diff --git a/codeflash/benchmarking/plugin/plugin.py b/codeflash/benchmarking/plugin/plugin.py index 8c502d143..b5639ddf5 100644 --- a/codeflash/benchmarking/plugin/plugin.py +++ b/codeflash/benchmarking/plugin/plugin.py @@ -200,7 +200,7 @@ def get_benchmark_timings(trace_path: Path) -> dict[BenchmarkKey, int]: # Pytest hooks @pytest.hookimpl - def pytest_sessionfinish(self, session, exitstatus) -> None: # noqa: ANN001 + def pytest_sessionfinish(self, session, exitstatus) -> None: """Execute after whole test run is completed.""" # Write any remaining benchmark timings to the database codeflash_trace.close() @@ -236,20 +236,20 @@ class Benchmark: # noqa: D106 def __init__(self, request: pytest.FixtureRequest) -> None: self.request = request - def __call__(self, func, *args, **kwargs): # noqa: ANN001, ANN002, ANN003, ANN204 + def __call__(self, func, *args, **kwargs): # noqa: ANN002, ANN003, ANN204 """Handle both direct function calls and decorator usage.""" if args or kwargs: # Used as benchmark(func, *args, **kwargs) return self._run_benchmark(func, *args, **kwargs) # Used as @benchmark decorator - def wrapped_func(*args, **kwargs): # noqa: ANN002, ANN003, ANN202 + def wrapped_func(*args, **kwargs): # noqa: ANN002, ANN003 return func(*args, **kwargs) self._run_benchmark(func) return wrapped_func - def _run_benchmark(self, func, *args, **kwargs): # noqa: ANN002, ANN003, ANN202 + def _run_benchmark(self, func, *args, **kwargs): # noqa: ANN002, ANN003 """Actual benchmark implementation.""" node_path = getattr(self.request.node, "path", None) or getattr(self.request.node, "fspath", None) if node_path is None: diff --git a/codeflash/cli_cmds/cmd_init.py b/codeflash/cli_cmds/cmd_init.py index 51ca1a4f2..7a83a9971 100644 --- a/codeflash/cli_cmds/cmd_init.py +++ b/codeflash/cli_cmds/cmd_init.py @@ -1474,11 +1474,7 @@ def customize_codeflash_yaml_content( return _customize_python_workflow_content(optimize_yml_content, git_root, benchmark_mode) -def _customize_python_workflow_content( - optimize_yml_content: str, - git_root: Path, - benchmark_mode: bool = False, # noqa: FBT001, FBT002 -) -> str: +def _customize_python_workflow_content(optimize_yml_content: str, git_root: Path, benchmark_mode: bool = False) -> str: """Customize workflow content for Python projects.""" # Get dependency installation commands toml_path = Path.cwd() / "pyproject.toml" @@ -1513,11 +1509,7 @@ def _customize_python_workflow_content( # TODO:{claude} Refactor and move to support for language specific -def _customize_js_workflow_content( - optimize_yml_content: str, - git_root: Path, - benchmark_mode: bool = False, # noqa: FBT001, FBT002 -) -> str: +def _customize_js_workflow_content(optimize_yml_content: str, git_root: Path, benchmark_mode: bool = False) -> str: """Customize workflow content for JavaScript/TypeScript projects.""" from codeflash.cli_cmds.init_javascript import ( get_js_codeflash_install_step, diff --git a/codeflash/cli_cmds/init_javascript.py b/codeflash/cli_cmds/init_javascript.py index 578b56ca5..22371982a 100644 --- a/codeflash/cli_cmds/init_javascript.py +++ b/codeflash/cli_cmds/init_javascript.py @@ -66,7 +66,7 @@ class JSSetupInfo: # Import theme from cmd_init to avoid duplication -def _get_theme(): # noqa: ANN202 +def _get_theme(): """Get the CodeflashTheme - imported lazily to avoid circular imports.""" from codeflash.cli_cmds.cmd_init import CodeflashTheme diff --git a/codeflash/code_utils/code_utils.py b/codeflash/code_utils/code_utils.py index bc23e844e..9244f6b11 100644 --- a/codeflash/code_utils/code_utils.py +++ b/codeflash/code_utils/code_utils.py @@ -436,7 +436,7 @@ def extract_unique_errors(pytest_output: str) -> set[str]: pattern = r"^E\s+(.*)$" for error_message in re.findall(pattern, pytest_output, re.MULTILINE): - error_message = error_message.strip() # noqa: PLW2901 + error_message = error_message.strip() if error_message: unique_errors.add(error_message) diff --git a/codeflash/code_utils/config_js.py b/codeflash/code_utils/config_js.py index 92f635c25..80cdbe216 100644 --- a/codeflash/code_utils/config_js.py +++ b/codeflash/code_utils/config_js.py @@ -105,7 +105,7 @@ def detect_module_root(project_root: Path, package_data: dict[str, Any]) -> str: return "." -def detect_test_runner(project_root: Path, package_data: dict[str, Any]) -> str: # noqa: ARG001 +def detect_test_runner(project_root: Path, package_data: dict[str, Any]) -> str: """Detect test runner from devDependencies or scripts.test. Detection order: @@ -144,7 +144,7 @@ def detect_test_runner(project_root: Path, package_data: dict[str, Any]) -> str: return "jest" -def detect_formatter(project_root: Path, package_data: dict[str, Any]) -> list[str] | None: # noqa: ARG001 +def detect_formatter(project_root: Path, package_data: dict[str, Any]) -> list[str] | None: """Detect formatter from devDependencies. Detection order: diff --git a/codeflash/code_utils/deduplicate_code.py b/codeflash/code_utils/deduplicate_code.py index 097fbbb71..a69c52ef3 100644 --- a/codeflash/code_utils/deduplicate_code.py +++ b/codeflash/code_utils/deduplicate_code.py @@ -14,10 +14,7 @@ def normalize_code( - code: str, - remove_docstrings: bool = True, - return_ast_dump: bool = False, - language: str | None = None, + code: str, remove_docstrings: bool = True, return_ast_dump: bool = False, language: str | None = None ) -> str: """Normalize code by parsing, cleaning, and normalizing variable names. diff --git a/codeflash/code_utils/instrument_existing_tests.py b/codeflash/code_utils/instrument_existing_tests.py index 6315830ce..4366468d0 100644 --- a/codeflash/code_utils/instrument_existing_tests.py +++ b/codeflash/code_utils/instrument_existing_tests.py @@ -89,7 +89,7 @@ def find_and_update_line_node( # it's much more efficient to visit nodes manually. We'll only descend into expressions/statements. # Helper for manual walk - def iter_ast_calls(node): # noqa: ANN202 + def iter_ast_calls(node): # Generator to yield each ast.Call in test_node, preserves node identity stack = [node] while stack: @@ -102,7 +102,7 @@ def iter_ast_calls(node): # noqa: ANN202 if isinstance(value, list): for item in reversed(value): if isinstance(item, ast.AST): - stack.append(item) # noqa: PERF401 + stack.append(item) elif isinstance(value, ast.AST): stack.append(value) diff --git a/codeflash/context/code_context_extractor.py b/codeflash/context/code_context_extractor.py index 4bafc0aeb..28141dcb9 100644 --- a/codeflash/context/code_context_extractor.py +++ b/codeflash/context/code_context_extractor.py @@ -46,8 +46,8 @@ def build_testgen_context( helpers_of_fto_dict: dict[Path, set[FunctionSource]], helpers_of_helpers_dict: dict[Path, set[FunctionSource]], project_root_path: Path, - remove_docstrings: bool, # noqa: FBT001 - include_imported_classes: bool, # noqa: FBT001 + remove_docstrings: bool, + include_imported_classes: bool, ) -> CodeStringsMarkdown: """Build testgen context with optional imported class definitions and external base inits.""" testgen_context = extract_code_markdown_context_from_files( diff --git a/codeflash/languages/javascript/module_system.py b/codeflash/languages/javascript/module_system.py index 6ed9d62f0..626cc2d32 100644 --- a/codeflash/languages/javascript/module_system.py +++ b/codeflash/languages/javascript/module_system.py @@ -185,7 +185,7 @@ def _get_relative_import_path(target_path: Path, source_path: Path) -> str: def add_js_extension(module_path: str) -> str: """Add .js extension to relative module paths for ESM compatibility.""" - if module_path.startswith(("./", "../")): # noqa: SIM102 + if module_path.startswith(("./", "../")): if not module_path.endswith(".js") and not module_path.endswith(".mjs"): return module_path + ".js" return module_path diff --git a/codeflash/languages/javascript/support.py b/codeflash/languages/javascript/support.py index 3ca13c88e..86c258b52 100644 --- a/codeflash/languages/javascript/support.py +++ b/codeflash/languages/javascript/support.py @@ -1872,7 +1872,7 @@ def instrument_source_for_line_profiler( # Write instrumented code to source file source_file_path.write_text(instrumented_source, encoding="utf-8") logger.debug("Wrote instrumented source to %s", source_file_path) - return True # noqa: TRY300 + return True except Exception as e: logger.warning("Failed to instrument source for line profiling: %s", e) return False diff --git a/codeflash/lsp/features/perform_optimization.py b/codeflash/lsp/features/perform_optimization.py index 7f84b2e0e..d1f413a7a 100644 --- a/codeflash/lsp/features/perform_optimization.py +++ b/codeflash/lsp/features/perform_optimization.py @@ -51,10 +51,10 @@ def sync_perform_optimization(server: CodeflashLanguageServer, cancel_event: thr ctx_tests = contextvars.copy_context() ctx_opts = contextvars.copy_context() - def run_generate_tests(): # noqa: ANN202 + def run_generate_tests(): return function_optimizer.generate_and_instrument_tests(code_context) - def run_generate_optimizations(): # noqa: ANN202 + def run_generate_optimizations(): return function_optimizer.generate_optimizations( read_writable_code=code_context.read_writable_code, read_only_context_code=code_context.read_only_context_code, diff --git a/codeflash/lsp/lsp_logger.py b/codeflash/lsp/lsp_logger.py index 8f522ba39..eb4f2fe43 100644 --- a/codeflash/lsp/lsp_logger.py +++ b/codeflash/lsp/lsp_logger.py @@ -127,7 +127,7 @@ def enhanced_log( # Configure logging to stderr for VS Code output channel def setup_logging() -> logging.Logger: - global root_logger # noqa: PLW0603 + global root_logger if root_logger: return root_logger # Clear any existing handlers to prevent conflicts diff --git a/codeflash/telemetry/posthog_cf.py b/codeflash/telemetry/posthog_cf.py index 15df7d509..1638f1ffc 100644 --- a/codeflash/telemetry/posthog_cf.py +++ b/codeflash/telemetry/posthog_cf.py @@ -20,7 +20,7 @@ def initialize_posthog(*, enabled: bool = True) -> None: if not enabled: return - global _posthog # noqa: PLW0603 + global _posthog _posthog = Posthog(project_api_key="phc_aUO790jHd7z1SXwsYCz8dRApxueplZlZWeDSpKc5hol", host="https://us.posthog.com") _posthog.log.setLevel(logging.CRITICAL) # Suppress PostHog logging ph("cli-telemetry-enabled") diff --git a/codeflash/verification/comparator.py b/codeflash/verification/comparator.py index f92b0d000..ad7c59ede 100644 --- a/codeflash/verification/comparator.py +++ b/codeflash/verification/comparator.py @@ -234,7 +234,7 @@ def comparator(orig: Any, new: Any, superset_obj=False) -> bool: try: insp = sqlalchemy.inspection.inspect(orig) - insp = sqlalchemy.inspection.inspect(new) # noqa: F841 + insp = sqlalchemy.inspection.inspect(new) orig_keys = orig.__dict__ new_keys = new.__dict__ for key in list(orig_keys.keys()): diff --git a/codeflash/verification/equivalence.py b/codeflash/verification/equivalence.py index 0ebd48fea..f660e35ea 100644 --- a/codeflash/verification/equivalence.py +++ b/codeflash/verification/equivalence.py @@ -28,9 +28,7 @@ def safe_repr(obj: object) -> str: def compare_test_results( - original_results: TestResults, - candidate_results: TestResults, - pass_fail_only: bool = False, # noqa: FBT001, FBT002 + original_results: TestResults, candidate_results: TestResults, pass_fail_only: bool = False ) -> tuple[bool, list[TestDiff]]: # This is meant to be only called with test results for the first loop index if len(original_results) == 0 or len(candidate_results) == 0: diff --git a/codeflash/verification/pytest_plugin.py b/codeflash/verification/pytest_plugin.py index 40324dbcb..0b7144356 100644 --- a/codeflash/verification/pytest_plugin.py +++ b/codeflash/verification/pytest_plugin.py @@ -383,7 +383,7 @@ def pytest_runtestloop(self, session: Session) -> bool: count += 1 loop_start = _ORIGINAL_PERF_COUNTER_NS() for index, item in enumerate(session.items): - item: pytest.Item = item # noqa: PLW0127, PLW2901 + item: pytest.Item = item # noqa: PLW0127 item._report_sections.clear() # clear reports for new test # noqa: SLF001 if total_time > SHORTEST_AMOUNT_OF_TIME: diff --git a/codeflash/version.py b/codeflash/version.py index ec305ddad..6225467e3 100644 --- a/codeflash/version.py +++ b/codeflash/version.py @@ -1,2 +1,2 @@ # These version placeholders will be replaced by uv-dynamic-versioning during build. -__version__ = "0.20.0.post91.dev0+28f8eb18" +__version__ = "0.20.0" diff --git a/docs/codeflash-concepts/benchmarking-gpu-code.mdx b/docs/codeflash-concepts/benchmarking-gpu-code.mdx new file mode 100644 index 000000000..41d4f1d89 --- /dev/null +++ b/docs/codeflash-concepts/benchmarking-gpu-code.mdx @@ -0,0 +1,117 @@ +--- +title: "How Codeflash Measures Code Runtime on GPUs" +description: "Learn how Codeflash accurately measures code performance on GPUs" +icon: "microchip" +sidebarTitle: "GPU Benchmarking" +keywords: ["benchmarking", "performance", "timing", "measurement", "runtime", "noise reduction", "GPU", "MPS"] +--- + +## Accurate Benchmarking on GPU devices + +When a GPU (Graphics Processing Unit) operation is executed, it executes **asynchronously**. This means the CPU (Central Processing Unit) queues up work for the GPU and immediately continues to the next line of code - it doesn't wait for the GPU to finish. Accurate measurement of code execution on GPUs involves the insertion of synchronization barriers to ensure no pending GPU tasks are executing before and after the timing measurements are made. + +## Illustration + +### Without Synchronization + +```mermaid actions={false} +%%{init: {'gantt': {'useWidth': 1200}}}%% +gantt + title CPU vs CUDA Stream Timeline (Without Synchronization) + dateFormat X + axisFormat %s + + section CPU + Timer Start :milestone, m1, 0, 0 + Launch Kernel 1 :active, cpu0, 0, 4 + Launch Kernel 2 :active, cpu1, 4, 8 + Launch Kernel 3 :active, cpu2, 8, 12 + Timer End :milestone, m2, 12, 12 + + section CUDA Stream + Waiting :done, wait, 0, 4 + Kernel 1 :active, k1, 4, 11 + Kernel 2 :active, k2, 11, 18 + Kernel 3 :active, k3, 18, 25 + + section Problem + Timer ends too early :done, p1, after m2, 25 +``` + +Here you can see that the timing statements are measuring the duration up till the end of the final kernel launch. The GPU computation hasn't completed yet, which means the timing measurement is not accurate and would affect any future inference based on this information. + +### With Synchronization + +```mermaid actions={false} +%%{init: {'gantt': {'useWidth': 1200}}}%% +gantt + title CPU vs CUDA Stream Timeline (With Synchronization) + dateFormat X + axisFormat %s + + section CPU + Device Synchronization :done, wait, 0, 4 + Timer Start :milestone, m1, 4, 4 + Launch Kernel 1 :active, cpu0, 4, 8 + Launch Kernel 2 :active, cpu1, 8, 12 + Launch Kernel 3 :active, cpu2, 12, 16 + Device Synchronization :done, wait, 16, 33 + Timer End :milestone, m2, 33, 33 + + section CUDA Stream + Previous Work :done, wait, 0, 4 + Waiting :done, wait, 4, 8 + Kernel 1 :active, k1, 8, 15 + Kernel 2 :active, k2, 15, 22 + Kernel 3 :active, k3, 22, 33 +``` + +Here you can see that a device synchronization call is made before executing the code, this ensures that the CPU waits for any pending GPU tasks to finish before starting the timer. After the launch of the final kernel, another device synchronization call is made which ensures all pending GPU tasks are finished before measuring the runtime. + + + +## Pytorch Example + +Execute the following code in your Python Interpreter to get the kernel launch time (Replace `cuda` with `mps` everywhere to run on your Mac). +```python +import torch +import time +device = "cuda" +x = torch.randn(8192, 8192, device=device) +y = torch.randn(8192, 8192, device=device) +t0 = time.perf_counter_ns() +z = torch.matmul(x, y) +t1 = time.perf_counter_ns() +print(f"Without synchronize: {(t1 - t0) / 1e6:.3f} ms") +``` + +Now, **Restart** your interpreter and execute the following code to get the kernel execution time (Replace `cuda` with `mps` everywhere to run on your Mac). +```python +import torch +import time +device = "cuda" +x = torch.randn(8192, 8192, device=device) +y = torch.randn(8192, 8192, device=device) +torch.cuda.synchronize() # clear any pending work +t0 = time.perf_counter_ns() +z = torch.matmul(x, y) +torch.cuda.synchronize() # wait for GPU to finish +t1 = time.perf_counter_ns() +print(f"With synchronize: {(t1 - t0) / 1e6:.3f} ms") +``` + + +Expected Output on CUDA + +``` +Without synchronize: 69.157 ms +With synchronize: 152.277 ms +``` + +# How Codeflash measures execution time involving GPUs + +Codeflash automatically inserts synchronization barriers before measuring performance. It currently supports GPU code written in `Pytorch`, `Tensorflow` and `JAX` for NVIDIA GPUs (`CUDA`) and MacOS Metal Performance Shaders (`MPS`). + +- **PyTorch**: Uses `torch.cuda.synchronize()` (`CUDA`) or `torch.mps.synchronize()` (`MPS`) depending on the device. +- **JAX**: Uses `jax.block_until_ready()` to wait for computation to complete. It works for both `CUDA` and `MPS` devices. +- **TensorFlow**: Uses `tf.test.experimental.sync_devices()` for device synchronization. It works for both `CUDA` and `MPS` devices. diff --git a/docs/docs.json b/docs/docs.json index 579a8355c..fdc33ef30 100644 --- a/docs/docs.json +++ b/docs/docs.json @@ -58,7 +58,9 @@ "group": "🧠 Core Concepts", "pages": [ "codeflash-concepts/how-codeflash-works", - "codeflash-concepts/benchmarking" + "codeflash-concepts/benchmarking", + "codeflash-concepts/benchmarking-gpu-code", + "support-for-jit/index" ] }, { diff --git a/docs/support-for-jit/index.mdx b/docs/support-for-jit/index.mdx new file mode 100644 index 000000000..0da08dacc --- /dev/null +++ b/docs/support-for-jit/index.mdx @@ -0,0 +1,257 @@ +--- +title: "Just-in-Time Compilation" +description: "Learn how Codeflash optimizes code using JIT compilation with Numba, PyTorch, TensorFlow, and JAX" +icon: "bolt" +sidebarTitle: "JIT Compilation" +keywords: ["JIT", "just-in-time", "numba", "pytorch", "tensorflow", "jax", "GPU", "CUDA", "MPS", "compilation", "performance"] +--- + +# Just-in-Time Compilation + +Just-in-time (JIT) compilation is a runtime technique where code is compiled into machine code on the fly, right before it is executed, to improve performance. Codeflash supports optimizing numerical code using Just-in-Time (JIT) compilation via leveraging JIT compilers from the **Numba**, **PyTorch**, **TensorFlow**, and **JAX** frameworks. + +## How Codeflash Optimizes with JIT + +When Codeflash identifies a function that could benefit from JIT compilation, it: + +1. Rewrites the code in a JIT-compatible format, which may involve breaking down complex functions into separate JIT-compiled components. +2. Generates appropriate tests that are compatible with JIT-compiled code, carefully handling data types since JIT compilers have stricter input type requirements. +3. Disables JIT compilation when running coverage and tracer. This ensures accurate coverage and trace data, since both rely on Python bytecode execution. JIT-compiled code bypasses Python bytecode, so it would prevent proper tracking. +4. Disables the Line Profiler for JIT compiled code. It could be possible to disable JIT compilation and run the line profiler, but that would lead to inaccurate information which could misguide the optimization process. + +## Configuration + +JIT compilation support is **enabled automatically** in Codeflash. You don't need to modify any configuration to enable JIT-based optimizations. Codeflash will automatically detect when JIT compilation could improve performance and suggest appropriate optimizations. + +## When JIT Compilation Helps + +JIT compilation is most effective for: + +- Numerical computations with loops that can't be easily vectorized. +- Custom algorithms not covered by existing optimized libraries. +- Functions that are called repeatedly with consistent input types. +- Code that benefits from hardware-specific optimizations (SIMD acceleration). + +### Example + +#### Function Definition + +```python +import torch +def complex_activation(x): + """A custom activation with many small operations - compile makes a huge difference""" + # Many sequential element-wise ops create kernel launch overhead + x = torch.sin(x) + x = x * torch.cos(x) + x = x + torch.exp(-x.abs()) + x = x / (1 + x.pow(2)) + x = torch.tanh(x) * torch.sigmoid(x) + x = x - 0.5 * x.pow(3) + return x +``` + +#### Benchmarking Snippet (replace `cuda` with `mps` to run on your Mac) + +```python +import time +# Create compiled version +complex_activation_compiled = torch.compile(complex_activation) + +# Benchmark +x = torch.randn(1000, 1000, device='cuda') + +# Warmup steps are slower as the JIT compiler is understanding the function execution to compile into machine code +for _ in range(10): + _ = complex_activation(x) + _ = complex_activation_compiled(x) + +# Time uncompiled +torch.cuda.synchronize() +start = time.time() +for _ in range(100): + y = complex_activation(x) +torch.cuda.synchronize() +uncompiled_time = time.time() - start + +# Time compiled +torch.cuda.synchronize() +start = time.time() +for _ in range(100): + y = complex_activation_compiled(x) +torch.cuda.synchronize() +compiled_time = time.time() - start + +print(f"Uncompiled: {uncompiled_time:.4f}s") +print(f"Compiled: {compiled_time:.4f}s") +print(f"Speedup: {uncompiled_time/compiled_time:.2f}x") +``` + +Expected Output on CUDA + +``` +Uncompiled: 0.0176s +Compiled: 0.0063s +Speedup: 2.80x +``` + +Here, JIT compilation via `torch.compile` is the only viable option because +1. Already vectorized - All operations are already PyTorch tensor ops. +2. Multiple Kernel Launches - Uncompiled code launches ~10 separate kernels. `torch.compile` fuses them into 1-2 kernels, eliminating kernel launch overhead. +3. No algorithmic improvement - The computation itself is already optimal. +4. Python overhead elimination - Removes Python interpreter overhead between operations. + + +## When JIT Compilation May Not Help + +JIT compilation may not provide speedups when: + +- The code already uses highly optimized libraries (e.g., `NumPy` with `MKL`, `cuBLAS`, `cuDNN`). +- Functions have variable input types or shapes that prevent effective compilation. +- The compilation overhead exceeds the runtime savings for short-running functions. +- The code relies heavily on Python objects or dynamic features that JIT compilers can't optimize. + +### Example + +#### Function Definition + +```python +def adaptive_processing(x, threshold=0.5): + """Function with data-dependent control flow - compile struggles here""" + # Check how many values exceed threshold (data-dependent!) + mask = x > threshold + num_large = mask.sum().item() # .item() causes graph break + + if num_large > x.numel() * 0.3: + # Path 1: Many large values - use expensive operation + result = torch.matmul(x, x.T) # Already optimized by cuBLAS + result = result.mean(dim=0) + else: + # Path 2: Few large values - use cheap operation + result = x.mean(dim=1) + + return result +``` + +#### Benchmarking Snippet (replace `cuda` with `mps` to run on your Mac) + +```python +# Create compiled version +adaptive_processing_compiled = torch.compile(adaptive_processing) + +# Test with data that causes branch variation +x = torch.randn(500, 500, device='cuda') + +# Warmup steps are slower as the JIT compiler is understanding the function execution to compile into machine code +for _ in range(10): + _ = adaptive_processing(x) + _ = adaptive_processing_compiled(x) + +# Benchmark with varying data (causes recompilation) +torch.cuda.synchronize() +start = time.time() +for i in range(100): + # Vary the data to trigger different branches + x_test = torch.randn(500, 500, device='cuda') + (i % 2) + y = adaptive_processing(x_test) +torch.cuda.synchronize() +uncompiled_time = time.time() - start + +torch.cuda.synchronize() +start = time.time() +for i in range(100): + x_test = torch.randn(500, 500, device='cuda') + (i % 2) + y = adaptive_processing_compiled(x_test) # Recompiles frequently! +torch.cuda.synchronize() +compiled_time = time.time() - start + +print(f"Uncompiled: {uncompiled_time:.4f}s") +print(f"Compiled: {compiled_time:.4f}s") +print(f"Slowdown: {compiled_time/uncompiled_time:.2f}x") +``` + +Expected Output on CUDA + +``` +Uncompiled: 0.0296s +Compiled: 0.2847s +Slowdown: 9.63x +``` + +Why `torch.compile` is detrimental here: + +1. Graph breaks - `.item()` forces a graph break, negating compile benefits. +2. Recompilation overhead - Different branches cause expensive recompilation each time. +3. Dynamic control flow - Data-dependent conditionals can't be optimized away. +4. Already optimized ops - `matmul` already uses `cuBLAS`; compile adds overhead without benefit. + +#### Better Optimization Strategy + +```python +def optimized_version(x, threshold=0.5): + """Remove data-dependent control flow - vectorize instead""" + mask = (x > threshold).float() + weight = (mask.mean() > 0.3).float() # Keep on GPU + + # Compute both paths, blend based on weight (branchless) + expensive = torch.matmul(x, x.T).mean(dim=0) + cheap = x.mean(dim=1).squeeze() + + # Pad cheap result to match expensive dimensions + cheap_padded = cheap.expand(expensive.shape[0]) + + result = weight * expensive + (1 - weight) * cheap_padded + return result +``` + +Expected Output on CUDA + +``` +Optimized: 0.0277s +Speedup compared to Uncompiled: 1.57x +``` + +Key improvements: + +1. Eliminate `.item()` - Keep computation on GPU. +2. Branchless execution - Compute both paths, blend results. +3. Vectorization - Replace conditionals with masked operations. +4. Reduce Python overhead - Minimize host-device synchronization. + +## Supported JIT Frameworks + +Each framework uses different compilation strategies to accelerate Python code: + +### Numba (CPU Code) + +Numba compiles Python functions to optimized machine code using the LLVM compiler infrastructure. Codeflash can suggest Numba optimizations that use: + +- **`@jit`** - General-purpose JIT compilation with optional flags. + - **`nopython=True`** - Compiles to machine code without falling back to the Python interpreter. + - **`fastmath=True`** - Uses aggressive floating-point optimizations via LLVM's fastmath flag. + - **`cache=True`** - cache compiled function to disk which reduces future runtimes. + - **`parallel=True`** - Parallelizes code inside loops. + +### PyTorch + +PyTorch provides JIT compilation through `torch.compile()`, the recommended compilation API introduced in PyTorch 2.0. It uses TorchDynamo to capture Python bytecode and TorchInductor to generate optimized kernels. + +- **`torch.compile()`** - Compiles a function or module for optimized execution. + - **`mode`** - Controls the compilation strategy: + - `"default"` - Balanced compilation with moderate optimization. + - `"reduce-overhead"` - Minimizes Python overhead using CUDA graphs, ideal for small batches. + - `"max-autotune"` - Spends more time auto-tuning to find the fastest kernels. + - **`fullgraph=True`** - Requires the entire function to be captured as a single graph. Raises an error if graph breaks occur, useful for ensuring complete optimization. + - **`dynamic=True`** - Enables dynamic shape support, allowing the compiled function to handle varying input sizes without recompilation. + +### TensorFlow + +TensorFlow uses `@tf.function` to compile Python functions into optimized TensorFlow graphs. When combined with XLA (Accelerated Linear Algebra), it can generate highly optimized machine code for both CPU and GPU. + +- **`@tf.function`** - Converts Python functions into TensorFlow graphs for optimized execution. + - **`jit_compile=True`** - Enables XLA compilation, which performs whole-function optimization including operation fusion, memory layout optimization, and target-specific code generation. + +### JAX + +JAX uses XLA to JIT compile pure functions into optimized machine code. It emphasizes functional programming patterns and captures side-effect-free operations for optimization. + +- **`@jax.jit`** - JIT compiles functions using XLA with automatic operation fusion. \ No newline at end of file