release: bump version to 0.8.0

New features in v0.8.0:
- Apple Instruments (xctrace) integration for Metal GPU profiling
- Cross-platform device utilities (CUDA/MPS/ROCm/CPU)
- Enhanced Python examples with device selection
- Test runner for multi-device testing

All 8 Python example tests pass on MPS.

Author: chenxingqiang

CHANGELOG.md

@@ -5,6 +5,37 @@ All notable changes to TraceSmith will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [0.8.0] - 2025-12-07
+
+### Added
+- **Apple Instruments (xctrace) Integration**: Real Metal GPU profiling on macOS
+  - `XCTraceProfiler`: Python wrapper for xctrace
+  - Automatic Metal GPU event parsing from Instruments traces
+  - Support for multiple Instruments templates (Metal System Trace, GPU Driver, etc.)
+  - CLI: `--xctrace`, `--xctrace-template`, `--keep-trace` options
+  - Captures real GPU events (13K+ events in tests)
+
+- **Cross-Platform Device Utilities**: Unified device management for examples
+  - `DeviceManager`: Auto-detect and manage CUDA/MPS/ROCm/CPU devices
+  - `benchmark()`: Cross-platform benchmarking with proper synchronization
+  - Test decorators: `skip_if_no_gpu`, `skip_if_not_cuda`, `skip_if_not_mps`
+  - `run_tests.py`: Test runner for all examples across devices
+
+- **Enhanced Python Examples**:
+  - All examples now support `--device` flag for device selection
+  - Automatic device detection and fallback
+  - Comprehensive test coverage on MPS and CPU
+
+### Changed
+- Updated C++ CLI `profile` command with xctrace support on macOS
+- Improved Python CLI with xctrace backend option
+- Enhanced README with cross-platform device support documentation
+
+### Fixed
+- PyTorch hooks profiling backward pass compatibility
+- Multi-GPU profiling cluster availability checks
+- Real-time tracing session configuration
+
 ## [0.7.1] - 2025-12-04
 
 ### Added

CMakeLists.txt

@@ -1,6 +1,6 @@
 cmake_minimum_required(VERSION 3.16)
 project(TraceSmith 
-    VERSION 0.6.0
+    VERSION 0.8.0
     DESCRIPTION "Cross-platform GPU Profiling & Replay System"
     LANGUAGES CXX
 )

README.md

@@ -269,7 +269,7 @@ TraceSmith provides a comprehensive CLI with ASCII banner and colored output:
    ██║   ██║  ██║██║  ██║╚██████╗███████╗███████║██║ ╚═╝ ██║██║   ██║   ██║  ██║
    ╚═╝   ╚═╝  ╚═╝╚═╝  ╚═╝ ╚═════╝╚══════╝╚══════╝╚═╝     ╚═╝╚═╝   ╚═╝   ╚═╝  ╚═╝
 
-                    GPU Profiling & Replay System v0.7.1
+                    GPU Profiling & Replay System v0.8.0
 ```
 
 **Available Commands:**
@@ -708,7 +708,7 @@ make benchmark_10k_stacks -j8
 
 ```bash
 # Basic installation
-pip install tracesmith==0.7.1
+pip install tracesmith==0.8.0
 
 # With CuPy for real GPU profiling in Python CLI (choose one):
 pip install tracesmith[cuda12]    # CUDA 12.x
@@ -795,7 +795,8 @@ for (int i = 0; i < 10000; ++i) {
 
 | Version | Date | Highlights |
 |---------|------|------------|
-| **v0.7.1** | 2025-12 | **Multi-GPU Phase 2** - TimeSync, NCCLTracker, ClockCorrelator, CommAnalysis |
+| **v0.8.0** | 2025-12 | **xctrace Integration** - Apple Instruments, Cross-Platform Device Utils, Enhanced Examples |
+| v0.7.1 | 2025-12 | **Multi-GPU Phase 2** - TimeSync, NCCLTracker, ClockCorrelator, CommAnalysis |
 | **v0.7.0** | 2025-12 | **Multi-GPU Cluster** - GPUTopology, MultiGPUProfiler, GitHub Actions CI/CD |
 | v0.6.9 | 2025-12 | Include reorganization - Directory structure matches `src/` layout |
 | v0.6.8 | 2025-12 | Enhanced CLI - ASCII banner, all commands, Python CLI |

docs/README.md

@@ -87,7 +87,7 @@ docs/
 
 | Metric | Value |
 |--------|-------|
-| **Version** | 0.7.1 |
+| **Version** | 0.8.0 |
 | **C++ Standard** | C++17 |
 | **Python** | 3.9 - 3.12 |
 | **Platforms** | Linux, macOS, Windows |

include/tracesmith/common/types.hpp

@@ -11,8 +11,8 @@ namespace tracesmith {
 
 /// Version information
 constexpr uint32_t VERSION_MAJOR = 0;
-constexpr uint32_t VERSION_MINOR = 7;
-constexpr uint32_t VERSION_PATCH = 1;
+constexpr uint32_t VERSION_MINOR = 8;
+constexpr uint32_t VERSION_PATCH = 0;
 
 /// Event types that can be captured
 enum class EventType : uint8_t {

python/src/bindings.cpp

@@ -57,7 +57,7 @@ PYBIND11_MODULE(_tracesmith, m) {
     m.doc() = "TraceSmith GPU Profiling & Replay System";
 
     // Version info
-    m.attr("__version__") = "0.7.1";
+    m.attr("__version__") = "0.8.0";
     m.attr("VERSION_MAJOR") = VERSION_MAJOR;
     m.attr("VERSION_MINOR") = VERSION_MINOR;
     m.attr("VERSION_PATCH") = VERSION_PATCH;

python/tracesmith/__init__.py

@@ -4,7 +4,7 @@
 A cross-platform GPU profiling and replay system for AI compilers,
 deep learning frameworks, and GPU driver engineers.
 
-Version: 0.7.1
+Version: 0.8.0
 
 Features:
 - Cross-platform GPU profiling (CUDA via CUPTI, ROCm, Metal)

setup.py

@@ -182,7 +182,7 @@ def build_extension(self, ext):
 
 setup(
     name='tracesmith',
-    version='0.7.1',
+    version='0.8.0',
     author='Xingqiang Chen',
     author_email='chenxingqiang@gmail.com',
     description='Cross-platform GPU Profiling & Replay System',