cbalint13 commented on code in PR #13621:
URL: https://github.com/apache/tvm/pull/13621#discussion_r1058171563
##########
tests/python/integration/test_auto_tensorize.py:
##########
@@ -278,6 +283,17 @@ def test_vnni_dense():
)
+# TODO(vvchernov): need schedule rules and postprocs for avx512
Review Comment:
* Can reuse/add them here on par with VNNI one ?
* Would be great to check auto-tensorizer along with the TOPI template.
##########
tests/python/unittest/test_meta_schedule_trace_apply.py:
##########
@@ -2502,6 +2503,7 @@ def apply_trace(sch):
verify(Conv2dInt8, apply_trace, Conv2dInt8_target, "cuda",
Conv2dInt8_tensorcore_scheduled)
+# TODO(vvchernov): test int8 conv2d foravx512 without VNNI
Review Comment:
* Can add the mentioned avx512 test case ?
##########
tests/python/unittest/test_meta_schedule_trace_apply.py:
##########
@@ -1131,6 +1131,7 @@ def main(p0: T.Buffer[(1, 32, 7, 7, 16), "uint8"], p1:
T.Buffer[(128, 32, 1, 1,
T_cast[ax0, ax1, ax2, ax3, ax4] = T.cast(compute_2[ax0, ax1,
ax2, ax3, ax4], "int32")
+# TODO(vvchernov): construct avx512 reference module (without vnni)
Review Comment:
* Can add mentioned reference module ?
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
