You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@tvm.apache.org by GitBox <gi...@apache.org> on 2021/04/14 06:51:59 UTC

[GitHub] [tvm] Hzfengsy opened a new pull request #7848: [TensorIR][M1c] LCA detector

Hzfengsy opened a new pull request #7848:
URL: https://github.com/apache/tvm/pull/7848


   Note that the LCA detector is a helper analysis for lowering, which can help find where the buffer should be allocated.
   
   cc @tqchen @junrushao1994 @jroesch @comaniac 


-- 
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.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org



[GitHub] [tvm] Hzfengsy commented on a change in pull request #7848: [TensorIR][M1c] LCA detector

Posted by GitBox <gi...@apache.org>.
Hzfengsy commented on a change in pull request #7848:
URL: https://github.com/apache/tvm/pull/7848#discussion_r612998482



##########
File path: src/tir/analysis/buffer_access_lca_detector.cc
##########
@@ -0,0 +1,173 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *   http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ * KIND, either express or implied.  See the License for the
+ * specific language governing permissions and limitations
+ * under the License.
+ */
+
+/*!
+ * \file tir/analysis/buffer_access_lca_detector.cc
+ * \brief Detect LCA of buffer access
+ */
+
+#include <tvm/tir/analysis.h>
+#include <tvm/tir/stmt_functor.h>
+
+#include "../../support/arena.h"
+
+namespace tvm {
+namespace tir {
+
+/*!
+ * \brief Detect the LCA position of Buffer access.

Review comment:
       Great suggestion! 




-- 
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.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org



[GitHub] [tvm] junrushao1994 merged pull request #7848: [TensorIR][M1c] LCA detector

Posted by GitBox <gi...@apache.org>.
junrushao1994 merged pull request #7848:
URL: https://github.com/apache/tvm/pull/7848


   


-- 
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.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org



[GitHub] [tvm] tqchen commented on pull request #7848: [TensorIR][M1c] LCA detector

Posted by GitBox <gi...@apache.org>.
tqchen commented on pull request #7848:
URL: https://github.com/apache/tvm/pull/7848#issuecomment-819512867


   @Hzfengsy please fix the lint error


-- 
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.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org



[GitHub] [tvm] junrushao1994 commented on a change in pull request #7848: [TensorIR][M1c] LCA detector

Posted by GitBox <gi...@apache.org>.
junrushao1994 commented on a change in pull request #7848:
URL: https://github.com/apache/tvm/pull/7848#discussion_r612990969



##########
File path: python/tvm/tir/analysis/analysis.py
##########
@@ -16,9 +16,14 @@
 # under the License.
 """Wrapping existing analysis utils."""
 # pylint: disable=invalid-name
-
+from __future__ import annotations

Review comment:
       I hit an issue a while ago that the current TVM version probably didn't support this :-(




-- 
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.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org



[GitHub] [tvm] junrushao1994 commented on a change in pull request #7848: [TensorIR][M1c] LCA detector

Posted by GitBox <gi...@apache.org>.
junrushao1994 commented on a change in pull request #7848:
URL: https://github.com/apache/tvm/pull/7848#discussion_r612990969



##########
File path: python/tvm/tir/analysis/analysis.py
##########
@@ -16,9 +16,14 @@
 # under the License.
 """Wrapping existing analysis utils."""
 # pylint: disable=invalid-name
-
+from __future__ import annotations

Review comment:
       I hit an issue a while ago that the current python version of the TVM repo probably didn't support this :-(




-- 
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.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org



[GitHub] [tvm] Hzfengsy commented on a change in pull request #7848: [TensorIR][M1c] LCA detector

Posted by GitBox <gi...@apache.org>.
Hzfengsy commented on a change in pull request #7848:
URL: https://github.com/apache/tvm/pull/7848#discussion_r612997463



##########
File path: tests/python/unittest/test_tir_analysis_detect_buffer_access_lca.py
##########
@@ -0,0 +1,107 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements.  See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership.  The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License.  You may obtain a copy of the License at
+#
+#   http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied.  See the License for the
+# specific language governing permissions and limitations
+# under the License.
+import tvm
+from tvm import tir
+from tvm.script import ty
+
+
+@tvm.script.tir
+def buffer_laod_store_func(a: ty.handle, b: ty.handle) -> None:
+    A = tir.match_buffer(a, (128, 128), "float32")
+    B = tir.match_buffer(b, (128, 128), "float32")
+    C = tir.alloc_buffer((128, 128), "float32")
+    D = tir.alloc_buffer((128, 128), "float32")
+    with tir.block([128, 128]) as [i, j]:
+        A[i, j] = tir.float32(0)
+    with tir.block([32, 32, tir.reduce_axis(0, 32)]) as [i, j, k]:
+        with tir.init():
+            for ii, jj in tir.grid(4, 4):
+                B[i * 4 + ii, j * 4 + jj] = A[i * 4 + ii, j * 4 + jj]
+        for ii, jj in tir.grid(4, 4):
+            for kk in range(0, 4):
+                B[i * 4 + ii, j * 4 + jj] += C[i * 4 + ii, k * 4 + kk]
+            for kk in range(0, 4):
+                B[i * 4 + ii, j * 4 + jj] += D[j * 4 + jj, k * 4 + kk] * C[i * 4 + ii, k * 4 + kk]
+
+
+@tvm.script.tir
+def buffer_opaque_access(b: ty.handle, c: ty.handle) -> None:
+    B = tir.match_buffer(b, [16, 16], "float32")
+    C = tir.match_buffer(c, [16, 16], "float32")
+
+    with tir.block([]):
+        tir.reads([])
+        tir.writes(B[0:16, 0:16])
+        A = tir.allocate([256], "float32", "global")
+        for i, j in tir.grid(16, 16):
+            tir.store(A, i * 16 + j, 1)
+        for i in range(0, 16):
+            for j in range(0, 16):
+                tir.evaluate(tir.load("float32", A, i * 16 + j))
+            for j in range(0, 16):
+                tir.evaluate(
+                    tir.tvm_fill_fragment(B.data, 16, 16, 16, 0, tir.float32(0), dtype="handle")
+                )
+
+    for i, j in tir.grid(16, 16):
+        with tir.block([16, 16]) as [vi, vj]:
+            tir.bind(vi, i)
+            tir.bind(vj, j)
+            C[vi, vj] = B[vi, vj]
+
+
+def test_buffer_laod_store():

Review comment:
       Thanks for pointing out the typo.




-- 
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.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org



[GitHub] [tvm] xqdan commented on a change in pull request #7848: [TensorIR][M1c] LCA detector

Posted by GitBox <gi...@apache.org>.
xqdan commented on a change in pull request #7848:
URL: https://github.com/apache/tvm/pull/7848#discussion_r612985703



##########
File path: tests/python/unittest/test_tir_analysis_detect_buffer_access_lca.py
##########
@@ -0,0 +1,107 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements.  See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership.  The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License.  You may obtain a copy of the License at
+#
+#   http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied.  See the License for the
+# specific language governing permissions and limitations
+# under the License.
+import tvm
+from tvm import tir
+from tvm.script import ty
+
+
+@tvm.script.tir
+def buffer_laod_store_func(a: ty.handle, b: ty.handle) -> None:
+    A = tir.match_buffer(a, (128, 128), "float32")
+    B = tir.match_buffer(b, (128, 128), "float32")
+    C = tir.alloc_buffer((128, 128), "float32")
+    D = tir.alloc_buffer((128, 128), "float32")
+    with tir.block([128, 128]) as [i, j]:
+        A[i, j] = tir.float32(0)
+    with tir.block([32, 32, tir.reduce_axis(0, 32)]) as [i, j, k]:
+        with tir.init():
+            for ii, jj in tir.grid(4, 4):
+                B[i * 4 + ii, j * 4 + jj] = A[i * 4 + ii, j * 4 + jj]
+        for ii, jj in tir.grid(4, 4):
+            for kk in range(0, 4):
+                B[i * 4 + ii, j * 4 + jj] += C[i * 4 + ii, k * 4 + kk]
+            for kk in range(0, 4):
+                B[i * 4 + ii, j * 4 + jj] += D[j * 4 + jj, k * 4 + kk] * C[i * 4 + ii, k * 4 + kk]
+
+
+@tvm.script.tir
+def buffer_opaque_access(b: ty.handle, c: ty.handle) -> None:
+    B = tir.match_buffer(b, [16, 16], "float32")
+    C = tir.match_buffer(c, [16, 16], "float32")
+
+    with tir.block([]):
+        tir.reads([])
+        tir.writes(B[0:16, 0:16])
+        A = tir.allocate([256], "float32", "global")
+        for i, j in tir.grid(16, 16):
+            tir.store(A, i * 16 + j, 1)
+        for i in range(0, 16):
+            for j in range(0, 16):
+                tir.evaluate(tir.load("float32", A, i * 16 + j))
+            for j in range(0, 16):
+                tir.evaluate(
+                    tir.tvm_fill_fragment(B.data, 16, 16, 16, 0, tir.float32(0), dtype="handle")
+                )
+
+    for i, j in tir.grid(16, 16):
+        with tir.block([16, 16]) as [vi, vj]:
+            tir.bind(vi, i)
+            tir.bind(vj, j)
+            C[vi, vj] = B[vi, vj]
+
+
+def test_buffer_laod_store():

Review comment:
       laod ->load

##########
File path: tests/python/unittest/test_tir_analysis_detect_buffer_access_lca.py
##########
@@ -0,0 +1,107 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements.  See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership.  The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License.  You may obtain a copy of the License at
+#
+#   http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied.  See the License for the
+# specific language governing permissions and limitations
+# under the License.
+import tvm
+from tvm import tir
+from tvm.script import ty
+
+
+@tvm.script.tir
+def buffer_laod_store_func(a: ty.handle, b: ty.handle) -> None:

Review comment:
       laod -> load

##########
File path: src/tir/analysis/buffer_access_lca_detector.cc
##########
@@ -0,0 +1,173 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *   http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ * KIND, either express or implied.  See the License for the
+ * specific language governing permissions and limitations
+ * under the License.
+ */
+
+/*!
+ * \file tir/analysis/buffer_access_lca_detector.cc
+ * \brief Detect LCA of buffer access
+ */
+
+#include <tvm/tir/analysis.h>
+#include <tvm/tir/stmt_functor.h>
+
+#include "../../support/arena.h"
+
+namespace tvm {
+namespace tir {
+
+/*!
+ * \brief Detect the LCA position of Buffer access.

Review comment:
       LCA = LowestCommonAncestor, or life-cycle analysis? Add the full name of LCA would be better.




-- 
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.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org