APP_SRCS... main.c BUILD_MODEL_SQ8BIT/networkKernels.c /gap_sdk/libs/gap_lib/img_io/ImgIO.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Activation_SQ8.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Bias_Linear_SQ8.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Conv_SQ8.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Pooling_SQ8.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Conv_DW_SQ8.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_MatAlgebra_SQ8.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_SoftMax_SQ8.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_AT_Misc.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/RNN_SQ8.c /gap_sdk/tools/nntool//autotiler/kernels/norm_transpose.c /gap_sdk/tools/nntool//autotiler/kernels/copy.c APP_CFLAGS... -g -O3 -w -mno-memcpy -fno-tree-loop-distribute-patterns -I. -I/gap_sdk/libs/gap_lib/include -I/gap_sdk/tools/autotiler_v3/Emulation -I/gap_sdk/tools/autotiler_v3/Autotiler -I/gap_sdk/tools/autotiler_v3/CNN_Libraries -I/gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8 -I/gap_sdk/tools/nntool//autotiler/kernels -IBUILD_MODEL_SQ8BIT -DAT_MODEL_PREFIX=network -DAT_INPUT_HEIGHT=200 -DAT_INPUT_WIDTH=200 -DAT_INPUT_COLORS=1 -DSTACK_SIZE=6096 -DSLAVE_STACK_SIZE=1024 -DAT_IMAGE=/module/data/images/frame_2.pgm -DPERF -DMODEL_ID= -DFREQ_FC=250 -DFREQ_CL=175 -DAT_CONSTRUCT=networkCNN_Construct -DAT_DESTRUCT=networkCNN_Destruct -DAT_CNN=networkCNN -DAT_L3_ADDR=network_L3_Flash common/model_rules.mk:28: warning: overriding recipe for target 'BUILD_MODEL_SQ8BIT' common/model_rules.mk:28: warning: ignoring old recipe for target 'BUILD_MODEL_SQ8BIT' common/model_rules.mk:31: warning: overriding recipe for target 'BUILD_MODEL_SQ8BIT/network.onnx' common/model_rules.mk:31: warning: ignoring old recipe for target 'BUILD_MODEL_SQ8BIT/network.onnx' common/model_rules.mk:35: warning: overriding recipe for target 'BUILD_MODEL_SQ8BIT/network.json' common/model_rules.mk:35: warning: ignoring old recipe for target 'BUILD_MODEL_SQ8BIT/network.json' common/model_rules.mk:40: warning: overriding recipe for target 'nntool_model_evaluation' common/model_rules.mk:40: warning: ignoring old recipe for target 'nntool_model_evaluation' common/model_rules.mk:48: warning: overriding recipe for target 'nntool_output/networkModel.c' common/model_rules.mk:48: warning: ignoring old recipe for target 'nntool_output/networkModel.c' common/model_rules.mk:55: warning: overriding recipe for target 'BUILD_MODEL_SQ8BIT/GenTile' common/model_rules.mk:55: warning: ignoring old recipe for target 'BUILD_MODEL_SQ8BIT/GenTile' common/model_rules.mk:62: warning: overriding recipe for target 'BUILD_MODEL_SQ8BIT/networkKernels.c' common/model_rules.mk:62: warning: ignoring old recipe for target 'BUILD_MODEL_SQ8BIT/networkKernels.c' common/model_rules.mk:69: warning: overriding recipe for target 'clean_model' common/model_rules.mk:69: warning: ignoring old recipe for target 'clean_model' APP_SRCS... main.c BUILD_MODEL_SQ8BIT/networkKernels.c /gap_sdk/libs/gap_lib/img_io/ImgIO.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Activation_SQ8.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Bias_Linear_SQ8.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Conv_SQ8.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Pooling_SQ8.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Conv_DW_SQ8.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_MatAlgebra_SQ8.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_SoftMax_SQ8.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_AT_Misc.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/RNN_SQ8.c /gap_sdk/tools/nntool//autotiler/kernels/norm_transpose.c /gap_sdk/tools/nntool//autotiler/kernels/copy.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Activation_SQ8.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Bias_Linear_SQ8.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Conv_SQ8.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Pooling_SQ8.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Conv_DW_SQ8.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_MatAlgebra_SQ8.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_SoftMax_SQ8.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_AT_Misc.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/RNN_SQ8.c /gap_sdk/tools/nntool//autotiler/kernels/norm_transpose.c /gap_sdk/tools/nntool//autotiler/kernels/copy.c main.c BUILD_MODEL_SQ8BIT/networkKernels.c /gap_sdk/libs/gap_lib/img_io/ImgIO.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Activation_SQ8.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Bias_Linear_SQ8.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Conv_SQ8.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Pooling_SQ8.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Conv_DW_SQ8.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_MatAlgebra_SQ8.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_SoftMax_SQ8.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_AT_Misc.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/RNN_SQ8.c /gap_sdk/tools/nntool//autotiler/kernels/norm_transpose.c /gap_sdk/tools/nntool//autotiler/kernels/copy.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Activation_SQ8.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Bias_Linear_SQ8.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Conv_SQ8.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Pooling_SQ8.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Conv_DW_SQ8.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_MatAlgebra_SQ8.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_SoftMax_SQ8.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_AT_Misc.c /gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/RNN_SQ8.c /gap_sdk/tools/nntool//autotiler/kernels/norm_transpose.c /gap_sdk/tools/nntool//autotiler/kernels/copy.c APP_CFLAGS... -g -O3 -w -mno-memcpy -fno-tree-loop-distribute-patterns -I. -I/gap_sdk/libs/gap_lib/include -I/gap_sdk/tools/autotiler_v3/Emulation -I/gap_sdk/tools/autotiler_v3/Autotiler -I/gap_sdk/tools/autotiler_v3/CNN_Libraries -I/gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8 -I/gap_sdk/tools/nntool//autotiler/kernels -IBUILD_MODEL_SQ8BIT -DAT_MODEL_PREFIX=network -DAT_INPUT_HEIGHT=200 -DAT_INPUT_WIDTH=200 -DAT_INPUT_COLORS=1 -DSTACK_SIZE=6096 -DSLAVE_STACK_SIZE=1024 -DAT_IMAGE=/module/data/images/frame_2.pgm -DPERF -DMODEL_ID= -DFREQ_FC=250 -DFREQ_CL=175 -DAT_CONSTRUCT=networkCNN_Construct -DAT_DESTRUCT=networkCNN_Destruct -DAT_CNN=networkCNN -DAT_L3_ADDR=network_L3_Flash -g -O3 -w -mno-memcpy -fno-tree-loop-distribute-patterns -I. -I/gap_sdk/libs/gap_lib/include -I/gap_sdk/tools/autotiler_v3/Emulation -I/gap_sdk/tools/autotiler_v3/Autotiler -I/gap_sdk/tools/autotiler_v3/CNN_Libraries -I/gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8 -I/gap_sdk/tools/nntool//autotiler/kernels -IBUILD_MODEL_SQ8BIT -DAT_MODEL_PREFIX=network -DAT_INPUT_HEIGHT=200 -DAT_INPUT_WIDTH=200 -DAT_INPUT_COLORS=1 -DSTACK_SIZE=6096 -DSLAVE_STACK_SIZE=1024 -DAT_IMAGE=/module/data/images/frame_2.pgm -DPERF -DMODEL_ID= -DFREQ_FC=250 -DFREQ_CL=175 -DAT_CONSTRUCT=networkCNN_Construct -DAT_DESTRUCT=networkCNN_Destruct -DAT_CNN=networkCNN -DAT_L3_ADDR=network_L3_Flash /gap_sdk/tools/rules/pulp_rules.mk:199: warning: overriding recipe for target '/module/data/BUILD/GAP8_V2/GCC_RISCV' /gap_sdk/tools/rules/pulp_rules.mk:199: warning: ignoring old recipe for target '/module/data/BUILD/GAP8_V2/GCC_RISCV' /gap_sdk/tools/rules/pulp_rules.mk:204: warning: overriding recipe for target '/module/data/BUILD/GAP8_V2/GCC_RISCV/main.o' /gap_sdk/tools/rules/pulp_rules.mk:204: warning: ignoring old recipe for target '/module/data/BUILD/GAP8_V2/GCC_RISCV/main.o' /gap_sdk/tools/rules/pulp_rules.mk:204: warning: overriding recipe for target '/module/data/BUILD/GAP8_V2/GCC_RISCV/BUILD_MODEL_SQ8BIT/networkKernels.o' /gap_sdk/tools/rules/pulp_rules.mk:204: warning: ignoring old recipe for target '/module/data/BUILD/GAP8_V2/GCC_RISCV/BUILD_MODEL_SQ8BIT/networkKernels.o' /gap_sdk/tools/rules/pulp_rules.mk:204: warning: overriding recipe for target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/libs/gap_lib/img_io/ImgIO.o' /gap_sdk/tools/rules/pulp_rules.mk:204: warning: ignoring old recipe for target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/libs/gap_lib/img_io/ImgIO.o' /gap_sdk/tools/rules/pulp_rules.mk:204: warning: overriding recipe for target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Activation_SQ8.o' /gap_sdk/tools/rules/pulp_rules.mk:204: warning: ignoring old recipe for target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Activation_SQ8.o' /gap_sdk/tools/rules/pulp_rules.mk:204: warning: overriding recipe for target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Bias_Linear_SQ8.o' /gap_sdk/tools/rules/pulp_rules.mk:204: warning: ignoring old recipe for target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Bias_Linear_SQ8.o' /gap_sdk/tools/rules/pulp_rules.mk:204: warning: overriding recipe for target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Conv_SQ8.o' /gap_sdk/tools/rules/pulp_rules.mk:204: warning: ignoring old recipe for target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Conv_SQ8.o' /gap_sdk/tools/rules/pulp_rules.mk:204: warning: overriding recipe for target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Pooling_SQ8.o' /gap_sdk/tools/rules/pulp_rules.mk:204: warning: ignoring old recipe for target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Pooling_SQ8.o' /gap_sdk/tools/rules/pulp_rules.mk:204: warning: overriding recipe for target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Conv_DW_SQ8.o' /gap_sdk/tools/rules/pulp_rules.mk:204: warning: ignoring old recipe for target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Conv_DW_SQ8.o' /gap_sdk/tools/rules/pulp_rules.mk:204: warning: overriding recipe for target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_MatAlgebra_SQ8.o' /gap_sdk/tools/rules/pulp_rules.mk:204: warning: ignoring old recipe for target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_MatAlgebra_SQ8.o' /gap_sdk/tools/rules/pulp_rules.mk:204: warning: overriding recipe for target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_SoftMax_SQ8.o' /gap_sdk/tools/rules/pulp_rules.mk:204: warning: ignoring old recipe for target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_SoftMax_SQ8.o' /gap_sdk/tools/rules/pulp_rules.mk:204: warning: overriding recipe for target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_AT_Misc.o' /gap_sdk/tools/rules/pulp_rules.mk:204: warning: ignoring old recipe for target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_AT_Misc.o' /gap_sdk/tools/rules/pulp_rules.mk:204: warning: overriding recipe for target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/RNN_SQ8.o' /gap_sdk/tools/rules/pulp_rules.mk:204: warning: ignoring old recipe for target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/RNN_SQ8.o' /gap_sdk/tools/rules/pulp_rules.mk:204: warning: overriding recipe for target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/nntool//autotiler/kernels/norm_transpose.o' /gap_sdk/tools/rules/pulp_rules.mk:204: warning: ignoring old recipe for target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/nntool//autotiler/kernels/norm_transpose.o' /gap_sdk/tools/rules/pulp_rules.mk:204: warning: overriding recipe for target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/nntool//autotiler/kernels/copy.o' /gap_sdk/tools/rules/pulp_rules.mk:204: warning: ignoring old recipe for target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/nntool//autotiler/kernels/copy.o' /gap_sdk/tools/rules/pulp_rules.mk:203: target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Activation_SQ8.o' given more than once in the same rule /gap_sdk/tools/rules/pulp_rules.mk:203: target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Bias_Linear_SQ8.o' given more than once in the same rule /gap_sdk/tools/rules/pulp_rules.mk:203: target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Conv_SQ8.o' given more than once in the same rule /gap_sdk/tools/rules/pulp_rules.mk:203: target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Pooling_SQ8.o' given more than once in the same rule /gap_sdk/tools/rules/pulp_rules.mk:203: target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Conv_DW_SQ8.o' given more than once in the same rule /gap_sdk/tools/rules/pulp_rules.mk:203: target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_MatAlgebra_SQ8.o' given more than once in the same rule /gap_sdk/tools/rules/pulp_rules.mk:203: target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_SoftMax_SQ8.o' given more than once in the same rule /gap_sdk/tools/rules/pulp_rules.mk:203: target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_AT_Misc.o' given more than once in the same rule /gap_sdk/tools/rules/pulp_rules.mk:203: target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/RNN_SQ8.o' given more than once in the same rule /gap_sdk/tools/rules/pulp_rules.mk:203: target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/nntool//autotiler/kernels/norm_transpose.o' given more than once in the same rule /gap_sdk/tools/rules/pulp_rules.mk:203: target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/nntool//autotiler/kernels/copy.o' given more than once in the same rule /gap_sdk/tools/rules/pulp_rules.mk:203: target '/module/data/BUILD/GAP8_V2/GCC_RISCV/main.o' given more than once in the same rule /gap_sdk/tools/rules/pulp_rules.mk:203: target '/module/data/BUILD/GAP8_V2/GCC_RISCV/BUILD_MODEL_SQ8BIT/networkKernels.o' given more than once in the same rule /gap_sdk/tools/rules/pulp_rules.mk:203: target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/libs/gap_lib/img_io/ImgIO.o' given more than once in the same rule /gap_sdk/tools/rules/pulp_rules.mk:203: target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Activation_SQ8.o' given more than once in the same rule /gap_sdk/tools/rules/pulp_rules.mk:203: target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Bias_Linear_SQ8.o' given more than once in the same rule /gap_sdk/tools/rules/pulp_rules.mk:203: target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Conv_SQ8.o' given more than once in the same rule /gap_sdk/tools/rules/pulp_rules.mk:203: target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Pooling_SQ8.o' given more than once in the same rule /gap_sdk/tools/rules/pulp_rules.mk:203: target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Conv_DW_SQ8.o' given more than once in the same rule /gap_sdk/tools/rules/pulp_rules.mk:203: target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_MatAlgebra_SQ8.o' given more than once in the same rule /gap_sdk/tools/rules/pulp_rules.mk:203: target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_SoftMax_SQ8.o' given more than once in the same rule /gap_sdk/tools/rules/pulp_rules.mk:203: target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_AT_Misc.o' given more than once in the same rule /gap_sdk/tools/rules/pulp_rules.mk:203: target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/RNN_SQ8.o' given more than once in the same rule /gap_sdk/tools/rules/pulp_rules.mk:203: target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/nntool//autotiler/kernels/norm_transpose.o' given more than once in the same rule /gap_sdk/tools/rules/pulp_rules.mk:203: target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/nntool//autotiler/kernels/copy.o' given more than once in the same rule /gap_sdk/tools/rules/pulp_rules.mk:203: target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Activation_SQ8.o' given more than once in the same rule /gap_sdk/tools/rules/pulp_rules.mk:203: target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Bias_Linear_SQ8.o' given more than once in the same rule /gap_sdk/tools/rules/pulp_rules.mk:203: target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Conv_SQ8.o' given more than once in the same rule /gap_sdk/tools/rules/pulp_rules.mk:203: target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Pooling_SQ8.o' given more than once in the same rule /gap_sdk/tools/rules/pulp_rules.mk:203: target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_Conv_DW_SQ8.o' given more than once in the same rule /gap_sdk/tools/rules/pulp_rules.mk:203: target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_MatAlgebra_SQ8.o' given more than once in the same rule /gap_sdk/tools/rules/pulp_rules.mk:203: target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_SoftMax_SQ8.o' given more than once in the same rule /gap_sdk/tools/rules/pulp_rules.mk:203: target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/CNN_AT_Misc.o' given more than once in the same rule /gap_sdk/tools/rules/pulp_rules.mk:203: target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8/RNN_SQ8.o' given more than once in the same rule /gap_sdk/tools/rules/pulp_rules.mk:203: target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/nntool//autotiler/kernels/norm_transpose.o' given more than once in the same rule /gap_sdk/tools/rules/pulp_rules.mk:203: target '/module/data/BUILD/GAP8_V2/GCC_RISCV//gap_sdk/tools/nntool//autotiler/kernels/copy.o' given more than once in the same rule /gap_sdk/tools/rules/pulp_rules.mk:208: warning: overriding recipe for target '/module/data/BUILD/GAP8_V2/GCC_RISCV/pulp-os/conf.o' /gap_sdk/tools/rules/pulp_rules.mk:208: warning: ignoring old recipe for target '/module/data/BUILD/GAP8_V2/GCC_RISCV/pulp-os/conf.o' /gap_sdk/tools/rules/pulp_rules.mk:212: warning: overriding recipe for target '/module/data/BUILD/GAP8_V2/GCC_RISCV/imagenet' /gap_sdk/tools/rules/pulp_rules.mk:212: warning: ignoring old recipe for target '/module/data/BUILD/GAP8_V2/GCC_RISCV/imagenet' /gap_sdk/tools/rules/pulp_rules.mk:229: warning: overriding recipe for target 'flash' /gap_sdk/tools/rules/pulp_rules.mk:229: warning: ignoring old recipe for target 'flash' /gap_sdk/tools/rules/pulp_rules.mk:232: warning: overriding recipe for target 'flash_fs' /gap_sdk/tools/rules/pulp_rules.mk:232: warning: ignoring old recipe for target 'flash_fs' /gap_sdk/tools/rules/pulp_rules.mk:235: warning: overriding recipe for target 'image' /gap_sdk/tools/rules/pulp_rules.mk:235: warning: ignoring old recipe for target 'image' /gap_sdk/tools/rules/pulp_rules.mk:238: warning: overriding recipe for target 'run.prepare' /gap_sdk/tools/rules/pulp_rules.mk:238: warning: ignoring old recipe for target 'run.prepare' /gap_sdk/tools/rules/pulp_rules.mk:241: warning: overriding recipe for target 'run.exec' /gap_sdk/tools/rules/pulp_rules.mk:241: warning: ignoring old recipe for target 'run.exec' /gap_sdk/tools/rules/pulp_rules.mk:244: warning: overriding recipe for target 'run' /gap_sdk/tools/rules/pulp_rules.mk:244: warning: ignoring old recipe for target 'run' /gap_sdk/tools/rules/pulp_rules.mk:248: warning: overriding recipe for target 'profiler' /gap_sdk/tools/rules/pulp_rules.mk:248: warning: ignoring old recipe for target 'profiler' /gap_sdk/tools/rules/pulp_rules.mk:255: warning: overriding recipe for target '/module/data/BUILD/GAP8_V2/GCC_RISCV/imagenet.s' /gap_sdk/tools/rules/pulp_rules.mk:255: warning: ignoring old recipe for target '/module/data/BUILD/GAP8_V2/GCC_RISCV/imagenet.s' /gap_sdk/tools/rules/pulp_rules.mk:262: warning: overriding recipe for target 'version' /gap_sdk/tools/rules/pulp_rules.mk:262: warning: ignoring old recipe for target 'version' rm -f BUILD_MODEL_SQ8BIT/GenTile rm -f -rf BUILD_MODEL_SQ8BIT rm -f -rf nntool_output mkdir BUILD_MODEL_SQ8BIT cp nntool_input/models_onnx/model_original_himax.onnx BUILD_MODEL_SQ8BIT/network.onnx echo "GENERATING NNTOOL STATE FILE" GENERATING NNTOOL STATE FILE echo BUILD_MODEL_SQ8BIT BUILD_MODEL_SQ8BIT nntool -s nntool_input/nntool_scripts/nntool_script_deployment BUILD_MODEL_SQ8BIT/network.onnx -q -s is not a recognized command, alias, or macro nntool_input/nntool_scripts/nntool_script_deployment is not a recognized command, alias, or macro BUILD_MODEL_SQ8BIT/network.onnx is not a recognized command, alias, or macro -q is not a recognized command, alias, or macro open - opening graph file BUILD_MODEL_SQ8BIT/network.onnx onnx - unable to determine batch dimension. if the graph fails to import properly set it to 1 or a variable. nngraph - update graph dimensions set_aliases - looking for aliased edges nngraph - calculate liveness debug - was: False now: True adjust_order - adding transposes to correct tensor order for AT kernels set_aliases - looking for aliased edges eliminate_transposes - eliminating unnecessary transposes eliminate_transposes - search for transposes eliminate_transposes - no transposes to eliminate found nngraph - update graph dimensions set_aliases - looking for aliased edges nngraph - calculate liveness eliminate_transposes - no further transpose sequences found set_aliases - looking for aliased edges nngraph - adjusted order nngraph - update graph dimensions set_aliases - looking for aliased edges nngraph - calculate liveness matcher - fusions - start remove_relus matcher - fusions - start remove_noops matcher - fusions - start fuse_external_bias_sq8 matcher - fusions - start fuse_pad matcher - fusions - start unused_concats matcher - fusions - start gather_to_split gather_to_split - gathers from Gemm_31[0] converted to a split matcher - fusions - gather_to_split modified graph nngraph - update graph dimensions set_aliases - looking for aliased edges nngraph - calculate liveness matcher - fusions - start find_missing_quantization matcher - fusions - start rnn_reverse nngraph - update graph dimensions set_aliases - looking for aliased edges nngraph - calculate liveness matcher - fusions - start rnn_unpack matcher - fusions - start match_far_hsigmoid matcher - fusions - start match_close_hsigmoid matcher - fusions - start expand_transposes matcher - fusions - start move_pooling_scale8 matcher - fusions - start move_activations_scale8 move_node_up - Node Sigmoid_36 cannot be moved matcher - fusions - start fuse_gap_convs matcher - fusions - start match_conv_active_pool matcher - fusions - start match_conv_pool_active matcher - fusions - start match_conv_active match_gap_conv - fusing nodes Conv_2,Clip_4 match_gap_conv - fusing nodes Conv_5,Clip_7 match_gap_conv - fusing nodes Conv_8,Clip_9 match_gap_conv - fusing nodes Conv_11,Clip_13 match_gap_conv - fusing nodes Conv_14,Clip_16 match_gap_conv - fusing nodes Conv_17,Clip_18 match_gap_conv - fusing nodes Conv_20,Clip_22 match_gap_conv - fusing nodes Conv_23,Clip_25 match_gap_conv - fusing nodes Conv_26,Clip_27 matcher - fusions - match_conv_active modified graph nngraph - update graph dimensions set_aliases - looking for aliased edges nngraph - calculate liveness matcher - fusions - start match_conv_pool match_gap_conv - fusing nodes Conv_0,MaxPool_1 matcher - fusions - match_conv_pool modified graph nngraph - update graph dimensions set_aliases - looking for aliased edges nngraph - calculate liveness matcher - fusions - start fuse_gap_linear matcher - fusions - start fuse_op_activation_scale8 matcher - fusions - fuse_op_activation_scale8 modified graph nngraph - update graph dimensions set_aliases - looking for aliased edges nngraph - calculate liveness matcher - fusions - start propagate_softmax_sym_qrec matcher - fusions - start equalize_sm_concats matcher - fusions - start filter_bigger_than_input matcher - fusions - start insert_copies nngraph - update graph dimensions set_aliases - looking for aliased edges nngraph - calculate liveness matcher - fusions - start propagate_up_rnn_in_qs nngraph - update graph dimensions set_aliases - looking for aliased edges nngraph - calculate liveness input_norm_func - was: '' now: 'x:x/255' aquant - input file ['nntool_input/quantization_files/1479425441232704425.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/1479425443582977575.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/frame_00164.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/1479425448633847925.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/frame_00079.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/frame_00135.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/frame_00064.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/frame_00014.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/frame_00143.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/1479425443382879593.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/1479425443833012205.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/frame_00037.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/frame_00033.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/frame_00191.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/frame_00194.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/1479425443432903742.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/frame_00013.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/frame_00136.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/frame_00078.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/1479425441282730750.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/frame_00192.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/frame_00193.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/1479425443782898808.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/1479425448133784412.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/frame_00139.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/frame_00001.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/1479425451134304083.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/frame_00016.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/frame_00036.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/frame_00151.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/frame_00038.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/1479425444633286162.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/frame_00017.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/1479425447083596868.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/frame_00035.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/1479425447133467484.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/frame_00058.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/frame_00195.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/frame_00121.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/frame_00034.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/frame_00140.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/frame_00015.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/1479425443682982187.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/frame_00131.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/frame_00018.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/frame_00158.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/1479425441182877835.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/1479425445983503068.jpg'] graph_executer - execute uncached: quantization mode none aquant - input file ['nntool_input/quantization_files/frame_00032.jpg'] graph_executer - execute uncached: quantization mode none aquant - Quantization set. Use qshow command to see it. +------+----------------+------------------+------------------+------------------+----------+----------+-------+-------+ | Step | Name | In | Out | Weights | Bias | Mulbias | Calc | Acc | +======+================+==================+==================+==================+==========+==========+=======+=======+ | 0 | input_1 | | -1.01>chan | Q32.0 | Q32.0 | | | | 402<1.00 | 554<4.23 | n | | | | | +------+----------------+------------------+------------------+------------------+----------+----------+-------+-------+ | 1 | MaxPool_1 | -4.27>chan | Q32.0 | Q32.0 | | | | 554<4.23 | 409<6.00 | n | | | | | +------+----------------+------------------+------------------+------------------+----------+----------+-------+-------+ | 2 | Clip_4 | -6.05>chan | Q32.0 | Q32.0 | | | | 409<6.00 | 409<6.00 | n | | | | | +------+----------------+------------------+------------------+------------------+----------+----------+-------+-------+ | 3 | Clip_7 | -6.05>chan | Q32.0 | Q32.0 | | | | 554<4.23 | 656<4.85 | n | | | | | +------+----------------+------------------+------------------+------------------+----------+----------+-------+-------+ | 4 | Clip_9 | -4.89>chan | Q32.0 | Q32.0 | | | | 409<6.00 | 409<6.00 | n | | | | | +------+----------------+------------------+------------------+------------------+----------+----------+-------+-------+ | 6 | Clip_13 | -6.05>chan | Q32.0 | Q32.0 | | | | 409<6.00 | 409<6.00 | n | | | | | +------+----------------+------------------+------------------+------------------+----------+----------+-------+-------+ | 7 | Clip_16 | -6.05>chan | Q32.0 | Q32.0 | | | | 409<6.00 | 075<4.87 | n | | | | | +------+----------------+------------------+------------------+------------------+----------+----------+-------+-------+ | 8 | Clip_18 | -4.91>chan | Q32.0 | Q32.0 | | | | 409<6.00 | 409<6.00 | n | | | | | +------+----------------+------------------+------------------+------------------+----------+----------+-------+-------+ | 10 | Clip_22 | -6.05>chan | Q32.0 | Q32.0 | | | | 409<6.00 | 456<1.13 | n | | | | | +------+----------------+------------------+------------------+------------------+----------+----------+-------+-------+ | 11 | Clip_25 | -1.14>chan | Q32.0 | Q32.0 | | | | 409<6.00 | 486<0.05 | n | | | | | +------+----------------+------------------+------------------+------------------+----------+----------+-------+-------+ | 12 | Clip_27 | -0.05>chan | Q32.0 | Q32.0 | | | | 912<2.26 | 5582<40.56 | n | | | | | +------+----------------+------------------+------------------+------------------+----------+----------+-------+-------+ | 16 | Gemm_31_split | -40.88 Feat: 1 W: 200, H: 1 Out => Feat: 1, W: 200, H: 1 KerName: CNN_NormBW_shift_fps Nb Oper : 200 ==== Process Tiling For User Kernel: S1_Op_input_1_formatter ======================= S1_Op_input_1_formatter Partition[0] Size = 808 (Min: 0, Max: 816), Fraction: 1.00, Giving: 46736 bytes out of 46736 bytes Kernel: S1_Op_input_1_formatter, Total Raw Memory: 400 fits into L1 memory 46736. Promoting all kernel arguments to initialized buffers. S1_Op_input_1_formatter For Iter Space: 0 Iteration count: 1, Given L1 Memory: 46736, Used L1 Memory: 400, Reusable Memory: 46336, Used L2 Memory: 0 ================================================================================================= InFeat: 1, OutFeat: 32 Conv => W: 200, Pad:[2,1] PadT:[2,1] => Wc: 100, Filter:[5,5] => H: 200, Pad:[2,1] PadT:[2,1] => Hc: 100 Pool => Wc: 100, Pad:[0,0] => Wo: 50, Filter:[2,2] => Hc: 100, Pad:[0,0] => Ho: 50 OverlapC: 3 OverlapP: 0 TileCons: 2 UsedIn : [200 x 200] UsedC : [100 x 100] SetBiasKerName: KerParSetBiasB32_SQ8 ConvKerName: KerParConv5x5Stride2_SQ8 DPReductionKerName: KerParReductIO_CC_SQ8 PoolKerName: KerParPool2x2Stride2_SQ8 Nb Oper : 8320000 ==== Process Tiling For User Kernel: S2_Conv2d_32x1x5x5_MaxPool_2x2 ======================= S2_Conv2d_32x1x5x5_MaxPool_2x2 Partition[0] Size = 1290425 (Min: 2000, Max: 1523449), Fraction: 1.00, Giving: 46736 bytes out of 46736 bytes S2_Conv2d_32x1x5x5_MaxPool_2x2 Full buffering on Arg: Bias, was using 256 bytes will require 128 bytes buffer S2_Conv2d_32x1x5x5_MaxPool_2x2 Full buffering on Arg: Scale, was using 64 bytes will require 32 bytes buffer S2_Conv2d_32x1x5x5_MaxPool_2x2 Full buffering on Arg: ScaleN, was using 64 bytes will require 32 bytes buffer S2_Conv2d_32x1x5x5_MaxPool_2x2, TiledSpace: Tile0 Iteration Count: 50 Parametric Space: [D1, M0=32] Parametric Space: [D0, M1=1] In : Ratio: 4.000000, FixDim: 200, VarDim: 7 [ 200], Size: 2800, Total: 2800, Move: 69400 (Decl x 1.735000) L2 * Bias : Ratio: 0.000000, Size: 128, Total: 2928, Move: 128 (Decl x 1.000000) L2 * Scale : Ratio: 0.000000, Size: 32, Total: 2960, Move: 32 (Decl x 1.000000) L2 * ScaleN : Ratio: 0.000000, Size: 32, Total: 2992, Move: 32 (Decl x 1.000000) L2 @ Filter : Ratio: 0.000000, Size: 800, Total: 3792, Move: 800 (Decl x 1.000000) L2 Out : Ratio: 1.000000, FixDim: 50, VarDim: 1 [ 50], Size: 3200, Total: 6992, Move: 80000 (Decl x 1.000000) L2 * ConvOut : Ratio: 2.000000, FixDim: 100, VarDim: 2 [ 100], Size: 25600, Total: 32592, Move: 0 (Decl x 0.000000) L2 * Infos : Ratio: 0.000000, Size: 12, Total: 32604, Move: 9 (Decl x 1.000000) L2 S2_Conv2d_32x1x5x5_MaxPool_2x2 - IterSpace: Tile0 - L1 Memory: 32604, L2Move: 150401, L3Move: 0, Tiling Overhead: 1.242973 S2_Conv2d_32x1x5x5_MaxPool_2x2 Partial buffering on Arg: Filter, From: D0 To: D1. Current is (Par) 1 x [W:1, H:1] x 25 => Partial buffer size is 1600 bytes S2_Conv2d_32x1x5x5_MaxPool_2x2 Found Parametric value for space D1 (Initial: 32, Div: 8) = 32 [32*1 + 0] and space D0 (Initial: 1, Div: 4) = 1 [1*1 + 0], Iteration for Tiled Space: 50 S2_Conv2d_32x1x5x5_MaxPool_2x2 For Iter Space: 0 Iteration count: 50, Given L1 Memory: 46736, Used L1 Memory: 33404, Reusable Memory: 13332, Used L2 Memory: 0 ================================================================================================= InFeat: 32, OutFeat: 32 Conv => W: 50, Pad:[1,0] PadT:[1,0] => Wc: 25, Filter:[3,3] => H: 50, Pad:[1,0] PadT:[1,0] => Hc: 25 Pool => Wc: 25, Pad:[0,0] => Wo: 25, Filter:[1,1] => Hc: 25, Pad:[0,0] => Ho: 25 OverlapC: 1 OverlapP: 0 TileCons: 2 UsedIn : [50 x 50] UsedC : [25 x 25] SetBiasKerName: KerParSetBiasB32_SQ8 ConvKerName: KerParConv3x3Stride2_SQ8 DPReductionKerName: KerParReduct_CC_ReLU_SQ8 Nb Oper : 5780000 ==== Process Tiling For User Kernel: S3_Conv2d_32x32x3x3_Relu6 ======================= S3_Conv2d_32x32x3x3_Relu6 Partition[0] Size = 93801 (Min: 300, Max: 284425), Fraction: 1.00, Giving: 46736 bytes out of 46736 bytes S3_Conv2d_32x32x3x3_Relu6 Full buffering on Arg: Bias, was using 256 bytes will require 128 bytes buffer S3_Conv2d_32x32x3x3_Relu6 Full buffering on Arg: Scale, was using 64 bytes will require 32 bytes buffer S3_Conv2d_32x32x3x3_Relu6 Full buffering on Arg: ScaleN, was using 64 bytes will require 32 bytes buffer S3_Conv2d_32x32x3x3_Relu6, TiledSpace: Tile0 Iteration Count: 7 Parametric Space: [D1, M0=32] Parametric Space: [D0, M1=8] In : Ratio: 2.000000, FixDim: 50, VarDim: 9 [ 50], Size: 7200, Total: 7200, Move: 89600 (Decl x 1.120000) L2 * Bias : Ratio: 0.000000, Size: 128, Total: 7328, Move: 128 (Decl x 1.000000) L2 * Scale : Ratio: 0.000000, Size: 32, Total: 7360, Move: 32 (Decl x 1.000000) L2 * ScaleN : Ratio: 0.000000, Size: 32, Total: 7392, Move: 32 (Decl x 1.000000) L2 @ Filter : Ratio: 0.000000, Size: 9216, Total: 16608, Move: 9216 (Decl x 1.000000) L2 Out : Ratio: 1.000000, FixDim: 25, VarDim: 4 [ 25], Size: 6400, Total: 23008, Move: 20000 (Decl x 1.000000) L2 * ConvOut : Ratio: 1.000000, FixDim: 25, VarDim: 4 [ 25], Size: 12800, Total: 35808, Move: 0 (Decl x 0.000000) L2 * Infos : Ratio: 0.000000, Size: 12, Total: 35820, Move: 9 (Decl x 1.000000) L2 S3_Conv2d_32x32x3x3_Relu6 - IterSpace: Tile0 - L1 Memory: 35820, L2Move: 119017, L3Move: 0, Tiling Overhead: 1.087738 S3_Conv2d_32x32x3x3_Relu6 Partial buffering on Arg: Filter, From: D0 To: D1. Current is (Par) 32 x [W:1, H:1] x 9 => Partial buffer size is 18432 bytes S3_Conv2d_32x32x3x3_Relu6 Found Parametric value for space D1 (Initial: 32, Div: 8) = 32 [32*1 + 0] and space D0 (Initial: 32, Div: 4) = 8 [8*4 + 0], Iteration for Tiled Space: 7 S3_Conv2d_32x32x3x3_Relu6 For Iter Space: 0 Iteration count: 7 (Last one is truncated), Given L1 Memory: 46736, Used L1 Memory: 45036, Reusable Memory: 1700, Used L2 Memory: 0 ================================================================================================= InFeat: 32, OutFeat: 32 Conv => W: 25, Pad:[1,1] PadT:[1,1] => Wc: 25, Filter:[3,3] => H: 25, Pad:[1,1] PadT:[1,1] => Hc: 25 Pool => Wc: 25, Pad:[0,0] => Wo: 25, Filter:[1,1] => Hc: 25, Pad:[0,0] => Ho: 25 OverlapC: 2 OverlapP: 0 TileCons: 1 UsedIn : [25 x 25] UsedC : [25 x 25] SetBiasKerName: KerParSetBiasB32_SQ8 ConvKerName: KerParConv3x3Stride1_SQ8 DPReductionKerName: KerParReduct_CC_ReLU_SQ8 Nb Oper : 5780000 ==== Process Tiling For User Kernel: S4_Conv2d_32x32x3x3_Relu6 ======================= S4_Conv2d_32x32x3x3_Relu6 Partition[0] Size = 87401 (Min: 150, Max: 164425), Fraction: 1.00, Giving: 46736 bytes out of 46736 bytes S4_Conv2d_32x32x3x3_Relu6 Full buffering on Arg: Bias, was using 256 bytes will require 128 bytes buffer S4_Conv2d_32x32x3x3_Relu6 Full buffering on Arg: Scale, was using 64 bytes will require 32 bytes buffer S4_Conv2d_32x32x3x3_Relu6 Full buffering on Arg: ScaleN, was using 64 bytes will require 32 bytes buffer S4_Conv2d_32x32x3x3_Relu6, TiledSpace: Tile0 Iteration Count: 5 Parametric Space: [D1, M0=32] Parametric Space: [D0, M1=8] In : Ratio: 1.000000, FixDim: 25, VarDim: 7 [ 25], Size: 2800, Total: 2800, Move: 26400 (Decl x 1.320000) L2 * Bias : Ratio: 0.000000, Size: 128, Total: 2928, Move: 128 (Decl x 1.000000) L2 * Scale : Ratio: 0.000000, Size: 32, Total: 2960, Move: 32 (Decl x 1.000000) L2 * ScaleN : Ratio: 0.000000, Size: 32, Total: 2992, Move: 32 (Decl x 1.000000) L2 @ Filter : Ratio: 0.000000, Size: 9216, Total: 12208, Move: 9216 (Decl x 1.000000) L2 Out : Ratio: 1.000000, FixDim: 25, VarDim: 5 [ 25], Size: 8000, Total: 20208, Move: 20000 (Decl x 1.000000) L2 * ConvOut : Ratio: 1.000000, FixDim: 25, VarDim: 5 [ 25], Size: 16000, Total: 36208, Move: 0 (Decl x 0.000000) L2 * Infos : Ratio: 0.000000, Size: 12, Total: 36220, Move: 9 (Decl x 1.000000) L2 S4_Conv2d_32x32x3x3_Relu6 - IterSpace: Tile0 - L1 Memory: 36220, L2Move: 55817, L3Move: 0, Tiling Overhead: 1.129510 S4_Conv2d_32x32x3x3_Relu6 Partial buffering on Arg: Filter, From: D0 To: D1. Current is (Par) 32 x [W:1, H:1] x 9 => Partial buffer size is 18432 bytes S4_Conv2d_32x32x3x3_Relu6 Found Parametric value for space D1 (Initial: 32, Div: 8) = 32 [32*1 + 0] and space D0 (Initial: 32, Div: 4) = 8 [8*4 + 0], Iteration for Tiled Space: 5 S4_Conv2d_32x32x3x3_Relu6 For Iter Space: 0 Iteration count: 5, Given L1 Memory: 46736, Used L1 Memory: 45436, Reusable Memory: 1300, Used L2 Memory: 0 ================================================================================================= InFeat: 32, OutFeat: 32 Conv => W: 50, Pad:[0,0] PadT:[0,0] => Wc: 25, Filter:[1,1] => H: 50, Pad:[0,0] PadT:[0,0] => Hc: 25 Pool => Wc: 25, Pad:[0,0] => Wo: 25, Filter:[1,1] => Hc: 25, Pad:[0,0] => Ho: 25 OverlapC: -1 OverlapP: 0 TileCons: 2 UsedIn : [49 x 49] UsedC : [25 x 25] SetBiasKerName: KerParSetBiasB32_SQ8 ConvKerName: KerParConv1x1Stride2_SQ8 DPReductionKerName: KerParReduct_CC_ReLU_SQ8 Nb Oper : 660000 Mapping this convolution to matrix multiplication with small first operand CNN_MatMulSmallM1_SQ8: S5_Conv2d_32x32x1x1_Relu6 In1 => W: 32, H: 32 In2 => W: 2500, H: 32, w: 50, h: 50, Sx: 2, Sy: 2, TileCons: 100 Out => W: 625, H: 32 MatMulKerName: KerParMatMulB32_ReLU_SF_SQ8 MatTransKerName: CNN_TransposeSxSy_fps Act: ReLU Nb Oper : 640000 ==== Process Tiling For User Kernel: S5_Conv2d_32x32x1x1_Relu6 ======================= S5_Conv2d_32x32x1x1_Relu6 Partition[0] Size = 34057 (Min: 6400, Max: 221481), Fraction: 1.00, Giving: 46736 bytes out of 46736 bytes S5_Conv2d_32x32x1x1_Relu6, TiledSpace: Tile0 Iteration Count: 5 * In1 : Ratio: 0.000000, Size: 1024, Total: 1024, Move: 1024 (Decl x 1.000000) L2 In2 : Ratio: 4.000000, FixDim: 32, VarDim: 500 [ 2500], Size: 32000, Total: 33024, Move: 80000 (Decl x 1.000000) L2 * TransIn2 : Ratio: 1.000000, FixDim: 32, VarDim: 125 [ 625], Size: 4000, Total: 37024, Move: 0 (Decl x 0.000000) L2 * Bias : Ratio: 0.000000, Size: 128, Total: 37152, Move: 128 (Decl x 1.000000) L2 Out : Ratio: 1.000000, FixDim: 32, VarDim: 125 [ 625], Size: 8000, Total: 45152, Move: 20000 (Decl x 1.000000) L2 * Scale : Ratio: 0.000000, Size: 32, Total: 45184, Move: 32 (Decl x 1.000000) L2 * ScaleN : Ratio: 0.000000, Size: 32, Total: 45216, Move: 32 (Decl x 1.000000) L2 * Infos : Ratio: 0.000000, Size: 12, Total: 45228, Move: 9 (Decl x 1.000000) L2 S5_Conv2d_32x32x1x1_Relu6 - IterSpace: Tile0 - L1 Memory: 45228, L2Move: 101225, L3Move: 0, Tiling Overhead: 1.000000 S5_Conv2d_32x32x1x1_Relu6 Iteration for Tiled Space: 5 S5_Conv2d_32x32x1x1_Relu6 For Iter Space: 0 Iteration count: 5, Given L1 Memory: 46736, Used L1 Memory: 45228, Reusable Memory: 1508, Used L2 Memory: 0 ================================================================================================= ==== Process Tiling For User Kernel: S6_MatAdd_32x25x25 ======================= S6_MatAdd_32x25x25 Partition[0] Size = 4825 (Min: 0, Max: 120073), Fraction: 1.00, Giving: 46736 bytes out of 46736 bytes S6_MatAdd_32x25x25, TiledSpace: Tile0 Iteration Count: 3 Parametric Space: [D0, M0=32] In1 : Ratio: 1.000000, FixDim: 25, VarDim: 9 [ 25], Size: 14400, Total: 14400, Move: 20000 (Decl x 1.000000) L2 In2 : Ratio: 1.000000, FixDim: 25, VarDim: 9 [ 25], Size: 14400, Total: 28800, Move: 20000 (Decl x 1.000000) L2 Out : Ratio: 1.000000, FixDim: 25, VarDim: 9 [ 25], Size: 14400, Total: 43200, Move: 20000 (Decl x 1.000000) L2 * Infos : Ratio: 0.000000, Size: 12, Total: 43212, Move: 9 (Decl x 1.000000) L2 S6_MatAdd_32x25x25 - IterSpace: Tile0 - L1 Memory: 43212, L2Move: 60009, L3Move: 0, Tiling Overhead: 1.000000 S6_MatAdd_32x25x25 Found Parametric value for space D0 (Initial: 32, Div: 8) = 32 [32*1 + 0], Iteration for Tiled Space: 3 S6_MatAdd_32x25x25 For Iter Space: 0 Iteration count: 3 (Last one is truncated), Given L1 Memory: 46736, Used L1 Memory: 43212, Reusable Memory: 3524, Used L2 Memory: 0 ================================================================================================= InFeat: 32, OutFeat: 64 Conv => W: 25, Pad:[1,1] PadT:[1,1] => Wc: 13, Filter:[3,3] => H: 25, Pad:[1,1] PadT:[1,1] => Hc: 13 Pool => Wc: 13, Pad:[0,0] => Wo: 13, Filter:[1,1] => Hc: 13, Pad:[0,0] => Ho: 13 OverlapC: 1 OverlapP: 0 TileCons: 2 UsedIn : [25 x 25] UsedC : [13 x 13] SetBiasKerName: KerParSetBiasB32_SQ8 ConvKerName: KerParConv3x3Stride2_SQ8 DPReductionKerName: KerParReduct_CC_ReLU_SQ8 Nb Oper : 3125824 ==== Process Tiling For User Kernel: S7_Conv2d_64x32x3x3_Relu6 ======================= S7_Conv2d_64x32x3x3_Relu6 Partition[0] Size = 53353 (Min: 150, Max: 110281), Fraction: 1.00, Giving: 46736 bytes out of 46736 bytes S7_Conv2d_64x32x3x3_Relu6 Full buffering on Arg: Bias, was using 512 bytes will require 256 bytes buffer S7_Conv2d_64x32x3x3_Relu6 Full buffering on Arg: Scale, was using 128 bytes will require 64 bytes buffer S7_Conv2d_64x32x3x3_Relu6 Full buffering on Arg: ScaleN, was using 128 bytes will require 64 bytes buffer S7_Conv2d_64x32x3x3_Relu6, TiledSpace: Tile0 Iteration Count: 13 Parametric Space: [D1, M0=64] Parametric Space: [D0, M1=24] In : Ratio: 2.000000, FixDim: 25, VarDim: 3 [ 25], Size: 3600, Total: 3600, Move: 29600 (Decl x 1.480000) L2 * Bias : Ratio: 0.000000, Size: 256, Total: 3856, Move: 256 (Decl x 1.000000) L2 * Scale : Ratio: 0.000000, Size: 64, Total: 3920, Move: 64 (Decl x 1.000000) L2 * ScaleN : Ratio: 0.000000, Size: 64, Total: 3984, Move: 64 (Decl x 1.000000) L2 @ Filter : Ratio: 0.000000, Size: 18432, Total: 22416, Move: 18432 (Decl x 1.000000) L2 Out : Ratio: 1.000000, FixDim: 13, VarDim: 1 [ 13], Size: 1664, Total: 24080, Move: 10816 (Decl x 1.000000) L2 * ConvOut : Ratio: 1.000000, FixDim: 13, VarDim: 1 [ 13], Size: 3328, Total: 27408, Move: 0 (Decl x 0.000000) L2 * Infos : Ratio: 0.000000, Size: 12, Total: 27420, Move: 9 (Decl x 1.000000) L2 S7_Conv2d_64x32x3x3_Relu6 - IterSpace: Tile0 - L1 Memory: 27420, L2Move: 59241, L3Move: 0, Tiling Overhead: 1.193389 S7_Conv2d_64x32x3x3_Relu6 Partial buffering on Arg: Filter, From: D0 To: D1. Current is (Par) 32 x [W:1, H:1] x 9 => Partial buffer size is 36864 bytes S7_Conv2d_64x32x3x3_Relu6 Found Parametric value for space D1 (Initial: 64, Div: 8) = 64 [64*1 + 0] and space D0 (Initial: 32, Div: 4) = 24 [24*1 + 8], Iteration for Tiled Space: 13 S7_Conv2d_64x32x3x3_Relu6 For Iter Space: 0 Iteration count: 13, Given L1 Memory: 46736, Used L1 Memory: 45852, Reusable Memory: 884, Used L2 Memory: 0 ================================================================================================= InFeat: 64, OutFeat: 64 Conv => W: 13, Pad:[1,1] PadT:[1,1] => Wc: 13, Filter:[3,3] => H: 13, Pad:[1,1] PadT:[1,1] => Hc: 13 Pool => Wc: 13, Pad:[0,0] => Wo: 13, Filter:[1,1] => Hc: 13, Pad:[0,0] => Ho: 13 OverlapC: 2 OverlapP: 0 TileCons: 1 UsedIn : [13 x 13] UsedC : [13 x 13] SetBiasKerName: KerParSetBiasB32_SQ8 ConvKerName: KerParConv3x3Stride1_SQ8 DPReductionKerName: KerParReduct_CC_ReLU_SQ8 Nb Oper : 6240832 ==== Process Tiling For User Kernel: S8_Conv2d_64x64x3x3_Relu6 ======================= S8_Conv2d_64x64x3x3_Relu6 Partition[0] Size = 51881 (Min: 78, Max: 92041), Fraction: 1.00, Giving: 46736 bytes out of 46736 bytes S8_Conv2d_64x64x3x3_Relu6 Full buffering on Arg: Bias, was using 320 bytes will require 256 bytes buffer S8_Conv2d_64x64x3x3_Relu6 Full buffering on Arg: Scale, was using 80 bytes will require 64 bytes buffer S8_Conv2d_64x64x3x3_Relu6 Full buffering on Arg: ScaleN, was using 80 bytes will require 64 bytes buffer S8_Conv2d_64x64x3x3_Relu6, TiledSpace: Tile0 Iteration Count: 1 Parametric Space: [D1, M0=40] Parametric Space: [D0, M1=4] In : Ratio: 1.000000, FixDim: 13, VarDim: 13 [ 13], Size: 1352, Total: 1352, Move: 21632 (Decl x 2.000000) L2 * Bias : Ratio: 0.000000, Size: 256, Total: 1608, Move: 256 (Decl x 1.000000) L2 * Scale : Ratio: 0.000000, Size: 64, Total: 1672, Move: 64 (Decl x 1.000000) L2 * ScaleN : Ratio: 0.000000, Size: 64, Total: 1736, Move: 64 (Decl x 1.000000) L2 Filter : Ratio: 0.000000, Size: 2880, Total: 4616, Move: 36864 (Decl x 1.000000) L2 Out : Ratio: 1.000000, FixDim: 13, VarDim: 13 [ 13], Size: 13520, Total: 18136, Move: 10816 (Decl x 1.000000) L2 * ConvOut : Ratio: 1.000000, FixDim: 13, VarDim: 13 [ 13], Size: 27040, Total: 45176, Move: 0 (Decl x 0.000000) L2 * Infos : Ratio: 0.000000, Size: 12, Total: 45188, Move: 9 (Decl x 1.000000) L2 S8_Conv2d_64x64x3x3_Relu6 - IterSpace: Tile0 - L1 Memory: 45188, L2Move: 69705, L3Move: 0, Tiling Overhead: 1.183668 S8_Conv2d_64x64x3x3_Relu6 Found Parametric value for space D1 (Initial: 64, Div: 8) = 40 [40*1 + 24] and space D0 (Initial: 64, Div: 4) = 4 [4*16 + 0], Iteration for Tiled Space: 1 S8_Conv2d_64x64x3x3_Relu6 For Iter Space: 0 Iteration count: 1, Given L1 Memory: 46736, Used L1 Memory: 45188, Reusable Memory: 1548, Used L2 Memory: 0 ================================================================================================= InFeat: 32, OutFeat: 64 Conv => W: 25, Pad:[0,0] PadT:[0,0] => Wc: 13, Filter:[1,1] => H: 25, Pad:[0,0] PadT:[0,0] => Hc: 13 Pool => Wc: 13, Pad:[0,0] => Wo: 13, Filter:[1,1] => Hc: 13, Pad:[0,0] => Ho: 13 OverlapC: -1 OverlapP: 0 TileCons: 2 UsedIn : [25 x 25] UsedC : [13 x 13] SetBiasKerName: KerParSetBiasB32_SQ8 ConvKerName: KerParConv1x1Stride2_SQ8 DPReductionKerName: KerParReduct_CC_ReLU_SQ8 Nb Oper : 356928 Mapping this convolution to matrix multiplication with small first operand CNN_MatMulSmallM1_SQ8: S9_Conv2d_64x32x1x1_Relu6 In1 => W: 32, H: 64 In2 => W: 625, H: 32, w: 25, h: 25, Sx: 2, Sy: 2, TileCons: 50 Out => W: 169, H: 64 MatMulKerName: KerParMatMulB32_ReLU_SF_SQ8 MatTransKerName: CNN_TransposeSxSy_fps Act: ReLU Nb Oper : 346112 ==== Process Tiling For User Kernel: S9_Conv2d_64x32x1x1_Relu6 ======================= S9_Conv2d_64x32x1x1_Relu6 Partition[0] Size = 17481 (Min: 3200, Max: 69737), Fraction: 1.00, Giving: 46736 bytes out of 46736 bytes Kernel: S9_Conv2d_64x32x1x1_Relu6, Total Raw Memory: 38668 fits into L1 memory 46736. Promoting all kernel arguments to initialized buffers. S9_Conv2d_64x32x1x1_Relu6 For Iter Space: 0 Iteration count: 1, Given L1 Memory: 46736, Used L1 Memory: 38668, Reusable Memory: 8068, Used L2 Memory: 0 ================================================================================================= ==== Process Tiling For User Kernel: S10_MatAdd_64x13x13 ======================= S10_MatAdd_64x13x13 Partition[0] Size = 5017 (Min: 0, Max: 64969), Fraction: 1.00, Giving: 46736 bytes out of 46736 bytes Kernel: S10_MatAdd_64x13x13, Total Raw Memory: 32460 fits into L1 memory 46736. Promoting all kernel arguments to initialized buffers. S10_MatAdd_64x13x13 For Iter Space: 0 Iteration count: 1, Given L1 Memory: 46736, Used L1 Memory: 32460, Reusable Memory: 14276, Used L2 Memory: 0 ================================================================================================= InFeat: 64, OutFeat: 128 Conv => W: 13, Pad:[1,1] PadT:[1,1] => Wc: 7, Filter:[3,3] => H: 13, Pad:[1,1] PadT:[1,1] => Hc: 7 Pool => Wc: 7, Pad:[0,0] => Wo: 7, Filter:[1,1] => Hc: 7, Pad:[0,0] => Ho: 7 OverlapC: 1 OverlapP: 0 TileCons: 2 UsedIn : [13 x 13] UsedC : [7 x 7] SetBiasKerName: KerParSetBiasB32_SQ8 ConvKerName: KerParConv3x3Stride2_SQ8 DPReductionKerName: KerParReduct_CC_ReLU_SQ8 Nb Oper : 3618944 ==== Process Tiling For User Kernel: S11_Conv2d_128x64x3x3_Relu6 ======================= S11_Conv2d_128x64x3x3_Relu6 Partition[0] Size = 37545 (Min: 78, Max: 66697), Fraction: 1.00, Giving: 46736 bytes out of 46736 bytes S11_Conv2d_128x64x3x3_Relu6 Full buffering on Arg: Bias, was using 896 bytes will require 512 bytes buffer S11_Conv2d_128x64x3x3_Relu6 Full buffering on Arg: Scale, was using 224 bytes will require 128 bytes buffer S11_Conv2d_128x64x3x3_Relu6 Full buffering on Arg: ScaleN, was using 224 bytes will require 128 bytes buffer S11_Conv2d_128x64x3x3_Relu6, TiledSpace: Tile0 Iteration Count: 1 Parametric Space: [D1, M0=112] Parametric Space: [D0, M1=4] In : Ratio: 2.000000, FixDim: 13, VarDim: 13 [ 13], Size: 1352, Total: 1352, Move: 21632 (Decl x 2.000000) L2 * Bias : Ratio: 0.000000, Size: 512, Total: 1864, Move: 512 (Decl x 1.000000) L2 * Scale : Ratio: 0.000000, Size: 128, Total: 1992, Move: 128 (Decl x 1.000000) L2 * ScaleN : Ratio: 0.000000, Size: 128, Total: 2120, Move: 128 (Decl x 1.000000) L2 Filter : Ratio: 0.000000, Size: 8064, Total: 10184, Move: 73728 (Decl x 1.000000) L2 Out : Ratio: 1.000000, FixDim: 7, VarDim: 7 [ 7], Size: 10976, Total: 21160, Move: 6272 (Decl x 1.000000) L2 * ConvOut : Ratio: 1.000000, FixDim: 7, VarDim: 7 [ 7], Size: 21952, Total: 43112, Move: 0 (Decl x 0.000000) L2 * Infos : Ratio: 0.000000, Size: 12, Total: 43124, Move: 9 (Decl x 1.000000) L2 S11_Conv2d_128x64x3x3_Relu6 - IterSpace: Tile0 - L1 Memory: 43124, L2Move: 102409, L3Move: 0, Tiling Overhead: 1.118088 S11_Conv2d_128x64x3x3_Relu6 Found Parametric value for space D1 (Initial: 128, Div: 8) = 112 [112*1 + 16] and space D0 (Initial: 64, Div: 4) = 4 [4*16 + 0], Iteration for Tiled Space: 1 S11_Conv2d_128x64x3x3_Relu6 For Iter Space: 0 Iteration count: 1, Given L1 Memory: 46736, Used L1 Memory: 43124, Reusable Memory: 3612, Used L2 Memory: 0 ================================================================================================= InFeat: 128, OutFeat: 128 Conv => W: 7, Pad:[1,1] PadT:[1,1] => Wc: 7, Filter:[3,3] => H: 7, Pad:[1,1] PadT:[1,1] => Hc: 7 Pool => Wc: 7, Pad:[0,0] => Wo: 7, Filter:[1,1] => Hc: 7, Pad:[0,0] => Ho: 7 OverlapC: 2 OverlapP: 0 TileCons: 1 UsedIn : [7 x 7] UsedC : [7 x 7] SetBiasKerName: KerParSetBiasB32_SQ8 ConvKerName: KerParConv3x3Stride1_SQ8 DPReductionKerName: KerParReduct_CC_ReLU_SQ8 Nb Oper : 7231616 ==== Process Tiling For User Kernel: S12_Conv2d_128x128x3x3_Relu6 ======================= S12_Conv2d_128x128x3x3_Relu6 Partition[0] Size = 36137 (Min: 42, Max: 57865), Fraction: 1.00, Giving: 46736 bytes out of 46736 bytes S12_Conv2d_128x128x3x3_Relu6 Full buffering on Arg: Bias, was using 960 bytes will require 512 bytes buffer S12_Conv2d_128x128x3x3_Relu6 Full buffering on Arg: Scale, was using 240 bytes will require 128 bytes buffer S12_Conv2d_128x128x3x3_Relu6 Full buffering on Arg: ScaleN, was using 240 bytes will require 128 bytes buffer S12_Conv2d_128x128x3x3_Relu6, TiledSpace: Tile0 Iteration Count: 1 Parametric Space: [D1, M0=120] Parametric Space: [D0, M1=4] In : Ratio: 1.000000, FixDim: 7, VarDim: 7 [ 7], Size: 392, Total: 392, Move: 12544 (Decl x 2.000000) L2 * Bias : Ratio: 0.000000, Size: 512, Total: 904, Move: 512 (Decl x 1.000000) L2 * Scale : Ratio: 0.000000, Size: 128, Total: 1032, Move: 128 (Decl x 1.000000) L2 * ScaleN : Ratio: 0.000000, Size: 128, Total: 1160, Move: 128 (Decl x 1.000000) L2 Filter : Ratio: 0.000000, Size: 8640, Total: 9800, Move: 147456 (Decl x 1.000000) L2 Out : Ratio: 1.000000, FixDim: 7, VarDim: 7 [ 7], Size: 11760, Total: 21560, Move: 6272 (Decl x 1.000000) L2 * ConvOut : Ratio: 1.000000, FixDim: 7, VarDim: 7 [ 7], Size: 23520, Total: 45080, Move: 0 (Decl x 0.000000) L2 * Infos : Ratio: 0.000000, Size: 12, Total: 45092, Move: 9 (Decl x 1.000000) L2 S12_Conv2d_128x128x3x3_Relu6 - IterSpace: Tile0 - L1 Memory: 45092, L2Move: 167049, L3Move: 0, Tiling Overhead: 1.039011 S12_Conv2d_128x128x3x3_Relu6 Found Parametric value for space D1 (Initial: 128, Div: 8) = 120 [120*1 + 8] and space D0 (Initial: 128, Div: 4) = 4 [4*32 + 0], Iteration for Tiled Space: 1 S12_Conv2d_128x128x3x3_Relu6 For Iter Space: 0 Iteration count: 1, Given L1 Memory: 46736, Used L1 Memory: 45092, Reusable Memory: 1644, Used L2 Memory: 0 ================================================================================================= InFeat: 64, OutFeat: 128 Conv => W: 13, Pad:[0,0] PadT:[0,0] => Wc: 7, Filter:[1,1] => H: 13, Pad:[0,0] PadT:[0,0] => Hc: 7 Pool => Wc: 7, Pad:[0,0] => Wo: 7, Filter:[1,1] => Hc: 7, Pad:[0,0] => Ho: 7 OverlapC: -1 OverlapP: 0 TileCons: 2 UsedIn : [13 x 13] UsedC : [7 x 7] SetBiasKerName: KerParSetBiasB32_SQ8 ConvKerName: KerParConv1x1Stride2_SQ8 DPReductionKerName: KerParReduct_CC_ReLU_SQ8 Nb Oper : 407680 Mapping this convolution to matrix multiplication CNN_MatMul_SQ8: S13_Conv2d_128x64x1x1_Relu6 In1 => W: 64, H: 128 In2 => W: 169, H: 64, w: 13, h: 13, Sx: 2, Sy: 2 Out => W: 49, H: 128 => Column first MatMulKerName: KerParMatMulSxSyB32_ReLU_SQ8 ==== Process Tiling For User Kernel: S13_Conv2d_128x64x1x1_Relu6 ======================= S13_Conv2d_128x64x1x1_Relu6 Partition[0] Size = 1824 (Min: 1024, Max: 17280), Fraction: 0.15, Giving: 12358 bytes out of 46736 bytes S13_Conv2d_128x64x1x1_Relu6 Partition[1] Size = 10089 (Min: 3328, Max: 34377), Fraction: 0.85, Giving: 34377 bytes out of 46736 bytes Kernel: S13_Conv2d_128x64x1x1_Relu6, Total Raw Memory: 17164 fits into L1 memory 34377. Promoting all kernel arguments to initialized buffers. S13_Conv2d_128x64x1x1_Relu6 For Iter Space: 1 Iteration count: 1, Given L1 Memory: 34377, Used L1 Memory: 17164, Reusable Memory: 17212, Used L2 Memory: 0 Kernel: S13_Conv2d_128x64x1x1_Relu6, Total Raw Memory: 8960 fits into L1 memory 12358. Promoting all kernel arguments to initialized buffers. S13_Conv2d_128x64x1x1_Relu6 For Iter Space: 0 Iteration count: 1, Given L1 Memory: 12358, Used L1 Memory: 8960, Reusable Memory: 3396, Used L2 Memory: 0 ================================================================================================= ==== Process Tiling For User Kernel: S14_MatAdd_128x7x7_Relu6 ======================= S14_MatAdd_128x7x7_Relu6 Partition[0] Size = 5401 (Min: 0, Max: 37705), Fraction: 1.00, Giving: 46736 bytes out of 46736 bytes Kernel: S14_MatAdd_128x7x7_Relu6, Total Raw Memory: 18828 fits into L1 memory 46736. Promoting all kernel arguments to initialized buffers. S14_MatAdd_128x7x7_Relu6 For Iter Space: 0 Iteration count: 1, Given L1 Memory: 46736, Used L1 Memory: 18828, Reusable Memory: 27908, Used L2 Memory: 0 ================================================================================================= Linear Layer S16_Linear_2x6272, Linear: InDim: 6272, OutDim: 2, Activation: None Linear Kernel: KerParLinearLayerFullFeatB32_SQ8 ==== Process Tiling For User Kernel: S16_Linear_2x6272 ======================= S16_Linear_2x6272 Partition[0] Size = 31425 (Min: 0, Max: 31593), Fraction: 1.00, Giving: 46736 bytes out of 46736 bytes Kernel: S16_Linear_2x6272, Total Raw Memory: 18848 fits into L1 memory 46736. Promoting all kernel arguments to initialized buffers. S16_Linear_2x6272 For Iter Space: 0 Iteration count: 1, Given L1 Memory: 46736, Used L1 Memory: 18848, Reusable Memory: 27888, Used L2 Memory: 0 ================================================================================================= Act => W: 1, Wo: 1 => H: 1, Ho: 1 ActKerName: KerPar_HSigmoid_SQ8 Nb Oper : 1 ==== Process Tiling For User Kernel: S19_Act_Hsigmoid ======================= S19_Act_Hsigmoid Partition[0] Size = 25 (Min: 0, Max: 49), Fraction: 1.00, Giving: 46736 bytes out of 46736 bytes Kernel: S19_Act_Hsigmoid, Total Raw Memory: 20 fits into L1 memory 46736. Promoting all kernel arguments to initialized buffers. S19_Act_Hsigmoid For Iter Space: 0 Iteration count: 1, Given L1 Memory: 46736, Used L1 Memory: 20, Reusable Memory: 46716, Used L2 Memory: 0 ================================================================================================= Error: Graph stacked tensor S16_Output, input tensor Output_1 is also defined as a graph local or graph argument Execution aborted riscv32-unknown-elf-gcc -g -O3 -w -mno-memcpy -fno-tree-loop-distribute-patterns -I. -I/gap_sdk/libs/gap_lib/include -I/gap_sdk/tools/autotiler_v3/Emulation -I/gap_sdk/tools/autotiler_v3/Autotiler -I/gap_sdk/tools/autotiler_v3/CNN_Libraries -I/gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8 -I/gap_sdk/tools/nntool//autotiler/kernels -IBUILD_MODEL_SQ8BIT -DAT_MODEL_PREFIX=network -DAT_INPUT_HEIGHT=200 -DAT_INPUT_WIDTH=200 -DAT_INPUT_COLORS=1 -DSTACK_SIZE=6096 -DSLAVE_STACK_SIZE=1024 -DAT_IMAGE=/module/data/images/frame_2.pgm -DPERF -DMODEL_ID= -DFREQ_FC=250 -DFREQ_CL=175 -DAT_CONSTRUCT=networkCNN_Construct -DAT_DESTRUCT=networkCNN_Destruct -DAT_CNN=networkCNN -DAT_L3_ADDR=network_L3_Flash -g -O3 -w -mno-memcpy -fno-tree-loop-distribute-patterns -I. -I/gap_sdk/libs/gap_lib/include -I/gap_sdk/tools/autotiler_v3/Emulation -I/gap_sdk/tools/autotiler_v3/Autotiler -I/gap_sdk/tools/autotiler_v3/CNN_Libraries -I/gap_sdk/tools/autotiler_v3/CNN_Libraries_SQ8 -I/gap_sdk/tools/nntool//autotiler/kernels -IBUILD_MODEL_SQ8BIT -DAT_MODEL_PREFIX=network -DAT_INPUT_HEIGHT=200 -DAT_INPUT_WIDTH=200 -DAT_INPUT_COLORS=1 -DSTACK_SIZE=6096 -DSLAVE_STACK_SIZE=1024 -DAT_IMAGE=/module/data/images/frame_2.pgm -DPERF -DMODEL_ID= -DFREQ_FC=250 -DFREQ_CL=175 -DAT_CONSTRUCT=networkCNN_Construct -DAT_DESTRUCT=networkCNN_Destruct -DAT_CNN=networkCNN -DAT_L3_ADDR=network_L3_Flash -DCONFIG_AI_DECK -DCONFIG_AI_DECK -DRT_FC_STACK_SIZE=2048 -D__PLATFORM_GVSOC__ -fno-jump-tables -fno-tree-loop-distribute-patterns -fdata-sections -ffunction-sections -mchip=gap8 -mPE=8 -mFC=1 -D__riscv__ -D__GAP__ -D__GAP8__ -DCHIP_VERSION=2 -mnativeomp -mnativeomp -D__pulp__ -DCONFIG_GAP -D__PULP_OS__ -D__pulp__ -DCONFIG_GAP -D__PULP_OS__ -MMD -MP -c main.c -I/gap_sdk/install/GAP8_V2/include -I/gap_sdk/install/GAP8_V2/include/io -I/gap_sdk/install/workstation/include -I/gap_sdk/tools/autotiler_v3/Emulation -I/gap_sdk/tools/autotiler_v3/Emulation -I/gap_sdk/tools/autotiler_v3/Emulation -I/gap_sdk/tools/autotiler_v3/Emulation -include /gap_sdk/install/GAP8_V2/include/rt/chips/gap_rev1/config.h -MD -MF /module/data/BUILD/GAP8_V2/GCC_RISCV/main.d -o /module/data/BUILD/GAP8_V2/GCC_RISCV/main.o main.c:39:10: fatal error: networkKernels.h: No such file or directory #include "networkKernels.h" ^~~~~~~~~~~~~~~~~~ compilation terminated. /gap_sdk/tools/rules/pulp_rules.mk:204: recipe for target '/module/data/BUILD/GAP8_V2/GCC_RISCV/main.o' failed make: *** [/module/data/BUILD/GAP8_V2/GCC_RISCV/main.o] Error 1 ​​​ Shift + Enter 换行