Conversation
|
Test failure in 0.6 is due to #38. |
src/execution.jl
Outdated
| iters = 2 | ||
| while (time() - start_time) < params.seconds && iters ≤ params.samples | ||
| params.gcsample && BenchmarkTools.gcscrub() | ||
| params.gcsample && BenchmarkTools.gcscrub() |
There was a problem hiding this comment.
Why is this getting done twice here?
|
Yay, tests are passing in the rebased branch. |
|
Rebased. |
|
Thanks! |
|
See also JuliaLang/julia#20173 |
|
|
| @@ -1,3 +1,3 @@ | |||
| julia 0.4 | |||
| Compat 0.8.0 | |||
| Compat 0.9.5 | |||
There was a problem hiding this comment.
what is this using from 0.9.5?
There was a problem hiding this comment.
I can't remember, maybe it was the Symbol constructor? I definitely upgraded the Compat requirement for a reason.
There was a problem hiding this comment.
Symbol is older I believe - probably JuliaLang/Compat.jl@2c11855
Maybe an earlier version of this was using one of the show rewrites? doesn't look like the final version is though
There was a problem hiding this comment.
I think maybe an earlier version tried to use a buffer and String(take!(buf)) for the @btime test, which requires Compat 0.9.5. I then rewrote it to use redirect_stdout(f) with a temporary file, which requires Compat 0.9.0 (JuliaLang/Compat.jl@2c11855)
This PR adds two macros,
@btimeand@belapsed, that are drop-in replacements for@timeand@elapsedfrom Base except that they use the@benchmarkapparatus. For example:It also changes the memory output to use SI prefixes. Previously, it would output something like
17.34 kb, which is ambiguous — is a kb 1000 or 1024 bytes (we use 1024)? TheSIISQ prefix is "KiB" for kibibyte, which unambiguously means 1024 bytes.Also, it now prints
24 bytesand not24.00 bytes— the decimal point was confusing since (presumably) we will never be able to allocate a fractional number of bytes.Closes #35.