From 50796f3818ba0647141b7ea9b5563d2df0fbf89b Mon Sep 17 00:00:00 2001
From: Hendrik Borras <hendrikborras@web.de>
Date: Thu, 19 Aug 2021 12:41:28 +0100
Subject: [PATCH] Implemented cppsim functionality for upsampling (some issues
remain).
---
src/finn/custom_op/fpgadataflow/upsampler.py | 48 ++++---------------
.../fpgadataflow/convert_to_hls_layers.py | 5 +-
2 files changed, 13 insertions(+), 40 deletions(-)
diff --git a/src/finn/custom_op/fpgadataflow/upsampler.py b/src/finn/custom_op/fpgadataflow/upsampler.py
index 6adb929f0..04beb1b1c 100644
--- a/src/finn/custom_op/fpgadataflow/upsampler.py
+++ b/src/finn/custom_op/fpgadataflow/upsampler.py
@@ -58,24 +58,12 @@ class UpsampleNearestNeighbour_Batch(HLSCustomOp):
return oshape
def get_folded_input_shape(self):
- raise NotImplementedError
normal_ishape = list(self.get_normal_input_shape())
- ifm_ch = self.get_nodeattr("NumChannels")
- simd = self.get_nodeattr("SIMD")
- assert ifm_ch % simd == 0, "SIMD must divide input channels"
- fold = int(normal_ishape[-1] / simd)
- folded_ishape = normal_ishape[:-1] + [fold, simd]
- return tuple(folded_ishape)
+ return tuple(normal_ishape)
def get_folded_output_shape(self):
- raise NotImplementedError
normal_oshape = list(self.get_normal_output_shape())
- ifm_ch = self.get_nodeattr("NumChannels")
- simd = self.get_nodeattr("SIMD")
- assert ifm_ch % simd == 0, "SIMD must divide input channels"
- fold = int(normal_oshape[-1] / simd)
- folded_oshape = normal_oshape[:-1] + [fold, simd]
- return tuple(folded_oshape)
+ return tuple(normal_oshape)
def make_shape_compatible_op(self, model):
exp_ishape = self.get_normal_input_shape()
@@ -122,21 +110,16 @@ class UpsampleNearestNeighbour_Batch(HLSCustomOp):
return ret
def get_output_datatype(self):
- raise NotImplementedError
"""Returns FINN DataType of output. (Same as input datatype)"""
return self.get_input_datatype()
def get_instream_width(self):
- raise NotImplementedError
ibits = self.get_input_datatype().bitwidth()
- simd = self.get_nodeattr("SIMD")
- return ibits * simd
+ return ibits
def get_outstream_width(self):
- raise NotImplementedError
obits = self.get_output_datatype().bitwidth()
- simd = self.get_nodeattr("SIMD")
- return obits * simd
+ return obits
def get_number_output_values(self):
raise NotImplementedError
@@ -144,11 +127,9 @@ class UpsampleNearestNeighbour_Batch(HLSCustomOp):
return np.prod(folded_oshape[:-1])
def global_includes(self):
- raise NotImplementedError
- self.code_gen_dict["$GLOBALS$"] = ['#include "slidingwindow.h"']
+ self.code_gen_dict["$GLOBALS$"] = ['#include "upsample.hpp"']
def defines(self, var):
- raise NotImplementedError
self.code_gen_dict["$DEFINES$"] = []
ifm_ch = self.get_nodeattr("NumChannels")
@@ -157,20 +138,16 @@ class UpsampleNearestNeighbour_Batch(HLSCustomOp):
ibits = self.get_input_datatype().bitwidth()
self.code_gen_dict["$DEFINES$"] += ["#define Input_precision {}".format(ibits)]
- idim = self.get_nodeattr("ImgDim")
+ idim = self.get_nodeattr("IFMDim")
self.code_gen_dict["$DEFINES$"] += ["#define IFMDim {}".format(idim)]
- simd = self.get_nodeattr("SIMD")
- self.code_gen_dict["$DEFINES$"] += ["#define SIMD {}".format(simd)]
-
- stride = self.get_nodeattr("Stride")
- self.code_gen_dict["$DEFINES$"] += ["#define Stride {}".format(stride)]
+ odim = self.get_nodeattr("OFMDim")
+ self.code_gen_dict["$DEFINES$"] += ["#define OFMDim {}".format(odim)]
batch_size = self.get_nodeattr("numInputVectors")
self.code_gen_dict["$DEFINES$"] += ["#define numReps {}".format(batch_size)]
def read_npy_data(self):
- raise NotImplementedError
code_gen_dir = self.get_nodeattr("code_gen_dir_cppsim")
dtype = self.get_input_datatype()
if dtype == DataType.BIPOLAR:
@@ -189,7 +166,6 @@ class UpsampleNearestNeighbour_Batch(HLSCustomOp):
)
def strm_decl(self):
- raise NotImplementedError
self.code_gen_dict["$STREAMDECLARATIONS$"] = []
self.code_gen_dict["$STREAMDECLARATIONS$"].append(
'hls::stream<ap_uint<{}>> in0 ("in0");'.format(self.get_instream_width())
@@ -199,14 +175,12 @@ class UpsampleNearestNeighbour_Batch(HLSCustomOp):
)
def docompute(self):
- raise NotImplementedError
self.code_gen_dict["$DOCOMPUTE$"] = [
- """ConvolutionInputGenerator_kernel1<IFMChannels, Input_precision,
- IFMDim, SIMD,Stride> (in0, out, numReps);"""
+ """UpsampleNearestNeighbour_Batch<OFMDim, IFMDim, IFMChannels,
+ ap_uint<Input_precision> > (in0, out, numReps);"""
]
def dataoutstrm(self):
- raise NotImplementedError
code_gen_dir = self.get_nodeattr("code_gen_dir_cppsim")
dtype = self.get_output_datatype()
if dtype == DataType.BIPOLAR:
@@ -234,7 +208,6 @@ class UpsampleNearestNeighbour_Batch(HLSCustomOp):
]
def save_as_npy(self):
- raise NotImplementedError
self.code_gen_dict["$SAVEASCNPY$"] = []
def blackboxfunction(self):
@@ -247,7 +220,6 @@ class UpsampleNearestNeighbour_Batch(HLSCustomOp):
]
def pragmas(self):
- raise NotImplementedError
self.code_gen_dict["$PRAGMAS$"] = ["#pragma HLS INTERFACE axis port=in0"]
self.code_gen_dict["$PRAGMAS$"].append("#pragma HLS INTERFACE axis port=out")
self.code_gen_dict["$PRAGMAS$"].append(
diff --git a/src/finn/transformation/fpgadataflow/convert_to_hls_layers.py b/src/finn/transformation/fpgadataflow/convert_to_hls_layers.py
index 8c1e58ceb..78c843209 100644
--- a/src/finn/transformation/fpgadataflow/convert_to_hls_layers.py
+++ b/src/finn/transformation/fpgadataflow/convert_to_hls_layers.py
@@ -220,8 +220,9 @@ class InferConvInpGen(Transformation):
class InferUpsample(Transformation):
- """Convert Upsample and Resize nodes to layers
- to UpsampleNearestNeighbour_Batch nodes."""
+ """
+ Convert Upsample and Resize nodes to layers to UpsampleNearestNeighbour_Batch nodes.
+ """
def apply(self, model):
graph = model.graph
--
GitLab