gh-137838: Move _PyUOpInstruction buffer[UOP_MAX_TRACE_LENGTH] to _PyThreadStateImpl

Author: corona10

Include/internal/pycore_optimizer.h

@@ -11,6 +11,7 @@ extern "C" {
 #include "pycore_typedefs.h"      // _PyInterpreterFrame
 #include "pycore_uop_ids.h"
 #include "pycore_stackref.h"      // _PyStackRef
+#include "pycore_tstate.h"       // _PyUOpInstruction
 #include <stdbool.h>
 
 
@@ -41,32 +42,6 @@ typedef struct {
     PyCodeObject *code;  // Weak (NULL if no corresponding ENTER_EXECUTOR).
 } _PyVMData;
 
-/* Depending on the format,
- * the 32 bits between the oparg and operand are:
- * UOP_FORMAT_TARGET:
- *    uint32_t target;
- * UOP_FORMAT_JUMP
- *    uint16_t jump_target;
- *    uint16_t error_target;
- */
-typedef struct {
-    uint16_t opcode:15;
-    uint16_t format:1;
-    uint16_t oparg;
-    union {
-        uint32_t target;
-        struct {
-            uint16_t jump_target;
-            uint16_t error_target;
-        };
-    };
-    uint64_t operand0;  // A cache entry
-    uint64_t operand1;
-#ifdef Py_STATS
-    uint64_t execution_count;
-#endif
-} _PyUOpInstruction;
-
 typedef struct _PyExitData {
     uint32_t target;
     uint16_t index;
@@ -118,9 +93,6 @@ PyAPI_FUNC(void) _Py_Executors_InvalidateCold(PyInterpreterState *interp);
 // trace_run_counter is greater than this value.
 #define JIT_CLEANUP_THRESHOLD 100000
 
-// This is the length of the trace we project initially.
-#define UOP_MAX_TRACE_LENGTH 1200
-
 #define TRACE_STACK_SIZE 5
 
 int _Py_uop_analyze_and_optimize(_PyInterpreterFrame *frame,

Include/internal/pycore_tstate.h

@@ -13,14 +13,43 @@ extern "C" {
 #include "pycore_mimalloc.h"        // struct _mimalloc_thread_state
 #include "pycore_qsbr.h"            // struct qsbr
 
-
 #ifdef Py_GIL_DISABLED
 struct _gc_thread_state {
     /* Thread-local allocation count. */
     Py_ssize_t alloc_count;
 };
 #endif
 
+/* Depending on the format,
+ * the 32 bits between the oparg and operand are:
+ * UOP_FORMAT_TARGET:
+ *    uint32_t target;
+ * UOP_FORMAT_JUMP
+ *    uint16_t jump_target;
+ *    uint16_t error_target;
+ */
+typedef struct _PyUOpInstruction{
+    uint16_t opcode:15;
+    uint16_t format:1;
+    uint16_t oparg;
+    union {
+        uint32_t target;
+        struct {
+            uint16_t jump_target;
+            uint16_t error_target;
+        };
+    };
+    uint64_t operand0;  // A cache entry
+    uint64_t operand1;
+#ifdef Py_STATS
+    uint64_t execution_count;
+#endif
+} _PyUOpInstruction;
+
+// This is the length of the trace we project initially.
+#define UOP_MAX_TRACE_LENGTH 1200
+
+
 // Every PyThreadState is actually allocated as a _PyThreadStateImpl. The
 // PyThreadState fields are exposed as part of the C API, although most fields
 // are intended to be private. The _PyThreadStateImpl fields not exposed.
@@ -75,6 +104,7 @@ typedef struct _PyThreadStateImpl {
 #if defined(Py_REF_DEBUG) && defined(Py_GIL_DISABLED)
     Py_ssize_t reftotal;  // this thread's total refcount operations
 #endif
+    struct _PyUOpInstruction buffer[1200];
 
 } _PyThreadStateImpl;
 

Misc/NEWS.d/next/Core_and_Builtins/2025-09-15-13-14-20.gh-issue-137838.Ju-vRW.rst

@@ -0,0 +1,2 @@
+Move _PyUOpInstruction buffer[UOP_MAX_TRACE_LENGTH] to _PyThreadStateImpl.
+Patch By Donghee Na.

Python/optimizer.c

@@ -1280,7 +1280,8 @@ uop_optimize(
 {
     _PyBloomFilter dependencies;
     _Py_BloomFilter_Init(&dependencies);
-    _PyUOpInstruction buffer[UOP_MAX_TRACE_LENGTH];
+    _PyThreadStateImpl *tstate = (_PyThreadStateImpl *)_PyThreadState_GET();
+    _PyUOpInstruction *buffer = tstate->buffer;
     OPT_STAT_INC(attempts);
     char *env_var = Py_GETENV("PYTHON_UOPS_OPTIMIZE");
     bool is_noopt = true;