2021-12-03 08:52:01 +08:00
|
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
//
|
|
|
|
// 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
|
|
|
|
// Also available under a BSD-style license. See LICENSE.
|
|
|
|
//
|
|
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
|
|
|
|
#include "torch-mlir/Conversion/TorchToTosa/TosaLegalizeUtils.h"
|
|
|
|
#include "mlir/Dialect/Tosa/IR/TosaOps.h" // from @llvm-project
|
|
|
|
#include "mlir/Dialect/Tosa/Utils/QuantUtils.h" // from @llvm-project
|
|
|
|
#include "torch-mlir/Conversion/TorchToTosa/TosaLegalizeCommon.h"
|
|
|
|
|
|
|
|
namespace mlir {
|
|
|
|
namespace tosa {
|
|
|
|
|
|
|
|
// Create a TOSA rescale op from input framework tensor, zero points and
|
|
|
|
// rounding mode
|
|
|
|
Value buildRescale(PatternRewriter &rewriter, Operation *op,
|
|
|
|
ShapedType output_type, Value input_val, double scale,
|
|
|
|
int64_t input_zp, int64_t output_zp, bool double_round,
|
|
|
|
bool scale32) {
|
|
|
|
int32_t multiplier;
|
|
|
|
int32_t shift;
|
|
|
|
|
|
|
|
int32_t scale_width = scale32 ? 32 : 16;
|
|
|
|
|
|
|
|
computeMultiplierAndShift(scale, multiplier, shift, scale_width);
|
|
|
|
|
|
|
|
auto rescale_op = CreateOpAndInfer<tosa::RescaleOp>(
|
|
|
|
rewriter, op->getLoc(), output_type, input_val,
|
|
|
|
rewriter.getI32IntegerAttr(static_cast<int32_t>(input_zp)),
|
|
|
|
rewriter.getI32IntegerAttr(static_cast<int32_t>(output_zp)),
|
2023-01-11 07:07:19 +08:00
|
|
|
rewriter.getDenseI32ArrayAttr({multiplier}),
|
2024-01-15 23:12:12 +08:00
|
|
|
rewriter.getDenseI8ArrayAttr({static_cast<int8_t>(shift)}),
|
|
|
|
rewriter.getBoolAttr(scale32), rewriter.getBoolAttr(double_round),
|
|
|
|
rewriter.getBoolAttr(false));
|
2021-12-03 08:52:01 +08:00
|
|
|
|
|
|
|
return rescale_op.getResult();
|
|
|
|
}
|
|
|
|
|
|
|
|
// Creates TOSA rescale op with int32 output
|
|
|
|
Value buildRescaleToInt32(PatternRewriter &rewriter, Operation *op,
|
|
|
|
Value input_val, double input_scale,
|
|
|
|
int64_t input_zp) {
|
|
|
|
// Output is always int32 type
|
|
|
|
auto input_type = input_val.getType().dyn_cast<mlir::ShapedType>();
|
|
|
|
assert(input_type);
|
|
|
|
auto output_type = input_type.clone(rewriter.getI32Type());
|
|
|
|
|
|
|
|
return buildRescale(rewriter, op, output_type, input_val, input_scale,
|
|
|
|
input_zp, 0, false, true);
|
|
|
|
}
|
|
|
|
|
2022-01-27 11:16:13 +08:00
|
|
|
// Creates a TOSA rescale op based on conv2d parameters.
|
|
|
|
Value buildRescaleOpConvOutput(PatternRewriter &rewriter, Operation *op,
|
|
|
|
Value conv_val, ShapedType input_type,
|
|
|
|
ShapedType weight_type, ShapedType output_type) {
|
|
|
|
auto input_qtype =
|
|
|
|
input_type.getElementType().dyn_cast<mlir::quant::UniformQuantizedType>();
|
|
|
|
auto output_qtype = output_type.getElementType()
|
|
|
|
.dyn_cast<mlir::quant::UniformQuantizedType>();
|
|
|
|
|
|
|
|
double input_scale = input_qtype.getScale();
|
|
|
|
|
|
|
|
int64_t output_zp = output_qtype.getZeroPoint();
|
|
|
|
double output_scale = output_qtype.getScale();
|
|
|
|
|
|
|
|
bool scale32 = isScale32(output_qtype);
|
|
|
|
int32_t scale_width = scale32 ? 32 : 16;
|
|
|
|
|
|
|
|
if (auto weight_per_tensor_qtype =
|
|
|
|
weight_type.getElementType()
|
|
|
|
.dyn_cast<mlir::quant::UniformQuantizedType>()) {
|
|
|
|
// Per-tensor quantization
|
|
|
|
double weight_scale = weight_per_tensor_qtype.getScale();
|
|
|
|
|
|
|
|
int32_t multiplier;
|
|
|
|
int32_t shift;
|
|
|
|
|
|
|
|
double op_tensor_scale = (input_scale * weight_scale) / output_scale;
|
|
|
|
|
|
|
|
computeMultiplierAndShift(op_tensor_scale, multiplier, shift, scale_width);
|
|
|
|
|
|
|
|
auto rescale_op = CreateOpAndInfer<tosa::RescaleOp>(
|
|
|
|
rewriter, op->getLoc(), output_type, conv_val,
|
|
|
|
rewriter.getI32IntegerAttr(0), rewriter.getI32IntegerAttr(output_zp),
|
2023-01-11 07:07:19 +08:00
|
|
|
rewriter.getDenseI32ArrayAttr({multiplier}),
|
2024-01-15 23:12:12 +08:00
|
|
|
rewriter.getDenseI8ArrayAttr({static_cast<int8_t>(shift)}),
|
|
|
|
rewriter.getBoolAttr(scale32), rewriter.getBoolAttr(true),
|
|
|
|
rewriter.getBoolAttr(false));
|
2022-01-27 11:16:13 +08:00
|
|
|
|
|
|
|
return rescale_op.getResult();
|
|
|
|
|
|
|
|
} else if (auto weight_per_channel_qtype =
|
|
|
|
weight_type.getElementType()
|
|
|
|
.dyn_cast<mlir::quant::UniformQuantizedPerAxisType>()) {
|
|
|
|
// Per-channel quantization
|
|
|
|
SmallVector<int32_t> multiplier_arr;
|
2024-01-15 23:12:12 +08:00
|
|
|
SmallVector<int8_t> shift_arr;
|
2022-01-27 11:16:13 +08:00
|
|
|
|
|
|
|
SmallVector<double> weight_scale_arr(
|
|
|
|
weight_per_channel_qtype.getScales().begin(),
|
|
|
|
weight_per_channel_qtype.getScales().end());
|
|
|
|
|
|
|
|
int64_t output_zp = output_qtype.getZeroPoint();
|
|
|
|
double output_scale = output_qtype.getScale();
|
|
|
|
|
|
|
|
for (double weight_scale : weight_scale_arr) {
|
|
|
|
int32_t multiplier;
|
|
|
|
int32_t shift;
|
|
|
|
|
|
|
|
double op_channel_scale = (input_scale * weight_scale) / output_scale;
|
|
|
|
|
|
|
|
computeMultiplierAndShift(op_channel_scale, multiplier, shift,
|
|
|
|
scale_width);
|
|
|
|
|
|
|
|
multiplier_arr.push_back(multiplier);
|
2024-01-15 23:12:12 +08:00
|
|
|
shift_arr.push_back(static_cast<int8_t>(shift));
|
2022-01-27 11:16:13 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
auto rescale_op = CreateOpAndInfer<tosa::RescaleOp>(
|
|
|
|
rewriter, op->getLoc(), output_type, conv_val,
|
|
|
|
rewriter.getI32IntegerAttr(0), rewriter.getI32IntegerAttr(output_zp),
|
2023-01-11 07:07:19 +08:00
|
|
|
rewriter.getDenseI32ArrayAttr(multiplier_arr),
|
2024-01-15 23:12:12 +08:00
|
|
|
rewriter.getDenseI8ArrayAttr(shift_arr), rewriter.getBoolAttr(scale32),
|
2022-01-27 11:16:13 +08:00
|
|
|
rewriter.getBoolAttr(true), rewriter.getBoolAttr(true));
|
|
|
|
|
|
|
|
return rescale_op.getResult();
|
|
|
|
|
|
|
|
} else {
|
|
|
|
op->emitOpError("buildConvRescaleOp: unknown weight quantized type");
|
|
|
|
return nullptr;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
// Check if scale32 mode is used for given output_element_type
|
|
|
|
bool isScale32(mlir::quant::UniformQuantizedType output_element_type) {
|
|
|
|
return (output_element_type.getStorageTypeIntegralWidth() == 8);
|
|
|
|
}
|
|
|
|
|
2021-12-03 08:52:01 +08:00
|
|
|
// Create a 32-bit float constant operator from a float
|
|
|
|
Value getTosaConstTensorSingleF32(PatternRewriter &rewriter, Operation *op,
|
|
|
|
float val) {
|
|
|
|
auto const_type = RankedTensorType::get({}, rewriter.getF32Type());
|
|
|
|
auto const_attr = DenseElementsAttr::get(const_type, val);
|
|
|
|
|
|
|
|
auto const_op =
|
|
|
|
rewriter.create<tosa::ConstOp>(op->getLoc(), const_type, const_attr);
|
|
|
|
return const_op.getResult();
|
|
|
|
}
|
|
|
|
|
2023-05-19 08:12:18 +08:00
|
|
|
// Create a zero constant tensor of the desired type and shape.
|
|
|
|
std::optional<Value> getZerosLikeTensor(PatternRewriter &rewriter,
|
|
|
|
Operation *op, Type type) {
|
|
|
|
RankedTensorType resultType = type.dyn_cast<RankedTensorType>();
|
|
|
|
|
|
|
|
if (!resultType) {
|
|
|
|
(void)rewriter.notifyMatchFailure(op, "not ranked tensor type");
|
|
|
|
return std::nullopt;
|
|
|
|
}
|
|
|
|
|
|
|
|
auto resultShape = resultType.getShape();
|
|
|
|
ShapedType zeroType =
|
|
|
|
RankedTensorType::get(resultShape, resultType.getElementType());
|
|
|
|
Attribute zeroAttr = rewriter.getZeroAttr(zeroType);
|
|
|
|
|
|
|
|
return CreateOpAndInfer<tosa::ConstOp>(rewriter, op->getLoc(), zeroType,
|
|
|
|
zeroAttr.cast<ElementsAttr>())
|
|
|
|
.getResult();
|
|
|
|
}
|
|
|
|
|
2022-01-15 05:57:27 +08:00
|
|
|
// Templated function to create a constant op for given type and shape.
|
|
|
|
// T: storage C type.
|
|
|
|
// Default template creates a constant tensor in T.
|
|
|
|
template <typename T>
|
2022-12-20 18:17:27 +08:00
|
|
|
std::optional<Value> getConstTensor(PatternRewriter &rewriter, Operation *op,
|
2024-01-30 01:59:33 +08:00
|
|
|
ArrayRef<T> vec, ArrayRef<int64_t> shape,
|
|
|
|
std::optional<Type> dtype) {
|
2022-01-15 05:57:27 +08:00
|
|
|
uint64_t num_total_elements = 1;
|
|
|
|
for (int64_t a : shape) {
|
|
|
|
num_total_elements *= a;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (vec.size() != num_total_elements) {
|
|
|
|
op->emitOpError("getConstTensor(): number of elements mismatch.");
|
2022-12-14 16:06:39 +08:00
|
|
|
return std::nullopt;
|
2022-01-15 05:57:27 +08:00
|
|
|
}
|
|
|
|
|
2023-07-17 15:49:04 +08:00
|
|
|
auto width = sizeof(T) * 8;
|
2024-01-30 01:59:33 +08:00
|
|
|
if constexpr (std::is_same_v<T, bool>)
|
2023-07-17 15:49:04 +08:00
|
|
|
width = 1;
|
|
|
|
|
2022-01-15 05:57:27 +08:00
|
|
|
auto const_type =
|
2023-07-17 15:49:04 +08:00
|
|
|
RankedTensorType::get(shape, rewriter.getIntegerType(width));
|
2022-01-15 05:57:27 +08:00
|
|
|
auto const_attr = DenseElementsAttr::get(const_type, vec);
|
|
|
|
|
|
|
|
auto const_op =
|
|
|
|
rewriter.create<tosa::ConstOp>(op->getLoc(), const_type, const_attr);
|
2023-06-16 15:51:24 +08:00
|
|
|
|
|
|
|
if (dtype) {
|
2024-01-30 01:59:33 +08:00
|
|
|
return rewriter.createOrFold<tosa::CastOp>(
|
2023-06-16 15:51:24 +08:00
|
|
|
op->getLoc(), RankedTensorType::get(shape, *dtype), const_op);
|
|
|
|
}
|
2022-01-15 05:57:27 +08:00
|
|
|
return const_op.getResult();
|
|
|
|
}
|
|
|
|
|
|
|
|
// Template specialization for APInt
|
|
|
|
template <>
|
2022-12-20 18:17:27 +08:00
|
|
|
std::optional<Value> getConstTensor<APInt>(PatternRewriter &rewriter,
|
|
|
|
Operation *op, ArrayRef<APInt> vec,
|
2024-01-30 01:59:33 +08:00
|
|
|
ArrayRef<int64_t> shape,
|
|
|
|
std::optional<Type> dtype) {
|
2022-01-15 05:57:27 +08:00
|
|
|
uint64_t num_total_elements = 1;
|
|
|
|
for (int64_t a : shape) {
|
|
|
|
num_total_elements *= a;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (vec.size() != num_total_elements) {
|
|
|
|
op->emitOpError("getConstTensor(): number of elements mismatch.");
|
2022-12-14 16:06:39 +08:00
|
|
|
return std::nullopt;
|
2022-01-15 05:57:27 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
auto const_type = RankedTensorType::get(
|
|
|
|
shape, rewriter.getIntegerType(vec[0].getBitWidth()));
|
|
|
|
auto const_attr = DenseElementsAttr::get(const_type, vec);
|
|
|
|
|
|
|
|
auto const_op =
|
|
|
|
rewriter.create<tosa::ConstOp>(op->getLoc(), const_type, const_attr);
|
2023-06-16 15:51:24 +08:00
|
|
|
|
|
|
|
if (dtype) {
|
2024-01-30 01:59:33 +08:00
|
|
|
return rewriter.createOrFold<tosa::CastOp>(
|
2023-06-16 15:51:24 +08:00
|
|
|
op->getLoc(), RankedTensorType::get(shape, *dtype), const_op);
|
|
|
|
}
|
2022-01-15 05:57:27 +08:00
|
|
|
return const_op.getResult();
|
|
|
|
}
|
|
|
|
|
|
|
|
// Template specialization for float
|
|
|
|
template <>
|
2022-12-20 18:17:27 +08:00
|
|
|
std::optional<Value> getConstTensor<float>(PatternRewriter &rewriter,
|
|
|
|
Operation *op, ArrayRef<float> vec,
|
2024-01-30 01:59:33 +08:00
|
|
|
ArrayRef<int64_t> shape,
|
|
|
|
std::optional<Type> dtype) {
|
2022-01-15 05:57:27 +08:00
|
|
|
uint64_t num_total_elements = 1;
|
|
|
|
for (int64_t a : shape) {
|
|
|
|
num_total_elements *= a;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (vec.size() != num_total_elements) {
|
|
|
|
op->emitOpError("getConstTensor(): number of elements mismatch.");
|
2022-12-14 16:06:39 +08:00
|
|
|
return std::nullopt;
|
2022-01-15 05:57:27 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
auto const_type = RankedTensorType::get(shape, rewriter.getF32Type());
|
|
|
|
auto const_attr = DenseElementsAttr::get(const_type, vec);
|
|
|
|
|
|
|
|
auto const_op =
|
|
|
|
rewriter.create<tosa::ConstOp>(op->getLoc(), const_type, const_attr);
|
2023-06-16 15:51:24 +08:00
|
|
|
|
|
|
|
if (dtype) {
|
2024-01-30 01:59:33 +08:00
|
|
|
return rewriter.createOrFold<tosa::CastOp>(
|
2023-06-16 15:51:24 +08:00
|
|
|
op->getLoc(), RankedTensorType::get(shape, *dtype), const_op);
|
|
|
|
}
|
2022-01-15 05:57:27 +08:00
|
|
|
return const_op.getResult();
|
|
|
|
}
|
|
|
|
|
2022-10-04 21:05:59 +08:00
|
|
|
static LogicalResult checkValidityOfCast(Type src, Type dest) {
|
2023-06-13 22:17:23 +08:00
|
|
|
if ((src == dest) || (src.isInteger(64) && dest.isInteger(32)) ||
|
2022-11-16 04:23:39 +08:00
|
|
|
(src.isInteger(64) && dest.isInteger(8)) ||
|
2022-10-04 21:05:59 +08:00
|
|
|
(src.isInteger(64) && dest.isInteger(1)) ||
|
2023-01-04 00:30:16 +08:00
|
|
|
(src.isInteger(64) && dest.isF32()) ||
|
2022-11-16 04:23:39 +08:00
|
|
|
(src.isInteger(32) && dest.isInteger(64)) ||
|
2022-10-04 21:05:59 +08:00
|
|
|
(src.isInteger(32) && dest.isInteger(1)) ||
|
2023-01-04 00:30:16 +08:00
|
|
|
(src.isInteger(32) && dest.isF32()) ||
|
2023-05-13 06:18:23 +08:00
|
|
|
(src.isInteger(32) && dest.isBF16()) ||
|
|
|
|
(src.isInteger(16) && dest.isBF16()) ||
|
2022-10-04 21:05:59 +08:00
|
|
|
(src.isInteger(8) && dest.isInteger(1)) ||
|
2023-05-13 06:18:23 +08:00
|
|
|
(src.isInteger(8) && dest.isBF16()) ||
|
2023-01-28 01:21:06 +08:00
|
|
|
(src.isInteger(1) && dest.isInteger(64)) ||
|
2023-06-13 22:17:23 +08:00
|
|
|
(src.isInteger(1) && dest.isF32()) || (src.isF32() && dest.isF64()) ||
|
|
|
|
(src.isF32() && dest.isBF16()) || (src.isF64() && dest.isF32()) ||
|
|
|
|
(src.isF64() && dest.isBF16()) || (src.isF32() && dest.isInteger(8)) ||
|
2023-04-08 00:40:56 +08:00
|
|
|
(src.isF32() && dest.isInteger(64)) ||
|
2023-05-13 06:18:23 +08:00
|
|
|
(src.isF32() && dest.isInteger(1)) ||
|
|
|
|
(src.isBF16() && dest.isInteger(8)) ||
|
|
|
|
(src.isBF16() && dest.isInteger(16)) ||
|
2023-06-13 22:17:23 +08:00
|
|
|
(src.isBF16() && dest.isInteger(32)) || (src.isBF16() && dest.isF32())) {
|
2022-10-04 21:05:59 +08:00
|
|
|
return success();
|
|
|
|
}
|
|
|
|
return failure();
|
|
|
|
}
|
|
|
|
|
|
|
|
// Template specialization for float
|
|
|
|
LogicalResult tosaCastTensorToType(PatternRewriter &rewriter, Operation *op,
|
|
|
|
Value src, Type destType, Value &result) {
|
|
|
|
|
|
|
|
Type srcElemTy = src.getType().dyn_cast<TensorType>().getElementType();
|
|
|
|
Type destElemTy = destType.dyn_cast<TensorType>().getElementType();
|
|
|
|
|
|
|
|
if (failed(checkValidityOfCast(srcElemTy, destElemTy)))
|
|
|
|
return rewriter.notifyMatchFailure(
|
|
|
|
op, "casting to result dtype is invalid or unsupported");
|
|
|
|
|
|
|
|
if (destElemTy.isInteger(1)) {
|
|
|
|
auto srcType = src.getType().dyn_cast<TensorType>();
|
|
|
|
SmallVector<int64_t> srcShape(srcType.getShape());
|
|
|
|
uint64_t num_total_elements = 1;
|
|
|
|
for (int64_t a : srcShape)
|
|
|
|
num_total_elements *= a;
|
|
|
|
|
2022-12-20 18:17:27 +08:00
|
|
|
std::optional<Value> constOp;
|
2022-10-04 21:05:59 +08:00
|
|
|
if (srcElemTy.isInteger(64)) {
|
|
|
|
SmallVector<int64_t> values(num_total_elements, 0);
|
|
|
|
constOp =
|
|
|
|
tosa::getConstTensor<int64_t>(rewriter, op, values, srcShape).value();
|
|
|
|
} else if (srcElemTy.isInteger(32)) {
|
|
|
|
SmallVector<int32_t> values(num_total_elements, 0);
|
|
|
|
constOp =
|
|
|
|
tosa::getConstTensor<int32_t>(rewriter, op, values, srcShape).value();
|
|
|
|
} else if (srcElemTy.isF32()) {
|
|
|
|
SmallVector<float> values(num_total_elements, 0.0);
|
|
|
|
constOp =
|
|
|
|
tosa::getConstTensor<float>(rewriter, op, values, srcShape).value();
|
|
|
|
} else if (srcElemTy.isInteger(8)) {
|
|
|
|
SmallVector<int8_t> values(num_total_elements, 0);
|
|
|
|
constOp =
|
|
|
|
tosa::getConstTensor<int8_t>(rewriter, op, values, srcShape).value();
|
|
|
|
}
|
|
|
|
Value equalToZero = rewriter.create<tosa::EqualOp>(op->getLoc(), destType,
|
|
|
|
src, constOp.value());
|
|
|
|
result = rewriter.create<tosa::LogicalNotOp>(op->getLoc(), destType,
|
|
|
|
equalToZero);
|
|
|
|
} else {
|
|
|
|
result = rewriter.create<tosa::CastOp>(op->getLoc(), destType, src);
|
|
|
|
}
|
|
|
|
return success();
|
|
|
|
}
|
|
|
|
|
2023-01-12 06:40:03 +08:00
|
|
|
Value promoteType(PatternRewriter &rewriter, Value input, TensorType outType) {
|
|
|
|
Operation *op = input.getDefiningOp();
|
|
|
|
TensorType inType = input.getType().cast<TensorType>();
|
|
|
|
|
|
|
|
if (inType.getElementType() != outType.getElementType()) {
|
|
|
|
TensorType promotedType =
|
|
|
|
inType.cloneWith(inType.getShape(), outType.getElementType());
|
|
|
|
return rewriter.create<tosa::CastOp>(op->getLoc(), promotedType, input);
|
|
|
|
}
|
|
|
|
return input;
|
|
|
|
}
|
|
|
|
|
2022-01-15 05:57:27 +08:00
|
|
|
// Template instantiation
|
2024-01-30 01:59:33 +08:00
|
|
|
template std::optional<Value>
|
|
|
|
getConstTensor<bool>(PatternRewriter &, Operation *, ArrayRef<bool> vec,
|
|
|
|
ArrayRef<int64_t> shape, std::optional<Type> dtype);
|
|
|
|
|
|
|
|
template std::optional<Value>
|
|
|
|
getConstTensor<int32_t>(PatternRewriter &, Operation *, ArrayRef<int32_t> vec,
|
|
|
|
ArrayRef<int64_t> shape, std::optional<Type> dtype);
|
|
|
|
|
|
|
|
template std::optional<Value>
|
|
|
|
getConstTensor<int64_t>(PatternRewriter &, Operation *, ArrayRef<int64_t> vec,
|
|
|
|
ArrayRef<int64_t> shape, std::optional<Type> dtype);
|
2023-06-13 22:17:23 +08:00
|
|
|
|
|
|
|
LogicalResult getAvgPool2dAccType(PatternRewriter &rewriter, Value input,
|
|
|
|
TypeAttr &accType) {
|
|
|
|
auto inputTy = llvm::dyn_cast<ShapedType>(input.getType());
|
|
|
|
if (!inputTy)
|
|
|
|
return failure();
|
|
|
|
auto inputETy = inputTy.getElementType();
|
|
|
|
|
|
|
|
if (auto quantType =
|
|
|
|
llvm::dyn_cast<mlir::quant::UniformQuantizedType>(inputETy))
|
|
|
|
inputETy = quantType.getStorageType();
|
|
|
|
|
|
|
|
// Tosa supports FP16 and FP32 accumulator type for FP16 input. When the time
|
|
|
|
// FP16 is supported, the accumulator type can be selected based on trade-off
|
|
|
|
// between performance and accuracy. Set to FP32 by default.
|
|
|
|
accType = inputETy.isa<FloatType>()
|
|
|
|
? mlir::TypeAttr::get(rewriter.getF32Type())
|
|
|
|
: mlir::TypeAttr::get(rewriter.getIntegerType(32));
|
|
|
|
|
|
|
|
return success();
|
|
|
|
}
|
|
|
|
|
2021-12-03 08:52:01 +08:00
|
|
|
} // namespace tosa
|
|
|
|
} // namespace mlir
|