You are viewing a plain text version of this content. The canonical link for it is here.
Posted to github@arrow.apache.org by GitBox <gi...@apache.org> on 2020/08/12 17:04:12 UTC
[GitHub] [arrow] pitrou commented on a change in pull request #7940: ARROW-9702: [C++] Move bpacking SIMD to runtime path.
pitrou commented on a change in pull request #7940:
URL: https://github.com/apache/arrow/pull/7940#discussion_r469406589
##########
File path: cpp/src/arrow/util/bpacking.cc
##########
@@ -0,0 +1,174 @@
+// 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.
+
+#include <mutex>
+
+#include "arrow/util/bpacking.h"
+#include "arrow/util/bpacking_default.h"
+#include "arrow/util/cpu_info.h"
+#include "arrow/util/logging.h"
+
+#if defined(ARROW_HAVE_RUNTIME_AVX2)
+#include "arrow/util/bpacking_avx2.h"
+#endif
+#if defined(ARROW_HAVE_RUNTIME_AVX512)
+#include "arrow/util/bpacking_avx512.h"
+#endif
+
+namespace arrow {
+namespace internal {
+
+int unpack32_default(const uint32_t* in, uint32_t* out, int batch_size, int num_bits) {
+ batch_size = batch_size / 32 * 32;
+ int num_loops = batch_size / 32;
+
+ switch (num_bits) {
+ case 0:
+ for (int i = 0; i < num_loops; ++i) in = nullunpacker32(in, out + i * 32);
+ break;
+ case 1:
+ for (int i = 0; i < num_loops; ++i) in = unpack1_32(in, out + i * 32);
+ break;
+ case 2:
+ for (int i = 0; i < num_loops; ++i) in = unpack2_32(in, out + i * 32);
+ break;
+ case 3:
+ for (int i = 0; i < num_loops; ++i) in = unpack3_32(in, out + i * 32);
+ break;
+ case 4:
+ for (int i = 0; i < num_loops; ++i) in = unpack4_32(in, out + i * 32);
+ break;
+ case 5:
+ for (int i = 0; i < num_loops; ++i) in = unpack5_32(in, out + i * 32);
+ break;
+ case 6:
+ for (int i = 0; i < num_loops; ++i) in = unpack6_32(in, out + i * 32);
+ break;
+ case 7:
+ for (int i = 0; i < num_loops; ++i) in = unpack7_32(in, out + i * 32);
+ break;
+ case 8:
+ for (int i = 0; i < num_loops; ++i) in = unpack8_32(in, out + i * 32);
+ break;
+ case 9:
+ for (int i = 0; i < num_loops; ++i) in = unpack9_32(in, out + i * 32);
+ break;
+ case 10:
+ for (int i = 0; i < num_loops; ++i) in = unpack10_32(in, out + i * 32);
+ break;
+ case 11:
+ for (int i = 0; i < num_loops; ++i) in = unpack11_32(in, out + i * 32);
+ break;
+ case 12:
+ for (int i = 0; i < num_loops; ++i) in = unpack12_32(in, out + i * 32);
+ break;
+ case 13:
+ for (int i = 0; i < num_loops; ++i) in = unpack13_32(in, out + i * 32);
+ break;
+ case 14:
+ for (int i = 0; i < num_loops; ++i) in = unpack14_32(in, out + i * 32);
+ break;
+ case 15:
+ for (int i = 0; i < num_loops; ++i) in = unpack15_32(in, out + i * 32);
+ break;
+ case 16:
+ for (int i = 0; i < num_loops; ++i) in = unpack16_32(in, out + i * 32);
+ break;
+ case 17:
+ for (int i = 0; i < num_loops; ++i) in = unpack17_32(in, out + i * 32);
+ break;
+ case 18:
+ for (int i = 0; i < num_loops; ++i) in = unpack18_32(in, out + i * 32);
+ break;
+ case 19:
+ for (int i = 0; i < num_loops; ++i) in = unpack19_32(in, out + i * 32);
+ break;
+ case 20:
+ for (int i = 0; i < num_loops; ++i) in = unpack20_32(in, out + i * 32);
+ break;
+ case 21:
+ for (int i = 0; i < num_loops; ++i) in = unpack21_32(in, out + i * 32);
+ break;
+ case 22:
+ for (int i = 0; i < num_loops; ++i) in = unpack22_32(in, out + i * 32);
+ break;
+ case 23:
+ for (int i = 0; i < num_loops; ++i) in = unpack23_32(in, out + i * 32);
+ break;
+ case 24:
+ for (int i = 0; i < num_loops; ++i) in = unpack24_32(in, out + i * 32);
+ break;
+ case 25:
+ for (int i = 0; i < num_loops; ++i) in = unpack25_32(in, out + i * 32);
+ break;
+ case 26:
+ for (int i = 0; i < num_loops; ++i) in = unpack26_32(in, out + i * 32);
+ break;
+ case 27:
+ for (int i = 0; i < num_loops; ++i) in = unpack27_32(in, out + i * 32);
+ break;
+ case 28:
+ for (int i = 0; i < num_loops; ++i) in = unpack28_32(in, out + i * 32);
+ break;
+ case 29:
+ for (int i = 0; i < num_loops; ++i) in = unpack29_32(in, out + i * 32);
+ break;
+ case 30:
+ for (int i = 0; i < num_loops; ++i) in = unpack30_32(in, out + i * 32);
+ break;
+ case 31:
+ for (int i = 0; i < num_loops; ++i) in = unpack31_32(in, out + i * 32);
+ break;
+ case 32:
+ for (int i = 0; i < num_loops; ++i) in = unpack32_32(in, out + i * 32);
+ break;
+ default:
+ DCHECK(false) << "Unsupported num_bits";
+ }
+
+ return batch_size;
+}
+
+typedef int (*unpack32_func_t)(const uint32_t* in, uint32_t* out, int batch_size,
+ int num_bits);
+
+static std::once_flag unpack32_ptr_initialized;
Review comment:
At some point it would be nice to have a facility to avoid writing this boilerplate time and time again. Something like this (just a sketch):
```c++
class Unpack32Function : public DynamicDispatch<Unpack32Function> {
using FunctionType = unpack32_func_t;
static std::vector<std::pair<CpuFlag, FunctionType>> implementations() {
return {{NONE, unpack32_default}, {AVX2, unpack32_avx2}, {AVX512, unpack32_avx512}};
}
};
int unpack32(const uint32_t* in, uint32_t* out, int batch_size, int num_bits) {
static Unpack32Function func;
return func(in, ou, batch_size, num_bits);
}
```
----------------------------------------------------------------
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