Use divide and conquer approach...

True but this doesnt't work, wouldn't pass any tests and doesn't conform to the tests

using System.Linq;class M{static T x(params T[]p)=>p.Max();}

You're right, vectorMax should be initialized to that to account for negatives.

Copying the vectors isn't an issue. You need two vectors to intrinsically compare two vectors. Doing this alone already compensates for the time spent copying some memory around. Memory is cheap. In fact, doing most things in C# involve allocating memory. A foreach loop even creates copies of values.

From benchmarks I have done a while ago, this method of comparing values in an array is magnitudes faster than any scalar method.

You iterate through every vector which would not any faster than copying a very small block of memory after using abstracted intrinsics to compare two vectors. Under the hood, the Max method allows for "jit intrinsic expansions" so you should save the entirety of the compared vectors until you're done with vectors

Thanks for explaining your choices :)

var vectorMax = Vector<int>.Zero; but var vectorMax = Vector<int>(int.MinValue); should be, I think

var blockVector = new Vector<int>(nums, blockIndex); copy memory

vectorMax = Vector.Max(vectorMax, blockVector); copy memory

2 memory coping. Is it faster then classic loop?

if (items == null || items.Length == 0) C# compiler makes check on null them self, so if (items.Length == 0) is enought.
max = items[i].CompareTo(max) > 0 ? items[i] : max; is slower then if (max.CompareTo(items[i]) < 0) max = items[i];

I love var more then T because code can be copied and paste to another place without changing.

C translation (author gone)

It even has random tests !

added

added useful error messages to JS

tests now use assert_equals

1000 Go completions so it's unlikely that the tests would crash

fixed for python