gh-114058: The Tier2 Optimizer

.gitattributes

@@ -94,6 +94,7 @@ Programs/test_frozenmain.h                          generated
 Python/Python-ast.c                                 generated
 Python/executor_cases.c.h                           generated
 Python/generated_cases.c.h                          generated
+Python/abstract_interp_cases.c.h                    generated
 Python/opcode_targets.h                             generated
 Python/stdlib_module_names.h                        generated
 Tools/peg_generator/pegen/grammar_parser.py         generated

Include/cpython/pystats.h

@@ -120,6 +120,14 @@ typedef struct _optimization_stats {
     uint64_t trace_length_hist[_Py_UOP_HIST_SIZE];
     uint64_t trace_run_length_hist[_Py_UOP_HIST_SIZE];
     uint64_t optimized_trace_length_hist[_Py_UOP_HIST_SIZE];
+    uint64_t optimizer_attempts;
+    uint64_t optimizer_successes;
+    uint64_t optimizer_failure_reason_null_function;
+    uint64_t optimizer_failure_reason_no_memory;
+    uint64_t optimizer_failure_reason_no_writebuffer;
+    uint64_t loop_body_duplication_attempts;
+    uint64_t loop_body_duplication_successes;
+    uint64_t loop_body_duplication_no_mem;
 } OptimizationStats;
 
 typedef struct _rare_event_stats {

Include/internal/pycore_function.h

@@ -16,7 +16,7 @@ extern PyObject* _PyFunction_Vectorcall(
 
 #define FUNC_MAX_WATCHERS 8
 
-#define FUNC_VERSION_CACHE_SIZE (1<<12)  /* Must be a power of 2 */
+#define FUNC_VERSION_CACHE_SIZE (1<<14)  /* Must be a power of 2 */
 struct _py_func_state {
     uint32_t next_version;
     // Borrowed references to function objects whose

Include/internal/pycore_optimizer.h

@@ -8,10 +8,21 @@ extern "C" {
 #  error "this header requires Py_BUILD_CORE define"
 #endif
 
+#include "pycore_uop_ids.h"
+
+// This is the length of the trace we project initially.
+#define UOP_MAX_TRACE_LENGTH 512
+// This the above + additional working space we need.
+#define UOP_MAX_TRACE_WORKING_LENGTH (UOP_MAX_TRACE_LENGTH * 2)
+
+#define TRACE_STACK_SIZE 5
+
 int _Py_uop_analyze_and_optimize(_PyInterpreterFrame *frame,
     _PyUOpInstruction *trace, int trace_len, int curr_stackentries,
     _PyBloomFilter *dependencies);
 
+
+
 extern PyTypeObject _PyCounterExecutor_Type;
 extern PyTypeObject _PyCounterOptimizer_Type;
 extern PyTypeObject _PyDefaultOptimizer_Type;