Missing optimization for findmax/findmin on SparseVector

[kulowf]

using SparseArrays

a = spzeros(2^32)
a[rand(UInt16,6)] = rand(6)

# this variant is very slow, on bigger Arrays error
@time (m,i)=findmax(a)

# this variant is quick, also on bigger Arrays
#   todo correct result if only negative values in a
@time begin
    (m1,tmp) = findmax(findnz(a)[2])
    i1 = findnz(a)[1][tmp]        
end
@show m i m1 i1

versioninfo()

Output:

21.400888 seconds (9 allocations: 224 bytes)
0.000028 seconds (14 allocations: 784 bytes)
m = 0.8275944812507452
i = 51128
m1 = 0.8275944812507452
i1 = 51128
51128

Julia Version 1.6.0
Commit f9720dc (2021-03-24 12:55 UTC)
Platform Info:
OS: Windows (x86_64-w64-mingw32)
CPU: Intel(R) Core(TM) i7-7820HQ CPU @ 2.90GHz
WORD_SIZE: 64
LIBM: libopenlibm
LLVM: libLLVM-11.0.1 (ORCJIT, skylake)
Environment:
JULIA_EDITOR = code
JULIA_NUM_THREADS =