diff --git a/llvm/projects/hpvm-tensor-rt/docs/Changes.md b/llvm/projects/hpvm-tensor-rt/docs/Changes.md
new file mode 100644
index 0000000000000000000000000000000000000000..ebf0295ed664bf618195d79a461b58b576b0c071
--- /dev/null
+++ b/llvm/projects/hpvm-tensor-rt/docs/Changes.md
@@ -0,0 +1,62 @@
+# Header Files Needing Modification
+
+- llvm/include/IR/IntrisicsVISC.td
+ - Add Tensor intrinsics here
+
+- bechmarks/common/include/visc.h - Benchmarks header
+ - Include tensor intrinsic declarations
+
+- Header files for "PROMISE_TARGET" and "CUDNN_TARGET" Hints
+ - /include/SupportVISC/DFG2LLVM.h
+ - /include/SupportVISC/VISCUtils.h
+ - /include/SupportVISC/VISCHint.h
+
+
+
+
+# Passes Needing Modification
+
+- GenVISC
+ * Handle Tensor Intrinsics
+ * Add Declarations for Intrinsic Functions called (at top)
+ * Modify runOnModule to include rules for functions -> intrinsics
+
+
+- ClearDFG
+ * Incorporate new changes for visc.node.id
+
+
+# New Passes to Incorporate in LLVM-9 Branch
+
+- lib/Transforms/DFG2LLVM_CUDNN
+- lib/Transforms/DFG2LLVM_WrapperAPI
+- lib/Transforms/FusedNodesHPVM
+
+** Changes needed:
+- Modify paths to tensor_runtime.ll
+- Use "hpvm" instead of "visc" intrinsics
+
+
+# Projects to Move
+
+- hpvm-tensor-rt
+ * Make sure to not move 'tuner_results' and 'model_params' subdirectories
+
+- gpu_profiler
+
+- soc_simlator
+
+- Keras (wait on it)
+
+
+
+# Build System
+
+- Add rules to /lib/Transforms/CMakeLists.txt
+- Add rules to /llvm/projects/CmakeLists.txt
+- Automate the generation of 'tensor_runtime.ll`
+- Add CUDNN, CUDA paths to the template environment setup file
+- Move to using Cmake-3.17 (earlier 3.15)
+
+
+
diff --git a/llvm/projects/hpvm-tensor-rt/lib/tensor_runtime.ll b/llvm/projects/hpvm-tensor-rt/lib/tensor_runtime.ll
index 42325406bd3308a7bc5f165b491da85cd1024f27..efef5f2c2f53d4812edfa4b6d231ca5a1cc52371 100644
--- a/llvm/projects/hpvm-tensor-rt/lib/tensor_runtime.ll
+++ b/llvm/projects/hpvm-tensor-rt/lib/tensor_runtime.ll
@@ -67,6 +67,10 @@ entry:
%FCLayer = alloca i8*, align 8
%ConvLayer_ = alloca i8*, align 8
%ConvLayer2 = alloca i8*, align 8
+<<<<<<< HEAD
+=======
+ %GroupConvLayer = alloca i8*, align 8
+>>>>>>> 2a31471f4de89edd9180689d139be7ca65b0df08
%FCLayer2 = alloca i8*, align 8
%AddWrapper = alloca i8*, align 8
%ReluWrapper = alloca i8*, align 8
@@ -122,6 +126,10 @@ entry:
store i8* bitcast (i8* (i8*, float, float, i8*, float, float, i8*, float, float, i32, float, float, i32)* @FCLayer_PROMISE to i8*), i8** %FCLayer, align 8
store i8* bitcast (i8* (i8*, i8*, i8*, i8*, i32, i32, i32, i32, i32, i32, i32, float, float)* @wrapper_ConvLayer to i8*), i8** %ConvLayer_, align 8
store i8* bitcast (i8* (i8*, i8*, i8*, i8*, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, float, float)* @wrapper_ConvLayer2 to i8*), i8** %ConvLayer2, align 8
+<<<<<<< HEAD
+=======
+ store i8* bitcast (i8* (i8*, i8*, i8*, i32, i32, i32, i32, i32, i32)* @wrapper_tensorGroupConvolution to i8*), i8** %GroupConvLayer, align 8
+>>>>>>> 2a31471f4de89edd9180689d139be7ca65b0df08
store i8* bitcast (i8* (i8*, i8*, i8*, i8*, i32, float, float)* @wrapper_FCLayer to i8*), i8** %FCLayer2, align 8
store i8* bitcast (i8* (i8*, i8*, i8*)* @wrapper_tensorAdd to i8*), i8** %AddWrapper, align 8
store i8* bitcast (i8* (i8*, i8*)* @wrapper_tensorRelu to i8*), i8** %ReluWrapper, align 8
@@ -219,6 +227,8 @@ declare i8* @wrapper_ConvLayer(i8*, i8*, i8*, i8*, i32, i32, i32, i32, i32, i32,
declare i8* @wrapper_ConvLayer2(i8*, i8*, i8*, i8*, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, float, float) #1
+declare i8* @wrapper_tensorGroupConvolution(i8*, i8*, i8*, i32, i32, i32, i32, i32, i32) #1
+
declare i8* @wrapper_FCLayer(i8*, i8*, i8*, i8*, i32, float, float) #1
declare i8* @wrapper_tensorAdd(i8*, i8*, i8*) #1
diff --git a/llvm/projects/hpvm-tensor-rt/tensor_runtime/include/tensor_signatures.cc b/llvm/projects/hpvm-tensor-rt/tensor_runtime/include/tensor_signatures.cc
index 9c4cf97908a129e421e61d29177e7a4359a860e6..478a7ba037702fd74d6dffb489ca093a49405449 100644
--- a/llvm/projects/hpvm-tensor-rt/tensor_runtime/include/tensor_signatures.cc
+++ b/llvm/projects/hpvm-tensor-rt/tensor_runtime/include/tensor_signatures.cc
@@ -47,6 +47,10 @@ void dummyFunction() {
void *ConvLayer_ = (void *)&wrapper_ConvLayer;
void *ConvLayer2 = (void *)&wrapper_ConvLayer2;
+<<<<<<< HEAD
+=======
+ void *GroupConvLayer = (void *)&wrapper_tensorGroupConvolution;
+>>>>>>> 2a31471f4de89edd9180689d139be7ca65b0df08
void *FCLayer2 = (void *)&wrapper_FCLayer;
void *AddWrapper = (void *)&wrapper_tensorAdd;
diff --git a/llvm/projects/keras/cmake_template/CMakeLists.txt b/llvm/projects/keras/cmake_template/CMakeLists.txt
index 67ce4a848b4de67c15cc2a8de032f6dd063d6fc7..0cdee697ce2d663775f0283e96f35c45fd467986 100644
--- a/llvm/projects/keras/cmake_template/CMakeLists.txt
+++ b/llvm/projects/keras/cmake_template/CMakeLists.txt
@@ -50,7 +50,7 @@ find_library(TENSOR_LIB
)
-set(LINK_LIBS cudart cudnn cufft cublas stdc++fs curand)
+set(LINK_LIBS cudart cudnn cufft cublas stdc++fs curand -pthread)
#### Image Processing Benchmarks
diff --git a/llvm/projects/keras/frontend/approxhpvm_translator.py b/llvm/projects/keras/frontend/approxhpvm_translator.py
index bd30ad2a44392bd71395958c8b58c24d8d9d66ad..e60d6adb994ee4da56765a1fa365b14a792863d2 100644
--- a/llvm/projects/keras/frontend/approxhpvm_translator.py
+++ b/llvm/projects/keras/frontend/approxhpvm_translator.py
@@ -21,6 +21,7 @@ class DFG:
def hasSingleInput(self, layer):
+
layer_name = layer.__class__.__name__
singleInLayers = {}
@@ -57,12 +58,12 @@ class DFG:
inbound_node_name = inbound_node_name.split("/")[0]
if inbound_node_name in self.node_map:
inbound_node = self.node_map[inbound_node_name]
- print (inbound_node_name, " found!")
+ DEBUG (inbound_node_name, " found!")
inbound_node.add_output(dfg_node)
dfg_node.add_input(inbound_node)
else:
- print ("--inbound node NOT FOUND!")
+ DEBUG ("--inbound node NOT FOUND!")
@@ -75,12 +76,12 @@ class DFG:
if self.hasMultipleInputs(layer):
for j in range(len(layer.input)):
- print(type(layer.input[j]))
- print(layer.input[j].op.name)
+ DEBUG (type(layer.input[j]))
+ DEBUG (layer.input[j].op.name)
self.add_dfg_edge(layer.input[j].op.name, dfg_node)
else:
- print (layer.input.name)
+ DEBUG (layer.input.name)
self.add_dfg_edge(layer.input.name, dfg_node)
# Adding DFG node to name mapping
@@ -104,8 +105,8 @@ class DFG:
return
if self.predVisited(cur_node, visited_nodes):
- print(cur_node.layer_type)
- print(cur_node.layer_name)
+ DEBUG (cur_node.layer_type)
+ DEBUG (cur_node.layer_name)
visited_nodes[cur_node.layer_name] = True
# Invoking traversal on outbound nodes
@@ -119,7 +120,7 @@ class DFG:
#Build and Print the DFG in reverse postorder
def buildDFG(self):
- print ("\n\n ****** Traversing and Printing DFG ******* \n\n")
+ DEBUG ("\n\n ****** Traversing and Printing DFG ******* \n\n")
visited_nodes = {}
# Starting traversal at the DFG root node
self.traverseNode(self.root_node, visited_nodes)
@@ -144,42 +145,42 @@ class DFGNode:
layer_type = layer.__class__.__name__
self.layer_type = layer_type # layer type e.g., conv2d, add, dense
self.layer_name = layer.name # unique layer identifier
- print (self.layer_name)
+ DEBUG (self.layer_name)
if layer_type == "Conv2D" or layer_type == "DepthwiseConv2D" or layer_type == "Dense":
self.weights = layer.get_weights()[0]
- print("\t", self.weights.shape)
+ DEBUG ("\t", self.weights.shape)
self.use_bias = layer.use_bias
if layer.use_bias:
self.use_bias = layer.use_bias
self.bias_weights = layer.get_weights()[1]
- print("\t", self.bias_weights.shape)
+ DEBUG ("\t", self.bias_weights.shape)
if layer_type == "Conv2D" or layer_type == "DepthwiseConv2D":
self.padding = layer.padding
self.strides = layer.strides
- print("\t", self.strides)
- print("\tPadding = ", self.padding)
+ DEBUG ("\t", self.strides)
+ DEBUG ("\tPadding = ", self.padding)
if layer_type == "MaxPooling2D" or layer_type == "AveragePooling2D":
self.pool_size = layer.pool_size
self.strides = layer.strides
- print("\t pool_size = ", self.pool_size)
- print("\t strides = ", self.strides)
+ DEBUG ("\t pool_size = ", self.pool_size)
+ DEBUG ("\t strides = ", self.strides)
if layerHasActivationAttr(self):
self.activation_type = layer.activation.__name__
- print ("\t Activation = ", self.activation_type)
+ DEBUG ("\t Activation = ", self.activation_type)
if layer_type == "ZeroPadding2D":
- print ("***ZeroPaddding \n");
+ DEBUG ("***ZeroPaddding \n");
self.padding = layer.padding
- print ("padding = ", self.padding);
+ DEBUG ("padding = ", self.padding);
if layer_type == "BatchNormalization":
self.epsilon = layer.epsilon
@@ -264,12 +265,12 @@ class TensorRtTranslator:
def getSingleInputName(self, cur_node):
- print (cur_node.layer_name)
+ DEBUG (cur_node.layer_name)
# Assumption: If no inputs, the previous layer must be input layer
if len(cur_node.inputs) == 0:
return "input"
- print ("Input_type = ", cur_node.inputs[0].layer_type)
+ DEBUG ("Input_type = ", cur_node.inputs[0].layer_type)
# NOTE: Assuming the 'inference' phase - hence skipping Dropout
pred_layer_type = cur_node.inputs[0].layer_type
@@ -295,12 +296,12 @@ class TensorRtTranslator:
def getPrevLayerPadding(self, cur_node):
- print (cur_node.layer_name)
+ DEBUG (cur_node.layer_name)
# Assumption: If no inputs, the previous layer must be input layer
if len(cur_node.inputs) == 0:
return None
- print ("Input_type = ", cur_node.inputs[0].layer_type)
+ DEBUG ("Input_type = ", cur_node.inputs[0].layer_type)
if cur_node.inputs[0].layer_type == "ZeroPadding2D":
pred_padding = cur_node.inputs[0].padding
return pred_padding
@@ -483,9 +484,7 @@ class TensorRtTranslator:
self.program_str += inst_str
-
-
-
+
def codegenNode(self, dfg, cur_node, visited_nodes):
@@ -494,12 +493,11 @@ class TensorRtTranslator:
if cur_node.layer_name in visited_nodes:
return
- print ("-visiting = ", cur_node.layer_name, "\n")
+ DEBUG ("-visiting = ", cur_node.layer_name, "\n")
if dfg.predVisited(cur_node, visited_nodes):
visited_nodes[cur_node.layer_name] = True
-
self.genNodeCalls(cur_node)
# Invoking traversal on outbound nodes
@@ -510,11 +508,13 @@ class TensorRtTranslator:
# Print the DFG in reverse postorder
def codegen(self, dfg):
- print ("\n\n ****** Codegen for HPVM Tensor Rt ******* \n\n")
+ print ("\n *** Starting Codegen for HPVM Tensor Rt *** \n")
visited_nodes = {}
# Starting traversal at the DFG root node
self.codegenNode(dfg, dfg.root_node, visited_nodes)
+ print ("\n\n --- Codegen Completed --- \n\n")
+
@@ -531,7 +531,7 @@ class TensorRtTranslator:
w_name = layer_name + "_w"
self.filter_names[w_name] = 1
- print (weights.shape, w_name)
+ DEBUG (weights.shape, w_name)
N = weights.shape[3]
C = weights.shape[2]
@@ -563,7 +563,7 @@ class TensorRtTranslator:
b_name = layer_name + "_b"
self.filter_names[b_name] = 1
- print (bias_weights.shape, b_name)
+ DEBUG (bias_weights.shape, b_name)
unique_file_name = b_name + ".bin"
dumpFcBias(prefix + unique_file_name, bias_weights, bias_weights.shape[0])
@@ -584,7 +584,7 @@ class TensorRtTranslator:
w_name = layer_name + "_w"
self.filter_names[w_name] = 1
- print (weights.shape, w_name)
+ DEBUG (weights.shape, w_name)
H = weights.shape[0]
W = weights.shape[1]
@@ -606,7 +606,7 @@ class TensorRtTranslator:
b_name = layer_name + "_b"
self.filter_names[b_name] = 1
- print (bias_weights.shape, b_name)
+ DEBUG (bias_weights.shape, b_name)
unique_file_name = b_name + ".bin"
dumpFcBias(prefix + unique_file_name, bias_weights, bias_weights.shape[0])
@@ -862,7 +862,7 @@ def reloadModelParams(model, reload_dir, x_test, y_test):
for i in range(len(model.layers)):
layer = model.layers[i]
layer_name = layer.name
- print ("*layer_name = ", layer_name)
+ DEBUG ("*layer_name = ", layer_name)
if "conv" not in layer_name and "dense" not in layer_name:
continue
@@ -870,9 +870,6 @@ def reloadModelParams(model, reload_dir, x_test, y_test):
w_path = reload_dir + layer_name + "_w.bin"
b_path = reload_dir + layer_name + "_b.bin"
- print ("** w_path = ", w_path)
- print ("** b_path = ", b_path)
-
w_arr = np.fromfile(w_path, dtype='float32')
b_arr = np.fromfile(b_path, dtype='float32')
@@ -885,7 +882,7 @@ def reloadModelParams(model, reload_dir, x_test, y_test):
b_arr = np.reshape(b_arr, b_shape)
w_arr = np.transpose(w_arr, (2,3,1,0))
- print ("old_shape = ", w_shape, " new_shape = ", w_arr.shape)
+ DEBUG ("old_shape = ", w_shape, " new_shape = ", w_arr.shape)
if "dense" in layer_name:
w_arr = np.reshape(w_arr, w_shape)
@@ -921,7 +918,6 @@ def getUniquePath(weights_dir):
weights_dir = getUniquePath(weights_dir)
- #print (weights_dir)
return weights_dir
@@ -950,8 +946,8 @@ def translate_to_approxhpvm(model, weights_dir, test_data=None, test_labels=None
dfg.buildDFG()
- print ("test_data.shape = ", test_data.shape, "\n")
- print ("test_labels.shape = ", test_labels.shape, "\n")
+ DEBUG ("test_data.shape = ", test_data.shape, "\n")
+ DEBUG ("test_labels.shape = ", test_labels.shape, "\n")
tensorRtTranslator = TensorRtTranslator(dfg)
tensorRtTranslator.translate(model, weights_dir, test_data, test_labels)
@@ -959,13 +955,17 @@ def translate_to_approxhpvm(model, weights_dir, test_data=None, test_labels=None
input_str = tensorRtTranslator.getInputStr()
- promiseRtTranslator = PromiseRtTranslator(dfg, weight_str)
- promiseRtTranslator.translate(model, weights_dir, test_data)
+ #promiseRtTranslator = PromiseRtTranslator(dfg, weight_str)
+ #promiseRtTranslator.translate(model, weights_dir, test_data)
filter_names = tensorRtTranslator.getFilterNames()
hpvmTranslator = HPVMTranslator(dfg, weight_str, input_str, filter_names)
hpvmTranslator.translate(model, weights_dir, test_data)
-
+
+ print ("-- Weight Files Under : ", weights_dir)
+ print ("-- TensorRT src : ", weights_dir + "/src.cc")
+ print ("-- ApproxHPVM src : ", weights_dir + "approxhpvm_src.cc")
+
return weights_dir
diff --git a/llvm/projects/keras/frontend/hpvm_dfg_translator.py b/llvm/projects/keras/frontend/hpvm_dfg_translator.py
index 233652600a564f2ca2533030b1f623b4aab06f48..65574a98881f010bf7cd67df344517803de8c67c 100644
--- a/llvm/projects/keras/frontend/hpvm_dfg_translator.py
+++ b/llvm/projects/keras/frontend/hpvm_dfg_translator.py
@@ -59,12 +59,12 @@ class HPVMTranslator:
def getSingleInputName(self, cur_node):
- print (cur_node.layer_name)
+ DEBUG (cur_node.layer_name)
# Assumption: If no inputs, the previous layer must be input layer
if len(cur_node.inputs) == 0:
return "input"
- print ("Input_type = ", cur_node.inputs[0].layer_type)
+ DEBUG ("Input_type = ", cur_node.inputs[0].layer_type)
pred_layer_type = cur_node.inputs[0].layer_type
# NOTE: Assuming the 'inference' phase - hence skipping Dropout
@@ -90,12 +90,12 @@ class HPVMTranslator:
def getPrevLayerPadding(self, cur_node):
- print (cur_node.layer_name)
+ DEBUG (cur_node.layer_name)
# Assumption: If no inputs, the previous layer must be input layer
if len(cur_node.inputs) == 0:
return None
- print ("Input_type = ", cur_node.inputs[0].layer_type)
+ DEBUG ("Input_type = ", cur_node.inputs[0].layer_type)
if cur_node.inputs[0].layer_type == "ZeroPadding2D":
pred_padding = cur_node.inputs[0].padding
return pred_padding
@@ -166,7 +166,7 @@ class HPVMTranslator:
node_header_str += ", "
node_header_str += ") { \n"
- node_header_str += " __visc__hint(visc::CUDNN_TARGET); \n"
+ node_header_str += " __visc__hint(visc::PROMISE_TARGET); \n"
node_header_str += " __visc__attributes(" + str(num_params) + ", "
for i in range(num_params):
@@ -215,8 +215,8 @@ class HPVMTranslator:
# Fix: replace deprecated genHpvmNodeEdges with genHpvmEdges
def genHpvmNodeEdges(self, out_var_name, input_var_name, input_var_name2):
- print ("input_var_name2 = ", input_var_name2)
- print ("input_var_name = ", input_var_name)
+ DEBUG ("input_var_name2 = ", input_var_name2)
+ DEBUG ("input_var_name = ", input_var_name)
hpvm_edge_str = "\n void* " + out_var_name + " = "
hpvm_edge_str += "__visc__createNodeND(0, " + out_var_name + "_node); \n\n"
@@ -462,7 +462,7 @@ class HPVMTranslator:
if layer_type == "MaxPooling2D":
func_name = "__visc__tensor_pool_max"
if layer_type == "AveragePooling2D":
- func_name = "__visc__tensor_pool_avg"
+ func_name = "__visc__tensor_pool_mean"
inst_str += " void* r = " + func_name + "(t1, "
inst_str += str(pool_size[0]) + ", " + str(pool_size[1]) + ", "
@@ -531,11 +531,13 @@ class HPVMTranslator:
# Print the DFG in reverse postorder
def codegen(self, dfg):
- print ("\n\n ****** Codegen for ApproxHPVM DFG Representation ******* \n\n")
+ print ("\n *** Starting Codegen for ApproxHPVM DFG Representation *** \n")
visited_nodes = {}
# Starting traversal at the DFG root node
self.codegenNode(dfg, dfg.root_node, visited_nodes)
+ print ("\n --- Codegen Completed --- \n")
+
@@ -648,7 +650,7 @@ class HPVMTranslator:
program_str = self.file_header_str + self.node_str + self.root_str
program_str += self.root_struct_str + self.main_func_str
- print (program_str)
+ DEBUG (program_str)
f = open(dir_prefix + "/approxhpvm_src.cc", "w+")
f.write(program_str)
diff --git a/llvm/projects/keras/frontend/promise_translator.py b/llvm/projects/keras/frontend/promise_translator.py
index 54982bd7c7f906d5fe27ddb052ae303575c260ba..30acf47cd8aeedaaae8ee1ba0fda637a7d931940 100644
--- a/llvm/projects/keras/frontend/promise_translator.py
+++ b/llvm/projects/keras/frontend/promise_translator.py
@@ -386,11 +386,8 @@ class PromiseRtTranslator:
if len(cur_node.inputs) == 0:
return "input"
- #print ("Input_type = ", cur_node.inputs[0].layer_type)
-
pred_layer_type = cur_node.inputs[0].layer_type
# FIXME: Assuming the 'inference' phase - hence skipping Dropout
- #if pred_layer_type == "Flatten" or pred_layer_type == "Dropout":
if self.isSkipLayer(pred_layer_type):
cur_node = self.getPrevActiveLayer(cur_node)
@@ -1056,8 +1053,6 @@ class PromiseRtTranslator:
# Inference over test set
layer_outs = functor([x_test[start:end], 1.])
-
- ###layer_outs = functor([x_test[start:end]])
# NOTE: Saving quant ranges for input
min_val = np.amin(x_test)
@@ -1096,12 +1091,9 @@ class PromiseRtTranslator:
min_val = range[0]
if range[1] > max_val:
max_val = range[1]
-
- # print ("****** min = ", range[0], " max = ", range[1])
self.quant_ranges[layer_name] = (min_val, max_val)
-
print ("---- min = ", min_val, " max = ", max_val, " ----- \n\n")
ind += 1
@@ -1123,20 +1115,6 @@ class PromiseRtTranslator:
print ("layer_shape = ", layer.input.shape)
self.layer_input_sizes[layer_name] = layer.input.shape
- #inp = model.input # input placeholder
- #inputs = []
- # functor = K.function([inp, K.learning_phase()], outputs ) # evaluation function
- # functor2 = K.function([inp, K.learning_phase()], inputs ) # evaluation function
- # Testing
- # layer_outs = functor([x_test, 1.])
- # layer_inputs = functor2([x_test, 1.])
-
- #index = 0
- #for layer_in in layer_inputs:
- # print ("layer_in = ", layer_in.shape)
- # layer_name = model.layers[index].name
- # self.layer_input_sizes[layer_name] = layer_in.shape
- # index += 1
@@ -1323,7 +1301,6 @@ class PromiseRtTranslator:
self.findQuantizeRanges(model, x_test)
- #self.traverseSuccessors(root_node, state)
self.handleLayers(root_node, state)
print ("\n *** Generated PROMISE Layers **** \n ")
@@ -1335,4 +1312,4 @@ class PromiseRtTranslator:
- #print (self.layer_size_str)
+
diff --git a/llvm/projects/keras/frontend/utils.py b/llvm/projects/keras/frontend/utils.py
index ffc338c19ea60df7c53430ac38b613c2daef402e..9a2c13715ee351b61f8d7aaffc148a90a45ee233 100644
--- a/llvm/projects/keras/frontend/utils.py
+++ b/llvm/projects/keras/frontend/utils.py
@@ -1,6 +1,13 @@
+def DEBUG(str, *args):
+
+ debug = False
+ if debug:
+ print (str, *args)
+
+
def nodeHasBias(cur_node):
@@ -45,7 +52,6 @@ def genActivationCallStr(input_var, output_var, activation_type):
inst_str = "void* " + output_var + " = "
inst_str += "tensor" + func_name + "(" + input_var + "); \n"
- print ("***** inst_str = ", inst_str, "\n")
return inst_str
diff --git a/llvm/projects/keras/frontend/weight_utils.py b/llvm/projects/keras/frontend/weight_utils.py
index 69c1ee5469f317a4b5d8fd3e080ddae6bb8c83d9..dd22765386e2172572ad0feec201c7dec407a909 100644
--- a/llvm/projects/keras/frontend/weight_utils.py
+++ b/llvm/projects/keras/frontend/weight_utils.py
@@ -6,17 +6,16 @@ from keras.optimizers import Adam
def dumpLabels(file_name, Y_test):
+
+ print ("Dumping Labels File = ", file_name)
- f = open(file_name, "wb")
-
+ f = open(file_name, "wb")
labels_map = {}
for label in Y_test:
label_val = 0
if len(Y_test.shape) > 1:
- #label_val = np.int8(label[0])
label_val = np.int32(label[0])
else:
- #label_val = np.int8(label)
label_val = np.int32(label)
if label_val not in labels_map:
@@ -28,7 +27,7 @@ def dumpLabels(file_name, Y_test):
f.close()
-
+"""
def dumpData(file_name, X_test):
N = X_test.shape[0]
@@ -37,8 +36,8 @@ def dumpData(file_name, X_test):
W = X_test.shape[3]
print ("*DumpData")
- print("-min_val = ", np.amin(X_test))
- print("-max_val = ", np.amax(X_test))
+ #print("-min_val = ", np.amin(X_test))
+ #print("-max_val = ", np.amax(X_test))
f = open(file_name, "wb")
for i in range(N):
@@ -50,35 +49,34 @@ def dumpData(file_name, X_test):
f.close()
+"""
+
-"""
def dumpData(file_name, X_test):
- N = X_test.shape[0]
- C = X_test.shape[1]
- H = X_test.shape[2]
- W = X_test.shape[3]
-
- print ("*DumpData")
- print("-min_val = ", np.amin(X_test))
- print("-max_val = ", np.amax(X_test))
+ print ("*Dumping Input File = ", file_name)
+ #print("-min_val = ", np.amin(X_test))
+ #print("-max_val = ", np.amax(X_test))
f = open(file_name, "wb")
+ X_test = X_test.flatten()
+ X_test = X_test.astype(np.float32)
X_test.tofile(f)
f.close()
-"""
+
-
+"""
def dumpConvWeights(file_name, weights, N, C, H, W):
print (weights.shape)
print ("*DumpConvWeights")
- print("-min_val = ", np.amin(weights))
- print("-max_val = ", np.amax(weights))
+
+ #print("-min_val = ", np.amin(weights))
+ #print("-max_val = ", np.amax(weights))
f = open(file_name, "wb")
@@ -90,15 +88,33 @@ def dumpConvWeights(file_name, weights, N, C, H, W):
f.close()
+"""
+
+
+def dumpConvWeights(file_name, X_test, N, C, H, W):
+
+ print ("*Dumping Conv Weights to file = ", file_name)
+ #print("-min_val = ", np.amin(X_test))
+ #print("-max_val = ", np.amax(X_test))
+
+ f = open(file_name, "wb")
+
+ X_test = np.transpose(X_test, (3, 2, 0, 1))
+ X_test = X_test.flatten()
+ X_test = X_test.astype(np.float32)
+ X_test.tofile(f)
+ f.close()
+
+
+
def dumpFcWeights(file_name, weights, H, W):
- print (weights.shape)
- print ("*DumpFcWeights")
- print("-min_val = ", np.amin(weights))
- print("-max_val = ", np.amax(weights))
-
+ print ("*Dumping FC weights to = ", file_name)
+
+ #print("-min_val = ", np.amin(weights))
+ #print("-max_val = ", np.amax(weights))
f = open(file_name, "wb")
for i in range(H):
@@ -111,11 +127,7 @@ def dumpFcWeights(file_name, weights, H, W):
def dumpFcBias(file_name, bias, W):
- print (bias.shape)
- print ("*DumpFcBias")
- print("-min_val = ", np.amin(bias))
- print("-max_val = ", np.amax(bias))
-
+ print ("*Dump Bias Weights = ", file_name)
f = open(file_name, "wb")
for i in range(W):
@@ -159,19 +171,19 @@ def dumpCalibrationData2(file_name, test_data, labels_fname, test_labels):
# Loads Existing HPVM FP32 weights
-def dumpHPVMToKerasModel(model, reload_dir, output_model, X_test, Y_test):
+def reloadHPVMWeights(model, reload_dir, output_model, X_test, Y_test):
print ("***** Reloading pre-trained HPVM weights ****")
for i in range(len(model.layers)):
layer = model.layers[i]
layer_name = layer.name
- print ("*layer_name = ", layer_name)
+ #-- print ("*layer_name = ", layer_name)
if "conv" not in layer_name and "dense" not in layer_name:
continue
w_path = reload_dir + layer_name + "_w.bin"
- print ("** w_path = ", w_path)
+ #-- print ("** w_path = ", w_path)
w_arr = np.fromfile(w_path, dtype='float32')
b_path = reload_dir + layer_name + "_b.bin"
diff --git a/llvm/projects/keras/keras_environment.yml b/llvm/projects/keras/keras_environment.yml
index b0aa451b76f83618f69e8eb95ff86163b658a33a..1f56f758bef1f136f4fe71f04fe17ed0e770f7db 100644
--- a/llvm/projects/keras/keras_environment.yml
+++ b/llvm/projects/keras/keras_environment.yml
@@ -4,316 +4,29 @@ channels:
- conda-forge
- defaults
dependencies:
- - absl-py=0.6.1=py36_0
- - anaconda-project=0.8.2=py36_0
- - asn1crypto=0.24.0=py36_0
- - automat=0.7.0=py36_0
- - babel=2.6.0=py36_0
- - backports=1.0=py36_1
- - backports.os=0.1.1=py36_0
- - beautifulsoup4=4.6.3=py36_0
- - bkcharts=0.2=py36_0
- - blaze=0.11.3=py36_0
- - conda=4.5.11=py36_0
- - conda-env=2.6.0=1
- - contextlib2=0.5.5=py36_0
- - cycler=0.10.0=py36_0
- - dill=0.2.8.2=py36_0
- - docutils=0.14=py36_0
- - entrypoints=0.2.3=py36_2
- - et_xmlfile=1.0.1=py36_0
- - idna=2.7=py36_0
- - imageio=2.4.1=py36_0
- - importlib_metadata=0.6=py36_0
- - ipython_genutils=0.2.0=py36_0
- - isort=4.3.4=py36_0
- - jdcal=1.4=py36_0
- - jedi=0.13.1=py36_0
- - jinja2=2.10=py36_0
- - jmespath=0.9.3=py36_0
- - jsonschema=2.6.0=py36_0
- - keyring=16.0.0=py36_0
- - libgcc=7.2.0=h69d50b8_2
- - libgfortran=3.0.0=1
- - locket=0.2.0=py36_1
- - more-itertools=4.3.0=py36_0
- - nbconvert=5.3.1=py36_0
- - nbformat=4.4.0=py36_0
- - nose=1.3.7=py36_2
- - notebook=5.7.0=py36_0
- - numpydoc=0.8.0=py36_0
- - odo=0.5.1=py36_0
- - pathlib2=2.3.2=py36_0
- - pexpect=4.6.0=py36_0
- - pickleshare=0.7.5=py36_0
- - ply=3.11=py36_0
- - ptyprocess=0.6.0=py36_0
- - pycodestyle=2.4.0=py36_0
- - pygments=2.2.0=py36_0
- - pylint=2.1.1=py36_0
- - pyopenssl=18.0.0=py36_0
- - qtconsole=4.4.2=py36_0
- - requests=2.19.1=py36_0
- - s3transfer=0.1.13=py36_0
- - secretstorage=3.1.0=py36_0
- - setuptools=40.5.0=py36_0
- - singledispatch=3.4.0.3=py36_0
- - six=1.11.0=py36_1
- - snowballstemmer=1.2.1=py36_0
- - sortedcollections=1.0.1=py36_0
- - sphinx=1.8.1=py36_0
- - spyder=3.3.1=py36_1
- - sympy=1.3=py36_0
- - tblib=1.3.2=py36_0
- - termcolor=1.1.0=py36_1
- - terminado=0.8.1=py36_1
- - testpath=0.4.2=py36_0
- torchvision=0.2.1=py36_0
- - traitlets=4.3.2=py36_0
- - typing=3.6.4=py36_0
- - unicodecsv=0.14.1=py36_0
- - urllib3=1.23=py36_0
- - wcwidth=0.1.7=py36_0
- - wheel=0.32.2=py36_0
- - widgetsnbextension=3.4.2=py36_0
- - xlwt=1.3.0=py36_0
- - _license=1.1=py36_1
- _tflow_select=2.1.0=gpu
- - alabaster=0.7.12=py36_0
- - anaconda-client=1.7.2=py36_0
- - anaconda=custom=py36hbbc8b67_0
- - anaconda-navigator=1.9.2=py36_0
- - appdirs=1.4.3=py36h28b3542_0
- - astor=0.7.1=py36_0
- - astroid=2.0.4=py36_0
- - astropy=3.0.5=py36h7b6447c_0
- - atomicwrites=1.2.1=py36_0
- - attrs=18.2.0=py36h28b3542_0
- - backcall=0.1.0=py36_0
- - backports.shutil_get_terminal_size=1.0.0=py36_2
- - bitarray=0.8.3=py36h14c3975_0
- - blas=1.0=mkl
- - bleach=3.0.2=py36_0
- - blosc=1.14.4=hdbcaa40_0
- - bokeh=1.0.1=py36_0
- - boto=2.49.0=py36_0
- - boto3=1.9.35=py36_0
- - botocore=1.12.35=py36_0
- - bottleneck=1.2.1=py36h035aef0_1
- - bz2file=0.98=py36_1
- - bzip2=1.0.6=h14c3975_5
- - ca-certificates=2018.03.07=0
- - cairo=1.14.12=h8948797_3
- - certifi=2018.10.15=py36_0
- - cffi=1.11.5=py36he75722e_1
- - chardet=3.0.4=py36_1
- - chest=0.2.3=py36_1
- - click=7.0=py36_0
- - cloudpickle=0.6.1=py36_0
- - clyent=1.2.2=py36_1
- - colorama=0.4.0=py36_0
- - configobj=5.0.6=py36_1
- - constantly=15.1.0=py36h28b3542_0
- - cryptography=2.3.1=py36hc365091_0
- - cudatoolkit=9.0=h13b8566_0
- - cudnn=7.1.2=cuda9.0_0
- - cupti=9.0.176=0
- - curl=7.61.0=h84994c4_0
- - cython=0.29=py36he6710b0_0
- - cytoolz=0.9.0.1=py36h14c3975_1
- - dask=0.20.0=py36_0
- - dask-core=0.20.0=py36_0
- - datashape=0.5.4=py36_1
- - dbus=1.13.2=h714fa37_1
- - decorator=4.3.0=py36_0
- - defusedxml=0.5.0=py36_1
- - distributed=1.24.0=py36_0
- - expat=2.2.6=he6710b0_0
- - fastcache=1.0.2=py36h14c3975_2
- - filelock=3.0.10=py36_0
- - flask=1.0.2=py36_1
- - flask-cors=3.0.6=py36_0
- - fontconfig=2.13.0=h9420a91_0
- - freetype=2.9.1=h8a8886c_1
- - fribidi=1.0.5=h7b6447c_0
- - gast=0.2.0=py36_0
- - gensim=3.4.0=py36h14c3975_0
- - get_terminal_size=1.0.0=haa9412d_0
- - gevent=1.3.7=py36h7b6447c_1
- - glib=2.56.2=hd408876_0
- - glob2=0.6=py36_1
- - gmp=6.1.2=h6c8ec71_1
- - gmpy2=2.0.8=py36h10f8cd9_2
- - graphite2=1.3.12=h23475e2_2
- - greenlet=0.4.15=py36h7b6447c_0
- - grpcio=1.12.1=py36hdbcaa40_0
- - gst-plugins-base=1.14.0=hbbd80ab_1
- - gstreamer=1.14.0=hb453b48_1
- - h5py=2.8.0=py36h989c5e5_3
- - harfbuzz=1.8.8=hffaf4a1_0
- - hdf5=1.10.2=hba1933b_1
- - heapdict=1.0.0=py36_2
- - html5lib=1.0.1=py36_0
- - hyperlink=18.0.0=py36_0
- - icu=58.2=h9c2bf20_1
- - imagesize=1.1.0=py36_0
- - incremental=17.5.0=py36_0
- - ipykernel=5.1.0=py36h39e3cac_0
- - ipython=7.1.1=py36h39e3cac_0
- - ipywidgets=7.4.2=py36_0
- - itsdangerous=1.1.0=py36_0
- - jbig=2.1=hdba287a_0
- - jeepney=0.4=py36_0
- - jpeg=9b=h024ee3a_2
- keras=2.1.6=py36_0
- keras-applications=1.0.6=py36_0
- keras-preprocessing=1.0.5=py36_0
- - kiwisolver=1.0.1=py36hf484d3e_0
- - lazy-object-proxy=1.3.1=py36h14c3975_2
- - libcurl=7.61.0=h1ad7b7a_0
- - libedit=3.1.20170329=h6b74fdf_2
- - libffi=3.2.1=hd88cf55_4
- - libgcc-ng=8.2.0=hdf63c60_1
- - libgfortran-ng=7.3.0=hdf63c60_0
- - libiconv=1.15=h63c8f33_5
- - libpng=1.6.35=hbc83047_0
- - libprotobuf=3.6.1=hd408876_0
- - libsodium=1.0.16=h1bed415_0
- - libssh2=1.8.0=h9cfc8f7_4
- - libstdcxx-ng=8.2.0=hdf63c60_1
- - libtiff=4.0.9=he85c1e1_2
- - libtool=2.4.6=h7b6447c_5
- - libuuid=1.0.3=h1bed415_2
- - libxcb=1.13=h1bed415_1
- - libxml2=2.9.8=h26e45fe_1
- - libxslt=1.1.32=h1312cb7_0
- - llvmlite=0.25.0=py36hd408876_0
- - lxml=4.2.5=py36hefd8a0e_0
- - lzo=2.10=h49e0be7_2
- - markdown=3.0.1=py36_0
- - markupsafe=1.0=py36h14c3975_1
- - matplotlib=3.0.1=py36h5429711_0
- - mccabe=0.6.1=py36_1
- - mistune=0.8.4=py36h7b6447c_0
- mkl=2018.0.3=1
- mkl-service=1.1.2=py36h90e4bf4_5
- mkl_fft=1.0.6=py36h7dd41cf_0
- mkl_random=1.0.1=py36h4414c95_1
- - mpc=1.1.0=h10f8cd9_1
- - mpfr=4.0.1=hdf1c602_3
- - mpmath=1.0.0=py36_2
- - msgpack-python=0.5.6=py36h6bb024c_1
- - multipledispatch=0.6.0=py36_0
- - navigator-updater=0.2.1=py36_0
- - nccl=1.3.5=cuda9.0_0
- ncurses=6.1=hf484d3e_0
- - networkx=2.2=py36_1
- - ninja=1.8.2=py36h6bb024c_1
- - nltk=3.3.0=py36_0
- - numba=0.40.0=py36h962f231_0
- - numexpr=2.6.8=py36hd89afb7_0
- numpy=1.15.3=py36h1d66e8a_0
- numpy-base=1.15.3=py36h81de0dd_0
- - olefile=0.46=py36_0
- - openpyxl=2.5.9=py36_0
- - openssl=1.0.2p=h14c3975_0
- - packaging=18.0=py36_0
- pandas=0.23.4=py36h04863e7_0
- - pandoc=2.2.3.2=0
- - pandocfilters=1.4.2=py36_1
- - pango=1.42.4=h049681c_0
- - parso=0.3.1=py36_0
- - partd=0.3.9=py36_0
- - patchelf=0.9=he6710b0_3
- - path.py=11.5.0=py36_0
- - patsy=0.5.1=py36_0
- - pcre=8.42=h439df22_0
- - pep8=1.7.1=py36_0
- - pillow=5.3.0=py36h34e0f95_0
- pip=18.1=py36_0
- - pixman=0.34.0=hceecf20_3
- - pkginfo=1.4.2=py36_1
- - pluggy=0.8.0=py36_0
- - prometheus_client=0.4.2=py36_0
- - prompt_toolkit=2.0.7=py36_0
- - protobuf=3.6.1=py36he6710b0_0
- psutil=5.4.8=py36h7b6447c_0
- py=1.7.0=py36_0
- - pyasn1=0.4.4=py36h28b3542_0
- - pyasn1-modules=0.2.2=py36_0
- - pycosat=0.6.3=py36h14c3975_0
- - pycparser=2.19=py36_0
- - pycrypto=2.6.1=py36h14c3975_9
- - pycurl=7.43.0.2=py36hb7f436b_0
- - pyflakes=2.0.0=py36_0
- - pyhamcrest=1.9.0=py36_2
- - pyodbc=4.0.24=py36he6710b0_0
- - pyparsing=2.2.2=py36_0
- - pyqt=5.9.2=py36h05f1152_2
- - pysocks=1.6.8=py36_0
- - pytables=3.4.4=py36ha205bf6_0
- - pytest=3.9.3=py36_0
- - pytest-arraydiff=0.2=py36h39e3cac_0
- - pytest-astropy=0.4.0=py36_0
- - pytest-doctestplus=0.1.3=py36_0
- - pytest-openfiles=0.3.0=py36_0
- - pytest-remotedata=0.3.1=py36_0
- python=3.6.6=h6e4f718_2
- - python-dateutil=2.7.5=py36_0
- pytorch=0.4.1=py36ha74772b_0
- - pytz=2018.7=py36_0
- - pywavelets=1.0.1=py36hdd07704_0
- - pyyaml=3.13=py36h14c3975_0
- - pyzmq=17.1.2=py36h14c3975_0
- - qt=5.9.6=h8703b6f_2
- - qtawesome=0.5.2=py36_0
- - qtpy=1.5.2=py36_0
- - readline=7.0=h7b6447c_5
- - redis=5.0.0=h7b6447c_0
- - redis-py=2.10.6=py36_0
- - rope=0.11.0=py36_0
- - ruamel_yaml=0.15.46=py36h14c3975_0
- - scikit-image=0.14.0=py36hf484d3e_1
- - scikit-learn=0.20.0=py36h4989274_1
- scipy=1.1.0=py36hfa4b5c9_1
- - seaborn=0.9.0=py36_0
- - send2trash=1.5.0=py36_0
- - service_identity=17.0.0=py36h28b3542_0
- - simplegeneric=0.8.1=py36_2
- - sip=4.19.8=py36hf484d3e_0
- - smart_open=1.7.1=py36_0
- - snappy=1.1.7=hbae5bb6_3
- - sockjs-tornado=1.0.6=py36_0
- - sortedcontainers=2.0.5=py36_0
- - sphinxcontrib=1.0=py36_1
- - sphinxcontrib-websupport=1.1.0=py36_1
- - spyder-kernels=0.2.6=py36_0
- - sqlalchemy=1.2.12=py36h7b6447c_0
- - sqlite=3.25.2=h7b6447c_0
- - statsmodels=0.9.0=py36h035aef0_0
- - tensorboard=1.11.0=py36hf484d3e_0
- - tensorflow=1.11.0=gpu_py36h4459f94_0
- - tensorflow-base=1.11.0=gpu_py36h8e0ae2d_0
- - tensorflow-gpu=1.11.0=h0d30ee6_0
- - tk=8.6.8=hbc83047_0
- - toolz=0.9.0=py36_0
- - tornado=5.1.1=py36h7b6447c_0
- - tqdm=4.28.1=py36h28b3542_0
- - twisted=18.9.0=py36h7b6447c_0
- - typed-ast=1.1.0=py36h14c3975_0
- - unixodbc=2.3.7=h14c3975_0
- - webencodings=0.5.1=py36_1
- - werkzeug=0.14.1=py36_0
- - wrapt=1.10.11=py36h14c3975_2
- - xlrd=1.1.0=py36_1
- - xlsxwriter=1.1.2=py36_0
- - xz=5.2.4=h14c3975_4
- - yaml=0.1.7=had09818_2
- - zeromq=4.2.5=hf484d3e_1
- - zict=0.1.3=py36_0
- - zlib=1.2.11=ha838bed_2
- - zope=1.0=py36_1
- - zope.interface=4.6.0=py36h7b6447c_0
+ - tensorboard=1.14
+ - tensorflow=1.14
+ - tensorflow-base=1.14
+ - tensorflow-gpu=1.14
- cuda91=1.0=h4c16780_0
- pip:
- msgpack==0.5.6
diff --git a/llvm/projects/keras/keras_environment_minimal.yml b/llvm/projects/keras/keras_environment_minimal.yml
deleted file mode 100644
index 1f56f758bef1f136f4fe71f04fe17ed0e770f7db..0000000000000000000000000000000000000000
--- a/llvm/projects/keras/keras_environment_minimal.yml
+++ /dev/null
@@ -1,35 +0,0 @@
-name: approxhpvm_keras
-channels:
- - pytorch
- - conda-forge
- - defaults
-dependencies:
- - torchvision=0.2.1=py36_0
- - _tflow_select=2.1.0=gpu
- - keras=2.1.6=py36_0
- - keras-applications=1.0.6=py36_0
- - keras-preprocessing=1.0.5=py36_0
- - mkl=2018.0.3=1
- - mkl-service=1.1.2=py36h90e4bf4_5
- - mkl_fft=1.0.6=py36h7dd41cf_0
- - mkl_random=1.0.1=py36h4414c95_1
- - ncurses=6.1=hf484d3e_0
- - numpy=1.15.3=py36h1d66e8a_0
- - numpy-base=1.15.3=py36h81de0dd_0
- - pandas=0.23.4=py36h04863e7_0
- - pip=18.1=py36_0
- - psutil=5.4.8=py36h7b6447c_0
- - py=1.7.0=py36_0
- - python=3.6.6=h6e4f718_2
- - pytorch=0.4.1=py36ha74772b_0
- - scipy=1.1.0=py36hfa4b5c9_1
- - tensorboard=1.14
- - tensorflow=1.14
- - tensorflow-base=1.14
- - tensorflow-gpu=1.14
- - cuda91=1.0=h4c16780_0
- - pip:
- - msgpack==0.5.6
- - tables==3.4.4
- - torch==0.4.1
-
diff --git a/llvm/projects/keras/generated/vgg16/approxhpvm_src.cc b/llvm/projects/keras/legacy/generated/vgg16/approxhpvm_src.cc
similarity index 100%
rename from llvm/projects/keras/generated/vgg16/approxhpvm_src.cc
rename to llvm/projects/keras/legacy/generated/vgg16/approxhpvm_src.cc
diff --git a/llvm/projects/keras/generated/vgg16/src.cc b/llvm/projects/keras/legacy/generated/vgg16/src.cc
similarity index 100%
rename from llvm/projects/keras/generated/vgg16/src.cc
rename to llvm/projects/keras/legacy/generated/vgg16/src.cc
diff --git a/llvm/projects/keras/legacy/keras_environment.yml b/llvm/projects/keras/legacy/keras_environment.yml
new file mode 100644
index 0000000000000000000000000000000000000000..caa3a773dfd8c7a82571a923bb1941997ba59ca9
--- /dev/null
+++ b/llvm/projects/keras/legacy/keras_environment.yml
@@ -0,0 +1,321 @@
+name: approxhpvm_keras
+channels:
+ - pytorch
+ - conda-forge
+ - defaults
+dependencies:
+ - absl-py=0.6.1=py36_0
+ - anaconda-project=0.8.2=py36_0
+ - asn1crypto=0.24.0=py36_0
+ - automat=0.7.0=py36_0
+ - babel=2.6.0=py36_0
+ - backports=1.0=py36_1
+ - backports.os=0.1.1=py36_0
+ - beautifulsoup4=4.6.3=py36_0
+ - bkcharts=0.2=py36_0
+ - blaze=0.11.3=py36_0
+ - conda=4.5.11=py36_0
+ - conda-env=2.6.0=1
+ - contextlib2=0.5.5=py36_0
+ - cycler=0.10.0=py36_0
+ - dill=0.2.8.2=py36_0
+ - docutils=0.14=py36_0
+ - entrypoints=0.2.3=py36_2
+ - et_xmlfile=1.0.1=py36_0
+ - idna=2.7=py36_0
+ - imageio=2.4.1=py36_0
+ - importlib_metadata=0.6=py36_0
+ - ipython_genutils=0.2.0=py36_0
+ - isort=4.3.4=py36_0
+ - jdcal=1.4=py36_0
+ - jedi=0.13.1=py36_0
+ - jinja2=2.10=py36_0
+ - jmespath=0.9.3=py36_0
+ - jsonschema=2.6.0=py36_0
+ - keyring=16.0.0=py36_0
+ - libgcc=7.2.0=h69d50b8_2
+ - libgfortran=3.0.0=1
+ - locket=0.2.0=py36_1
+ - more-itertools=4.3.0=py36_0
+ - nbconvert=5.3.1=py36_0
+ - nbformat=4.4.0=py36_0
+ - nose=1.3.7=py36_2
+ - notebook=5.7.0=py36_0
+ - numpydoc=0.8.0=py36_0
+ - odo=0.5.1=py36_0
+ - pathlib2=2.3.2=py36_0
+ - pexpect=4.6.0=py36_0
+ - pickleshare=0.7.5=py36_0
+ - ply=3.11=py36_0
+ - ptyprocess=0.6.0=py36_0
+ - pycodestyle=2.4.0=py36_0
+ - pygments=2.2.0=py36_0
+ - pylint=2.1.1=py36_0
+ - pyopenssl=18.0.0=py36_0
+ - qtconsole=4.4.2=py36_0
+ - requests=2.19.1=py36_0
+ - s3transfer=0.1.13=py36_0
+ - secretstorage=3.1.0=py36_0
+ - setuptools=40.5.0=py36_0
+ - singledispatch=3.4.0.3=py36_0
+ - six=1.11.0=py36_1
+ - snowballstemmer=1.2.1=py36_0
+ - sortedcollections=1.0.1=py36_0
+ - sphinx=1.8.1=py36_0
+ - spyder=3.3.1=py36_1
+ - sympy=1.3=py36_0
+ - tblib=1.3.2=py36_0
+ - termcolor=1.1.0=py36_1
+ - terminado=0.8.1=py36_1
+ - testpath=0.4.2=py36_0
+ - torchvision=0.2.1=py36_0
+ - traitlets=4.3.2=py36_0
+ - typing=3.6.4=py36_0
+ - unicodecsv=0.14.1=py36_0
+ - urllib3=1.23=py36_0
+ - wcwidth=0.1.7=py36_0
+ - wheel=0.32.2=py36_0
+ - widgetsnbextension=3.4.2=py36_0
+ - xlwt=1.3.0=py36_0
+ - _tflow_select=2.1.0=gpu
+ - alabaster=0.7.12=py36_0
+ - anaconda-client=1.7.2=py36_0
+ - anaconda=custom=py36hbbc8b67_0
+ - anaconda-navigator=1.9.2=py36_0
+ - appdirs=1.4.3=py36h28b3542_0
+ - astor=0.7.1=py36_0
+ - astroid=2.0.4=py36_0
+ - astropy=3.0.5=py36h7b6447c_0
+ - atomicwrites=1.2.1=py36_0
+ - attrs=18.2.0=py36h28b3542_0
+ - backcall=0.1.0=py36_0
+ - backports.shutil_get_terminal_size=1.0.0=py36_2
+ - bitarray=0.8.3=py36h14c3975_0
+ - blas=1.0=mkl
+ - bleach=3.0.2=py36_0
+ - blosc=1.14.4=hdbcaa40_0
+ - bokeh=1.0.1=py36_0
+ - boto=2.49.0=py36_0
+ - boto3=1.9.35=py36_0
+ - botocore=1.12.35=py36_0
+ - bottleneck=1.2.1=py36h035aef0_1
+ - bz2file=0.98=py36_1
+ - bzip2=1.0.6=h14c3975_5
+ - ca-certificates=2018.03.07=0
+ - cairo=1.14.12=h8948797_3
+ - certifi=2018.10.15=py36_0
+ - cffi=1.11.5=py36he75722e_1
+ - chardet=3.0.4=py36_1
+ - chest=0.2.3=py36_1
+ - click=7.0=py36_0
+ - cloudpickle=0.6.1=py36_0
+ - clyent=1.2.2=py36_1
+ - colorama=0.4.0=py36_0
+ - configobj=5.0.6=py36_1
+ - constantly=15.1.0=py36h28b3542_0
+ - cryptography=2.3.1=py36hc365091_0
+ - cudatoolkit=9.0=h13b8566_0
+ - cudnn=7.1.2=cuda9.0_0
+ - cupti=9.0.176=0
+ - curl=7.61.0=h84994c4_0
+ - cython=0.29=py36he6710b0_0
+ - cytoolz=0.9.0.1=py36h14c3975_1
+ - dask=0.20.0=py36_0
+ - dask-core=0.20.0=py36_0
+ - datashape=0.5.4=py36_1
+ - dbus=1.13.2=h714fa37_1
+ - decorator=4.3.0=py36_0
+ - defusedxml=0.5.0=py36_1
+ - distributed=1.24.0=py36_0
+ - expat=2.2.6=he6710b0_0
+ - fastcache=1.0.2=py36h14c3975_2
+ - filelock=3.0.10=py36_0
+ - flask=1.0.2=py36_1
+ - flask-cors=3.0.6=py36_0
+ - fontconfig=2.13.0=h9420a91_0
+ - freetype=2.9.1=h8a8886c_1
+ - fribidi=1.0.5=h7b6447c_0
+ - gast=0.2.0=py36_0
+ - gensim=3.4.0=py36h14c3975_0
+ - get_terminal_size=1.0.0=haa9412d_0
+ - gevent=1.3.7=py36h7b6447c_1
+ - glib=2.56.2=hd408876_0
+ - glob2=0.6=py36_1
+ - gmp=6.1.2=h6c8ec71_1
+ - gmpy2=2.0.8=py36h10f8cd9_2
+ - graphite2=1.3.12=h23475e2_2
+ - greenlet=0.4.15=py36h7b6447c_0
+ - grpcio=1.12.1=py36hdbcaa40_0
+ - gst-plugins-base=1.14.0=hbbd80ab_1
+ - gstreamer=1.14.0=hb453b48_1
+ - h5py=2.8.0=py36h989c5e5_3
+ - harfbuzz=1.8.8=hffaf4a1_0
+ - hdf5=1.10.2=hba1933b_1
+ - heapdict=1.0.0=py36_2
+ - html5lib=1.0.1=py36_0
+ - hyperlink=18.0.0=py36_0
+ - icu=58.2=h9c2bf20_1
+ - imagesize=1.1.0=py36_0
+ - incremental=17.5.0=py36_0
+ - ipykernel=5.1.0=py36h39e3cac_0
+ - ipython=7.1.1=py36h39e3cac_0
+ - ipywidgets=7.4.2=py36_0
+ - itsdangerous=1.1.0=py36_0
+ - jbig=2.1=hdba287a_0
+ - jeepney=0.4=py36_0
+ - jpeg=9b=h024ee3a_2
+ - keras=2.1.6=py36_0
+ - keras-applications=1.0.6=py36_0
+ - keras-preprocessing=1.0.5=py36_0
+ - kiwisolver=1.0.1=py36hf484d3e_0
+ - lazy-object-proxy=1.3.1=py36h14c3975_2
+ - libcurl=7.61.0=h1ad7b7a_0
+ - libedit=3.1.20170329=h6b74fdf_2
+ - libffi=3.2.1=hd88cf55_4
+ - libgcc-ng=8.2.0=hdf63c60_1
+ - libgfortran-ng=7.3.0=hdf63c60_0
+ - libiconv=1.15=h63c8f33_5
+ - libpng=1.6.35=hbc83047_0
+ - libprotobuf=3.6.1=hd408876_0
+ - libsodium=1.0.16=h1bed415_0
+ - libssh2=1.8.0=h9cfc8f7_4
+ - libstdcxx-ng=8.2.0=hdf63c60_1
+ - libtiff=4.0.9=he85c1e1_2
+ - libtool=2.4.6=h7b6447c_5
+ - libuuid=1.0.3=h1bed415_2
+ - libxcb=1.13=h1bed415_1
+ - libxml2=2.9.8=h26e45fe_1
+ - libxslt=1.1.32=h1312cb7_0
+ - llvmlite=0.25.0=py36hd408876_0
+ - lxml=4.2.5=py36hefd8a0e_0
+ - lzo=2.10=h49e0be7_2
+ - markdown=3.0.1=py36_0
+ - markupsafe=1.0=py36h14c3975_1
+ - matplotlib=3.0.1=py36h5429711_0
+ - mccabe=0.6.1=py36_1
+ - mistune=0.8.4=py36h7b6447c_0
+ - mkl=2018.0.3=1
+ - mkl-service=1.1.2=py36h90e4bf4_5
+ - mkl_fft=1.0.6=py36h7dd41cf_0
+ - mkl_random=1.0.1=py36h4414c95_1
+ - mpc=1.1.0=h10f8cd9_1
+ - mpfr=4.0.1=hdf1c602_3
+ - mpmath=1.0.0=py36_2
+ - msgpack-python=0.5.6=py36h6bb024c_1
+ - multipledispatch=0.6.0=py36_0
+ - navigator-updater=0.2.1=py36_0
+ - nccl=1.3.5=cuda9.0_0
+ - ncurses=6.1=hf484d3e_0
+ - networkx=2.2=py36_1
+ - ninja=1.8.2=py36h6bb024c_1
+ - nltk=3.3.0=py36_0
+ - numba=0.40.0=py36h962f231_0
+ - numexpr=2.6.8=py36hd89afb7_0
+ - numpy=1.15.3=py36h1d66e8a_0
+ - numpy-base=1.15.3=py36h81de0dd_0
+ - olefile=0.46=py36_0
+ - openpyxl=2.5.9=py36_0
+ - openssl=1.0.2p=h14c3975_0
+ - packaging=18.0=py36_0
+ - pandas=0.23.4=py36h04863e7_0
+ - pandoc=2.2.3.2=0
+ - pandocfilters=1.4.2=py36_1
+ - pango=1.42.4=h049681c_0
+ - parso=0.3.1=py36_0
+ - partd=0.3.9=py36_0
+ - patchelf=0.9=he6710b0_3
+ - path.py=11.5.0=py36_0
+ - patsy=0.5.1=py36_0
+ - pcre=8.42=h439df22_0
+ - pep8=1.7.1=py36_0
+ - pillow=5.3.0=py36h34e0f95_0
+ - pip=18.1=py36_0
+ - pixman=0.34.0=hceecf20_3
+ - pkginfo=1.4.2=py36_1
+ - pluggy=0.8.0=py36_0
+ - prometheus_client=0.4.2=py36_0
+ - prompt_toolkit=2.0.7=py36_0
+ - protobuf=3.6.1=py36he6710b0_0
+ - psutil=5.4.8=py36h7b6447c_0
+ - py=1.7.0=py36_0
+ - pyasn1=0.4.4=py36h28b3542_0
+ - pyasn1-modules=0.2.2=py36_0
+ - pycosat=0.6.3=py36h14c3975_0
+ - pycparser=2.19=py36_0
+ - pycrypto=2.6.1=py36h14c3975_9
+ - pycurl=7.43.0.2=py36hb7f436b_0
+ - pyflakes=2.0.0=py36_0
+ - pyhamcrest=1.9.0=py36_2
+ - pyodbc=4.0.24=py36he6710b0_0
+ - pyparsing=2.2.2=py36_0
+ - pyqt=5.9.2=py36h05f1152_2
+ - pysocks=1.6.8=py36_0
+ - pytables=3.4.4=py36ha205bf6_0
+ - pytest=3.9.3=py36_0
+ - pytest-arraydiff=0.2=py36h39e3cac_0
+ - pytest-astropy=0.4.0=py36_0
+ - pytest-doctestplus=0.1.3=py36_0
+ - pytest-openfiles=0.3.0=py36_0
+ - pytest-remotedata=0.3.1=py36_0
+ - python=3.6.6=h6e4f718_2
+ - python-dateutil=2.7.5=py36_0
+ - pytorch=0.4.1=py36ha74772b_0
+ - pytz=2018.7=py36_0
+ - pywavelets=1.0.1=py36hdd07704_0
+ - pyyaml=3.13=py36h14c3975_0
+ - pyzmq=17.1.2=py36h14c3975_0
+ - qt=5.9.6=h8703b6f_2
+ - qtawesome=0.5.2=py36_0
+ - qtpy=1.5.2=py36_0
+ - readline=7.0=h7b6447c_5
+ - redis=5.0.0=h7b6447c_0
+ - redis-py=2.10.6=py36_0
+ - rope=0.11.0=py36_0
+ - ruamel_yaml=0.15.46=py36h14c3975_0
+ - scikit-image=0.14.0=py36hf484d3e_1
+ - scikit-learn=0.20.0=py36h4989274_1
+ - scipy=1.1.0=py36hfa4b5c9_1
+ - seaborn=0.9.0=py36_0
+ - send2trash=1.5.0=py36_0
+ - service_identity=17.0.0=py36h28b3542_0
+ - simplegeneric=0.8.1=py36_2
+ - sip=4.19.8=py36hf484d3e_0
+ - smart_open=1.7.1=py36_0
+ - snappy=1.1.7=hbae5bb6_3
+ - sockjs-tornado=1.0.6=py36_0
+ - sortedcontainers=2.0.5=py36_0
+ - sphinxcontrib=1.0=py36_1
+ - sphinxcontrib-websupport=1.1.0=py36_1
+ - spyder-kernels=0.2.6=py36_0
+ - sqlalchemy=1.2.12=py36h7b6447c_0
+ - sqlite=3.25.2=h7b6447c_0
+ - statsmodels=0.9.0=py36h035aef0_0
+ - tensorboard=1.11.0=py36hf484d3e_0
+ - tensorflow=1.11.0=gpu_py36h4459f94_0
+ - tensorflow-base=1.11.0=gpu_py36h8e0ae2d_0
+ - tensorflow-gpu=1.11.0=h0d30ee6_0
+ - tk=8.6.8=hbc83047_0
+ - toolz=0.9.0=py36_0
+ - tornado=5.1.1=py36h7b6447c_0
+ - tqdm=4.28.1=py36h28b3542_0
+ - twisted=18.9.0=py36h7b6447c_0
+ - typed-ast=1.1.0=py36h14c3975_0
+ - unixodbc=2.3.7=h14c3975_0
+ - webencodings=0.5.1=py36_1
+ - werkzeug=0.14.1=py36_0
+ - wrapt=1.10.11=py36h14c3975_2
+ - xlrd=1.1.0=py36_1
+ - xlsxwriter=1.1.2=py36_0
+ - xz=5.2.4=h14c3975_4
+ - yaml=0.1.7=had09818_2
+ - zeromq=4.2.5=hf484d3e_1
+ - zict=0.1.3=py36_0
+ - zlib=1.2.11=ha838bed_2
+ - zope=1.0=py36_1
+ - zope.interface=4.6.0=py36h7b6447c_0
+ - cuda91=1.0=h4c16780_0
+ - pip:
+ - msgpack==0.5.6
+ - tables==3.4.4
+ - torch==0.4.1
+
diff --git a/llvm/projects/keras/keras_environment_deps.yml b/llvm/projects/keras/legacy/keras_environment_deps.yml
similarity index 100%
rename from llvm/projects/keras/keras_environment_deps.yml
rename to llvm/projects/keras/legacy/keras_environment_deps.yml
diff --git a/llvm/projects/keras/src/Benchmark.py b/llvm/projects/keras/src/Benchmark.py
index fc3b483fbc9b34dc99455ba6e61fcc7eea79f24a..e83a78f4ea5d776a21ed6a6b47ccdb840f42c129 100644
--- a/llvm/projects/keras/src/Benchmark.py
+++ b/llvm/projects/keras/src/Benchmark.py
@@ -7,7 +7,7 @@ from keras.utils.np_utils import to_categorical
from keras.models import load_model
from frontend.approxhpvm_translator import translate_to_approxhpvm
from frontend.weight_utils import dumpCalibrationData
-from frontend.weight_utils import dumpHPVMToKerasModel
+from frontend.weight_utils import reloadHPVMWeights
# Every CNN Benchmark must inherit from Benchmark class
@@ -47,14 +47,19 @@ class Benchmark:
# Cmake ../
# make
-
- def run(self, argv):
- if len(argv) < 2:
- print ("Usage: python ${benchmark.py} [hpvm_reload|keras_reload|train] [frontend] [compile]")
+ def printUsage(self):
+
+ print ("Usage: python ${benchmark.py} [hpvm_reload|keras_reload|train] [frontend] [compile]")
sys.exit(0)
+
+ def run(self, argv):
+
+ if len(argv) < 2:
+ self.printUsage()
+
# Virtual method call implemented by each CNN
model = self.buildModel()
@@ -62,14 +67,19 @@ class Benchmark:
X_train, Y_train, X_test, Y_test = self.data_preprocess()
if argv[1] == "hpvm_reload":
- model = dumpHPVMToKerasModel(model, self.reload_dir, self.keras_model_file, X_test, Y_test)
+ print ("loading weights .....\n\n")
+ model = reloadHPVMWeights(model, self.reload_dir, self.keras_model_file, X_test, Y_test)
- if argv[1] == "keras_reload":
+ elif argv[1] == "keras_reload":
model = load_model(self.keras_model_file)
- if argv[1] == "train":
+ elif argv[1] == "train":
model = self.trainModel(model)
+ else:
+ self.printUsage()
+
+
score = model.evaluate(X_test, to_categorical(Y_test, self.num_classes), verbose=0)
print('Test accuracy2:', score[1])
@@ -81,10 +91,15 @@ class Benchmark:
# Main call to ApproxHPVM-Keras Frontend
working_dir = translate_to_approxhpvm(model, self.hpvm_dir, X_test, Y_test, self.num_classes)
- print ("*** working_dir = ", working_dir)
+ #-- print ("*** working_dir = ", working_dir)
if len(argv) > 3 and argv[3] == "compile":
self.compileSource(working_dir)
-
+ else:
+ self.printUsage()
+
+ elif len(argv) > 2:
+ self.printUsage()
+
diff --git a/llvm/projects/keras/src/alexnet.py b/llvm/projects/keras/src/alexnet.py
index d7848a2b6388e506583772caf9f39efd787df27d..9bfe80a156ae21e9befea3a6872b63758e37e2a3 100644
--- a/llvm/projects/keras/src/alexnet.py
+++ b/llvm/projects/keras/src/alexnet.py
@@ -19,9 +19,7 @@ import keras
import numpy as np
import os
from Benchmark import Benchmark
-from frontend.approxhpvm_translator import translate_to_approxhpvm
-from frontend.weight_utils import dumpCalibrationData
-from frontend.weight_utils import dumpHPVMToKerasModel
+
@@ -46,6 +44,8 @@ class AlexNet(Benchmark):
def buildModel(self):
+ print ("BuildModel ...")
+
activation_type = "tanh"
weight_decay = 1e-4
@@ -138,16 +138,20 @@ class AlexNet(Benchmark):
def data_preprocess(self):
+ print ("Data Preprocess... \n")
+
(X_train, Y_train), (X_test, Y_test) = cifar10.load_data()
+ print ("Data Loaded... \n")
+
X_train = X_train / 255.0
X_test = X_test / 255.0
-
+
mean = np.mean(X_train,axis=(0,1,2,3))
std = np.std(X_train,axis=(0,1,2,3))
X_train = (X_train-mean)/(std+1e-7)
X_test = (X_test-mean)/(std+1e-7)
-
+
return X_train, Y_train, X_test, Y_test
diff --git a/llvm/projects/keras/src/alexnet_imagenet.py b/llvm/projects/keras/src/alexnet_imagenet.py
index 0e8425b8259a7361064fb189c77af0dc389cda66..41cf2d837cf8ba764b614512f40a0f6c0522ab3d 100644
--- a/llvm/projects/keras/src/alexnet_imagenet.py
+++ b/llvm/projects/keras/src/alexnet_imagenet.py
@@ -30,10 +30,10 @@ data_format = 'channels_first'
IMAGENET_DIR = '/home/nz11/ILSVRC2012/'
OUTPUT_DIR = 'data/alexnet_imagenet_tune/'
-WEIGHTS_PATH = 'data/weights.h5'
+WEIGHTS_PATH = 'data/alexnet_imagenet_tune/weights.h5'
NUM_CLASSES = 200
-IMAGES_PER_CLASS = 40
+IMAGES_PER_CLASS = 50
# VAL_SIZE = 100
@@ -183,6 +183,9 @@ y_true = np.array(y_true)
X_tune = np.array(X_tune)
y_tune = np.array(y_tune)
+print ('tune size', len(X_tune))
+print ('test size', len(X_test))
+
@@ -233,16 +236,16 @@ model.compile(optimizer=keras.optimizers.Adam(lr=0.00001), loss='categorical_cro
if os.path.exists(WEIGHTS_PATH):
model.load_weights(WEIGHTS_PATH)
else:
- model.fit_generator(generate(), steps_per_epoch=1000, validation_data=(X_test, to_categorical(y_true, num_classes=1000)), epochs=2)
- K.set_value(model.optimizer.lr, 0.000001)
- model.fit_generator(generate(), steps_per_epoch=1000, validation_data=(X_test, to_categorical(y_true, num_classes=1000)), epochs=6)
- model.save_weights('data/weights.h5')
+ pass
+# model.fit_generator(generate(), steps_per_epoch=1000, validation_data=(X_test, to_categorical(y_true, num_classes=1000)), epochs=3)
+# K.set_value(model.optimizer.lr, 0.000001)
+# model.fit_generator(generate(), steps_per_epoch=1000, validation_data=(X_test, to_categorical(y_true, num_classes=1000)), epochs=3)
translate_to_approxhpvm(model, OUTPUT_DIR, X_tune, y_tune, 1000)
-# dumpCalibrationData2(OUTPUT_DIR + 'test_input_10K.bin', X_test, OUTPUT_DIR + 'test_labels_10K.bin', y_true)
-dumpCalibrationData2(OUTPUT_DIR + 'tune_input.bin', X_tune, OUTPUT_DIR + 'tune_labels.bin', y_tune)
-dumpCalibrationData2(OUTPUT_DIR + 'test_input.bin', X_test, OUTPUT_DIR + 'test_labels.bin', y_true)
+# # dumpCalibrationData2(OUTPUT_DIR + 'test_input_10K.bin', X_test, OUTPUT_DIR + 'test_labels_10K.bin', y_true)
+# dumpCalibrationData2(OUTPUT_DIR + 'tune_input.bin', X_tune, OUTPUT_DIR + 'tune_labels.bin', y_tune)
+# dumpCalibrationData2(OUTPUT_DIR + 'test_input.bin', X_test, OUTPUT_DIR + 'test_labels.bin', y_true)
pred = np.argmax(model.predict(X_test), axis=1)
@@ -250,4 +253,7 @@ print ('val accuracy', np.sum(pred == y_true.ravel()) / len(X_test))
pred = np.argmax(model.predict(X_tune), axis=1)
print ('val accuracy', np.sum(pred == y_tune.ravel()) / len(X_tune))
+
+model.save_weights(OUTPUT_DIR + '/weights.h5')
+
\ No newline at end of file
diff --git a/llvm/projects/keras/src/lenet_conv_test.py b/llvm/projects/keras/src/legacy/lenet_conv_test.py
similarity index 100%
rename from llvm/projects/keras/src/lenet_conv_test.py
rename to llvm/projects/keras/src/legacy/lenet_conv_test.py
diff --git a/llvm/projects/keras/src/mobilenet_imagenet.py b/llvm/projects/keras/src/legacy/mobilenet_imagenet.py
similarity index 100%
rename from llvm/projects/keras/src/mobilenet_imagenet.py
rename to llvm/projects/keras/src/legacy/mobilenet_imagenet.py
diff --git a/llvm/projects/keras/src/mobilenet_shallow.py b/llvm/projects/keras/src/legacy/mobilenet_shallow.py
similarity index 100%
rename from llvm/projects/keras/src/mobilenet_shallow.py
rename to llvm/projects/keras/src/legacy/mobilenet_shallow.py
diff --git a/llvm/projects/keras/src/mobilenetv2_cifar10.py b/llvm/projects/keras/src/legacy/mobilenetv2_cifar10.py
similarity index 100%
rename from llvm/projects/keras/src/mobilenetv2_cifar10.py
rename to llvm/projects/keras/src/legacy/mobilenetv2_cifar10.py
diff --git a/llvm/projects/keras/src/resnet.py b/llvm/projects/keras/src/resnet18_cifar10.py
similarity index 100%
rename from llvm/projects/keras/src/resnet.py
rename to llvm/projects/keras/src/resnet18_cifar10.py
diff --git a/llvm/projects/keras/src/resnet50_imagenet.py b/llvm/projects/keras/src/resnet50_imagenet.py
index b4d9925a085e6e510c4d2707f9564cfe28a53765..55d0918b7a526ba1a9866d2d8e3b0e2e8608cc25 100644
--- a/llvm/projects/keras/src/resnet50_imagenet.py
+++ b/llvm/projects/keras/src/resnet50_imagenet.py
@@ -29,7 +29,7 @@ data_format = 'channels_first'
IMAGENET_DIR = '/home/nz11/ILSVRC2012/'
-OUTPUT_DIR = 'data/resnet50_imagenet_tune/'
+OUTPUT_DIR = 'data/resnet50_imagenet_tune_regenerate/'
WEIGHTS_PATH = 'data/resnet50_imagenet/weights.h5'
NUM_CLASSES = 200
@@ -217,6 +217,9 @@ y_true = np.array(y_true)
X_tune = np.array(X_tune)
y_tune = np.array(y_tune)
+print ('tune size', len(X_tune))
+print ('test size', len(X_test))
+
@@ -267,19 +270,20 @@ model.compile(optimizer=keras.optimizers.Adam(lr=0.00001), loss='categorical_cro
if os.path.exists(WEIGHTS_PATH):
model.load_weights(WEIGHTS_PATH)
else:
- model.fit_generator(generate(), steps_per_epoch=1000, validation_data=(X_test, to_categorical(y_true, num_classes=1000)), epochs=6)
- model.save_weights(OUTPUT_DIR + 'weights.h5')
+ pass
+# model.fit_generator(generate(), steps_per_epoch=1000, validation_data=(X_test, to_categorical(y_true, num_classes=1000)), epochs=6)
+# model.save_weights(OUTPUT_DIR + 'weights.h5')
translate_to_approxhpvm(model, OUTPUT_DIR, X_tune, y_tune, 1000)
-# dumpCalibrationData2(OUTPUT_DIR + 'test_input_10K.bin', X_test, OUTPUT_DIR + 'test_labels_10K.bin', y_true)
-dumpCalibrationData2(OUTPUT_DIR + 'tune_input.bin', X_tune, OUTPUT_DIR + 'tune_labels.bin', y_tune)
-dumpCalibrationData2(OUTPUT_DIR + 'test_input.bin', X_test, OUTPUT_DIR + 'test_labels.bin', y_true)
+# # dumpCalibrationData2(OUTPUT_DIR + 'test_input_10K.bin', X_test, OUTPUT_DIR + 'test_labels_10K.bin', y_true)
+# dumpCalibrationData2(OUTPUT_DIR + 'tune_input.bin', X_tune, OUTPUT_DIR + 'tune_labels.bin', y_tune)
+# dumpCalibrationData2(OUTPUT_DIR + 'test_input.bin', X_test, OUTPUT_DIR + 'test_labels.bin', y_true)
pred = np.argmax(model.predict(X_test), axis=1)
print ('val accuracy', np.sum(pred == y_true.ravel()) / len(X_test))
-pred = np.argmax(model.predict(X_tune), axis=1)
-print ('val accuracy', np.sum(pred == y_tune.ravel()) / len(X_tune))
+# pred = np.argmax(model.predict(X_tune), axis=1)
+# print ('val accuracy', np.sum(pred == y_tune.ravel()) / len(X_tune))
\ No newline at end of file
diff --git a/llvm/projects/keras/src/vgg16_imagenet.py b/llvm/projects/keras/src/vgg16_imagenet.py
index 4f24e6b184911537dc425f4db0200f3f0836a48c..b41df8b83a966495d6a1c8281745e91181a66c41 100644
--- a/llvm/projects/keras/src/vgg16_imagenet.py
+++ b/llvm/projects/keras/src/vgg16_imagenet.py
@@ -32,7 +32,7 @@ IMAGENET_DIR = '/home/nz11/ILSVRC2012/'
OUTPUT_DIR = 'data/vgg16_imagenet_tune/'
NUM_CLASSES = 200
-IMAGES_PER_CLASS = 40
+IMAGES_PER_CLASS = 50
# VAL_SIZE = 100
@@ -212,18 +212,22 @@ y_true = np.array(y_true)
X_tune = np.array(X_tune)
y_tune = np.array(y_tune)
+print ('tune size', len(X_tune))
+print ('test size', len(X_test))
+
translate_to_approxhpvm(model, OUTPUT_DIR, X_tune, y_tune, 1000)
-# dumpCalibrationData2(OUTPUT_DIR + 'test_input_10K.bin', X_test, OUTPUT_DIR + 'test_labels_10K.bin', y_true)
-dumpCalibrationData2(OUTPUT_DIR + 'tune_input.bin', X_tune, OUTPUT_DIR + 'tune_labels.bin', y_tune)
-dumpCalibrationData2(OUTPUT_DIR + 'test_input.bin', X_test, OUTPUT_DIR + 'test_labels.bin', y_true)
+# # dumpCalibrationData2(OUTPUT_DIR + 'test_input_10K.bin', X_test, OUTPUT_DIR + 'test_labels_10K.bin', y_true)
+# dumpCalibrationData2(OUTPUT_DIR + 'tune_input.bin', X_tune, OUTPUT_DIR + 'tune_labels.bin', y_tune)
+# dumpCalibrationData2(OUTPUT_DIR + 'test_input.bin', X_test, OUTPUT_DIR + 'test_labels.bin', y_true)
pred = np.argmax(model.predict(X_test), axis=1)
print ('val accuracy', np.sum(pred == y_true.ravel()) / len(X_test))
-pred = np.argmax(model.predict(X_tune), axis=1)
-print ('val accuracy', np.sum(pred == y_tune.ravel()) / len(X_tune))
+# pred = np.argmax(model.predict(X_tune), axis=1)
+# print ('val accuracy', np.sum(pred == y_tune.ravel()) / len(X_tune))
+
diff --git a/llvm/test/VISC/DNN_Benchmarks/benchmarks/alexnet/Makefile b/llvm/test/VISC/DNN_Benchmarks/benchmarks/alexnet/Makefile
index fb23f70f66e85a77ae057c4d2edb989097723987..32cc431640742ce8e8323bec1ea74f0a1f5e202d 100644
--- a/llvm/test/VISC/DNN_Benchmarks/benchmarks/alexnet/Makefile
+++ b/llvm/test/VISC/DNN_Benchmarks/benchmarks/alexnet/Makefile
@@ -1,6 +1,4 @@
DNN_BENCHMARK_ROOT = $(LLVM_SRC_ROOT)/test/VISC/DNN_Benchmarks
-# NOTE: can configure build directory
-#HPVM_BUILD_DIR = $(LLVM_SRC_ROOT)/../build_hpvm/
HPVM_BUILD_DIR = $(LLVM_BUILD_ROOT)
CC = $(HPVM_BUILD_DIR)/bin/clang++
@@ -49,36 +47,28 @@ TARGET = $(BUILD_DIR)/$(APP).opt.bc
SOURCES = $(SRC_DIR)/$(APP).cpp
VISC_RT_PATH = $(LLVM_SRC_ROOT)/../build/projects/visc-rt/visc-rt.ll
-#OBJS = $(BUILD_DIR)/$(wildcabrd *.ll)
+
.PRECIOUS: $(BUILD_DIR)/$(APP).ll $(BUILD_DIR)/$(APP).visc.ll
default: $(BUILD_DIR) $(TARGET)
$(BUILD_DIR)/%.ll: $(SRC_DIR)/%.cpp
$(CC) $(CC_FLAGS) -emit-llvm src/$(APP).cpp -S -o $(BUILD_DIR)/$(APP).ll
- $(CC) $(CC_FLAGS) -emit-llvm src/$(APP)_promise.cpp -S -o $(BUILD_DIR)/$(APP)_promise.ll
$(CC) $(CC_FLAGS) -emit-llvm src/$(APP)_loop.cpp -S -o $(BUILD_DIR)/$(APP)_loop.ll
$(BUILD_DIR)/%.opt.bc: $(BUILD_DIR)/%.ll
$(OPT) -load LLVMGenVISC.so -genvisc -globaldce $(BUILD_DIR)/$(APP).ll -S -o $(BUILD_DIR)/$(APP).visc.ll
- $(OPT) -load LLVMGenVISC.so -genvisc -globaldce $(BUILD_DIR)/$(APP)_promise.ll -S -o $(BUILD_DIR)/$(APP)_promise.visc.ll
$(OPT) -load LLVMGenVISC.so -genvisc -globaldce $(BUILD_DIR)/$(APP)_loop.ll -S -o $(BUILD_DIR)/$(APP)_loop.visc.ll
$(OPT) $(VISC_OPTFLAGS) $(BUILD_DIR)/$(APP).visc.ll -o $(BUILD_DIR)/$(APP)_cudnn.bc
- #$(OPT) $(VISC_OPTFLAGS2) $(BUILD_DIR)/$(APP)_promise.visc.ll -o $(BUILD_DIR)/$(APP)_promise.bc
- $(OPT) $(VISC_OPTFLAGS3) $(BUILD_DIR)/$(APP)_promise.visc.ll -o $(BUILD_DIR)/$(APP)_wrapperapi.bc
$(OPT) $(VISC_OPTFLAGS3) $(BUILD_DIR)/$(APP)_loop.visc.ll -o $(BUILD_DIR)/$(APP)_loop_wrapperapi.bc
$(OPT) $(VISC_PRED_OPTFLAGS3) $(BUILD_DIR)/$(APP)_promise.visc.ll -o $(BUILD_DIR)/$(APP)_pred_wrapperapi.bc
$(OPT) $(VISC_PRED_OPTFLAGS3) $(BUILD_DIR)/$(APP)_loop.visc.ll -o $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_cudnn.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_cudnn_linked.bc
- #$(LLVM_LINK) $(BUILD_DIR)/$(APP)_promise.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_promise_linked.bc
- $(LLVM_LINK) $(BUILD_DIR)/$(APP)_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_wrapperapi_linked.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_loop_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_loop_wrapperapi_linked.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_pred_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_pred_wrapperapi_linked.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi_linked.bc
$(CC) $(BUILD_DIR)/$(APP)_cudnn_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_cudnn_linked $(LINKER_FLAGS)
- #$(CC) $(BUILD_DIR)/$(APP)_promise_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_promise_linked $(LINKER_FLAGS)
- $(CC) $(BUILD_DIR)/$(APP)_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_wrapperapi_linked $(LINKER_FLAGS)
$(CC) $(BUILD_DIR)/$(APP)_loop_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_loop_wrapperapi_linked $(LINKER_FLAGS)
$(CC) $(BUILD_DIR)/$(APP)_pred_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_pred_wrapperapi_linked $(LINKER_FLAGS)
$(CC) $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi_linked $(LINKER_FLAGS)
diff --git a/llvm/test/VISC/DNN_Benchmarks/benchmarks/alexnet/src/alexnet_loop.cpp b/llvm/test/VISC/DNN_Benchmarks/benchmarks/alexnet/src/alexnet_loop.cpp
index 09e6e25cffe18e375575bf83ebd4283e4295e7e4..246fa5f32d1bd1e2ab29867b25ce648558b0cb20 100644
--- a/llvm/test/VISC/DNN_Benchmarks/benchmarks/alexnet/src/alexnet_loop.cpp
+++ b/llvm/test/VISC/DNN_Benchmarks/benchmarks/alexnet/src/alexnet_loop.cpp
@@ -440,12 +440,11 @@ int main(){
int test_input_size = 5000;
int batch_count = test_input_size / batch_size;
- // void* input = create4DTensor(0,nchw,batch_size,3,32,32);
startMemTracking();
startProfiling();
- for (int j = 0; j < 14; j++){
+ for (int j = 0; j < 1; j++){
for (int i = 0; i < batch_count; i++){
int start = i * batch_size;
diff --git a/llvm/test/VISC/DNN_Benchmarks/benchmarks/alexnet2/Makefile b/llvm/test/VISC/DNN_Benchmarks/benchmarks/alexnet2/Makefile
index cae7df33de24e4d20e2d0d2b0977709a1865c9a8..4d4821a922507a257c9040af24adbcabe9dbf673 100644
--- a/llvm/test/VISC/DNN_Benchmarks/benchmarks/alexnet2/Makefile
+++ b/llvm/test/VISC/DNN_Benchmarks/benchmarks/alexnet2/Makefile
@@ -1,6 +1,4 @@
DNN_BENCHMARK_ROOT = $(LLVM_SRC_ROOT)/test/VISC/DNN_Benchmarks
-# NOTE: can configure build directory
-#HPVM_BUILD_DIR = $(LLVM_SRC_ROOT)/../build_hpvm/
HPVM_BUILD_DIR = $(LLVM_BUILD_ROOT)
CC = $(HPVM_BUILD_DIR)/bin/clang++
@@ -49,36 +47,28 @@ TARGET = $(BUILD_DIR)/$(APP).opt.bc
SOURCES = $(SRC_DIR)/$(APP).cpp
VISC_RT_PATH = $(LLVM_SRC_ROOT)/../build/projects/visc-rt/visc-rt.ll
-#OBJS = $(BUILD_DIR)/$(wildcabrd *.ll)
+
.PRECIOUS: $(BUILD_DIR)/$(APP).ll $(BUILD_DIR)/$(APP).visc.ll
default: $(BUILD_DIR) $(TARGET)
$(BUILD_DIR)/%.ll: $(SRC_DIR)/%.cpp
$(CC) $(CC_FLAGS) -emit-llvm src/$(APP).cpp -S -o $(BUILD_DIR)/$(APP).ll
- $(CC) $(CC_FLAGS) -emit-llvm src/$(APP)_promise.cpp -S -o $(BUILD_DIR)/$(APP)_promise.ll
$(CC) $(CC_FLAGS) -emit-llvm src/$(APP)_loop.cpp -S -o $(BUILD_DIR)/$(APP)_loop.ll
$(BUILD_DIR)/%.opt.bc: $(BUILD_DIR)/%.ll
$(OPT) -load LLVMGenVISC.so -genvisc -globaldce $(BUILD_DIR)/$(APP).ll -S -o $(BUILD_DIR)/$(APP).visc.ll
- $(OPT) -load LLVMGenVISC.so -genvisc -globaldce $(BUILD_DIR)/$(APP)_promise.ll -S -o $(BUILD_DIR)/$(APP)_promise.visc.ll
$(OPT) -load LLVMGenVISC.so -genvisc -globaldce $(BUILD_DIR)/$(APP)_loop.ll -S -o $(BUILD_DIR)/$(APP)_loop.visc.ll
$(OPT) $(VISC_OPTFLAGS) $(BUILD_DIR)/$(APP).visc.ll -o $(BUILD_DIR)/$(APP)_cudnn.bc
- #$(OPT) $(VISC_OPTFLAGS2) $(BUILD_DIR)/$(APP)_promise.visc.ll -o $(BUILD_DIR)/$(APP)_promise.bc
- $(OPT) $(VISC_OPTFLAGS3) $(BUILD_DIR)/$(APP)_promise.visc.ll -o $(BUILD_DIR)/$(APP)_wrapperapi.bc
$(OPT) $(VISC_OPTFLAGS3) $(BUILD_DIR)/$(APP)_loop.visc.ll -o $(BUILD_DIR)/$(APP)_loop_wrapperapi.bc
$(OPT) $(VISC_PRED_OPTFLAGS3) $(BUILD_DIR)/$(APP)_promise.visc.ll -o $(BUILD_DIR)/$(APP)_pred_wrapperapi.bc
$(OPT) $(VISC_PRED_OPTFLAGS3) $(BUILD_DIR)/$(APP)_loop.visc.ll -o $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_cudnn.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_cudnn_linked.bc
- #$(LLVM_LINK) $(BUILD_DIR)/$(APP)_promise.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_promise_linked.bc
- $(LLVM_LINK) $(BUILD_DIR)/$(APP)_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_wrapperapi_linked.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_loop_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_loop_wrapperapi_linked.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_pred_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_pred_wrapperapi_linked.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi_linked.bc
$(CC) $(BUILD_DIR)/$(APP)_cudnn_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_cudnn_linked $(LINKER_FLAGS)
- #$(CC) $(BUILD_DIR)/$(APP)_promise_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_promise_linked $(LINKER_FLAGS)
- $(CC) $(BUILD_DIR)/$(APP)_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_wrapperapi_linked $(LINKER_FLAGS)
$(CC) $(BUILD_DIR)/$(APP)_loop_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_loop_wrapperapi_linked $(LINKER_FLAGS)
$(CC) $(BUILD_DIR)/$(APP)_pred_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_pred_wrapperapi_linked $(LINKER_FLAGS)
$(CC) $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi_linked $(LINKER_FLAGS)
diff --git a/llvm/test/VISC/DNN_Benchmarks/benchmarks/alexnet2/src/alexnet2_loop.cpp b/llvm/test/VISC/DNN_Benchmarks/benchmarks/alexnet2/src/alexnet2_loop.cpp
index 77a5fde7f8d448f8b9c14b959696f65a574205b2..555b62869ff457596b3e435bb22937c5fa920e20 100644
--- a/llvm/test/VISC/DNN_Benchmarks/benchmarks/alexnet2/src/alexnet2_loop.cpp
+++ b/llvm/test/VISC/DNN_Benchmarks/benchmarks/alexnet2/src/alexnet2_loop.cpp
@@ -486,7 +486,7 @@ int main(){
args->dense_1_b_bytes = 0;
- int batch_size = 500;
+ int batch_size = 250;
int test_input_size = 5000;
int batch_count = test_input_size / batch_size;
@@ -496,7 +496,7 @@ int main(){
startMemTracking();
startProfiling();
- for (int j = 0; j < 14; j++){
+ for (int j = 0; j < 1; j++){
for (int i = 0; i < batch_count; i++){
int start = i * batch_size;
diff --git a/llvm/test/VISC/DNN_Benchmarks/benchmarks/alexnet_imagenet/Makefile b/llvm/test/VISC/DNN_Benchmarks/benchmarks/alexnet_imagenet/Makefile
index 0e6d9e4f08d33fb62bfb20ac8eb4f86ea4ca87de..fdc988b6858b15942f1bc16e9dbc8fcdb6c60b0f 100644
--- a/llvm/test/VISC/DNN_Benchmarks/benchmarks/alexnet_imagenet/Makefile
+++ b/llvm/test/VISC/DNN_Benchmarks/benchmarks/alexnet_imagenet/Makefile
@@ -1,6 +1,4 @@
DNN_BENCHMARK_ROOT = $(LLVM_SRC_ROOT)/test/VISC/DNN_Benchmarks
-# NOTE: can configure build directory
-#HPVM_BUILD_DIR = $(LLVM_SRC_ROOT)/../build_hpvm/
HPVM_BUILD_DIR = $(LLVM_BUILD_ROOT)
CC = $(HPVM_BUILD_DIR)/bin/clang++
@@ -49,22 +47,21 @@ TARGET = $(BUILD_DIR)/$(APP).opt.bc
SOURCES = $(SRC_DIR)/$(APP).cpp
VISC_RT_PATH = $(LLVM_SRC_ROOT)/../build/projects/visc-rt/visc-rt.ll
-#OBJS = $(BUILD_DIR)/$(wildcabrd *.ll)
+
.PRECIOUS: $(BUILD_DIR)/$(APP).ll $(BUILD_DIR)/$(APP).visc.ll
default: $(BUILD_DIR) $(TARGET)
$(BUILD_DIR)/%.ll: $(SRC_DIR)/%.cpp
$(CC) $(CC_FLAGS) -emit-llvm src/$(APP).cpp -S -o $(BUILD_DIR)/$(APP).ll
- $(CC) $(CC_FLAGS) -emit-llvm src/$(APP)_promise.cpp -S -o $(BUILD_DIR)/$(APP)_promise.ll
$(CC) $(CC_FLAGS) -emit-llvm src/$(APP)_loop.cpp -S -o $(BUILD_DIR)/$(APP)_loop.ll
$(BUILD_DIR)/%.opt.bc: $(BUILD_DIR)/%.ll
$(OPT) -load LLVMGenVISC.so -genvisc -globaldce $(BUILD_DIR)/$(APP).ll -S -o $(BUILD_DIR)/$(APP).visc.ll
- $(OPT) -load LLVMGenVISC.so -genvisc -globaldce $(BUILD_DIR)/$(APP)_promise.ll -S -o $(BUILD_DIR)/$(APP)_promise.visc.ll
$(OPT) -load LLVMGenVISC.so -genvisc -globaldce $(BUILD_DIR)/$(APP)_loop.ll -S -o $(BUILD_DIR)/$(APP)_loop.visc.ll
$(OPT) $(VISC_OPTFLAGS) $(BUILD_DIR)/$(APP).visc.ll -o $(BUILD_DIR)/$(APP)_cudnn.bc
+<<<<<<< HEAD
#$(OPT) $(VISC_OPTFLAGS2) $(BUILD_DIR)/$(APP)_promise.visc.ll -o $(BUILD_DIR)/$(APP)_promise.bc
$(OPT) $(VISC_OPTFLAGS3) $(BUILD_DIR)/$(APP)_promise.visc.ll -o $(BUILD_DIR)/$(APP)_wrapperapi.bc
$(OPT) $(VISC_OPTFLAGS3) $(BUILD_DIR)/$(APP)_loop.visc.ll -o $(BUILD_DIR)/$(APP)_loop_wrapperapi.bc
@@ -73,12 +70,19 @@ $(BUILD_DIR)/%.opt.bc: $(BUILD_DIR)/%.ll
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_cudnn.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_cudnn_linked.bc
#$(LLVM_LINK) $(BUILD_DIR)/$(APP)_promise.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_promise_linked.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_wrapperapi_linked.bc
+=======
+ $(OPT) $(VISC_OPTFLAGS3) $(BUILD_DIR)/$(APP)_loop.visc.ll -o $(BUILD_DIR)/$(APP)_loop_wrapperapi.bc
+ $(LLVM_LINK) $(BUILD_DIR)/$(APP)_cudnn.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_cudnn_linked.bc
+>>>>>>> 2a31471f4de89edd9180689d139be7ca65b0df08
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_loop_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_loop_wrapperapi_linked.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_pred_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_pred_wrapperapi_linked.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi_linked.bc
$(CC) $(BUILD_DIR)/$(APP)_cudnn_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_cudnn_linked $(LINKER_FLAGS)
+<<<<<<< HEAD
#$(CC) $(BUILD_DIR)/$(APP)_promise_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_promise_linked $(LINKER_FLAGS)
$(CC) $(BUILD_DIR)/$(APP)_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_wrapperapi_linked $(LINKER_FLAGS)
+=======
+>>>>>>> 2a31471f4de89edd9180689d139be7ca65b0df08
$(CC) $(BUILD_DIR)/$(APP)_loop_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_loop_wrapperapi_linked $(LINKER_FLAGS)
$(CC) $(BUILD_DIR)/$(APP)_pred_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_pred_wrapperapi_linked $(LINKER_FLAGS)
$(CC) $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi_linked $(LINKER_FLAGS)
diff --git a/llvm/test/VISC/DNN_Benchmarks/benchmarks/alexnet_imagenet/src/alexnet_imagenet_loop.cpp b/llvm/test/VISC/DNN_Benchmarks/benchmarks/alexnet_imagenet/src/alexnet_imagenet_loop.cpp
index 8423886e50000cc4e65527c19cc85f3a96e41efe..e9e3e6e0fd344f205784dc7d1fbd0e159f50d72a 100644
--- a/llvm/test/VISC/DNN_Benchmarks/benchmarks/alexnet_imagenet/src/alexnet_imagenet_loop.cpp
+++ b/llvm/test/VISC/DNN_Benchmarks/benchmarks/alexnet_imagenet/src/alexnet_imagenet_loop.cpp
@@ -555,7 +555,7 @@ int main(){
startMemTracking();
startProfiling();
- for (int j = 0; j < 14; j++){
+ for (int j = 0; j < 1; j++){
for (int i = 0; i < batch_count; i++){
int start = i * batch_size;
diff --git a/llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet_shallow/Makefile b/llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/mobilenet_shallow/Makefile
similarity index 100%
rename from llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet_shallow/Makefile
rename to llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/mobilenet_shallow/Makefile
diff --git a/llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet_shallow/data/autotuner_data/tuner_confs_batch220.txt b/llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/mobilenet_shallow/data/autotuner_data/tuner_confs_batch220.txt
similarity index 100%
rename from llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet_shallow/data/autotuner_data/tuner_confs_batch220.txt
rename to llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/mobilenet_shallow/data/autotuner_data/tuner_confs_batch220.txt
diff --git a/llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet_shallow/data/autotuner_data/tuner_pareto_confs_batch220.txt b/llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/mobilenet_shallow/data/autotuner_data/tuner_pareto_confs_batch220.txt
similarity index 100%
rename from llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet_shallow/data/autotuner_data/tuner_pareto_confs_batch220.txt
rename to llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/mobilenet_shallow/data/autotuner_data/tuner_pareto_confs_batch220.txt
diff --git a/llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet_shallow/data/autotuner_data/tuner_promise_confs_batch220_multi.txt b/llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/mobilenet_shallow/data/autotuner_data/tuner_promise_confs_batch220_multi.txt
similarity index 100%
rename from llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet_shallow/data/autotuner_data/tuner_promise_confs_batch220_multi.txt
rename to llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/mobilenet_shallow/data/autotuner_data/tuner_promise_confs_batch220_multi.txt
diff --git a/llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet_shallow/data/autotuner_data/tuner_promise_confs_batch220_multi2.txt b/llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/mobilenet_shallow/data/autotuner_data/tuner_promise_confs_batch220_multi2.txt
similarity index 100%
rename from llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet_shallow/data/autotuner_data/tuner_promise_confs_batch220_multi2.txt
rename to llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/mobilenet_shallow/data/autotuner_data/tuner_promise_confs_batch220_multi2.txt
diff --git a/llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet_shallow/data/autotuner_data/tuner_promise_confs_batch220_single.txt b/llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/mobilenet_shallow/data/autotuner_data/tuner_promise_confs_batch220_single.txt
similarity index 100%
rename from llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet_shallow/data/autotuner_data/tuner_promise_confs_batch220_single.txt
rename to llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/mobilenet_shallow/data/autotuner_data/tuner_promise_confs_batch220_single.txt
diff --git a/llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet_shallow/data/quant_ranges.txt b/llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/mobilenet_shallow/data/quant_ranges.txt
similarity index 100%
rename from llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet_shallow/data/quant_ranges.txt
rename to llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/mobilenet_shallow/data/quant_ranges.txt
diff --git a/llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet_shallow/data/quant_ranges_rt.txt b/llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/mobilenet_shallow/data/quant_ranges_rt.txt
similarity index 100%
rename from llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet_shallow/data/quant_ranges_rt.txt
rename to llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/mobilenet_shallow/data/quant_ranges_rt.txt
diff --git a/llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet_shallow/data/tuner_confs_base.txt b/llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/mobilenet_shallow/data/tuner_confs_base.txt
similarity index 100%
rename from llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet_shallow/data/tuner_confs_base.txt
rename to llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/mobilenet_shallow/data/tuner_confs_base.txt
diff --git a/llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet_shallow/src/mobilenet_shallow.cpp b/llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/mobilenet_shallow/src/mobilenet_shallow.cpp
similarity index 100%
rename from llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet_shallow/src/mobilenet_shallow.cpp
rename to llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/mobilenet_shallow/src/mobilenet_shallow.cpp
diff --git a/llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet_shallow/src/mobilenet_shallow_promise.cpp b/llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/mobilenet_shallow/src/mobilenet_shallow_promise.cpp
similarity index 100%
rename from llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet_shallow/src/mobilenet_shallow_promise.cpp
rename to llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/mobilenet_shallow/src/mobilenet_shallow_promise.cpp
diff --git a/llvm/test/VISC/DNN_Benchmarks/benchmarks/alexnet2/src/alexnet2_promise.cpp b/llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/promise_src/alexnet2_promise.cpp
similarity index 100%
rename from llvm/test/VISC/DNN_Benchmarks/benchmarks/alexnet2/src/alexnet2_promise.cpp
rename to llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/promise_src/alexnet2_promise.cpp
diff --git a/llvm/test/VISC/DNN_Benchmarks/benchmarks/alexnet_imagenet/src/alexnet_imagenet_promise.cpp b/llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/promise_src/alexnet_imagenet_promise.cpp
similarity index 100%
rename from llvm/test/VISC/DNN_Benchmarks/benchmarks/alexnet_imagenet/src/alexnet_imagenet_promise.cpp
rename to llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/promise_src/alexnet_imagenet_promise.cpp
diff --git a/llvm/test/VISC/DNN_Benchmarks/benchmarks/alexnet/src/alexnet_promise.cpp b/llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/promise_src/alexnet_promise.cpp
similarity index 100%
rename from llvm/test/VISC/DNN_Benchmarks/benchmarks/alexnet/src/alexnet_promise.cpp
rename to llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/promise_src/alexnet_promise.cpp
diff --git a/llvm/test/VISC/DNN_Benchmarks/benchmarks/lenet_mnist/src/lenet_mnist_promise.cpp b/llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/promise_src/lenet_mnist_promise.cpp
similarity index 100%
rename from llvm/test/VISC/DNN_Benchmarks/benchmarks/lenet_mnist/src/lenet_mnist_promise.cpp
rename to llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/promise_src/lenet_mnist_promise.cpp
diff --git a/llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet/src/mobilenet_promise.cpp b/llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/promise_src/mobilenet_promise.cpp
similarity index 100%
rename from llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet/src/mobilenet_promise.cpp
rename to llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/promise_src/mobilenet_promise.cpp
diff --git a/llvm/test/VISC/DNN_Benchmarks/benchmarks/resnet18/src/resnet18_promise.cpp b/llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/promise_src/resnet18_promise.cpp
similarity index 100%
rename from llvm/test/VISC/DNN_Benchmarks/benchmarks/resnet18/src/resnet18_promise.cpp
rename to llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/promise_src/resnet18_promise.cpp
diff --git a/llvm/test/VISC/DNN_Benchmarks/benchmarks/vgg16_cifar100/src/vgg16_cifar100_promise.cpp b/llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/promise_src/vgg16_cifar100_promise.cpp
similarity index 100%
rename from llvm/test/VISC/DNN_Benchmarks/benchmarks/vgg16_cifar100/src/vgg16_cifar100_promise.cpp
rename to llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/promise_src/vgg16_cifar100_promise.cpp
diff --git a/llvm/test/VISC/DNN_Benchmarks/benchmarks/vgg16_cifar10/src/vgg16_cifar10_promise.cpp b/llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/promise_src/vgg16_cifar10_promise.cpp
similarity index 100%
rename from llvm/test/VISC/DNN_Benchmarks/benchmarks/vgg16_cifar10/src/vgg16_cifar10_promise.cpp
rename to llvm/test/VISC/DNN_Benchmarks/benchmarks/legacy/promise_src/vgg16_cifar10_promise.cpp
diff --git a/llvm/test/VISC/DNN_Benchmarks/benchmarks/lenet_mnist/Makefile b/llvm/test/VISC/DNN_Benchmarks/benchmarks/lenet_mnist/Makefile
index 909ecfdca5a730e4d400cce70addee2d36573f1c..643715cbf57138926b222456f7b3b4257ca80f20 100644
--- a/llvm/test/VISC/DNN_Benchmarks/benchmarks/lenet_mnist/Makefile
+++ b/llvm/test/VISC/DNN_Benchmarks/benchmarks/lenet_mnist/Makefile
@@ -1,6 +1,9 @@
DNN_BENCHMARK_ROOT = $(LLVM_SRC_ROOT)/test/VISC/DNN_Benchmarks
+<<<<<<< HEAD
# NOTE: can configure build directory
#HPVM_BUILD_DIR = $(LLVM_SRC_ROOT)/../build_hpvm/
+=======
+>>>>>>> 2a31471f4de89edd9180689d139be7ca65b0df08
HPVM_BUILD_DIR = $(LLVM_BUILD_ROOT)
CC = $(HPVM_BUILD_DIR)/bin/clang++
@@ -49,36 +52,50 @@ TARGET = $(BUILD_DIR)/$(APP).opt.bc
SOURCES = $(SRC_DIR)/$(APP).cpp
VISC_RT_PATH = $(LLVM_SRC_ROOT)/../build/projects/visc-rt/visc-rt.ll
-#OBJS = $(BUILD_DIR)/$(wildcabrd *.ll)
+
.PRECIOUS: $(BUILD_DIR)/$(APP).ll $(BUILD_DIR)/$(APP).visc.ll
default: $(BUILD_DIR) $(TARGET)
$(BUILD_DIR)/%.ll: $(SRC_DIR)/%.cpp
$(CC) $(CC_FLAGS) -emit-llvm src/$(APP).cpp -S -o $(BUILD_DIR)/$(APP).ll
+<<<<<<< HEAD
$(CC) $(CC_FLAGS) -emit-llvm src/$(APP)_promise.cpp -S -o $(BUILD_DIR)/$(APP)_promise.ll
+=======
+>>>>>>> 2a31471f4de89edd9180689d139be7ca65b0df08
$(CC) $(CC_FLAGS) -emit-llvm src/$(APP)_loop.cpp -S -o $(BUILD_DIR)/$(APP)_loop.ll
$(BUILD_DIR)/%.opt.bc: $(BUILD_DIR)/%.ll
$(OPT) -load LLVMGenVISC.so -genvisc -globaldce $(BUILD_DIR)/$(APP).ll -S -o $(BUILD_DIR)/$(APP).visc.ll
+<<<<<<< HEAD
$(OPT) -load LLVMGenVISC.so -genvisc -globaldce $(BUILD_DIR)/$(APP)_promise.ll -S -o $(BUILD_DIR)/$(APP)_promise.visc.ll
$(OPT) -load LLVMGenVISC.so -genvisc -globaldce $(BUILD_DIR)/$(APP)_loop.ll -S -o $(BUILD_DIR)/$(APP)_loop.visc.ll
$(OPT) $(VISC_OPTFLAGS) $(BUILD_DIR)/$(APP).visc.ll -o $(BUILD_DIR)/$(APP)_cudnn.bc
#$(OPT) $(VISC_OPTFLAGS2) $(BUILD_DIR)/$(APP)_promise.visc.ll -o $(BUILD_DIR)/$(APP)_promise.bc
$(OPT) $(VISC_OPTFLAGS3) $(BUILD_DIR)/$(APP)_promise.visc.ll -o $(BUILD_DIR)/$(APP)_wrapperapi.bc
+=======
+ $(OPT) -load LLVMGenVISC.so -genvisc -globaldce $(BUILD_DIR)/$(APP)_loop.ll -S -o $(BUILD_DIR)/$(APP)_loop.visc.ll
+ $(OPT) $(VISC_OPTFLAGS) $(BUILD_DIR)/$(APP).visc.ll -o $(BUILD_DIR)/$(APP)_cudnn.bc
+>>>>>>> 2a31471f4de89edd9180689d139be7ca65b0df08
$(OPT) $(VISC_OPTFLAGS3) $(BUILD_DIR)/$(APP)_loop.visc.ll -o $(BUILD_DIR)/$(APP)_loop_wrapperapi.bc
$(OPT) $(VISC_PRED_OPTFLAGS3) $(BUILD_DIR)/$(APP)_promise.visc.ll -o $(BUILD_DIR)/$(APP)_pred_wrapperapi.bc
$(OPT) $(VISC_PRED_OPTFLAGS3) $(BUILD_DIR)/$(APP)_loop.visc.ll -o $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_cudnn.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_cudnn_linked.bc
+<<<<<<< HEAD
#$(LLVM_LINK) $(BUILD_DIR)/$(APP)_promise.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_promise_linked.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_wrapperapi_linked.bc
+=======
+>>>>>>> 2a31471f4de89edd9180689d139be7ca65b0df08
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_loop_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_loop_wrapperapi_linked.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_pred_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_pred_wrapperapi_linked.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi_linked.bc
$(CC) $(BUILD_DIR)/$(APP)_cudnn_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_cudnn_linked $(LINKER_FLAGS)
+<<<<<<< HEAD
#$(CC) $(BUILD_DIR)/$(APP)_promise_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_promise_linked $(LINKER_FLAGS)
$(CC) $(BUILD_DIR)/$(APP)_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_wrapperapi_linked $(LINKER_FLAGS)
+=======
+>>>>>>> 2a31471f4de89edd9180689d139be7ca65b0df08
$(CC) $(BUILD_DIR)/$(APP)_loop_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_loop_wrapperapi_linked $(LINKER_FLAGS)
$(CC) $(BUILD_DIR)/$(APP)_pred_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_pred_wrapperapi_linked $(LINKER_FLAGS)
$(CC) $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi_linked $(LINKER_FLAGS)
diff --git a/llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet/Makefile b/llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet/Makefile
index 3d678aae8feaf65bdfb9f3c04fafcb6a04505070..507d64f121e2448263ea294f6a94cd8974f0ff50 100644
--- a/llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet/Makefile
+++ b/llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet/Makefile
@@ -1,6 +1,4 @@
DNN_BENCHMARK_ROOT = $(LLVM_SRC_ROOT)/test/VISC/DNN_Benchmarks
-# NOTE: can configure build directory
-#HPVM_BUILD_DIR = $(LLVM_SRC_ROOT)/../build_hpvm/
HPVM_BUILD_DIR = $(LLVM_BUILD_ROOT)
CC = $(HPVM_BUILD_DIR)/bin/clang++
@@ -49,36 +47,28 @@ TARGET = $(BUILD_DIR)/$(APP).opt.bc
SOURCES = $(SRC_DIR)/$(APP).cpp
VISC_RT_PATH = $(LLVM_SRC_ROOT)/../build/projects/visc-rt/visc-rt.ll
-#OBJS = $(BUILD_DIR)/$(wildcabrd *.ll)
+
.PRECIOUS: $(BUILD_DIR)/$(APP).ll $(BUILD_DIR)/$(APP).visc.ll
default: $(BUILD_DIR) $(TARGET)
$(BUILD_DIR)/%.ll: $(SRC_DIR)/%.cpp
$(CC) $(CC_FLAGS) -emit-llvm src/$(APP).cpp -S -o $(BUILD_DIR)/$(APP).ll
- $(CC) $(CC_FLAGS) -emit-llvm src/$(APP)_promise.cpp -S -o $(BUILD_DIR)/$(APP)_promise.ll
$(CC) $(CC_FLAGS) -emit-llvm src/$(APP)_loop.cpp -S -o $(BUILD_DIR)/$(APP)_loop.ll
$(BUILD_DIR)/%.opt.bc: $(BUILD_DIR)/%.ll
$(OPT) -load LLVMGenVISC.so -genvisc -globaldce $(BUILD_DIR)/$(APP).ll -S -o $(BUILD_DIR)/$(APP).visc.ll
- $(OPT) -load LLVMGenVISC.so -genvisc -globaldce $(BUILD_DIR)/$(APP)_promise.ll -S -o $(BUILD_DIR)/$(APP)_promise.visc.ll
$(OPT) -load LLVMGenVISC.so -genvisc -globaldce $(BUILD_DIR)/$(APP)_loop.ll -S -o $(BUILD_DIR)/$(APP)_loop.visc.ll
$(OPT) $(VISC_OPTFLAGS) $(BUILD_DIR)/$(APP).visc.ll -o $(BUILD_DIR)/$(APP)_cudnn.bc
- #$(OPT) $(VISC_OPTFLAGS2) $(BUILD_DIR)/$(APP)_promise.visc.ll -o $(BUILD_DIR)/$(APP)_promise.bc
- $(OPT) $(VISC_OPTFLAGS3) $(BUILD_DIR)/$(APP)_promise.visc.ll -o $(BUILD_DIR)/$(APP)_wrapperapi.bc
$(OPT) $(VISC_OPTFLAGS3) $(BUILD_DIR)/$(APP)_loop.visc.ll -o $(BUILD_DIR)/$(APP)_loop_wrapperapi.bc
$(OPT) $(VISC_PRED_OPTFLAGS3) $(BUILD_DIR)/$(APP)_promise.visc.ll -o $(BUILD_DIR)/$(APP)_pred_wrapperapi.bc
$(OPT) $(VISC_PRED_OPTFLAGS3) $(BUILD_DIR)/$(APP)_loop.visc.ll -o $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_cudnn.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_cudnn_linked.bc
- #$(LLVM_LINK) $(BUILD_DIR)/$(APP)_promise.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_promise_linked.bc
- $(LLVM_LINK) $(BUILD_DIR)/$(APP)_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_wrapperapi_linked.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_loop_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_loop_wrapperapi_linked.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_pred_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_pred_wrapperapi_linked.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi_linked.bc
$(CC) $(BUILD_DIR)/$(APP)_cudnn_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_cudnn_linked $(LINKER_FLAGS)
- #$(CC) $(BUILD_DIR)/$(APP)_promise_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_promise_linked $(LINKER_FLAGS)
- $(CC) $(BUILD_DIR)/$(APP)_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_wrapperapi_linked $(LINKER_FLAGS)
$(CC) $(BUILD_DIR)/$(APP)_loop_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_loop_wrapperapi_linked $(LINKER_FLAGS)
$(CC) $(BUILD_DIR)/$(APP)_pred_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_pred_wrapperapi_linked $(LINKER_FLAGS)
$(CC) $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi_linked $(LINKER_FLAGS)
diff --git a/llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet/data/tuner_confs_base.txt b/llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet/data/tuner_confs_base.txt
index 8b5c1727ad0dc9e24310e4c86e116894051c84b3..ed02ddab0dbef2b21f785226b80f4eee7a1735cf 100644
--- a/llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet/data/tuner_confs_base.txt
+++ b/llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet/data/tuner_confs_base.txt
@@ -1,3 +1,4 @@
+1000
+++++
conf1 1 0 84.8 0
1 gpu conv fp32 1
diff --git a/llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet/src/mobilenet.cpp b/llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet/src/mobilenet.cpp
index 5266f216887445b41f7d63a4c2e92b134a09ccc5..2fbb5a6eb424ae93e0d301b70620185060f4b7f5 100644
--- a/llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet/src/mobilenet.cpp
+++ b/llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet/src/mobilenet.cpp
@@ -1826,7 +1826,7 @@ RootIn;
int main(){
- std::string dir_prefix = std::string("../../../../../../projects/hpvm-tensor-rt/model_params/mobilenet_quant/");
+ std::string dir_prefix = std::string("../../../../../../projects/hpvm-tensor-rt/model_params/mobilenet/");
std::string input_path = dir_prefix + std::string("input.bin");
std::string labels_path = dir_prefix + std::string("labels.bin");
diff --git a/llvm/test/VISC/DNN_Benchmarks/benchmarks/resnet18/Makefile b/llvm/test/VISC/DNN_Benchmarks/benchmarks/resnet18/Makefile
index 4de3565fce81f857beb3dd4ee60e7423bdd0322a..409358ba466c3d86f563f630d24fe6845cea7274 100644
--- a/llvm/test/VISC/DNN_Benchmarks/benchmarks/resnet18/Makefile
+++ b/llvm/test/VISC/DNN_Benchmarks/benchmarks/resnet18/Makefile
@@ -1,6 +1,5 @@
+LLVM_BUILD_ROOT=/home/hsharif3/Gitlab/hpvm/build_hpvm/
DNN_BENCHMARK_ROOT = $(LLVM_SRC_ROOT)/test/VISC/DNN_Benchmarks
-# NOTE: can configure build directory
-#HPVM_BUILD_DIR = $(LLVM_SRC_ROOT)/../build_hpvm/
HPVM_BUILD_DIR = $(LLVM_BUILD_ROOT)
CC = $(HPVM_BUILD_DIR)/bin/clang++
@@ -56,29 +55,21 @@ default: $(BUILD_DIR) $(TARGET)
$(BUILD_DIR)/%.ll: $(SRC_DIR)/%.cpp
$(CC) $(CC_FLAGS) -emit-llvm src/$(APP).cpp -S -o $(BUILD_DIR)/$(APP).ll
- $(CC) $(CC_FLAGS) -emit-llvm src/$(APP)_promise.cpp -S -o $(BUILD_DIR)/$(APP)_promise.ll
$(CC) $(CC_FLAGS) -emit-llvm src/$(APP)_loop.cpp -S -o $(BUILD_DIR)/$(APP)_loop.ll
$(BUILD_DIR)/%.opt.bc: $(BUILD_DIR)/%.ll
$(OPT) -load LLVMGenVISC.so -genvisc -globaldce $(BUILD_DIR)/$(APP).ll -S -o $(BUILD_DIR)/$(APP).visc.ll
- $(OPT) -load LLVMGenVISC.so -genvisc -globaldce $(BUILD_DIR)/$(APP)_promise.ll -S -o $(BUILD_DIR)/$(APP)_promise.visc.ll
$(OPT) -load LLVMGenVISC.so -genvisc -globaldce $(BUILD_DIR)/$(APP)_loop.ll -S -o $(BUILD_DIR)/$(APP)_loop.visc.ll
$(OPT) $(VISC_OPTFLAGS) $(BUILD_DIR)/$(APP).visc.ll -o $(BUILD_DIR)/$(APP)_cudnn.bc
- #$(OPT) $(VISC_OPTFLAGS2) $(BUILD_DIR)/$(APP)_promise.visc.ll -o $(BUILD_DIR)/$(APP)_promise.bc
- $(OPT) $(VISC_OPTFLAGS3) $(BUILD_DIR)/$(APP)_promise.visc.ll -o $(BUILD_DIR)/$(APP)_wrapperapi.bc
$(OPT) $(VISC_OPTFLAGS3) $(BUILD_DIR)/$(APP)_loop.visc.ll -o $(BUILD_DIR)/$(APP)_loop_wrapperapi.bc
$(OPT) $(VISC_PRED_OPTFLAGS3) $(BUILD_DIR)/$(APP)_promise.visc.ll -o $(BUILD_DIR)/$(APP)_pred_wrapperapi.bc
$(OPT) $(VISC_PRED_OPTFLAGS3) $(BUILD_DIR)/$(APP)_loop.visc.ll -o $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_cudnn.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_cudnn_linked.bc
- #$(LLVM_LINK) $(BUILD_DIR)/$(APP)_promise.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_promise_linked.bc
- $(LLVM_LINK) $(BUILD_DIR)/$(APP)_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_wrapperapi_linked.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_loop_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_loop_wrapperapi_linked.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_pred_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_pred_wrapperapi_linked.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi_linked.bc
$(CC) $(BUILD_DIR)/$(APP)_cudnn_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_cudnn_linked $(LINKER_FLAGS)
- #$(CC) $(BUILD_DIR)/$(APP)_promise_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_promise_linked $(LINKER_FLAGS)
- $(CC) $(BUILD_DIR)/$(APP)_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_wrapperapi_linked $(LINKER_FLAGS)
$(CC) $(BUILD_DIR)/$(APP)_loop_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_loop_wrapperapi_linked $(LINKER_FLAGS)
$(CC) $(BUILD_DIR)/$(APP)_pred_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_pred_wrapperapi_linked $(LINKER_FLAGS)
$(CC) $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi_linked $(LINKER_FLAGS)
diff --git a/llvm/test/VISC/DNN_Benchmarks/benchmarks/resnet50_imagenet/Makefile b/llvm/test/VISC/DNN_Benchmarks/benchmarks/resnet50_imagenet/Makefile
index 5fb4534594a1aef9077f74f53ffe06281bc2b23a..5180f55744d292e54042ebbf8227a3ebed99a3b4 100644
--- a/llvm/test/VISC/DNN_Benchmarks/benchmarks/resnet50_imagenet/Makefile
+++ b/llvm/test/VISC/DNN_Benchmarks/benchmarks/resnet50_imagenet/Makefile
@@ -1,3 +1,7 @@
+<<<<<<< HEAD
+=======
+LLVM_BUILD_ROOT=/home/hsharif3/Gitlab/hpvm/build_hpvm/
+>>>>>>> 2a31471f4de89edd9180689d139be7ca65b0df08
DNN_BENCHMARK_ROOT = $(LLVM_SRC_ROOT)/test/VISC/DNN_Benchmarks
# NOTE: can configure build directory
#HPVM_BUILD_DIR = $(LLVM_SRC_ROOT)/../build_hpvm/
@@ -49,36 +53,50 @@ TARGET = $(BUILD_DIR)/$(APP).opt.bc
SOURCES = $(SRC_DIR)/$(APP).cpp
VISC_RT_PATH = $(LLVM_SRC_ROOT)/../build/projects/visc-rt/visc-rt.ll
-#OBJS = $(BUILD_DIR)/$(wildcabrd *.ll)
+
.PRECIOUS: $(BUILD_DIR)/$(APP).ll $(BUILD_DIR)/$(APP).visc.ll
default: $(BUILD_DIR) $(TARGET)
$(BUILD_DIR)/%.ll: $(SRC_DIR)/%.cpp
$(CC) $(CC_FLAGS) -emit-llvm src/$(APP).cpp -S -o $(BUILD_DIR)/$(APP).ll
+<<<<<<< HEAD
$(CC) $(CC_FLAGS) -emit-llvm src/$(APP)_promise.cpp -S -o $(BUILD_DIR)/$(APP)_promise.ll
+=======
+>>>>>>> 2a31471f4de89edd9180689d139be7ca65b0df08
$(CC) $(CC_FLAGS) -emit-llvm src/$(APP)_loop.cpp -S -o $(BUILD_DIR)/$(APP)_loop.ll
$(BUILD_DIR)/%.opt.bc: $(BUILD_DIR)/%.ll
$(OPT) -load LLVMGenVISC.so -genvisc -globaldce $(BUILD_DIR)/$(APP).ll -S -o $(BUILD_DIR)/$(APP).visc.ll
+<<<<<<< HEAD
$(OPT) -load LLVMGenVISC.so -genvisc -globaldce $(BUILD_DIR)/$(APP)_promise.ll -S -o $(BUILD_DIR)/$(APP)_promise.visc.ll
$(OPT) -load LLVMGenVISC.so -genvisc -globaldce $(BUILD_DIR)/$(APP)_loop.ll -S -o $(BUILD_DIR)/$(APP)_loop.visc.ll
$(OPT) $(VISC_OPTFLAGS) $(BUILD_DIR)/$(APP).visc.ll -o $(BUILD_DIR)/$(APP)_cudnn.bc
#$(OPT) $(VISC_OPTFLAGS2) $(BUILD_DIR)/$(APP)_promise.visc.ll -o $(BUILD_DIR)/$(APP)_promise.bc
$(OPT) $(VISC_OPTFLAGS3) $(BUILD_DIR)/$(APP)_promise.visc.ll -o $(BUILD_DIR)/$(APP)_wrapperapi.bc
+=======
+ $(OPT) -load LLVMGenVISC.so -genvisc -globaldce $(BUILD_DIR)/$(APP)_loop.ll -S -o $(BUILD_DIR)/$(APP)_loop.visc.ll
+ $(OPT) $(VISC_OPTFLAGS) $(BUILD_DIR)/$(APP).visc.ll -o $(BUILD_DIR)/$(APP)_cudnn.bc
+>>>>>>> 2a31471f4de89edd9180689d139be7ca65b0df08
$(OPT) $(VISC_OPTFLAGS3) $(BUILD_DIR)/$(APP)_loop.visc.ll -o $(BUILD_DIR)/$(APP)_loop_wrapperapi.bc
$(OPT) $(VISC_PRED_OPTFLAGS3) $(BUILD_DIR)/$(APP)_promise.visc.ll -o $(BUILD_DIR)/$(APP)_pred_wrapperapi.bc
$(OPT) $(VISC_PRED_OPTFLAGS3) $(BUILD_DIR)/$(APP)_loop.visc.ll -o $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_cudnn.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_cudnn_linked.bc
+<<<<<<< HEAD
#$(LLVM_LINK) $(BUILD_DIR)/$(APP)_promise.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_promise_linked.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_wrapperapi_linked.bc
+=======
+>>>>>>> 2a31471f4de89edd9180689d139be7ca65b0df08
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_loop_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_loop_wrapperapi_linked.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_pred_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_pred_wrapperapi_linked.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi_linked.bc
$(CC) $(BUILD_DIR)/$(APP)_cudnn_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_cudnn_linked $(LINKER_FLAGS)
+<<<<<<< HEAD
#$(CC) $(BUILD_DIR)/$(APP)_promise_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_promise_linked $(LINKER_FLAGS)
$(CC) $(BUILD_DIR)/$(APP)_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_wrapperapi_linked $(LINKER_FLAGS)
+=======
+>>>>>>> 2a31471f4de89edd9180689d139be7ca65b0df08
$(CC) $(BUILD_DIR)/$(APP)_loop_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_loop_wrapperapi_linked $(LINKER_FLAGS)
$(CC) $(BUILD_DIR)/$(APP)_pred_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_pred_wrapperapi_linked $(LINKER_FLAGS)
$(CC) $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi_linked $(LINKER_FLAGS)
diff --git a/llvm/test/VISC/DNN_Benchmarks/benchmarks/vgg16_cifar10/Makefile b/llvm/test/VISC/DNN_Benchmarks/benchmarks/vgg16_cifar10/Makefile
index cf001e5dc9187b9562959af3223970ed9dce9b97..511cb5c8adf073956599b412ea18f7941c889ab9 100644
--- a/llvm/test/VISC/DNN_Benchmarks/benchmarks/vgg16_cifar10/Makefile
+++ b/llvm/test/VISC/DNN_Benchmarks/benchmarks/vgg16_cifar10/Makefile
@@ -1,6 +1,4 @@
DNN_BENCHMARK_ROOT = $(LLVM_SRC_ROOT)/test/VISC/DNN_Benchmarks
-# NOTE: can configure build directory
-#HPVM_BUILD_DIR = $(LLVM_SRC_ROOT)/../build_hpvm/
HPVM_BUILD_DIR = $(LLVM_BUILD_ROOT)
CC = $(HPVM_BUILD_DIR)/bin/clang++
@@ -49,36 +47,30 @@ TARGET = $(BUILD_DIR)/$(APP).opt.bc
SOURCES = $(SRC_DIR)/$(APP).cpp
VISC_RT_PATH = $(LLVM_SRC_ROOT)/../build/projects/visc-rt/visc-rt.ll
-#OBJS = $(BUILD_DIR)/$(wildcabrd *.ll)
+
.PRECIOUS: $(BUILD_DIR)/$(APP).ll $(BUILD_DIR)/$(APP).visc.ll
default: $(BUILD_DIR) $(TARGET)
$(BUILD_DIR)/%.ll: $(SRC_DIR)/%.cpp
$(CC) $(CC_FLAGS) -emit-llvm src/$(APP).cpp -S -o $(BUILD_DIR)/$(APP).ll
- $(CC) $(CC_FLAGS) -emit-llvm src/$(APP)_promise.cpp -S -o $(BUILD_DIR)/$(APP)_promise.ll
$(CC) $(CC_FLAGS) -emit-llvm src/$(APP)_loop.cpp -S -o $(BUILD_DIR)/$(APP)_loop.ll
$(BUILD_DIR)/%.opt.bc: $(BUILD_DIR)/%.ll
$(OPT) -load LLVMGenVISC.so -genvisc -globaldce $(BUILD_DIR)/$(APP).ll -S -o $(BUILD_DIR)/$(APP).visc.ll
- $(OPT) -load LLVMGenVISC.so -genvisc -globaldce $(BUILD_DIR)/$(APP)_promise.ll -S -o $(BUILD_DIR)/$(APP)_promise.visc.ll
$(OPT) -load LLVMGenVISC.so -genvisc -globaldce $(BUILD_DIR)/$(APP)_loop.ll -S -o $(BUILD_DIR)/$(APP)_loop.visc.ll
$(OPT) $(VISC_OPTFLAGS) $(BUILD_DIR)/$(APP).visc.ll -o $(BUILD_DIR)/$(APP)_cudnn.bc
- #$(OPT) $(VISC_OPTFLAGS2) $(BUILD_DIR)/$(APP)_promise.visc.ll -o $(BUILD_DIR)/$(APP)_promise.bc
- $(OPT) $(VISC_OPTFLAGS3) $(BUILD_DIR)/$(APP)_promise.visc.ll -o $(BUILD_DIR)/$(APP)_wrapperapi.bc
$(OPT) $(VISC_OPTFLAGS3) $(BUILD_DIR)/$(APP)_loop.visc.ll -o $(BUILD_DIR)/$(APP)_loop_wrapperapi.bc
$(OPT) $(VISC_PRED_OPTFLAGS3) $(BUILD_DIR)/$(APP)_promise.visc.ll -o $(BUILD_DIR)/$(APP)_pred_wrapperapi.bc
$(OPT) $(VISC_PRED_OPTFLAGS3) $(BUILD_DIR)/$(APP)_loop.visc.ll -o $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_cudnn.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_cudnn_linked.bc
- #$(LLVM_LINK) $(BUILD_DIR)/$(APP)_promise.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_promise_linked.bc
- $(LLVM_LINK) $(BUILD_DIR)/$(APP)_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_wrapperapi_linked.bc
+ #$(LLVM_LINK) $(BUILD_DIR)/$(APP)_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_wrapperapi_linked.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_loop_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_loop_wrapperapi_linked.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_pred_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_pred_wrapperapi_linked.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi_linked.bc
$(CC) $(BUILD_DIR)/$(APP)_cudnn_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_cudnn_linked $(LINKER_FLAGS)
- #$(CC) $(BUILD_DIR)/$(APP)_promise_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_promise_linked $(LINKER_FLAGS)
- $(CC) $(BUILD_DIR)/$(APP)_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_wrapperapi_linked $(LINKER_FLAGS)
+ #$(CC) $(BUILD_DIR)/$(APP)_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_wrapperapi_linked $(LINKER_FLAGS)
$(CC) $(BUILD_DIR)/$(APP)_loop_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_loop_wrapperapi_linked $(LINKER_FLAGS)
$(CC) $(BUILD_DIR)/$(APP)_pred_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_pred_wrapperapi_linked $(LINKER_FLAGS)
$(CC) $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi_linked $(LINKER_FLAGS)
diff --git a/llvm/test/VISC/DNN_Benchmarks/benchmarks/vgg16_cifar10/src/vgg16_cifar10_loop.cpp b/llvm/test/VISC/DNN_Benchmarks/benchmarks/vgg16_cifar10/src/vgg16_cifar10_loop.cpp
index 602164516f78462f7059ef7aa975bf2aa0d34bf0..a5f507e409d72f6f264eb9ff2b7e3dcc03fcc79a 100644
--- a/llvm/test/VISC/DNN_Benchmarks/benchmarks/vgg16_cifar10/src/vgg16_cifar10_loop.cpp
+++ b/llvm/test/VISC/DNN_Benchmarks/benchmarks/vgg16_cifar10/src/vgg16_cifar10_loop.cpp
@@ -974,12 +974,11 @@ int main(){
int test_input_size = 5000;
int batch_count = test_input_size / batch_size;
- // void* input = create4DTensor(0,nchw,batch_size,3,32,32);
startMemTracking();
startProfiling();
- for (int j = 0; j < 14; j++){
+ for (int j = 0; j < 1; j++){
for (int i = 0; i < batch_count; i++){
int start = i * batch_size;
diff --git a/llvm/test/VISC/DNN_Benchmarks/benchmarks/vgg16_cifar100/Makefile b/llvm/test/VISC/DNN_Benchmarks/benchmarks/vgg16_cifar100/Makefile
index 199d24b54f856be3b5c858ccd4ba8269a7e73328..052011292a3af053d23d0d6c627e5fd2d0231b13 100644
--- a/llvm/test/VISC/DNN_Benchmarks/benchmarks/vgg16_cifar100/Makefile
+++ b/llvm/test/VISC/DNN_Benchmarks/benchmarks/vgg16_cifar100/Makefile
@@ -1,6 +1,4 @@
DNN_BENCHMARK_ROOT = $(LLVM_SRC_ROOT)/test/VISC/DNN_Benchmarks
-# NOTE: can configure build directory
-#HPVM_BUILD_DIR = $(LLVM_SRC_ROOT)/../build_hpvm/
HPVM_BUILD_DIR = $(LLVM_BUILD_ROOT)
CC = $(HPVM_BUILD_DIR)/bin/clang++
@@ -49,36 +47,28 @@ TARGET = $(BUILD_DIR)/$(APP).opt.bc
SOURCES = $(SRC_DIR)/$(APP).cpp
VISC_RT_PATH = $(LLVM_SRC_ROOT)/../build/projects/visc-rt/visc-rt.ll
-#OBJS = $(BUILD_DIR)/$(wildcabrd *.ll)
+
.PRECIOUS: $(BUILD_DIR)/$(APP).ll $(BUILD_DIR)/$(APP).visc.ll
default: $(BUILD_DIR) $(TARGET)
$(BUILD_DIR)/%.ll: $(SRC_DIR)/%.cpp
$(CC) $(CC_FLAGS) -emit-llvm src/$(APP).cpp -S -o $(BUILD_DIR)/$(APP).ll
- $(CC) $(CC_FLAGS) -emit-llvm src/$(APP)_promise.cpp -S -o $(BUILD_DIR)/$(APP)_promise.ll
$(CC) $(CC_FLAGS) -emit-llvm src/$(APP)_loop.cpp -S -o $(BUILD_DIR)/$(APP)_loop.ll
$(BUILD_DIR)/%.opt.bc: $(BUILD_DIR)/%.ll
$(OPT) -load LLVMGenVISC.so -genvisc -globaldce $(BUILD_DIR)/$(APP).ll -S -o $(BUILD_DIR)/$(APP).visc.ll
- $(OPT) -load LLVMGenVISC.so -genvisc -globaldce $(BUILD_DIR)/$(APP)_promise.ll -S -o $(BUILD_DIR)/$(APP)_promise.visc.ll
$(OPT) -load LLVMGenVISC.so -genvisc -globaldce $(BUILD_DIR)/$(APP)_loop.ll -S -o $(BUILD_DIR)/$(APP)_loop.visc.ll
$(OPT) $(VISC_OPTFLAGS) $(BUILD_DIR)/$(APP).visc.ll -o $(BUILD_DIR)/$(APP)_cudnn.bc
- #$(OPT) $(VISC_OPTFLAGS2) $(BUILD_DIR)/$(APP)_promise.visc.ll -o $(BUILD_DIR)/$(APP)_promise.bc
- $(OPT) $(VISC_OPTFLAGS3) $(BUILD_DIR)/$(APP)_promise.visc.ll -o $(BUILD_DIR)/$(APP)_wrapperapi.bc
$(OPT) $(VISC_OPTFLAGS3) $(BUILD_DIR)/$(APP)_loop.visc.ll -o $(BUILD_DIR)/$(APP)_loop_wrapperapi.bc
$(OPT) $(VISC_PRED_OPTFLAGS3) $(BUILD_DIR)/$(APP)_promise.visc.ll -o $(BUILD_DIR)/$(APP)_pred_wrapperapi.bc
$(OPT) $(VISC_PRED_OPTFLAGS3) $(BUILD_DIR)/$(APP)_loop.visc.ll -o $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_cudnn.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_cudnn_linked.bc
- #$(LLVM_LINK) $(BUILD_DIR)/$(APP)_promise.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_promise_linked.bc
- $(LLVM_LINK) $(BUILD_DIR)/$(APP)_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_wrapperapi_linked.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_loop_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_loop_wrapperapi_linked.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_pred_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_pred_wrapperapi_linked.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi_linked.bc
$(CC) $(BUILD_DIR)/$(APP)_cudnn_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_cudnn_linked $(LINKER_FLAGS)
- #$(CC) $(BUILD_DIR)/$(APP)_promise_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_promise_linked $(LINKER_FLAGS)
- $(CC) $(BUILD_DIR)/$(APP)_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_wrapperapi_linked $(LINKER_FLAGS)
$(CC) $(BUILD_DIR)/$(APP)_loop_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_loop_wrapperapi_linked $(LINKER_FLAGS)
$(CC) $(BUILD_DIR)/$(APP)_pred_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_pred_wrapperapi_linked $(LINKER_FLAGS)
$(CC) $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi_linked $(LINKER_FLAGS)
diff --git a/llvm/test/VISC/DNN_Benchmarks/benchmarks/vgg16_imagenet/Makefile b/llvm/test/VISC/DNN_Benchmarks/benchmarks/vgg16_imagenet/Makefile
index 8875f2eddf39493321fb706b9eec45880819da16..f5e4d17b622deca90a98745c60df7586fd973435 100644
--- a/llvm/test/VISC/DNN_Benchmarks/benchmarks/vgg16_imagenet/Makefile
+++ b/llvm/test/VISC/DNN_Benchmarks/benchmarks/vgg16_imagenet/Makefile
@@ -56,29 +56,43 @@ default: $(BUILD_DIR) $(TARGET)
$(BUILD_DIR)/%.ll: $(SRC_DIR)/%.cpp
$(CC) $(CC_FLAGS) -emit-llvm src/$(APP).cpp -S -o $(BUILD_DIR)/$(APP).ll
+<<<<<<< HEAD
$(CC) $(CC_FLAGS) -emit-llvm src/$(APP)_promise.cpp -S -o $(BUILD_DIR)/$(APP)_promise.ll
+=======
+>>>>>>> 2a31471f4de89edd9180689d139be7ca65b0df08
$(CC) $(CC_FLAGS) -emit-llvm src/$(APP)_loop.cpp -S -o $(BUILD_DIR)/$(APP)_loop.ll
$(BUILD_DIR)/%.opt.bc: $(BUILD_DIR)/%.ll
$(OPT) -load LLVMGenVISC.so -genvisc -globaldce $(BUILD_DIR)/$(APP).ll -S -o $(BUILD_DIR)/$(APP).visc.ll
+<<<<<<< HEAD
$(OPT) -load LLVMGenVISC.so -genvisc -globaldce $(BUILD_DIR)/$(APP)_promise.ll -S -o $(BUILD_DIR)/$(APP)_promise.visc.ll
$(OPT) -load LLVMGenVISC.so -genvisc -globaldce $(BUILD_DIR)/$(APP)_loop.ll -S -o $(BUILD_DIR)/$(APP)_loop.visc.ll
$(OPT) $(VISC_OPTFLAGS) $(BUILD_DIR)/$(APP).visc.ll -o $(BUILD_DIR)/$(APP)_cudnn.bc
#$(OPT) $(VISC_OPTFLAGS2) $(BUILD_DIR)/$(APP)_promise.visc.ll -o $(BUILD_DIR)/$(APP)_promise.bc
$(OPT) $(VISC_OPTFLAGS3) $(BUILD_DIR)/$(APP)_promise.visc.ll -o $(BUILD_DIR)/$(APP)_wrapperapi.bc
+=======
+ $(OPT) -load LLVMGenVISC.so -genvisc -globaldce $(BUILD_DIR)/$(APP)_loop.ll -S -o $(BUILD_DIR)/$(APP)_loop.visc.ll
+ $(OPT) $(VISC_OPTFLAGS) $(BUILD_DIR)/$(APP).visc.ll -o $(BUILD_DIR)/$(APP)_cudnn.bc
+>>>>>>> 2a31471f4de89edd9180689d139be7ca65b0df08
$(OPT) $(VISC_OPTFLAGS3) $(BUILD_DIR)/$(APP)_loop.visc.ll -o $(BUILD_DIR)/$(APP)_loop_wrapperapi.bc
$(OPT) $(VISC_PRED_OPTFLAGS3) $(BUILD_DIR)/$(APP)_promise.visc.ll -o $(BUILD_DIR)/$(APP)_pred_wrapperapi.bc
$(OPT) $(VISC_PRED_OPTFLAGS3) $(BUILD_DIR)/$(APP)_loop.visc.ll -o $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_cudnn.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_cudnn_linked.bc
+<<<<<<< HEAD
#$(LLVM_LINK) $(BUILD_DIR)/$(APP)_promise.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_promise_linked.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_wrapperapi_linked.bc
+=======
+>>>>>>> 2a31471f4de89edd9180689d139be7ca65b0df08
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_loop_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_loop_wrapperapi_linked.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_pred_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_pred_wrapperapi_linked.bc
$(LLVM_LINK) $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi.bc $(VISC_RT_PATH) -o $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi_linked.bc
$(CC) $(BUILD_DIR)/$(APP)_cudnn_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_cudnn_linked $(LINKER_FLAGS)
+<<<<<<< HEAD
#$(CC) $(BUILD_DIR)/$(APP)_promise_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_promise_linked $(LINKER_FLAGS)
$(CC) $(BUILD_DIR)/$(APP)_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_wrapperapi_linked $(LINKER_FLAGS)
+=======
+>>>>>>> 2a31471f4de89edd9180689d139be7ca65b0df08
$(CC) $(BUILD_DIR)/$(APP)_loop_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_loop_wrapperapi_linked $(LINKER_FLAGS)
$(CC) $(BUILD_DIR)/$(APP)_pred_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_pred_wrapperapi_linked $(LINKER_FLAGS)
$(CC) $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi_linked.bc $(TENSOR_LIB_DIR) $(PROFILER_LIB_DIR) $(SOC_SIMULATOR_LIB_DIR) -o $(BUILD_DIR)/$(APP)_pred_loop_wrapperapi_linked $(LINKER_FLAGS)
diff --git a/llvm/test/VISC/DNN_Benchmarks/common/include/visc.h b/llvm/test/VISC/DNN_Benchmarks/common/include/visc.h
index 978f29e85a9aae8e310f528bc8fa7630c85d8542..9bf77b6ed654a6b309a8685acd51882ea1d95a5c 100644
--- a/llvm/test/VISC/DNN_Benchmarks/common/include/visc.h
+++ b/llvm/test/VISC/DNN_Benchmarks/common/include/visc.h
@@ -112,6 +112,10 @@ void* __visc__tensor_map3(void*, void*, void*, void*);
void* __visc__tensor_cosineT(void*);
void* __visc__tensor_stencil(void*);
+// New HPVM intrinsic for Setting Node ID
+void* __visc__node_id(int);
+
+
#include <unistd.h>
long get_global_id(int);