gh-112287: Speed up Tier 2 (uop) interpreter a little (#112286)

This makes the Tier 2 interpreter a little faster.
I calculated by about 3%,
though I hesitate to claim an exact number.

This starts by doubling the trace size limit (to 512),
making it more likely that loops fit in a trace.

The rest of the approach is to only load
`oparg` and `operand` in cases that use them.
The code generator know when these are used.

For `oparg`, it will conditionally emit
```
oparg = CURRENT_OPARG();
```
at the top of the case block.
(The `oparg` variable may be referenced multiple times
by the instructions code block, so it must be in a variable.)

For `operand`, it will use `CURRENT_OPERAND()` directly
instead of referencing the `operand` variable,
which no longer exists.
(There is only one place where this will be used.)

Author: gvanrossum

Include/internal/pycore_uops.h

@@ -10,7 +10,7 @@ extern "C" {
 
 #include "pycore_frame.h"         // _PyInterpreterFrame
 
-#define _Py_UOP_MAX_TRACE_LENGTH 256
+#define _Py_UOP_MAX_TRACE_LENGTH 512
 
 typedef struct {
     uint16_t opcode;

Misc/NEWS.d/next/Core and Builtins/2023-11-20-10-40-40.gh-issue-112287.15gWAK.rst

@@ -0,0 +1,3 @@
+Slightly optimize the Tier 2 (uop) interpreter by only loading ``oparg`` and
+``operand`` when needed. Also double the trace size limit again, to 512 this
+time.

Python/ceval.c

@@ -994,21 +994,18 @@ _PyEval_EvalFrameDefault(PyThreadState *tstate, _PyInterpreterFrame *frame, int
 
     OPT_STAT_INC(traces_executed);
     _PyUOpInstruction *next_uop = current_executor->trace;
-    uint64_t operand;
 #ifdef Py_STATS
     uint64_t trace_uop_execution_counter = 0;
 #endif
 
     for (;;) {
         opcode = next_uop->opcode;
-        oparg = next_uop->oparg;
-        operand = next_uop->operand;
         DPRINTF(3,
                 "%4d: uop %s, oparg %d, operand %" PRIu64 ", target %d, stack_level %d\n",
                 (int)(next_uop - current_executor->trace),
                 _PyUopName(opcode),
-                oparg,
-                operand,
+                next_uop->oparg,
+                next_uop->operand,
                 next_uop->target,
                 (int)(stack_pointer - _PyFrame_Stackbase(frame)));
         next_uop++;
@@ -1025,8 +1022,9 @@ _PyEval_EvalFrameDefault(PyThreadState *tstate, _PyInterpreterFrame *frame, int
             default:
 #ifdef Py_DEBUG
             {
-                fprintf(stderr, "Unknown uop %d, oparg %d, operand %" PRIu64 "\n",
-                        opcode, oparg, operand);
+                fprintf(stderr, "Unknown uop %d, oparg %d, operand %" PRIu64 " @ %d\n",
+                        opcode, next_uop[-1].oparg, next_uop[-1].operand,
+                        (int)(next_uop - current_executor->trace - 1));
                 Py_FatalError("Unknown uop");
             }
 #else
@@ -1055,7 +1053,7 @@ _PyEval_EvalFrameDefault(PyThreadState *tstate, _PyInterpreterFrame *frame, int
     STACK_SHRINK(1);
 error_tier_two:
     DPRINTF(2, "Error: [Uop %d (%s), oparg %d, operand %" PRIu64 ", target %d @ %d]\n",
-            opcode, _PyUopName(opcode), oparg, operand, next_uop[-1].target,
+            opcode, _PyUopName(opcode), next_uop[-1].oparg, next_uop[-1].operand, next_uop[-1].target,
             (int)(next_uop - current_executor->trace - 1));
     OPT_HIST(trace_uop_execution_counter, trace_run_length_hist);
     frame->return_offset = 0;  // Don't leave this random
@@ -1068,7 +1066,7 @@ _PyEval_EvalFrameDefault(PyThreadState *tstate, _PyInterpreterFrame *frame, int
     // On DEOPT_IF we just repeat the last instruction.
     // This presumes nothing was popped from the stack (nor pushed).
     DPRINTF(2, "DEOPT: [Uop %d (%s), oparg %d, operand %" PRIu64 ", target %d @ %d]\n",
-            opcode, _PyUopName(opcode), oparg, operand, next_uop[-1].target,
+            opcode, _PyUopName(opcode), next_uop[-1].oparg, next_uop[-1].operand, next_uop[-1].target,
             (int)(next_uop - current_executor->trace - 1));
     OPT_HIST(trace_uop_execution_counter, trace_run_length_hist);
     UOP_STAT_INC(opcode, miss);

Python/ceval_macros.h

@@ -397,3 +397,7 @@ stack_pointer = _PyFrame_GetStackPointer(frame);
 #define GOTO_TIER_TWO() goto enter_tier_two;
 
 #define GOTO_TIER_ONE() goto exit_trace;
+
+#define CURRENT_OPARG() (next_uop[-1].oparg)
+
+#define CURRENT_OPERAND() (next_uop[-1].operand)