Julia: Performance regression with multiple calls to `foreach`

[rdeits]

Here's a simple example:

function g(f, v)
    foreach(f, v)
    foreach(f, v)
end

using BenchmarkTools
@btime g(identity, foo) setup=(foo = rand(Int, 100))

On v0.6.1: 127.299 ns (0 allocations: 0 bytes)

On nightly (0.7.0-DEV.3234): 6.770 μs (400 allocations: 6.25 KiB)

Removing the second call to foreach fixes the performance issue. Replacing the foreach call with for x in v; f(x); end also fixes the issue.

versioninfo for 0.6.1 is:

Julia Version 0.6.1
Commit 0d7248e (2017-10-24 22:15 UTC)
Platform Info:
  OS: macOS (x86_64-apple-darwin14.5.0)
  CPU: Intel(R) Core(TM) i7-2860QM CPU @ 2.50GHz
  WORD_SIZE: 64
  BLAS: libopenblas (USE64BITINT DYNAMIC_ARCH NO_AFFINITY Sandybridge)
  LAPACK: libopenblas64_
  LIBM: libopenlibm
  LLVM: libLLVM-3.9.1 (ORCJIT, sandybridge)

versioninfo for 0.7.0-DEV.3234 is:

Julia Version 0.7.0-DEV.3234
Commit 2cc82d2 (2018-01-02 11:44 UTC)
Platform Info:
  OS: macOS (x86_64-apple-darwin15.6.0)
  CPU: Intel(R) Core(TM) i7-2860QM CPU @ 2.50GHz
  WORD_SIZE: 64
  BLAS: libopenblas (USE64BITINT DYNAMIC_ARCH NO_AFFINITY Sandybridge)
  LAPACK: libopenblas64_
  LIBM: libopenlibm
  LLVM: libLLVM-3.9.1 (ORCJIT, sandybridge)
Environment:

[rdeits]

I just checked today's nightly, and this issue is resolved. In fact, the code is about 3X faster on 0.7.0-beta2.43 than 0.6.3!