python · Fidget-Spinner · Sep 18, 2025 · Sep 19, 2025 · Sep 19, 2025 · Sep 19, 2025
@@ -57,9 +57,10 @@ jobs:
       fail-fast: false
       matrix:
         target:
-          - i686-pc-windows-msvc/msvc
-          - x86_64-pc-windows-msvc/msvc
-          - aarch64-pc-windows-msvc/msvc
+# To re-enable later when we support these.
+#          - i686-pc-windows-msvc/msvc
+#          - x86_64-pc-windows-msvc/msvc
+#          - aarch64-pc-windows-msvc/msvc
           - x86_64-apple-darwin/clang
           - aarch64-apple-darwin/clang
           - x86_64-unknown-linux-gnu/gcc
@@ -70,15 +71,16 @@ jobs:
         llvm:
           - 19
         include:
-          - target: i686-pc-windows-msvc/msvc
-            architecture: Win32
-            runner: windows-2022
-          - target: x86_64-pc-windows-msvc/msvc
-            architecture: x64
-            runner: windows-2022
-          - target: aarch64-pc-windows-msvc/msvc
-            architecture: ARM64
-            runner: windows-11-arm
+# To re-enable later when we support these.
+#          - target: i686-pc-windows-msvc/msvc
+#            architecture: Win32
+#            runner: windows-2022
+#          - target: x86_64-pc-windows-msvc/msvc
+#            architecture: x64
+#            runner: windows-2022
+#          - target: aarch64-pc-windows-msvc/msvc
+#            architecture: ARM64
+#            runner: windows-11-arm
           - target: x86_64-apple-darwin/clang
             architecture: x86_64
             runner: macos-15-intel

diff --git a/Include/internal/pycore_backoff.h b/Include/internal/pycore_backoff.h
@@ -99,8 +99,13 @@ backoff_counter_triggers(_Py_BackoffCounter counter)
 // Must be larger than ADAPTIVE_COOLDOWN_VALUE, otherwise when JIT code is
 // invalidated we may construct a new trace before the bytecode has properly
 // re-specialized:
-#define JUMP_BACKWARD_INITIAL_VALUE 4095
-#define JUMP_BACKWARD_INITIAL_BACKOFF 12
+// Note: this should be a prime number-1. This increases the likelihood of
+// finding a "good" loop iteration to trace.
+// For example, 4095 does not work for the nqueens benchmark on pyperformanc
-// For example, 4095 does not work for the nqueens benchmark on pyperformanc
+// For example, 4095 does not work for the nqueens benchmark on pyperformance
-// For example, 4095 does not work for the nqueens benchmark on pyperformanc
+// For example, 4095 does not work for the nqueens benchmark on pyperformance
+// as we always end up tracing the loop iteration's
+// exhaustion iteration. Which aborts our current tracer.
+#define JUMP_BACKWARD_INITIAL_VALUE 4000
+#define JUMP_BACKWARD_INITIAL_BACKOFF 14
 static inline _Py_BackoffCounter
 initial_jump_backoff_counter(void)
 {
@@ -112,7 +117,7 @@ initial_jump_backoff_counter(void)
  * Must be larger than ADAPTIVE_COOLDOWN_VALUE,
  * otherwise when a side exit warms up we may construct
  * a new trace before the Tier 1 code has properly re-specialized. */
-#define SIDE_EXIT_INITIAL_VALUE 4095
+#define SIDE_EXIT_INITIAL_VALUE 4000
 #define SIDE_EXIT_INITIAL_BACKOFF 12
 
 static inline _Py_BackoffCounter
@@ -129,6 +134,14 @@ initial_unreachable_backoff_counter(void)
     return make_backoff_counter(0, UNREACHABLE_BACKOFF);
 }
 
+// Required to not get stuck in infinite specialization loops due to specialization failure.
+// We use 2 here as there are a few scenarios:
+// 1. Freshly specialized from unspecialized, in which case the counter will be 1.
+// 2. Re-specialized from deopt, in which case the counter will be 1.
+// 3. Deopt -> Specialize -> Deopt -> Specialize, in which case the counter will be 2.
+// We do not want the 3rd case.
+#define MAX_SPECIALIZATION_TRIES 2
+
 #ifdef __cplusplus
 }
 #endif

diff --git a/Include/internal/pycore_interp_structs.h b/Include/internal/pycore_interp_structs.h
@@ -757,6 +757,29 @@ struct _Py_unique_id_pool {
 
 typedef _Py_CODEUNIT *(*_PyJitEntryFuncPtr)(struct _PyExecutorObject *exec, _PyInterpreterFrame *frame, _PyStackRef *stack_pointer, PyThreadState *tstate);
 
+typedef struct _PyJitTracerState {
+    bool dependencies_still_valid;
+    bool dynamic_jump_taken;
+    bool prev_instr_is_super;
+    int code_max_size;
+    int code_curr_size;
+    int initial_stack_depth;
+    int initial_chain_depth;
+    int prev_instr_oparg;
+    int prev_instr_stacklevel;
+    int specialize_counter;
+    _PyUOpInstruction *code_buffer;
+    _Py_CODEUNIT *insert_exec_instr;
+    _Py_CODEUNIT *close_loop_instr;
+    PyCodeObject *initial_code; // Strong
+    PyFunctionObject *initial_func; // Strong
+    _Py_CODEUNIT *prev_instr;
+    PyCodeObject *prev_instr_code; // Strong
+    struct _PyExitData *prev_exit;
+    _PyInterpreterFrame *prev_instr_frame;
+    _PyBloomFilter dependencies;
+} _PyJitTracerState;
+
 /* PyInterpreterState holds the global state for one of the runtime's
    interpreters.  Typically the initial (main) interpreter is the only one.
 
@@ -932,9 +955,9 @@ struct _is {
     struct types_state types;
     struct callable_cache callable_cache;
     PyObject *common_consts[NUM_COMMON_CONSTANTS];
+    _PyJitTracerState jit_state;
     bool jit;
     bool compiling;
-    struct _PyUOpInstruction *jit_uop_buffer;
     struct _PyExecutorObject *executor_list_head;
     struct _PyExecutorObject *executor_deletion_list_head;
     struct _PyExecutorObject *cold_executor;

diff --git a/Include/internal/pycore_magic_number.h b/Include/internal/pycore_magic_number.h
@@ -286,6 +286,7 @@ Known values:
     Python 3.15a1 3653 (Fix handling of opcodes that may leave operands on the stack when optimizing LOAD_FAST)
     Python 3.15a1 3654 (Fix missing exception handlers in logical expression)
     Python 3.15a1 3655 (Fix miscompilation of some module-level annotations)
+    Python 3.15a2 3656 (Add RECORD_PREVIOUS_INST for a tracing JIT)
 
 
     Python 3.16 will start with 3700
@@ -299,7 +300,7 @@ PC/launcher.c must also be updated.
 
 */
 
-#define PYC_MAGIC_NUMBER 3655
+#define PYC_MAGIC_NUMBER 3656
 /* This is equivalent to converting PYC_MAGIC_NUMBER to 2 bytes
    (little-endian) and then appending b'\r\n'. */
 #define PYC_MAGIC_NUMBER_TOKEN \

diff --git a/Include/internal/pycore_opcode_metadata.h b/Include/internal/pycore_opcode_metadata.h
diff --git a/Include/internal/pycore_optimizer.h b/Include/internal/pycore_optimizer.h
@@ -21,14 +21,6 @@ typedef struct _PyExecutorLinkListNode {
 } _PyExecutorLinkListNode;
 
 
-/* Bloom filter with m = 256
- * https://en.wikipedia.org/wiki/Bloom_filter */
-#define _Py_BLOOM_FILTER_WORDS 8
-
-typedef struct {
-    uint32_t bits[_Py_BLOOM_FILTER_WORDS];
-} _PyBloomFilter;
-
 typedef struct {
     uint8_t opcode;
     uint8_t oparg;
@@ -95,7 +87,8 @@ PyAPI_FUNC(void) _Py_Executors_InvalidateCold(PyInterpreterState *interp);
 
 #define TRACE_STACK_SIZE 5
 
-int _Py_uop_analyze_and_optimize(_PyInterpreterFrame *frame,
+int _Py_uop_analyze_and_optimize(
+    PyFunctionObject *initial_func,
     _PyUOpInstruction *trace, int trace_len, int curr_stackentries,
     _PyBloomFilter *dependencies);
 
@@ -257,6 +250,7 @@ struct _Py_UOpsAbstractFrame {
     int stack_len;
     int locals_len;
     PyFunctionObject *func;
+    PyCodeObject *code;
 
     JitOptRef *stack_pointer;
     JitOptRef *stack;
@@ -332,11 +326,11 @@ extern _Py_UOpsAbstractFrame *_Py_uop_frame_new(
     int curr_stackentries,
     JitOptRef *args,
     int arg_len);
-extern int _Py_uop_frame_pop(JitOptContext *ctx);
+extern int _Py_uop_frame_pop(JitOptContext *ctx, PyCodeObject *co, int curr_stackentries);
 
 PyAPI_FUNC(PyObject *) _Py_uop_symbols_test(PyObject *self, PyObject *ignored);
 
-PyAPI_FUNC(int) _PyOptimizer_Optimize(_PyInterpreterFrame *frame, _Py_CODEUNIT *start, _PyExecutorObject **exec_ptr, int chain_depth);
+PyAPI_FUNC(int) _PyOptimizer_Optimize(_PyInterpreterFrame *frame, PyThreadState *tstate);
 
 static inline _PyExecutorObject *_PyExecutor_FromExit(_PyExitData *exit)
 {
@@ -353,7 +347,8 @@ static inline int is_terminator(const _PyUOpInstruction *uop)
     int opcode = uop->opcode;
     return (
         opcode == _EXIT_TRACE ||
-        opcode == _JUMP_TO_TOP
+        opcode == _JUMP_TO_TOP ||
+        opcode == _DYNAMIC_EXIT
     );
 }
 
@@ -364,6 +359,17 @@ PyAPI_FUNC(int) _PyDumpExecutors(FILE *out);
 extern void _Py_ClearExecutorDeletionList(PyInterpreterState *interp);
 #endif
 
+int _PyJit_translate_single_bytecode_to_trace(PyThreadState *tstate, _PyInterpreterFrame *frame, _Py_CODEUNIT *next_instr);
+
+void
+_PyJit_InitializeTracing(PyThreadState *tstate, _PyInterpreterFrame *frame,
+    _Py_CODEUNIT *curr_instr, _Py_CODEUNIT *insert_exec_instr,
+    _Py_CODEUNIT *close_loop_instr, int curr_stackdepth, int chain_depth, _PyExitData *exit, int oparg);
+
+void _PyJit_FinalizeTracing(PyThreadState *tstate);
+
+void _PyJit_Tracer_InvalidateDependency(PyThreadState *old_tstate, void *obj);
+
 #ifdef __cplusplus
 }
 #endif

diff --git a/Include/internal/pycore_uop.h b/Include/internal/pycore_uop.h
@@ -35,10 +35,18 @@ typedef struct _PyUOpInstruction{
 #endif
 } _PyUOpInstruction;
 
-// This is the length of the trace we project initially.
-#define UOP_MAX_TRACE_LENGTH 1200
+// This is the length of the trace we translate initially.
+#define UOP_MAX_TRACE_LENGTH 3000
 #define UOP_BUFFER_SIZE (UOP_MAX_TRACE_LENGTH * sizeof(_PyUOpInstruction))
 
+/* Bloom filter with m = 256
+ * https://en.wikipedia.org/wiki/Bloom_filter */
+#define _Py_BLOOM_FILTER_WORDS 8
+
+typedef struct {
+    uint32_t bits[_Py_BLOOM_FILTER_WORDS];
+} _PyBloomFilter;
+
 #ifdef __cplusplus
 }
 #endif