github/tinygrad
1
1
Fork 0
mirror of https://github.com/tinygrad/tinygrad.git synced 2026-01-09 15:08:02 -05:00
Code Issues Packages Projects Releases Wiki Activity

usbgpu: 10 steps in benchmark to hit cache (#10273)

Browse Source
This commit is contained in:
nimlgen
2025-05-13 17:06:50 +03:00
committed by GitHub
parent ad5cb2717d
commit 6f42bf8b54
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
test/external/external_test_usb_asm24.py vendored
View File

@@ -46,7 +46,7 @@ class TestDevCopySpeeds(unittest.TestCase):
if not cls.dev.is_usb(): raise unittest.SkipTest("only test this on USB devices")
def testCopyCPUtoDefault(self):
for _ in range(3):
for _ in range(10):
t = Tensor.ones(self.sz, self.sz, device="CPU").contiguous().realize()
with Timing(f"copyin of {t.nbytes()/1e6:.2f} MB: ", on_exit=lambda ns: f" @ {t.nbytes()/ns * 1e3:.2f} MB/s"): # noqa: F821
t.to(Device.DEFAULT).realize()
@@ -55,7 +55,7 @@ class TestDevCopySpeeds(unittest.TestCase):
def testCopyDefaulttoCPU(self):
t = Tensor.ones(self.sz, self.sz).contiguous().realize()
for _ in range(3):
for _ in range(10):
with Timing(f"copyout of {t.nbytes()/1e6:.2f} MB: ", on_exit=lambda ns: f" @ {t.nbytes()/ns * 1e3:.2f} MB/s"):
t.to('CPU').realize()