Some performance data

[dibyendumajumdar]

Some numbers obtained from running ravi-simplex over netlib data.

• Lua 5.4.3 lua-results.txt
• Ravi interpreted ravi-nojit-results.txt
• Ravi JIT (from Lua bytecodes) ravi-oldjit-results.txt
• Ravi JIT (from source) ravi-newjit-results.txt
• Ravi AOT ravi-aot-results.txt

[lufeidefei]

i have tested the performace use https://github.com/gligneul/Lua-Benchmarks。（i download the code and compile by myself）。here is the result. maybe i miss something.
test lua-5.4.3 lua-5.3.5 ravi ravi-jit
ack 1.239 1.8 1.777 2.246
fixpoint-fact 1.874 1.995 1.697 1.736
heapsort 2.906 4.712 3.242 3.017
mandelbrot 2.649 2.799 2.854 2.97
juliaset 2.15 2.802 2.498 2.456
queen 2.981 4.085 3.043 3.111
sieve 1.715 2.741 2.248 2.236
binary 2.879 7.665 5.852 6.711
n-body 4.914 5.921 5.668 5.291
fannkuch 3.906 7.004 4.701 4.314
fasta 3.513 4.831 4.083 4.156
k-nucleotide 0.003 0.002 0.003 0.003
spectral-norm 4.062 6.061 4.882 5.3

[dibyendumajumdar]

hi @lufeidefei

You have will to modify the scripts for Ravi to get best performance. Ravi requires type annotations; code that use numeric arrays and numeric processing will show the best results.

There is no benefit running standard Lua programs in Ravi as performance of JIT is better only when there are type annotations.

Regards

[lufeidefei]

Thanks for reply! Got it. I will try.