Hi, Sergey!

thanks for the patch! LGTM

Sergey

On 12/26/25 12:18, Sergey Kaplun wrote:
> This patch adjusts the aforementioned test to use the benchmark
> framework introduced before. The default arguments are adjusted
> according to the <PARAM_x86.txt> file. The arguments to the script still
> can be provided in the command line run.
>
> The input for the test is redirected from the generated file
> <SUMCOL_5000.txt>. This file is the result of concatenation of the
> <SUMCOL_1.txt> 5000 times.
> ---
>   perf/LuaJIT-benches/sum-file.lua | 33 +++++++++++++++++++++++++++-----
>   1 file changed, 28 insertions(+), 5 deletions(-)
>
> diff --git a/perf/LuaJIT-benches/sum-file.lua b/perf/LuaJIT-benches/sum-file.lua
> index c9e618fd..6af2b4a5 100644
> --- a/perf/LuaJIT-benches/sum-file.lua
> +++ b/perf/LuaJIT-benches/sum-file.lua
> @@ -1,6 +1,29 @@
> +-- The benchmark to check the performance of reading lines from
> +-- stdin and sum the given numbers (the strings converted to
> +-- numbers by the VM automatically).
>   
> -local sum = 0
> -for line in io.lines() do
> -  sum = sum + line
> -end
> -io.write(sum, "\n")
> +local bench = require("bench").new(arg)
> +
> +-- XXX: The input file is generated from <SUMCOL_1.txt> by
> +-- repeating it 5000 times. The <SUMCOL_1.txt> contains 1000 lines
> +-- with the total sum of 500.
> +bench:add({
> +  name = "sum_file",
> +  payload = function()
> +    local sum = 0
> +    for line in io.lines() do
> +      sum = sum + line
> +    end
> +    -- Allow several iterations.
> +io.stdin:seek("set", 0)
> +    return sum
> +  end,
> +  checker = function(res)
> +    -- Precomputed result.
> +    return res == 2500000
> +  end,
> +  -- Fixed size of the file.
> +  items = 5e6,
> +})
> +
> +bench:run_and_report()