Procedure Expand_Pragma_Inspection_Point (N : Node_Id) Procedure Expand_Pragma_Import_Or_Interface (N : Node_Id) + procedure Expand_Pragma_CUDA_Execute (N : Node_Id) Procedure Expand_Pragma_Common_Object (N : Node_Id) Procedure Expand_Pragma_Check (N : Node_Id)
Procedure Expand_Pragma_Abort_Defer (N : Node_Id) With Exp_Util use -67,6 +68,7 package body Exp_Prag is Obtains the first element of the given element list or, if the list hasĭiff -git a/gcc/ada/exp_prag.adb b/gcc/ada/exp_prag.adb + - Create a new element list containing the given arguments.įunction First_Elmt (List : Elist_Id) return Elmt_Id + Elmt4 : Node_Or_Entity_Id) return Elist_Id + Elmt3 : Node_Or_Entity_Id) return Elist_Id + Elmt2 : Node_Or_Entity_Id) return Elist_Id
+ function New_Elmt_List (Elmt1 : Node_Or_Entity_Id) return Elist_Id list is then subsequently filled in using Append calls. a field in some other node which points to an element list where the + L : constant Elist_Id := New_Elmt_List (Elmt1, Elmt2, Elmt3) ĭiff -git a/gcc/ada/elists.ads b/gcc/ada/elists.ads
+ L : constant Elist_Id := New_Elmt_List (Elmt1, Elmt2) + L : constant Elist_Id := New_Elmt_List (Elmt1) + L : constant Elist_Id := New_Elmt_List + function New_Elmt_List (Elmt1 : Node_Or_Entity_Id) * rtsfind.adb: Extend Interfaces_Descendant to includeĭiff -git a/gcc/ada/elists.adb b/gcc/ada/elists.adb RE_Pop_Call_Configuration, RE_Launch_Kernel, RO_IC_Unsigned, * rtsfind.ads: Add CUDA.Internal, CUDA.Runtime, System.C (Expand_N_Pragma): Add call to Expand_Pragma_CUDA_Execute. (Expand_Pragma_CUDA_Execute): New function. * exp_prag.adb: Add dependency on Elists. * elists.adb (New_Elmt_List): New functions. * elists.ads (New_Elmt_List): New functions. Tested on x86_64-pc-linux-gnu, committed on trunk These functions requires putting arguments of CUDA_Execute in specificĭata structures and thus requires a declare block to hold these The hidden functions CUDA.Push_Call_Configuration andĬUDA.Pop_Call_Configuration but also to CUDA.Launch_Kernel. This commit implements expansion of the CUDA_Execute pragma.ĬUDA_Execute must result in multiple calls to the runtime API: calls to You can change the fields of grid and block with assignments like grid.Implement expansion of CUDA_Execute pragma public inbox for help / color / mirror / Atom feed * Implement expansion of CUDA_Execute pragma 7:35 Pierre-Marie de Rodat 0 siblings, 0 replies only message in threadįrom: Pierre-Marie de Rodat 7:35 UTC ( / raw) Any field not provided during initialization is initialized to 1. In the case of your interest, you will have dim3 grid(m*n) įrom the definition of dim3, it is not needed to explicitly initialize the fields of grid and block. You can define a dim3 variable as dim3 grid(256) // defines a grid of 256 x 1 x 1 blocksĭim3 block(512,512) // defines a block of 512 x 512 x 1 threads _host_ _device_ dim3(unsigned int vx = 1, unsigned int vy = 1, unsigned int vz = 1) : x(vx), y(vy), z(vz) Dim3 is an integer struct type defined in the file vector_types.h as struct _device_builtin_ dim3
0 kommentar(er)
