This commit is contained in:
@@ -40,7 +40,7 @@ cloneMappedHelperOperands(Operation* op, IRMapping& mapping, IRRewriter& rewrite
|
||||
continue;
|
||||
|
||||
if (auto constantOp = dyn_cast<arith::ConstantOp>(definingOp)) {
|
||||
mapping.map(operand, getOrCreateHostConstantLike(constantOp, constantFolder));
|
||||
mapping.map(operand, getOrCreateHostConstantLike(constantFolder, constantOp));
|
||||
continue;
|
||||
}
|
||||
|
||||
@@ -218,7 +218,7 @@ LogicalResult raptor::SpatialToPimPass::lowerComputeOp(spatial::SpatCompute comp
|
||||
continue;
|
||||
|
||||
if (auto constantOp = input.getDefiningOp<arith::ConstantOp>()) {
|
||||
blockArg->replaceAllUsesWith(getOrCreateHostConstantLike(constantOp, constantFolder));
|
||||
blockArg->replaceAllUsesWith(getOrCreateHostConstantLike(constantFolder, constantOp));
|
||||
continue;
|
||||
}
|
||||
|
||||
@@ -230,8 +230,8 @@ LogicalResult raptor::SpatialToPimPass::lowerComputeOp(spatial::SpatCompute comp
|
||||
PimMemCopyHostToDevOp::create(rewriter,
|
||||
loc,
|
||||
outputBuffer.getType(),
|
||||
getOrCreateHostIndexConstant(outputBuffer.getOperation(), 0, constantFolder),
|
||||
getOrCreateHostIndexConstant(outputBuffer.getOperation(), 0, constantFolder),
|
||||
getOrCreateHostIndexConstant(constantFolder, outputBuffer.getOperation(), 0),
|
||||
getOrCreateHostIndexConstant(constantFolder, outputBuffer.getOperation(), 0),
|
||||
outputBuffer,
|
||||
input,
|
||||
getTensorSizeInBytesAttr(rewriter, input))
|
||||
|
||||
@@ -326,7 +326,7 @@ cloneMappedHelperOperands(Operation* op, IRMapping& mapping, IRRewriter& rewrite
|
||||
continue;
|
||||
|
||||
if (auto constantOp = dyn_cast<arith::ConstantOp>(definingOp)) {
|
||||
mapping.map(operand, getOrCreateHostConstantLike(constantOp, constantFolder));
|
||||
mapping.map(operand, getOrCreateHostConstantLike(constantFolder, constantOp));
|
||||
continue;
|
||||
}
|
||||
|
||||
@@ -370,8 +370,8 @@ static Value emitHostCopy(IRRewriter& rewriter,
|
||||
OperationFolder& constantFolder) {
|
||||
Operation* anchorOp = sourceValue.getDefiningOp() ? sourceValue.getDefiningOp() : outputTensor.getDefiningOp();
|
||||
assert(anchorOp && "expected a concrete op anchor for return-path host copy constants");
|
||||
Value hostTargetOffsetValue = getOrCreateHostIndexConstant(anchorOp, hostTargetOffset, constantFolder);
|
||||
Value deviceSourceOffsetValue = getOrCreateHostIndexConstant(anchorOp, deviceSourceOffset, constantFolder);
|
||||
Value hostTargetOffsetValue = getOrCreateHostIndexConstant(constantFolder, anchorOp, hostTargetOffset);
|
||||
Value deviceSourceOffsetValue = getOrCreateHostIndexConstant(constantFolder, anchorOp, deviceSourceOffset);
|
||||
return PimMemCopyDevToHostOp::create(rewriter,
|
||||
loc,
|
||||
outputTensor.getType(),
|
||||
|
||||
@@ -81,7 +81,7 @@ static Value createZeroedDeviceHVector(IRRewriter& rewriter,
|
||||
auto outputBuffer = createEmptyTensorFromShaped(rewriter, loc, tensorType);
|
||||
auto zeroGlobal = getOrCreateZeroGlobal(rewriter, loc, tensorType);
|
||||
auto zeroValue = memref::GetGlobalOp::create(rewriter, loc, zeroGlobal.getType(), zeroGlobal.getName());
|
||||
auto zeroIndex = getOrCreateHostIndexConstant(outputBuffer.getOperation(), 0, constantFolder);
|
||||
auto zeroIndex = getOrCreateHostIndexConstant(constantFolder, outputBuffer.getOperation(), 0);
|
||||
auto sizeAttr = rewriter.getI32IntegerAttr(static_cast<int32_t>(getShapedTypeSizeInBytes(tensorType)));
|
||||
|
||||
if (outputBuffer->getParentOfType<PimCoreBatchOp>())
|
||||
@@ -333,9 +333,9 @@ LogicalResult raptor::SpatialToPimPass::allocateAndInitializeCoreLocalVariables(
|
||||
rewriter,
|
||||
loc,
|
||||
tensorType,
|
||||
getOrCreateHostIndexConstant(deviceTensor.getOperation(), 0, constantFolder),
|
||||
getOrCreateHostIndexConstant(
|
||||
deviceTensor.getOperation(), static_cast<int64_t>(elementsOffset * elementByteSize), constantFolder),
|
||||
getOrCreateHostIndexConstant(constantFolder, deviceTensor.getOperation(), 0),
|
||||
getOrCreateHostIndexConstant(constantFolder,
|
||||
deviceTensor.getOperation(), static_cast<int64_t>(elementsOffset * elementByteSize) ),
|
||||
deviceTensor,
|
||||
inputTensor,
|
||||
rewriter.getI32IntegerAttr(static_cast<int32_t>(tensorType.getNumElements() * elementByteSize)));
|
||||
|
||||
Reference in New Issue
Block a user