Commit de02a203 by Yizhi Liu Committed by Tianqi Chen

print import_llvm ir in tensorize tutorial (#2064)

parent c91ded32
...@@ -154,6 +154,12 @@ def gemv_impl(): ...@@ -154,6 +154,12 @@ def gemv_impl():
# The importing needs to happen before the tensorized GEMV being executed. # The importing needs to happen before the tensorized GEMV being executed.
# #
s[C].pragma(x, "import_llvm", gemv_impl()) s[C].pragma(x, "import_llvm", gemv_impl())
print(tvm.lower(s, [A, B, C], simple_mode=True))
######################################################################
# Finally we compare the tensorize version with that :code:`numpy.dot` produces,
# ensure our implementation is correct.
#
func = tvm.build(s, [A, B, C], target="llvm", name="gemv") func = tvm.build(s, [A, B, C], target="llvm", name="gemv")
from topi.util import get_const_tuple from topi.util import get_const_tuple
...@@ -166,12 +172,11 @@ func(tvm.nd.array(a, ctx), tvm.nd.array(b, ctx), c) ...@@ -166,12 +172,11 @@ func(tvm.nd.array(a, ctx), tvm.nd.array(b, ctx), c)
tvm.testing.assert_allclose(c.asnumpy(), np.dot(a, b.T), rtol=1e-3) tvm.testing.assert_allclose(c.asnumpy(), np.dot(a, b.T), rtol=1e-3)
###################################################################### ######################################################################
# We compare the tensorize version with that :code:`numpy.dot` produces,
# ensure our implementation is correct.
#
# Reduce-update for Tensorize # Reduce-update for Tensorize
# ------------------------------------ # ---------------------------
# Let's then move one step forward. # So far you have learned the basic idea of tensorize,
# now let's move one step forward to a more complicated case.
#
# Assume our accelerator could only multiply a vector by a square matrix, # Assume our accelerator could only multiply a vector by a square matrix,
# in which the vector size needs to be no larger than 16. # in which the vector size needs to be no larger than 16.
# Given such hardware constrain, now we need to split the reduce axis as following, # Given such hardware constrain, now we need to split the reduce axis as following,
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment