benchtests/README

   1 Using the glibc microbenchmark suite
   2 ====================================
   3
   4 The glibc microbenchmark suite automatically generates code for specified
   5 functions, builds and calls them repeatedly for given inputs to give some
   6 basic performance properties of the function.
   7
   8 Running the benchmark:
   9 =====================
  10
  11 The benchmark needs python 2.7 or later in addition to the
  12 dependencies required to build the GNU C Library.  One may run the
  13 benchmark by invoking make as follows:
  14
  15   $ make bench
  16
  17 This runs each function for 10 seconds and appends its output to
  18 benchtests/bench.out.  To ensure that the tests are rebuilt, one could run:
  19
  20   $ make bench-clean
  21
  22 The duration of each test can be configured setting the BENCH_DURATION variable
  23 in the call to make.  One should run `make bench-clean' before changing
  24 BENCH_DURATION.
  25
  26   $ make BENCH_DURATION=1 bench
  27
  28 The benchmark suite does function call measurements using architecture-specific
  29 high precision timing instructions whenever available.  When such support is
  30 not available, it uses clock_gettime (CLOCK_MONOTONIC).
  31
  32 On x86 processors, RDTSCP instruction provides more precise timing data
  33 than RDTSC instruction.  All x86 processors since 2010 support RDTSCP
  34 instruction.  One can force the benchmark to use RDTSCP by invoking make
  35 as follows:
  36
  37   $ make USE_RDTSCP=1 bench
  38
  39 One must run `make bench-clean' before changing the measurement method.
  40
  41 Running benchmarks on another target:
  42 ====================================
  43
  44 If the target where you want to run benchmarks is not capable of building the
  45 code or you're cross-building, you could build and execute the benchmark in
  46 separate steps.  On the build system run:
  47
  48   $ make bench-build
  49
  50 and then copy the source and build directories to the target and run the
  51 benchmarks from the build directory as usual:
  52
  53   $ make bench
  54
  55 make sure the copy preserves timestamps by using either rsync or scp -p
  56 otherwise the above command may try to build the benchmark again.  Benchmarks
  57 that require generated code to be executed during the build are skipped when
  58 cross-building.
  59
  60 Running subsets of benchmarks:
  61 ==============================
  62
  63 To run only a subset of benchmarks, one may invoke make as follows
  64
  65   $ make bench BENCHSET="bench-pthread bench-math malloc-thread"
  66
  67 where BENCHSET may be a space-separated list of the following values:
  68
  69     bench-math
  70     bench-pthread
  71     bench-string
  72     string-benchset
  73     wcsmbs-benchset
  74     stdlib-benchset
  75     stdio-common-benchset
  76     math-benchset
  77     malloc-thread
  78
  79 Adding a function to benchtests:
  80 ===============================
  81
  82 If the name of the function is `foo', then the following procedure should allow
  83 one to add `foo' to the bench tests:
  84
  85 - Append the function name to the bench variable in the Makefile.
  86
  87 - Make a file called `foo-inputs` to provide the definition and input for the
  88   function.  The file should have some directives telling the parser script
  89   about the function and then one input per line.  Directives are lines that
  90   have a special meaning for the parser and they begin with two hashes '##'.
  91   The following directives are recognized:
  92
  93   - args: This should be assigned a colon separated list of types of the input
  94     arguments.  This directive may be skipped if the function does not take any
  95     inputs.  One may identify output arguments by nesting them in <>.  The
  96     generator will create variables to get outputs from the calling function.
  97   - ret: This should be assigned the type that the function returns.  This
  98     directive may be skipped if the function does not return a value.
  99   - includes: This should be assigned a comma-separated list of headers that
 100     need to be included to provide declarations for the function and types it
 101     may need (specifically, this includes using "#include <header>").
 102   - include-sources: This should be assigned a comma-separated list of source
 103     files that need to be included to provide definitions of global variables
 104     and functions (specifically, this includes using "#include "source").
 105     See pthread_once-inputs and pthreads_once-source.c for an example of how
 106     to use this to benchmark a function that needs state across several calls.
 107   - init: Name of an initializer function to call to initialize the benchtest.
 108   - name: See following section for instructions on how to use this directive.
 109
 110   Lines beginning with a single hash '#' are treated as comments.  See
 111   pow-inputs for an example of an input file.
 112
 113 Multiple execution units per function:
 114 =====================================
 115
 116 Some functions have distinct performance characteristics for different input
 117 domains and it may be necessary to measure those separately.  For example, some
 118 math functions perform computations at different levels of precision (64-bit vs
 119 240-bit vs 768-bit) and mixing them does not give a very useful picture of the
 120 performance of these functions.  One could separate inputs for these domains in
 121 the same file by using the `name' directive that looks something like this:
 122
 123   ##name: 240bit
 124
 125 See the pow-inputs file for an example of what such a partitioned input file
 126 would look like.
 127
 128 It is also possible to measure throughput of a (partial) trace extracted from
 129 a real workload.  In this case the whole trace is iterated over multiple times
 130 rather than repeating every input multiple times.  This can be done via:
 131
 132   ##name: workload-<name>
 133
 134 Benchmark Sets:
 135 ==============
 136
 137 In addition to standard benchmarking of functions, one may also generate
 138 custom outputs for a set of functions.  This is currently used by string
 139 function benchmarks where the aim is to compare performance between
 140 implementations at various alignments and for various sizes.
 141
 142 To add a benchset for `foo':
 143
 144 - Add `foo' to the benchset variable.
 145 - Write your bench-foo.c that prints out the measurements to stdout.
 146 - On execution, a bench-foo.out is created in $(objpfx) with the contents of
 147   stdout.
 148
 149 Reading String Benchmark Results:
 150 ================================
 151
 152 Some of the string benchmark results are now in JSON to make it easier to read
 153 in scripts.  Use the benchtests/compare_strings.py script to show the results
 154 in a tabular format, generate graphs and more. Run
 155
 156     benchtests/scripts/compare_strings.py -h
 157
 158 for usage information.