[codegen/selection] tensor cores now used for transposed layotus
This commit is contained in:
@@ -25,6 +25,7 @@ def run_dot():
|
||||
# Test
|
||||
hresult = np.dot(ha.T, hb.T).T
|
||||
dif = np.abs(result - hresult)
|
||||
np.savetxt('dif.dat', dif, '%2.4f')
|
||||
print(hresult)
|
||||
print(result)
|
||||
print("dif: %f" % np.max(dif))
|
||||
|
Reference in New Issue
Block a user