Files
TheRock/rocm-libraries/shared/rocroller/lib/source/KernelGraph/Transformations/LoopOverTileNumbers.cpp
T
derek.holloway 7502bf64a4
CodeQL / Analyze (actions) (push) Canceled after 0s
CodeQL / Analyze (python) (push) Canceled after 0s
pre-commit / pre-commit (push) Canceled after 0s
Multi-Arch CI / setup (push) Canceled after 0s
Publish build_manylinux_x86_64 images / publish_build_manylinux_x86_64 (push) Canceled after 0s
Setup (Multi-Arch) / setup (push) Canceled after 0s
Multi-Arch CI / Linux::${{ fromJSON(needs.setup.outputs.linux_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubi10 image / publish_no_rocm_image_ubi10 (push) Canceled after 0s
Multi-Arch CI / Windows::${{ fromJSON(needs.setup.outputs.windows_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Multi-Arch CI / Manifest Diff (push) Canceled after 0s
TheRock Manifest Diff Report / Generate Manifest Diff Report (push) Canceled after 0s
Multi-Arch CI / CI Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_ocl_rt images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_rocgdb images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Unit Tests / Unit Tests :: ubuntu-24.04 (push) Canceled after 0s
Unit Tests / Unit Tests :: windows-2022 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_media (push) Canceled after 0s
Unit Tests / Unit Tests Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_openmpi (push) Canceled after 0s
Flatten subprojects into a single base monorepo
2026-09-02 16:34:47 -07:00

162 lines
6.5 KiB
C++

// Copyright Advanced Micro Devices, Inc., or its affiliates.
// SPDX-License-Identifier: MIT
#include <map>
#include <unordered_set>
#include <vector>
#include <rocRoller/Expression.hpp>
#include <rocRoller/ExpressionTransformations.hpp>
#include <rocRoller/KernelGraph/KernelGraph.hpp>
#include <rocRoller/KernelGraph/Transforms/LoopOverTileNumbers.hpp>
#include <rocRoller/KernelGraph/Utils.hpp>
#include <rocRoller/Utilities/Logging.hpp>
namespace rocRoller
{
namespace KernelGraph
{
using namespace ControlGraph;
using namespace CoordinateGraph;
using GD = rocRoller::Graph::Direction;
std::string LoopOverTileNumbers::name() const
{
return concatenate("LoopOverTileNumbers");
}
LoopOverTileNumbers::LoopOverTileNumbers(std::vector<int> const& dims,
std::vector<uint> const& tileNumberCoordSizes,
uint numIteratedTiles,
std::string const& topLoop,
ContextPtr context)
: m_dimensions(dims)
, m_topLoop(topLoop)
, m_context(context)
{
for(auto s : tileNumberCoordSizes)
m_tileNumberCoordSizes.push_back(Expression::literal(s));
m_numIteratedTileNumbers = Expression::literal(numIteratedTiles);
}
//
// Stage
//
// Look for all leaf MacroTileNumbers with matching
// sub-dimension.
//
// Matches are: tile->dim in m_dimensions.
//
void LoopOverTileNumbers::stage(KernelGraph const& graph)
{
// Find all dangling MacroTileNumber dimensions associated
// with the requested dimensions
for(auto dimension : m_dimensions)
{
auto danglingTileNumberPredicate = [&](int tag) {
auto maybeTileNumber = graph.coordinates.get<MacroTileNumber>(tag);
if(!maybeTileNumber)
return false;
if(maybeTileNumber->dim != dimension)
return false;
if(std::empty(graph.coordinates.getNeighbours<GD::Upstream>(tag))
|| std::empty(graph.coordinates.getNeighbours<GD::Downstream>(tag)))
return true;
return false;
};
m_tileNumberCoords[dimension]
= graph.coordinates.findElements(danglingTileNumberPredicate)
.to<std::unordered_set>();
}
}
//
// Commit
//
void LoopOverTileNumbers::commit(KernelGraph& graph)
{
auto numTotalTiles = Expression::literal(1u);
for(auto d : m_dimensions)
numTotalTiles = numTotalTiles * m_tileNumberCoordSizes[d];
numTotalTiles = simplify(numTotalTiles);
auto [forLoopCoord, forLoopOp]
= rangeFor(graph, m_numIteratedTileNumbers, "TileLoop", DataType::UInt32);
// Create forward/reverse tile-numbers for each dimension
// and attach to all staged tile-number coordinates
std::vector<int> tileNumbersUp, tileNumbersDown;
for(auto d : m_dimensions)
{
auto up = graph.coordinates.addElement(
MacroTileNumber(d, m_tileNumberCoordSizes[d], nullptr));
auto down = graph.coordinates.addElement(
MacroTileNumber(d, m_tileNumberCoordSizes[d], nullptr));
for(auto tileNumTag : m_tileNumberCoords.at(d))
{
if(std::empty(graph.coordinates.getNeighbours<GD::Upstream>(tileNumTag)))
graph.coordinates.addElement(PassThrough(), {down}, {tileNumTag});
if(std::empty(graph.coordinates.getNeighbours<GD::Downstream>(tileNumTag)))
graph.coordinates.addElement(PassThrough(), {tileNumTag}, {up});
}
tileNumbersUp.push_back(up);
tileNumbersDown.push_back(down);
}
// Create foward/reverse transforms from (Workgroup,
// ForLoop) to new tile-numbers
auto workgroups = simplify(numTotalTiles / m_numIteratedTileNumbers);
auto initUp = graph.coordinates.addElement(MacroTileNumber(0, workgroups, nullptr));
auto flatUp = graph.coordinates.addElement(Linear(numTotalTiles, nullptr));
graph.coordinates.addElement(Flatten(), {forLoopCoord, initUp}, {flatUp});
graph.coordinates.addElement(Tile(), std::vector<int>{flatUp}, tileNumbersDown);
auto initDown = graph.coordinates.addElement(MacroTileNumber(0, workgroups, nullptr));
auto flatDown = graph.coordinates.addElement(Linear(numTotalTiles, nullptr));
graph.coordinates.addElement(Tile(), {flatDown}, {forLoopCoord, initDown});
graph.coordinates.addElement(Flatten(), tileNumbersUp, std::vector<int>{flatDown});
// Insert ForLoop into graph
insertWithBody(graph, m_topLoopOp, forLoopOp);
}
//
// Apply
//
KernelGraph LoopOverTileNumbers::apply(KernelGraph const& original)
{
TIMER(t, "KernelGraph::LoopOverTileNumbers");
// Make sure we can find the top-for-loop location
auto findTopLoopPredicate = [&](int tag) -> bool {
auto maybeForLoop = original.control.get<ForLoopOp>(tag);
if(!maybeForLoop)
return false;
if(maybeForLoop->loopName == m_topLoop)
return true;
return false;
};
auto maybeTopLoopOp = original.control.findNodes(
*original.control.roots().begin(), findTopLoopPredicate, GD::Downstream);
if(maybeTopLoopOp.empty())
{
rocRoller::Log::getLogger()->warn(
"Unable to find ForLoop '{}' during LoopOverTileNumbers pass. "
"LoopOverTileNumbers transform skipped.",
m_topLoop);
return original;
}
m_topLoopOp = *maybeTopLoopOp.take(1).only();
auto graph = original;
stage(graph);
commit(graph);
return graph;
}
}
}