From 29b60014f4d0cd23ac2af408306cc54edd5de89d Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Ren=C3=A9=20He=C3=9F?= <rene.hess@iwr.uni-heidelberg.de>
Date: Wed, 3 Apr 2019 15:32:09 +0200
Subject: [PATCH] [skip ci] Autotune loop order in tensor contraction
---
.../dune/codegen/sumfact/transformations.py | 47 +++++++++++++++----
1 file changed, 39 insertions(+), 8 deletions(-)
diff --git a/python/dune/codegen/sumfact/transformations.py b/python/dune/codegen/sumfact/transformations.py
index b7ab65cf..8319890b 100644
--- a/python/dune/codegen/sumfact/transformations.py
+++ b/python/dune/codegen/sumfact/transformations.py
@@ -9,6 +9,7 @@ from dune.codegen.loopy.transformations.remove_reductions import remove_all_redu
from dune.codegen.options import get_form_option, get_option
from dune.codegen.pdelab.geometry import world_dimension
from dune.codegen.error import CodegenAutotuneError
+from dune.codegen.sumfact.autotune import autotune_realization
def move_zero_assignment_up(kernel, move_up_inames):
@@ -169,16 +170,46 @@ def reorder_loops_in_tensor_contraction(kernel, iname_order):
return kernel
+def tensor_contraction_loop_order_generator(kernel):
+ dim = world_dimension()
+ assert dim == 3
+ yield kernel
+
+ indices = ['l', 'k', 'i', 'j']
+ import itertools
+ for loop_order in itertools.permutations(indices):
+ loop_order = ''.join(loop_order)
+ new_kernel = reorder_loops_in_tensor_contraction(kernel, loop_order)
+ yield new_kernel
+
+
+def simple_autotuner(kernel_generator, signature):
+ # palpo TODO
+ from dune.codegen.options import set_option
+ set_option("autotune_google_benchmark", True)
+
+ kernel = next(kernel_generator)
+ best_cost = autotune_realization(kernel=kernel, signature=signature)
+ best_kernel = kernel
+ for kernel in kernel_generator:
+ cost = autotune_realization(kernel=kernel, signature=signature)
+ if cost < best_cost:
+ best_cost = cost
+ best_kernel = kernel
+ return best_kernel
+
+
+def autotune_tensor_contraction_loop_order(kernel, signature):
+ from dune.codegen.loopy.transformations.matchfma import match_fused_multiply_add
+ kernel = match_fused_multiply_add(kernel)
+
+ generator = tensor_contraction_loop_order_generator(kernel)
+ return simple_autotuner(generator, signature)
+
+
def sumfact_performance_transformations(kernel, signature):
if kernel.name.startswith('sfimpl'):
- # from dune.codegen.loopy.transformations.matchfma import match_fused_multiply_add
- # kernel = match_fused_multiply_add(kernel)
- # kernel = reorder_loops_in_tensor_contraction(kernel, 'ijlk')
-
- # from dune.codegen.sumfact.autotune import autotune_realization
- # from dune.codegen.options import set_option
- # set_option("autotune_google_benchmark", True)
- # test = autotune_realization(kernel=kernel, signature=signature)
+ # kernel = autotune_tensor_contraction_loop_order(kernel, signature)
pass
return kernel
--
GitLab