Make setscalar GPU-friendly

[blegat]

On CPU, it makes no difference:

using BenchmarkTools

function _setindex1!(x, v, i)
    x[i] = v
    return
end

function _setindex2!(x, v, i)
    I = reshape(i:i, ())
    view(x, I) .= v
    return
end

function _setindex3!(x, v, i)
    I = i:i
    view(x, I) .= v
    return
end

function setindex_bench()
    x = rand(10)
    i = 3
    v = 2.0
    @btime _setindex1!($x, $v, $i)
    @btime _setindex2!($x, $v, $i)
    @btime _setindex3!($x, $v, $i)
end

setindex_bench()
setindex_bench()
setindex_bench()
setindex_bench()

gives

  2.075 ns (0 allocations: 0 bytes)
  2.091 ns (0 allocations: 0 bytes)
  2.046 ns (0 allocations: 0 bytes)

  1.937 ns (0 allocations: 0 bytes)
  2.089 ns (0 allocations: 0 bytes)
  2.089 ns (0 allocations: 0 bytes)

  2.050 ns (0 allocations: 0 bytes)
  2.048 ns (0 allocations: 0 bytes)
  2.085 ns (0 allocations: 0 bytes)

  2.088 ns (0 allocations: 0 bytes)
  2.047 ns (0 allocations: 0 bytes)
  2.046 ns (0 allocations: 0 bytes)

But since CUDA disallow scalar indexing, this is a way to make CUDA not complain without needed to write CUDA.@allowscalar.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 89.88%. Comparing base (94ef9e2) to head (eca4231).

Additional details and impacted files

@@           Coverage Diff           @@
##             main      #52   +/-   ##
=======================================
  Coverage   89.87%   89.88%           
=======================================
  Files          23       23           
  Lines        2845     2847    +2     
=======================================
+ Hits         2557     2559    +2     
  Misses        288      288           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.