pythonGH-139109: Support switch/case dispatch with the tracing interpreter. (pythonGH-141703)

Author: markshannon

.github/workflows/jit.yml

@@ -57,10 +57,9 @@ jobs:
       fail-fast: false
       matrix:
         target:
-# To re-enable later when we support these.
-#          - i686-pc-windows-msvc/msvc
-#          - x86_64-pc-windows-msvc/msvc
-#          - aarch64-pc-windows-msvc/msvc
+          - 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
@@ -71,16 +70,15 @@ jobs:
         llvm:
           - 21
         include:
-# 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: 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

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 3.15a1 3656 (Add TRACE_RECORD instruction, for platforms with switch based interpreter)
 
 
     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 \

Include/internal/pycore_opcode_metadata.h

@@ -364,7 +364,7 @@ extern void _Py_ClearExecutorDeletionList(PyInterpreterState *interp);
 
 int _PyJit_translate_single_bytecode_to_trace(PyThreadState *tstate, _PyInterpreterFrame *frame, _Py_CODEUNIT *next_instr, int stop_tracing_opcode);
 
-int
+PyAPI_FUNC(int)
 _PyJit_TryInitializeTracing(PyThreadState *tstate, _PyInterpreterFrame *frame,
     _Py_CODEUNIT *curr_instr, _Py_CODEUNIT *start_instr,
     _Py_CODEUNIT *close_loop_instr, int curr_stackdepth, int chain_depth, _PyExitData *exit,

Include/internal/pycore_optimizer.h

@@ -5636,10 +5636,12 @@ dummy_func(
             DISPATCH();
         }
 
-        label(record_previous_inst) {
+        inst(TRACE_RECORD, (--)) {
 #if _Py_TIER2
             assert(IS_JIT_TRACING());
-            int opcode = next_instr->op.code;
+            next_instr = this_instr;
+            frame->instr_ptr = prev_instr;
+            opcode = next_instr->op.code;
             bool stop_tracing = (opcode == WITH_EXCEPT_START ||
                 opcode == RERAISE || opcode == CLEANUP_THROW ||
                 opcode == PUSH_EXC_INFO || opcode == INTERPRETER_EXIT);
@@ -5675,7 +5677,8 @@ dummy_func(
             }
             DISPATCH_GOTO_NON_TRACING();
 #else
-            Py_FatalError("JIT label executed in non-jit build.");
+            (void)prev_instr;
+            Py_FatalError("JIT instruction executed in non-jit build.");
 #endif
         }
 

Include/internal/pycore_uop_ids.h

@@ -1179,6 +1179,10 @@ _PyEval_EvalFrameDefault(PyThreadState *tstate, _PyInterpreterFrame *frame, int
     uint8_t opcode;    /* Current opcode */
     int oparg;         /* Current opcode argument, if any */
     assert(tstate->current_frame == NULL || tstate->current_frame->stackpointer != NULL);
+#if !USE_COMPUTED_GOTOS
+    uint8_t tracing_mode = 0;
+    uint8_t dispatch_code;
+#endif
 #endif
     _PyEntryFrame entry;
 

Include/opcode_ids.h

@@ -134,8 +134,8 @@
 #  define LABEL(name) name:
 #else
 #  define TARGET(op) case op: TARGET_##op:
-#  define DISPATCH_GOTO() goto dispatch_opcode
-#  define DISPATCH_GOTO_NON_TRACING() goto dispatch_opcode
+#  define DISPATCH_GOTO() dispatch_code = opcode | tracing_mode ; goto dispatch_opcode
+#  define DISPATCH_GOTO_NON_TRACING() dispatch_code = opcode; goto dispatch_opcode
 #  define JUMP_TO_LABEL(name) goto name;
 #  define JUMP_TO_PREDICTED(name) goto PREDICTED_##name;
 #  define LABEL(name) name:
@@ -148,9 +148,9 @@
 #  define LEAVE_TRACING() \
     DISPATCH_TABLE_VAR = DISPATCH_TABLE;
 #else
-#  define IS_JIT_TRACING() (0)
-#  define ENTER_TRACING()
-#  define LEAVE_TRACING()
+#  define IS_JIT_TRACING() (tracing_mode != 0)
+#  define ENTER_TRACING() tracing_mode = 255
+#  define LEAVE_TRACING() tracing_mode = 0
 #endif
 
 /* PRE_DISPATCH_GOTO() does lltrace if enabled. Normally a no-op */