614 lines
26 KiB
C++
614 lines
26 KiB
C++
//===- GPUTransformOps.cpp - Implementation of GPU transform ops ----------===//
|
|
//
|
|
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
|
|
// See https://llvm.org/LICENSE.txt for license information.
|
|
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
|
|
//
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
#include "mlir/Dialect/GPU/TransformOps/GPUTransformOps.h"
|
|
|
|
#include "mlir/Dialect/Arith/IR/Arith.h"
|
|
#include "mlir/Dialect/GPU/IR/GPUDialect.h"
|
|
#include "mlir/Dialect/GPU/TransformOps/GPUTransformOps.h"
|
|
#include "mlir/Dialect/PDL/IR/PDL.h"
|
|
#include "mlir/Dialect/SCF/IR/DeviceMappingInterface.h"
|
|
#include "mlir/Dialect/SCF/IR/SCF.h"
|
|
#include "mlir/Dialect/Transform/IR/TransformDialect.h"
|
|
#include "mlir/Dialect/Transform/IR/TransformInterfaces.h"
|
|
#include "mlir/Dialect/Transform/IR/TransformUtils.h"
|
|
#include "mlir/IR/IRMapping.h"
|
|
|
|
using namespace mlir;
|
|
using namespace mlir::gpu;
|
|
using namespace mlir::transform;
|
|
|
|
/// Check if given mapping attributes are one of the desired attributes
|
|
static DiagnosedSilenceableFailure
|
|
checkAttributeType(ArrayRef<DeviceMappingAttrInterface> threadMappingAttributes,
|
|
const std::optional<ArrayAttr> &foreachMapping,
|
|
std::optional<TransformOpInterface> transformOp) {
|
|
if (!foreachMapping.has_value())
|
|
return transformOp->emitSilenceableError() << "mapping must be present";
|
|
|
|
DenseSet<Attribute> seen;
|
|
for (Attribute map : foreachMapping->getValue()) {
|
|
if (!llvm::is_contained(threadMappingAttributes, map)) {
|
|
return transformOp->emitDefiniteFailure()
|
|
<< "mapping must be one of " << threadMappingAttributes;
|
|
}
|
|
if (llvm::is_contained(seen, map)) {
|
|
return transformOp->emitDefiniteFailure()
|
|
<< map
|
|
<< " is duplicated, cannot map different "
|
|
"loops to the same processor";
|
|
}
|
|
seen.insert(map);
|
|
}
|
|
|
|
return DiagnosedSilenceableFailure::success();
|
|
}
|
|
|
|
/// Determines if the size of the kernel configuration is supported by the GPU
|
|
/// architecture being used. It presently makes use of CUDA limitations, however
|
|
/// that aspect may be enhanced for other GPUs.
|
|
static DiagnosedSilenceableFailure checkGpuLimits(
|
|
TransformOpInterface transformOp, std::optional<int64_t> gridDimX,
|
|
std::optional<int64_t> gridDimY, std::optional<int64_t> gridDimZ,
|
|
std::optional<int64_t> blockDimX, std::optional<int64_t> blockDimY,
|
|
std::optional<int64_t> blockDimZ) {
|
|
|
|
static constexpr int maxTotalBlockdim = 1024;
|
|
static constexpr int maxBlockdimx = 1024;
|
|
static constexpr int maxBlockdimy = 1024;
|
|
static constexpr int maxBlockdimz = 64;
|
|
static constexpr int maxTotalGriddim = 2147483647;
|
|
static constexpr int maxGriddimx = 2147483647;
|
|
static constexpr int maxGriddimy = 65535;
|
|
static constexpr int maxGriddimz = 65535;
|
|
|
|
if ((blockDimX.value_or(1) * blockDimY.value_or(1) * blockDimZ.value_or(1)) >
|
|
maxTotalBlockdim ||
|
|
(gridDimX.value_or(1) * gridDimY.value_or(1) * gridDimZ.value_or(1)) >
|
|
maxTotalGriddim ||
|
|
blockDimX.value_or(1) > maxBlockdimx ||
|
|
blockDimY.value_or(1) > maxBlockdimy ||
|
|
blockDimZ.value_or(1) > maxBlockdimz ||
|
|
gridDimY.value_or(1) > maxGriddimy ||
|
|
gridDimZ.value_or(1) > maxGriddimz ||
|
|
gridDimX.value_or(1) > maxGriddimx) {
|
|
return transformOp.emitSilenceableError()
|
|
<< "Trying to launch a GPU kernel with gridDim = ("
|
|
<< gridDimX.value_or(1) << ", " << gridDimY.value_or(1) << ", "
|
|
<< gridDimZ.value_or(1) << ") blockDim = (" << blockDimX.value_or(1)
|
|
<< ", " << blockDimY.value_or(1) << ", " << blockDimZ.value_or(1)
|
|
<< "). It is larger than the limits.";
|
|
}
|
|
return DiagnosedSilenceableFailure::success();
|
|
}
|
|
|
|
/// Creates an empty-body gpu::LaunchOp using the provided kernel settings and
|
|
/// put a terminator within.
|
|
static DiagnosedSilenceableFailure
|
|
createGpuLaunch(RewriterBase &rewriter, Location loc,
|
|
TransformOpInterface transformOp, LaunchOp &launchOp,
|
|
std::optional<int64_t> gridDimX = std::nullopt,
|
|
std::optional<int64_t> gridDimY = std::nullopt,
|
|
std::optional<int64_t> gridDimZ = std::nullopt,
|
|
std::optional<int64_t> blockDimX = std::nullopt,
|
|
std::optional<int64_t> blockDimY = std::nullopt,
|
|
std::optional<int64_t> blockDimZ = std::nullopt) {
|
|
DiagnosedSilenceableFailure diag =
|
|
checkGpuLimits(transformOp, gridDimX, gridDimY, gridDimZ, blockDimX,
|
|
blockDimY, blockDimZ);
|
|
if (!diag.succeeded())
|
|
return diag;
|
|
|
|
auto createConst = [&](int dim) {
|
|
return rewriter.create<arith::ConstantIndexOp>(loc, dim);
|
|
};
|
|
OpBuilder::InsertionGuard guard(rewriter);
|
|
Value one = createConst(1);
|
|
Value gridSizeX = gridDimX.has_value() ? createConst(gridDimX.value()) : one;
|
|
Value gridSizeY = gridDimY.has_value() ? createConst(gridDimY.value()) : one;
|
|
Value gridSizeZ = gridDimZ.has_value() ? createConst(gridDimZ.value()) : one;
|
|
Value blkSizeX = blockDimX.has_value() ? createConst(blockDimX.value()) : one;
|
|
Value blkSizeY = blockDimY.has_value() ? createConst(blockDimY.value()) : one;
|
|
Value blkSizeZ = blockDimZ.has_value() ? createConst(blockDimZ.value()) : one;
|
|
launchOp = rewriter.create<LaunchOp>(loc, gridSizeX, gridSizeY, gridSizeZ,
|
|
blkSizeX, blkSizeY, blkSizeZ);
|
|
rewriter.setInsertionPointToEnd(&launchOp.getBody().front());
|
|
rewriter.create<TerminatorOp>(loc);
|
|
return DiagnosedSilenceableFailure::success();
|
|
}
|
|
|
|
/// Alter kernel configuration of the given kernel.
|
|
static DiagnosedSilenceableFailure
|
|
alterGpuLaunch(TrivialPatternRewriter &rewriter, LaunchOp gpuLaunch,
|
|
TransformOpInterface transformOp,
|
|
std::optional<int64_t> gridDimX = std::nullopt,
|
|
std::optional<int64_t> gridDimY = std::nullopt,
|
|
std::optional<int64_t> gridDimZ = std::nullopt,
|
|
std::optional<int64_t> blockDimX = std::nullopt,
|
|
std::optional<int64_t> blockDimY = std::nullopt,
|
|
std::optional<int64_t> blockDimZ = std::nullopt) {
|
|
DiagnosedSilenceableFailure diag =
|
|
checkGpuLimits(transformOp, gridDimX, gridDimY, gridDimZ, blockDimX,
|
|
blockDimY, blockDimZ);
|
|
if (!diag.succeeded())
|
|
return diag;
|
|
|
|
KernelDim3 currentBlockdim = gpuLaunch.getBlockSizeOperandValues();
|
|
OpBuilder::InsertionGuard guard(rewriter);
|
|
rewriter.setInsertionPointAfterValue(currentBlockdim.x);
|
|
auto createConstValue = [&](int dim) {
|
|
return rewriter.create<arith::ConstantIndexOp>(currentBlockdim.x.getLoc(),
|
|
dim);
|
|
};
|
|
|
|
if (gridDimX.has_value())
|
|
gpuLaunch.getGridSizeXMutable().assign(createConstValue(gridDimX.value()));
|
|
if (gridDimY.has_value())
|
|
gpuLaunch.getGridSizeYMutable().assign(createConstValue(gridDimY.value()));
|
|
if (gridDimZ.has_value())
|
|
gpuLaunch.getGridSizeZMutable().assign(createConstValue(gridDimZ.value()));
|
|
if (blockDimX.has_value())
|
|
gpuLaunch.getBlockSizeXMutable().assign(
|
|
createConstValue(blockDimX.value()));
|
|
if (blockDimY.has_value())
|
|
gpuLaunch.getBlockSizeYMutable().assign(
|
|
createConstValue(blockDimY.value()));
|
|
if (blockDimZ.has_value())
|
|
gpuLaunch.getBlockSizeZMutable().assign(
|
|
createConstValue(blockDimZ.value()));
|
|
return DiagnosedSilenceableFailure::success();
|
|
}
|
|
|
|
//===----------------------------------------------------------------------===//
|
|
// MapForeachToBlocks
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
DiagnosedSilenceableFailure mlir::transform::gpu::mapForeachToBlocksImpl(
|
|
RewriterBase &rewriter, scf::ForallOp forallOp,
|
|
function_ref<void(RewriterBase &, scf::ForallOp, SmallVectorImpl<Value> &)>
|
|
blockIdGenerator,
|
|
SmallVectorImpl<int64_t> &gridDims, TransformOpInterface transformOp,
|
|
const ArrayRef<DeviceMappingAttrInterface> &mappingAttributes) {
|
|
// Step 0. Target-specific verifications. There is no good place to anchor
|
|
// those right now: the ForallOp is target-independent and the
|
|
// transform op does not apply to individual ForallOp.
|
|
Location loc = forallOp->getLoc();
|
|
|
|
if (!forallOp.isNormalized())
|
|
return transformOp.emitSilenceableError()
|
|
<< "unsupported non-normalized loops";
|
|
if (forallOp.getNumResults() > 0)
|
|
return transformOp.emitSilenceableError()
|
|
<< "only bufferized scf.forall lowers to "
|
|
"gpu.block_id";
|
|
if (forallOp.getRank() > 3)
|
|
return transformOp.emitSilenceableError()
|
|
<< "scf.forall with rank > 3 does not lower to "
|
|
"gpu.block_id";
|
|
if (llvm::any_of(forallOp.getMixedUpperBound(), [](OpFoldResult ofr) {
|
|
return !getConstantIntValue(ofr).has_value();
|
|
})) {
|
|
return transformOp.emitSilenceableError()
|
|
<< "unsupported dynamic griddim size";
|
|
}
|
|
SmallVector<Attribute> blockMapping =
|
|
llvm::to_vector(forallOp.getMapping()->getValue());
|
|
|
|
// Step 1. Complete the blockMapping to a full mapping (with 1s) if necessary.
|
|
SmallVector<Value> numBlocks = forallOp.getUpperBound(rewriter);
|
|
// Ensure we have 3 block sizes, one for each id.
|
|
Value one;
|
|
for (auto attr : mappingAttributes) {
|
|
if (std::find(blockMapping.begin(), blockMapping.end(), attr) ==
|
|
blockMapping.end()) {
|
|
blockMapping.push_back(attr);
|
|
one = one ? one : rewriter.create<arith::ConstantIndexOp>(loc, 1);
|
|
numBlocks.push_back(one);
|
|
}
|
|
}
|
|
|
|
// Step 2. sort the values by the corresponding DeviceMappingAttrInterface.
|
|
auto comparator = [&](DeviceMappingAttrInterface a,
|
|
DeviceMappingAttrInterface b) -> bool {
|
|
return a.getMappingId() < b.getMappingId();
|
|
};
|
|
SmallVector<Value> gridDimValues =
|
|
scf::ForallOp::getValuesSortedByKey(blockMapping, numBlocks, comparator);
|
|
for (Value v : gridDimValues)
|
|
gridDims.push_back(v.getDefiningOp<arith::ConstantIndexOp>().value());
|
|
|
|
// Step 3. Generate the blockIds using the provided generator and map the
|
|
// induction variables to the newly created ops.
|
|
SmallVector<Value> blockOps;
|
|
blockIdGenerator(rewriter, forallOp, blockOps);
|
|
IRMapping bvm;
|
|
for (auto [blockIdx, blockDim] :
|
|
llvm::zip(forallOp.getInductionVars(), blockMapping)) {
|
|
bvm.map(blockIdx,
|
|
blockOps[static_cast<int64_t>(
|
|
blockDim.cast<DeviceMappingAttrInterface>().getMappingId())]);
|
|
}
|
|
|
|
// Step 4. Move the body of forallOp.
|
|
// Erase the terminator first, it will not be used since we are on buffers.
|
|
rewriter.eraseOp(forallOp.getTerminator());
|
|
Block *targetBlock = forallOp->getBlock();
|
|
Block::iterator insertionPoint = Block::iterator(forallOp);
|
|
Block &sourceBlock = forallOp.getRegion().front();
|
|
targetBlock->getOperations().splice(insertionPoint,
|
|
sourceBlock.getOperations());
|
|
|
|
// Step 5. RAUW thread indices to thread ops.
|
|
for (Value loopIndex : forallOp.getInductionVars()) {
|
|
Value blockIdx = bvm.lookup(loopIndex);
|
|
rewriter.replaceAllUsesWith(loopIndex, blockIdx);
|
|
}
|
|
|
|
// Step 6. Erase old op.
|
|
rewriter.eraseOp(forallOp);
|
|
|
|
return DiagnosedSilenceableFailure::success();
|
|
}
|
|
|
|
DiagnosedSilenceableFailure
|
|
mlir::transform::gpu::findTopLevelForallOp(Operation *target,
|
|
scf::ForallOp &topLevelForallOp,
|
|
TransformOpInterface transformOp) {
|
|
auto walkResult = target->walk([&](scf::ForallOp forallOp) {
|
|
if (forallOp->getParentOfType<scf::ForallOp>())
|
|
return WalkResult::advance();
|
|
if (topLevelForallOp)
|
|
// TODO: Handle multiple foreach if there is no dependences between them
|
|
return WalkResult::interrupt();
|
|
topLevelForallOp = forallOp;
|
|
return WalkResult::advance();
|
|
});
|
|
|
|
if (walkResult.wasInterrupted())
|
|
return transformOp.emitSilenceableError()
|
|
<< "could not find a unique topLevel scf.forall";
|
|
return DiagnosedSilenceableFailure::success();
|
|
}
|
|
|
|
/// This is a helper that is only used in
|
|
/// rewriteTopLevelForallToGpuBlocks. It generates GPU dialects
|
|
/// block_id.
|
|
static void generateGpuBlockIds(RewriterBase &rewriter, scf::ForallOp foreachOp,
|
|
SmallVectorImpl<Value> &blockOps) {
|
|
Location loc = foreachOp->getLoc();
|
|
OpBuilder::InsertionGuard guard(rewriter);
|
|
rewriter.setInsertionPoint(foreachOp);
|
|
IndexType indexType = rewriter.getIndexType();
|
|
blockOps = SmallVector<Value>{
|
|
rewriter.create<BlockIdOp>(loc, indexType, Dimension::x),
|
|
rewriter.create<BlockIdOp>(loc, indexType, Dimension::y),
|
|
rewriter.create<BlockIdOp>(loc, indexType, Dimension::z)};
|
|
}
|
|
|
|
DiagnosedSilenceableFailure
|
|
transform::MapForeachToBlocks::applyToOne(Operation *target,
|
|
ApplyToEachResultList &results,
|
|
transform::TransformState &state) {
|
|
LaunchOp gpuLaunch = dyn_cast<LaunchOp>(target);
|
|
TrivialPatternRewriter rewriter(getContext());
|
|
auto transformOp = cast<TransformOpInterface>(getOperation());
|
|
|
|
if (!getGenerateGpuLaunch() && !gpuLaunch) {
|
|
DiagnosedSilenceableFailure diag =
|
|
emitSilenceableError()
|
|
<< "Given target is not gpu.launch, set `generate_gpu_launch` "
|
|
"attribute";
|
|
diag.attachNote(target->getLoc()) << "when applied to this payload op";
|
|
return diag;
|
|
}
|
|
|
|
scf::ForallOp topLevelForallOp;
|
|
DiagnosedSilenceableFailure diag = mlir::transform::gpu::findTopLevelForallOp(
|
|
target, topLevelForallOp, transformOp);
|
|
if (!diag.succeeded()) {
|
|
diag.attachNote(target->getLoc()) << "when applied to this payload op";
|
|
return diag;
|
|
}
|
|
|
|
OpBuilder::InsertionGuard guard(rewriter);
|
|
rewriter.setInsertionPoint(topLevelForallOp);
|
|
|
|
// Generate gpu launch here and move the forall inside
|
|
if (getGenerateGpuLaunch()) {
|
|
DiagnosedSilenceableFailure diag =
|
|
createGpuLaunch(rewriter, target->getLoc(), transformOp, gpuLaunch);
|
|
if (!diag.succeeded()) {
|
|
return diag;
|
|
}
|
|
rewriter.setInsertionPointToStart(&gpuLaunch.getBody().front());
|
|
Operation *newForallOp = rewriter.clone(*topLevelForallOp);
|
|
rewriter.eraseOp(topLevelForallOp);
|
|
topLevelForallOp = cast<scf::ForallOp>(newForallOp);
|
|
}
|
|
|
|
SmallVector<int64_t> gridDim = extractFromI64ArrayAttr(getGridDim());
|
|
SmallVector<DeviceMappingAttrInterface> blockMappingAttributes = {
|
|
GPUBlockMappingAttr::get(getContext(), Blocks::DimX),
|
|
GPUBlockMappingAttr::get(getContext(), Blocks::DimY),
|
|
GPUBlockMappingAttr::get(getContext(), Blocks::DimZ)};
|
|
|
|
diag = checkAttributeType(blockMappingAttributes,
|
|
topLevelForallOp.getMapping(), transformOp);
|
|
if (diag.succeeded())
|
|
diag = mlir::transform::gpu::mapForeachToBlocksImpl(
|
|
rewriter, topLevelForallOp, generateGpuBlockIds, gridDim, transformOp,
|
|
blockMappingAttributes);
|
|
if (diag.succeeded()) {
|
|
diag = alterGpuLaunch(rewriter, gpuLaunch,
|
|
cast<TransformOpInterface>(getOperation()),
|
|
gridDim[0], gridDim[1], gridDim[2]);
|
|
}
|
|
|
|
results.push_back(gpuLaunch);
|
|
return diag;
|
|
}
|
|
|
|
//===----------------------------------------------------------------------===//
|
|
// MapNestedForeachToThreads
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
/// Searches `scf.forall` ops nested under `target` and maps each such
|
|
/// op to GPU threads. Mapping is one-to-one and the induction variables of
|
|
/// `scf.forall` are rewritten to gpu.thread_id according to the
|
|
/// thread_dim_mapping attribute. Sibling `scf.forall` are supported in
|
|
/// which case, the union of the number of threads is computed and may result
|
|
/// in predication. Dynamic, `scf.forall` trip counts are currently
|
|
/// not supported. Dynamic block dim sizes are currently not supported.
|
|
static DiagnosedSilenceableFailure rewriteOneForallToGpuThreads(
|
|
RewriterBase &rewriter, scf::ForallOp forallOp,
|
|
const SmallVectorImpl<int64_t> &globalBlockDims,
|
|
const SmallVectorImpl<Value> &threadOps, bool syncAfterDistribute,
|
|
std::optional<TransformOpInterface> transformOp,
|
|
const ArrayRef<DeviceMappingAttrInterface> &threadMappingAttributes) {
|
|
// Step 0. Target-specific verifications. There is no good place to anchor
|
|
// those right now: the ForallOp is target-independent and the
|
|
// transform op does not apply to individual ForallOp.
|
|
auto failureHelper =
|
|
[&](const Twine &message) -> DiagnosedSilenceableFailure {
|
|
if (transformOp.has_value()) {
|
|
return transformOp->emitSilenceableError() << message;
|
|
}
|
|
return emitDefiniteFailure(forallOp, message);
|
|
};
|
|
Location loc = forallOp->getLoc();
|
|
if (!forallOp.isNormalized())
|
|
return failureHelper("unsupported non-normalized loops");
|
|
if (forallOp.getNumResults() > 0)
|
|
return failureHelper("only bufferized scf.forall lowers to gpu.thread_id");
|
|
if (forallOp.getRank() > 3)
|
|
return failureHelper(
|
|
"scf.forall with rank > 3 does not lower to gpu.thread_id");
|
|
if (llvm::any_of(forallOp.getMixedUpperBound(), [](OpFoldResult ofr) {
|
|
return !getConstantIntValue(ofr).has_value();
|
|
})) {
|
|
return failureHelper("unsupported dynamic blockdim size");
|
|
}
|
|
if (!forallOp.getMapping().has_value())
|
|
return failureHelper("mapping must be present");
|
|
SmallVector<Attribute> threadMapping =
|
|
llvm::to_vector(forallOp.getMapping()->getValue());
|
|
|
|
// Step 1. Complete the threadMapping to a full mapping (with 1s) if
|
|
// necessary.
|
|
SmallVector<Value> numThreads = forallOp.getUpperBound(rewriter);
|
|
// Ensure we have 3 block sizes, one for each id.
|
|
Value one;
|
|
for (auto attr : threadMappingAttributes) {
|
|
if (std::find(threadMapping.begin(), threadMapping.end(), attr) ==
|
|
threadMapping.end()) {
|
|
threadMapping.push_back(attr);
|
|
one = one ? one : rewriter.create<arith::ConstantIndexOp>(loc, 1);
|
|
numThreads.push_back(one);
|
|
}
|
|
}
|
|
|
|
// Step 2. sort the values by the corresponding DeviceMappingAttrInterface.
|
|
auto comparator = [&](DeviceMappingAttrInterface a,
|
|
DeviceMappingAttrInterface b) -> bool {
|
|
return a.getMappingId() < b.getMappingId();
|
|
};
|
|
SmallVector<Value> blockDimValues = scf::ForallOp::getValuesSortedByKey(
|
|
threadMapping, numThreads, comparator);
|
|
SmallVector<int64_t> blockDims =
|
|
llvm::to_vector(llvm::map_range(blockDimValues, [](Value v) {
|
|
return v.getDefiningOp<arith::ConstantIndexOp>().value();
|
|
}));
|
|
|
|
// Step 3. Create the gpu.thread ops and map the induction variables to the
|
|
// newly created ops.
|
|
// Replace ids of dimension size 1 by zero to simplify the IR.
|
|
SmallVector<Value> threadOpsUpdated(threadOps.begin(), threadOps.end());
|
|
assert(threadOps.size() == globalBlockDims.size());
|
|
Value zero = rewriter.create<arith::ConstantIndexOp>(loc, 0);
|
|
for (size_t i : llvm::seq(size_t(0), globalBlockDims.size())) {
|
|
if (globalBlockDims[i] == 1)
|
|
threadOpsUpdated[i] = zero;
|
|
}
|
|
IRMapping bvm;
|
|
for (auto [blockIdx, blockDim] :
|
|
llvm::zip(forallOp.getInductionVars(), threadMapping)) {
|
|
bvm.map(blockIdx,
|
|
threadOpsUpdated[blockDim.cast<DeviceMappingAttrInterface>()
|
|
.getMappingId()]);
|
|
}
|
|
|
|
// Step 4. Maybe create conditionals to predicate the region.
|
|
Value predicate;
|
|
for (auto [threadId, blockDim, globalBlockDim] :
|
|
llvm::zip(threadOpsUpdated, blockDims, globalBlockDims)) {
|
|
if (blockDim > globalBlockDim) {
|
|
return failureHelper(
|
|
"The requested GPU threads are fewer than the number of loop trip "
|
|
"counts. Try to tile scf.forall before mapping or set "
|
|
"small blockDim.");
|
|
}
|
|
if (blockDim == globalBlockDim)
|
|
continue;
|
|
Value blockIdx = rewriter.create<arith::ConstantIndexOp>(loc, blockDim);
|
|
Value tmpPredicate = rewriter.create<arith::CmpIOp>(
|
|
loc, arith::CmpIPredicate::ult, threadId, blockIdx);
|
|
predicate =
|
|
predicate ? rewriter.create<arith::AndIOp>(loc, predicate, tmpPredicate)
|
|
: tmpPredicate;
|
|
}
|
|
|
|
// Step 5. Move the body of forallOp.
|
|
// Erase the terminator first, it will not be used.
|
|
rewriter.eraseOp(forallOp.getTerminator());
|
|
Block *targetBlock;
|
|
Block::iterator insertionPoint;
|
|
if (predicate) {
|
|
// Step 5.a. If predicated, move at the beginning.
|
|
auto ifOp =
|
|
rewriter.create<scf::IfOp>(loc, predicate, /*withElseRegion=*/false);
|
|
targetBlock = ifOp.thenBlock();
|
|
insertionPoint = ifOp.thenBlock()->begin();
|
|
} else {
|
|
// Step 5.b. Otherwise, move inline just before forallOp.
|
|
targetBlock = forallOp->getBlock();
|
|
insertionPoint = Block::iterator(forallOp);
|
|
}
|
|
Block &sourceBlock = forallOp.getRegion().front();
|
|
targetBlock->getOperations().splice(insertionPoint,
|
|
sourceBlock.getOperations());
|
|
|
|
// Step 6. RAUW thread indices to thread ops.
|
|
for (Value loopIndex : forallOp.getInductionVars()) {
|
|
Value threadIdx = bvm.lookup(loopIndex);
|
|
rewriter.replaceAllUsesWith(loopIndex, threadIdx);
|
|
}
|
|
|
|
// Step 7. syncthreads.
|
|
// TODO: Need warpsync
|
|
if (syncAfterDistribute)
|
|
rewriter.create<BarrierOp>(loc);
|
|
|
|
// Step 8. Erase old op.
|
|
rewriter.eraseOp(forallOp);
|
|
|
|
return DiagnosedSilenceableFailure::success();
|
|
}
|
|
|
|
DiagnosedSilenceableFailure mlir::transform::gpu::mapNestedForeachToThreadsImpl(
|
|
RewriterBase &rewriter, Operation *target,
|
|
const SmallVectorImpl<int64_t> &blockDim,
|
|
function_ref<void(RewriterBase &, scf::ForallOp, SmallVectorImpl<Value> &)>
|
|
threadIdGenerator,
|
|
bool syncAfterDistribute, std::optional<TransformOpInterface> transformOp,
|
|
const ArrayRef<DeviceMappingAttrInterface> &threadMappingAttributes) {
|
|
DiagnosedSilenceableFailure diag = DiagnosedSilenceableFailure::success();
|
|
target->walk([&](scf::ForallOp forallOp) {
|
|
// Ignore cases with different attributes.
|
|
for (Attribute map : forallOp.getMapping()->getValue()) {
|
|
if (!llvm::is_contained(threadMappingAttributes, map)) {
|
|
return WalkResult::skip();
|
|
}
|
|
}
|
|
diag = checkAttributeType(threadMappingAttributes, forallOp.getMapping(),
|
|
transformOp);
|
|
if (diag.succeeded()) {
|
|
rewriter.setInsertionPoint(forallOp);
|
|
SmallVector<Value> threadOps;
|
|
threadIdGenerator(rewriter, forallOp, threadOps);
|
|
diag = rewriteOneForallToGpuThreads(rewriter, forallOp, blockDim,
|
|
threadOps, syncAfterDistribute,
|
|
transformOp, threadMappingAttributes);
|
|
}
|
|
return diag.succeeded() ? WalkResult::advance() : WalkResult::interrupt();
|
|
});
|
|
return diag;
|
|
}
|
|
|
|
DiagnosedSilenceableFailure transform::MapNestedForeachToThreads::applyToOne(
|
|
Operation *target, ApplyToEachResultList &results, TransformState &state) {
|
|
LaunchOp gpuLaunch = dyn_cast<LaunchOp>(target);
|
|
auto transformOp = cast<TransformOpInterface>(getOperation());
|
|
|
|
if (!gpuLaunch) {
|
|
return emitSilenceableError() << "Given target is not gpu.launch";
|
|
}
|
|
|
|
SmallVector<int64_t> blockDim = extractFromI64ArrayAttr(getBlockDim());
|
|
blockDim.resize(/*size=*/3, /*value=*/1);
|
|
|
|
DiagnosedSilenceableFailure diag =
|
|
checkGpuLimits(transformOp, std::nullopt, std::nullopt, std::nullopt,
|
|
blockDim[0], blockDim[1], blockDim[2]);
|
|
if (diag.isSilenceableFailure()) {
|
|
diag.attachNote(getLoc()) << getBlockDimAttrName() << " is very large";
|
|
return diag;
|
|
}
|
|
|
|
MLIRContext *ctx = getContext();
|
|
TrivialPatternRewriter rewriter(ctx);
|
|
rewriter.setInsertionPoint(target);
|
|
|
|
SmallVector<DeviceMappingAttrInterface> threadMappingAttributes = {
|
|
GPUThreadMappingAttr::get(ctx, Threads::DimX),
|
|
GPUThreadMappingAttr::get(ctx, Threads::DimY),
|
|
GPUThreadMappingAttr::get(ctx, Threads::DimZ)};
|
|
auto threadIdGenerator = [](RewriterBase &rewriter, scf::ForallOp forallOp,
|
|
SmallVectorImpl<Value> &threadIds) {
|
|
IndexType indexType = rewriter.getIndexType();
|
|
threadIds.assign({rewriter.create<ThreadIdOp>(forallOp->getLoc(), indexType,
|
|
Dimension::x),
|
|
rewriter.create<ThreadIdOp>(forallOp->getLoc(), indexType,
|
|
Dimension::y),
|
|
rewriter.create<ThreadIdOp>(forallOp->getLoc(), indexType,
|
|
Dimension::z)});
|
|
};
|
|
diag = mlir::transform::gpu::mapNestedForeachToThreadsImpl(
|
|
rewriter, target, blockDim, threadIdGenerator, getSyncAfterDistribute(),
|
|
transformOp, threadMappingAttributes);
|
|
|
|
if (diag.succeeded()) {
|
|
diag = alterGpuLaunch(rewriter, gpuLaunch, transformOp, std::nullopt,
|
|
std::nullopt, std::nullopt, blockDim[0], blockDim[1],
|
|
blockDim[2]);
|
|
}
|
|
|
|
results.push_back(gpuLaunch.getOperation());
|
|
return diag;
|
|
}
|
|
|
|
//===----------------------------------------------------------------------===//
|
|
// Transform op registration
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
namespace {
|
|
/// Registers new ops and declares PDL as dependent dialect since the
|
|
/// additional ops are using PDL types for operands and results.
|
|
class GPUTransformDialectExtension
|
|
: public transform::TransformDialectExtension<
|
|
GPUTransformDialectExtension> {
|
|
public:
|
|
GPUTransformDialectExtension() {
|
|
declareDependentDialect<pdl::PDLDialect>();
|
|
declareGeneratedDialect<scf::SCFDialect>();
|
|
declareGeneratedDialect<arith::ArithDialect>();
|
|
declareGeneratedDialect<GPUDialect>();
|
|
registerTransformOps<
|
|
#define GET_OP_LIST
|
|
#include "mlir/Dialect/GPU/TransformOps/GPUTransformOps.cpp.inc"
|
|
>();
|
|
}
|
|
};
|
|
} // namespace
|
|
|
|
#define GET_OP_CLASSES
|
|
#include "mlir/Dialect/GPU/TransformOps/GPUTransformOps.cpp.inc"
|
|
|
|
void mlir::gpu::registerTransformDialectExtension(DialectRegistry ®istry) {
|
|
registry.addExtensions<GPUTransformDialectExtension>();
|
|
}
|