github/verilator
1
0
Fork 1
mirror of https://github.com/verilator/verilator.git synced 2025-02-08 14:41:48 +00:00
Code Issues Packages Projects Releases Wiki Activity
d5b03ab721
verilator/src/V3DfgPeephole.h

135 lines
7.5 KiB
C
Raw Normal View History

Introduce DFG based combinational logic optimizer (#3527) Added a new data-flow graph (DFG) based combinational logic optimizer. The capabilities of this covers a combination of V3Const and V3Gate, but is also more capable of transforming combinational logic into simplified forms and more. This entail adding a new internal representation, `DfgGraph`, and appropriate `astToDfg` and `dfgToAst` conversion functions. The graph represents some of the combinational equations (~continuous assignments) in a module, and for the duration of the DFG passes, it takes over the role of AstModule. A bulk of the Dfg vertices represent expressions. These vertex classes, and the corresponding conversions to/from AST are mostly auto-generated by astgen, together with a DfgVVisitor that can be used for dynamic dispatch based on vertex (operation) types. The resulting combinational logic graph (a `DfgGraph`) is then optimized in various ways. Currently we perform common sub-expression elimination, variable inlining, and some specific peephole optimizations, but there is scope for more optimizations in the future using the same representation. The optimizer is run directly before and after inlining. The pre inline pass can operate on smaller graphs and hence converges faster, but still has a chance of substantially reducing the size of the logic on some designs, making inlining both faster and less memory intensive. The post inline pass can then optimize across the inlined module boundaries. No optimization is performed across a module boundary. For debugging purposes, each peephole optimization can be disabled individually via the -fno-dfg-peepnole-<OPT> option, where <OPT> is one of the optimizations listed in V3DfgPeephole.h, for example -fno-dfg-peephole-remove-not-not. The peephole patterns currently implemented were mostly picked based on the design that inspired this work, and on that design the optimizations yields ~30% single threaded speedup, and ~50% speedup on 4 threads. As you can imagine not having to haul around redundant combinational networks in the rest of the compilation pipeline also helps with memory consumption, and up to 30% peak memory usage of Verilator was observed on the same design. Gains on other arbitrary designs are smaller (and can be improved by analyzing those designs). For example OpenTitan gains between 1-15% speedup depending on build type.
2022-09-23 15:46:22 +00:00
// -*- mode: C++; c-file-style: "cc-mode" -*-
//*************************************************************************
// DESCRIPTION: Verilator: Peephole optimizations over DfgGraph
//
// Code available from: https://verilator.org
//
//*************************************************************************
//
Copyright year update
2023-01-01 15:18:39 +00:00
// Copyright 2003-2023 by Wilson Snyder. This program is free software; you
Introduce DFG based combinational logic optimizer (#3527) Added a new data-flow graph (DFG) based combinational logic optimizer. The capabilities of this covers a combination of V3Const and V3Gate, but is also more capable of transforming combinational logic into simplified forms and more. This entail adding a new internal representation, `DfgGraph`, and appropriate `astToDfg` and `dfgToAst` conversion functions. The graph represents some of the combinational equations (~continuous assignments) in a module, and for the duration of the DFG passes, it takes over the role of AstModule. A bulk of the Dfg vertices represent expressions. These vertex classes, and the corresponding conversions to/from AST are mostly auto-generated by astgen, together with a DfgVVisitor that can be used for dynamic dispatch based on vertex (operation) types. The resulting combinational logic graph (a `DfgGraph`) is then optimized in various ways. Currently we perform common sub-expression elimination, variable inlining, and some specific peephole optimizations, but there is scope for more optimizations in the future using the same representation. The optimizer is run directly before and after inlining. The pre inline pass can operate on smaller graphs and hence converges faster, but still has a chance of substantially reducing the size of the logic on some designs, making inlining both faster and less memory intensive. The post inline pass can then optimize across the inlined module boundaries. No optimization is performed across a module boundary. For debugging purposes, each peephole optimization can be disabled individually via the -fno-dfg-peepnole-<OPT> option, where <OPT> is one of the optimizations listed in V3DfgPeephole.h, for example -fno-dfg-peephole-remove-not-not. The peephole patterns currently implemented were mostly picked based on the design that inspired this work, and on that design the optimizations yields ~30% single threaded speedup, and ~50% speedup on 4 threads. As you can imagine not having to haul around redundant combinational networks in the rest of the compilation pipeline also helps with memory consumption, and up to 30% peak memory usage of Verilator was observed on the same design. Gains on other arbitrary designs are smaller (and can be improved by analyzing those designs). For example OpenTitan gains between 1-15% speedup depending on build type.
2022-09-23 15:46:22 +00:00
// can redistribute it and/or modify it under the terms of either the GNU
// Lesser General Public License Version 3 or the Perl Artistic License
// Version 2.0.
// SPDX-License-Identifier: LGPL-3.0-only OR Artistic-2.0
//
//*************************************************************************
#ifndef VERILATOR_V3DFGPEEPHOLE_H_
#define VERILATOR_V3DFGPEEPHOLE_H_
#include "config_build.h"
#include <V3Stats.h>
#define _FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, arg) macro(arg, #arg)
DFG: Add more peephole patterns
2022-09-30 15:19:53 +00:00
// Enumeration of each peephole optimization. Must be kept in sorted order (enforced by tests).
Introduce DFG based combinational logic optimizer (#3527) Added a new data-flow graph (DFG) based combinational logic optimizer. The capabilities of this covers a combination of V3Const and V3Gate, but is also more capable of transforming combinational logic into simplified forms and more. This entail adding a new internal representation, `DfgGraph`, and appropriate `astToDfg` and `dfgToAst` conversion functions. The graph represents some of the combinational equations (~continuous assignments) in a module, and for the duration of the DFG passes, it takes over the role of AstModule. A bulk of the Dfg vertices represent expressions. These vertex classes, and the corresponding conversions to/from AST are mostly auto-generated by astgen, together with a DfgVVisitor that can be used for dynamic dispatch based on vertex (operation) types. The resulting combinational logic graph (a `DfgGraph`) is then optimized in various ways. Currently we perform common sub-expression elimination, variable inlining, and some specific peephole optimizations, but there is scope for more optimizations in the future using the same representation. The optimizer is run directly before and after inlining. The pre inline pass can operate on smaller graphs and hence converges faster, but still has a chance of substantially reducing the size of the logic on some designs, making inlining both faster and less memory intensive. The post inline pass can then optimize across the inlined module boundaries. No optimization is performed across a module boundary. For debugging purposes, each peephole optimization can be disabled individually via the -fno-dfg-peepnole-<OPT> option, where <OPT> is one of the optimizations listed in V3DfgPeephole.h, for example -fno-dfg-peephole-remove-not-not. The peephole patterns currently implemented were mostly picked based on the design that inspired this work, and on that design the optimizations yields ~30% single threaded speedup, and ~50% speedup on 4 threads. As you can imagine not having to haul around redundant combinational networks in the rest of the compilation pipeline also helps with memory consumption, and up to 30% peak memory usage of Verilator was observed on the same design. Gains on other arbitrary designs are smaller (and can be improved by analyzing those designs). For example OpenTitan gains between 1-15% speedup depending on build type.
2022-09-23 15:46:22 +00:00
// clang-format off
#define FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION(macro) \
DfgPeephole: Rework folding of associative operations Allow constant folding through adjacent nodes of all associative operations, for example '((a & 2) & 3)' or '(3 & (2 & a))' can now be folded into '(a & 2)' and '(2 & a)' respectively. Also improve speed of making associative expression trees right leaning by using rotation of the existing vertices whenever instead of allocation of new nodes.
2022-10-05 09:36:28 +00:00
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, FOLD_ASSOC_BINARY) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, FOLD_ASSOC_BINARY_LHS_OF_RHS) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, FOLD_ASSOC_BINARY_RHS_OF_LHS) \
DfgPeephole: add constant folding for all integer types Also added a testing only -fno-const-before-dfg option, as otherwise V3Const eats up a lot of the simple inputs. A lot of the things V3Const swallows in the simple cases can make it to DFG in complex cases, or DFG itself can create them during optimization. In any case to save complexity of testing DFG constant folding, we use this option to turn off V3Const prior to the DFG passes in the relevant test.
2022-10-04 20:18:39 +00:00
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, FOLD_BINARY) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, FOLD_SEL) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, FOLD_UNARY) \
DFG: Add more peephole patterns
2022-09-30 15:19:53 +00:00
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, INLINE_ARRAYSEL) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, PULL_NOTS_THROUGH_COND) \
Introduce DFG based combinational logic optimizer (#3527) Added a new data-flow graph (DFG) based combinational logic optimizer. The capabilities of this covers a combination of V3Const and V3Gate, but is also more capable of transforming combinational logic into simplified forms and more. This entail adding a new internal representation, `DfgGraph`, and appropriate `astToDfg` and `dfgToAst` conversion functions. The graph represents some of the combinational equations (~continuous assignments) in a module, and for the duration of the DFG passes, it takes over the role of AstModule. A bulk of the Dfg vertices represent expressions. These vertex classes, and the corresponding conversions to/from AST are mostly auto-generated by astgen, together with a DfgVVisitor that can be used for dynamic dispatch based on vertex (operation) types. The resulting combinational logic graph (a `DfgGraph`) is then optimized in various ways. Currently we perform common sub-expression elimination, variable inlining, and some specific peephole optimizations, but there is scope for more optimizations in the future using the same representation. The optimizer is run directly before and after inlining. The pre inline pass can operate on smaller graphs and hence converges faster, but still has a chance of substantially reducing the size of the logic on some designs, making inlining both faster and less memory intensive. The post inline pass can then optimize across the inlined module boundaries. No optimization is performed across a module boundary. For debugging purposes, each peephole optimization can be disabled individually via the -fno-dfg-peepnole-<OPT> option, where <OPT> is one of the optimizations listed in V3DfgPeephole.h, for example -fno-dfg-peephole-remove-not-not. The peephole patterns currently implemented were mostly picked based on the design that inspired this work, and on that design the optimizations yields ~30% single threaded speedup, and ~50% speedup on 4 threads. As you can imagine not having to haul around redundant combinational networks in the rest of the compilation pipeline also helps with memory consumption, and up to 30% peak memory usage of Verilator was observed on the same design. Gains on other arbitrary designs are smaller (and can be improved by analyzing those designs). For example OpenTitan gains between 1-15% speedup depending on build type.
2022-09-23 15:46:22 +00:00
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, PUSH_BITWISE_OP_THROUGH_CONCAT) \
DFG: Add more peephole patterns
2022-09-30 15:19:53 +00:00
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, PUSH_BITWISE_THROUGH_REDUCTION) \
Introduce DFG based combinational logic optimizer (#3527) Added a new data-flow graph (DFG) based combinational logic optimizer. The capabilities of this covers a combination of V3Const and V3Gate, but is also more capable of transforming combinational logic into simplified forms and more. This entail adding a new internal representation, `DfgGraph`, and appropriate `astToDfg` and `dfgToAst` conversion functions. The graph represents some of the combinational equations (~continuous assignments) in a module, and for the duration of the DFG passes, it takes over the role of AstModule. A bulk of the Dfg vertices represent expressions. These vertex classes, and the corresponding conversions to/from AST are mostly auto-generated by astgen, together with a DfgVVisitor that can be used for dynamic dispatch based on vertex (operation) types. The resulting combinational logic graph (a `DfgGraph`) is then optimized in various ways. Currently we perform common sub-expression elimination, variable inlining, and some specific peephole optimizations, but there is scope for more optimizations in the future using the same representation. The optimizer is run directly before and after inlining. The pre inline pass can operate on smaller graphs and hence converges faster, but still has a chance of substantially reducing the size of the logic on some designs, making inlining both faster and less memory intensive. The post inline pass can then optimize across the inlined module boundaries. No optimization is performed across a module boundary. For debugging purposes, each peephole optimization can be disabled individually via the -fno-dfg-peepnole-<OPT> option, where <OPT> is one of the optimizations listed in V3DfgPeephole.h, for example -fno-dfg-peephole-remove-not-not. The peephole patterns currently implemented were mostly picked based on the design that inspired this work, and on that design the optimizations yields ~30% single threaded speedup, and ~50% speedup on 4 threads. As you can imagine not having to haul around redundant combinational networks in the rest of the compilation pipeline also helps with memory consumption, and up to 30% peak memory usage of Verilator was observed on the same design. Gains on other arbitrary designs are smaller (and can be improved by analyzing those designs). For example OpenTitan gains between 1-15% speedup depending on build type.
2022-09-23 15:46:22 +00:00
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, PUSH_COMPARE_OP_THROUGH_CONCAT) \
DFG: Add more peephole patterns
2022-09-30 15:19:53 +00:00
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, PUSH_CONCAT_THROUGH_NOTS) \
Introduce DFG based combinational logic optimizer (#3527) Added a new data-flow graph (DFG) based combinational logic optimizer. The capabilities of this covers a combination of V3Const and V3Gate, but is also more capable of transforming combinational logic into simplified forms and more. This entail adding a new internal representation, `DfgGraph`, and appropriate `astToDfg` and `dfgToAst` conversion functions. The graph represents some of the combinational equations (~continuous assignments) in a module, and for the duration of the DFG passes, it takes over the role of AstModule. A bulk of the Dfg vertices represent expressions. These vertex classes, and the corresponding conversions to/from AST are mostly auto-generated by astgen, together with a DfgVVisitor that can be used for dynamic dispatch based on vertex (operation) types. The resulting combinational logic graph (a `DfgGraph`) is then optimized in various ways. Currently we perform common sub-expression elimination, variable inlining, and some specific peephole optimizations, but there is scope for more optimizations in the future using the same representation. The optimizer is run directly before and after inlining. The pre inline pass can operate on smaller graphs and hence converges faster, but still has a chance of substantially reducing the size of the logic on some designs, making inlining both faster and less memory intensive. The post inline pass can then optimize across the inlined module boundaries. No optimization is performed across a module boundary. For debugging purposes, each peephole optimization can be disabled individually via the -fno-dfg-peepnole-<OPT> option, where <OPT> is one of the optimizations listed in V3DfgPeephole.h, for example -fno-dfg-peephole-remove-not-not. The peephole patterns currently implemented were mostly picked based on the design that inspired this work, and on that design the optimizations yields ~30% single threaded speedup, and ~50% speedup on 4 threads. As you can imagine not having to haul around redundant combinational networks in the rest of the compilation pipeline also helps with memory consumption, and up to 30% peak memory usage of Verilator was observed on the same design. Gains on other arbitrary designs are smaller (and can be improved by analyzing those designs). For example OpenTitan gains between 1-15% speedup depending on build type.
2022-09-23 15:46:22 +00:00
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, PUSH_NOT_THROUGH_COND) \
DFG: Add more peephole patterns
2022-09-30 15:19:53 +00:00
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, PUSH_REDUCTION_THROUGH_CONCAT) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, PUSH_REDUCTION_THROUGH_COND_WITH_CONST_BRANCH) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, PUSH_SEL_THROUGH_CONCAT) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, PUSH_SEL_THROUGH_COND) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, PUSH_SEL_THROUGH_NOT) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, PUSH_SEL_THROUGH_REPLICATE) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, PUSH_SEL_THROUGH_SHIFTL) \
Introduce DFG based combinational logic optimizer (#3527) Added a new data-flow graph (DFG) based combinational logic optimizer. The capabilities of this covers a combination of V3Const and V3Gate, but is also more capable of transforming combinational logic into simplified forms and more. This entail adding a new internal representation, `DfgGraph`, and appropriate `astToDfg` and `dfgToAst` conversion functions. The graph represents some of the combinational equations (~continuous assignments) in a module, and for the duration of the DFG passes, it takes over the role of AstModule. A bulk of the Dfg vertices represent expressions. These vertex classes, and the corresponding conversions to/from AST are mostly auto-generated by astgen, together with a DfgVVisitor that can be used for dynamic dispatch based on vertex (operation) types. The resulting combinational logic graph (a `DfgGraph`) is then optimized in various ways. Currently we perform common sub-expression elimination, variable inlining, and some specific peephole optimizations, but there is scope for more optimizations in the future using the same representation. The optimizer is run directly before and after inlining. The pre inline pass can operate on smaller graphs and hence converges faster, but still has a chance of substantially reducing the size of the logic on some designs, making inlining both faster and less memory intensive. The post inline pass can then optimize across the inlined module boundaries. No optimization is performed across a module boundary. For debugging purposes, each peephole optimization can be disabled individually via the -fno-dfg-peepnole-<OPT> option, where <OPT> is one of the optimizations listed in V3DfgPeephole.h, for example -fno-dfg-peephole-remove-not-not. The peephole patterns currently implemented were mostly picked based on the design that inspired this work, and on that design the optimizations yields ~30% single threaded speedup, and ~50% speedup on 4 threads. As you can imagine not having to haul around redundant combinational networks in the rest of the compilation pipeline also helps with memory consumption, and up to 30% peak memory usage of Verilator was observed on the same design. Gains on other arbitrary designs are smaller (and can be improved by analyzing those designs). For example OpenTitan gains between 1-15% speedup depending on build type.
2022-09-23 15:46:22 +00:00
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REMOVE_AND_WITH_ONES) \
DFG: Add more peephole patterns
2022-09-30 15:19:53 +00:00
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REMOVE_CONCAT_OF_ADJOINING_SELS) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REMOVE_COND_WITH_FALSE_CONDITION) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REMOVE_COND_WITH_TRUE_CONDITION) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REMOVE_FULL_WIDTH_SEL) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REMOVE_NOT_NOT) \
Introduce DFG based combinational logic optimizer (#3527) Added a new data-flow graph (DFG) based combinational logic optimizer. The capabilities of this covers a combination of V3Const and V3Gate, but is also more capable of transforming combinational logic into simplified forms and more. This entail adding a new internal representation, `DfgGraph`, and appropriate `astToDfg` and `dfgToAst` conversion functions. The graph represents some of the combinational equations (~continuous assignments) in a module, and for the duration of the DFG passes, it takes over the role of AstModule. A bulk of the Dfg vertices represent expressions. These vertex classes, and the corresponding conversions to/from AST are mostly auto-generated by astgen, together with a DfgVVisitor that can be used for dynamic dispatch based on vertex (operation) types. The resulting combinational logic graph (a `DfgGraph`) is then optimized in various ways. Currently we perform common sub-expression elimination, variable inlining, and some specific peephole optimizations, but there is scope for more optimizations in the future using the same representation. The optimizer is run directly before and after inlining. The pre inline pass can operate on smaller graphs and hence converges faster, but still has a chance of substantially reducing the size of the logic on some designs, making inlining both faster and less memory intensive. The post inline pass can then optimize across the inlined module boundaries. No optimization is performed across a module boundary. For debugging purposes, each peephole optimization can be disabled individually via the -fno-dfg-peepnole-<OPT> option, where <OPT> is one of the optimizations listed in V3DfgPeephole.h, for example -fno-dfg-peephole-remove-not-not. The peephole patterns currently implemented were mostly picked based on the design that inspired this work, and on that design the optimizations yields ~30% single threaded speedup, and ~50% speedup on 4 threads. As you can imagine not having to haul around redundant combinational networks in the rest of the compilation pipeline also helps with memory consumption, and up to 30% peak memory usage of Verilator was observed on the same design. Gains on other arbitrary designs are smaller (and can be improved by analyzing those designs). For example OpenTitan gains between 1-15% speedup depending on build type.
2022-09-23 15:46:22 +00:00
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REMOVE_OR_WITH_ZERO) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REMOVE_REDUNDANT_ZEXT_ON_RHS_OF_SHIFT) \
DfgPeephole: add constant folding for all integer types Also added a testing only -fno-const-before-dfg option, as otherwise V3Const eats up a lot of the simple inputs. A lot of the things V3Const swallows in the simple cases can make it to DFG in complex cases, or DFG itself can create them during optimization. In any case to save complexity of testing DFG constant folding, we use this option to turn off V3Const prior to the DFG passes in the relevant test.
2022-10-04 20:18:39 +00:00
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REMOVE_REPLICATE_ONCE) \
Introduce DFG based combinational logic optimizer (#3527) Added a new data-flow graph (DFG) based combinational logic optimizer. The capabilities of this covers a combination of V3Const and V3Gate, but is also more capable of transforming combinational logic into simplified forms and more. This entail adding a new internal representation, `DfgGraph`, and appropriate `astToDfg` and `dfgToAst` conversion functions. The graph represents some of the combinational equations (~continuous assignments) in a module, and for the duration of the DFG passes, it takes over the role of AstModule. A bulk of the Dfg vertices represent expressions. These vertex classes, and the corresponding conversions to/from AST are mostly auto-generated by astgen, together with a DfgVVisitor that can be used for dynamic dispatch based on vertex (operation) types. The resulting combinational logic graph (a `DfgGraph`) is then optimized in various ways. Currently we perform common sub-expression elimination, variable inlining, and some specific peephole optimizations, but there is scope for more optimizations in the future using the same representation. The optimizer is run directly before and after inlining. The pre inline pass can operate on smaller graphs and hence converges faster, but still has a chance of substantially reducing the size of the logic on some designs, making inlining both faster and less memory intensive. The post inline pass can then optimize across the inlined module boundaries. No optimization is performed across a module boundary. For debugging purposes, each peephole optimization can be disabled individually via the -fno-dfg-peepnole-<OPT> option, where <OPT> is one of the optimizations listed in V3DfgPeephole.h, for example -fno-dfg-peephole-remove-not-not. The peephole patterns currently implemented were mostly picked based on the design that inspired this work, and on that design the optimizations yields ~30% single threaded speedup, and ~50% speedup on 4 threads. As you can imagine not having to haul around redundant combinational networks in the rest of the compilation pipeline also helps with memory consumption, and up to 30% peak memory usage of Verilator was observed on the same design. Gains on other arbitrary designs are smaller (and can be improved by analyzing those designs). For example OpenTitan gains between 1-15% speedup depending on build type.
2022-09-23 15:46:22 +00:00
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REMOVE_SEL_FROM_LHS_OF_CONCAT) \
DFG: Add more peephole patterns
2022-09-30 15:19:53 +00:00
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REMOVE_SEL_FROM_RHS_OF_CONCAT) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REMOVE_SUB_ZERO) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REMOVE_WIDTH_ONE_REDUCTION) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REMOVE_XOR_WITH_ZERO) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REPLACE_AND_OF_NOT_AND_NEQ) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REPLACE_AND_OF_NOT_AND_NOT) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REPLACE_AND_WITH_ZERO) \
Introduce DFG based combinational logic optimizer (#3527) Added a new data-flow graph (DFG) based combinational logic optimizer. The capabilities of this covers a combination of V3Const and V3Gate, but is also more capable of transforming combinational logic into simplified forms and more. This entail adding a new internal representation, `DfgGraph`, and appropriate `astToDfg` and `dfgToAst` conversion functions. The graph represents some of the combinational equations (~continuous assignments) in a module, and for the duration of the DFG passes, it takes over the role of AstModule. A bulk of the Dfg vertices represent expressions. These vertex classes, and the corresponding conversions to/from AST are mostly auto-generated by astgen, together with a DfgVVisitor that can be used for dynamic dispatch based on vertex (operation) types. The resulting combinational logic graph (a `DfgGraph`) is then optimized in various ways. Currently we perform common sub-expression elimination, variable inlining, and some specific peephole optimizations, but there is scope for more optimizations in the future using the same representation. The optimizer is run directly before and after inlining. The pre inline pass can operate on smaller graphs and hence converges faster, but still has a chance of substantially reducing the size of the logic on some designs, making inlining both faster and less memory intensive. The post inline pass can then optimize across the inlined module boundaries. No optimization is performed across a module boundary. For debugging purposes, each peephole optimization can be disabled individually via the -fno-dfg-peepnole-<OPT> option, where <OPT> is one of the optimizations listed in V3DfgPeephole.h, for example -fno-dfg-peephole-remove-not-not. The peephole patterns currently implemented were mostly picked based on the design that inspired this work, and on that design the optimizations yields ~30% single threaded speedup, and ~50% speedup on 4 threads. As you can imagine not having to haul around redundant combinational networks in the rest of the compilation pipeline also helps with memory consumption, and up to 30% peak memory usage of Verilator was observed on the same design. Gains on other arbitrary designs are smaller (and can be improved by analyzing those designs). For example OpenTitan gains between 1-15% speedup depending on build type.
2022-09-23 15:46:22 +00:00
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REPLACE_CONCAT_SEL_BOTTOM_AND_ZERO_WITH_SHIFTL) \
DFG: Add more peephole patterns
2022-09-30 15:19:53 +00:00
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REPLACE_CONCAT_ZERO_AND_SEL_TOP_WITH_SHIFTR) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REPLACE_COND_DEC) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REPLACE_COND_INC) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REPLACE_COND_WITH_ELSE_BRANCH_ONES) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REPLACE_COND_WITH_ELSE_BRANCH_ZERO) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REPLACE_COND_WITH_THEN_BRANCH_ONES) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REPLACE_COND_WITH_THEN_BRANCH_ZERO) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REPLACE_CONTRADICTORY_AND) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REPLACE_EXTEND) \
Introduce DFG based combinational logic optimizer (#3527) Added a new data-flow graph (DFG) based combinational logic optimizer. The capabilities of this covers a combination of V3Const and V3Gate, but is also more capable of transforming combinational logic into simplified forms and more. This entail adding a new internal representation, `DfgGraph`, and appropriate `astToDfg` and `dfgToAst` conversion functions. The graph represents some of the combinational equations (~continuous assignments) in a module, and for the duration of the DFG passes, it takes over the role of AstModule. A bulk of the Dfg vertices represent expressions. These vertex classes, and the corresponding conversions to/from AST are mostly auto-generated by astgen, together with a DfgVVisitor that can be used for dynamic dispatch based on vertex (operation) types. The resulting combinational logic graph (a `DfgGraph`) is then optimized in various ways. Currently we perform common sub-expression elimination, variable inlining, and some specific peephole optimizations, but there is scope for more optimizations in the future using the same representation. The optimizer is run directly before and after inlining. The pre inline pass can operate on smaller graphs and hence converges faster, but still has a chance of substantially reducing the size of the logic on some designs, making inlining both faster and less memory intensive. The post inline pass can then optimize across the inlined module boundaries. No optimization is performed across a module boundary. For debugging purposes, each peephole optimization can be disabled individually via the -fno-dfg-peepnole-<OPT> option, where <OPT> is one of the optimizations listed in V3DfgPeephole.h, for example -fno-dfg-peephole-remove-not-not. The peephole patterns currently implemented were mostly picked based on the design that inspired this work, and on that design the optimizations yields ~30% single threaded speedup, and ~50% speedup on 4 threads. As you can imagine not having to haul around redundant combinational networks in the rest of the compilation pipeline also helps with memory consumption, and up to 30% peak memory usage of Verilator was observed on the same design. Gains on other arbitrary designs are smaller (and can be improved by analyzing those designs). For example OpenTitan gains between 1-15% speedup depending on build type.
2022-09-23 15:46:22 +00:00
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REPLACE_NESTED_CONCAT_OF_ADJOINING_SELS_ON_LHS) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REPLACE_NESTED_CONCAT_OF_ADJOINING_SELS_ON_RHS) \
DFG: Add more peephole patterns
2022-09-30 15:19:53 +00:00
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REPLACE_NOT_EQ) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REPLACE_NOT_NEQ) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REPLACE_OR_OF_CONCAT_LHS_ZERO_AND_CONCAT_ZERO_RHS) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REPLACE_OR_OF_CONCAT_ZERO_LHS_AND_CONCAT_RHS_ZERO) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REPLACE_OR_OF_NOT_AND_NEQ) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REPLACE_OR_OF_NOT_AND_NOT) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REPLACE_OR_WITH_ONES) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REPLACE_SEL_FROM_SEL) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REPLACE_SUB_WITH_NOT) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REPLACE_TAUTOLOGICAL_OR) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, REPLACE_XOR_WITH_ONES) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, RIGHT_LEANING_ASSOC) \
Introduce DFG based combinational logic optimizer (#3527) Added a new data-flow graph (DFG) based combinational logic optimizer. The capabilities of this covers a combination of V3Const and V3Gate, but is also more capable of transforming combinational logic into simplified forms and more. This entail adding a new internal representation, `DfgGraph`, and appropriate `astToDfg` and `dfgToAst` conversion functions. The graph represents some of the combinational equations (~continuous assignments) in a module, and for the duration of the DFG passes, it takes over the role of AstModule. A bulk of the Dfg vertices represent expressions. These vertex classes, and the corresponding conversions to/from AST are mostly auto-generated by astgen, together with a DfgVVisitor that can be used for dynamic dispatch based on vertex (operation) types. The resulting combinational logic graph (a `DfgGraph`) is then optimized in various ways. Currently we perform common sub-expression elimination, variable inlining, and some specific peephole optimizations, but there is scope for more optimizations in the future using the same representation. The optimizer is run directly before and after inlining. The pre inline pass can operate on smaller graphs and hence converges faster, but still has a chance of substantially reducing the size of the logic on some designs, making inlining both faster and less memory intensive. The post inline pass can then optimize across the inlined module boundaries. No optimization is performed across a module boundary. For debugging purposes, each peephole optimization can be disabled individually via the -fno-dfg-peepnole-<OPT> option, where <OPT> is one of the optimizations listed in V3DfgPeephole.h, for example -fno-dfg-peephole-remove-not-not. The peephole patterns currently implemented were mostly picked based on the design that inspired this work, and on that design the optimizations yields ~30% single threaded speedup, and ~50% speedup on 4 threads. As you can imagine not having to haul around redundant combinational networks in the rest of the compilation pipeline also helps with memory consumption, and up to 30% peak memory usage of Verilator was observed on the same design. Gains on other arbitrary designs are smaller (and can be improved by analyzing those designs). For example OpenTitan gains between 1-15% speedup depending on build type.
2022-09-23 15:46:22 +00:00
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, SWAP_COND_WITH_NEQ_CONDITION) \
DFG: Add more peephole patterns
2022-09-30 15:19:53 +00:00
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, SWAP_COND_WITH_NOT_CONDITION) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, SWAP_CONST_IN_COMMUTATIVE_BINARY) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, SWAP_NOT_IN_COMMUTATIVE_BINARY) \
_FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION_APPLY(macro, SWAP_VAR_IN_COMMUTATIVE_BINARY)
Introduce DFG based combinational logic optimizer (#3527) Added a new data-flow graph (DFG) based combinational logic optimizer. The capabilities of this covers a combination of V3Const and V3Gate, but is also more capable of transforming combinational logic into simplified forms and more. This entail adding a new internal representation, `DfgGraph`, and appropriate `astToDfg` and `dfgToAst` conversion functions. The graph represents some of the combinational equations (~continuous assignments) in a module, and for the duration of the DFG passes, it takes over the role of AstModule. A bulk of the Dfg vertices represent expressions. These vertex classes, and the corresponding conversions to/from AST are mostly auto-generated by astgen, together with a DfgVVisitor that can be used for dynamic dispatch based on vertex (operation) types. The resulting combinational logic graph (a `DfgGraph`) is then optimized in various ways. Currently we perform common sub-expression elimination, variable inlining, and some specific peephole optimizations, but there is scope for more optimizations in the future using the same representation. The optimizer is run directly before and after inlining. The pre inline pass can operate on smaller graphs and hence converges faster, but still has a chance of substantially reducing the size of the logic on some designs, making inlining both faster and less memory intensive. The post inline pass can then optimize across the inlined module boundaries. No optimization is performed across a module boundary. For debugging purposes, each peephole optimization can be disabled individually via the -fno-dfg-peepnole-<OPT> option, where <OPT> is one of the optimizations listed in V3DfgPeephole.h, for example -fno-dfg-peephole-remove-not-not. The peephole patterns currently implemented were mostly picked based on the design that inspired this work, and on that design the optimizations yields ~30% single threaded speedup, and ~50% speedup on 4 threads. As you can imagine not having to haul around redundant combinational networks in the rest of the compilation pipeline also helps with memory consumption, and up to 30% peak memory usage of Verilator was observed on the same design. Gains on other arbitrary designs are smaller (and can be improved by analyzing those designs). For example OpenTitan gains between 1-15% speedup depending on build type.
2022-09-23 15:46:22 +00:00
// clang-format on
class VDfgPeepholePattern final {
public:
enum en : unsigned {
#define OPTIMIZATION_ID(id, name) id,
FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION(OPTIMIZATION_ID)
#undef OPTIMIZATION_ID
_ENUM_END
};
enum en m_e;
const char* ascii() const {
static const char* const names[] = {
#define OPTIMIZATION_NAME(id, name) name,
FOR_EACH_DFG_PEEPHOLE_OPTIMIZATION(OPTIMIZATION_NAME)
#undef OPTIMIZATION_NAME
"_ENUM_END" //
};
return names[m_e];
}
// cppcheck-suppress noExplicitConstructor
VDfgPeepholePattern(en _e)
: m_e{_e} {}
operator en() const { return m_e; }
};
struct V3DfgPeepholeContext final {
const std::string m_label; // Label to apply to stats
// Enable flags for each optimization
bool m_enabled[VDfgPeepholePattern::_ENUM_END];
// Count of applications for each optimization (for statistics)
VDouble0 m_count[VDfgPeepholePattern::_ENUM_END];
Internals: cppcheck cleanups. No functional change intended.
2022-10-03 03:04:55 +00:00
explicit V3DfgPeepholeContext(const std::string& label);
Introduce DFG based combinational logic optimizer (#3527) Added a new data-flow graph (DFG) based combinational logic optimizer. The capabilities of this covers a combination of V3Const and V3Gate, but is also more capable of transforming combinational logic into simplified forms and more. This entail adding a new internal representation, `DfgGraph`, and appropriate `astToDfg` and `dfgToAst` conversion functions. The graph represents some of the combinational equations (~continuous assignments) in a module, and for the duration of the DFG passes, it takes over the role of AstModule. A bulk of the Dfg vertices represent expressions. These vertex classes, and the corresponding conversions to/from AST are mostly auto-generated by astgen, together with a DfgVVisitor that can be used for dynamic dispatch based on vertex (operation) types. The resulting combinational logic graph (a `DfgGraph`) is then optimized in various ways. Currently we perform common sub-expression elimination, variable inlining, and some specific peephole optimizations, but there is scope for more optimizations in the future using the same representation. The optimizer is run directly before and after inlining. The pre inline pass can operate on smaller graphs and hence converges faster, but still has a chance of substantially reducing the size of the logic on some designs, making inlining both faster and less memory intensive. The post inline pass can then optimize across the inlined module boundaries. No optimization is performed across a module boundary. For debugging purposes, each peephole optimization can be disabled individually via the -fno-dfg-peepnole-<OPT> option, where <OPT> is one of the optimizations listed in V3DfgPeephole.h, for example -fno-dfg-peephole-remove-not-not. The peephole patterns currently implemented were mostly picked based on the design that inspired this work, and on that design the optimizations yields ~30% single threaded speedup, and ~50% speedup on 4 threads. As you can imagine not having to haul around redundant combinational networks in the rest of the compilation pipeline also helps with memory consumption, and up to 30% peak memory usage of Verilator was observed on the same design. Gains on other arbitrary designs are smaller (and can be improved by analyzing those designs). For example OpenTitan gains between 1-15% speedup depending on build type.
2022-09-23 15:46:22 +00:00
~V3DfgPeepholeContext();
};
#endif
Reference in New Issue Copy Permalink