Always use generated aligned_sizeof

[christiangnrd]

See #2790 (comment)

Close #2790

[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/CUDA.jl b/src/CUDA.jl
index 8a82201a0..97402ecd5 100644
--- a/src/CUDA.jl
+++ b/src/CUDA.jl
@@ -54,7 +54,7 @@ using Printf
 # - Base.aligned_sizeof is the size of an object in an array/inline alloced
 # Both of them are equivalent for immutable objects, but differ for mutable singtons and Symbol
 # We use `aligned_sizeof` since we care about the size of a type in an array
-@generated function aligned_sizeof(::Type{T}) where T
+@generated function aligned_sizeof(::Type{T}) where {T}
     return :($(Base.aligned_sizeof(T)))
 end
 

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 89.77%. Comparing base (205c238) to head (659d67f).
⚠️ Report is 6 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #2838      +/-   ##
==========================================
- Coverage   89.78%   89.77%   -0.02%     
==========================================
  Files         150      150              
  Lines       13229    13229              
==========================================
- Hits        11878    11876       -2     
- Misses       1351     1353       +2     

☔ 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

Benchmark suite	Current: 659d67f	Previous: c05359d	Ratio
latency/precompile	42971664043.5 ns	42922336650.5 ns	1.00
latency/ttfp	6993096215 ns	7015168424 ns	1.00
latency/import	3570402743 ns	3571269514 ns	1.00
integration/volumerhs	9607559 ns	9608723 ns	1.00
integration/byval/slices=1	146787 ns	146920.5 ns	1.00
integration/byval/slices=3	425733 ns	425845 ns	1.00
integration/byval/reference	144985 ns	145020 ns	1.00
integration/byval/slices=2	286327 ns	286380 ns	1.00
integration/cudadevrt	103478 ns	103554 ns	1.00
kernel/indexing	14271 ns	14235 ns	1.00
kernel/indexing_checked	14751 ns	14711 ns	1.00
kernel/occupancy	673.8924050632911 ns	672.5506329113924 ns	1.00
kernel/launch	2185.4444444444443 ns	2270.3333333333335 ns	0.96
kernel/rand	14825 ns	14669 ns	1.01
array/reverse/1d	19626 ns	19682 ns	1.00
array/reverse/2d	23222.5 ns	23613.5 ns	0.98
array/reverse/1d_inplace	10009 ns	10461 ns	0.96
array/reverse/2d_inplace	11700 ns	13212 ns	0.89
array/copy	20785 ns	20972 ns	0.99
array/iteration/findall/int	157667 ns	157808 ns	1.00
array/iteration/findall/bool	139824 ns	139837 ns	1.00
array/iteration/findfirst/int	157411 ns	164937 ns	0.95
array/iteration/findfirst/bool	157968 ns	165868 ns	0.95
array/iteration/scalar	72335 ns	73041 ns	0.99
array/iteration/logical	213993.5 ns	214850 ns	1.00
array/iteration/findmin/1d	46243 ns	46704 ns	0.99
array/iteration/findmin/2d	96178.5 ns	96962.5 ns	0.99
array/reductions/reduce/Int64/1d	43151.5 ns	46033 ns	0.94
array/reductions/reduce/Int64/dims=1	47530.5 ns	55193 ns	0.86
array/reductions/reduce/Int64/dims=2	61834.5 ns	62917 ns	0.98
array/reductions/reduce/Int64/dims=1L	88681 ns	88869 ns	1.00
array/reductions/reduce/Int64/dims=2L	87096 ns	87079 ns	1.00
array/reductions/reduce/Float32/1d	34057 ns	34606 ns	0.98
array/reductions/reduce/Float32/dims=1	41224 ns	43875 ns	0.94
array/reductions/reduce/Float32/dims=2	59333 ns	59705 ns	0.99
array/reductions/reduce/Float32/dims=1L	52069 ns	52260 ns	1.00
array/reductions/reduce/Float32/dims=2L	69431 ns	70051.5 ns	0.99
array/reductions/mapreduce/Int64/1d	43068 ns	42671.5 ns	1.01
array/reductions/mapreduce/Int64/dims=1	50494 ns	45980 ns	1.10
array/reductions/mapreduce/Int64/dims=2	61817.5 ns	62143.5 ns	0.99
array/reductions/mapreduce/Int64/dims=1L	88523 ns	88812 ns	1.00
array/reductions/mapreduce/Int64/dims=2L	86431.5 ns	86818 ns	1.00
array/reductions/mapreduce/Float32/1d	33770 ns	34742 ns	0.97
array/reductions/mapreduce/Float32/dims=1	41333.5 ns	43090.5 ns	0.96
array/reductions/mapreduce/Float32/dims=2	59688 ns	60061 ns	0.99
array/reductions/mapreduce/Float32/dims=1L	52355 ns	52528 ns	1.00
array/reductions/mapreduce/Float32/dims=2L	69878 ns	70191 ns	1.00
array/broadcast	20001.5 ns	20155 ns	0.99
array/copyto!/gpu_to_gpu	11257 ns	11294 ns	1.00
array/copyto!/cpu_to_gpu	215839 ns	216503 ns	1.00
array/copyto!/gpu_to_cpu	283928 ns	284237 ns	1.00
array/accumulate/Int64/1d	124425 ns	125529 ns	0.99
array/accumulate/Int64/dims=1	83088 ns	84037 ns	0.99
array/accumulate/Int64/dims=2	157834 ns	159166 ns	0.99
array/accumulate/Int64/dims=1L	1708959.5 ns	1720376 ns	0.99
array/accumulate/Int64/dims=2L	966206.5 ns	968348 ns	1.00
array/accumulate/Float32/1d	108670 ns	109984 ns	0.99
array/accumulate/Float32/dims=1	80472 ns	81082 ns	0.99
array/accumulate/Float32/dims=2	147446 ns	148760 ns	0.99
array/accumulate/Float32/dims=1L	1618565 ns	1629307.5 ns	0.99
array/accumulate/Float32/dims=2L	697859 ns	701479 ns	0.99
array/construct	1292.8 ns	1287.2 ns	1.00
array/random/randn/Float32	47586 ns	44176 ns	1.08
array/random/randn!/Float32	24637 ns	24930 ns	0.99
array/random/rand!/Int64	27389 ns	27547 ns	0.99
array/random/rand!/Float32	8754.333333333334 ns	8724.666666666666 ns	1.00
array/random/rand/Int64	29575 ns	30114 ns	0.98
array/random/rand/Float32	12759 ns	13059 ns	0.98
array/permutedims/4d	59734 ns	60761 ns	0.98
array/permutedims/2d	53974.5 ns	54037 ns	1.00
array/permutedims/3d	54693.5 ns	54954 ns	1.00
array/sorting/1d	2775235.5 ns	2756544 ns	1.01
array/sorting/by	3368263.5 ns	3343249 ns	1.01
array/sorting/2d	1087793 ns	1080799 ns	1.01
cuda/synchronization/stream/auto	1035.9 ns	1040.3 ns	1.00
cuda/synchronization/stream/nonblocking	7060.9 ns	7220 ns	0.98
cuda/synchronization/stream/blocking	831.2708333333334 ns	802.3333333333334 ns	1.04
cuda/synchronization/context/auto	1161.2 ns	1203.5 ns	0.96
cuda/synchronization/context/nonblocking	7108.299999999999 ns	7276.700000000001 ns	0.98
cuda/synchronization/context/blocking	881.2641509433962 ns	900.4347826086956 ns	0.98

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

[maleadt]

cc @vchuravy

[vchuravy]

@christiangnrd perhaps we should open an issue with Julia, but it also just might mean that this relies on effect analysis to be elided and therefore bleh.

[vchuravy]

This is also likely the same as JuliaGPU/GPUCompiler.jl#712