Take into account unified memory allocations for GC pressure

[Tortar]

Fixes #3013

Please take into consideration that I'm not actually sure about the correctness of this since I have no experience working with the internals of this lib even if it does seem to solve the problem, an LLM was used to produce it

[github-actions]

Your PR requires formatting changes to meet the project's style guidelines.
Please consider running Runic (git runic master) to apply these changes.

Click here to view the suggested changes.

diff --git a/src/memory.jl b/src/memory.jl
index 78ed633c1..8181aca79 100644
--- a/src/memory.jl
+++ b/src/memory.jl
@@ -670,9 +670,9 @@ end
     mem
 end
 @inline function _pool_alloc(::Type{UnifiedMemory}, sz)
-  mem = alloc(UnifiedMemory, sz)
-  account!(memory_stats(), sz)
-  mem
+    mem = alloc(UnifiedMemory, sz)
+    account!(memory_stats(), sz)
+    return mem
 end
 @inline function _pool_alloc(::Type{HostMemory}, sz)
   alloc(HostMemory, sz)
@@ -727,8 +727,8 @@ end
     account!(memory_stats(mem.dev), -sizeof(mem))
 end
 @inline function _pool_free(mem::UnifiedMemory, stream::CuStream)
-  account!(memory_stats(), -sizeof(mem))
-  free(mem)
+    account!(memory_stats(), -sizeof(mem))
+    return free(mem)
 end
 @inline _pool_free(mem::HostMemory, stream::CuStream) = free(mem)
 

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 89.46%. Comparing base (0c00b83) to head (5050835).
⚠️ Report is 1 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #3014      +/-   ##
==========================================
+ Coverage   89.43%   89.46%   +0.02%     
==========================================
  Files         148      148              
  Lines       12991    12995       +4     
==========================================
+ Hits        11619    11626       +7     
+ Misses       1372     1369       -3     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[github-actions]

CUDA.jl Benchmarks

Details

Benchmark suite	Current: 5050835	Previous: 0c00b83	Ratio
latency/precompile	55440528447 ns	55103441406.5 ns	1.01
latency/ttfp	7811786159.5 ns	7854970810.5 ns	0.99
latency/import	4135630131 ns	4142920660.5 ns	1.00
integration/volumerhs	9611393 ns	9624895.5 ns	1.00
integration/byval/slices=1	146970 ns	147201 ns	1.00
integration/byval/slices=3	425954 ns	426000 ns	1.00
integration/byval/reference	145011 ns	145105 ns	1.00
integration/byval/slices=2	286518.5 ns	286632 ns	1.00
integration/cudadevrt	103528 ns	103846 ns	1.00
kernel/indexing	14146 ns	14265 ns	0.99
kernel/indexing_checked	14722 ns	14925 ns	0.99
kernel/occupancy	683.1148648648649 ns	783.1441441441441 ns	0.87
kernel/launch	2226.5555555555557 ns	2262.3333333333335 ns	0.98
kernel/rand	14753 ns	16624 ns	0.89
array/reverse/1d	19859 ns	20261 ns	0.98
array/reverse/2dL_inplace	66580 ns	66981 ns	0.99
array/reverse/1dL	70083 ns	70447 ns	0.99
array/reverse/2d	21937.5 ns	22244 ns	0.99
array/reverse/1d_inplace	9633 ns	11580 ns	0.83
array/reverse/2d_inplace	13446 ns	13344 ns	1.01
array/reverse/2dL	73826 ns	74284 ns	0.99
array/reverse/1dL_inplace	66890 ns	67021 ns	1.00
array/copy	20366 ns	20733 ns	0.98
array/iteration/findall/int	157825 ns	159065 ns	0.99
array/iteration/findall/bool	139945.5 ns	141350 ns	0.99
array/iteration/findfirst/int	160523.5 ns	162741 ns	0.99
array/iteration/findfirst/bool	161829 ns	164024 ns	0.99
array/iteration/scalar	71865.5 ns	72819 ns	0.99
array/iteration/logical	214670.5 ns	220064.5 ns	0.98
array/iteration/findmin/1d	90973 ns	56834 ns	1.60
array/iteration/findmin/2d	120862 ns	98602 ns	1.23
array/reductions/reduce/Int64/1d	43196 ns	44369 ns	0.97
array/reductions/reduce/Int64/dims=1	45261.5 ns	46082 ns	0.98
array/reductions/reduce/Int64/dims=2	61434.5 ns	62261 ns	0.99
array/reductions/reduce/Int64/dims=1L	88991 ns	89560 ns	0.99
array/reductions/reduce/Int64/dims=2L	87848.5 ns	88709 ns	0.99
array/reductions/reduce/Float32/1d	37083 ns	38170.5 ns	0.97
array/reductions/reduce/Float32/dims=1	51660.5 ns	43662 ns	1.18
array/reductions/reduce/Float32/dims=2	59862 ns	60196 ns	0.99
array/reductions/reduce/Float32/dims=1L	52305 ns	52890 ns	0.99
array/reductions/reduce/Float32/dims=2L	71875 ns	72931 ns	0.99
array/reductions/mapreduce/Int64/1d	43420 ns	44486 ns	0.98
array/reductions/mapreduce/Int64/dims=1	45021 ns	51105 ns	0.88
array/reductions/mapreduce/Int64/dims=2	61289 ns	61916 ns	0.99
array/reductions/mapreduce/Int64/dims=1L	89081 ns	89497 ns	1.00
array/reductions/mapreduce/Int64/dims=2L	87923.5 ns	88980 ns	0.99
array/reductions/mapreduce/Float32/1d	36873.5 ns	37944 ns	0.97
array/reductions/mapreduce/Float32/dims=1	44834 ns	52429 ns	0.86
array/reductions/mapreduce/Float32/dims=2	60114 ns	60388 ns	1.00
array/reductions/mapreduce/Float32/dims=1L	52540 ns	53049 ns	0.99
array/reductions/mapreduce/Float32/dims=2L	72469 ns	72843 ns	0.99
array/broadcast	19847 ns	20274 ns	0.98
array/copyto!/gpu_to_gpu	10996 ns	11225 ns	0.98
array/copyto!/cpu_to_gpu	214160 ns	218396.5 ns	0.98
array/copyto!/gpu_to_cpu	284329.5 ns	284648 ns	1.00
array/accumulate/Int64/1d	124290 ns	125449 ns	0.99
array/accumulate/Int64/dims=1	84400 ns	84251 ns	1.00
array/accumulate/Int64/dims=2	158340 ns	158690 ns	1.00
array/accumulate/Int64/dims=1L	1710794 ns	1709941.5 ns	1.00
array/accumulate/Int64/dims=2L	966620.5 ns	967026.5 ns	1.00
array/accumulate/Float32/1d	108697 ns	109856 ns	0.99
array/accumulate/Float32/dims=1	80376 ns	81373 ns	0.99
array/accumulate/Float32/dims=2	147865.5 ns	148536 ns	1.00
array/accumulate/Float32/dims=1L	1619116 ns	1619811 ns	1.00
array/accumulate/Float32/dims=2L	698567 ns	699285.5 ns	1.00
array/construct	1276.4 ns	1296.2 ns	0.98
array/random/randn/Float32	43461 ns	48633 ns	0.89
array/random/randn!/Float32	24860 ns	25237 ns	0.99
array/random/rand!/Int64	27194 ns	27465 ns	0.99
array/random/rand!/Float32	8822.333333333334 ns	8946 ns	0.99
array/random/rand/Int64	31068.5 ns	30454 ns	1.02
array/random/rand/Float32	13132 ns	13364.5 ns	0.98
array/permutedims/4d	55347.5 ns	55600 ns	1.00
array/permutedims/2d	53864 ns	54423 ns	0.99
array/permutedims/3d	54831 ns	55435 ns	0.99
array/sorting/1d	2758123 ns	2759622.5 ns	1.00
array/sorting/by	3344447.5 ns	3345835 ns	1.00
array/sorting/2d	1080757 ns	1082443 ns	1.00
cuda/synchronization/stream/auto	1058.1 ns	1033 ns	1.02
cuda/synchronization/stream/nonblocking	8113 ns	7095.4 ns	1.14
cuda/synchronization/stream/blocking	842.40625 ns	848.7717391304348 ns	0.99
cuda/synchronization/context/auto	1178.9 ns	1163.5 ns	1.01
cuda/synchronization/context/nonblocking	7050.4 ns	7702.2 ns	0.92
cuda/synchronization/context/blocking	895.5102040816327 ns	910.1 ns	0.98

This comment was automatically generated by workflow using github-action-benchmark.