do
cat "${test_input:-/dev/null}" | \
/usr/bin/time --format="%e" "$@" 2>&1 | \
+ tee "${1}.out" | \
tail -n 1
done | avgstddev > "$tmp"
read avg1 stddev1 < "$tmp"
for p in 1 2 4
do
- echo "$VG --tool=exp-drd $@ -p$p"
+ echo "$VG --tool=exp-drd --check-stack-var=yes $@ -p$p"
for ((i=0;i<3;i++))
do
cat "${test_input:-/dev/null}" | \
/usr/bin/time --format="%e" $VG --tool=exp-drd "$@" -p$p 2>&1 | \
+ tee "${1}-drd-with-stack-var-${p}.out" | \
+ tail -n 1
+ done | avgstddev > "$tmp"
+ read avg2 stddev2 < "$tmp"
+ echo "Average time: ${avg2} +/- ${stddev2} seconds"
+ awk "END{print "'"'"Ratio ="'"'", ${avg2}/${avg1}, "'"'"+/-"'"'", ${avg2}/${avg1}*(${stddev1}/${avg1}+${stddev2}/${avg2})}" </dev/null
+
+ echo "$VG --tool=exp-drd --check-stack-var=no $@ -p$p"
+ for ((i=0;i<3;i++))
+ do
+ cat "${test_input:-/dev/null}" | \
+ /usr/bin/time --format="%e" $VG --tool=exp-drd "$@" -p$p 2>&1 | \
+ tee "${1}-drd-without-stack-var-${p}.out" | \
tail -n 1
done | avgstddev > "$tmp"
read avg2 stddev2 < "$tmp"
exit 1
fi
-# Results: (-p1) (-p2) (-p3) (-p4) ITC (-p4) ITC (-p4)
-# original w/ filter
-# .........................................................................
-# Cholesky 40 47 ? 82 239 82
-# FFT 16 17 N/A 47 90 41
-# LU, contiguous blocks 39 41 ? 45 428 128
-# LU, non-contiguous blocks 39 41 ? 49 428 128
-# Ocean, contiguous partitions 17 19 N/A 25 90 28
-# Ocean, non-continguous partns 18 21 N/A 30 90 28
-# Radiosity 78 78 ? 78 485 163
-# Radix 10 12 ? 15 222 56
-# Raytrace 56 56 ? 56 172 53
-# Water-n2 34 34 ? 34 189 39
-# Water-sp 33 33 ? 33 183 34
-#
-# Hardware: dual-core Intel Xeon 5130, 1.995 MHz, 4 MB L2 cache, 4 GB RAM.
-# Software: Ubuntu 7.10 server, 64-bit (includes gcc 4.1.3).
-
-cache_size=$(get_cache_size)
+##############################################################################
+# Results: (-p1) (-p2) (-p4) ITC (-p4) ITC (-p4)
+# original w/ filter
+# ............................................................................
+# Cholesky 40 47 82 239 82
+# FFT 16 17 47 90 41
+# LU, contiguous blocks 39 41 45 428 128
+# LU, non-contiguous blocks 39 41 49 428 128
+# Ocean, contiguous partitions 17 19 25 90 28
+# Ocean, non-continguous partns 18 21 30 90 28
+# Radiosity 78 78 78 485 163
+# Radix 10 12 15 222 56
+# Raytrace 56 56 56 172 53
+# Water-n2 34 34 34 189 39
+# Water-sp 33 33 33 183 34
+# ............................................................................
+# Hardware: dual-core Intel Xeon 5130, 2.0 GHz, 4 MB L2 cache, 4 GB RAM.
+# Software: Ubuntu 7.10 server, 64-bit, gcc 4.1.3.
+##############################################################################
+# Results: native DRD (-p4) DRD (-p4) ITC (-p4) ITC (-p4)
+# time original w/ filter original w/ filter
+# ............................................................................
+# Cholesky 0.21 74 76 239 82
+# FFT 0.16 12 11 90 41
+# LU, contiguous blocks 0.57 49 49 428 128
+# LU, non-contiguous blocks 0.60 52 52 428 128
+# Ocean, contiguous partitions 14.42 24 24 90 28
+# Ocean, non-continguous partns 0.21 30 30 90 28
+# Radiosity 2.35 75 76 485 163
+# Radix 2.80 15 16 222 56
+# Raytrace 90.79 55 54 172 53
+# Water-n2 0.15 27 27 189 39
+# Water-sp 0.15 26 26 183 34
+# ............................................................................
+# Hardware: dual-core Intel Core2 Duo E6750, 2.66 GHz, 4 MB L2 cache, 2 GB RAM.
+# Software: openSUSE 10.3, 64-bit, gcc 4.2.1.
+##############################################################################
+
+cache_size=$(($(get_cache_size)/2))
log2_cache_size=$(log2 ${cache_size})
# Cholesky
do
gzip -cd <$f >${f%.Z}
done
- run_test ../CHOLESKY -C${cache_size} -n1024 tk29.O
+ run_test ../CHOLESKY -C${cache_size} tk29.O
)
# FFT
-run_test ${SPLASH2}/codes/kernels/fft/FFT -t -l${log2_cache_size} -m20
+run_test ${SPLASH2}/codes/kernels/fft/FFT -t -l${log2_cache_size} -m18
# LU, contiguous blocks.
run_test ${SPLASH2}/codes/kernels/lu/contiguous_blocks/LU -n1024