Add _throw_dmrs device override for reshape of views

[Abdelrahman912]

Problem

reshape(@view(data[1:n*n]), (n, n)) fails to compile on the GPU. @view creates a SubArray, which has no specialized reshape method on the device, so it falls back to Base's generic _reshape. That path calls _throw_dmrs, which tries to construct a DimensionMismatch string — unsupported on the GPU.

Reason: unsupported call to an external C function
Stacktrace:
 [1] _string_n
   @ ./strings/string.jl:109
 [2] StringMemory
   @ ./iobuffer.jl:167
 [3] dec
   @ ./intfuncs.jl:918
 [4] #string#403
   @ ./intfuncs.jl:1000
 [5] multiple call sites
   @ unknown:0
   

function _reshape(parent::AbstractArray, dims::Dims)
  n = length(parent)
  prod(dims) == n || _throw_dmrs(n, "size", dims)
  __reshape((parent, IndexStyle(parent)), dims)
end

@noinline function _throw_dmrs(n, str, dims)
  throw(DimensionMismatch("parent has $n elements, which is incompatible with $str $dims")) ## THIS IS THE CULPRIT
end

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 90.32%. Comparing base (5065018) to head (14712fa).

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #3095   +/-   ##
=======================================
  Coverage   90.31%   90.32%           
=======================================
  Files         141      141           
  Lines       12165    12165           
=======================================
+ Hits        10987    10988    +1     
+ Misses       1178     1177    -1     

☔ 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: 14712fa	Previous: 5065018	Ratio
array/accumulate/Float32/1d	101002 ns	101779 ns	0.99
array/accumulate/Float32/dims=1	76209 ns	76943 ns	0.99
array/accumulate/Float32/dims=1L	1584594 ns	1585563 ns	1.00
array/accumulate/Float32/dims=2	143217 ns	144109 ns	0.99
array/accumulate/Float32/dims=2L	657221.5 ns	658119 ns	1.00
array/accumulate/Int64/1d	118204 ns	118854 ns	0.99
array/accumulate/Int64/dims=1	79400.5 ns	80389 ns	0.99
array/accumulate/Int64/dims=1L	1693357.5 ns	1694708 ns	1.00
array/accumulate/Int64/dims=2	155741 ns	156166 ns	1.00
array/accumulate/Int64/dims=2L	961351 ns	961594 ns	1.00
array/broadcast	20395.5 ns	20512 ns	0.99
array/construct	1335 ns	1320.1 ns	1.01
array/copy	18666.5 ns	18914 ns	0.99
array/copyto!/cpu_to_gpu	212677 ns	214662 ns	0.99
array/copyto!/gpu_to_cpu	282913 ns	285022 ns	0.99
array/copyto!/gpu_to_gpu	11190 ns	11357 ns	0.99
array/iteration/findall/bool	132091 ns	132106 ns	1.00
array/iteration/findall/int	149004 ns	148958 ns	1.00
array/iteration/findfirst/bool	81362.5 ns	82321 ns	0.99
array/iteration/findfirst/int	83892 ns	84247.5 ns	1.00
array/iteration/findmin/1d	84612 ns	85453 ns	0.99
array/iteration/findmin/2d	116960 ns	117211 ns	1.00
array/iteration/logical	199494.5 ns	200370 ns	1.00
array/iteration/scalar	69041 ns	69501 ns	0.99
array/permutedims/2d	52311.5 ns	52561 ns	1.00
array/permutedims/3d	52770 ns	53080 ns	0.99
array/permutedims/4d	51628 ns	51880.5 ns	1.00
array/random/rand/Float32	13019 ns	12980 ns	1.00
array/random/rand/Int64	37146 ns	29766 ns	1.25
array/random/rand!/Float32	8592.666666666666 ns	8371 ns	1.03
array/random/rand!/Int64	34432 ns	33996 ns	1.01
array/random/randn/Float32	43891 ns	38221 ns	1.15
array/random/randn!/Float32	31481 ns	31486 ns	1.00
array/reductions/mapreduce/Float32/1d	34717 ns	35499.5 ns	0.98
array/reductions/mapreduce/Float32/dims=1	39448.5 ns	39750 ns	0.99
array/reductions/mapreduce/Float32/dims=1L	51882 ns	51923 ns	1.00
array/reductions/mapreduce/Float32/dims=2	56756 ns	56902 ns	1.00
array/reductions/mapreduce/Float32/dims=2L	69674 ns	70119.5 ns	0.99
array/reductions/mapreduce/Int64/1d	42941.5 ns	43301 ns	0.99
array/reductions/mapreduce/Int64/dims=1	43018 ns	53296 ns	0.81
array/reductions/mapreduce/Int64/dims=1L	87516 ns	87758 ns	1.00
array/reductions/mapreduce/Int64/dims=2	59713.5 ns	60300 ns	0.99
array/reductions/mapreduce/Int64/dims=2L	84918 ns	85288 ns	1.00
array/reductions/reduce/Float32/1d	34837.5 ns	35511 ns	0.98
array/reductions/reduce/Float32/dims=1	39636 ns	49667 ns	0.80
array/reductions/reduce/Float32/dims=1L	51785 ns	52100 ns	0.99
array/reductions/reduce/Float32/dims=2	56779 ns	56986 ns	1.00
array/reductions/reduce/Float32/dims=2L	69644 ns	70384 ns	0.99
array/reductions/reduce/Int64/1d	43093 ns	43476 ns	0.99
array/reductions/reduce/Int64/dims=1	50624.5 ns	42765.5 ns	1.18
array/reductions/reduce/Int64/dims=1L	87761 ns	87783 ns	1.00
array/reductions/reduce/Int64/dims=2	59541 ns	59896 ns	0.99
array/reductions/reduce/Int64/dims=2L	84343 ns	85038 ns	0.99
array/reverse/1d	18184 ns	18408 ns	0.99
array/reverse/1dL	68791 ns	69079 ns	1.00
array/reverse/1dL_inplace	65905 ns	65864 ns	1.00
array/reverse/1d_inplace	10261.666666666666 ns	10281.333333333334 ns	1.00
array/reverse/2d	20690 ns	20739 ns	1.00
array/reverse/2dL	72630 ns	72716 ns	1.00
array/reverse/2dL_inplace	65964 ns	65932 ns	1.00
array/reverse/2d_inplace	11041 ns	10105.5 ns	1.09
array/sorting/1d	2734368 ns	2734769 ns	1.00
array/sorting/2d	1067985 ns	1076074 ns	0.99
array/sorting/by	3302830 ns	3327052 ns	0.99
cuda/synchronization/context/auto	1166.7 ns	1184.7 ns	0.98
cuda/synchronization/context/blocking	961.2222222222222 ns	940.969696969697 ns	1.02
cuda/synchronization/context/nonblocking	8403 ns	8469.4 ns	0.99
cuda/synchronization/stream/auto	998.2 ns	1021.1 ns	0.98
cuda/synchronization/stream/blocking	842.3979591836735 ns	832.7631578947369 ns	1.01
cuda/synchronization/stream/nonblocking	7281.5 ns	7582.5 ns	0.96
integration/byval/reference	143927 ns	144056 ns	1.00
integration/byval/slices=1	145747 ns	145820 ns	1.00
integration/byval/slices=2	284359 ns	284683 ns	1.00
integration/byval/slices=3	423018 ns	423445 ns	1.00
integration/cudadevrt	102569 ns	102525.5 ns	1.00
integration/volumerhs	23436276.5 ns	23421325.5 ns	1.00
kernel/indexing	13354 ns	13267 ns	1.01
kernel/indexing_checked	14129 ns	14039 ns	1.01
kernel/launch	2199.5555555555557 ns	2141.4444444444443 ns	1.03
kernel/occupancy	668.5875 ns	695.4267515923567 ns	0.96
kernel/rand	18046 ns	15508 ns	1.16
latency/import	3832774972.5 ns	3810464780.5 ns	1.01
latency/precompile	4616111922 ns	4591174243 ns	1.01
latency/ttfp	4415613716 ns	4390561943.5 ns	1.01

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

[maleadt]

Please don't continuously merge master. That's a waste of CI resources.