linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [PATCH v3] perf jevents: Parse metrics during conversion
@ 2022-12-01  3:41 Ian Rogers
  2022-12-05 12:53 ` Arnaldo Carvalho de Melo
  2022-12-05 15:24 ` John Garry
  0 siblings, 2 replies; 6+ messages in thread
From: Ian Rogers @ 2022-12-01  3:41 UTC (permalink / raw)
  To: Peter Zijlstra, Ingo Molnar, Arnaldo Carvalho de Melo,
	Mark Rutland, Alexander Shishkin, Jiri Olsa, Namhyung Kim,
	Ian Rogers, John Garry, Sumanth Korikkar, Thomas Richter,
	linux-kernel, linux-perf-users
  Cc: Stephane Eranian

Currently the 'MetricExpr' json value is passed from the json
file to the pmu-events.c. This change introduces an expression
tree that is parsed into. The parsing is done largely by using
operator overloading and python's 'eval' function. Two advantages
in doing this are:

1) Broken metrics fail at compile time rather than relying on
   `perf test` to detect. `perf test` remains relevant for checking
   event encoding and actual metric use.

2) The conversion to a string from the tree can minimize the metric's
   string size, for example, preferring 1e6 over 1000000, avoiding
   multiplication by 1 and removing unnecessary whitespace. On x86
   this reduces the string size by 3,050bytes (0.07%).

In future changes it would be possible to programmatically
generate the json expressions (a single line of text and so a
pain to write manually) for an architecture using the expression
tree. This could avoid copy-pasting metrics for all architecture
variants.

Signed-off-by: Ian Rogers <irogers@google.com>

v3. Avoids generic types on standard types like set that aren't
    supported until Python 3.9, fixing an issue with Python 3.6
    reported-by John Garry. v3 also fixes minor pylint issues and adds
    a call to Simplify on the read expression tree.
v2. Improvements to type information.
---
 tools/perf/pmu-events/Build          |   2 +-
 tools/perf/pmu-events/jevents.py     |  12 +-
 tools/perf/pmu-events/metric.py      | 498 +++++++++++++++++++++++++++
 tools/perf/pmu-events/metric_test.py | 148 ++++++++
 4 files changed, 656 insertions(+), 4 deletions(-)
 create mode 100644 tools/perf/pmu-events/metric.py
 create mode 100644 tools/perf/pmu-events/metric_test.py

diff --git a/tools/perf/pmu-events/Build b/tools/perf/pmu-events/Build
index 04ef95174660..15b9e8fdbffa 100644
--- a/tools/perf/pmu-events/Build
+++ b/tools/perf/pmu-events/Build
@@ -21,7 +21,7 @@ $(OUTPUT)pmu-events/pmu-events.c: pmu-events/empty-pmu-events.c
 	$(call rule_mkdir)
 	$(Q)$(call echo-cmd,gen)cp $< $@
 else
-$(OUTPUT)pmu-events/pmu-events.c: $(JSON) $(JSON_TEST) $(JEVENTS_PY)
+$(OUTPUT)pmu-events/pmu-events.c: $(JSON) $(JSON_TEST) $(JEVENTS_PY) pmu-events/metric.py
 	$(call rule_mkdir)
 	$(Q)$(call echo-cmd,gen)$(PYTHON) $(JEVENTS_PY) $(JEVENTS_ARCH) pmu-events/arch $@
 endif
diff --git a/tools/perf/pmu-events/jevents.py b/tools/perf/pmu-events/jevents.py
index 0daa3e007528..4c398e0eeb2f 100755
--- a/tools/perf/pmu-events/jevents.py
+++ b/tools/perf/pmu-events/jevents.py
@@ -4,6 +4,7 @@
 import argparse
 import csv
 import json
+import metric
 import os
 import sys
 from typing import (Callable, Dict, Optional, Sequence, Set, Tuple)
@@ -268,9 +269,10 @@ class JsonEvent:
     self.metric_name = jd.get('MetricName')
     self.metric_group = jd.get('MetricGroup')
     self.metric_constraint = jd.get('MetricConstraint')
-    self.metric_expr = jd.get('MetricExpr')
-    if self.metric_expr:
-      self.metric_expr = self.metric_expr.replace('\\', '\\\\')
+    self.metric_expr = None
+    if 'MetricExpr' in jd:
+       self.metric_expr = metric.ParsePerfJson(jd['MetricExpr']).Simplify()
+
     arch_std = jd.get('ArchStdEvent')
     if precise and self.desc and '(Precise Event)' not in self.desc:
       extra_desc += ' (Must be precise)' if precise == '2' else (' (Precise '
@@ -322,6 +324,10 @@ class JsonEvent:
     s = ''
     for attr in _json_event_attributes:
       x = getattr(self, attr)
+      if x and attr == 'metric_expr':
+        # Convert parsed metric expressions into a string. Slashes
+        # must be doubled in the file.
+        x = x.ToPerfJson().replace('\\', '\\\\')
       s += f'{x}\\000' if x else '\\000'
     return s
 
diff --git a/tools/perf/pmu-events/metric.py b/tools/perf/pmu-events/metric.py
new file mode 100644
index 000000000000..65025a95de00
--- /dev/null
+++ b/tools/perf/pmu-events/metric.py
@@ -0,0 +1,498 @@
+# SPDX-License-Identifier: (LGPL-2.1 OR BSD-2-Clause)
+"""Parse or generate representations of perf metrics."""
+import ast
+import decimal
+import json
+import re
+from typing import Dict, List, Optional, Set, Union
+
+
+class Expression:
+  """Abstract base class of elements in a metric expression."""
+
+  def ToPerfJson(self) -> str:
+    """Returns a perf json file encoded representation."""
+    raise NotImplementedError()
+
+  def ToPython(self) -> str:
+    """Returns a python expr parseable representation."""
+    raise NotImplementedError()
+
+  def Simplify(self):
+    """Returns a simplified version of self."""
+    raise NotImplementedError()
+
+  def Equals(self, other) -> bool:
+    """Returns true when two expressions are the same."""
+    raise NotImplementedError()
+
+  def __str__(self) -> str:
+    return self.ToPerfJson()
+
+  def __or__(self, other: Union[int, float, 'Expression']) -> 'Operator':
+    return Operator('|', self, other)
+
+  def __ror__(self, other: Union[int, float, 'Expression']) -> 'Operator':
+    return Operator('|', other, self)
+
+  def __xor__(self, other: Union[int, float, 'Expression']) -> 'Operator':
+    return Operator('^', self, other)
+
+  def __and__(self, other: Union[int, float, 'Expression']) -> 'Operator':
+    return Operator('&', self, other)
+
+  def __lt__(self, other: Union[int, float, 'Expression']) -> 'Operator':
+    return Operator('<', self, other)
+
+  def __gt__(self, other: Union[int, float, 'Expression']) -> 'Operator':
+    return Operator('>', self, other)
+
+  def __add__(self, other: Union[int, float, 'Expression']) -> 'Operator':
+    return Operator('+', self, other)
+
+  def __radd__(self, other: Union[int, float, 'Expression']) -> 'Operator':
+    return Operator('+', other, self)
+
+  def __sub__(self, other: Union[int, float, 'Expression']) -> 'Operator':
+    return Operator('-', self, other)
+
+  def __rsub__(self, other: Union[int, float, 'Expression']) -> 'Operator':
+    return Operator('-', other, self)
+
+  def __mul__(self, other: Union[int, float, 'Expression']) -> 'Operator':
+    return Operator('*', self, other)
+
+  def __rmul__(self, other: Union[int, float, 'Expression']) -> 'Operator':
+    return Operator('*', other, self)
+
+  def __truediv__(self, other: Union[int, float, 'Expression']) -> 'Operator':
+    return Operator('/', self, other)
+
+  def __rtruediv__(self, other: Union[int, float, 'Expression']) -> 'Operator':
+    return Operator('/', other, self)
+
+  def __mod__(self, other: Union[int, float, 'Expression']) -> 'Operator':
+    return Operator('%', self, other)
+
+
+def _Constify(val: Union[bool, int, float, Expression]) -> Expression:
+  """Used to ensure that the nodes in the expression tree are all Expression."""
+  if isinstance(val, bool):
+    return Constant(1 if val else 0)
+  if isinstance(val, (int, float)):
+    return Constant(val)
+  return val
+
+
+# Simple lookup for operator precedence, used to avoid unnecessary
+# brackets. Precedence matches that of python and the simple expression parser.
+_PRECEDENCE = {
+    '|': 0,
+    '^': 1,
+    '&': 2,
+    '<': 3,
+    '>': 3,
+    '+': 4,
+    '-': 4,
+    '*': 5,
+    '/': 5,
+    '%': 5,
+}
+
+
+class Operator(Expression):
+  """Represents a binary operator in the parse tree."""
+
+  def __init__(self, operator: str, lhs: Union[int, float, Expression],
+               rhs: Union[int, float, Expression]):
+    self.operator = operator
+    self.lhs = _Constify(lhs)
+    self.rhs = _Constify(rhs)
+
+  def Bracket(self,
+              other: Expression,
+              other_str: str,
+              rhs: bool = False) -> str:
+    """If necessary brackets the given other value.
+
+    If ``other`` is an operator then a bracket is necessary when
+    this/self operator has higher precedence. Consider: '(a + b) * c',
+    ``other_str`` will be 'a + b'. A bracket is necessary as without
+    the bracket 'a + b * c' will evaluate 'b * c' first. However, '(a
+    * b) + c' doesn't need a bracket as 'a * b' will always be
+    evaluated first. For 'a / (b * c)' (ie the same precedence level
+    operations) then we add the bracket to best match the original
+    input, but not for '(a / b) * c' where the bracket is unnecessary.
+
+    Args:
+      other (Expression): is a lhs or rhs operator
+      other_str (str): ``other`` in the appropriate string form
+      rhs (bool):  is ``other`` on the RHS
+
+    Returns:
+      str: possibly bracketed other_str
+    """
+    if isinstance(other, Operator):
+      if _PRECEDENCE.get(self.operator, -1) > _PRECEDENCE.get(
+          other.operator, -1):
+        return f'({other_str})'
+      if rhs and _PRECEDENCE.get(self.operator, -1) == _PRECEDENCE.get(
+          other.operator, -1):
+        return f'({other_str})'
+    return other_str
+
+  def ToPerfJson(self):
+    return (f'{self.Bracket(self.lhs, self.lhs.ToPerfJson())} {self.operator} '
+            f'{self.Bracket(self.rhs, self.rhs.ToPerfJson(), True)}')
+
+  def ToPython(self):
+    return (f'{self.Bracket(self.lhs, self.lhs.ToPython())} {self.operator} '
+            f'{self.Bracket(self.rhs, self.rhs.ToPython(), True)}')
+
+  def Simplify(self) -> Expression:
+    lhs = self.lhs.Simplify()
+    rhs = self.rhs.Simplify()
+    if isinstance(lhs, Constant) and isinstance(rhs, Constant):
+      return Constant(ast.literal_eval(lhs + self.operator + rhs))
+
+    if isinstance(self.lhs, Constant):
+      if self.operator in ('+', '|') and lhs.value == '0':
+        return rhs
+
+      if self.operator == '*' and lhs.value == '0':
+        return Constant(0)
+
+      if self.operator == '*' and lhs.value == '1':
+        return rhs
+
+    if isinstance(rhs, Constant):
+      if self.operator in ('+', '|') and rhs.value == '0':
+        return lhs
+
+      if self.operator == '*' and rhs.value == '0':
+        return Constant(0)
+
+      if self.operator == '*' and self.rhs.value == '1':
+        return lhs
+
+    return Operator(self.operator, lhs, rhs)
+
+  def Equals(self, other: Expression) -> bool:
+    if isinstance(other, Operator):
+      return self.operator == other.operator and self.lhs.Equals(
+          other.lhs) and self.rhs.Equals(other.rhs)
+    return False
+
+
+class Select(Expression):
+  """Represents a select ternary in the parse tree."""
+
+  def __init__(self, true_val: Union[int, float, Expression],
+               cond: Union[int, float, Expression],
+               false_val: Union[int, float, Expression]):
+    self.true_val = _Constify(true_val)
+    self.cond = _Constify(cond)
+    self.false_val = _Constify(false_val)
+
+  def ToPerfJson(self):
+    true_str = self.true_val.ToPerfJson()
+    cond_str = self.cond.ToPerfJson()
+    false_str = self.false_val.ToPerfJson()
+    return f'({true_str} if {cond_str} else {false_str})'
+
+  def ToPython(self):
+    return (f'Select({self.true_val.ToPython()}, {self.cond.ToPython()}, '
+            f'{self.false_val.ToPython()})')
+
+  def Simplify(self) -> Expression:
+    cond = self.cond.Simplify()
+    true_val = self.true_val.Simplify()
+    false_val = self.false_val.Simplify()
+    if isinstance(cond, Constant):
+      return false_val if cond.value == '0' else true_val
+
+    if true_val.Equals(false_val):
+      return true_val
+
+    return Select(true_val, cond, false_val)
+
+  def Equals(self, other: Expression) -> bool:
+    if isinstance(other, Select):
+      return self.cond.Equals(other.cond) and self.false_val.Equals(
+          other.false_val) and self.true_val.Equals(other.true_val)
+    return False
+
+
+class Function(Expression):
+  """A function in an expression like min, max, d_ratio."""
+
+  def __init__(self,
+               fn: str,
+               lhs: Union[int, float, Expression],
+               rhs: Optional[Union[int, float, Expression]] = None):
+    self.fn = fn
+    self.lhs = _Constify(lhs)
+    self.rhs = _Constify(rhs)
+
+  def ToPerfJson(self):
+    if self.rhs:
+      return f'{self.fn}({self.lhs.ToPerfJson()}, {self.rhs.ToPerfJson()})'
+    return f'{self.fn}({self.lhs.ToPerfJson()})'
+
+  def ToPython(self):
+    if self.rhs:
+      return f'{self.fn}({self.lhs.ToPython()}, {self.rhs.ToPython()})'
+    return f'{self.fn}({self.lhs.ToPython()})'
+
+  def Simplify(self) -> Expression:
+    lhs = self.lhs.Simplify()
+    rhs = self.rhs.Simplify() if self.rhs else None
+    if isinstance(lhs, Constant) and isinstance(rhs, Constant):
+      if self.fn == 'd_ratio':
+        if rhs.value == '0':
+          return Constant(0)
+        Constant(ast.literal_eval(f'{lhs} / {rhs}'))
+      return Constant(ast.literal_eval(f'{self.fn}({lhs}, {rhs})'))
+
+    return Function(self.fn, lhs, rhs)
+
+  def Equals(self, other: Expression) -> bool:
+    if isinstance(other, Function):
+      return self.fn == other.fn and self.lhs.Equals(
+          other.lhs) and self.rhs.Equals(other.rhs)
+    return False
+
+
+def _FixEscapes(s: str) -> str:
+  s = re.sub(r'([^\\]),', r'\1\\,', s)
+  return re.sub(r'([^\\])=', r'\1\\=', s)
+
+
+class Event(Expression):
+  """An event in an expression."""
+
+  def __init__(self, name: str, legacy_name: str = ''):
+    self.name = _FixEscapes(name)
+    self.legacy_name = _FixEscapes(legacy_name)
+
+  def ToPerfJson(self):
+    result = re.sub('/', '@', self.name)
+    return result
+
+  def ToPython(self):
+    return f'Event(r"{self.name}")'
+
+  def Simplify(self) -> Expression:
+    return self
+
+  def Equals(self, other: Expression) -> bool:
+    return isinstance(other, Event) and self.name == other.name
+
+
+class Constant(Expression):
+  """A constant within the expression tree."""
+
+  def __init__(self, value: Union[float, str]):
+    ctx = decimal.Context()
+    ctx.prec = 20
+    dec = ctx.create_decimal(repr(value) if isinstance(value, float) else value)
+    self.value = dec.normalize().to_eng_string()
+    self.value = self.value.replace('+', '')
+    self.value = self.value.replace('E', 'e')
+
+  def ToPerfJson(self):
+    return self.value
+
+  def ToPython(self):
+    return f'Constant({self.value})'
+
+  def Simplify(self) -> Expression:
+    return self
+
+  def Equals(self, other: Expression) -> bool:
+    return isinstance(other, Constant) and self.value == other.value
+
+
+class Literal(Expression):
+  """A runtime literal within the expression tree."""
+
+  def __init__(self, value: str):
+    self.value = value
+
+  def ToPerfJson(self):
+    return self.value
+
+  def ToPython(self):
+    return f'Literal({self.value})'
+
+  def Simplify(self) -> Expression:
+    return self
+
+  def Equals(self, other: Expression) -> bool:
+    return isinstance(other, Literal) and self.value == other.value
+
+
+def min(lhs: Union[int, float, Expression], rhs: Union[int, float,
+                                                       Expression]) -> Function:
+  # pylint: disable=redefined-builtin
+  # pylint: disable=invalid-name
+  return Function('min', lhs, rhs)
+
+
+def max(lhs: Union[int, float, Expression], rhs: Union[int, float,
+                                                       Expression]) -> Function:
+  # pylint: disable=redefined-builtin
+  # pylint: disable=invalid-name
+  return Function('max', lhs, rhs)
+
+
+def d_ratio(lhs: Union[int, float, Expression],
+            rhs: Union[int, float, Expression]) -> Function:
+  # pylint: disable=redefined-builtin
+  # pylint: disable=invalid-name
+  return Function('d_ratio', lhs, rhs)
+
+
+def source_count(event: Event) -> Function:
+  # pylint: disable=redefined-builtin
+  # pylint: disable=invalid-name
+  return Function('source_count', event)
+
+
+class Metric:
+  """An individual metric that will specifiable on the perf command line."""
+  groups: Set[str]
+  expr: Expression
+  scale_unit: str
+  constraint: bool
+
+  def __init__(self,
+               name: str,
+               description: str,
+               expr: Expression,
+               scale_unit: str,
+               constraint: bool = False):
+    self.name = name
+    self.description = description
+    self.expr = expr.Simplify()
+    # Workraound valid_only_metric hiding certain metrics based on unit.
+    scale_unit = scale_unit.replace('/sec', ' per sec')
+    if scale_unit[0].isdigit():
+      self.scale_unit = scale_unit
+    else:
+      self.scale_unit = f'1{scale_unit}'
+    self.constraint = constraint
+    self.groups = set()
+
+  def __lt__(self, other):
+    """Sort order."""
+    return self.name < other.name
+
+  def AddToMetricGroup(self, group):
+    """Callback used when being added to a MetricGroup."""
+    self.groups.add(group.name)
+
+  def Flatten(self) -> Set['Metric']:
+    """Return a leaf metric."""
+    return set([self])
+
+  def ToPerfJson(self) -> Dict[str, str]:
+    """Return as dictionary for Json generation."""
+    result = {
+        'MetricName': self.name,
+        'MetricGroup': ';'.join(sorted(self.groups)),
+        'BriefDescription': self.description,
+        'MetricExpr': self.expr.ToPerfJson(),
+        'ScaleUnit': self.scale_unit
+    }
+    if self.constraint:
+      result['MetricConstraint'] = 'NO_NMI_WATCHDOG'
+
+    return result
+
+
+class _MetricJsonEncoder(json.JSONEncoder):
+  """Special handling for Metric objects."""
+
+  def default(self, o):
+    if isinstance(o, Metric):
+      return o.ToPerfJson()
+    return json.JSONEncoder.default(self, o)
+
+
+class MetricGroup:
+  """A group of metrics.
+
+  Metric groups may be specificd on the perf command line, but within
+  the json they aren't encoded. Metrics may be in multiple groups
+  which can facilitate arrangements similar to trees.
+  """
+
+  def __init__(self, name: str, metric_list: List[Union[Metric,
+                                                        'MetricGroup']]):
+    self.name = name
+    self.metric_list = metric_list
+    for metric in metric_list:
+      metric.AddToMetricGroup(self)
+
+  def AddToMetricGroup(self, group):
+    """Callback used when a MetricGroup is added into another."""
+    for metric in self.metric_list:
+      metric.AddToMetricGroup(group)
+
+  def Flatten(self) -> Set[Metric]:
+    """Returns a set of all leaf metrics."""
+    result = set()
+    for x in self.metric_list:
+      result = result.union(x.Flatten())
+
+    return result
+
+  def ToPerfJson(self) -> str:
+    return json.dumps(sorted(self.Flatten()), indent=2, cls=_MetricJsonEncoder)
+
+  def __str__(self) -> str:
+    return self.ToPerfJson()
+
+
+class _RewriteIfExpToSelect(ast.NodeTransformer):
+
+  def visit_IfExp(self, node):
+    # pylint: disable=invalid-name
+    call = ast.Call(
+        func=ast.Name(id='Select', ctx=ast.Load()),
+        args=[node.body, node.test, node.orelse],
+        keywords=[])
+    ast.copy_location(call, node.test)
+    return call
+
+
+def ParsePerfJson(orig: str) -> Expression:
+  """A simple json metric expression decoder.
+
+  Converts a json encoded metric expression by way of python's ast and
+  eval routine. First tokens are mapped to Event calls, then
+  accidentally converted keywords or literals are mapped to their
+  appropriate calls. Python's ast is used to match if-else that can't
+  be handled via operator overloading. Finally the ast is evaluated.
+
+  Args:
+    orig (str): String to parse.
+
+  Returns:
+    Expression: The parsed string.
+  """
+  # pylint: disable=eval-used
+  py = orig.strip()
+  py = re.sub(r'([a-zA-Z][^-+/\* \\\(\),]*(?:\\.[^-+/\* \\\(\),]*)*)',
+              r'Event(r"\1")', py)
+  py = re.sub(r'#Event\(r"([^"]*)"\)', r'Literal("#\1")', py)
+  py = re.sub(r'([0-9]+)Event\(r"(e[0-9]+)"\)', r'\1\2', py)
+  keywords = ['if', 'else', 'min', 'max', 'd_ratio', 'source_count']
+  for kw in keywords:
+    py = re.sub(rf'Event\(r"{kw}"\)', kw, py)
+
+  parsed = ast.parse(py, mode='eval')
+  _RewriteIfExpToSelect().visit(parsed)
+  parsed = ast.fix_missing_locations(parsed)
+  return _Constify(eval(compile(parsed, orig, 'eval')))
diff --git a/tools/perf/pmu-events/metric_test.py b/tools/perf/pmu-events/metric_test.py
new file mode 100644
index 000000000000..66be55893de0
--- /dev/null
+++ b/tools/perf/pmu-events/metric_test.py
@@ -0,0 +1,148 @@
+# SPDX-License-Identifier: (LGPL-2.1 OR BSD-2-Clause)
+import unittest
+from metric import Constant
+from metric import Event
+from metric import ParsePerfJson
+
+
+class TestMetricExpressions(unittest.TestCase):
+
+  def test_Operators(self):
+    a = Event('a')
+    b = Event('b')
+    self.assertEqual((a | b).ToPerfJson(), 'a | b')
+    self.assertEqual((a ^ b).ToPerfJson(), 'a ^ b')
+    self.assertEqual((a & b).ToPerfJson(), 'a & b')
+    self.assertEqual((a < b).ToPerfJson(), 'a < b')
+    self.assertEqual((a > b).ToPerfJson(), 'a > b')
+    self.assertEqual((a + b).ToPerfJson(), 'a + b')
+    self.assertEqual((a - b).ToPerfJson(), 'a - b')
+    self.assertEqual((a * b).ToPerfJson(), 'a * b')
+    self.assertEqual((a / b).ToPerfJson(), 'a / b')
+    self.assertEqual((a % b).ToPerfJson(), 'a % b')
+    one = Constant(1)
+    self.assertEqual((a + one).ToPerfJson(), 'a + 1')
+
+  def test_Brackets(self):
+    a = Event('a')
+    b = Event('b')
+    c = Event('c')
+    self.assertEqual((a * b + c).ToPerfJson(), 'a * b + c')
+    self.assertEqual((a + b * c).ToPerfJson(), 'a + b * c')
+    self.assertEqual(((a + a) + a).ToPerfJson(), 'a + a + a')
+    self.assertEqual(((a + b) * c).ToPerfJson(), '(a + b) * c')
+    self.assertEqual((a + (b * c)).ToPerfJson(), 'a + b * c')
+    self.assertEqual(((a / b) * c).ToPerfJson(), 'a / b * c')
+    self.assertEqual((a / (b * c)).ToPerfJson(), 'a / (b * c)')
+
+  def test_ParsePerfJson(self):
+    # Based on an example of a real metric.
+    before = '(a + b + c + d) / (2 * e)'
+    after = before
+    self.assertEqual(ParsePerfJson(before).ToPerfJson(), after)
+
+    # Parsing should handle events with '-' in their name. Note, in
+    # the json file the '\' are doubled to '\\'.
+    before = r'topdown\-fe\-bound / topdown\-slots - 1'
+    after = before
+    self.assertEqual(ParsePerfJson(before).ToPerfJson(), after)
+
+    # Parsing should handle escaped modifiers. Note, in the json file
+    # the '\' are doubled to '\\'.
+    before = r'arb@event\=0x81\,umask\=0x1@ + arb@event\=0x84\,umask\=0x1@'
+    after = before
+    self.assertEqual(ParsePerfJson(before).ToPerfJson(), after)
+
+    # Parsing should handle exponents in numbers.
+    before = r'a + 1e12 + b'
+    after = before
+    self.assertEqual(ParsePerfJson(before).ToPerfJson(), after)
+
+  def test_IfElseTests(self):
+    # if-else needs rewriting to Select and back.
+    before = r'Event1 if #smt_on else Event2'
+    after = f'({before})'
+    self.assertEqual(ParsePerfJson(before).ToPerfJson(), after)
+
+    before = r'Event1 if 0 else Event2'
+    after = f'({before})'
+    self.assertEqual(ParsePerfJson(before).ToPerfJson(), after)
+
+    before = r'Event1 if 1 else Event2'
+    after = f'({before})'
+    self.assertEqual(ParsePerfJson(before).ToPerfJson(), after)
+
+    # Ensure the select is evaluate last.
+    before = r'Event1 + 1 if Event2 < 2 else Event3 + 3'
+    after = (r'Select(Event(r"Event1") + Constant(1), Event(r"Event2") < '
+             r'Constant(2), Event(r"Event3") + Constant(3))')
+    self.assertEqual(ParsePerfJson(before).ToPython(), after)
+
+    before = r'Event1 > 1 if Event2 < 2 else Event3 > 3'
+    after = (r'Select(Event(r"Event1") > Constant(1), Event(r"Event2") < '
+             r'Constant(2), Event(r"Event3") > Constant(3))')
+    self.assertEqual(ParsePerfJson(before).ToPython(), after)
+
+    before = r'min(a + b if c > 1 else c + d, e + f)'
+    after = r'min((a + b if c > 1 else c + d), e + f)'
+    self.assertEqual(ParsePerfJson(before).ToPerfJson(), after)
+
+  def test_ToPython(self):
+    # pylint: disable=eval-used
+    # Based on an example of a real metric.
+    before = '(a + b + c + d) / (2 * e)'
+    py = ParsePerfJson(before).ToPython()
+    after = eval(py).ToPerfJson()
+    self.assertEqual(before, after)
+
+  def test_Simplify(self):
+    before = '1 + 2 + 3'
+    after = '6'
+    self.assertEqual(ParsePerfJson(before).Simplify().ToPerfJson(), after)
+
+    before = 'a + 0'
+    after = 'a'
+    self.assertEqual(ParsePerfJson(before).Simplify().ToPerfJson(), after)
+
+    before = '0 + a'
+    after = 'a'
+    self.assertEqual(ParsePerfJson(before).Simplify().ToPerfJson(), after)
+
+    before = 'a | 0'
+    after = 'a'
+    self.assertEqual(ParsePerfJson(before).Simplify().ToPerfJson(), after)
+
+    before = '0 | a'
+    after = 'a'
+    self.assertEqual(ParsePerfJson(before).Simplify().ToPerfJson(), after)
+
+    before = 'a * 0'
+    after = '0'
+    self.assertEqual(ParsePerfJson(before).Simplify().ToPerfJson(), after)
+
+    before = '0 * a'
+    after = '0'
+    self.assertEqual(ParsePerfJson(before).Simplify().ToPerfJson(), after)
+
+    before = 'a * 1'
+    after = 'a'
+    self.assertEqual(ParsePerfJson(before).Simplify().ToPerfJson(), after)
+
+    before = '1 * a'
+    after = 'a'
+    self.assertEqual(ParsePerfJson(before).Simplify().ToPerfJson(), after)
+
+    before = 'a if 0 else b'
+    after = 'b'
+    self.assertEqual(ParsePerfJson(before).Simplify().ToPerfJson(), after)
+
+    before = 'a if 1 else b'
+    after = 'a'
+    self.assertEqual(ParsePerfJson(before).Simplify().ToPerfJson(), after)
+
+    before = 'a if b else a'
+    after = 'a'
+    self.assertEqual(ParsePerfJson(before).Simplify().ToPerfJson(), after)
+
+if __name__ == '__main__':
+  unittest.main()
-- 
2.38.1.584.g0f3c55d4c2-goog


^ permalink raw reply related	[flat|nested] 6+ messages in thread

* Re: [PATCH v3] perf jevents: Parse metrics during conversion
  2022-12-01  3:41 [PATCH v3] perf jevents: Parse metrics during conversion Ian Rogers
@ 2022-12-05 12:53 ` Arnaldo Carvalho de Melo
  2022-12-05 12:56   ` John Garry
  2022-12-05 15:24 ` John Garry
  1 sibling, 1 reply; 6+ messages in thread
From: Arnaldo Carvalho de Melo @ 2022-12-05 12:53 UTC (permalink / raw)
  To: Ian Rogers
  Cc: Peter Zijlstra, Ingo Molnar, Mark Rutland, Alexander Shishkin,
	Jiri Olsa, Namhyung Kim, John Garry, Sumanth Korikkar,
	Thomas Richter, linux-kernel, linux-perf-users, Stephane Eranian

Em Wed, Nov 30, 2022 at 07:41:38PM -0800, Ian Rogers escreveu:
> Currently the 'MetricExpr' json value is passed from the json
> file to the pmu-events.c. This change introduces an expression
> tree that is parsed into. The parsing is done largely by using
> operator overloading and python's 'eval' function. Two advantages
> in doing this are:
> 
> 1) Broken metrics fail at compile time rather than relying on
>    `perf test` to detect. `perf test` remains relevant for checking
>    event encoding and actual metric use.
> 
> 2) The conversion to a string from the tree can minimize the metric's
>    string size, for example, preferring 1e6 over 1000000, avoiding
>    multiplication by 1 and removing unnecessary whitespace. On x86
>    this reduces the string size by 3,050bytes (0.07%).
> 
> In future changes it would be possible to programmatically
> generate the json expressions (a single line of text and so a
> pain to write manually) for an architecture using the expression
> tree. This could avoid copy-pasting metrics for all architecture
> variants.
> 
> Signed-off-by: Ian Rogers <irogers@google.com>
> 
> v3. Avoids generic types on standard types like set that aren't
>     supported until Python 3.9, fixing an issue with Python 3.6
>     reported-by John Garry. v3 also fixes minor pylint issues and adds
>     a call to Simplify on the read expression tree.

Cool, John looked reviewed it (may I add the tag?). Applying to my local
tree, will do some light testing.

- Arnaldo

> v2. Improvements to type information.
> ---
>  tools/perf/pmu-events/Build          |   2 +-
>  tools/perf/pmu-events/jevents.py     |  12 +-
>  tools/perf/pmu-events/metric.py      | 498 +++++++++++++++++++++++++++
>  tools/perf/pmu-events/metric_test.py | 148 ++++++++
>  4 files changed, 656 insertions(+), 4 deletions(-)
>  create mode 100644 tools/perf/pmu-events/metric.py
>  create mode 100644 tools/perf/pmu-events/metric_test.py
> 
> diff --git a/tools/perf/pmu-events/Build b/tools/perf/pmu-events/Build
> index 04ef95174660..15b9e8fdbffa 100644
> --- a/tools/perf/pmu-events/Build
> +++ b/tools/perf/pmu-events/Build
> @@ -21,7 +21,7 @@ $(OUTPUT)pmu-events/pmu-events.c: pmu-events/empty-pmu-events.c
>  	$(call rule_mkdir)
>  	$(Q)$(call echo-cmd,gen)cp $< $@
>  else
> -$(OUTPUT)pmu-events/pmu-events.c: $(JSON) $(JSON_TEST) $(JEVENTS_PY)
> +$(OUTPUT)pmu-events/pmu-events.c: $(JSON) $(JSON_TEST) $(JEVENTS_PY) pmu-events/metric.py
>  	$(call rule_mkdir)
>  	$(Q)$(call echo-cmd,gen)$(PYTHON) $(JEVENTS_PY) $(JEVENTS_ARCH) pmu-events/arch $@
>  endif
> diff --git a/tools/perf/pmu-events/jevents.py b/tools/perf/pmu-events/jevents.py
> index 0daa3e007528..4c398e0eeb2f 100755
> --- a/tools/perf/pmu-events/jevents.py
> +++ b/tools/perf/pmu-events/jevents.py
> @@ -4,6 +4,7 @@
>  import argparse
>  import csv
>  import json
> +import metric
>  import os
>  import sys
>  from typing import (Callable, Dict, Optional, Sequence, Set, Tuple)
> @@ -268,9 +269,10 @@ class JsonEvent:
>      self.metric_name = jd.get('MetricName')
>      self.metric_group = jd.get('MetricGroup')
>      self.metric_constraint = jd.get('MetricConstraint')
> -    self.metric_expr = jd.get('MetricExpr')
> -    if self.metric_expr:
> -      self.metric_expr = self.metric_expr.replace('\\', '\\\\')
> +    self.metric_expr = None
> +    if 'MetricExpr' in jd:
> +       self.metric_expr = metric.ParsePerfJson(jd['MetricExpr']).Simplify()
> +
>      arch_std = jd.get('ArchStdEvent')
>      if precise and self.desc and '(Precise Event)' not in self.desc:
>        extra_desc += ' (Must be precise)' if precise == '2' else (' (Precise '
> @@ -322,6 +324,10 @@ class JsonEvent:
>      s = ''
>      for attr in _json_event_attributes:
>        x = getattr(self, attr)
> +      if x and attr == 'metric_expr':
> +        # Convert parsed metric expressions into a string. Slashes
> +        # must be doubled in the file.
> +        x = x.ToPerfJson().replace('\\', '\\\\')
>        s += f'{x}\\000' if x else '\\000'
>      return s
>  
> diff --git a/tools/perf/pmu-events/metric.py b/tools/perf/pmu-events/metric.py
> new file mode 100644
> index 000000000000..65025a95de00
> --- /dev/null
> +++ b/tools/perf/pmu-events/metric.py
> @@ -0,0 +1,498 @@
> +# SPDX-License-Identifier: (LGPL-2.1 OR BSD-2-Clause)
> +"""Parse or generate representations of perf metrics."""
> +import ast
> +import decimal
> +import json
> +import re
> +from typing import Dict, List, Optional, Set, Union
> +
> +
> +class Expression:
> +  """Abstract base class of elements in a metric expression."""
> +
> +  def ToPerfJson(self) -> str:
> +    """Returns a perf json file encoded representation."""
> +    raise NotImplementedError()
> +
> +  def ToPython(self) -> str:
> +    """Returns a python expr parseable representation."""
> +    raise NotImplementedError()
> +
> +  def Simplify(self):
> +    """Returns a simplified version of self."""
> +    raise NotImplementedError()
> +
> +  def Equals(self, other) -> bool:
> +    """Returns true when two expressions are the same."""
> +    raise NotImplementedError()
> +
> +  def __str__(self) -> str:
> +    return self.ToPerfJson()
> +
> +  def __or__(self, other: Union[int, float, 'Expression']) -> 'Operator':
> +    return Operator('|', self, other)
> +
> +  def __ror__(self, other: Union[int, float, 'Expression']) -> 'Operator':
> +    return Operator('|', other, self)
> +
> +  def __xor__(self, other: Union[int, float, 'Expression']) -> 'Operator':
> +    return Operator('^', self, other)
> +
> +  def __and__(self, other: Union[int, float, 'Expression']) -> 'Operator':
> +    return Operator('&', self, other)
> +
> +  def __lt__(self, other: Union[int, float, 'Expression']) -> 'Operator':
> +    return Operator('<', self, other)
> +
> +  def __gt__(self, other: Union[int, float, 'Expression']) -> 'Operator':
> +    return Operator('>', self, other)
> +
> +  def __add__(self, other: Union[int, float, 'Expression']) -> 'Operator':
> +    return Operator('+', self, other)
> +
> +  def __radd__(self, other: Union[int, float, 'Expression']) -> 'Operator':
> +    return Operator('+', other, self)
> +
> +  def __sub__(self, other: Union[int, float, 'Expression']) -> 'Operator':
> +    return Operator('-', self, other)
> +
> +  def __rsub__(self, other: Union[int, float, 'Expression']) -> 'Operator':
> +    return Operator('-', other, self)
> +
> +  def __mul__(self, other: Union[int, float, 'Expression']) -> 'Operator':
> +    return Operator('*', self, other)
> +
> +  def __rmul__(self, other: Union[int, float, 'Expression']) -> 'Operator':
> +    return Operator('*', other, self)
> +
> +  def __truediv__(self, other: Union[int, float, 'Expression']) -> 'Operator':
> +    return Operator('/', self, other)
> +
> +  def __rtruediv__(self, other: Union[int, float, 'Expression']) -> 'Operator':
> +    return Operator('/', other, self)
> +
> +  def __mod__(self, other: Union[int, float, 'Expression']) -> 'Operator':
> +    return Operator('%', self, other)
> +
> +
> +def _Constify(val: Union[bool, int, float, Expression]) -> Expression:
> +  """Used to ensure that the nodes in the expression tree are all Expression."""
> +  if isinstance(val, bool):
> +    return Constant(1 if val else 0)
> +  if isinstance(val, (int, float)):
> +    return Constant(val)
> +  return val
> +
> +
> +# Simple lookup for operator precedence, used to avoid unnecessary
> +# brackets. Precedence matches that of python and the simple expression parser.
> +_PRECEDENCE = {
> +    '|': 0,
> +    '^': 1,
> +    '&': 2,
> +    '<': 3,
> +    '>': 3,
> +    '+': 4,
> +    '-': 4,
> +    '*': 5,
> +    '/': 5,
> +    '%': 5,
> +}
> +
> +
> +class Operator(Expression):
> +  """Represents a binary operator in the parse tree."""
> +
> +  def __init__(self, operator: str, lhs: Union[int, float, Expression],
> +               rhs: Union[int, float, Expression]):
> +    self.operator = operator
> +    self.lhs = _Constify(lhs)
> +    self.rhs = _Constify(rhs)
> +
> +  def Bracket(self,
> +              other: Expression,
> +              other_str: str,
> +              rhs: bool = False) -> str:
> +    """If necessary brackets the given other value.
> +
> +    If ``other`` is an operator then a bracket is necessary when
> +    this/self operator has higher precedence. Consider: '(a + b) * c',
> +    ``other_str`` will be 'a + b'. A bracket is necessary as without
> +    the bracket 'a + b * c' will evaluate 'b * c' first. However, '(a
> +    * b) + c' doesn't need a bracket as 'a * b' will always be
> +    evaluated first. For 'a / (b * c)' (ie the same precedence level
> +    operations) then we add the bracket to best match the original
> +    input, but not for '(a / b) * c' where the bracket is unnecessary.
> +
> +    Args:
> +      other (Expression): is a lhs or rhs operator
> +      other_str (str): ``other`` in the appropriate string form
> +      rhs (bool):  is ``other`` on the RHS
> +
> +    Returns:
> +      str: possibly bracketed other_str
> +    """
> +    if isinstance(other, Operator):
> +      if _PRECEDENCE.get(self.operator, -1) > _PRECEDENCE.get(
> +          other.operator, -1):
> +        return f'({other_str})'
> +      if rhs and _PRECEDENCE.get(self.operator, -1) == _PRECEDENCE.get(
> +          other.operator, -1):
> +        return f'({other_str})'
> +    return other_str
> +
> +  def ToPerfJson(self):
> +    return (f'{self.Bracket(self.lhs, self.lhs.ToPerfJson())} {self.operator} '
> +            f'{self.Bracket(self.rhs, self.rhs.ToPerfJson(), True)}')
> +
> +  def ToPython(self):
> +    return (f'{self.Bracket(self.lhs, self.lhs.ToPython())} {self.operator} '
> +            f'{self.Bracket(self.rhs, self.rhs.ToPython(), True)}')
> +
> +  def Simplify(self) -> Expression:
> +    lhs = self.lhs.Simplify()
> +    rhs = self.rhs.Simplify()
> +    if isinstance(lhs, Constant) and isinstance(rhs, Constant):
> +      return Constant(ast.literal_eval(lhs + self.operator + rhs))
> +
> +    if isinstance(self.lhs, Constant):
> +      if self.operator in ('+', '|') and lhs.value == '0':
> +        return rhs
> +
> +      if self.operator == '*' and lhs.value == '0':
> +        return Constant(0)
> +
> +      if self.operator == '*' and lhs.value == '1':
> +        return rhs
> +
> +    if isinstance(rhs, Constant):
> +      if self.operator in ('+', '|') and rhs.value == '0':
> +        return lhs
> +
> +      if self.operator == '*' and rhs.value == '0':
> +        return Constant(0)
> +
> +      if self.operator == '*' and self.rhs.value == '1':
> +        return lhs
> +
> +    return Operator(self.operator, lhs, rhs)
> +
> +  def Equals(self, other: Expression) -> bool:
> +    if isinstance(other, Operator):
> +      return self.operator == other.operator and self.lhs.Equals(
> +          other.lhs) and self.rhs.Equals(other.rhs)
> +    return False
> +
> +
> +class Select(Expression):
> +  """Represents a select ternary in the parse tree."""
> +
> +  def __init__(self, true_val: Union[int, float, Expression],
> +               cond: Union[int, float, Expression],
> +               false_val: Union[int, float, Expression]):
> +    self.true_val = _Constify(true_val)
> +    self.cond = _Constify(cond)
> +    self.false_val = _Constify(false_val)
> +
> +  def ToPerfJson(self):
> +    true_str = self.true_val.ToPerfJson()
> +    cond_str = self.cond.ToPerfJson()
> +    false_str = self.false_val.ToPerfJson()
> +    return f'({true_str} if {cond_str} else {false_str})'
> +
> +  def ToPython(self):
> +    return (f'Select({self.true_val.ToPython()}, {self.cond.ToPython()}, '
> +            f'{self.false_val.ToPython()})')
> +
> +  def Simplify(self) -> Expression:
> +    cond = self.cond.Simplify()
> +    true_val = self.true_val.Simplify()
> +    false_val = self.false_val.Simplify()
> +    if isinstance(cond, Constant):
> +      return false_val if cond.value == '0' else true_val
> +
> +    if true_val.Equals(false_val):
> +      return true_val
> +
> +    return Select(true_val, cond, false_val)
> +
> +  def Equals(self, other: Expression) -> bool:
> +    if isinstance(other, Select):
> +      return self.cond.Equals(other.cond) and self.false_val.Equals(
> +          other.false_val) and self.true_val.Equals(other.true_val)
> +    return False
> +
> +
> +class Function(Expression):
> +  """A function in an expression like min, max, d_ratio."""
> +
> +  def __init__(self,
> +               fn: str,
> +               lhs: Union[int, float, Expression],
> +               rhs: Optional[Union[int, float, Expression]] = None):
> +    self.fn = fn
> +    self.lhs = _Constify(lhs)
> +    self.rhs = _Constify(rhs)
> +
> +  def ToPerfJson(self):
> +    if self.rhs:
> +      return f'{self.fn}({self.lhs.ToPerfJson()}, {self.rhs.ToPerfJson()})'
> +    return f'{self.fn}({self.lhs.ToPerfJson()})'
> +
> +  def ToPython(self):
> +    if self.rhs:
> +      return f'{self.fn}({self.lhs.ToPython()}, {self.rhs.ToPython()})'
> +    return f'{self.fn}({self.lhs.ToPython()})'
> +
> +  def Simplify(self) -> Expression:
> +    lhs = self.lhs.Simplify()
> +    rhs = self.rhs.Simplify() if self.rhs else None
> +    if isinstance(lhs, Constant) and isinstance(rhs, Constant):
> +      if self.fn == 'd_ratio':
> +        if rhs.value == '0':
> +          return Constant(0)
> +        Constant(ast.literal_eval(f'{lhs} / {rhs}'))
> +      return Constant(ast.literal_eval(f'{self.fn}({lhs}, {rhs})'))
> +
> +    return Function(self.fn, lhs, rhs)
> +
> +  def Equals(self, other: Expression) -> bool:
> +    if isinstance(other, Function):
> +      return self.fn == other.fn and self.lhs.Equals(
> +          other.lhs) and self.rhs.Equals(other.rhs)
> +    return False
> +
> +
> +def _FixEscapes(s: str) -> str:
> +  s = re.sub(r'([^\\]),', r'\1\\,', s)
> +  return re.sub(r'([^\\])=', r'\1\\=', s)
> +
> +
> +class Event(Expression):
> +  """An event in an expression."""
> +
> +  def __init__(self, name: str, legacy_name: str = ''):
> +    self.name = _FixEscapes(name)
> +    self.legacy_name = _FixEscapes(legacy_name)
> +
> +  def ToPerfJson(self):
> +    result = re.sub('/', '@', self.name)
> +    return result
> +
> +  def ToPython(self):
> +    return f'Event(r"{self.name}")'
> +
> +  def Simplify(self) -> Expression:
> +    return self
> +
> +  def Equals(self, other: Expression) -> bool:
> +    return isinstance(other, Event) and self.name == other.name
> +
> +
> +class Constant(Expression):
> +  """A constant within the expression tree."""
> +
> +  def __init__(self, value: Union[float, str]):
> +    ctx = decimal.Context()
> +    ctx.prec = 20
> +    dec = ctx.create_decimal(repr(value) if isinstance(value, float) else value)
> +    self.value = dec.normalize().to_eng_string()
> +    self.value = self.value.replace('+', '')
> +    self.value = self.value.replace('E', 'e')
> +
> +  def ToPerfJson(self):
> +    return self.value
> +
> +  def ToPython(self):
> +    return f'Constant({self.value})'
> +
> +  def Simplify(self) -> Expression:
> +    return self
> +
> +  def Equals(self, other: Expression) -> bool:
> +    return isinstance(other, Constant) and self.value == other.value
> +
> +
> +class Literal(Expression):
> +  """A runtime literal within the expression tree."""
> +
> +  def __init__(self, value: str):
> +    self.value = value
> +
> +  def ToPerfJson(self):
> +    return self.value
> +
> +  def ToPython(self):
> +    return f'Literal({self.value})'
> +
> +  def Simplify(self) -> Expression:
> +    return self
> +
> +  def Equals(self, other: Expression) -> bool:
> +    return isinstance(other, Literal) and self.value == other.value
> +
> +
> +def min(lhs: Union[int, float, Expression], rhs: Union[int, float,
> +                                                       Expression]) -> Function:
> +  # pylint: disable=redefined-builtin
> +  # pylint: disable=invalid-name
> +  return Function('min', lhs, rhs)
> +
> +
> +def max(lhs: Union[int, float, Expression], rhs: Union[int, float,
> +                                                       Expression]) -> Function:
> +  # pylint: disable=redefined-builtin
> +  # pylint: disable=invalid-name
> +  return Function('max', lhs, rhs)
> +
> +
> +def d_ratio(lhs: Union[int, float, Expression],
> +            rhs: Union[int, float, Expression]) -> Function:
> +  # pylint: disable=redefined-builtin
> +  # pylint: disable=invalid-name
> +  return Function('d_ratio', lhs, rhs)
> +
> +
> +def source_count(event: Event) -> Function:
> +  # pylint: disable=redefined-builtin
> +  # pylint: disable=invalid-name
> +  return Function('source_count', event)
> +
> +
> +class Metric:
> +  """An individual metric that will specifiable on the perf command line."""
> +  groups: Set[str]
> +  expr: Expression
> +  scale_unit: str
> +  constraint: bool
> +
> +  def __init__(self,
> +               name: str,
> +               description: str,
> +               expr: Expression,
> +               scale_unit: str,
> +               constraint: bool = False):
> +    self.name = name
> +    self.description = description
> +    self.expr = expr.Simplify()
> +    # Workraound valid_only_metric hiding certain metrics based on unit.
> +    scale_unit = scale_unit.replace('/sec', ' per sec')
> +    if scale_unit[0].isdigit():
> +      self.scale_unit = scale_unit
> +    else:
> +      self.scale_unit = f'1{scale_unit}'
> +    self.constraint = constraint
> +    self.groups = set()
> +
> +  def __lt__(self, other):
> +    """Sort order."""
> +    return self.name < other.name
> +
> +  def AddToMetricGroup(self, group):
> +    """Callback used when being added to a MetricGroup."""
> +    self.groups.add(group.name)
> +
> +  def Flatten(self) -> Set['Metric']:
> +    """Return a leaf metric."""
> +    return set([self])
> +
> +  def ToPerfJson(self) -> Dict[str, str]:
> +    """Return as dictionary for Json generation."""
> +    result = {
> +        'MetricName': self.name,
> +        'MetricGroup': ';'.join(sorted(self.groups)),
> +        'BriefDescription': self.description,
> +        'MetricExpr': self.expr.ToPerfJson(),
> +        'ScaleUnit': self.scale_unit
> +    }
> +    if self.constraint:
> +      result['MetricConstraint'] = 'NO_NMI_WATCHDOG'
> +
> +    return result
> +
> +
> +class _MetricJsonEncoder(json.JSONEncoder):
> +  """Special handling for Metric objects."""
> +
> +  def default(self, o):
> +    if isinstance(o, Metric):
> +      return o.ToPerfJson()
> +    return json.JSONEncoder.default(self, o)
> +
> +
> +class MetricGroup:
> +  """A group of metrics.
> +
> +  Metric groups may be specificd on the perf command line, but within
> +  the json they aren't encoded. Metrics may be in multiple groups
> +  which can facilitate arrangements similar to trees.
> +  """
> +
> +  def __init__(self, name: str, metric_list: List[Union[Metric,
> +                                                        'MetricGroup']]):
> +    self.name = name
> +    self.metric_list = metric_list
> +    for metric in metric_list:
> +      metric.AddToMetricGroup(self)
> +
> +  def AddToMetricGroup(self, group):
> +    """Callback used when a MetricGroup is added into another."""
> +    for metric in self.metric_list:
> +      metric.AddToMetricGroup(group)
> +
> +  def Flatten(self) -> Set[Metric]:
> +    """Returns a set of all leaf metrics."""
> +    result = set()
> +    for x in self.metric_list:
> +      result = result.union(x.Flatten())
> +
> +    return result
> +
> +  def ToPerfJson(self) -> str:
> +    return json.dumps(sorted(self.Flatten()), indent=2, cls=_MetricJsonEncoder)
> +
> +  def __str__(self) -> str:
> +    return self.ToPerfJson()
> +
> +
> +class _RewriteIfExpToSelect(ast.NodeTransformer):
> +
> +  def visit_IfExp(self, node):
> +    # pylint: disable=invalid-name
> +    call = ast.Call(
> +        func=ast.Name(id='Select', ctx=ast.Load()),
> +        args=[node.body, node.test, node.orelse],
> +        keywords=[])
> +    ast.copy_location(call, node.test)
> +    return call
> +
> +
> +def ParsePerfJson(orig: str) -> Expression:
> +  """A simple json metric expression decoder.
> +
> +  Converts a json encoded metric expression by way of python's ast and
> +  eval routine. First tokens are mapped to Event calls, then
> +  accidentally converted keywords or literals are mapped to their
> +  appropriate calls. Python's ast is used to match if-else that can't
> +  be handled via operator overloading. Finally the ast is evaluated.
> +
> +  Args:
> +    orig (str): String to parse.
> +
> +  Returns:
> +    Expression: The parsed string.
> +  """
> +  # pylint: disable=eval-used
> +  py = orig.strip()
> +  py = re.sub(r'([a-zA-Z][^-+/\* \\\(\),]*(?:\\.[^-+/\* \\\(\),]*)*)',
> +              r'Event(r"\1")', py)
> +  py = re.sub(r'#Event\(r"([^"]*)"\)', r'Literal("#\1")', py)
> +  py = re.sub(r'([0-9]+)Event\(r"(e[0-9]+)"\)', r'\1\2', py)
> +  keywords = ['if', 'else', 'min', 'max', 'd_ratio', 'source_count']
> +  for kw in keywords:
> +    py = re.sub(rf'Event\(r"{kw}"\)', kw, py)
> +
> +  parsed = ast.parse(py, mode='eval')
> +  _RewriteIfExpToSelect().visit(parsed)
> +  parsed = ast.fix_missing_locations(parsed)
> +  return _Constify(eval(compile(parsed, orig, 'eval')))
> diff --git a/tools/perf/pmu-events/metric_test.py b/tools/perf/pmu-events/metric_test.py
> new file mode 100644
> index 000000000000..66be55893de0
> --- /dev/null
> +++ b/tools/perf/pmu-events/metric_test.py
> @@ -0,0 +1,148 @@
> +# SPDX-License-Identifier: (LGPL-2.1 OR BSD-2-Clause)
> +import unittest
> +from metric import Constant
> +from metric import Event
> +from metric import ParsePerfJson
> +
> +
> +class TestMetricExpressions(unittest.TestCase):
> +
> +  def test_Operators(self):
> +    a = Event('a')
> +    b = Event('b')
> +    self.assertEqual((a | b).ToPerfJson(), 'a | b')
> +    self.assertEqual((a ^ b).ToPerfJson(), 'a ^ b')
> +    self.assertEqual((a & b).ToPerfJson(), 'a & b')
> +    self.assertEqual((a < b).ToPerfJson(), 'a < b')
> +    self.assertEqual((a > b).ToPerfJson(), 'a > b')
> +    self.assertEqual((a + b).ToPerfJson(), 'a + b')
> +    self.assertEqual((a - b).ToPerfJson(), 'a - b')
> +    self.assertEqual((a * b).ToPerfJson(), 'a * b')
> +    self.assertEqual((a / b).ToPerfJson(), 'a / b')
> +    self.assertEqual((a % b).ToPerfJson(), 'a % b')
> +    one = Constant(1)
> +    self.assertEqual((a + one).ToPerfJson(), 'a + 1')
> +
> +  def test_Brackets(self):
> +    a = Event('a')
> +    b = Event('b')
> +    c = Event('c')
> +    self.assertEqual((a * b + c).ToPerfJson(), 'a * b + c')
> +    self.assertEqual((a + b * c).ToPerfJson(), 'a + b * c')
> +    self.assertEqual(((a + a) + a).ToPerfJson(), 'a + a + a')
> +    self.assertEqual(((a + b) * c).ToPerfJson(), '(a + b) * c')
> +    self.assertEqual((a + (b * c)).ToPerfJson(), 'a + b * c')
> +    self.assertEqual(((a / b) * c).ToPerfJson(), 'a / b * c')
> +    self.assertEqual((a / (b * c)).ToPerfJson(), 'a / (b * c)')
> +
> +  def test_ParsePerfJson(self):
> +    # Based on an example of a real metric.
> +    before = '(a + b + c + d) / (2 * e)'
> +    after = before
> +    self.assertEqual(ParsePerfJson(before).ToPerfJson(), after)
> +
> +    # Parsing should handle events with '-' in their name. Note, in
> +    # the json file the '\' are doubled to '\\'.
> +    before = r'topdown\-fe\-bound / topdown\-slots - 1'
> +    after = before
> +    self.assertEqual(ParsePerfJson(before).ToPerfJson(), after)
> +
> +    # Parsing should handle escaped modifiers. Note, in the json file
> +    # the '\' are doubled to '\\'.
> +    before = r'arb@event\=0x81\,umask\=0x1@ + arb@event\=0x84\,umask\=0x1@'
> +    after = before
> +    self.assertEqual(ParsePerfJson(before).ToPerfJson(), after)
> +
> +    # Parsing should handle exponents in numbers.
> +    before = r'a + 1e12 + b'
> +    after = before
> +    self.assertEqual(ParsePerfJson(before).ToPerfJson(), after)
> +
> +  def test_IfElseTests(self):
> +    # if-else needs rewriting to Select and back.
> +    before = r'Event1 if #smt_on else Event2'
> +    after = f'({before})'
> +    self.assertEqual(ParsePerfJson(before).ToPerfJson(), after)
> +
> +    before = r'Event1 if 0 else Event2'
> +    after = f'({before})'
> +    self.assertEqual(ParsePerfJson(before).ToPerfJson(), after)
> +
> +    before = r'Event1 if 1 else Event2'
> +    after = f'({before})'
> +    self.assertEqual(ParsePerfJson(before).ToPerfJson(), after)
> +
> +    # Ensure the select is evaluate last.
> +    before = r'Event1 + 1 if Event2 < 2 else Event3 + 3'
> +    after = (r'Select(Event(r"Event1") + Constant(1), Event(r"Event2") < '
> +             r'Constant(2), Event(r"Event3") + Constant(3))')
> +    self.assertEqual(ParsePerfJson(before).ToPython(), after)
> +
> +    before = r'Event1 > 1 if Event2 < 2 else Event3 > 3'
> +    after = (r'Select(Event(r"Event1") > Constant(1), Event(r"Event2") < '
> +             r'Constant(2), Event(r"Event3") > Constant(3))')
> +    self.assertEqual(ParsePerfJson(before).ToPython(), after)
> +
> +    before = r'min(a + b if c > 1 else c + d, e + f)'
> +    after = r'min((a + b if c > 1 else c + d), e + f)'
> +    self.assertEqual(ParsePerfJson(before).ToPerfJson(), after)
> +
> +  def test_ToPython(self):
> +    # pylint: disable=eval-used
> +    # Based on an example of a real metric.
> +    before = '(a + b + c + d) / (2 * e)'
> +    py = ParsePerfJson(before).ToPython()
> +    after = eval(py).ToPerfJson()
> +    self.assertEqual(before, after)
> +
> +  def test_Simplify(self):
> +    before = '1 + 2 + 3'
> +    after = '6'
> +    self.assertEqual(ParsePerfJson(before).Simplify().ToPerfJson(), after)
> +
> +    before = 'a + 0'
> +    after = 'a'
> +    self.assertEqual(ParsePerfJson(before).Simplify().ToPerfJson(), after)
> +
> +    before = '0 + a'
> +    after = 'a'
> +    self.assertEqual(ParsePerfJson(before).Simplify().ToPerfJson(), after)
> +
> +    before = 'a | 0'
> +    after = 'a'
> +    self.assertEqual(ParsePerfJson(before).Simplify().ToPerfJson(), after)
> +
> +    before = '0 | a'
> +    after = 'a'
> +    self.assertEqual(ParsePerfJson(before).Simplify().ToPerfJson(), after)
> +
> +    before = 'a * 0'
> +    after = '0'
> +    self.assertEqual(ParsePerfJson(before).Simplify().ToPerfJson(), after)
> +
> +    before = '0 * a'
> +    after = '0'
> +    self.assertEqual(ParsePerfJson(before).Simplify().ToPerfJson(), after)
> +
> +    before = 'a * 1'
> +    after = 'a'
> +    self.assertEqual(ParsePerfJson(before).Simplify().ToPerfJson(), after)
> +
> +    before = '1 * a'
> +    after = 'a'
> +    self.assertEqual(ParsePerfJson(before).Simplify().ToPerfJson(), after)
> +
> +    before = 'a if 0 else b'
> +    after = 'b'
> +    self.assertEqual(ParsePerfJson(before).Simplify().ToPerfJson(), after)
> +
> +    before = 'a if 1 else b'
> +    after = 'a'
> +    self.assertEqual(ParsePerfJson(before).Simplify().ToPerfJson(), after)
> +
> +    before = 'a if b else a'
> +    after = 'a'
> +    self.assertEqual(ParsePerfJson(before).Simplify().ToPerfJson(), after)
> +
> +if __name__ == '__main__':
> +  unittest.main()
> -- 
> 2.38.1.584.g0f3c55d4c2-goog

-- 

- Arnaldo

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: [PATCH v3] perf jevents: Parse metrics during conversion
  2022-12-05 12:53 ` Arnaldo Carvalho de Melo
@ 2022-12-05 12:56   ` John Garry
  0 siblings, 0 replies; 6+ messages in thread
From: John Garry @ 2022-12-05 12:56 UTC (permalink / raw)
  To: Arnaldo Carvalho de Melo, Ian Rogers
  Cc: Peter Zijlstra, Ingo Molnar, Mark Rutland, Alexander Shishkin,
	Jiri Olsa, Namhyung Kim, Sumanth Korikkar, Thomas Richter,
	linux-kernel, linux-perf-users, Stephane Eranian

On 05/12/2022 12:53, Arnaldo Carvalho de Melo wrote:
> Em Wed, Nov 30, 2022 at 07:41:38PM -0800, Ian Rogers escreveu:
>> Currently the 'MetricExpr' json value is passed from the json
>> file to the pmu-events.c. This change introduces an expression
>> tree that is parsed into. The parsing is done largely by using
>> operator overloading and python's 'eval' function. Two advantages
>> in doing this are:
>>
>> 1) Broken metrics fail at compile time rather than relying on
>>     `perf test` to detect. `perf test` remains relevant for checking
>>     event encoding and actual metric use.
>>
>> 2) The conversion to a string from the tree can minimize the metric's
>>     string size, for example, preferring 1e6 over 1000000, avoiding
>>     multiplication by 1 and removing unnecessary whitespace. On x86
>>     this reduces the string size by 3,050bytes (0.07%).
>>
>> In future changes it would be possible to programmatically
>> generate the json expressions (a single line of text and so a
>> pain to write manually) for an architecture using the expression
>> tree. This could avoid copy-pasting metrics for all architecture
>> variants.
>>
>> Signed-off-by: Ian Rogers<irogers@google.com>
>>
>> v3. Avoids generic types on standard types like set that aren't
>>      supported until Python 3.9, fixing an issue with Python 3.6
>>      reported-by John Garry. v3 also fixes minor pylint issues and adds
>>      a call to Simplify on the read expression tree.
> Cool, John looked reviewed it (may I add the tag?). Applying to my local
> tree, will do some light testing.

Actually I just got as far as testing v2 (which gave rise to this v3). I 
will look at it further today.

Thanks,
John

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: [PATCH v3] perf jevents: Parse metrics during conversion
  2022-12-01  3:41 [PATCH v3] perf jevents: Parse metrics during conversion Ian Rogers
  2022-12-05 12:53 ` Arnaldo Carvalho de Melo
@ 2022-12-05 15:24 ` John Garry
  2022-12-06 19:29   ` Ian Rogers
  1 sibling, 1 reply; 6+ messages in thread
From: John Garry @ 2022-12-05 15:24 UTC (permalink / raw)
  To: Ian Rogers, Peter Zijlstra, Ingo Molnar,
	Arnaldo Carvalho de Melo, Mark Rutland, Alexander Shishkin,
	Jiri Olsa, Namhyung Kim, Sumanth Korikkar, Thomas Richter,
	linux-kernel, linux-perf-users
  Cc: Stephane Eranian

On 01/12/2022 03:41, Ian Rogers wrote:
> Currently the 'MetricExpr' json value is passed from the json
> file to the pmu-events.c. This change introduces an expression
> tree that is parsed into. The parsing is done largely by using
> operator overloading and python's 'eval' function. Two advantages
> in doing this are:
> 
> 1) Broken metrics fail at compile time rather than relying on
>     `perf test` to detect. `perf test` remains relevant for checking
>     event encoding and actual metric use.

Do we still require the code to "resolve metrics" in resolve_metric()? 
But I'm not sure it even ever had any users.

> 
> 2) The conversion to a string from the tree can minimize the metric's
>     string size, for example, preferring 1e6 over 1000000, avoiding
>     multiplication by 1 and removing unnecessary whitespace. On x86
>     this reduces the string size by 3,050bytes (0.07%).

Out of curiosity, did you try the exponent change on its own (to see the 
impact on size)?

Nit:

Unrelated, really, I notice that sometimes we lose the parenthesis and 
sometimes never had them, like:

/* offset=11526 */ "\000\000metrics\000Ave [...] 0\000( 1000000000 * ( 
UNC_CHA
/* offset=11207 */ "\000\000metrics\000Ave [...] 0\0001e9 * (UNC_CHA_TOR

To me, it seems neater to have the expression contained within (a 
parenthesis) ever since we moved to this "big string". This seems to be 
a preexisting feature.

Thanks,
John


> 
> In future changes it would be possible to programmatically
> generate the json expressions (a single line of text and so a
> pain to write manually) for an architecture using the expression
> tree. This could avoid copy-pasting metrics for all architecture
> variants.
> 
> Signed-off-by: Ian Rogers<irogers@google.com>


^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: [PATCH v3] perf jevents: Parse metrics during conversion
  2022-12-05 15:24 ` John Garry
@ 2022-12-06 19:29   ` Ian Rogers
  2022-12-07 11:21     ` John Garry
  0 siblings, 1 reply; 6+ messages in thread
From: Ian Rogers @ 2022-12-06 19:29 UTC (permalink / raw)
  To: John Garry
  Cc: Peter Zijlstra, Ingo Molnar, Arnaldo Carvalho de Melo,
	Mark Rutland, Alexander Shishkin, Jiri Olsa, Namhyung Kim,
	Sumanth Korikkar, Thomas Richter, linux-kernel, linux-perf-users,
	Stephane Eranian

On Mon, Dec 5, 2022 at 7:24 AM John Garry <john.g.garry@oracle.com> wrote:
>
> On 01/12/2022 03:41, Ian Rogers wrote:
> > Currently the 'MetricExpr' json value is passed from the json
> > file to the pmu-events.c. This change introduces an expression
> > tree that is parsed into. The parsing is done largely by using
> > operator overloading and python's 'eval' function. Two advantages
> > in doing this are:
> >
> > 1) Broken metrics fail at compile time rather than relying on
> >     `perf test` to detect. `perf test` remains relevant for checking
> >     event encoding and actual metric use.
>
> Do we still require the code to "resolve metrics" in resolve_metric()?
> But I'm not sure it even ever had any users.

We use metrics referencing other metrics for topdown metrics on x86.
For example:

https://git.kernel.org/pub/scm/linux/kernel/git/acme/linux.git/tree/tools/perf/pmu-events/arch/x86/icelakex/icx-metrics.json?h=perf/core#n34
    {
        "BriefDescription": "This metric represents fraction of cycles
the CPU was stalled due to Branch Resteers",
        "MetricExpr": "INT_MISC.CLEAR_RESTEER_CYCLES / CLKS +
tma_unknown_branches",
        "MetricGroup": "FetchLat;TopdownL3;tma_fetch_latency_group",
        "MetricName": "tma_branch_resteers",
        "PublicDescription": "This metric represents fraction of
cycles the CPU was stalled due to Branch Resteers. Branch Resteers
estimates the Frontend delay in fetching operations from corrected
path; following all sorts of miss-predicted branches. For example;
branchy code with lots of miss-predictions might get categorized under
Branch Resteers. Note the value of this node may overlap with its
siblings. Sample with: BR_MISP_RETIRED.ALL_BRANCHES",
        "ScaleUnit": "100%"
    },
...
    {
        "BriefDescription": "This metric represents fraction of cycles
the CPU was stalled due to new branch address clears",
        "MetricExpr": "10 * BACLEARS.ANY / CLKS",
        "MetricGroup": "BigFoot;FetchLat;TopdownL4;tma_branch_resteers_group",
        "MetricName": "tma_unknown_branches",
        "PublicDescription": "This metric represents fraction of
cycles the CPU was stalled due to new branch address clears. These are
fetched branches the Branch Prediction Unit was unable to recognize
(First fetch or hitting BPU capacity limit). Sample with:
BACLEARS.ANY",
        "ScaleUnit": "100%"
    },

> >
> > 2) The conversion to a string from the tree can minimize the metric's
> >     string size, for example, preferring 1e6 over 1000000, avoiding
> >     multiplication by 1 and removing unnecessary whitespace. On x86
> >     this reduces the string size by 3,050bytes (0.07%).
>
> Out of curiosity, did you try the exponent change on its own (to see the
> impact on size)?

The file size savings are very modest. Without removing the "1 * " the
savings were roughly 2KB, perhaps 1KB was shrinking the constant
exponents.

> Nit:
>
> Unrelated, really, I notice that sometimes we lose the parenthesis and
> sometimes never had them, like:
>
> /* offset=11526 */ "\000\000metrics\000Ave [...] 0\000( 1000000000 * (
> UNC_CHA
> /* offset=11207 */ "\000\000metrics\000Ave [...] 0\0001e9 * (UNC_CHA_TOR
>
> To me, it seems neater to have the expression contained within (a
> parenthesis) ever since we moved to this "big string". This seems to be
> a preexisting feature.

You can also read the metrics through "perf list --detail", we could
add parentheses there if it helps readability. We can also expand out
what the big string values are for comments. Fwiw, I want to start
refactoring jevents.py in follow up work and that would impact
readability. Some thoughts there are:

1) we shouldn't parse all json events for all PMUs in prior to parsing
events, we should initialize a PMU when an event references it and
then possibly then go through the json events. To facilitate this it
would be useful to organize events by their PMU.
2) metrics and events should be separated at least in the C code.
Currently on x86 ScaleUnit in the json will apply both to an event and
its metric, even though the uses of an event and a metric should have
different units.
3) for some operating systems with limited disk, it would be nice to
be able to have the build exclude models.

Let me know if there's anything more outstanding to fix on this patch set.

Thanks,
Ian

> Thanks,
> John
>
>
> >
> > In future changes it would be possible to programmatically
> > generate the json expressions (a single line of text and so a
> > pain to write manually) for an architecture using the expression
> > tree. This could avoid copy-pasting metrics for all architecture
> > variants.
> >
> > Signed-off-by: Ian Rogers<irogers@google.com>
>

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: [PATCH v3] perf jevents: Parse metrics during conversion
  2022-12-06 19:29   ` Ian Rogers
@ 2022-12-07 11:21     ` John Garry
  0 siblings, 0 replies; 6+ messages in thread
From: John Garry @ 2022-12-07 11:21 UTC (permalink / raw)
  To: Ian Rogers
  Cc: Peter Zijlstra, Ingo Molnar, Arnaldo Carvalho de Melo,
	Mark Rutland, Alexander Shishkin, Jiri Olsa, Namhyung Kim,
	Sumanth Korikkar, Thomas Richter, linux-kernel, linux-perf-users,
	Stephane Eranian

On 06/12/2022 19:29, Ian Rogers wrote:
> Dec 5, 2022 at 7:24 AM John Garry<john.g.garry@oracle.com>  wrote:
>> On 01/12/2022 03:41, Ian Rogers wrote:
>>> Currently the 'MetricExpr' json value is passed from the json
>>> file to the pmu-events.c. This change introduces an expression
>>> tree that is parsed into. The parsing is done largely by using
>>> operator overloading and python's 'eval' function. Two advantages
>>> in doing this are:
>>>
>>> 1) Broken metrics fail at compile time rather than relying on
>>>      `perf test` to detect. `perf test` remains relevant for checking
>>>      event encoding and actual metric use.
>> Do we still require the code to "resolve metrics" in resolve_metric()?
>> But I'm not sure it even ever had any users.
> We use metrics referencing other metrics for topdown metrics on x86.
> For example:

ok, I just wasn't sure if there were ever any metrics which did require 
"resolving". Now I know.

> 
> https://urldefense.com/v3/__https://git.kernel.org/pub/scm/linux/kernel/git/acme/linux.git/tree/tools/perf/pmu-events/arch/x86/icelakex/icx-metrics.json?h=perf*core*n34__;LyM!!ACWV5N9M2RV99hQ!PMKcRLro8XREBI-072XYolfLHVvOm4P-HBWTpvu8IxJzkE0NWydgW9wi2PclFvUrdQcuC-4uvubPf5RgWYI$  
>      {
>          "BriefDescription": "This metric represents fraction of cycles
> the CPU was stalled due to Branch Resteers",
>          "MetricExpr": "INT_MISC.CLEAR_RESTEER_CYCLES / CLKS +
> tma_unknown_branches",
>          "MetricGroup": "FetchLat;TopdownL3;tma_fetch_latency_group",
>          "MetricName": "tma_branch_resteers",
>          "PublicDescription": "This metric represents fraction of
> cycles the CPU was stalled due to Branch Resteers. Branch Resteers
> estimates the Frontend delay in fetching operations from corrected
> path; following all sorts of miss-predicted branches. For example;
> branchy code with lots of miss-predictions might get categorized under
> Branch Resteers. Note the value of this node may overlap with its
> siblings. Sample with: BR_MISP_RETIRED.ALL_BRANCHES",
>          "ScaleUnit": "100%"
>      },
> ...
>      {
>          "BriefDescription": "This metric represents fraction of cycles
> the CPU was stalled due to new branch address clears",
>          "MetricExpr": "10 * BACLEARS.ANY / CLKS",
>          "MetricGroup": "BigFoot;FetchLat;TopdownL4;tma_branch_resteers_group",
>          "MetricName": "tma_unknown_branches",
>          "PublicDescription": "This metric represents fraction of
> cycles the CPU was stalled due to new branch address clears. These are
> fetched branches the Branch Prediction Unit was unable to recognize
> (First fetch or hitting BPU capacity limit). Sample with:
> BACLEARS.ANY",
>          "ScaleUnit": "100%"
>      },
> 
>>> 2) The conversion to a string from the tree can minimize the metric's
>>>      string size, for example, preferring 1e6 over 1000000, avoiding
>>>      multiplication by 1 and removing unnecessary whitespace. On x86
>>>      this reduces the string size by 3,050bytes (0.07%).
>> Out of curiosity, did you try the exponent change on its own (to see the
>> impact on size)?
> The file size savings are very modest. Without removing the "1 * " the
> savings were roughly 2KB, perhaps 1KB was shrinking the constant
> exponents.
> 
>> Nit:
>>
>> Unrelated, really, I notice that sometimes we lose the parenthesis and
>> sometimes never had them, like:
>>
>> /* offset=11526 */ "\000\000metrics\000Ave [...] 0\000( 1000000000 * (
>> UNC_CHA
>> /* offset=11207 */ "\000\000metrics\000Ave [...] 0\0001e9 * (UNC_CHA_TOR
>>
>> To me, it seems neater to have the expression contained within (a
>> parenthesis) ever since we moved to this "big string". This seems to be
>> a preexisting feature.
> You can also read the metrics through "perf list --detail", we could
> add parentheses there if it helps readability.

At least being consistent would be nice, whichever way you want to go.

  We can also expand out
> what the big string values are for comments. 

Maybe a comment at the top of the array would be nice to tell which 
member is per column.

> Fwiw, I want to start
> refactoring jevents.py in follow up work and that would impact
> readability. Some thoughts there are:
> 
> 1) we shouldn't parse all json events for all PMUs in prior to parsing
> events, we should initialize a PMU when an event references it and
> then possibly then go through the json events. To facilitate this it
> would be useful to organize events by their PMU.
> 2) metrics and events should be separated at least in the C code.
> Currently on x86 ScaleUnit in the json will apply both to an event and
> its metric, even though the uses of an event and a metric should have
> different units.
> 3) for some operating systems with limited disk, it would be nice to
> be able to have the build exclude models.

Eh, do you mean an option to build just for the host system? If so, 
seems reasonable.

> 
> Let me know if there's anything more outstanding to fix on this patch set.

It seems fine. FWIW,

Reviewed-by: John Garry <john.g.garry@oracle.com>

^ permalink raw reply	[flat|nested] 6+ messages in thread

end of thread, other threads:[~2022-12-07 11:22 UTC | newest]

Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-12-01  3:41 [PATCH v3] perf jevents: Parse metrics during conversion Ian Rogers
2022-12-05 12:53 ` Arnaldo Carvalho de Melo
2022-12-05 12:56   ` John Garry
2022-12-05 15:24 ` John Garry
2022-12-06 19:29   ` Ian Rogers
2022-12-07 11:21     ` John Garry

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).