_argument(
             "--results-dir",
             type=Path,
-            default=Path("carm_results"),
-            help="Results directory root (default: carm_results, relative to current working directory)",
+            default=default_results_root(),
+            help="Results directory root (default: <user-data-directory>/carm)",
         )
         parser.add_argument("--gui-host", default="127.0.0.1", help="Host address for the Dash server")
         parser.add_argument("--gui-port", type=int, default=8050, help="Port for the Dash server")

{carm_roofline-1.0.0.dev2 → carm_roofline-1.0.0.dev3}/gui/dashboard.py

@@ -25,9 +25,11 @@ import plotly.graph_objects as go
 from dash import ALL, DiskcacheManager, Input, Output, State, callback_context, dcc, html, no_update
 from dash.exceptions import PreventUpdate
 
-# Local Python Scripts
 import utils as ut
 
+# Local Python Scripts
+from results_paths import default_results_root
+
 from . import gui_utils as gut
 
 log = logging.getLogger("werkzeug")
@@ -40,7 +42,7 @@ cache = diskcache.Cache(str(cache_dir))
 background_callback_manager = DiskcacheManager(cache)
 
 
-DEFAULT_RESULTS_ROOT = (Path.cwd() / "carm_results").resolve()
+DEFAULT_RESULTS_ROOT = default_results_root()
 RESULTS_ROOT = DEFAULT_RESULTS_ROOT
 roof_results_path = str((RESULTS_ROOT / "roofline").resolve())
 csv_files = []
@@ -2423,6 +2425,18 @@ _register_filter_dropdown_callbacks("")
 _register_filter_dropdown_callbacks("2")
 
 
+@app.callback(
+    [
+        *[Output(f"{field_id}-dynamic-dropdown", "value") for field_id, _ in FILTER_DROPDOWN_FIELDS],
+        *[Output(f"{field_id}-dynamic-dropdown2", "value") for field_id, _ in FILTER_DROPDOWN_FIELDS],
+    ],
+    Input("filename", "value"),
+    prevent_initial_call=True,
+)
+def reset_filter_dropdown_values(_selected_file: str | None) -> list[None]:
+    return [None] * (len(FILTER_DROPDOWN_FIELDS) * 2)
+
+
 @app.callback(
     [
         Output(component_id="graphs", component_property="figure"),
@@ -2538,7 +2552,6 @@ def analysis(
     min_gflops2 = 1e99
     change_annotation = 0
 
-    # Get queries for both sets of inputs
     query1 = construct_query(isa, precision, threads, loads, stores, interleaved, dram_bytes, fp_inst, date)
     query2 = construct_query(isa2, precision2, threads2, loads2, stores2, interleaved2, dram_bytes2, fp_inst2, date2)
 

{carm_roofline-1.0.0.dev2 → carm_roofline-1.0.0.dev3}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "carm-roofline"
-version = "1.0.0.dev2"
+version = "1.0.0.dev3"
 description = "CARM: Cache-Aware Roofline Model benchmarking and visualization toolkit"
 readme = "README.md"
 requires-python = ">=3.9"
@@ -23,9 +23,10 @@ classifiers = [
 dependencies = [
     "rich>=14.3.2",
     "rich-argparse>=1.7.0",
-    "numpy>=1.24.4",
+    "numpy>=2.0.0",
     "matplotlib>=3.7.2",
     "tomli>=2.0.0",
+    "platformdirs>=4.0.0",
 ]
 
 [project.optional-dependencies]
@@ -78,6 +79,7 @@ py-modules = [
     "error",
     "exec_interface",
     "output_utils",
+    "results_paths",
     "run_config",
     "units",
     "utils",
@@ -195,7 +197,7 @@ markers = [
 # ============================================================================
 
 [tool.mypy]
-python_version = "3.10"
+python_version = "3.9"
 strict = true
 strict_equality = true
 show_error_codes = true

carm_roofline-1.0.0.dev3/results_paths.py

@@ -0,0 +1,13 @@
+from __future__ import annotations
+
+from pathlib import Path
+
+from platformdirs import user_data_dir
+
+
+def user_data_dir_for_results() -> Path:
+    return Path(user_data_dir("carm", appauthor=None)).expanduser()
+
+
+def default_results_root() -> Path:
+    return user_data_dir_for_results()

{carm_roofline-1.0.0.dev2 → carm_roofline-1.0.0.dev3}/run_config.py

@@ -6,6 +6,7 @@ from pathlib import Path
 from arguments import InsertsArguments, enum_action
 from benchmark.output.base import OutputKind
 from output_utils import warn
+from results_paths import default_results_root
 
 
 class RunConfig(InsertsArguments):
@@ -39,15 +40,15 @@ class RunConfig(InsertsArguments):
         )
         parser.add_argument(
             "--output-dir",
-            default=Path("carm_results"),
+            default=default_results_root(),
             nargs="?",
             type=Path,
-            help="Directory to write result files (default: carm_results)",
+            help="Directory to write result files (default: platform user data dir for app 'carm')",
         )
         parser.add_argument(
             "--output-fmt",
             nargs="+",
             action=enum_action(OutputKind),
-            default={OutputKind.TABLE, OutputKind.PLOT, OutputKind.JSON},
-            help="Output format(s): table, plot, json, csv (default: table plot json)",
+            default={OutputKind.TABLE, OutputKind.PLOT, OutputKind.JSON, OutputKind.CSV},
+            help="Output format(s): table, plot, json, csv (default: table plot json csv)",
         )

{carm_roofline-1.0.0.dev2 → carm_roofline-1.0.0.dev3}/test_bench/builder.py

@@ -364,6 +364,9 @@ def run_microbenchmarks(
             error_msg += f"stderr:\n{result.stderr}"
         raise RuntimeError(error_msg)
 
+    if result.stderr:
+        debug(f"Benchmark debug output:\n{result.stderr}")
+
     detail("Microbenchmarks completed successfully")
     return str(result.stdout)