Refined packing

[YingboMa]

julia> m = k = n = 400; C = rand(m, n); A = rand(m, k); B = rand(k, n);

julia> @btime LinearAlgebra.mul!($(copy(C)), $A, $B);
  3.373 ms (0 allocations: 0 bytes)

julia> @btime MaBLAS.mul!($(copy(C)), $A, $B; packing=(false, false)); # no packing
  4.511 ms (0 allocations: 0 bytes)

julia> @btime MaBLAS.mul!($(copy(C)), $A, $B; packing=(true, true)); # pack A and B
  3.572 ms (3 allocations: 240 bytes)

julia> @btime MaBLAS.mul!($(copy(C)), $A, $B; packing=(true, false)); # pack A
  3.370 ms (2 allocations: 160 bytes)

julia> @btime MaBLAS.mul!($(copy(C)), $A, $B; packing=(false, true)); # pack B
  4.723 ms (2 allocations: 160 bytes)

[codecov-io]

Codecov Report

Merging #4 into master will increase coverage by 0.91%. The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master       #4      +/-   ##
==========================================
+ Coverage   84.21%   85.12%   +0.91%     
==========================================
  Files           3        3              
  Lines         228      242      +14     
==========================================
+ Hits          192      206      +14     
  Misses         36       36              

Impacted Files	Coverage Δ
src/gemm.jl	86.08% <100.00%> (+0.90%)	:arrow_up:

---

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update ca1305e...9d513d3. Read the comment docs.