gh-116968: Reimplement Tier 2 counters (#117144)

Introduce a unified 16-bit backoff counter type (``_Py_BackoffCounter``),
shared between the Tier 1 adaptive specializer and the Tier 2 optimizer. The
API used for adaptive specialization counters is changed but the behavior is
(supposed to be) identical.

The behavior of the Tier 2 counters is changed:
- There are no longer dynamic thresholds (we never varied these).
- All counters now use the same exponential backoff.
- The counter for ``JUMP_BACKWARD`` starts counting down from 16.
- The ``temperature`` in side exits starts counting down from 64.

2 files changed, 12 insertions, 14 deletions

diff --git a/Include/cpython/code.h b/Include/cpython/code.h
index d5dac17..b0e226e 100644
--- a/Include/cpython/code.h
+++ b/Include/cpython/code.h
@@ -24,6 +24,16 @@ typedef struct _Py_GlobalMonitors {
     uint8_t tools[_PY_MONITORING_UNGROUPED_EVENTS];
 } _Py_GlobalMonitors;
 
+typedef struct {
+    union {
+        struct {
+            uint16_t backoff : 4;
+            uint16_t value : 12;
+        };
+        uint16_t as_counter;  // For printf("%#x", ...)
+    };
+} _Py_BackoffCounter;
+
 /* Each instruction in a code object is a fixed-width value,
  * currently 2 bytes: 1-byte opcode + 1-byte oparg.  The EXTENDED_ARG
  * opcode allows for larger values but the current limit is 3 uses
@@ -39,6 +49,7 @@ typedef union {
         uint8_t code;
         uint8_t arg;
     } op;
+    _Py_BackoffCounter counter;  // First cache entry of specializable op
 } _Py_CODEUNIT;
 
 
diff --git a/Include/cpython/optimizer.h b/Include/cpython/optimizer.h
index bc960c5..819251a 100644
--- a/Include/cpython/optimizer.h
+++ b/Include/cpython/optimizer.h
@@ -89,7 +89,7 @@ static inline uint16_t uop_get_error_target(const _PyUOpInstruction *inst)
 
 typedef struct _exit_data {
     uint32_t target;
-    int16_t temperature;
+    _Py_BackoffCounter temperature;
     const struct _PyExecutorObject *executor;
 } _PyExitData;
 
@@ -115,11 +115,6 @@ typedef int (*optimize_func)(
 struct _PyOptimizerObject {
     PyObject_HEAD
     optimize_func optimize;
-    /* These thresholds are treated as signed so do not exceed INT16_MAX
-     * Use INT16_MAX to indicate that the optimizer should never be called */
-    uint16_t resume_threshold;
-    uint16_t side_threshold;
-    uint16_t backedge_threshold;
     /* Data needed by the optimizer goes here, but is opaque to the VM */
 };
 
@@ -151,14 +146,6 @@ extern void _Py_Executors_InvalidateAll(PyInterpreterState *interp, int is_inval
 PyAPI_FUNC(PyObject *)PyUnstable_Optimizer_NewCounter(void);
 PyAPI_FUNC(PyObject *)PyUnstable_Optimizer_NewUOpOptimizer(void);
 
-#define OPTIMIZER_BITS_IN_COUNTER 4
-/* Minimum of 16 additional executions before retry */
-#define MIN_TIER2_BACKOFF 4
-#define MAX_TIER2_BACKOFF (15 - OPTIMIZER_BITS_IN_COUNTER)
-#define OPTIMIZER_BITS_MASK ((1 << OPTIMIZER_BITS_IN_COUNTER) - 1)
-/* A value <= UINT16_MAX but large enough that when shifted is > UINT16_MAX */
-#define OPTIMIZER_UNREACHABLE_THRESHOLD UINT16_MAX
-
 #define _Py_MAX_ALLOWED_BUILTINS_MODIFICATIONS 3
 #define _Py_MAX_ALLOWED_GLOBALS_MODIFICATIONS 6