Only warn about missed transformations when debug output is enabled.

[maleadt]

julia> function main()
           function kernel(a)
               @loopinfo unroll for i = eachindex(a)
                   a[i] = 0
               end
               return
           end

           arr = CUDA.zeros(10)
           @cuda kernel(arr)
       end
main (generic function with 1 method)

julia> empty!(CUDA.compiler_cache(context())); main()
CUDA.HostKernel for kernel(CuDeviceVector{Float32, 1})

julia> ENV["JULIA_DEBUG"] = "GPUCompiler"
"GPUCompiler"

julia> empty!(CUDA.compiler_cache(context())); main()
warning: abstractarray.jl:699:0: loop not unrolled: the optimizer was unable to perform the requested transformation; the transformation might be disabled or specified as part of an unsupported transformation ordering
CUDA.HostKernel for kernel(CuDeviceVector{Float32, 1})

cc @pchintalapudi

[codecov]

Codecov Report

Patch coverage: 66.66% and project coverage change: +0.82% 🎉

Comparison is base (edfdc1a) 83.18% compared to head (ce900c4) 84.00%.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #521      +/-   ##
==========================================
+ Coverage   83.18%   84.00%   +0.82%     
==========================================
  Files          24       24              
  Lines        3300     3302       +2     
==========================================
+ Hits         2745     2774      +29     
+ Misses        555      528      -27     

Files Changed	Coverage Δ
src/optim.jl	88.51% <50.00%> (-0.24%)	⬇️
src/utils.jl	94.00% <100.00%> (+0.12%)	⬆️

... and 1 file with indirect coverage changes

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