Commit 0dfc6228 authored by Marc's avatar Marc
Browse files

Added versions directory with executables for older versions of the module and...

Added versions directory with executables for older versions of the module and additonal information
parent 48c160ca
export CURRENT_DIR=$PWD
if [ -z "$2" ]
then export par1=4
else export par1=$2
fi
cd ../../software/marcmod/leon3/ && make $1 && cd - > /dev/null
Matrix multiplication (saturation)
4x4 -> 3.185
8x8 -> 21.013
16x16 -> 158.917
32x32 -> 1.228.738
120x120-> 69.457.937
Matrix multiplication (integer)
4x4 -> 2.755
8x8 -> 16.425
16x16 -> 119.401
32x32 -> 918.297
120x120-> 72.831.362
Matrix multiplication (transposed + saturation)
4x4 -> 3.102
8x8 -> 20.445
16x16 -> 154.873
32x32 -> 1.196.077
120x120-> 69.457.937
Matrix multiplication (transpose + integer)
4x4 -> 2.603
8x8 -> 15.771
16x16 -> 115.026
32x32 -> 884.536
120x120-> 72.831.362
This source diff could not be displayed because it is too large. You can view the blob instead.
export CURRENT_DIR=$PWD
if [ -z "$2" ]
then export par1=4
else export par1=$2
fi
cd ../../software/marcmod/leon3/ && make $1 && cd - > /dev/null
Matrix multiplication (SIMD)
4x4 -> 811
8x8 -> 3.244
16x16 -> 27.560
32x32 -> 198.012
120x120-> 13.731.096
Matrix multiplication (saturation)
4x4 -> 3.185
8x8 -> 21.013
16x16 -> 158.917
32x32 -> 1.228.738
120x120-> 69.457.937
Matrix multiplication (integer)
4x4 -> 2.755
8x8 -> 16.425
16x16 -> 119.401
32x32 -> 918.297
120x120-> 72.831.362
Matrix multiplication (transposed + saturation)
4x4 -> 3.102
8x8 -> 20.445
16x16 -> 154.873
32x32 -> 1.196.077
120x120-> 69.457.937
Matrix multiplication (transpose + integer)
4x4 -> 2.603
8x8 -> 15.771
16x16 -> 115.026
32x32 -> 884.536
120x120-> 72.831.362
This source diff could not be displayed because it is too large. You can view the blob instead.
Versions directory containing the binary files and experiment results for each leon3-minimal version. Below each version characteristics are explained.
- v0.0: Baseline leon3
- v1.0: First version of a working SIMD module integrated in the leon3 with timing constrains of 10ns met.
Two stages in the module being the second reduction operatins.
All operatins take one cycle.
Mask and configuration parameters hardcoded in WRMSK instruction.
Supports Markdown
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment