mukesh
84a1325e5c
Charge per-tier bandwidth + setup overhead at IPCQ slot WRITE
(receiver inbound DMA, in pe_dma._handle_ipcq_inbound) and slot
READ (recv consume, in pe_ipcq._handle_recv). Tier table
(common/ipcq_types.py):
tcm : 512 GB/s, 0 ns
sram : 128 GB/s, 2 ns
hbm : 32 GB/s, 6 ns
Before this change, slot read/write was free regardless of
buffer_kind, making memory-tier choice invisible in simulated
latency. After the change, swapping buffer_kind in ccl.yaml
produces measurable per-tier separation in allreduce latency.
Tests:
test_ipcq_buffer_kind_latency.py — three micro-tests asserting
tcm < sram < hbm ordering, payload-scaling, and that
buffer_kind sensitivity grows with payload (credit-only path
stays fabric-bound).
test_allreduce_buffer_kind_sweep.py — 12-config parametrized
sweep emitting buffer_kind_sweep.png (3 lines, torus_2d).
conftest sessionfinish hook generalised to dispatch multiple
sweep aggregators (allreduce + buffer-kind).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
75 lines
2.3 KiB
Python
75 lines
2.3 KiB
Python
"""Shared pytest fixtures for the kernbench test suite.
|
|
|
|
Session-scoped topology caching: ``resolve_topology("topology.yaml")`` is
|
|
pure (no side effects), so we cache the result across all tests in a
|
|
worker process. Each test still builds its own ``GraphEngine`` (which is
|
|
stateful/SimPy-event-consuming and MUST NOT be shared).
|
|
"""
|
|
from __future__ import annotations
|
|
|
|
import os
|
|
|
|
import pytest
|
|
|
|
from kernbench.topology.builder import resolve_topology
|
|
|
|
|
|
def pytest_sessionfinish(session, exitstatus):
|
|
"""Aggregate parametrized sweep rows into combined CSV + PNG plots.
|
|
|
|
Runs on the controller node only (xdist worker processes set
|
|
``PYTEST_XDIST_WORKER``; we skip those). Idempotent — does nothing
|
|
if no sweep rows are present (e.g., when the sweep was filtered out).
|
|
"""
|
|
if os.environ.get("PYTEST_XDIST_WORKER"):
|
|
return
|
|
import importlib.util
|
|
import sys
|
|
from pathlib import Path
|
|
|
|
def _exec(name: str, attr: str) -> None:
|
|
mod_path = Path(__file__).parent / name
|
|
if not mod_path.exists():
|
|
return
|
|
s = importlib.util.spec_from_file_location(
|
|
f"_{name.removesuffix('.py')}_for_aggregate", mod_path,
|
|
)
|
|
if s is None or s.loader is None:
|
|
return
|
|
mod = importlib.util.module_from_spec(s)
|
|
sys.modules[s.name] = mod
|
|
try:
|
|
s.loader.exec_module(mod)
|
|
fn = getattr(mod, attr, None)
|
|
if fn is not None:
|
|
fn()
|
|
except Exception as e:
|
|
print(f"[conftest] aggregator {attr}() in {name} failed: {e}")
|
|
|
|
_exec("test_allreduce_multidevice.py", "_aggregate_sweep_plots")
|
|
_exec("test_allreduce_buffer_kind_sweep.py", "aggregate_buffer_kind_plot")
|
|
|
|
|
|
@pytest.fixture(scope="session")
|
|
def topology():
|
|
"""Session-scoped parsed topology (immutable graph + spec).
|
|
|
|
Usage in tests::
|
|
|
|
def test_foo(topology):
|
|
engine = GraphEngine(topology.topology_obj, enable_data=True)
|
|
"""
|
|
return resolve_topology("topology.yaml")
|
|
|
|
|
|
@pytest.fixture(scope="session")
|
|
def topology_obj(topology):
|
|
"""The TopologyGraph inside the handle (convenience shortcut)."""
|
|
return topology.topology_obj
|
|
|
|
|
|
@pytest.fixture(scope="session")
|
|
def spec(topology):
|
|
"""Topology spec dict (convenience shortcut)."""
|
|
return topology.topology_obj.spec
|