Benchmark cleanup #56

benchmark/results/gen_text.sh

@@ -0,0 +1,118 @@
+#!/bin/bash
+
+ROUTINEs="Rdot"
+
+MP1s="_Float128 _Float64x dd"
+MP2s="mpfr gmp qd"
+
+for _routine in $ROUTINEs; do
+    pos=-1
+    if [ $_routine = "Rgemm" ]; then
+        pos=4
+    fi
+    if [ $_routine = "Rgemv" ]; then
+        pos=3
+    fi
+    if [ $_routine = "Rgetrf" ]; then
+        pos=3
+    fi
+    if [ $_routine = "Rsyrk" ]; then
+        pos=3
+    fi
+    if [ $_routine = "Rpotrf" ]; then
+        pos=2
+    fi
+    if [ $_routine = "Raxpy" ]; then
+        pos=2
+    fi
+
+    if [ $_routine = "Rdot" ]; then
+        pos=2
+    fi
+
+    if [ $pos = -1 ]; then
+        echo "error"
+        exit
+    fi
+    echo "\subsection{{\tt $_routine} benchmarks}"
+    while read line
+    do
+        IFS=, list=(${line})
+        abbriv=${list[0]}
+        corename=${list[1]}
+        _dirname=${list[2]}
+        mkdir -p ~/Desktop/eps
+        (pushd 2022/$_dirname/logs ; grep -v "'double'" ${_routine}1.plt  | grep -v "set title" > ll ; sed -i 's/th 1/th 6/g' ll ; sed -i 's/set terminal pdf/set terminal eps/g' ll ; sed -i 's/plot/set key above\nplot/' ll ; gnuplot ll > ~/Desktop/eps/${_routine}1.${abbriv}.eps ; popd) >& /dev/null
+        (pushd 2022/$_dirname/logs ; grep -v "'double'" ${_routine}2.plt  | grep -v "set title" > ll ; sed -i 's/th 1/th 6/g' ll ; sed -i 's/set terminal pdf/set terminal eps/g' ll ; sed -i 's/plot/set key above\nplot/' ll ; gnuplot ll > ~/Desktop/eps/${_routine}2.${abbriv}.eps ; popd) >& /dev/null
+        echo "%%%%% $corename %%%%"
+        echo "\subsubsection{{\tt $_routine} on ${corename}}"
+        echo "In Figure~\ref{${_routine}1.${abbriv}}, we show the result of {\tt ${_routine}} performance for {\tt \_Float128}, {\tt \_Float64x} and {\tt double-double}, and in Figure~\ref{${_routine}2.${abbriv}} we show the result of {\tt ${_routine}} performance for {\tt MPFR}, {\tt GMP} and {\tt quad-double} on ${corename}."
+        IFS=" "
+        PEAKS=""
+        for _mp1 in $MP1s; do
+            _PEAK=`sort -k${pos}n 2022/$_dirname/logs/log.${_routine}.${_mp1} | tail -1 | awk "{print \\$${pos}}"`
+            if [ `echo "$_PEAK < 100" | bc` = 1 ]; then
+                PEAK=`printf "%3.1f\n" ${_PEAK}`
+            else
+                PEAK=`printf "%3.0f\n" ${_PEAK}`
+            fi
+            PEAKS="$PEAKS ${PEAK} MFlops,"
+        done
+        echo "The peak performances of the reference {\tt ${_routine}}s of {\tt \_Float128}, {\tt \_Float64x} and {\tt double-double} are$PEAKS respectively."
+
+        PEAKS=""
+        for _mp1 in $MP1s; do
+            _PEAK=`sort -k${pos}n 2022/$_dirname/logs/log.${_routine}.${_mp1}_opt | tail -1 | awk "{print \\$${pos}}"`
+            if [ `echo "$_PEAK < 100" | bc` = 1 ]; then
+                PEAK=`printf "%3.1f\n" ${_PEAK}`
+            else
+                PEAK=`printf "%3.0f\n" ${_PEAK}`
+            fi
+            PEAKS="$PEAKS ${PEAK} MFlops,"
+        done
+        echo "The peak performances of simple OpenMP parallelized {\tt ${_routine}}s of {\tt \_Float128}, {\tt \_Float64x} and {\tt double-double} are$PEAKS respectively."
+
+        PEAKS=""
+        for _mp2 in $MP2s; do
+            _PEAK=`sort -k${pos}n 2022/$_dirname/logs/log.${_routine}.${_mp2} | tail -1 | awk "{print \\$${pos}}"`
+            if [ `echo "$_PEAK < 100" | bc` = 1 ]; then
+                PEAK=`printf "%3.1f\n" ${_PEAK}`
+            else
+                PEAK=`printf "%3.0f\n" ${_PEAK}`
+            fi
+            PEAKS="$PEAKS ${PEAK} MFlops,"
+        done
+        echo "The peak performances of the reference {\tt ${_routine}}s of {\tt MPFR 512bit}, {\tt GMP 512bit} and {\tt quad-double} are$PEAKS respectively."
+
+        PEAKS=""
+        for _mp2 in $MP2s; do
+            _PEAK=`sort -k${pos}n 2022/$_dirname/logs/log.${_routine}.${_mp2}_opt | tail -1 | awk "{print \\$${pos}}"`
+            if [ `echo "$_PEAK < 100" | bc` = 1 ]; then
+                PEAK=`printf "%3.1f\n" ${_PEAK}`
+            else
+                PEAK=`printf "%3.0f\n" ${_PEAK}`
+            fi
+            PEAKS="$PEAKS ${PEAK} MFlops,"
+        done
+        echo "The peak performances of simple OpenMP parallelized {\tt MPFR 512bit}, {\tt GMP 512bit} and {\tt quad-double} are$PEAKS respectively.\\\\"
+
+
+        echo "\begin{figure}"
+        echo "\caption{ {\tt ${_routine}} performance on ${corename} for {\tt \_Float128}, {\tt \_Float64x} and {\tt double-double} with/without simple OpenMP acceleration. }" 
+        echo "\label{${_routine}1.${abbriv}}"
+        echo "\begin{center}"
+        echo "\includegraphics{${_routine}1.${abbriv}.eps}"
+        echo "\end{center}"
+        echo "\end{figure}"
+
+        echo "\begin{figure}"
+        echo "\caption{{\tt ${_routine}} performance on ${corename} for {\tt MPFR 512bit}, {\tt GMP 512bit} and {\tt quad-double} with/without simple OpenMP acceleration. }"
+        echo "\label{${_routine}2.${abbriv}}"
+        echo "\begin{center}"
+        echo "\includegraphics{${_routine}2.${abbriv}.eps}"
+        echo "\end{center}"
+        echo "\end{figure}"
+        echo ""
+        echo ""
+    done < table_abbriv.txt
+done

benchmark/results/table_abbriv.txt

@@ -0,0 +1,4 @@
+A,AMD Ryzen 3970X,x86_64-pc-linux-gnu-Ryzen-Threadripper-3970X-RTX3080
+B,Intel Xeon E5 2623,x86_64-pc-linux-gnu-Xeon-E5-2623-TeslaV100
+C,Intel Core i5-8500B,x86_64-apple-darwin20.6.0-Corei5-8500B
+G,Raspberry Pi4 ARM Cortex A72,aarch64-unknown-linux-gnu-raspberry-pi4