GH-91719: Make MSVC generate somewhat faster switch code (#91718)

Apparently a switch on an 8-bit quantity where all cases are
present generates a more efficient jump (doing only one indexed
memory load instead of two).

So we make opcode and use_tracing uint8_t, and generate a macro
full of extra `case NNN:` lines for all unused opcodes.

See https://github.com/faster-cpython/ideas/issues/321#issuecomment-1103263673

4 files changed, 90 insertions, 3 deletions

diff --git a/Include/cpython/pystate.h b/Include/cpython/pystate.h
index 1af21a2..2bd4606 100644
--- a/Include/cpython/pystate.h
+++ b/Include/cpython/pystate.h
@@ -46,7 +46,7 @@ typedef struct _PyCFrame {
      * discipline and make sure that instances of this struct cannot
      * accessed outside of their lifetime.
      */
-    int use_tracing;
+    uint8_t use_tracing;  // 0 or 255 (or'ed into opcode, hence 8-bit type)
     /* Pointer to the currently executing frame (it can be NULL) */
     struct _PyInterpreterFrame *current_frame;
     struct _PyCFrame *previous;
diff --git a/Include/opcode.h b/Include/opcode.h
index 8db4238..89f56e2 100644
--- a/Include/opcode.h
+++ b/Include/opcode.h
@@ -707,6 +707,84 @@ static const char *const _PyOpcode_OpName[256] = {
 };
 #endif
 
+#define EXTRA_CASES \
+    case 180: \
+    case 181: \
+    case 182: \
+    case 183: \
+    case 184: \
+    case 185: \
+    case 186: \
+    case 187: \
+    case 188: \
+    case 189: \
+    case 190: \
+    case 191: \
+    case 192: \
+    case 193: \
+    case 194: \
+    case 195: \
+    case 196: \
+    case 197: \
+    case 198: \
+    case 199: \
+    case 200: \
+    case 201: \
+    case 202: \
+    case 203: \
+    case 204: \
+    case 205: \
+    case 206: \
+    case 207: \
+    case 208: \
+    case 209: \
+    case 210: \
+    case 211: \
+    case 212: \
+    case 213: \
+    case 214: \
+    case 215: \
+    case 216: \
+    case 217: \
+    case 218: \
+    case 219: \
+    case 220: \
+    case 221: \
+    case 222: \
+    case 223: \
+    case 224: \
+    case 225: \
+    case 226: \
+    case 227: \
+    case 228: \
+    case 229: \
+    case 230: \
+    case 231: \
+    case 232: \
+    case 233: \
+    case 234: \
+    case 235: \
+    case 236: \
+    case 237: \
+    case 238: \
+    case 239: \
+    case 240: \
+    case 241: \
+    case 242: \
+    case 243: \
+    case 244: \
+    case 245: \
+    case 246: \
+    case 247: \
+    case 248: \
+    case 249: \
+    case 250: \
+    case 251: \
+    case 252: \
+    case 253: \
+    case 254: \
+        ;
+
 #define HAS_ARG(op) ((op) >= HAVE_ARGUMENT)
 
 /* Reserve some bytecodes for internal use in the compiler.
diff --git a/Python/ceval.c b/Python/ceval.c
index 12b0aef..a80583f 100644
--- a/Python/ceval.c
+++ b/Python/ceval.c
@@ -1662,7 +1662,9 @@ _PyEval_EvalFrameDefault(PyThreadState *tstate, _PyInterpreterFrame *frame, int
 #ifdef Py_STATS
     int lastopcode = 0;
 #endif
-    int opcode;        /* Current opcode */
+    // opcode is an 8-bit value to improve the code generated by MSVC
+    // for the big switch below (in combination with the EXTRA_CASES macro).
+    uint8_t opcode;        /* Current opcode */
     int oparg;         /* Current opcode argument, if any */
     _Py_atomic_int * const eval_breaker = &tstate->interp->ceval.eval_breaker;
 
@@ -5645,7 +5647,7 @@ handle_eval_breaker:
 #if USE_COMPUTED_GOTOS
         _unknown_opcode:
 #else
-        default:
+        EXTRA_CASES  // From opcode.h, a 'case' for each unused opcode
 #endif
             fprintf(stderr, "XXX lineno: %d, opcode: %d\n",
                     _PyInterpreterFrame_GetLine(frame),  opcode);
diff --git a/Tools/scripts/generate_opcode_h.py b/Tools/scripts/generate_opcode_h.py
index a13fa77..1b45020 100644
--- a/Tools/scripts/generate_opcode_h.py
+++ b/Tools/scripts/generate_opcode_h.py
@@ -129,6 +129,13 @@ def main(opcode_py, outfile='Include/opcode.h'):
         fobj.write("};\n")
         fobj.write("#endif\n")
 
+        fobj.write("\n")
+        fobj.write("#define EXTRA_CASES \\\n")
+        for i, flag in enumerate(used):
+            if not flag:
+                fobj.write(f"    case {i}: \\\n")
+        fobj.write("        ;\n")
+
         fobj.write(footer)