argsort — Return one-based permutation indices using RunMat sort options.
argsort(X) returns the permutation indices that order X the same way sort(X) would. This RunMat convenience operation uses the index behavior of [~, I] = sort(X, ...) for supported dimensions and directions.
Syntax
I = argsort(A)
I = argsort(A, arg1)
I = argsort(A, arg1, arg2)
I = argsort(A, ..., "ComparisonMethod", method)
I = argsort(A, ..., "MissingPlacement", placement)Inputs
A- Type
Any- Required
- Yes
- Description
- Input array.
arg1- Type
Any- Required
- Yes
- Description
- Dimension selector or direction token.
arg1- Type
Any- Required
- Yes
- Description
- Dimension selector, placeholder, or direction token.
arg2- Type
Any- Required
- Yes
- Description
- Dimension selector or direction token.
arg- Type
Any- Required
- Variadic
- Description
- Optional dimension/direction arguments.
name- Type
StringScalar- Required
- Yes
- Default
"ComparisonMethod"- Description
- Name-value option key.
method- Type
StringScalar- Required
- Yes
- Default
"auto"- Description
- Comparison method: 'auto', 'real', or 'abs'.
name- Type
StringScalar- Required
- Yes
- Default
"MissingPlacement"- Description
- Name-value option key.
placement- Type
StringScalar- Required
- Yes
- Default
"auto"- Description
- Requested NaN placement option (currently unsupported).
Returns
I- Type
NumericArray- Description
- One-based permutation indices that sort each slice.
Errors
RunMat:sort:InvalidDimension- When
- Dimension argument is non-positive, non-integer, or otherwise invalid.
- Message
sort: invalid dimension argument
RunMat:sort:ComparisonMethodRequiresString- When
- ComparisonMethod option value is not string-like.
- Message
sort: 'ComparisonMethod' requires a string value
RunMat:sort:ComparisonMethodUnknown- When
- ComparisonMethod option value is not one of 'auto'/'real'/'abs'.
- Message
sort: unsupported ComparisonMethod
Behavior
Operates along the first non-singleton dimension by default. Pass a dimension argument to override.
Accepts the same direction keywords as sort: 'ascend' (default) or 'descend'.
Supports 'ComparisonMethod' values 'auto', 'real', and 'abs' for real and complex inputs.
Returns indices as double-precision tensors using MATLAB's one-based indexing.
Treats NaN values as missing: they appear at the end for ascending permutations and at the beginning for descending permutations.
Acts as a residency sink. GPU tensors are gathered when the active provider does not expose a specialised sort kernel.
Types, placement and limitations
argsort publishes only one-based double permutation indices. It currently uses the same evaluation as sort, including computing sorted values. Resident indices are restored to the input owner; a provider hook may stage work through host memory.
Resident execution
These operations are eager fusion boundaries, not elementwise kernels or reductions, and do not broadcast. Provider availability does not imply a device-native sorting or set kernel: current providers may stage host work. An unsupported input can use exact host fallback before provider work starts. Errors after an attempt, including cancellation, are returned without repeating the operation on the host. Supported resident outputs return to their owning provider; predicates and all-indices cells have the host placement described above. Distributed execution is not supported.
Examples
Getting indices that sort a vector
A = [4; 1; 3];
idx = argsort(A)Expected output:
idx =
2
3
1Reordering data with the permutation indices
A = [3 9 1 5];
idx = argsort(A);
sorted = A(idx)Expected output:
sorted =
1 3 5 9Sorting along a specific dimension
A = [1 6 4; 2 3 5];
idx = argsort(A, 2)Expected output:
idx =
1 3 2
1 2 3Descending order permutations
A = [10 4 7 9];
idx = argsort(A, 'descend')Expected output:
idx =
1 4 3 2Using ComparisonMethod to sort by magnitude
A = [-8 -1 3 -2];
idx = argsort(A, 'ComparisonMethod', 'abs')Expected output:
idx =
2 4 3 1Handling NaN values during permutation
A = [NaN 4 1 2];
idx = argsort(A)Expected output:
idx =
3 4 2 1Argsort on GPU tensors falls back gracefully
G = gpuArray(randn(5, 1));
idx = argsort(G)Using argsort with coding agents
Open a RunMat example with live inputs, then ask the agent to explain how argsort changes the result.
Run a small argsort example, explain the result, then change one input and compare the output.
FAQ
How is argsort different from sort?⌄
argsort returns only the permutation indices. It behaves like calling [~, I] = sort(X, ...) and publishes only the indices; the current evaluation also computes sorted values.
Are the indices one-based like MATLAB?⌄
Yes. All indices follow MATLAB's one-based convention so they can be used directly with subsequent indexing operations.
Does argsort support the same arguments as sort?⌄
Yes. Dimension arguments, direction keywords, and 'ComparisonMethod' behave exactly like they do for sort.
How are NaN values ordered?⌄
NaNs are treated as missing. They appear at the end for ascending permutations and at the beginning for descending permutations, matching MATLAB.
Can I call argsort on GPU arrays?⌄
argsort publishes only one-based double permutation indices. It currently uses the same evaluation as sort, including computing sorted values. Resident indices are restored to the input owner; a provider hook may stage work through host memory. These operations are eager fusion boundaries, not elementwise kernels or reductions, and do not broadcast. Provider availability does not imply a device-native sorting or set kernel: current providers may stage host work. An unsupported input can use exact host fallback before provider work starts. Errors after an attempt, including cancellation, are returned without repeating the operation on the host. Supported resident outputs return to their owning provider; predicates and all-indices cells have the host placement described above. Distributed execution is not supported.
Is the permutation stable?⌄
Yes. Equal elements keep their relative order so that argsort remains consistent with MATLAB's stable sorting semantics.
What type is returned?⌄
A double-precision tensor (or scalar) with the same shape as the input, containing permutation indices.
Does argsort mutate its input?⌄
No. It only returns indices. Combine the result with indexing (A(idx)) to obtain reordered values when needed.
Related Array functions
Sorting Sets
intersect · ismember · ismembertol · issorted · issortedrows · setdiff · setxor · sort · sortrows · union · unique
Accumulation
Shape
blkdiag · cat · circshift · diag · flip · fliplr · flipud · horzcat · ipermute · kron · permute · repelem · repmat · reshape · rot90 · shiftdim · squeeze · toeplitz · tril · triu · vertcat
Creation
colon · createArray · empty · eye · false · full · inf · int16.empty · int32.empty · int64.empty · int8.empty · linspace · logspace · magic · meshgrid · nan · ndgrid · nonzeros · ones · peaks · rand · randi · randn · randperm · range · sparse · spdiags · speye · spones · sprand · true · uint16.empty · uint32.empty · uint64.empty · uint8.empty · zeros
Combinatorics
Binning
Grouping
Source you can read
Every RunMat function's source is public. Read exactly how argsort is executed, line by line, in Rust.
- View the source for argsort in Rust on GitHub
- Learn how the RunMat runtime works
- Found a bug? Open an issue with a minimal reproduction.
About RunMat
RunMat is a runtime that executes MATLAB-syntax code fast on any GPU. The runtime is free for individuals, including freelancers; students and faculty at an academic institution, for teaching or research there; contributors to an open-source project, for work on that project; organizations with 250 or fewer people and US$10M or less in annual revenue; and anyone, for non-production use. See runmat.com/license.
- RunMat automatically optimizes your math for GPU execution on Apple, Nvidia, and AMD hardware. No code changes needed. Simulations that took hours now take minutes.
- Start running code in seconds. RunMat runs in the browser, on the desktop, or from the CLI. No license server, no IT ticket.