Skip to content

Optimise integer division #645

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Draft
neworderofjamie wants to merge 7 commits into
base: master
Choose a base branch
from
Draft

Optimise integer division #645

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
1b301d2
WIP fast constant divide support
neworderofjamie Oct 14, 2024
aac227a
fixed typo
neworderofjamie Oct 14, 2024
263981a
used division results to calculate modulus
neworderofjamie Oct 14, 2024
24062ad
fixed one issue with fast divide in custom updates
neworderofjamie Dec 3, 2024
33ba02e
Tweaks
neworderofjamie Dec 3, 2024
b8919d4
fixed bug in batched indexing
neworderofjamie Dec 3, 2024
698fdd5
Switched out bespoke code for libdivide
neworderofjamie Dec 3, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 3 additions & 0 deletions include/genn/backends/cuda/backend.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -213,6 +213,9 @@ class BACKEND_EXPORT Backend : public BackendSIMT
//! Get type of population RNG
virtual Type::ResolvedType getPopulationRNGType() const final;

//! Generate a fast uint32 divide
virtual std::string getFastU32Divide(const std::string &numerator, const std::string &divisorVar) const final;

//--------------------------------------------------------------------------
// CodeGenerator::BackendBase virtuals
//--------------------------------------------------------------------------
Expand Down
3 changes: 3 additions & 0 deletions include/genn/genn/code_generator/backendSIMT.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -116,6 +116,9 @@ class GENN_EXPORT BackendSIMT : public BackendBase
//! Get type of population RNG
virtual Type::ResolvedType getPopulationRNGType() const = 0;

//! Generate a fast uint32 divide
virtual std::string getFastU32Divide(const std::string &numerator, const std::string &divisorVar) const = 0;

//------------------------------------------------------------------------
// BackendBase virtuals
//------------------------------------------------------------------------
Expand Down
32 changes: 32 additions & 0 deletions include/genn/genn/code_generator/environment.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,6 +6,12 @@
#include <unordered_map>
#include <variant>

// Lib divide includes
extern "C"
{
#include "libdivide.h"
}

// GeNN includes
#include "gennExport.h"
#include "gennUtils.h"
Expand Down Expand Up @@ -409,6 +415,7 @@ class EnvironmentGroupMergedField : public EnvironmentExternalDynamicBase<Enviro
Runtime::MergedDynamicFieldDestinations&>>;
using GetFieldNonNumericValueFunc = std::function<NonNumericFieldValue(Runtime::Runtime&, const GroupInternal&, size_t)>;
using GetFieldNumericValueFunc = std::function<Type::NumericValue(const GroupInternal&, size_t)>;
using GetFieldUint32ValueFunc = std::function<uint32_t(const GroupInternal&, size_t)>;
using IsDynamicFn = bool (GroupInternal::*)(const std::string&) const;
using IsVarInitHeterogeneousFn = bool (G::*)(const std::string&, const std::string&) const;
using GetParamValuesFn = const std::map<std::string, Type::NumericValue> &(GroupInternal::*)(void) const;
Expand Down Expand Up @@ -500,6 +507,31 @@ class EnvironmentGroupMergedField : public EnvironmentExternalDynamicBase<Enviro
addField(type, name, type, fieldName, getFieldValue, indexSuffix, mergedFieldType, initialisers);
}

void addFastDivideField(const std::string &name, const std::string &fieldName,
GetFieldUint32ValueFunc getFieldValue)
{
// Add main field
addField(GeNN::Type::Uint32.addConst(), name,
GeNN::Type::Uint32, fieldName,
getFieldValue);

// Add "magic" field
addField(GeNN::Type::Uint32.addConst(), name + "_magic",
GeNN::Type::Uint32, fieldName + "Magic",
[getFieldValue](const GroupInternal &g, size_t i)
{
return libdivide_u32_branchfree_gen(getFieldValue(g, i)).magic;
});

// Add "more" field
addField(GeNN::Type::Uint8.addConst(), name + "_more",
GeNN::Type::Uint8, fieldName + "More",
[getFieldValue](const GroupInternal &g, size_t i)
{
return libdivide_u32_branchfree_gen(getFieldValue(g, i)).more;
});
}

void addParams(const Snippet::Base::ParamVec &params, const std::string &fieldSuffix,
GetParamValuesFn getParamValues, IsDynamicFn isDynamic)
{
Expand Down
Loading