show number of waves in each SE/CU (#13491)

* show number of waves in each SE/CU

* update to test_ones
This commit is contained in:
qazal
2025-11-30 22:29:16 +08:00
committed by GitHub
parent 455dd88236
commit 9023ca30ef
2 changed files with 7 additions and 5 deletions

View File

@@ -2,7 +2,9 @@ import os
os.environ["PYTHONPATH"] = "."
os.environ["SQTT"] = "1"
if "DEV" not in os.environ: os.environ["DEV"] = "AMD"
os.environ["VIZ"] = "1"
os.environ["PROFILE"] = "1"
# VIZ=1 to launch server
# os.environ["VIZ"] = "1"
os.environ["AMD_LLVM"] = "0"
import unittest
@@ -129,13 +131,13 @@ class TestTiming(unittest.TestCase):
for w in waves:
print(f"{w.wave_id:<2} {w.simd=} {w.cu=} {w.se=} @ clk {w.begin_time}")
def test_multiple_runs(self):
N = 4096
def test_ones(self):
N = getenv("N", 4096)
CNT = getenv("CNT", 2)
with save_sqtt() as sqtt:
for _ in range(CNT):
Tensor.ones(N, N).contiguous().realize()
self.assertEqual(len(sqtt), 2)
self.assertEqual(len(sqtt), CNT)
if __name__ == "__main__":
unittest.main()

View File

@@ -256,7 +256,7 @@ def load_sqtt(profile:list[ProfileEvent]) -> None:
steps.append(create_step(kernel.name if kernel is not None else name.prg, ("/counters", len(ctxs), len(steps)), {"src":src}, depth=1))
for cu in prg_cu:
events = [ProfilePointEvent(unit, "start", unit, ts=Decimal(0)) for unit in units]+cu_events[cu]
steps.append(create_step(cu, ("/counters", len(ctxs), len(steps)),
steps.append(create_step(f"{cu} {len(cu_events[cu])}", ("/counters", len(ctxs), len(steps)),
{"value":get_profile(events, sort_fn=row_tuple), "content_type":"application/octet-stream"}, depth=2))
for k in sorted(wave_insts.get(cu, []), key=row_tuple):
data = wave_insts[cu][k]