!"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQR S T U V W X Y Z[\]^_`abc d e f g h i j k l m n o p q r s t u v w x y z { |}~ & Safe-InferredA region in the 2d plane. "Size of a region in the 2d plane. An offset in the 2d plane. 'Create a new region of the given size. Offset a region. HPartition a region into inner and border regions for the given stencil. Size of array Size of stencil Focus of stencil NoneEElement types that can be used with the blockwise filling functions. (This class is mainly used to define the ! method. This is used internally L in the imeplementation of Repa to prevent let-binding from being floated 3 inappropriately by the GHC simplifier. Doing a  sometimes isn' t enough, S because the GHC simplifier can erase these, and still move around the bindings. FPlace a demand on a value at a particular point in an IO computation. +Generic zero value, helpful for debugging. *Generic one value, helpful for debugging.  Safe-Inferred=Class of types that can be used as array shapes and indices. )Get the number of dimensions in a shape. FThe shape of an array of size zero, with a particular dimensionality. GThe shape of an array with size one, with a particular dimensionality. (Compute the intersection of two shapes. 0Add the coordinates of two shapes componentwise >Get the total number of elements in an array with this shape. :Check whether this shape is small enough so that its flat  indices an be represented as . If this returns  then your = array is too big. Mostly used for writing QuickCheck tests. FConvert an index into its equivalent flat, linear, row-major version.  Inverse of  . 0Check whether an index is within a given shape. -Convert a shape into its list of dimensions. (Convert a list of dimensions to a shape -Ensure that a shape is completely evaluated. 3Check whether an index is a part of a given shape. "Nicely format a shape as a string  Shape of the array. Index into the array. Shape of the array. "Index into linear representation. Start index for range. Final index for range. Index to check for. Shape of the array. Index.      Safe-Inferred ?Class of array representations that we can read elements from. *O(1). Take the extent (size) of an array. "O(1). Shape polymorphic indexing. "O(1). Shape polymorphic indexing. HO(1). Linear indexing into underlying, row-major, array representation. HO(1). Linear indexing into underlying, row-major, array representation. Ensure an array'%s data structure is fully evaluated. O(1). Alias for  "O(n). Convert an array to a list. Apply  to up to four arrays.  None MClass of manifest array representations that can be constructed in parallel. !'Mutable version of the representation. "0Allocate a new mutable array of the given size. #)Write an element into the mutable array. $7Freeze the mutable array into an immutable Repa array. %;Ensure the strucure of a mutable array is fully evaluated. &.Ensure the array is still live at this point. C Needed when the mutable array is a ForeignPtr with a finalizer. '.O(n). Construct a manifest array from a list.  !"#$%&' !"#$%&' !"#$%&'None(MCompute a range of elements defined by an array and write them to a fillable  representation. )'Fill a range of an array sequentially. *&Fill a range of an array in parallel. +FCompute all elements defined by an array and write them to a manifest  target representation. DNote that instances require that the source array to have a delayed  representation such as D or C$. If you want to use a pre-existing % manifest array as the source then delay it first. ,#Fill an entire array sequentially. -"Fill an entire array in parallel. ()*+,-()*+,-()*+,-None.AHints that evaluating this array is only a small amount of work. D It will be evaluated sequentially in the main thread, instead of L in parallel on the gang. This avoids the associated scheduling overhead. /%Wrap an array with a smallness hint. ././././None62Our index type, used for both shapes and indices. 8An index of dimension zero :Helper for index construction. /Use this instead of explicit constructors like (Z :. (x :: Int)). / The this is sometimes needed to ensure that x is constrained to  be in Int. 0123456789:;<=>0123456789:;<=>8967543210:;<=>0123456789:;<=>None?-Class of index types that can map to slices. @:Map an index of a full shape onto an index of some slice. A9Map an index of a slice onto an index of the full shape. BSMap the type of an index in the slice, to the type of the index in the full shape. CRMap a type of the index in the full shape, to the type of the index in the slice. D%Place holder for any possible shape. F*Select all indices at a certain position. ?@ABCDEFG ?@ABCDEFG FGDECB?@A ?@ABCDEFGNoneH=Represents a convolution stencil that we can apply to array. ; Only statically known stencils are supported right now. I:Static stencils are used when the coefficients are fixed,  and known at compile time. MGHow to handle the case when the stencil lies partly outside the array. N:Clamp points outside to the same value as the edge pixel. O;Treat points outside the array as having a constant value. P&Use a fixed value for border regions. QDMake a stencil from a function yielding coefficients at each index. R Wrapper for Q that requires a DIM2 stencil. HIJKLMNOPQExtent of stencil. #Get the coefficient at this index. Rextent of stencil #Get the coefficient at this index. HIJKLMNOPQRHIJKLMPONQR NoneS6QuasiQuoter for producing a static stencil defintion. A definition like    [stencil2| 0 1 0  1 0 1  0 1 0 |] Is converted to:   makeStencil2 (Z:.3:.3)  (\ix -> case ix of + Z :. -1 :. 0 -> Just 1 + Z :. 0 :. -1 -> Just 1 + Z :. 0 :. 1 -> Just 1 + Z :. 1 :. 0 -> Just 1 - _ -> Nothing) Parse a stencil definition.  TODO: make this more robust. SSS NoneTA T= is a group of threads that execute arbitrary work requests. Number of threads in the gang. +Workers listen for requests on these vars. )Workers put their results in these vars. !Indicates that the gang is busy. The C type encapsulates work requests for individual members of a gang. Tell the worker that we're shutting the gang down. # The worker should signal that it's receieved the request by H writing to its result var before returning to the caller (forkGang). -Instruct the worker to run the given action. UKThis globally shared gang is auto-initialised at startup and shared by all  Repa computations. DIn a data parallel setting, it does not help to have multiple gangs D running at the same time. This is because a single data parallel J computation should already be able to keep all threads busy. If we had M multiple gangs running at the same time, then the system as a whole would M run slower as the gangs would contend for cache and thrash the scheduler. EIf, due to laziness or otherwise, you try to start multiple parallel G Repa computations at the same time, then you will get the following ! warning on stderr at runtime: FData.Array.Repa: Performing nested parallel computation sequentially.  You've probably called the compute or copy function while another H instance was already running. This can happen if the second version . was suspended due to lazy evaluation. Use  deepSeqArray to ensure that - each array is fully evaluated before you compute the next one. V)O(1). Yield the number of threads in the T. WFork a T0 with the given number of threads (at least 1). The worker thread of a T. > The threads blocks on the MVar waiting for a work request. Finaliser for worker threads. 7 We want to shutdown the corresponding thread when it's MVar becomes  unreachable. 1 Without this Repa programs can complain about "Blocked indefinitely on an MVar9 because worker threads are still blocked on the request G MVars when the program ends. Whether the finalizer is called or not = is very racey. It happens about 1 in 10 runs when for the > repa-edgedetect benchmark, and less often with the others. We'7re relying on the comment in System.Mem.Weak that says  If there are no other threads to run, the runtime system will check for runnablefinalizers before declaring the system to be deadlocked. =If we were creating and destroying the gang cleanly we wouldn't need D this, but theGang is created with a top-level unsafePerformIO. & Hacks beget hacks beget hacks... XIssue work requests for the T and wait until they complete. 4If the gang is already busy then print a warning to  and just : run the actions sequentially in the requesting thread. (Run an action on the gang sequentially. 'Run an action on the gang in parallel. YSame as X but in the  monad. TUVWXYTUVWXYUTWVXY TUVWXY!NonePSequential reduction of a multidimensional array along the innermost dimension.  NParallel reduction of a multidimensional array along the innermost dimension. L Each output value is computed by a single thread, with the output values 5 distributed evenly amongst the available threads.  6Sequential reduction of all the elements in an array.  LParallel tree reduction of an array to a single value. Each thread takes an O equally sized chunk of the data and computes a partial sum. The main thread ? then reduces the array of partial sums to the final result. We don'Ft require that the initial value be a neutral element, so each thread K computes a fold1 on its chunk of the data, and the seed element is only ( applied in the final reduction step.  *This is the primitive reduction function. G We use manual specialisations and rewrite rules to avoid the result * being boxed up in the final iteration.  5Sequentially reduce values between the given indices vector to write elements into 0function to get an element from the given index (binary associative combination function 'starting value (typically an identity) &inner dimension (length to fold over)  vector to write elements into 0function to get an element from the given index )binary associative combination operator -starting value. Must be neutral with respect  ^ to the operator. eg  0 + a = a. &inner dimension (length to fold over)  0function to get an element from the given index &binary associative combining function starting value number of elements  0function to get an element from the given index &binary associative combining function starting value number of elements  Get data from the array. Function to combine elements. Starting value. Starting index in array. Ending index in array. Result.           "NoneZ%Select indices matching a predicate. ? This primitive can be useful for writing filtering functions. [2Select indices matching a predicate, in parallel. ? This primitive can be useful for writing filtering functions. F The array is split into linear chunks, with one chunk being given to  each thread. H The number of elements in the result array depends on how many threads  you're running the program with. Z&Update function to write into result. See if this predicate matches. &.. and apply fn to the matching index )Extent of indices to apply to predicate. 1Number of elements written to destination array. [See if this predicate matches. Z[Z[#None\,Fill a block in a rank-2 array in parallel. G Blockwise filling can be more cache-efficient than linear filling for  rank-2 arrays. 9 Coordinates given are of the filled edges of the block. G We divide the block into columns, and give one column to each thread. > Each column is filled in row major order from top to bottom. ],Fill a block in a rank-2 array in parallel. V Blockwise filling can be more cache-efficient than linear filling for rank-2 arrays. R Using cursor functions can help to expose inter-element indexing computations to " the GHC and LLVM optimisers. 9 Coordinates given are of the filled edges of the block. G We divide the block into columns, and give one column to each thread. > Each column is filled in row major order from top to bottom. ^.Fill a block in a rank-2 array, sequentially. V Blockwise filling can be more cache-efficient than linear filling for rank-2 arrays. R Using cursor functions can help to expose inter-element indexing computations to " the GHC and LLVM optimisers. 9 Coordinates given are of the filled edges of the block. < The block is filled in row major order from top to bottom. \-Update function to write into result buffer. .Function to evaluate the element at an index. Width of the whole array. &x0 lower left corner of block to fill y0 w0 width of block to fill. h0 height of block to fill. ] -Update function to write into result buffer. 'Make a cursor to a particular element. Shift the cursor by an offset. .Function to evaluate the element at an index. Width of the whole array. &x0 lower left corner of block to fill y0 w0 width of block to fill h0 height of block to fill ^ -Update function to write into result buffer. 'Make a cursor to a particular element. Shift the cursor by an offset. 4Function to evaluate an element at the given index. Width of the whole array. 'x0 lower left corner of block to fill. y0 w0 width of block to fill h0 height of block to fill \]^\]^$None_Fill something sequentially. 6 The array is filled linearly from start to finish. `.Fill a block in a rank-2 array, sequentially. G Blockwise filling can be more cache-efficient than linear filling for  rank-2 arrays. < The block is filled in row major order from top to bottom. aFill something in parallel. ( The array is split into linear chunks, / and each thread linearly fills one chunk. bHFill something in parallel, using a separate IO action for each thread. ( The array is split into linear chunks, / and each thread linearly fills one chunk. _Number of elements. -Update function to write into result buffer. &Fn to get the value at a given index. `-Update function to write into result buffer. (Fn to get the value at the given index. Width of the whole array. 'x0 lower left corner of block to fill. y0 w0 width of block to fill h0 height of block to fill aNumber of elements. -Update function to write into result buffer. &Fn to get the value at a given index. bNumber of elements. 'Update fn to write into result buffer. /Create a fn to get the value at a given index.  The first * is the thread number, so you can do some  per-thread initialisation. _`ab_`ab NonecMDelayed arrays are represented as functions from the index to element value. HEvery time you index into a delayed array the element at that position  is recomputed. d*O(1). Wrap a function as a delayed array. eDO(1). Produce the extent of an array, and a function to retrieve an  arbitrary element. fO(1). Delay an array. @ This wraps the internal representation to be a function from ( indices to elements, so consumers don't need to worry about ) what the previous representation was. /Compute a range of elements in a rank-2 array. "Compute all elements in an array. %Compute elements of a delayed array. cdefcdefcdefcdef NonegUnstructured traversal. hUnstructured traversal. i0Unstructured traversal over two arrays at once. j0Unstructured traversal over two arrays at once. k2Unstructured traversal over three arrays at once. l2Unstructured traversal over three arrays at once. m1Unstructured traversal over four arrays at once. n1Unstructured traversal over four arrays at once. gSource array. .Function to produce the extent of the result. ,Function to produce elements of the result. A It is passed a lookup function to get elements of the source. hSource array. .Function to produce the extent of the result. ,Function to produce elements of the result. A It is passed a lookup function to get elements of the source. iFirst source array. Second source array. .Function to produce the extent of the result. ,Function to produce elements of the result. 8 It is passed lookup functions to get elements of the  source arrays. jFirst source array. Second source array. .Function to produce the extent of the result. ,Function to produce elements of the result. 8 It is passed lookup functions to get elements of the  source arrays. klmnghijklmnghijklmnghijklmn None o0Impose a new shape on the elements of an array. > The new extent must be the same size as the original, else . pAppend two arrays. qAppend two arrays. r1Transpose the lowest two dimensions of an array. 1 Transposing an array twice yields the original. s/Extract a sub-range of elements from an array. t!Backwards permutation of an array' s elements. u!Backwards permutation of an array' s elements. v)Default backwards permutation of an array' s elements.  If the function returns ' then the value at that index is taken  from the default array (arrDft) w)Default backwards permutation of an array' s elements.  If the function returns ' then the value at that index is taken  from the default array (arrDft) x;Extend an array, according to a given slice specification. BFor example, to replicate the rows of an array use the following: #extend arr (Any :. (5::Int) :. All)y;Extend an array, according to a given slice specification. BFor example, to replicate the rows of an array use the following: #extend arr (Any :. (5::Int) :. All)z@Take a slice from an array, according to a given specification. <For example, to take a row from a matrix use the following:  "slice arr (Any :. (5::Int) :. All)To take a column use: slice arr (Any :. (5::Int)){@Take a slice from an array, according to a given specification. <For example, to take a row from a matrix use the following:  "slice arr (Any :. (5::Int) :. All)To take a column use: slice arr (Any :. (5::Int))opqrsStarting index. Size of result. tExtent of result array. 0Function mapping each index in the result array " to an index of the source array. Source array. uExtent of result array. 0Function mapping each index in the result array " to an index of the source array. Source array. vDefault values (arrDft) 0Function mapping each index in the result array " to an index in the source array. Source array. wDefault values (arrDft) 0Function mapping each index in the result array " to an index in the source array. Source array. xyz{ opqrstuvwxyz{ opqrstuvwxyz{opqrstuvwxyz{%Nonehjlnuwy{uw{yhjlnNone|'Interleave the elements of two arrays. 8 All the input arrays must have the same extent, else . M The lowest dimension of the result array is twice the size of the inputs.  - interleave2 a1 a2 b1 b2 => a1 b1 a2 b2 - a3 a4 b3 b4 a3 b3 a4 b4 })Interleave the elements of three arrays. ~(Interleave the elements of four arrays. |}~|}~|}~|}~NoneIStrict ByteStrings arrays are represented as ForeignPtr buffers of Word8  O(1). Wrap a  as an array. O(1). Unpack a  from an array. Read elements from a . None5Arrays represented as foreign buffers in the C heap.  O(1). Wrap a  as an array. O(1). Unpack a  from an array. DCompute an array sequentially and write the elements into a foreign > buffer without intermediate copying. If you want to copy a 8 pre-existing manifest array to a foreign buffer then f it first. CCompute an array in parallel and write the elements into a foreign > buffer without intermediate copying. If you want to copy a 8 pre-existing manifest array to a foreign buffer then f it first. Filling foreign buffers.  %Read elements from a foreign buffer.  ! &NoneFill something in parallel. H The array is split into linear chunks and each thread fills one chunk. Number of elements. -Update function to write into result buffer. &Fn to get the value at a given index. None?Hints that computing this array will be an unbalanced workload @ and evaluation should be interleaved between the processors. +Wrap an array with a unbalanced-ness hint. "#$"#None(Parallel computation of array elements. : The source array must have a delayed representation like c, C or P, 3 and the result a manifest representation like U or F. D If you want to copy data between manifest representations then use   instead. J If you want to convert a manifest array back to a delayed representation  then use f instead. *Sequential computation of array elements. 2Suspended parallel computation of array elements. EThis version creates a thunk that will evaluate the array on demand. H If you force it when another parallel computation is already running L then you will get a runtime warning and evaluation will be sequential.  Use  and ( to ensure that each array is evaluated ' before proceeding to the next one. ,If unsure then just use the monadic version . This one ensures 9 that each array is fully evaluated before continuing. Parallel copying of arrays. 7 This is a wrapper that delays an array before calling . A You can use it to copy manifest arrays between representations. Sequential copying of arrays. +Suspended parallel copy of array elements. Monadic version of . DForces an suspended array computation to be completed at this point  in a monadic computation. ' do let arr2 = suspendedComputeP arr1  ...  arr3 < - now $ arr2  ... # !"#$%&'()*+,-Z[\]^_`ab# !"#$%&+,-()*'_ab\`^]Z[NonePartitioned arrays. % The last partition takes priority These are produced by Repa'3s support functions and allow arrays to be defined : using a different element function for each partition. The basic idea is described in ``&Efficient Parallel Stencil Convolution'', L Ben Lippmeier and Gabriele Keller, Haskell 2011 -- though the underlying D array representation has changed since this paper was published. 2Check whether an index is within the given range. %(Read elements from a partitioned array. &'%(&'%None3Unboxed arrays are represented as unboxed vectors. The implementation uses Data.Vector.Unboxed which is based on type I families and picks an efficient, specialised representation for every I element type. In particular, unboxed vectors of pairs are represented  as pairs of unboxed vectors. A This is the most efficient representation for numerical data. +Sequential computation of array elements..  This is an alias for  with a more specific type. (Parallel computation of array elements.  This is an alias for  with a more specific type. 1O(n). Convert a list to an unboxed vector array.  This is an alias for ' with a more specific type. *O(1). Wrap an unboxed vector as an array. .O(1). Unpack an unboxed vector from an array. O(1). Zip some unboxed arrays. + The shapes must be identical else . O(1). Zip some unboxed arrays. + The shapes must be identical else . O(1). Zip some unboxed arrays. + The shapes must be identical else . O(1). Zip some unboxed arrays. + The shapes must be identical else . O(1). Zip some unboxed arrays. + The shapes must be identical else . O(1). Unzip an unboxed array. O(1). Unzip an unboxed array. O(1). Unzip an unboxed array. O(1). Unzip an unboxed array. O(1). Unzip an unboxed array. )"Filling of unboxed vector arrays. *,Read elements from an unboxed vector array. )*+)*NoneAProduce an array by applying a predicate to a range of integers. F If the predicate matches, then use the second function to generate  the element. < This is a low-level function helpful for writing filtering  operations on arrays. 0 Use the integer as the index into the array you're filtering. #If the Int matches this predicate, /... then pass it to this fn to produce a value "Range between 0 and this maximum. "Array containing produced values. None"An array with undefined elements. K This is normally used as the last representation in a partitioned array, G as the previous partitions are expected to provide full coverage. ,1Undefined array elements. Inspecting them yields . -,.-,NoneCursored Arrays.  These are produced by Repa')s stencil functions, and help the fusion @ framework to share index compuations between array elements. The basic idea is described in ``&Efficient Parallel Stencil Convolution'', L Ben Lippmeier and Gabriele Keller, Haskell 2011 -- though the underlying D array representation has changed since this paper was published. Define a new cursored array. //Compute a range of elements in a rank-2 array. 0*Compute all elements in an rank-2 array. 1&Compute elements of a cursored array. Create a cursor for an index. Shift a cursor by an offset. #Compute the element at the cursor. /01 .23456/01NoneStructured versions of map and zipWith" that preserve the representation ( of cursored and partitioned arrays. For cursored (C:) arrays, the cursoring of the source array is preserved. For partitioned (P7) arrays, the worker function is fused with each array I partition separately, instead of treating the whole array as a single  bulk object. Preserving the cursored and/*or paritioned representation of an array K is will make follow-on computation more efficient than if the array was # converted to a vanilla Delayed (D) array as with plain  and . 8If the source array is not cursored or partitioned then  and  ' are identical to the plain functions. "The target result representation.  Structured map.  Structured zipWith. G If you have a cursored or partitioned source array then use that as ( the third argument (corresponding to r1 here) 6Apply a worker function to each element of an array, . yielding a new array with the same extent. :Combine two arrays, element-wise, with a binary operator. 2 If the extent of the two array arguments differ,  then the resulting array' s extent is their intersection. 789:;<=>? 789:;<=>?None0Check if an index lies inside the given extent.  As opposed to  from Data.Array.Repa.Index, K this is a short-circuited test that checks that lowest dimension first. 1Check if an index lies outside the given extent.  As opposed to  from Data.Array.Repa.Index, J this is a short-circuited test that checks the lowest dimension first. GGiven the extent of an array, clamp the components of an index so they I lie within the given array. Outlying indices are clamped to the index " of the nearest border element. IMake a 2D partitioned array from two others, one to produce the elements M in the internal region, and one to produce elements in the border region. - The two arrays must have the same extent. 3 The border must be the same width on all sides. Extent of array. Index to check. Extent of array. Index to check. Extent of array. Index to clamp. Extent of array. Width of border. Array for internal elements. Array for border elements.  None@A index into the flat array. 3 Should be abstract outside the stencil modules. Like " but with the parameters flipped. 0Apply a stencil to every element of a 2D array. ANLike above, but treat elements outside the array has having a constant value. BLLike above, but clamp out of bounds array values to the closest real value. C&Data template for stencils up to 7x7. @D)How to handle the boundary of the array. Stencil to apply. Array to apply stencil to. EABCRSRS@DEABCNone LSequential reduction of the innermost dimension of an arbitrary rank array. Combine this with  transpose to fold any other dimension. IParallel reduction of the innermost dimension of an arbitray rank array. CThe first argument needs to be an associative sequential operator. J The starting element must be neutral with respect to the operator, for  example 0 is neutral with respect to (+) as  0 + a = a. J These restrictions are required to support parallel evaluation, as the S starting element may be used multiple times depending on the number of threads. MSequential reduction of an array of arbitrary rank to a single scalar value. KParallel reduction of an array of arbitrary rank to a single scalar value. CThe first argument needs to be an associative sequential operator. F The starting element must be neutral with respect to the operator,  for example 0 is neutral with respect to (+) as  0 + a = a. J These restrictions are required to support parallel evaluation, as the S starting element may be used multiple times depending on the number of threads. 4Sequential sum the innermost dimension of an array. 2Parallel sum the innermost dimension of an array. 0Sequential sum of all the elements of an array. +Parallel sum all the elements of an array. ICheck whether two arrays have the same shape and contain equal elements,  in parallel. ICheck whether two arrays have the same shape and contain equal elements,  sequentially. F FNone%Arrays represented as boxed vectors. DThis representation should only be used when your element type doesn't  have an Unbox- instsance. If it does, then use the Unboxed U " representation will be faster. *Sequential computation of array elements.  This is an alias for compute with a more specific type. (Parallel computation of array elements. .O(n). Convert a list to a boxed vector array.  This is an alias for ' with a more specific type. 'O(1). Wrap a boxed vector as an array. +O(1). Unpack a boxed vector from an array. G Filling of boxed vector arrays. H)Read elements from a boxed vector array. GHIGHNonet 0123456789:;<=>?@ABCDEFGcdefgikmopqrstvxz|}~!$(+.23456I>cdefopqsrtvzxgikm|}~None HIJKLMNOPQ HIJKLMPONQJ'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^__``abcdefghijkkllmnopqrstuvw x y z { | } ~""###$$$$  &               ! " # $ %&'( ) *&+, y -!!!!!.!/!0!1!2!3!4!5 6 7 8 9&:;&<= >?@AB&CDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcde f g h i f jklmno repa-3.2.2.2Data.Array.Repa.Repr.UnboxedData.Array.Repa.EvalData.Array.Repa.ShapeData.Array.RepaData.Array.Repa.Repr.HintSmallData.Array.Repa.IndexData.Array.Repa.SliceData.Array.Repa.StencilData.Array.Repa.Stencil.Dim2Data.Array.Repa.Eval.GangData.Array.Repa.Repr.Delayed#Data.Array.Repa.Operators.Traversal$Data.Array.Repa.Operators.IndexSpace$Data.Array.Repa.Operators.InterleaveData.Array.Repa.Repr.ByteStringData.Array.Repa.Repr.ForeignPtr#Data.Array.Repa.Repr.HintInterleave Data.Array.Repa.Repr.Partitioned#Data.Array.Repa.Operators.SelectionData.Array.Repa.Repr.UndefinedData.Array.Repa.Repr.Cursored!Data.Array.Repa.Operators.Mapping Data.Array.Repa.Specialised.Dim2#Data.Array.Repa.Operators.ReductionData.Array.Repa.Repr.Vector!Data.Array.Repa.Stencil.PartitionData.Array.Repa.Eval.EltData.Array.Repa.BaseData.Array.Repa.Eval.TargetData.Array.Repa.Eval.LoadData.Array.Repa.Stencil.Base Data.Array.Repa.Stencil.TemplateData.Array.Repa.Eval.ReductionData.Array.Repa.Eval.SelectionData.Array.Repa.Eval.CursoredData.Array.Repa.Eval.ChunkedData.Array.Repa.Unsafe Data.Array.Repa.Eval.Interleaved vector-0.9.1Data.Vector.Unboxed.BaseUnboxElttouchzerooneShaperankzeroDimunitDim intersectDimaddDimsize sizeIsValidtoIndex fromIndex inShapeRange listOfShape shapeOfListdeepSeqinShape showShapeSourceArrayextentindex unsafeIndex linearIndexunsafeLinearIndex deepSeqArray!toList deepSeqArraysTargetMVecnewMVecunsafeWriteMVecunsafeFreezeMVec deepSeqMVec touchMVecfromList LoadRange loadRangeS loadRangePLoadloadSloadPS hintSmallDIM5DIM4DIM3DIM2DIM1DIM0:.Zix1ix2ix3ix4ix5Slice sliceOfFull fullOfSlice SliceShape FullShapeAnyAllStencil StencilStatic stencilExtent stencilZero stencilAccBoundary BoundClamp BoundConst BoundFixed makeStencil makeStencil2stencil2GangtheGanggangSizeforkGanggangIOgangSTselectChunkedSselectChunkedP fillBlock2PfillCursoredBlock2PfillCursoredBlock2S fillLinearS fillBlock2S fillChunkedPfillChunkedIOPD fromFunction toFunctiondelaytraverseunsafeTraverse traverse2unsafeTraverse2 traverse3unsafeTraverse3 traverse4unsafeTraverse4reshapeappend++ transposeextract backpermuteunsafeBackpermutebackpermuteDftunsafeBackpermuteDftextend unsafeExtendslice unsafeSlice interleave2 interleave3 interleave4BfromByteString toByteStringFfromForeignPtr toForeignPtr computeIntoS computeIntoPfillInterleavedPIhintInterleavecomputePcomputeSsuspendedComputePcopyPcopySsuspendedCopyPnowRangePinRangeUcomputeUnboxedScomputeUnboxedPfromListUnboxed fromUnboxed toUnboxedzipzip3zip4zip5zip6unzipunzip3unzip4unzip5unzip6selectPXC makeCursored StructuredTRsmapszipWithmapzipWith+^-^*^/^ isInside2 isOutside2clampToBorder2 makeBordered2PC5 forStencil2 mapStencil2foldSfoldPfoldAllSfoldAllPsumSsumPsumAllSsumAllPequalsPequalsSVcomputeVectorScomputeVectorPfromListVector fromVectortoVectorRegionSizeOffset regionOfSize offsetRegionpartitionForStencilregionXregionY regionWidth regionHeightghc-primGHC.Primseq $fElt(,,,,,) $fElt(,,,,) $fElt(,,,) $fElt(,,)$fElt(,) $fEltWord64 $fEltWord32 $fEltWord16 $fEltWord8 $fEltWord $fEltInt64 $fEltInt32 $fEltInt16 $fEltInt8$fEltInt $fEltDouble $fEltFloat $fEltBool GHC.TypesIntFalse$fLoadRangeSshe $fLoadSshe $fSourceSaASmallstage $fShape:.$fShapeZ $fSlice:. $fSlice:.0 $fSliceAny$fSliceZ parseStencil2 makeStencil2' _gangThreads_gangRequestVars_gangResultVars _gangBusyReq ReqShutdownReqDo gangWorkerfinaliseWorkerbaseGHC.IO.Handle.FDstderrseqIOparIOGHC.STST $fShowGangreduce reduceAny reduceInt reduceFloat reduceDoubleunboxInt unboxFloat unboxDouble$fLoadRangeD:.e $fLoadDshe $fSourceDaADelayedGHC.Errerror Data.MaybeNothingbytestring-0.10.0.0Data.ByteString.Internal ByteString$fSourceBWord8 AByteStringGHC.ForeignPtr ForeignPtr $fTargetFe $fSourceFa AForeignPtr $fLoadIshe $fSourceIa AInterleave $fSourcePe deepSeqRange $fLoadPsheAPart $fTargetUe $fSourceUaAUnboxed $fSourceXe $fLoadXshe AUndefined$fLoadRangeC:.e $fLoadC:.e $fSourceCa ACursoredcursoredExtent makeCursor shiftCursor loadCursor$fStructuredXab$fStructuredUab$fStructuredIab$fStructuredSab$fStructuredPab$fStructuredFab$fStructuredDab$fStructuredCab$fStructuredBWord8bCursorunsafeAppStencilCursor2_constunsafeAppStencilCursor2_clamp template7x7unsafeAppStencilCursor2 $fEqArray $fTargetVe $fSourceVaAVector