Replace dispatch heuristics with profiling-derived thresholds; fix AVX-512/MSVC build

CMakeLists.txt

@@ -1092,19 +1092,28 @@ endif()
 # source-file-specific flags (cmake/SIMDDetection.cmake) - All platforms: Definitions are set by
 # SIMDDetection.cmake based on detection
 
-# Windows compilers: Use global SIMD flags for compatibility
+# Windows compilers: Use highest detected SIMD level as global flag.
+# SIMDDetection.cmake has already run by this point and set LIBSTATS_HAS_AVX512 etc.
 if(CMAKE_CXX_COMPILER_ID STREQUAL "MSVC" AND CMAKE_SIZEOF_VOID_P EQUAL 8)
-    # MSVC x64 has comprehensive SIMD support
-    add_compile_options(/arch:AVX2)
-    message(STATUS "Applied MSVC x64 SIMD flags: /arch:AVX2")
+    if(LIBSTATS_HAS_AVX512)
+        add_compile_options(/arch:AVX512)
+        message(STATUS "Applied MSVC x64 SIMD flags: /arch:AVX512")
+    else()
+        add_compile_options(/arch:AVX2)
+        message(STATUS "Applied MSVC x64 SIMD flags: /arch:AVX2")
+    endif()
 
 elseif(
     CMAKE_CXX_COMPILER_ID MATCHES "Clang"
     AND WIN32
     AND CMAKE_SIZEOF_VOID_P EQUAL 8)
-    # Clang-cl on Windows x64
-    add_compile_options(-mavx2)
-    message(STATUS "Applied Clang-cl x64 SIMD flags: -mavx2")
+    if(LIBSTATS_HAS_AVX512)
+        add_compile_options(-mavx512f)
+        message(STATUS "Applied Clang-cl x64 SIMD flags: -mavx512f")
+    else()
+        add_compile_options(-mavx2)
+        message(STATUS "Applied Clang-cl x64 SIMD flags: -mavx2")
+    endif()
 endif()
 
 # IMPORTANT: SIMD compile definitions are handled by cmake/SIMDDetection.cmake That system detects
@@ -1169,12 +1178,20 @@ endif()
 
 # SIMD Status Messages (compiler-specific)
 if(CMAKE_CXX_COMPILER_ID STREQUAL "MSVC" AND CMAKE_SIZEOF_VOID_P EQUAL 8)
-    message(STATUS "SIMD: AVX2/AVX/SSE2 enabled (MSVC x64)")
+    if(LIBSTATS_HAS_AVX512)
+        message(STATUS "SIMD: AVX-512/AVX2/AVX/SSE2 enabled (MSVC x64)")
+    else()
+        message(STATUS "SIMD: AVX2/AVX/SSE2 enabled (MSVC x64)")
+    endif()
 elseif(
     CMAKE_CXX_COMPILER_ID MATCHES "Clang"
     AND WIN32
     AND CMAKE_SIZEOF_VOID_P EQUAL 8)
-    message(STATUS "SIMD: AVX2/AVX/SSE2 enabled (ClangCL x64)")
+    if(LIBSTATS_HAS_AVX512)
+        message(STATUS "SIMD: AVX-512/AVX2/AVX/SSE2 enabled (ClangCL x64)")
+    else()
+        message(STATUS "SIMD: AVX2/AVX/SSE2 enabled (ClangCL x64)")
+    endif()
 elseif(CMAKE_CXX_COMPILER_ID MATCHES "GNU")
     # GCC SIMD status messages
     set(SIMD_FEATURES "")
@@ -1225,10 +1242,8 @@ set(LIBSTATS_CORE_UTILITIES_SOURCES
 )
 
 # Level 2: Platform Capabilities (Depends on Level 0-1)
-set(LIBSTATS_PLATFORM_SOURCES
-    src/parallel_thresholds.cpp # Architecture-specific parallel thresholds
-    src/thread_pool.cpp # Thread pool implementation
-    src/work_stealing_pool.cpp # Advanced work-stealing thread pool
+set(LIBSTATS_PLATFORM_SOURCES src/thread_pool.cpp # Thread pool implementation
+                              src/work_stealing_pool.cpp # Advanced work-stealing thread pool
 )
 
 # Level 3: Advanced Infrastructure (Depends on Level 0-2)
@@ -1799,7 +1814,7 @@ if(LIBSTATS_BUILD_TESTS)
             test_student_t_enhanced
             test_beta_enhanced
             test_performance_dispatcher
-            test_system_capabilities   # runs live SIMD/threading/bandwidth benchmarks
+            test_system_capabilities # runs live SIMD/threading/bandwidth benchmarks
             PROPERTIES LABELS "timing")
     endif()
     set_tests_properties(benchmark_simd_all PROPERTIES LABELS "benchmark")
@@ -1933,11 +1948,8 @@ if(LIBSTATS_BUILD_TOOLS)
     add_standalone_tool(cpp20_features_inspector cpp20_features_inspector.cpp)
 
     # Performance & Benchmarking Tools
-    add_libstats_tool(parallel_threshold_benchmark parallel_threshold_benchmark.cpp)
     add_libstats_tool(parallel_batch_fitting_benchmark parallel_batch_fitting_benchmark.cpp)
-    add_libstats_tool(performance_dispatcher_tool performance_dispatcher_tool.cpp)
-    add_libstats_tool(learning_analyzer learning_analyzer.cpp)
-    add_libstats_tool(empirical_characteristics_demo empirical_characteristics_demo.cpp)
+    add_libstats_tool(strategy_profile strategy_profile.cpp)
     add_libstats_tool(simd_verification simd_verification.cpp)
     add_libstats_tool(parallel_correctness_verification parallel_correctness_verification.cpp)
 
@@ -1950,25 +1962,13 @@ if(LIBSTATS_BUILD_TOOLS)
         STATUS
             "  - cpp20_features_inspector: Comprehensive C++20 compiler and standard library feature detection with detailed functionality tests"
     )
-    message(
-        STATUS
-            "  - parallel_threshold_benchmark: Enhanced distribution-specific threshold optimization with adaptive learning"
-    )
     message(
         STATUS
             "  - parallel_batch_fitting_benchmark: Comprehensive parallel batch fitting performance analysis across all distributions with scalability testing"
     )
     message(
         STATUS
-            "  - performance_dispatcher_tool: Interactive Phase 3 performance framework demonstration"
-    )
-    message(
-        STATUS
-            "  - learning_analyzer: Unified adaptive learning analysis with both educational simulation and real execution data (consolidates threshold_learning_demo and adaptive_learning_analyzer)"
-    )
-    message(
-        STATUS
-            "  - empirical_characteristics_demo: Demonstration of empirical distribution characteristics system replacing assumption-based performance models"
+            "  - strategy_profile: Canonical forced-strategy profiler for dispatcher threshold tuning across distributions, operations, and batch sizes"
     )
     message(
         STATUS

PROJECT_CONCEPT.md

@@ -139,9 +139,8 @@ These help validate correctness, SIMD behavior, thresholds, and runtime capabili
 Examples:
 - `system_inspector`
 - `simd_verification`
-- `parallel_threshold_benchmark`
-- `performance_dispatcher_tool`
-- `learning_analyzer`
+- `strategy_profile`
+- `parallel_batch_fitting_benchmark`
 
 ### Historical or specialized analysis tools
 These support specific refactors or investigations and should be documented as such when retained.

README.md

@@ -183,9 +183,8 @@ libstats/
 ### 🔧 **Analysis Tools** (`tools/` directory)
 - `system_inspector` - CPU capabilities and system information
 - `simd_verification` - SIMD correctness and speedup verification
-- `parallel_threshold_benchmark` - Architecture-aware parallel threshold analysis
-- `performance_dispatcher_tool` - Dispatch strategy inspection and comparison
-- `learning_analyzer` - Performance-learning and threshold-analysis support
+- `strategy_profile` - Canonical forced-strategy profiler for dispatcher threshold tuning
+- `parallel_batch_fitting_benchmark` - Parallel batch fitting performance analysis
 
 
 ## Testing

WARP.md

@@ -184,7 +184,7 @@ The active SIMD level changes fundamentally between machines:
 SIMD code paths, performance thresholds, and test results are architecture-dependent. If the machine has changed since the last session:
 - Note the change explicitly
 - Verify the build directory is current for this architecture (`cmake ..` may be needed)
-- Threshold values in `src/parallel_thresholds.cpp` may need review
+- Dispatch thresholds in `include/core/dispatch_thresholds.h` are architecture-specific
 - Benchmark results are not comparable across architectures
 
 ## Essential Build Commands
@@ -233,9 +233,11 @@ cmake -DCMAKE_BUILD_TYPE=MSVCStrict ..
 ./build/tools/cpp20_features_inspector
 
 # Performance analysis
-./build/tools/parallel_threshold_benchmark
+./build/tools/strategy_profile
 ./build/tools/simd_verification
-./build/tools/performance_dispatcher_tool
+
+# Dispatcher profiling bundle capture
+./scripts/capture_dispatcher_profile.sh
 
 # Cross-compiler compatibility testing
 ./scripts/test-cross-compiler.sh --clean
@@ -429,7 +431,7 @@ include/
 ```
 src/
 ├── [Level 0-1] Foundation and utilities (cpu_detection.cpp, safety.cpp)
-├── [Level 2] Platform capabilities (thread_pool.cpp, parallel_thresholds.cpp)
+├── [Level 2] Platform capabilities (thread_pool.cpp, work_stealing_pool.cpp)
 ├── [Level 3] Infrastructure (benchmark.cpp, performance_dispatcher.cpp)
 ├── [Level 4] Framework (distribution_base.cpp)
 └── [Level 5] Distributions (gaussian.cpp, exponential.cpp, etc.)
@@ -464,7 +466,8 @@ The CMake system uses dependency-aware object libraries for parallel compilation
 #### Parallel Processing
 - Auto-dispatch API: `getProbability(std::span<const double>, std::span<double>, hint)`
 - Explicit control: `getProbabilityWithStrategy(spans, Strategy::PARALLEL)`
-- Performance thresholds: <8 elements (scalar), 8-1000 (SIMD), >1000 (parallel)
+- Dispatch thresholds are per-(architecture, distribution, operation) in `dispatch_thresholds.h`
+- Thresholds derived from four-architecture profiling data in `data/profiles/dispatcher/`
 
 ### Build System Customization
 
@@ -545,8 +548,8 @@ when the machine is loaded. This is a measurement problem, not a correctness pro
 # Verify SIMD operations and performance
 ./build/tools/simd_verification
 
-# Analyze parallel thresholds
-./build/tools/parallel_threshold_benchmark
+# Profile forced strategies for threshold tuning
+./build/tools/strategy_profile
 
 # System capability analysis
 ./build/tools/system_inspector --performance