diff options
author | Chris Wilson <chris@chris-wilson.co.uk> | 2015-07-22 23:15:39 +0100 |
---|---|---|
committer | Martin Peres <martin.peres@linux.intel.com> | 2015-07-27 16:31:40 +0300 |
commit | 1da30df32d3bfdb2b2e37e7f7b36bf027d22ef89 (patch) | |
tree | c513ec8b27bae81fa63e32aec19229af3ffbc396 /tests.d | |
parent | 2667720508d4c04d086625b4597e1010455e8d8c (diff) |
Report gem_exec_nop in terms of ops per second
In keeping with the bigger is better mantra style for reporting
inter-run variances, convert the gem_exec_nop summary from a time in us
to operations per second.
Diffstat (limited to 'tests.d')
-rw-r--r-- | tests.d/gem/gem_exec_nop.test | 13 |
1 files changed, 5 insertions, 8 deletions
diff --git a/tests.d/gem/gem_exec_nop.test b/tests.d/gem/gem_exec_nop.test index 70f3b6f..f7dbf9f 100644 --- a/tests.d/gem/gem_exec_nop.test +++ b/tests.d/gem/gem_exec_nop.test @@ -1,8 +1,10 @@ test_name=gem_exec_nop test_exec_time=18 -# This outputs a graph of time(N), the time it takes to execute N empty batches. +# This outputs a graph of time(N), the us it takes to execute N empty batches. # e.g. 110 59 34 22 16 13 11 10 10 10 9 4 1.4 1.4 1.2 0.9 0.8 +# As a summary, print the number of nop/s (so that it matches the bigger is +# better motif of fps). # # Ideally we want to run this per ring, # gem_exec_nop:rcs, gem_exec_nop:bcs, gem_exec_nop:vcs @@ -11,10 +13,5 @@ test_exec_time=18 [ -e $IGT_BENCHMARKS/gem_exec_nop ] || return 1 sudo -n true || return 1 -gem_exec_nop_run() { - sudo $IGT_BENCHMARKS/gem_exec_nop -e 1 -r $1 -} - -gem_exec_nop_process() { - echo "${@: -1}" -} +gem_exec_nop_run() { sudo $IGT_BENCHMARKS/gem_exec_nop -e 1 -r $1; } +gem_exec_nop_process() { bc -l <<< "1000000 / ${@: -1}"; } |