Merge pull request #118 from omlins/nested-functions

Support early returns in nested functions

src/ParallelKernel/kernel_language.jl

@@ -86,12 +86,24 @@ Call a macro analogue to `Base.@println`, compatible with the package for parall
 macro pk_println(args...) check_initialized(); esc(pk_println(args...)); end
 
 
+##
+macro return_value(args...) check_initialized(); checksinglearg(args...); esc(return_value(args...)); end
+
+
+##
+macro return_nothing(args...) check_initialized(); checknoargs(args...); esc(return_nothing(args...)); end
+
+
 ## ARGUMENT CHECKS
 
 function checknoargs(args...)
     if (length(args) != 0) @ArgumentError("no arguments allowed.") end
 end
 
+function checksinglearg(args...)
+    if (length(args) != 1) @ArgumentError("wrong number of arguments.") end
+end
+
 function checkargs_sharedMem(args...)
     if !(2 <= length(args) <= 3) @ArgumentError("wrong number of arguments.") end
 end
@@ -177,6 +189,17 @@ function pk_println(args...; package::Symbol=get_package())
 end
 
 
+## FUNCTION FOR DIVERSE TASKS
+
+function return_value(value)
+    return :(return $value)
+end
+
+function return_nothing()
+    return :(return)
+end
+
+
 ## CPU TARGET IMPLEMENTATIONS
 
 macro gridDim_cpu()   esc(:(ParallelStencil.ParallelKernel.Dim3($(RANGELENGTHS_VARNAMES[1]), $(RANGELENGTHS_VARNAMES[2]), $(RANGELENGTHS_VARNAMES[3])))) end

src/ParallelKernel/parallel.jl

@@ -156,7 +156,7 @@ function parallel_kernel(caller::Module, package::Symbol, numbertype::DataType,
     if (!isa(indices,Symbol) && !isa(indices.head,Symbol)) @ArgumentError("@parallel_indices: argument 'indices' must be a tuple of indices or a single index (e.g. (ix, iy, iz) or (ix, iy) or ix ).") end
     indices = extract_tuple(indices)
     body = get_body(kernel)
-    body = remove_returns(body)
+    body = remove_return(body)
     use_aliases = !all(indices .== INDICES[1:length(indices)])
     if use_aliases # NOTE: we treat explicit parallel indices as aliases to the statically retrievable indices INDICES.
         indices_aliases = indices

src/ParallelKernel/shared.jl

@@ -142,36 +142,36 @@ function push_to_signature!(kernel::Expr, arg::Expr)
     return kernel
 end
 
-function remove_returns(body::Expr)
+function remove_return(body::Expr)
     if !(body.args[end] in [:(return), :(return nothing), :(nothing)])
         @ArgumentError("invalid kernel in @parallel kernel definition: the last statement must be a `return nothing` statement ('return' or 'return nothing' or 'nothing') as required for any GPU kernels.")
     end
-    body = make_nested_returns_implicit(body)
+    body = disguise_nested_returns(body)
     remainder = copy(body)
     remainder.args = body.args[1:end-2]
     if inexpr_walk(remainder, :return) @ArgumentError("invalid kernel in @parallel kernel definition: only one return statement is allowed in the kernel (exception: nested function definitions) and it must return nothing and be the last statement (required to ensure equal behaviour with different packages for parallellization).") end
     return remainder
 end
 
-function make_nested_returns_implicit(body::Expr)
+function disguise_nested_returns(body::Expr)
     return postwalk(body) do ex
         if isdef(ex)
             f_elems = splitdef(ex)
             body = f_elems[:body]
-            f_elems[:body] = make_returns_implicit(body)
+            f_elems[:body] = disguise_returns(body)
             return combinedef(f_elems)
         else
             return ex
         end
     end
 end
 
-function make_returns_implicit(body::Expr)
+function disguise_returns(body::Expr)
     return postwalk(body) do ex
         if @capture(ex, return x_)
-            return x
+            return :(ParallelStencil.ParallelKernel.@return_value($x))
         elseif @capture(ex, return)
-            return nothing
+            return :(ParallelStencil.ParallelKernel.@return_nothing)
         else
             return ex
         end

src/kernel_language.jl

@@ -12,10 +12,6 @@ macro memopt(args...) check_initialized(); checkargs_memopt(args...); esc(memopt
 macro shortif(args...) check_initialized(); checktwoargs(args...); esc(shortif(args...)); end
 
 
-##
-macro return_nothing(args...) check_initialized(); checknoargs(args...); esc(return_nothing(args...)); end
-
-
 ## ARGUMENT CHECKS
 
 function checknoargs(args...)
@@ -487,11 +483,6 @@ function shortif(else_val, if_expr; package::Symbol=get_package())
 end
 
 
-function return_nothing()
-    return :(return)
-end
-
-
 ## FUNCTIONS FOR SHARED MEMORY ALLOCATION
 
 

src/parallel.jl

@@ -1,4 +1,4 @@
-import .ParallelKernel: get_name, set_name, get_body, set_body!, add_return, remove_returns, extract_kwargs, split_parallel_args, extract_tuple, substitute, literaltypes, push_to_signature!, add_loop, add_threadids, promote_maxsize
+import .ParallelKernel: get_name, set_name, get_body, set_body!, add_return, remove_return, extract_kwargs, split_parallel_args, extract_tuple, substitute, literaltypes, push_to_signature!, add_loop, add_threadids, promote_maxsize
 
 const PARALLEL_DOC = """
     @parallel kernel
@@ -185,7 +185,7 @@ function parallel_indices_memopt(metadata_module::Module, metadata_function::Exp
     if (!isa(indices,Symbol) && !isa(indices.head,Symbol)) @ArgumentError("@parallel_indices: argument 'indices' must be a tuple of indices, a single index or a variable followed by the splat operator representing a tuple of indices (e.g. (ix, iy, iz) or (ix, iy) or ix or I...).") end
     if (!isa(optvars,Symbol) && !isa(optvars.head,Symbol)) @ArgumentError("@parallel_indices: argument 'optvars' must be a tuple of optvars or a single optvar (e.g. (A, B, C) or A ).") end
     body = get_body(kernel)
-    body = remove_returns(body)
+    body = remove_return(body)
     body = add_memopt(metadata_module, is_parallel_kernel, caller, package, body, indices, optvars, loopdim, loopsize, optranges, useshmemhalos, optimize_halo_read)
     body = add_return(body)
     set_body!(kernel, body)
@@ -198,7 +198,7 @@ function parallel_kernel(metadata_module::Module, metadata_function::Expr, calle
     memopt = haskey(kwargs, :memopt) ? kwargs.memopt : get_memopt()
     indices = get_indices_expr(ndims).args
     body = get_body(kernel)
-    body = remove_returns(body)
+    body = remove_return(body)
     validate_body(body)
     kernelargs = splitarg.(extract_kernel_args(kernel)[1])
     argvars = (arg[1] for arg in kernelargs)

src/shared.jl

@@ -9,7 +9,7 @@ elseif ENABLE_AMDGPU
     using AMDGPU
 end
 import MacroTools: @capture, postwalk, splitarg # NOTE: inexpr_walk used instead of MacroTools.inexpr
-import .ParallelKernel: eval_arg, split_args, split_kwargs, extract_posargs_init, extract_kernel_args, is_kernel, is_call, gensym_world, isgpu, @isgpu, substitute, inexpr_walk, cast, @ranges, @rangelengths
+import .ParallelKernel: eval_arg, split_args, split_kwargs, extract_posargs_init, extract_kernel_args, is_kernel, is_call, gensym_world, isgpu, @isgpu, substitute, inexpr_walk, cast, @ranges, @rangelengths, @return_value, @return_nothing
 import .ParallelKernel: PKG_CUDA, PKG_AMDGPU, PKG_THREADS, PKG_NONE, NUMBERTYPE_NONE, SUPPORTED_NUMBERTYPES, SUPPORTED_PACKAGES, ERRMSG_UNSUPPORTED_PACKAGE, INT_CUDA, INT_AMDGPU, INT_THREADS, INDICES, PKNumber, RANGES_VARNAME, RANGES_TYPE, RANGELENGTH_XYZ_TYPE, RANGELENGTHS_VARNAMES, THREADIDS_VARNAMES, GENSYM_SEPARATOR, AD_SUPPORTED_ANNOTATIONS
 import .ParallelKernel: @require, @symbols, symbols, longnameof, @prettyexpand, @prettystring, prettystring, @gorgeousexpand, @gorgeousstring, gorgeousstring