Threading fixes

[maleadt]

Fixes #706

[codecov]

Codecov Report

Merging #734 (ec6ce32) into master (c6bfe1d) will increase coverage by 13.38%.
The diff coverage is 75.00%.

@@             Coverage Diff             @@
##           master     #734       +/-   ##
===========================================
+ Coverage   65.35%   78.74%   +13.38%     
===========================================
  Files         123      124        +1     
  Lines        7427     7466       +39     
===========================================
+ Hits         4854     5879     +1025     
+ Misses       2573     1587      -986     

Impacted Files	Coverage Δ
lib/cudadrv/context.jl	89.04% <ø> (ø)
src/pool/utils.jl	57.69% <0.00%> (+1.69%)	⬆️
src/state.jl	87.00% <ø> (+6.59%)	⬆️
lib/cudadrv/stream.jl	81.25% <75.00%> (+1.25%)	⬆️
lib/curand/CURAND.jl	98.41% <100.00%> (+1.69%)	⬆️
examples/wmma/high-level.jl	11.11% <0.00%> (-88.89%)	⬇️
examples/wmma/low-level.jl	14.28% <0.00%> (-85.72%)	⬇️
src/compiler/exceptions.jl	94.44% <0.00%> (-5.56%)	⬇️
lib/cufft/fft.jl	85.03% <0.00%> (-3.85%)	⬇️
lib/cufft/error.jl	27.27% <0.00%> (-2.73%)	⬇️
... and 39 more

---

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 c6bfe1d...7f8aae9. Read the comment docs.

[norci]

This fix works. Great!

btw, there is a undefined variable pool_name .
I guess there's no unit test for memory_status :)

diff --git a/src/pool.jl b/src/pool.jl
index fbb239d9..b3aabb8f 100644
--- a/src/pool.jl
+++ b/src/pool.jl
@@ -584,6 +584,10 @@ function memory_status(io::IO=stdout)
   alloc_used_bytes = used_memory()
   alloc_cached_bytes = cached_memory()
   alloc_total_bytes = alloc_used_bytes + alloc_cached_bytes
+
+  default_pool = version() >= v"11.2" ? "cuda" : "binned"
+  pool_name = get(ENV, "JULIA_CUDA_MEMORY_POOL", default_pool)
+
   @printf(io, "%s usage: %s (%s allocated, %s cached)\n", pool_name,
               Base.format_bytes(alloc_total_bytes), Base.format_bytes(alloc_used_bytes),
               Base.format_bytes(alloc_cached_bytes))

[maleadt]

btw, there is a undefined variable pool_name .
I guess there's no unit test for memory_status :)

Thanks, fixed!