Details | Last modification | View Log | RSS feed
| Rev | Author | Line No. | Line |
|---|---|---|---|
| 14 | pmbaty | 1 | //===- CodeLayout.h - Code layout/placement algorithms ---------*- C++ -*-===// |
| 2 | // |
||
| 3 | // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. |
||
| 4 | // See https://llvm.org/LICENSE.txt for license information. |
||
| 5 | // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception |
||
| 6 | // |
||
| 7 | //===----------------------------------------------------------------------===// |
||
| 8 | // |
||
| 9 | /// \file |
||
| 10 | /// Declares methods and data structures for code layout algorithms. |
||
| 11 | // |
||
| 12 | //===----------------------------------------------------------------------===// |
||
| 13 | |||
| 14 | #ifndef LLVM_TRANSFORMS_UTILS_CODELAYOUT_H |
||
| 15 | #define LLVM_TRANSFORMS_UTILS_CODELAYOUT_H |
||
| 16 | |||
| 17 | #include "llvm/ADT/DenseMap.h" |
||
| 18 | |||
| 19 | #include <vector> |
||
| 20 | |||
| 21 | namespace llvm { |
||
| 22 | |||
| 23 | using EdgeT = std::pair<uint64_t, uint64_t>; |
||
| 24 | using EdgeCountT = std::pair<EdgeT, uint64_t>; |
||
| 25 | |||
| 26 | /// Find a layout of nodes (basic blocks) of a given CFG optimizing jump |
||
| 27 | /// locality and thus processor I-cache utilization. This is achieved via |
||
| 28 | /// increasing the number of fall-through jumps and co-locating frequently |
||
| 29 | /// executed nodes together. |
||
| 30 | /// The nodes are assumed to be indexed by integers from [0, |V|) so that the |
||
| 31 | /// current order is the identity permutation. |
||
| 32 | /// \p NodeSizes: The sizes of the nodes (in bytes). |
||
| 33 | /// \p NodeCounts: The execution counts of the nodes in the profile. |
||
| 34 | /// \p EdgeCounts: The execution counts of every edge (jump) in the profile. The |
||
| 35 | /// map also defines the edges in CFG and should include 0-count edges. |
||
| 36 | /// \returns The best block order found. |
||
| 37 | std::vector<uint64_t> |
||
| 38 | applyExtTspLayout(const std::vector<uint64_t> &NodeSizes, |
||
| 39 | const std::vector<uint64_t> &NodeCounts, |
||
| 40 | const std::vector<EdgeCountT> &EdgeCounts); |
||
| 41 | |||
| 42 | /// Estimate the "quality" of a given node order in CFG. The higher the score, |
||
| 43 | /// the better the order is. The score is designed to reflect the locality of |
||
| 44 | /// the given order, which is anti-correlated with the number of I-cache misses |
||
| 45 | /// in a typical execution of the function. |
||
| 46 | double calcExtTspScore(const std::vector<uint64_t> &Order, |
||
| 47 | const std::vector<uint64_t> &NodeSizes, |
||
| 48 | const std::vector<uint64_t> &NodeCounts, |
||
| 49 | const std::vector<EdgeCountT> &EdgeCounts); |
||
| 50 | |||
| 51 | /// Estimate the "quality" of the current node order in CFG. |
||
| 52 | double calcExtTspScore(const std::vector<uint64_t> &NodeSizes, |
||
| 53 | const std::vector<uint64_t> &NodeCounts, |
||
| 54 | const std::vector<EdgeCountT> &EdgeCounts); |
||
| 55 | |||
| 56 | } // end namespace llvm |
||
| 57 | |||
| 58 | #endif // LLVM_TRANSFORMS_UTILS_CODELAYOUT_H |