diff --git a/Lib/test/test_generated_cases.py b/Lib/test/test_generated_cases.py
index 4b17bd5c797375..c93d6fc3d237a2 100644
--- a/Lib/test/test_generated_cases.py
+++ b/Lib/test/test_generated_cases.py
@@ -412,7 +412,7 @@ def test_predictions(self):
frame->instr_ptr = next_instr;
next_instr += 1;
INSTRUCTION_STATS(OP1);
- PREDICTED(OP1);
+ PREDICTED_OP1:;
_PyStackRef res;
res = Py_None;
stack_pointer[-1] = res;
@@ -646,7 +646,7 @@ def test_macro_instruction(self):
frame->instr_ptr = next_instr;
next_instr += 6;
INSTRUCTION_STATS(OP);
- PREDICTED(OP);
+ PREDICTED_OP:;
_Py_CODEUNIT* const this_instr = next_instr - 6;
(void)this_instr;
_PyStackRef left;
diff --git a/Python/ceval_macros.h b/Python/ceval_macros.h
index c37e1cf3afa60e..c0b7c0c637a4c8 100644
--- a/Python/ceval_macros.h
+++ b/Python/ceval_macros.h
@@ -165,35 +165,6 @@ GETITEM(PyObject *v, Py_ssize_t i) {
#define JUMPBY(x) (next_instr += (x))
#define SKIP_OVER(x) (next_instr += (x))
-/* OpCode prediction macros
- Some opcodes tend to come in pairs thus making it possible to
- predict the second code when the first is run. For example,
- COMPARE_OP is often followed by POP_JUMP_IF_FALSE or POP_JUMP_IF_TRUE.
-
- Verifying the prediction costs a single high-speed test of a register
- variable against a constant. If the pairing was good, then the
- processor's own internal branch predication has a high likelihood of
- success, resulting in a nearly zero-overhead transition to the
- next opcode. A successful prediction saves a trip through the eval-loop
- including its unpredictable switch-case branch. Combined with the
- processor's internal branch prediction, a successful PREDICT has the
- effect of making the two opcodes run as if they were a single new opcode
- with the bodies combined.
-
- If collecting opcode statistics, your choices are to either keep the
- predictions turned-on and interpret the results as if some opcodes
- had been combined or turn-off predictions so that the opcode frequency
- counter updates for both opcodes.
-
- Opcode prediction is disabled with threaded code, since the latter allows
- the CPU to record separate branch prediction information for each
- opcode.
-
-*/
-
-#define PREDICT_ID(op) PRED_##op
-#define PREDICTED(op) PREDICT_ID(op):
-
/* Stack manipulation macros */
@@ -259,8 +230,6 @@ GETITEM(PyObject *v, Py_ssize_t i) {
GETLOCAL(i) = value; \
PyStackRef_XCLOSE(tmp); } while (0)
-#define GO_TO_INSTRUCTION(op) goto PREDICT_ID(op)
-
#ifdef Py_STATS
#define UPDATE_MISS_STATS(INSTNAME) \
do { \
@@ -280,7 +249,7 @@ GETITEM(PyObject *v, Py_ssize_t i) {
/* This is only a single jump on release builds! */ \
UPDATE_MISS_STATS((INSTNAME)); \
assert(_PyOpcode_Deopt[opcode] == (INSTNAME)); \
- GO_TO_INSTRUCTION(INSTNAME); \
+ goto PREDICTED_##INSTNAME; \
}
diff --git a/Python/generated_cases.c.h b/Python/generated_cases.c.h
index ec3397e1cf6e5f..4eba3b297e24af 100644
--- a/Python/generated_cases.c.h
+++ b/Python/generated_cases.c.h
@@ -13,7 +13,7 @@
frame->instr_ptr = next_instr;
next_instr += 6;
INSTRUCTION_STATS(BINARY_OP);
- PREDICTED(BINARY_OP);
+ PREDICTED_BINARY_OP:;
_Py_CODEUNIT* const this_instr = next_instr - 6;
(void)this_instr;
_PyStackRef lhs;
@@ -484,7 +484,7 @@
frame->instr_ptr = next_instr;
next_instr += 2;
INSTRUCTION_STATS(BINARY_SUBSCR);
- PREDICTED(BINARY_SUBSCR);
+ PREDICTED_BINARY_SUBSCR:;
_Py_CODEUNIT* const this_instr = next_instr - 2;
(void)this_instr;
_PyStackRef container;
@@ -931,7 +931,7 @@
frame->instr_ptr = next_instr;
next_instr += 4;
INSTRUCTION_STATS(CALL);
- PREDICTED(CALL);
+ PREDICTED_CALL:;
_Py_CODEUNIT* const this_instr = next_instr - 4;
(void)this_instr;
_PyStackRef *callable;
@@ -1707,7 +1707,7 @@
frame->instr_ptr = next_instr;
next_instr += 1;
INSTRUCTION_STATS(CALL_FUNCTION_EX);
- PREDICTED(CALL_FUNCTION_EX);
+ PREDICTED_CALL_FUNCTION_EX:;
_Py_CODEUNIT* const this_instr = next_instr - 1;
(void)this_instr;
_PyStackRef func;
@@ -1960,7 +1960,7 @@
frame->instr_ptr = next_instr;
next_instr += 4;
INSTRUCTION_STATS(CALL_KW);
- PREDICTED(CALL_KW);
+ PREDICTED_CALL_KW:;
_Py_CODEUNIT* const this_instr = next_instr - 4;
(void)this_instr;
_PyStackRef *callable;
@@ -3299,7 +3299,7 @@
frame->instr_ptr = next_instr;
next_instr += 2;
INSTRUCTION_STATS(COMPARE_OP);
- PREDICTED(COMPARE_OP);
+ PREDICTED_COMPARE_OP:;
_Py_CODEUNIT* const this_instr = next_instr - 2;
(void)this_instr;
_PyStackRef left;
@@ -3479,7 +3479,7 @@
frame->instr_ptr = next_instr;
next_instr += 2;
INSTRUCTION_STATS(CONTAINS_OP);
- PREDICTED(CONTAINS_OP);
+ PREDICTED_CONTAINS_OP:;
_Py_CODEUNIT* const this_instr = next_instr - 2;
(void)this_instr;
_PyStackRef left;
@@ -4000,7 +4000,7 @@
frame->instr_ptr = next_instr;
next_instr += 2;
INSTRUCTION_STATS(FOR_ITER);
- PREDICTED(FOR_ITER);
+ PREDICTED_FOR_ITER:;
_Py_CODEUNIT* const this_instr = next_instr - 2;
(void)this_instr;
_PyStackRef iter;
@@ -4631,7 +4631,8 @@
frame->instr_ptr = next_instr;
next_instr += 1;
INSTRUCTION_STATS(INSTRUMENTED_CALL_FUNCTION_EX);
- GO_TO_INSTRUCTION(CALL_FUNCTION_EX);
+
+ goto PREDICTED_CALL_FUNCTION_EX;
}
TARGET(INSTRUMENTED_CALL_KW) {
@@ -4655,7 +4656,7 @@
stack_pointer = _PyFrame_GetStackPointer(frame);
if (err) goto error;
PAUSE_ADAPTIVE_COUNTER(this_instr[1].counter);
- GO_TO_INSTRUCTION(CALL_KW);
+ goto PREDICTED_CALL_KW;
}
TARGET(INSTRUMENTED_END_FOR) {
@@ -4845,7 +4846,7 @@
// cancel out the decrement that will happen in LOAD_SUPER_ATTR; we
// don't want to specialize instrumented instructions
PAUSE_ADAPTIVE_COUNTER(this_instr[1].counter);
- GO_TO_INSTRUCTION(LOAD_SUPER_ATTR);
+ goto PREDICTED_LOAD_SUPER_ATTR;
}
TARGET(INSTRUMENTED_LOAD_SUPER_METHOD) {
@@ -4857,7 +4858,7 @@
// cancel out the decrement that will happen in LOAD_SUPER_ATTR; we
// don't want to specialize instrumented instructions
PAUSE_ADAPTIVE_COUNTER(this_instr[1].counter);
- GO_TO_INSTRUCTION(LOAD_SUPER_METHOD);
+ goto PREDICTED_LOAD_SUPER_METHOD;
}
TARGET(INSTRUMENTED_NOT_TAKEN) {
@@ -5312,7 +5313,7 @@
frame->instr_ptr = next_instr;
next_instr += 10;
INSTRUCTION_STATS(LOAD_ATTR);
- PREDICTED(LOAD_ATTR);
+ PREDICTED_LOAD_ATTR:;
_Py_CODEUNIT* const this_instr = next_instr - 10;
(void)this_instr;
_PyStackRef owner;
@@ -5829,7 +5830,7 @@
frame->instr_ptr = next_instr;
next_instr += 1;
INSTRUCTION_STATS(LOAD_CONST);
- PREDICTED(LOAD_CONST);
+ PREDICTED_LOAD_CONST:;
_Py_CODEUNIT* const this_instr = next_instr - 1;
(void)this_instr;
_PyStackRef value;
@@ -6072,7 +6073,7 @@
frame->instr_ptr = next_instr;
next_instr += 5;
INSTRUCTION_STATS(LOAD_GLOBAL);
- PREDICTED(LOAD_GLOBAL);
+ PREDICTED_LOAD_GLOBAL:;
_Py_CODEUNIT* const this_instr = next_instr - 5;
(void)this_instr;
_PyStackRef *res;
@@ -6222,7 +6223,7 @@
frame->instr_ptr = next_instr;
next_instr += 10;
INSTRUCTION_STATS(LOAD_METHOD);
- PREDICTED(LOAD_METHOD);
+ PREDICTED_LOAD_METHOD:;
_Py_CODEUNIT* const this_instr = next_instr - 10;
(void)this_instr;
_PyStackRef owner;
@@ -6482,7 +6483,7 @@
frame->instr_ptr = next_instr;
next_instr += 2;
INSTRUCTION_STATS(LOAD_SUPER_ATTR);
- PREDICTED(LOAD_SUPER_ATTR);
+ PREDICTED_LOAD_SUPER_ATTR:;
_Py_CODEUNIT* const this_instr = next_instr - 2;
(void)this_instr;
_PyStackRef global_super_st;
@@ -6612,7 +6613,7 @@
frame->instr_ptr = next_instr;
next_instr += 2;
INSTRUCTION_STATS(LOAD_SUPER_METHOD);
- PREDICTED(LOAD_SUPER_METHOD);
+ PREDICTED_LOAD_SUPER_METHOD:;
_Py_CODEUNIT* const this_instr = next_instr - 2;
(void)this_instr;
_PyStackRef global_super_st;
@@ -7182,7 +7183,7 @@
frame->instr_ptr = next_instr;
next_instr += 1;
INSTRUCTION_STATS(RESUME);
- PREDICTED(RESUME);
+ PREDICTED_RESUME:;
_Py_CODEUNIT* const this_instr = next_instr - 1;
(void)this_instr;
// _LOAD_BYTECODE
@@ -7335,7 +7336,7 @@
frame->instr_ptr = next_instr;
next_instr += 2;
INSTRUCTION_STATS(SEND);
- PREDICTED(SEND);
+ PREDICTED_SEND:;
_Py_CODEUNIT* const this_instr = next_instr - 2;
(void)this_instr;
_PyStackRef receiver;
@@ -7576,7 +7577,7 @@
frame->instr_ptr = next_instr;
next_instr += 5;
INSTRUCTION_STATS(STORE_ATTR);
- PREDICTED(STORE_ATTR);
+ PREDICTED_STORE_ATTR:;
_Py_CODEUNIT* const this_instr = next_instr - 5;
(void)this_instr;
_PyStackRef owner;
@@ -7940,7 +7941,7 @@
frame->instr_ptr = next_instr;
next_instr += 2;
INSTRUCTION_STATS(STORE_SUBSCR);
- PREDICTED(STORE_SUBSCR);
+ PREDICTED_STORE_SUBSCR:;
_Py_CODEUNIT* const this_instr = next_instr - 2;
(void)this_instr;
_PyStackRef container;
@@ -8070,7 +8071,7 @@
frame->instr_ptr = next_instr;
next_instr += 4;
INSTRUCTION_STATS(TO_BOOL);
- PREDICTED(TO_BOOL);
+ PREDICTED_TO_BOOL:;
_Py_CODEUNIT* const this_instr = next_instr - 4;
(void)this_instr;
_PyStackRef value;
@@ -8306,7 +8307,7 @@
frame->instr_ptr = next_instr;
next_instr += 2;
INSTRUCTION_STATS(UNPACK_SEQUENCE);
- PREDICTED(UNPACK_SEQUENCE);
+ PREDICTED_UNPACK_SEQUENCE:;
_Py_CODEUNIT* const this_instr = next_instr - 2;
(void)this_instr;
_PyStackRef seq;
diff --git a/Tools/cases_generator/generators_common.py b/Tools/cases_generator/generators_common.py
index c441569b7e70dc..9edf3d4898eb4d 100644
--- a/Tools/cases_generator/generators_common.py
+++ b/Tools/cases_generator/generators_common.py
@@ -127,6 +127,7 @@ def __init__(self, out: CWriter):
"DISPATCH": self.dispatch,
"INSTRUCTION_SIZE": self.instruction_size,
"POP_INPUT": self.pop_input,
+ "GO_TO_INSTRUCTION": self.go_to_instruction,
}
self.out = out
@@ -402,6 +403,23 @@ def sync_sp(
self._print_storage(storage)
return True
+ def go_to_instruction(
+ self,
+ tkn: Token,
+ tkn_iter: TokenIterator,
+ uop: Uop,
+ storage: Storage,
+ inst: Instruction | None,
+ ) -> bool:
+ next(tkn_iter)
+ name = next(tkn_iter)
+ next(tkn_iter)
+ next(tkn_iter)
+ assert name.kind == "IDENTIFIER"
+ self.emit("\n")
+ self.emit(f"goto PREDICTED_{name.text};\n")
+ return True
+
def emit_save(self, storage: Storage) -> None:
storage.save(self.out)
self._print_storage(storage)
diff --git a/Tools/cases_generator/tier1_generator.py b/Tools/cases_generator/tier1_generator.py
index 95876c387bd745..f4409da3d83ccd 100644
--- a/Tools/cases_generator/tier1_generator.py
+++ b/Tools/cases_generator/tier1_generator.py
@@ -158,7 +158,7 @@ def generate_tier1(
out.emit(f"next_instr += {inst.size};\n")
out.emit(f"INSTRUCTION_STATS({name});\n")
if inst.is_target:
- out.emit(f"PREDICTED({name});\n")
+ out.emit(f"PREDICTED_{name}:;\n")
if needs_this:
out.emit(f"_Py_CODEUNIT* const this_instr = next_instr - {inst.size};\n")
out.emit(unused_guard)
pFad - Phonifier reborn
Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.
Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.
Alternative Proxies:
Alternative Proxy
pFad Proxy
pFad v3 Proxy
pFad v4 Proxy