Merge commit '7d9102373d60bd159920abfe96d636420afedd67'

README.adoc

@@ -1210,7 +1210,11 @@ You can make QEMU or gem5 <<gem5-vs-qemu-performance,run faster>> by passing ena
 ./run -K
 ....
 
-but it was broken in gem5 with pending patches: https://www.mail-archive.com/gem5-users@gem5.org/msg15046.html
+but it was broken in gem5 with pending patches: https://www.mail-archive.com/gem5-users@gem5.org/msg15046.html It fails immediately on:
+
+....
+panic: KVM: Failed to enter virtualized mode (hw reason: 0x80000021)
+....
 
 KVM uses the link:https://en.wikipedia.org/wiki/Kernel-based_Virtual_Machine[KVM Linux kernel feature] of the host to run most instructions natively.
 
@@ -2401,17 +2405,17 @@ arch=aarch64
 
 # Generate a checkpoint after Linux boots.
 # The boot takes a while, be patient young Padawan.
-printf 'm5 exit' >data/readfile
+printf 'm5 exit' > data/readfile
 ./run -a "$arch" -g -E 'm5 checkpoint;m5 readfile > a.sh;sh a.sh'
 
 # Restore the checkpoint, and run the benchmark with parameter 1.000.
 # We skip the boot completely, saving time!
-printf 'm5 resetstats;dhrystone 1000;m5 exit' >data/readfile
+printf 'm5 resetstats;dhrystone 1000;m5 exit' > data/readfile
 ./run -a "$arch" -g -- -r 1
 ./gem5-stat -a "$arch"
 
 # Now with another parameter 10.000.
-printf 'm5 resetstats;dhrystone 10000;m5 exit' >data/readfile
+printf 'm5 resetstats;dhrystone 10000;m5 exit' > data/readfile
 ./run -a "$arch" -g -- -r 1
 ./gem5-stat -a "$arch"
 ....
@@ -2550,7 +2554,39 @@ So we take a performance measurement approach instead:
 cat out/aarch64/gem5/bench-cache.txt
 ....
 
-TODO: sort out HPI, and then paste results here, why the `--cpu-type=HPI` there always generates a `switch_cpu`, even if the original run was also on HPI?
+which gives:
+
+[options="header",cols="3*>"]
+|===
+|test size
+|cache size
+|cycle count
+
+|1000
+|1k
+|52.432.956
+
+|1000
+|1M
+|6.328.325
+
+|10000
+|1k
+|141.637.834
+
+|10000
+|1M
+|16.969.057
+
+|100000
+|1k
+|1.034.500.724
+
+|100000
+|1M
+|121.728.035
+
+|===
 
 ===== gem5 memory latency
 
@@ -3024,7 +3060,7 @@ Internals:
 * <<m5>> is a guest utility present inside the gem5 tree which we cross-compiled and installed into the guest
 
 [[gem5-restore-new-scrip]]
-===== gem5 checkpoint restore and run a different script
+==== gem5 checkpoint restore and run a different script
 
 You want to automate running several tests from a single pristine post-boot state.
 
@@ -3036,11 +3072,11 @@ The problem is that after the checkpoint, the memory and disk states are fixed,
 There is however one loophole: <<m5-readfile>>, which reads whatever is present on the host, so we can do it like:
 
 ....
-printf 'echo "setup run";m5 exit' >data/readfile
+printf 'echo "setup run";m5 exit' > data/readfile
 ./run -a aarch64 -g -E 'm5 checkpoint;m5 readfile > a.sh;sh a.sh'
-printf 'echo "first benchmark";m5 exit' >data/readfile
+printf 'echo "first benchmark";m5 exit' > data/readfile
 ./run -a aarch64 -g -- -r 1
-printf 'echo "second benchmark";m5 exit' >data/readfile
+printf 'echo "second benchmark";m5 exit' > data/readfile
 ./run -a aarch64 -g -- -r 1
 ....
 

gem5-bench-cache

@@ -1,11 +1,15 @@
 #!/usr/bin/env bash
 set -eu
 . common
-while getopts a:gh OPT; do
+generate_checkpoints=true
+while getopts a:C OPT; do
   case "$OPT" in
     a)
       arch="$OPTARG"
       ;;
+    C)
+      generate_checkpoints=false
+      ;;
   esac
 done
 shift "$(($OPTIND - 1))"
@@ -13,51 +17,80 @@ shift "$(($OPTIND - 1))"
 # Vars
 set_common_vars "$arch" true
 cmd="./run -a $arch -g"
-cpt="-E 'm5 checkpoint;m5 readfile > a.sh;sh a.sh'"
+cache_small='--caches --l2cache --l1d_size=1024   --l1i_size=1024   --l2_size=1024   --l3_size=1024  '
-cache_small='--caches --l2cache --l1d_size=1024   --l1i_size=1024   --l2_size=1024   --l3_size=1024'
 cache_large='--caches --l2cache --l1d_size=1024kB --l1i_size=1024kB --l2_size=1024kB --l3_size=1024kB'
 result_file="${gem5_out_dir}/bench-cache.txt"
 
 bench() (
   cmd="$1"
+  printf 'cmd ' >> "$result_file"
   ./eeval "$cmd" "$result_file"
-  ./gem5-ncycles -a "$arch" >> "$result_file"
+  {
+    printf 'cycles '
+    ./gem5-stat -a "$arch"
+    # RESTORE_INVESTIGATION
+    #cycles_switch="$(./gem5-stat -a "$arch" system.switch_cpus.numCycles)"
+    #if [ -n "$cycles_switch" ]; then
+    #  printf "cycles_switch ${cycles_switch}\n"
+    #fi
+  } >> "$result_file"
 )
 
 bench-all() (
-  bench "$cmd -- -r 1"
+  bench "${cmd} -- -r 1 ${cache_small} --cpu-type=HPI --restore-with-cpu=HPI"
-  bench "$cmd -- -r 2 $cache_small"
+  bench "${cmd} -- -r 1 ${cache_large} --cpu-type=HPI --restore-with-cpu=HPI"
-  bench "$cmd -- -r 3 $cache_large"
+  # RESTORE_INVESTIGATION
-  bench "$cmd -- -r 4 $cache_small --cpu-type=HPI"
+  # These were mostly to investigate what happens on restore:
-  bench "$cmd -- -r 5 $cache_large --cpu-type=HPI"
+  # https://stackoverflow.com/questions/49011096/how-to-switch-cpu-models-in-gem5-after-restoring-a-checkpoint-and-then-observe-t
+  #bench "$cmd -- -r 1"
+  #bench "$cmd -- -r 1 $cache_small"
+  #bench "$cmd -- -r 1 $cache_large"
+  #bench "$cmd -- -r 2 $cache_small"
+  #bench "$cmd -- -r 3 $cache_large"
+  #bench "$cmd -- -r 4 $cache_small --cpu-type=HPI"
+  #bench "$cmd -- -r 5 $cache_large --cpu-type=HPI"
+  ## Restore from AtomicSimpleCPU to HPI.
+  #bench "$cmd -- -r 2 $cache_small --cpu-type=HPI --restore-with-cpu=HPI"
+  #bench "$cmd -- -r 3 $cache_large --cpu-type=HPI --restore-with-cpu=HPI"
+  #bench "$cmd -- -r 2 $cache_small --restore-with-cpu=HPI"
+  #bench "$cmd -- -r 3 $cache_large --restore-with-cpu=HPI"
+  #bench "$cmd -- -r 2 $cache_small --cpu-type=HPI"
+  #bench "$cmd -- -r 3 $cache_large --cpu-type=HPI"
+  ## Restore HPI with different cache sizes and see if it is used.
+  #bench "$cmd -- -r 4 $cache_large --cpu-type=HPI"
+  #bench "$cmd -- -r 5 $cache_small --cpu-type=HPI"
+  #bench "$cmd -- -r 2 $cache_large --cpu-type=HPI"
+  #bench "$cmd -- -r 3 $cache_small --cpu-type=HPI"
 )
 
-# Files.
-rm -rf \
-  "$result_file" \
-  "${m5out_dir}/cpt.*" \
-;
 
-# Create the checkpoints after the kernel boot.
+if "$generate_checkpoints"; then
-printf 'm5 exit' >"${readfile_file}"
+  # Create the checkpoints after the kernel boot.
-./eeval "$cmd $cpt"
+  rm -rf "${m5out_dir}/cpt.*";
-./eeval "$cmd $cpt -- $cache_small"
+  printf 'm5 exit' >"${readfile_file}"
-./eeval "$cmd $cpt -- $cache_large"
+  cpt_cmd="-E 'm5 checkpoint;m5 readfile > a.sh;sh a.sh'"
-./eeval "$cmd $cpt -- $cache_small --cpu-type=HPI"
+  # 1
-./eeval "$cmd $cpt -- $cache_large --cpu-type=HPI"
+  ./eeval "$cmd $cpt_cmd"
+  # RESTORE_INVESTIGATION
+  ## 2
+  #./eeval "$cmd $cpt_cmd -- $cache_small"
+  ## 3
+  #./eeval "$cmd $cpt_cmd -- $cache_large"
+  ## 4
+  #./eeval "$cmd $cpt_cmd -- $cache_small --cpu-type=HPI"
+  ## 5
+  #./eeval "$cmd $cpt_cmd -- $cache_large --cpu-type=HPI"
+fi
 
-# dhrystone 1.000
+# Restore and run benchmarks.
+rm -f "$result_file"
 printf '#!/bin/sh
 m5 resetstats
-dhrystone 1000
+dhrystone XXX
 m5 exit
 ' >"${readfile_file}"
-bench-all
+for n in 1000 10000 100000; do
-
+  printf "n ${n}\n" >> "$result_file"
-# dhrystone 10.000
+  sed -Ei "s/^dhrystone .*/dhrystone ${n}/" "${readfile_file}"
-sed -Ei 's/^dhrystone .*/dhrystone 10000/' "${readfile_file}"
+  bench-all
-bench-all
+done
-
-# dhrystone 100.000
-sed -Ei 's/^dhrystone .*/dhrystone 100000/' "${readfile_file}"
-bench-all