The cost can vary, see performances_over_time.md.
This result is computed by instructions_cost.lua.
The value of witness represent the testing algorithm cost.
It is already substracted from all tests.
Results are for 1000 instructions (string.format missbehave with small floats).
Read times may be innacurate because variables had to be used inside additional code.
witness: 0ms ttl - 0.0ms/1000i function call: 70ms ttl - 0.7ms/1000i function call 1 arg: 69ms ttl - 0.69ms/1000i function call 4 args: 105ms ttl - 1.05ms/1000i function call return: 63ms ttl - 0.63ms/1000i tonumber('NNNN'): 74ms ttl - 0.74ms/1000i tostring(NNNN): 100ms ttl - 1.0ms/1000i math.random(): 44ms ttl - 0.44ms/1000i read local: 18ms ttl - 0.18ms/1000i read global: 91ms ttl - 0.91ms/1000i assign local: 11ms ttl - 0.11ms/1000i assign global: 101ms ttl - 1.01ms/1000i read array entry: 31ms ttl - 0.31ms/1000i assign array entry: 28ms ttl - 0.28ms/1000i read table entry: 39ms ttl - 0.39ms/1000i assign table entry: 19ms ttl - 0.19ms/1000i assign #10 string: 22ms ttl - 0.22ms/1000i assign #20 string: 23ms ttl - 0.23ms/1000i 'ABCD' .. 'EFGH': 46ms ttl - 0.46ms/1000i for ipairs(#10 array):42ms ttl - 0.42ms/1000i for pairs(#10 array): 48ms ttl - 0.48ms/1000i for pairs(#10 table): 51ms ttl - 0.51ms/1000i