CUDA Fortran

Programming Guide and Reference

Release 2011

The Portland Group

While every precaution has been taken in the preparation of this document, The Portland Group® (PGI®), a wholly-owned subsidiary of STMicroelectronics, Inc., makes no warranty for the use of its products and assumes no responsibility for any errors that may appear, or for damages resulting from the use of the information contained herein. The Portland Group retains the right to make changes to this information at any time, without notice. The software described in this document is distributed under license from STMicroelectronics and/or The Portland Group and may be used or copied only in accordance with the terms of the license agreement ("EULA"). PGI Workstation, PGI Server, PGI Accelerator, PGF95, PGF90, PGFORTRAN, and PGI Unified Binary are trademarks; and PGI, PGHPF, PGF77, PGCC, PGC++, PGI Visual Fortran, PVF, PGI CDK, Cluster Development Kit, PGPROF, PGDBG, and The Portland Group are registered trademarks of The Portland Group Incorporated. No part of this document may be reproduced or transmitted in any form or by any means, for any purpose other than the purchaser's or the end user's personal use without the express written permission of STMicroelectronics and/or The Portland Group.

Copyright © 2010 The Portland Group® and STMicroelectronics, Inc.All rights reserved. Printed in the United States of America Release 2010, version 10.8, August 2010 Release 2010, version 10.9, September 2010 Release 2011, version 11.0, November 2010

Technical support: trs@pgroup.com Sales: sales@pgroup.com Web: www.pgroup.com

ID: 103151117

Contents
Preface ...................................................................................................................................... ix
Intended Audience ................................................................................................................. ix Organization .......................................................................................................................... ix Conventions ........................................................................................................................... ix Terminology ........................................................................................................................... x Related Publications ................................................................................................................ x

1. Introduction ........................................................................................................................ 1 2. Programming Guide .......................................................................................................... 3
CUDA Fortran Kernels .............................................................................................................. 3 Thread Blocks ........................................................................................................................ 4 Memory Hierarchy .................................................................................................................. 4 Subroutine / Function Qualifiers ................................................................................................ 4 Attributes(host) .............................................................................................................. 5 Attributes(global) ............................................................................................................ 5 Attributes(device) ........................................................................................................... 5 Restrictions .................................................................................................................... 5 Variable Qualifiers .................................................................................................................. 5 Attributes(device) ........................................................................................................... 5 Attributes(constant) ........................................................................................................ 6 Attributes(shared) ........................................................................................................... 6 Attributes(pinned) .......................................................................................................... 6 Datatypes in Device Subprograms ............................................................................................. 6 Predefined Variables in Device Subprograms .............................................................................. 7 Execution Configuration ........................................................................................................... 7 Asynchronous concurrent execution .......................................................................................... 8 Concurrent Host and Device Execution .............................................................................. 8 Concurrent Stream Execution ........................................................................................... 8 Kernel Loop Directive .............................................................................................................. 8 Restrictions on the CUF kernel directive ........................................................................... 10 Building a CUDA Fortran Program ........................................................................................... 11 Emulation Mode ................................................................................................................... 11 iii

3. Reference ........................................................................................................................... 13
New Subroutine and Function Attributes ................................................................................... Host Subroutines and Functions ...................................................................................... Global Subroutines ........................................................................................................ Device Subroutines and Functions ................................................................................... Restrictions on Device Subprograms ................................................................................ Variable Attributes ................................................................................................................. Device data .................................................................................................................. Constant data ................................................................................................................ Shared data .................................................................................................................. Value dummy arguments ................................................................................................ Pinned arrays ............................................................................................................... Allocating Device and Pinned Arrays ........................................................................................ Allocating Device Memory .............................................................................................. Allocating Device Memory Using Runtime Routines ............................................................ Allocating Pinned Memory .............................................................................................. Data transfer between host and device memory ......................................................................... Data Transfer Using Assignment Statements ....................................................................... Implicit Data Transfer in Expressions ............................................................................... Data Transfer Using Runtime Routines ............................................................................. Invoking a kernel subroutine .................................................................................................. Device code ......................................................................................................................... Datatypes allowed ......................................................................................................... Builtin variables ............................................................................................................ Fortran intrinsics .......................................................................................................... New Intrinsic Functions ................................................................................................. Warp-Vote Operations .................................................................................................... Atomic Functions .......................................................................................................... Restrictions .................................................................................................................. Host code ............................................................................................................................ SIZEOF Intrinsic ............................................................................................................ 13 13 13 14 14 14 14 15 16 17 17 17 17 18 18 19 19 19 20 20 21 21 22 22 24 26 27 29 29 29

4. Runtime APIs .................................................................................................................... 31
Initialization ......................................................................................................................... Device Management ............................................................................................................... cudaGetDeviceCount ...................................................................................................... cudaSetDevice ............................................................................................................... cudaSetDeviceFlags ........................................................................................................ cudaGetDevice .............................................................................................................. cudaGetDeviceProperties ................................................................................................ cudaChooseDevice ......................................................................................................... Thread Management .............................................................................................................. cudaThreadSynchronize ................................................................................................. cudaThreadExit ............................................................................................................. Memory Management ............................................................................................................ cudaMalloc .................................................................................................................. iv 31 31 31 31 31 32 32 32 32 32 32 32 33

cudaMallocPitch ............................................................................................................ cudaFree ...................................................................................................................... cudaMallocArray ........................................................................................................... cudaFreeArray .............................................................................................................. cudaMemset ................................................................................................................. cudaMemset2D ............................................................................................................. cudaMemcpy ................................................................................................................ cudaMemcpyAsync ........................................................................................................ cudaMemcpy2D ............................................................................................................ cudaMemcpy2DAsync .................................................................................................... cudaMemcpyToArray ...................................................................................................... cudaMemcpy2DToArray .................................................................................................. cudaMemcpyFromArray .................................................................................................. cudaMemcpy2DFromArray .............................................................................................. cudaMemcpyArrayToArray .............................................................................................. cudaMemcpy2DArrayToArray .......................................................................................... cudaMalloc3D .............................................................................................................. cudaMalloc3DArray ....................................................................................................... cudaMemset3D ............................................................................................................. cudaMemcpy3D ............................................................................................................ cudaMemcpy3DAsync .................................................................................................... cudaMemcpyToSymbol ................................................................................................... cudaMemcpyFromSymbol ............................................................................................... cudaMemcpyToSymbolAsync ........................................................................................... cudaMemcpyFromSymbolAsync ....................................................................................... cudaGetSymbolAddress ................................................................................................... cudaGetSymbolSize ........................................................................................................ cudaHostAlloc ............................................................................................................... cudaHostGetDevicePointer .............................................................................................. cudaMallocHost ............................................................................................................ cudaFreeHost ................................................................................................................ Stream Management .............................................................................................................. cudaStreamCreate .......................................................................................................... cudaStreamQuery .......................................................................................................... cudaStreamSynchronize .................................................................................................. cudaStreamDestroy ........................................................................................................ Event Management ................................................................................................................ cudaEventCreate ............................................................................................................ cudaEventRecord .......................................................................................................... cudaEventQuery ............................................................................................................ cudaEventSynchronize .................................................................................................... cudaEventDestroy .......................................................................................................... cudaEventElapsedTime ................................................................................................... Error Handling ..................................................................................................................... cudaGetLastError .......................................................................................................... cudaGetErrorString ........................................................................................................

33 33 33 33 34 34 34 34 34 35 35 35 35 35 36 36 36 36 36 36 37 37 37 37 37 38 38 38 38 39 39 39 39 39 39 39 40 40 40 40 40 40 40 41 41 41 v

............................................................................................................ Examples ....Version Management ........................................................................................ 41 5............................................................................. 43 Matrix Multiplication Example ............................................................................................................................. 47 vi ................................................................................................................................................................................................................................. 41 cudaDriverGetVersion ........................................................... 41 cudaRuntimeGetVersion .................. Contact Information ........................ 45 6........... 43 Source Code Description .......... 43 Source Code Listing ...................................................................................

...................... 22 3.....................................10............1............... 28 3...........................5..................................................9....... Fortran Bit Manipulation Intrinsics .............................. 21 3............ Fortran Numeric and Logical Intrinsics .................. Fortran Random Number Intrinsics .................. Counting Atomic Functions ........................................ Device Code Intrinsic Datatypes ...... 23 3................................................... 6 3...6.............. Intrinsic Datatypes ................. 24 3...................................... Fortran Numeric Inquiry Intrinsics .. 24 3.......................................................................................Tables 2.......2..... Compare and Swap Atomic Function ................................. Arithmetic and Bitwise Atomic Functions ...........................................................4.............................. 27 3...................... 29 vii ....................................8........................................................................................ 23 3..................................3........7......................... Fortran Mathematical Intrinsics ............ 23 3.....................................1................................................................................................................................ Fortran Reduction Intrinsics ......................................

viii .

C. Conventions This guide uses the following conventions: italic is used for emphasis. a small set of extensions to Fortran that supports and is built upon the CUDA computing architecture. “Runtime APIs” describes the interface between CUDA Fortran and the CUDA Runtime API Chapter 5. and Intel 64 hardware platforms. This guide assumes familiarity with basic operating system usage. Organization The organization of this document is as follows: Chapter 1. The PGI tools are available on a variety of operating systems for the X86. “Programming Guide” serves as a programming guide for CUDA Fortran Chapter 3. and/or C++ languages. AMD64.Preface This document describes CUDA Fortran. ix . “Reference” describes the CUDA Fortran language reference Chapter 4. “Introduction” contains a general introduction Chapter 2. “Examples” provides sample code and an explanation of the simple example. scientists and engineers proficient in programming with the Fortran. Intended Audience This guide is intended for application programmers.

htm Related Publications The following documents contain additional information related to CUDA Fortran programming. • NVIDIA CUDA Programming Guide. 2009. NVIDIA. arguments. examples. • ISO/IEC 1539-1:1997. Available online at http:// www. options.com/cuda. June 2010. 7/21/2010. Version 3. x . square brackets are required to specify a p/t-set. The PGI compilers and tools are supported on both 32-bit and 64-bit variants of the Linux and Windows operating systems on a variety of x86-compatible processors.1. multiple filenames are allowed.com/cuda. [ item1 ] in general. including assembly language statements. directories.Terminology Constant Width is used for filenames. September. Available online at http://www. Version 3. PGI provides a glossary of terms which you can access at www.pdf.com/doc/pgiug. Information Technology – Programming Languages – FORTRAN.com/support/definitions.nvidia.pgroup. In this example. and for language statements in the text. There are a wide variety of releases and distributions of each of these types of operating systems. you must select either item2 or item3. Zero or more of the preceding item may occur. C/C++ C/C++ language statements are shown in the test of this guide using a reduced fixed point size. In this case. In this case item1 is optional.pgroup.. ellipsis indicate a repetition. NVIDIA. • NVIDIA CUDA Compute Unified Device Architecture Reference Manual. 1997 (Fortran 95). filename .. In the context of p/tsets. square brackets indicate optional items. • PGI Users Guide. Bold is used for commands. The Portland Group. Available online at http:// www. Release 10.nvidia. Terminology If there are terms in this guide with which you are unfamiliar.1.9. Geneva. { item2 | item 3} braces indicate that a selection is required.1. FORTRAN Fortran language statements are shown in the text of this guide using a reduced fixed point size.

which can cooperate using shared memory and barrier synchronization. CUDA comes with an extended C compiler. A properly designed CUDA program will run on any CUDA-enabled GPU. shared memory and barrier synchronization within thread groups. A CUDA programmer must partition the program into coarse grain blocks that can be executed in parallel. Compared to the PGI Accelerator directives-based model and compilers. highly parallel computational units with very high memory bandwidth. but are general enough to be useful in many data-parallel.Chapter 1. Each block is partitioned into fine grain threads. PGI 2011 includes support for CUDA Fortran on Linux. NVIDIA introduced CUDA™. CUDA Fortran is a lower-level explicit programming model with substantial runtime library components that give expert programmers direct control of all aspects of GPGPU programming. a small set of extensions to Fortran that supports and is built upon the CUDA computing architecture. Introduction Welcome to Release 2011 of PGI CUDA Fortran. and tremendous potential for many applications. CUDA Fortran is an analog to NVIDIA's CUDA C compiler. and coordinated independent thread groups organized into a grid. a general purpose parallel programming architecture. regardless of the number of available processor cores. CUDA Fortran includes a Fortran 2003 compiler and tool chain for programming NVIDIA GPUs using Fortran. and back • Writing subroutines and functions to execute on the GPU • Invoking GPU subroutines from the host • Allocating pinned memory on the host • Using asynchronous transfers between the host and GPU 1 . Mac OS X and Windows. with compilers and libraries to support the programming of NVIDIA GPUs. here called CUDA C. GPU designs are optimized for the computations found in graphics rendering. The programming model supports four key abstractions: cooperating threads organized into thread groups. allowing direct programming of the GPU from a high level language. Graphic processing units or GPUs have evolved into programmable. The CUDA Fortran extensions described in this document allow the following operations in a Fortran program: • Declaring variables that are allocated in the GPU device memory • Allocating dynamic memory in the GPU device memory • Copying data from the host memory to the GPU memory. compute-intensive programs.

2 .

and a local thread index within its thread block. In this example.Chapter 2. A kernel is defined using the attributes(global) specifier on the subroutine statement. y ) real. each thread performs one iteration of the common SAXPY loop operation. the call to the kernel ksaxpy specifies n/64 thread blocks. 64>>>( n. and a thread index accessed through threadidx. Each thread is assigned a thread block index accessed through the built-in blockidx variable. The CUDA threads are organized into thread blocks. a. y real :: a integer :: n ! call the kernel call ksaxpy<<<n/64. “Reference.y real. refer to Chapter 3. A call to a kernel specifies how many parallel instances of the kernel must be executed. Such a subroutine is called a device kernel or kernel. y ) end subroutine In this case. each instance will be executed by a different CUDA thread. value :: a integer. 3 . a.” on page 13. CUDA Fortran Kernels CUDA Fortran allows the definition of Fortran subroutines that execute in parallel on the GPU when called from the Fortran program which has been invoked and is running on the host. a kernel is called using special chevron syntax to specify the number of thread blocks and threads within each thread block: ! Kernel definition attributes(global) subroutine ksaxpy( n. device. Programming Guide This chapter introduces the CUDA programming model through examples written in CUDA Fortran. each with 64 threads. a. x. dimension(*) :: x. value :: n. and each thread has a global thread block index. For a reference for CUDA Fortran. x. y ) real. dimension(*) :: x. x. i i = (blockidx%x-1) * blockdim%x + threadidx%x if( i <= n ) y(i) = a * x(i) + y(i) end subroutine ! Host subroutine subroutine solve( n.

high bandwidth software managed cache memory. two thread blocks may be executed in parallel or one after the other. • A subprogram declared with attributes(global) or attributes(device) is called a device subprogram.or two-dimensional grid of blocks. and by synchronizing at a barrier using the SYNCTHREADS() intrinsic. Each thread can also have private local memory. but it is read-only to the threads and limited in size. For a two-dimensional thread block of size (Dx. by the same core or by different cores. the thread index for each dimension starts at one. and if the latter. Data in constant memory space is initialized by the host program. the host program can directly access data in the host main memory. On the device side. the first argument in the chevron <<<>>> syntax is the grid size.Dz). • A subprogram declared with attributes(host). Threads in the same thread block can access and share data in shared memory. the thread index is equal to the thread ID. data in shared memory has a lifetime of the thread block. The thread blocks are organized into a one. data in thread local memory may be implemented as processor registers or may be allocated in the global device memory. On the host side.Dy). and the second argument is the thread block size. again implemented using DMA access. The shared memory acts like a low-latency. each with the same thread block size. For a three-dimensional thread block of size (Dx. Memory Hierarchy CUDA Fortran programs have access to several memory spaces. so are slow relative to the host memory.Thread Blocks Thread Blocks Each thread is assigned a thread block index accessed through the built-in blockidx variable. In CUDA Fortran. such data copies require DMA access to the device. It can also directly copy data to and from the device global memory. or called from another device subprogram. A kernel may be invoked with many thread blocks. For a one-dimensional thread block. the thread ID is equal to (x+Dx(y-1)). so each thread has a thread index within the block. 4 . or three-dimensional index. Accesses to constant memory are typically faster than accesses to global memory. Currently. all threads can read data in constant memory. the thread ID is (x+Dx(y-1)+Dy(z-1)). Each thread in the block waits at the call to SYNCTHREADS() until all threads have reached that call. two-. data in global device memory can be read or written by all threads. The thread index may be a one-. The host can also set the values in the device constant memory. best performance will often be obtained when thread local data Subroutine / Function Qualifiers A subroutine or function in CUDA Fortran has an additional attribute. and a block index within the grid. designating whether it is executed on the host or on the device. called from the host. and a thread index accessed through threadidx. Thread blocks must be able to execute independently.Dy. When invoking a kernel. or with the host attribute by default. is called a host subprogram. A unique thread ID is assigned to each thread. whether it is a kernel. computed from the thread index. Threads in the same thread block may cooperate by using shared memory. the maximum number of threads in a thread block is 512.

” on page 7. which declares in which memory the data is allocated. or device attributes is specified. • A device subprogram must not contain variables with the SAVE attribute. Dummy arguments in a device subprogram must not have the pointer attribute. • A device subprogram must not have optional arguments. as described in “Predefined Variables in Device Subprograms. and may only be called from the host using a kernel call containing the chevron syntax and runtime mapping parameters. The default is attributes(host). • A kernel subroutine may not also have the device or host attribute. the host routine making the call continues to execute before the device has completed its execution of the kernel subroutine. shared. 5 . At most one of the device. By default. to be executed on the device. Attributes(host) The host attribute. if none of the host. constant.Chapter 2. specified on the subroutine or function statement. that is. declares that the subroutine or function is to be executed on the host. specified on the subroutine or function statement. Restrictions The following restrictions apply to subprograms: • A device subprogram must not be recursive. and is allocated in the device global memory. such a routine must be called from a subprogram with the global or device attribute. and may not contain any subroutines or functions. Attributes(global) The global attribute may only be specified on a subroutine statement. Such a call is asynchronous. Variable Qualifiers Variables in CUDA Fortran have a new attribute. • Device subprograms may not be contained in a host subroutine or function. global. the variable may be accessed by any subprogram in that module and by any subprogram that uses the module. • If declared in a module. variables declared in modules or host subprograms are allocated in the host main memory. or with data initialization. Attributes(device) A variable with the device attribute is called a device variable. declares that the subprogram is to be executed on the device. it declares that the subroutine is a kernel subroutine. Such a subprogram can only be called from another host subprogram. or pinned attributes may be specified for a variable. • Calls to a kernel subroutine must specify the execution configuration. Programming Guide • A subroutine declared with attributes(global) is also called a kernel subroutine. Attributes(device) The device attribute.

an allocatable array. it will be allocated in host pagelocked memory. Attributes(pinned) A variable with the pinned attributes is called a pinned variable. or by other device subprograms to which it is passed as an argument. It can be read or written by all threads in the block.4. A shared variable may only be declared in a device subprogram. A device array may be an explicit-shape array. or size of page-locked memory. from when it is allocated until it is deallocated. A pinned variable must be an allocatable array.4.8 1. the variable may be accessed by that subprogram or subprograms contained in that subprogram. the variable may be accessed by any subprogram in that module and by any subprogram that uses the module. If declared in a module.8 4. and have a lifetime of the entire application. The advantage of using pinned variables is that copies from page-locked memory to device memory are faster than copies from normal paged host memory.2. An allocatable device variable has a dynamic lifetime. or an assumed-shape dummy array. but may be modified in host subprograms.1. A shared variable may not be data initialized. Device constant data may not be assigned or modified in any device subprogram. availability.2. though a write in one thread is only guaranteed to be visible to other threads after the next call to the SYNCTHREADS() intrinsic. When a pinned variable is allocated. and may only be accessed within that subprogram. Device constant variables may not be allocatable. Intrinsic Datatypes Type integer logical real double precision Type Kind 1. Attributes(shared) A variable with the shared attributed is called a device shared variable or a shared variable. Device constant variables are allocated in the device constant memory space. A shared variable is allocated in the device shared memory for a thread block. Other device variables have a lifetime of the entire application. Datatypes in Device Subprograms The following intrinsic datatypes are allowed in device subprograms and device data: Table 2.Datatypes in Device Subprograms • If declared in a host subprogram. the variable will be allocated in the normal host paged memory.8 equivalent to real(kind=8) 6 . Attributes(constant) A variable with the constant attributes is called a device constant variable. if the allocation in page-locked memory fails. Some operating systems or installations may restrict the use. and has a lifetime of the thread block.

• The variable griddim contains the dimensions of the grid.or two-dimensional thread blocks. the value of grid%z must be one. in addition to the statically allocated shared memory. Threads are executed in groups of 32. the value of each component must be equal to or greater than one. two-.y. This memory is used for the 7 . • The variable is warpsize is declared to be type integer.Chapter 2. or are type(dim3).block[. the threadidx%y and/or threadidx%z components have the value one. and the product of the component values must be less than or equal to 512. and a stream identifier. blockidx%y has the value one.streamid]]>>>(arg1. The value of blockidx%z is always one. are not accessible in host subprograms. called warps. for one. • The value of bytes must be an integer.z end type These predefined variables. • If grid is type(dim3). warpsize contains the number of threads in a warp.8 1 Derived types may contain members with these intrinsic datatypes or other allowed derived types. as with threadidx. A kernel subroutine call looks like this: call kernel<<<grid. except for warpsize.bytes[. • The variable blockdim contains the dimensions of the thread block. for one.arg2.or two-dimensional grids and one-. to support concurrent stream execution on the device.…) where • grid and block are either integer expressions (for one-dimensional grids and thread blocks). for one-dimensional grids. blockdim has the same value for all thread blocks in the same grid. The execution configuration defines the dimensionality and extent of the grid and thread blocks that execute the subroutine. These variables are of type dim3. it specifies the number of bytes of shared memory to be allocated for each thread block. and block%x and block%y must be equal to or greater than one. the value of griddim%z is always one. in bytes. Predefined Variables in Device Subprograms Device subprograms have access to block and grid indices and dimensions through several builtin read-only variables. Programming Guide complex character(len=1) 4. the module cudafor defines the derived type dim3 as follows: type(dim3) integer(kind=4) :: x. Execution Configuration A call to a kernel subroutine must specify an execution configuration. It may also specify a dynamic shared memory extent. or three-dimensional thread blocks. • If block is type(dim3). • The variable threadidx contains the thread index within its thread block. • The variable blockidx contains the block index within the grid.

Concurrent Host and Device Execution When a host subprogram calls a kernel subroutine. stream zero is special in that operations on the stream zero queue will begin only after all preceding operations on all queues are complete. The one exception to this rule is a scalar reduction operation. Asynchronous concurrent execution There are two components to asynchronous concurrent execution with CUDA Fortran. and will only be initiated when all previous operations on that queue have been completed. For these operations. the host can continue executing. its value is treated as zero. the launch is asynchronous. used when no stream identifier is specified.3. If not specified. • The value of streamid must be an integer greater than or equal to zero. Each user-created stream manages its own queue. An operation is placed at the end of the stream queue. The call can be treated as a kernel launch operation. including kernel execution and data copies to and from device memory. In this way. and may execute concurrently with each other. Launch configuration and mapping of the loop iterations onto the hardware is controlled and specified as part of the directive body using the familiar CUDA chevron syntax. By calling the runtime routine cudaThreadSynchronize. where the launch actually corresponds to placing the kernel on a queue for execution by the device. the call actually returns to the host program before the kernel subroutine begins execution. An application can manage more concurrency by using multiple streams. is stream zero. The work in the loops specified by the directive is executed in parallel. Kernel Loop Directive CUDA Fortran allows automatic kernel generation and invocation from a region of host code containing one or more tightly nested loops. the host program can synchronize and wait for all previously launched or queued kernels. it is the programmer's responsibility to ensure that parallel execution is legal and produces the correct answer. Programmers must be careful when using concurrent host and device execution. The default stream. the host program should synchronize with the device to avoid erroneous results. The program can use cudaThreadSynchronize() or CUDA Events to wait for the completion of the kernel. it specifies the stream to which this call is associated. including calling or queueing more kernels for execution on the device. see Section 3. and no subsequent operations on any queue begin until the stream zero operation is complete. in cases where the host program reads or modifies device or constant data.Asynchronous concurrent execution assumed-size shared variables in the thread block. the compiler handles the generation of the final reduction kernel. As with any kernel. such as summing the values in a vector or matrix. 8 . are implemented using stream queues. operations on different stream queues may execute out-of-order with respect to when they were placed on the queues.2. Concurrent Stream Execution Operations involving the device. inserting synchronization into the kernel as appropriate. across the thread blocks and grid.

n a(i. (32. Alternatively. you could specify a configuration like the following in which the threads read and write the matrices in coalesced fashion.j) end do end do Without an explicit n on the do. The body of the doubly-nested loop is turned in the kernel body: ThreadIdx%x runs from 1 to 32 and is mapped onto the inner i loop. specified as (*. the directive defines a two-dimensional thread block of size 32x4. The grid and block values can be an integer scalar or a parenthesized list. 256 >>> do j = 1. is computed by the compiler and runtime by dividing the loop trip counts n and m by the thread block size. Programming Guide Syntax The general form of the kernel directive is: !$cuf kernel do[(n)] <<< grid. Alternatively.Chapter 2. n a(i. that is.*).j) +c(i. (256. the default value is 1.j) = b(i. The inner i dimension is run sequentially on each thread. m do i = 1.j) = b(i. so all iterations are computed.4) >>> do j = 1. only the outer j loop is run in parallel with a thread block size of 256.j) end do end do In this example. Example 2 !$cuf kernel do <<< *. starting at loop n and working out.j) = b(i. ThreadIdx%y runs from 1 to 4 and is mapped onto the outer j loop.j) +c(i. m do i = 1. In this case. block >>> The compiler maps the launch configuration specified by the grid and block values onto the outermost n loops. n a(i. You might consider if the code in Example 2 would perform better if the two loops were interchanged. the schedule applies just to the outermost loop. Loops which are not mapped onto the grid and block values are run sequentially on each thread. m do i = 1.*).j) end do end do 9 . !$cuf kernel do(2) <<< *. Example 1 !$cuf kernel do(2) <<< (*.1) >>> do j = 1. The grid shape.j) +c(i. using asterisks tells the compiler to choose a thread block shape and/or compute the grid shape from the thread block shape and the loop limits.

if they are allowed in device code. as illustrated here: !$cuf kernel do(2) <<< (1. • The tightly-nested DO loops must have invariant loop limits: the lower limit. in concert. IF statements.Kernel Loop Directive Example 3 In Example 2. one for each thread. it must be followed by at least that many tightly-nested DO loops. and increment must be invariant with respect to any other loop in the kernel do. • Scalars used or assigned in the loop must either have the device attribute. (256. • CUDA Fortran intrinsic functions are allowed.j) end do end do Now the threads in a thread block handle all values in the i direction. • Implicit loops and F90 array syntax are not allowed within the directive loops. upper limit. Restrictions on the CUF kernel directive The following restrictions apply to CUF kernel directives: • If the directive specifies n dimensions. To "unroll" more work into each thread. loops. • Only CUDA Fortran dataypes are allowed within the loops. specify non-asterisk values for the grid. n a(i. Except in the case of reductions. incrementing by 256.j) +c(i. are not. gridDim%x is equal to m. or the compiler will make a device copy of that variable live for the duration of the loops. • The body of the loops may contain assignment statements. atomic functions. • Subroutine and function calls to attributes(device) subprograms are allowed if they are in the same module as the code containing the directive. 10 . the 256 threads in each block each do one element of the matrix addition. m do i = 1. In CUDA built-in language.j) = b(i. One thread block is created for each j. Further expansion of the work along the i direction and all work across the j dimension is handled by the mapping onto the grid dimensions. • Arrays used or assigned in the loop must have the device attribute. and GOTO statements. when a reduction has a scalar target. but the device-specific intrinsics such as syncthreads. • There can be no GOTO or EXIT statements within or between any loops that have been mapped onto the grid and block configuration values. Specifically. Summation Example The simplest directive form for performing a dot product on two device arrays takes advantage of the properties for scalar use outlined previously. the j loop is mapped onto the grid x-dimension.*). the compiler generates a correct sequence of synchronized operations to produce one copy either in device global memory or on the host. etc.1) >>> do j = 1. because the compiler skips over the constant 1 in the i loop grid size.

allowing the programmer to use a host debugger. * >>> do i = 1. 11 . * >>> do i = 1. such as memory races. the compiler recognizes rsum as a scalar temporary which should be allocated locally on every thread. add the command line option –Mfixed.cuf extension specifies that the file is a free-format CUDA Fortran program. Building a CUDA Fortran Program CUDA Fortran is supported by the PGI Fortran compilers when the filename uses a CUDA Fortran extension. the . the host floating point units and intrinsics are used. However. compile and link with the –Mcuda=emu command line option. the compiler recognizes the use of the scalar and generates just one final result. This will not expose certain errors. In particular.CUF extension may also be used. To compile a fixed-format program. The . In emulation mode. n rsum = rsum + x(i) * y(i) end do For reductions. Emulation Mode PGI Fortran compilers support an emulation mode for program development on workstations or systems without a CUDA-enabled GPU and for debugging. CUDA Fortran extensions can be enabled in any Fortran source file by adding the –Mcuda command line option. This CUF kernel can be followed by another CUF kernel in the same subprogram: !cuf kernel do <<< *.Chapter 2. n rsum = x(i) * y(i) z(i) = rsum end do In this CUF kernel. in which case the program is processed by the preprocessor before being compiled. Programming Guide rsum = 0. the device code is compiled for and runs on the host. emulation mode may execute a single thread block at a time. To build a program using emulation mode. In emulation mode. which may produce slightly different answers than the device units and intrinsics.0 !cuf kernel do <<< *. It’s important to note that the emulation is far from exact. use of rsum on the host following this loop is undefined.

12 .

or elemental. Host Subroutines and Functions The host attribute may be explicitly specified on the Subroutine or Function statement as follows: attributes(host) subroutine sub(…) attributes(host) integer function func(…) integer attributes(host) function func(…) The host attributes prefix may be preceded or followed by any other allowable subroutine or function prefix specifiers (recursive. A kernel subroutine is compiled as a kernel for execution on the device. A subprogram with no attributes prefix has the host attribute by default. These attributes are specified using the attributes(attr) prefix on the Subroutine or Function statement. as described in the following sections. so no other subroutine prefixes are allowed. Global Subroutines The global attribute may be explicitly specified on the Subroutine statement as follows: attributes(global) subroutine sub(…) Functions may not have the global attribute. to be called from a host routine using an execution configuration. A kernel subroutine may not be recursive. or both. A kernel subroutine may not be contained in another subroutine or function. New Subroutine and Function Attributes CUDA Fortran adds new attributes to subroutines and functions. A subroutine with the global attribute is called a kernel subroutine.Chapter 3. and may not contain any other subprogram. A Function may have the host or device attribute. or a host subprogram. then default rules are used. if there is no attributes prefix on the subprogram statement. 13 . Reference This chapter is the CUDA Fortran Language Reference. elemental. their meaning and restrictions. pure. A Subroutine may have the host. or device attribute. A host subprogram is compiled for execution on the host processor. global. This chapter describes how to specify the new attributes. pure. function return datatype). A subroutine or function with the host attribute is called a host subroutine or function. or may have both host and device attribute.

The advantage of using pinned memory is that transfers between the device and pinned memory are faster and can be asynchronous. Variable Attributes CUDA Fortran adds new attributes for variables and arrays. in constant memory space. A subroutine or function with the device or kernel attribute is called a device subprogram. in page-locked host memory space. Variables in modules may be declared to be in device global memory or constant memory space. Variables declared in a device subprogram may have one of four attributes: they may be declared to be in device global memory. to be device arrays of size 100. that is. The device attribute can be specified with the attributes statement. • It may not contain another subprogram. CUDA Fortran adds a new attribute for allocatable arrays in host memory.Variable Attributes Device Subroutines and Functions The device attribute may be explicitly specified on the Subroutine or Function statement as follows: attributes(device) subroutine sub(…) attributes(device) datatype function func(…) datatype attributes(device) function func(…) A subroutine or function with the device attribute may not be recursive. and may only be called from device subprograms in the same module. or elemental. • It may not be pure or elemental. • It may not be contained in another subroutine or function. the array may be declared to be in pinned memory. nor have the pure or elemental prefix on the subprogram statement. A device subprogram is compiled for execution on the device. Restrictions on Device Subprograms A subroutine or function with the device or global attribute must satisfy the following restrictions: • It may not be recursive. A subroutine or function with the device attribute must appear within a Fortran module. so no other subroutine or function prefixes are allowed. pure. refer to “Device code. Device data A variable or array with the device attribute is defined to reside in the device global memory. nor have the recursive prefix on the subprogram statement. The following example declares two arrays. This section describes how to specify the new attributes and their meaning and restriction. or as an attribute on the type declaration statement. in the thread block shared memory. or in thread local memory. except for the function return datatype. For more information.” on page 21. 14 . a and b.

c and d. • Device variables and arrays may appear in modules. device :: b(100) These rules apply to device data: • An allocatable device array dynamically allocates device global memory. device data may only be used in the following manner: • In declaration statements • In Allocate and Deallocate statements • As an argument to the Allocated intrinsic function • As the source or destination in a data transfer assignment statement • As an actual argument to a kernel subroutine • As an actual argument to another host subprogram or runtime API call • As a dummy argument in a host subprogram A device array may have the allocatable attribute. to be constant arrays of size 100. 15 . constant :: d(100) These rules apply to constant data: • Constant data may not have the Pointer. in that case. • Device variables and arrays may be passed as actual arguments to host and device subprograms.Chapter 3. • Device variables and arrays declared in a host subprogram cannot have the Save attribute. Constant data A variable or array with the constant attribute is defined to reside in the device constant memory space. real :: c(100) attributes(constant) :: c real. or as an attribute on the type declaration statement. the subprogram interface must be explicit (in the Fortran sense). or may have adjustable extent. but may not be in a Common block or an Equivalence statement. Reference real :: a(100) attributes(device) :: a real. Target. • Members of a derived type may not have the device attribute unless they are allocatable. The following example declares two arrays. The constant attribute can be specified with the attributes statement. In host subprograms. or Allocatable attributes. but may not be in a Common block or an Equivalence statement. • Constant variables and arrays may appear in modules. and the matching dummy argument must also have the device attribute. • Device variables and arrays may not have the Pointer or Target attributes.

the value of the bytes argument in the execution configuration is used to specify the number of bytes of shared memory that is dynamically allocated for each thread block. or Allocatable attributes. • Within host subprograms. s and t. A shared variable or array may only be declared and used inside a device subprogram. real :: c(100) attributes(shared) :: c real. variables and arrays with the constant attribute may be read and written. as long as the subprogram interface is explicit. or as an attribute on the type declaration statement. The following example declares two arrays. that is. data with the constant attribute may only be used in the following manner: • In declaration statements • As the source or destination in a data transfer assignment statement • As an actual argument to another host subprogram • As a dummy argument in a host subprogram Shared data A variable or array with the shared attribute is defined to reside in the shared memory space of a thread block.Variable Attributes • Members of a derived type may not have the constant attribute. as long as the interface is explicit and the matching dummy argument has the shared attribute. • Arrays with the constant attribute must have fixed size. • Shared variables and arrays may be passed as actual arguments to from a device subprogram to another device subprogram. The shared attribute can be specified with the attributes statement. and the matching dummy argument also has the constant attribute. • Members of a derived type may not have the shared attribute. In host subprograms. Shared arrays that are not dummy arguments may be declared as assumed-size arrays. This memory is used for the 16 . Target. • Shared variables may not be in a Common block or Equivalence statement. shared :: x(*) Such an array has special significance. Its size is determined at run time by the call to the kernel. to be shared arrays of size 100. the last dimension of a shared array may have an asterisk as its upper bound: real. • Constant variables and arrays may be passed as actual arguments to host and device subprograms. When the kernel is called. • Within device subprograms. shared :: d(100) These rules apply to shared data: • Shared data may not have the Pointer. variables and arrays with the constant attribute may not be assigned or modified.

These arrays are dynamically allocated in host subprograms using the Allocate statement. value :: n In this case. real :: p(:) allocatable :: p attributes(pinned) :: p real. Value dummy arguments In device subprograms. dummy arguments are passed by default by reference. to be pinned allocatable arrays. dimension(n. If a shared array is not a dummy argument and not assumed-size. Reference assumed-size shared memory arrays in that thread block. The pinned attribute can be specified with the attributes statement. by adding the value attribute to the variable declaration. Scalar arguments can be passed by value. they are all implicitly equivalenced. b.Chapter 3. The following example declares two arrays.n) :: a. it will be automatically deallocated when the subprogram returns. p and q. If a device array declared in a host subprogram does not have the Save attribute. when such memory is available. Allocating Device and Pinned Arrays This section describes extensions to the Allocate statement. following the rules of Fortran. The variable arrays corresponding to the dummy arguments a and b must be set up before the call to reside on the device. as is done in C. pinned :: q(:) Pinned arrays may be passed as arguments to host subprograms regardless of whether the interface is explicit. An array with the pinned attribute may be declared in a module or in a host subprogram. the declaration for the array must still have the pinned attribute. or as an attribute on the type declaration statement. Pinned arrays An allocatable array with the pinned attribute will be allocated in special page-locked host memory. b integer. Where the array is deallocated. attributes(global) subroutine madd( a. Allocating Device Memory Device arrays can have the allocatable attribute. 17 . and other supported methods for allocating device memory. starting at the same shared memory address. and the address of the argument is passed to the subprogram. This means the actual argument must be stored in device global memory. Programmers must take this into account when coding. it must be fixed size. specifically for dynamically allocating device arrays and host pinned arrays. allocatable. if there is more than one assumed-size shared memory array. n ) real. the value of n can be passed from the host without needing to reside in device memory. or the deallocation may fail. and dynamically deallocated using the Deallocate statement. or whether the dummy argument has the pinned and allocatable attributes.

To use these. the compiler allocates the array in normal paged host memory. Mixing standard Fortran allocate/deallocate with the runtime Malloc/Free for a given array is not supported. if available. pinned allocatable arrays work and act like any other allocatable array on the host. device :: ndev allocate(ndev) ndev = 100 The language also supports the ability to create the equivalent of automatic and local device arrays without using the allocate statement. Allocating Pinned Memory Allocatable arrays with the pinned attribute are dynamically allocated using the Allocate statement.Allocating Device and Pinned Arrays real.c. The cudaMalloc function can be used to allocate single-dimensional arrays of the supported intrinsic datatypes. or if it is exhausted. 100) … istat = cudaFree(v) For a complete list of the memory management runtime routines. allocatable. refer to“Memory Management.” on page 32. These functions return memory which will bypass certain Fortran allocatable properties such as automatic deallocation. declare and initialize the scalar on the host as: integer. If no such memory space is available. allocatable. device :: adev(n) real. Otherwise.stat=istat) … if(allocated(b)) deallocate(b) Scalar variables can be allocated on the device using the Fortran 2003 allocatable scalar feature.stat=istat) … if(allocated(p)) deallocate(p) 18 . and thus the arrays are treated more like C malloc’ed areas. Fortran interfaces to the CUDA memory management runtime routines are provided. The compiler will generate code to allocate the array in host page-locked memory.n) real. device :: atmp(4) … end subroutine vfunc ! adev and atmp are deallocated Allocating Device Memory Using Runtime Routines For programmers comfortable with the CUDA C programming environment. and cudaFree can be used to free it: real. allocatable. pinned :: p(:) allocate(p(5000). device :: b(:) allocate(b(5024). These arrays will also have a lifetime of the subprogram as is usual with the Fortran language: subroutine vfunc(a. device :: v(:) istat = cudaMalloc(v. allocatable. real.

Note Using assignment statements to read or write device or constant data implicitly uses CUDA stream zero.not. meaning the data copy waits until all previous kernels and data copies complete. copies data from the device global memory to the host memory. Similarly. pinned=plog) if (. Implicit Data Transfer in Expressions Some limited data transfer can be enclosed within expressions. • An assignment statement where the left hand side is a device variable or device array or array section. and adev. and cdev are conforming device arrays. Reference To determine whether or not the allocation from page-locked memory was successful. .Chapter 3. In general. an additional PINNED keyword is added to the allocate statement. stat=istat. the following expressions are legal: a = adev adev = a b = a + adev c = x * adev + b 19 . . Data Transfer Using Assignment Statements You can copy variables and arrays from the host memory to the device memory by using simple assignment statements in host subprograms. plog) then . This means such data copies are synchronous. • An assignment statement where the left hand side is a host variable or host array or array section. if a. It returns a logical success value. and the right hand is a host variable or host array or array section. Compiler-generated temporary arrays are generated to accommodate the host copies of device data as needed. For instance. copies data from the host memory to the device global memory. b. Data transfer between host and device memory This section provides methods to transfer data between the host and device memory. which normally only allows one device array to appear on the righthand-side of an expression. bdev. you can use simple assignment statements to copy or assign variables or arrays with the constant attribute. • An assignment statement with a device variable or device array or array section on both sides of the assignment statement copies data between two device variables or arrays. and c are conforming host arrays. logical plog allocate(p(5000). the rule of thumb is all arithmetic or operations must occur on the host. and the right hand side is a device variable or device array or array section.

the value of grid%z must be one. and the implementation. Invoking a kernel subroutine A call to a kernel subroutine must give the execution configuration for the call. Data Transfer Using Runtime Routines For programmers comfortable with the CUDA C programming environment. bytes.” on page 32.1). or of type(dim3). it is converted to dim3(grid. bl • block is an integer. device :: wrk(1024) real cur(512) istat = cudaMemcpy(wrk. and their performance is dependent on the size of the slice. The execution configuration gives the size and shape of the grid and thread blocks that execute the function. Array slices are also supported. or of type(dim3). 20 . These functions can transfer data either from the host to device. Counts expressed in arguments to the Fortran runtime routines are expressed in terms of data type elements. or from one device array to another. it has the form: <<< grid. not bytes. and the associated stream. as well as the amount of shared memory to use for assumed-size shared memory arrays. it is converted to dim3(block. or would be more efficient written in another way. which must be less than the maximum supported by the device. amount of contiguous data in the slices. The product grid%x*grid%y gives the number of thread blocks to launch. If block is an integer. block. device to host.1.1. If grid is an integer. The execution configuration is specified after the subroutine name in the call statement. For a complete list of memory management runtime routines. cur. If it is type(dim3). or both: c = adev + bdev adev = adev + a b = sqrt(adev) Elemental transfers are supported by the language but will perform poorly. Fortran interfaces to the CUDA memory management runtime routines are provided. stream >>> • grid is an integer. The cudaMemcpy function can be used to copy data between the host and the GPU: real. the kind parameter to the Memcpy routines is optional in Fortran since the attributes of the arrays are explicitly declared. If it is type(dim3).Invoking a kernel subroutine The following expressions are not legal as they either promote a false impression of where the actual computation occurs. 512) For those familiar with the CUDA C routines.1). refer to “Memory Management. the number of threads per thread block is block%x*block%y*block%z.

It specifies the stream to which this call is enqueued. Device Code Intrinsic Datatypes Type integer logical real double precision complex character(len=1) Type Kind 1.2. defined as type(dim3) integer(kind=4) :: x. Device code Datatypes allowed Variables and arrays with the device.2.Chapter 3. See Section 4. and have a value of zero.8 1 (default) Additionally.” on page 16. Reference • bytes is optional. or shared attributes.8 equivalent to real(kind=8) 4(default). where the members of the derived type have one of the allowed intrinsic datatypes. If not specified. or declared in device subprograms. For instance. are limited to the types described in this section. if present. DB >>> ( A ) The function call will fail if the grid or block arguments are greater than the maximum sizes allowed. they may be of derived type. or another allowed derived type. if present. and specifies the number of bytes of shared memory to be allocated for each thread block to use for assumed-size shared memory arrays. The system module cudafor includes definitions of the derived type dim3. constant. it must be a scalar integer.4(default).y. For more information. a kernel subroutine attributes(global) subroutine sub( a ) can be called like: call sub <<< DG. such as the function arguments and execution configuration arguments.5 on page 41.4(default).8 1. or if bytes is greater than the shared memory available. it must be an integer. or a value returned by a call to cudaStreamCreate. • stream is optional.z end type 21 . allowing for static shared memory declared in the kernel and for other dedicated uses. refer to “Shared data. Table 3.8 4(default). the value zero is used.1. They may have any of the intrinsic datatypes in the following table.

Device code Builtin variables The system module cudafor declares several predefined variables. real. the value of griddim%z is always one. the threadidx%y and/or threadidx%z components have the value one. • The variable griddim contains the dimensions of the grid.or two-dimensional thread blocks. real integer. as with threadidx.2. the value of griddim%y is one for one-dimensional grids. real integer. real integer. blockidx. • The variables threadidx. real real Name int logical max min mod modulo nint real sign Argument Datatypes integer. • The variable blockdim contains the dimensions of the thread block. for onedimensional grids. currently defined to be 32. real real integer.or two-dimensional thread blocks. Fortran Numeric and Logical Intrinsics Name abs aimag aint anint ceiling cmplx conjg dim floor Argument Datatypes integer. Fortran intrinsics This section lists the Fortran intrinsic functions allowed in device subprograms.real) complex integer. real. The value of blockidx is the same for all threads in the same thread block. The value of blockidx%z is always one. blockdim. the value of griddim%z is always one. • The variable blockidx contains the block index within the grid. blockdim. complex integer. real. • The variable warpsize contains the number of threads in a warp. and griddim are available only in device subprograms. They are declared as follows: type(dim3) :: threadidx. for one. It has constant value. real 22 . The value of griddim is the same for all threads in the same grid. Fortran Numeric and Logical Intrinsics Table 3. for one. complex logical integer. These variables are read-only. blockidx. griddim integer(4) :: warpsize • The variable threadidx contains the thread index within its thread block. blockidx%y has the value one. blockdim has the same value for all threads in the same grid. the blockdim%y and/or blockdim %z components have the value one. complex complex real real real real or (real.

complex integer (integer. real real real Name precision radix range selected_int_kind selected_real_kind tiny Argument Datatypes real. real real integer.integer) real Fortran Bit Manipulation Intrinsics Table 3. complex real real. complex real real. complex real real.4. complex integer. Fortran Bit Manipulation Intrinsics Name btest iand ibclr ibits ibset ieor ior Argument Datatypes integer integer integer integer integer integer integer Name ishft ishftc leadz mvbits not popcnt poppar Argument Datatypes integer integer integer integer integer integer integer 23 . complex real real Fortran Numeric Inquiry Intrinsics Table 3. real. complex Name log log10 sin sinh sqrt tan tanh Argument Datatypes real.3. real integer.Chapter 3. Reference Fortran Mathematical Intrinsics Table 3.real) real.5. Fortran Numeric Inquiry Intrinsics Name bit_size digits epsilon huge maxexponent minexponent Argument Datatypes integer integer. Fortran Mathematical Intrinsics Name acos asin atan atan2 cos cosh exp Argument Datatypes real real real (real.

Fortran Reduction Intrinsics Name all any count maxloc maxval Argument Datatypes logical logical logical integer. real. real integer. To avoid these potential issues. refer to “Thread Management. This 24 . real integer. it has no arguments: void syncthreads() Sometimes threads within a block access the same addresses in shared or global memory. complex integer. Synchronization Functions The synchronization functions control the synchronization of various threads during execution of thread blocks. real Name minloc minval product sum Argument Datatypes integer.6. real integer. or write-after-write hazards for some of these memory accesses. thus creating potential read-after-write. . complex Fortran Random Number Intrinsics Table 3. write-after-read.7. Fortran Random Number Intrinsics Name random_number random_seed Argument Datatypes real integer New Intrinsic Functions This section describes the new intrinsic functions and subroutines supported in device subprograms. SYNCTHREADS The syncthreads intrinsic subroutine acts as a barrier synchronization for all threads in a single thread block.Device code Fortran Reduction Intrinsics Table 3. syncthreads syncthreads_count syncthreads_and syncthread_or threadfence threadfence_block threadfence_system For detailed information on these functions.” on page 32. real. use syncthreads()to specify synchronization points in the kernel.

acts as a barrier at which all threads in the block must wait before any thread is allowed to proceed. suppose you use a kernel to compute the sum of an array of N numbers in one call.Chapter 3. when a thread issues a series of writes to memory in a particular order. when a thread issues a series of writes to memory in a particular order. creating a wait. each block atomically increments a counter to signal that it is done with computing and storing its partial sum. threadfence_block(). other threads may see the effects of these memory writes in a different order. For example. In addition. syncthreads_count evaluates the integer argument int_value for all threads of the block and returns the number of threads for which int_value evaluates to non-zero. SYNCTHREADS_AND integer syncthreads_and(int_value) syncthreads_and. all threads must also reach and execute the same call statement. and threadfence_system() to create a memory fence to enforce ordering. In addition. Reference intrinsic acts as a barrier at which all threads in the block must wait before any thread is allowed to proceed. other threads may see the effects of these memory writes in a different order. syncthreads_and evaluates the integer argument int_value for all threads of the block and returns non-zero if and only if int_value evaluates to non-zero for all of them. acts a as a barrier at which all threads in the block must wait before any thread is allowed to proceed. acts as a barrier at which all threads in the block must wait before any thread is allowed to proceed. or the kernel fails to complete correctly. SYNCTHREADS_COUNT integer syncthreads_count(int_value) syncthreads_count. like syncthreads. the last block done reads each of these partial sums from global memory and sums them to obtain the final result. Each block first sums a subset of the array and stores the result in global memory. To determine which block is finished last. If any thread in a thread block issues a call to syncthreads. syncthreads_or evaluates the integer argument int_value for all threads of the block and returns non-zero if and only if int_value evaluates to non-zero for any of them. like syncthreads. THREADFENCE void threadfence() threadfence acts as a memory fence. SYNCTHREADS_OR integer syncthreads_or(int_value) syncthreads_or. Memory Fences In general. like syncthreads. In addition. 25 . Each thread in a thread block pauses at the syncthreads call until all threads have reached that call. You can use threadfence(). Typically. If no fence is placed between storing the partial sum and incrementing the counter. When all blocks are done. Threads within a block cooperate and share data by synchronizing their execution to coordinate memory accesses. the counter might increment before the partial sum is stored.

All global and shared memory accesses made by the calling thread prior to threadfence() are visible to: • All threads in the thread block for shared memory accesses • All threads in the device for global memory accesses THREADFENCE_BLOCK void threadfence_block() threadfence_block acts as a memory fence. The value of the function is .true.0) ) allneg = . as described in “cudaGetDeviceProperties. GPU_TIME The gpu_time intrinsic returns the value of the clock cycle counter on the GPU. 26 . creating a wait until all global and shared memory accesses made by the calling thread prior to threadfence_system() are visible to: • All threads in the thread block for shared memory accesses • All threads in the device for global memory accesses • Host threads for page-locked host memory accesses threadfence_system() is only supported by devices of compute capability 2. THREADFENCE_SYSTEM void threadfence_system() threadfence_system acts as a memory fence.Device code threadfence() is one method to enforce a specific order. The clock frequency can be determined by calling cudaGetDeviceProperties.” on page 32. Each of these functions has a single argument.2 and higher. The function allthreads evaluates its argument for all threads in the current warp.true. Warp-Vote Operations Warp-vote operations are only supported by devices with compute capability 1.true. It has a single argument: integer(8) clock call gpu_time(clock) The argument to gpu_time is set to the value of the clock cycle counter. creating a wait until all global and shared memory accesses made by the calling thread prior to threadfence_block() are visible to all threads in the thread block for all accesses. only if the value of the argument is .0 or higher. for all threads in the warp. ALLTHREADS The allthreads function is a warp-vote operation with a single scalar logical argument: if( allthreads(a(i)<0.

true.0. Example: if( ballot(int_value) ) allneg = .0) ) allneg = . only if the value of the argument is . Both arguments must be of type integer(kind=4). BALLOT The ballot function is a warp-vote operation with a single integer argument: unsigned integer ballot(int_value) The function ballot evaluates the argument int_value for all threads of the warp and returns an integer whose Nth bit is set if and only if int_value evaluates to non-zero for the Nth thread of the warp. The value of the function is . Atomic functions are only supported by devices with compute capability 1.2 or higher is required if the first argument has the shared attribute. Reference ANYTHREAD The anythread function is a warp-vote operation with a single scalar logical argument: if( anythread(a(i)<0.false. These functions are: Table 3.Chapter 3.1 and higher. value ) atomicsub( mem. value ) Additional Atomic Update mem = mem + value mem = mem – value 27 . The atomic functions return correct values even if multiple threads in the same or different thread blocks try to read and update the same location without any synchronization. Arithmetic and Bitwise Atomic Functions These atomic functions read and return the value of the first argument. Atomic Functions The atomic functions read and write the value of their first operand. depending on the function. They also combine that value with the value of the second argument. This function is only supported by devices of compute capability 2. The function anythread evaluates its argument for all threads in the current warp.false. for all threads in the warp. Note The return value for each of these functions is the first argument. Arithmetic and Bitwise Atomic Functions Function atomicadd( mem.true. mem. which must be a variable or array element in shared memory (with the shared attribute) or in device global memory (with the device attribute). and store the combined value back to the first argument location.8. Compute capability 1.

They also compare the first argument with the second argument. Both arguments must be of type integer(kind=4).value) mem = ior(mem. The function is: 28 .value) mem = value Counting Atomic Functions These atomic functions read and return the value of the first argument. depending on the result of the comparison. Note The return value for each of these functions is the first argument.value) mem = iand(mem.value) mem = min(mem. These functions are: Table 3.and. and atomically stores a new value back to the first argument location if the first and second argument are equal. imax ) Additional Atomic Update if (mem<imax) then mem = mem+1 else mem = 0 endif if (mem<imax . value ) atomicxor( mem. These functions are intended to implement circular counters. value ) atomicand( mem. Counting Atomic Functions Function atomicinc( mem. All three arguments must be of type integer(kind=4).value) mem = ieor(mem. Note The return value for this function is the first argument. mem. It also compares the first argument with the second argument. mem>0) then mem = mem-1 else mem = imax endif atomicdec( mem. value ) mem = max(mem. value ) atomicor( mem. and stores a new value back to the first argument location. counting up to or down from a maximum value specified in the second argument.Device code atomicmax( mem. mem.9. value ) atomicexch( mem. value ) atomicmin( mem. imax ) Compare and Swap Atomic Function This atomic function reads and returns the value of the first argument.

3 are not supported. character substrings are not supported. • Subroutine and function calls are supported only if they can be inlined. such as the new sizeof intrinsic function. FORMAT. integer(kind=4) :: i. OPEN. Host code Host subprograms may use intrinsic functions. returns the number of bytes required to hold the value of A. • Recursive subroutines and functions are not allowed. WRITE. j j = sizeof(i) ! this assigns the value 4 to j 29 . PRINT.6. Reference Table 3. Compare and Swap Atomic Function Function atomiccas(mem.Chapter 3. • STOP and PAUSE statements are not allowed. • Objects with character type must have LEN=1. BACKSPACE. CLOSE. where A is a variable or expression. • Automatic arrays must be fixed size. INQUIRE.comp. • Objects with the Pointer and Allocatable attribute are not allowed. ENDFILE. SIZEOF Intrinsic A call to sizeof(A).val) Additional Atomic Update if (mem == comp) then mem = val endif Restrictions This section lists restrictions on statements and features that can appear in device subprograms. • Input/Output statements are not allowed: READ. • Floating point exception handling is not supported. • Cray pointers are not supported. REWIND. • ENTRY statements are not allowed.10. • Fortran intrinsic functions not listed in Section 3. NAMELIST. • Optional arguments are not allowed. • Alternate return specifications are not allowed.

30 .

Most of the runtime API routines are integer functions that return an error code. Runtime APIs The system module cudafor defines the interfaces to the Runtime API routines. be aware that initialization can add some overhead. Tip When doing timing runs. Initialization No explicit initialization is required. refer to “Error Handling. intent(in) :: devnum cudaSetDevice selects the device to associate with this host thread. Device Management Use the functions in this section for device management. 31 . the runtime initializes and connects to the device the first time a runtime routine is called or a device array is allocated. intent(in) :: flags cudaSetDeviceFlags records how the CUDA runtime interacts with this host thread. cudaGetDeviceCount integer function cudaGetDeviceCount( numdev ) integer. To interpret the error codes.Chapter 4. they return a value of zero if the call was successful. and a nonzero value if there was an error. cudaSetDevice integer function cudaSetDevice( devnum ) integer.” on page 41. intent(out) :: numdev cudaGetDeviceCount assigns the number of available devices to its first argument. cudaSetDeviceFlags integer function cudaSetDevice( flags ) integer.

Thread Management cudaGetDevice integer function cudaGetDevice( devnum ) integer. intent(out) :: prop integer. Some can also take C types. intent(out) :: devnum type(cudadeviceprop). cudaChooseDevice integer function cudaChooseDevice ( devnum. To avoid these potential issues. cudaThreadExit integer function cudaThreadExit() cudaThreadExit explicitly cleans up all runtime-related CUDA resources associated with the host thread. defined in the cudafor module. or write-after-write hazards for some of these memory accesses. It may return an error condition if one of the preceding operations fails. Any subsequent CUDA calls or operations will reinitialize the runtime. Memory Management Many of the memory management routines can take device arrays as arguments. intent(in) :: devnum cudaGetDeviceProperties returns the properties of a given device. intent(out) :: devnum cudaGetDevice assigns the device number associated with this host thread to its first argument. as TYPE(C_DEVPTR). devnum ) type(cudadeviceprop). prop ) integer. intent(in) :: prop cudaChooseDevice assigns the device number that best matches the properties given in prop to its first argument. provided through the Fortran 2003 iso_c_binding module. Consistent use of TYPE(C_PTR) and 32 . cudaThreadSynchronize integer function cudaThreadSynchronize() cudaThreadSynchronize blocks execution of the host subprogram until all preceding kernels and operations are complete. as arguments to simplify interfacing to existing CUDA C code. cudaGetDeviceProperties integer function cudaGetDeviceProperties( prop. thus creating potential read-after-write. Thread Management Sometimes threads within a block access the same addresses in shared or global memory. use the functions in this section for thread management. CUDA Fortran has extended the F2003 derived type TYPE(C_PTR) by providing a C device pointer. it is implicitly called when the host thread exits. write-after-read. Calling cudaThreadExit is optional.

in the future. cudaMalloc integer function cudaMalloc(devptr. devptr may be of TYPE(C_DEVPTR). “Device Code Intrinsic Datatypes. The height is an integer. cdesc. Both of these features are subject to change when.” on page 21. Or. cudaFree integer function cudaFree(devptr) cudaFree deallocates data on the device. in which case the count is in bytes. an allocatable device array as the second argument. width. height) type(cudaArrayPtr) :: carray type(cudaChannelFormatDesc) :: cdesc integer :: width. cudaMallocPitch integer function cudaMallocPitch(devptr.1. and in effect transfer the allocation to the Fortran array. one-dimensional device array of a supported type specified in Table 3. “Device Code Intrinsic Datatypes. cudaMallocPitch may pad the data.” on page 21. two-dimensional device array of a supported type specified in Table 3. pitch. The count is in terms of elements. in which case the integer values are expressed in bytes. proper Fortran pointers for device data are supported. height) cudaMallocPitch allocates data on the device. Under CUDA Fortran. it is possible to construct a Fortran device array out of a TYPE(C_DEVPTR) by using an extension of the iso_c_binding subroutine c_f_pointer. should be of benefit. and the padded width is returned in the variable pitch. Runtime APIs TYPE(C_DEVPTR). Similarly. Or. devptr may be any allocatable. cudaFreeArray integer function cudaFreeArray(carray) type(cudaArrayPtr) :: carray 33 . devptr may be any allocatable device array of a supported type specified in Table 3. devptr may be any allocatable. cudaMallocArray integer function cudaMallocArray(carray. as well as consistency checks between Fortran device arrays and host arrays. a shape as the third argument. height cudaMallocArray allocates a data array on the device.Chapter 4.1. “Device Code Intrinsic Datatypes. devptr may also be of TYPE(C_DEVPTR). count) cudaMalloc allocates data on the device. Currently.1.” on page 21. c_f_pointer will take a TYPE(C_DEVPTR) as the first argument. width. The width is in terms of number of elements. devptr may be of TYPE(C_DEVPTR). Use the functions in this section for memory management. there is also a function C_DEVLOC() defined which will create a TYPE(C_DEVPTR) that holds the C address of the Fortran device array argument.

This function operates on page-locked host memory only. in which case the count is in term of bytes. “Device Code Intrinsic Datatypes. “Device Code Intrinsic Datatypes. Or. cudaMemcpy2D integer function cudaMemcpy2D(dst. The copy can be associated with a stream by passing a non-zero stream argument.” on page 21. and the lowest byte of value is used. for more information. devptr may be any device scalar or array of a supported type specified in Table 3. count. The count is in terms of elements.1. stream) cudaMemcpy copies data from one location to another.” on page 21. dst and src may be any device or host. value.1. height) cudaMemset2D sets an array to the specified value. cudaMemset2D integer function cudaMemset2D(devptr. dst and src may be any device or host array. and height are in terms of elements. “Device Code Intrinsic Datatypes. and the lowest byte of value is used. kdir. src. If kdir is specified. kdir) cudaMemcpy copies data from one location to another. width. cudaMemcpyDeviceToHost. scalar or array. The count is in terms of elements.” on page 20. or cudaMemcpyDeviceToDevice. dst and src may be of TYPE(C_DEVPTR) or TYPE(C_PTR). “Device Code Intrinsic Datatypes. If kdir is specified. height. scalar or array. Or. width. kdir may be optional. count. refer to “Data Transfer Using Runtime Routines. it must be one of the defined enums cudaMemcpyHostToDevice. cudaMemset integer function cudaMemset(devptr. in which case the count is in term of bytes. The value must match in type and kind. of a supported type specified in Table 3. The pitch.” on page 21. width. in which case the pitch. devptr may be any device array of a supported type specified in Table 3. count) cudaMemset sets a location or array to the specified value. dpitch. of a supported type specified in Table 3. refer to “Data Transfer Using Runtime Routines. pitch. otherwise the stream argument is optional and defaults to zero. dst and src may be of TYPE(C_DEVPTR) or TYPE(C_PTR). The value must match in type and kind. src. dst and src may be any device or host. Alternatively. value. in which case the count is in term of bytes. kdir may be optional.” on page 20.” on page 21.Memory Management cudaFreeArray frees an array that was allocated on the device. src. The width 34 . it must be one of the defined enums cudaMemcpyHostToDevice. “Device Code Intrinsic Datatypes. devptr may be of TYPE(C_DEVPTR). devptr may be of TYPE(C_DEVPTR). The count is in terms of elements. spitch. width.1. and height are in terms of bytes. of a supported type specified in Table 3.” on page 21.1. cudaMemcpyAsync integer function cudaMemcpyAsync(dst.1. Alternatively. cudaMemcpyDeviceToHost. cudaMemcpy integer function cudaMemcpy(dst. for more information. or cudaMemcpyDeviceToDevice. kdir) cudaMemcpy2D copies data from one location to another.

of a supported type specified in Table 3. src. stream) cudaMemcpy2D copies data from one location to another. spitch. kdir) type(cudaArrayPtr) :: dsta integer :: dstx. count.” on page 20. srca. dstx. cudaMemcpyDeviceToHost. The copy can be associated with a stream by passing a non-zero stream argument. kdir) type(cudaArrayPtr) :: dsta integer :: dstx. srcx. width. refer to “Data Transfer Using Runtime Routines.” on page 21. dsty. This function operates on page-locked host memory only. count. in which case the width and height are in term of bytes. cudaMemcpyDeviceToHost. cudaMemcpy2DAsync integer function cudaMemcpy2DAsync(dst. srcx. dst and src may be of TYPE(C_DEVPTR) or TYPE(C_PTR). kdir cudaMemcpy2DFromArray copies array data to and from the device. srcy. otherwise the stream argument is optional and defaults to zero. cudaMemcpy2DToArray integer function cudaMemcpy2DToArray(dsta. dsty. dst and src may be of TYPE(C_DEVPTR) or TYPE(C_PTR). cudaMemcpyFromArray integer function cudaMemcpyFromArray(dst. “Device Code Intrinsic Datatypes. it must be one of the defined enums cudaMemcpyHostToDevice. count. in which case the width and height are in term of bytes. kdir) type(cudaArrayPtr) :: srca integer :: dstx. height. cudaMemcpy2DFromArray integer function cudaMemcpy2DFromArray(dst. it must be one of the defined enums cudaMemcpyHostToDevice.Chapter 4. height. kdir cudaMemcpy2DToArray copies array data to and from the device. refer to “Data Transfer Using Runtime Routines. Alternatively. srcy. src. If kdir is specified. kdir may be optional. or cudaMemcpyDeviceToDevice. dstx. dpitch. for more information. kdir) type(cudaArrayPtr) :: srca integer :: dpitch. for more information.1. width. Runtime APIs and height are in terms of elements. cudaMemcpyToArray integer function cudaMemcpyToArray(dsta. 35 . height. dsty. count. kdir cudaMemcpyToArray copies array data to and from the device. dpitch. dst and src may be any device or host array. srca. spitch. width. If kdir is specified. height. kdir. dsty. width. dsty. spitch. width. src. kdir may be optional. or cudaMemcpyDeviceToDevice.” on page 20. srcx. height. The width and height are in terms of elements. srcy. kdir cudaMemcpyFromArray copies array data to and from the device. Alternatively.

height. cudaMemcpy3D integer function cudaMemcpy3D(p) type(cudaMemcpy3DParms) :: p cudaMemcpy3D copies elements from one 3D array to another as specified by the data held in the derived type p.Memory Management cudaMemcpyArrayToArray integer function cudaMemcpyArrayToArray(dsta. srcx. srca integer :: dstx.” on page 21. dsty. width. dstx. Alternatively. cudaMemset3D integer function cudaMemset3D(pitchptr. srcy. cext) type(cudaPitchedPtr) :: pitchptr integer :: value type(cudaExtent) :: cext cudaMemset3D sets elements of an array. dsty. dstx. kdir cudaMemcpyArrayToArray copies array data to and from the device. 36 . kdir) type(cudaArrayPtr) :: dsta. dsty. srcx. the extents in each dimension specified by cext. kdir) type(cudaArrayPtr) :: dsta. cext is also a derived type which holds the extents of the allocated array. intent(in) :: cext cudaMalloc3D allocates data on the device. width. cdesc. srcy. cudaMalloc3DArray integer function cudaMalloc3DArray(carray. srcy. srcy. srca integer :: dstx. three-dimensional device array of a supported type specified in “Datatypes allowed. srcx. pitchptr is a derived type defined in the cudafor module. srca. which was allocated with cudaMalloc3D to a specified value. intent(out) :: pitchptr type(cudaExtent). pitchptr may be any allocatable. cext) type(cudaArrayPtr) :: carray type(cudaChannelFormatDesc) :: cdesc type(cudaExtent) :: cext cudaMalloc3DArray allocates array data on the device. cext) type(cudaPitchedPtr). count. height. cudaMalloc3D integer function cudaMalloc3D(pitchptr. srca. dsty. srcx. cudaMemcpy2DArrayToArray integer function cudaMemcpy2DArrayToArray(dsta. count. value. kdir cudaMemcpy2DArrayToArray copies array data to and from the device.

symbol. kdir cudaMemcpyToSymbol copies data from the source to a device area in global or constant memory space referenced by a symbol. offset. src may be any host scalar or array of a supported type as specified in “Datatypes allowed. count. count. dst may be any host scalar or array of a supported type specified in “Datatypes allowed. kdir cudaMemcpyFromSymbol copies data from a device area in global or constant memory space referenced by a symbol to a destination on the host. kdir integer. kdir) type(cudaSymbol) :: symbol integer :: count. kdir. offset. cudaMemcpyFromSymbol integer function cudaMemcpyFromSymbol(dst. Runtime APIs cudaMemcpy3DAsync integer function cudaMemcpy3D(p. This function operates on page-locked host memory only. This function operates on page-locked host memory only.” on page 21. The count is in terms of elements. cudaMemcpyToSymbolAsync integer function cudaMemcpyToSymbolAsync(symbol. offset. count. stream) type(cudaMemcpy3DParms) :: p integer :: stream cudaMemcpy3DAsync copies elements from one 3D array to another as specified by the data held in the derived type p. The count is in terms of elements. kdir) type(cudaSymbol) :: symbol integer :: count. cudaMemcpyFromSymbolAsync integer function cudaMemcpyFromSymbolAsync(dst.” on page 21. offset. optional :: stream cudaMemcpyToSymbol copies data from the source to a device area in global or constant memory space referenced by a symbol. kdir integer. symbol. The count is in terms of elements. src may be any host scalar or array of a supported type specified in “Datatypes allowed. offset. The copy can be associated with a stream by passing a non-zero stream argument. count. stream) type(cudaSymbol) :: symbol integer :: count. stream) type(cudaSymbol) :: symbol integer :: count. The copy can be associated with a stream by passing a non-zero stream argument. optional :: stream 37 .” on page 21. offset.Chapter 4. src. kdir. src. offset. offset. cudaMemcpyToSymbol integer function cudaMemcpyToSymbol(symbol.

hostptr is a pinned memory buffer that was allocated via cudaHostAlloc(). a kernel which runs on 38 . The flags argument enables different options to be specified that affect the allocation. allocatable. Size is in bytes. 100) Vx = 0. The normal iso_c_binding subroutine c_f_pointer can be used to move the type(c_ptr) to a Fortran pointer. cudaGetSymbolAddress integer function cudaGetSymbolAddress(devptr. flags) type(C_PTR) :: hostptr integer :: size. dst may be any host scalar or array of a supported type specified in “Datatypes allowed.0 cudaGetSymbolSize integer function cudaGetSymbolSize(size. symbol) integer :: size type(cudaSymbol) :: symbol cudaGetSymbolSize sets the variable size to the size of a device area in global or constant memory space referenced by the symbol.” on page 21. This function operates on page-locked host memory only. The copy can be associated with a stream by passing a non-zero stream argument. flags) type(C_DEVPTR) :: devptr type(C_PTR) :: hostptr integer :: flags cudaHostGetDevicePointer returns a pointer to a device memory address corresponding to the pinned memory on the host.Memory Management cudaMemcpyFromSymbol copies data from a device area in global or constant memory space referenced by a symbol to a destination on the host. and read and written by. cudaHostGetDevicePointer integer function cudaHostGetDevicePointer(devptr. A symbol can be set to an external device name via a character string. symbol) type(C_DEVPTR) :: devptr type(cudaSymbol) :: symbol cudaGetSymbolAddress returns in the devptr argument the address of symbol on the device. It returns in devptr an address that can be passed to. device :: vx(:) csvx = “vx” Istat = cudaGetSymbolAddress(cdvx. hostptr. csvx) Call c_f_pointer(cdvx. vx. or returns an error if the memory is unavailable. flags cudaHostAlloc allocates pinned memory on the host. size. The following code sequence initializes a global device array “vx” from a CUDA C kernel: type(cudaSymbol) :: csvx type(c_devptr) :: cdvx real. cudaHostAlloc integer function cudaHostAlloc(hostptr. The count is in terms of elements. It returns in hostptr the address of the pagelocked allocation.

or returns an error if the memory is unavailable. The normal iso_c_binding subroutine c_f_pointer can be used to move the type(c_devptr)to a device array. and the value cudaErrorNotReady if not.Chapter 4. Runtime APIs the device. It may also return another error condition if some asynchronous operations failed. cudaStreamSynchronize integer function cudaStreamSynchronize( stream ) integer. intent(in) :: stream cudaStreamSynchronize blocks execution of the host subprogram until all preceding kernels and operations associated with the given stream are complete. It returns in hostptr the address of the page-locked allocation. cudaMallocHost integer function cudaMallocHost(hostptr. intent(in) :: stream cudaStreamDestroy releases any resources associated with the given stream. 39 . it returns zero (success) if all operations are complete. intent(in) :: stream cudaStreamQuery tests whether all operations enqueued to the selected stream are complete. The flags argument is provided for future releases. cudaStreamQuery integer function cudaStreamQuery( stream ) integer. asynchronous operations. size) type(C_PTR) :: hostptr integer :: size cudaMallocHost allocates pinned memory on the host. size is in bytes. It may return error codes from previous. intent(out) :: stream cudaStreamCreate creates an asynchronous stream and assigns its identifier to its first argument. cudaFreeHost integer function cudaFreeHost(hostptr) type(C_PTR) :: hostptr cudaFreeHost deallocates pinned memory on the host allocated with cudaMalloHost. Stream Management Use the functions in this section for stream management. The normal iso_c_binding subroutine c_f_pointer can be used to move the type(c_ptr) to a Fortran pointer. cudaStreamDestroy integer function cudaStreamDestroy( stream ) integer. cudaStreamCreate integer function cudaStreamCreate( stream ) integer.

cudaEventElapsedTime integer function cudaEventElapsedTime( time. end cudaEventElapsedTime computes the elapsed time between two events (in milliseconds). 40 . The event is recorded after all preceding operations in the stream are complete. It returns success (zero) if the event has been recorded. start. intent(in) :: event cudaEventSynchronize blocks until the event has been recorded. and cudaErrorNotReady if it has not. cudaEventQuery integer function cudaEventQuery( event ) type(cudaEvent). intent(in) :: stream cudaEventRecord issues an operation to the given stream to record an event. intent(in) :: event cudaEventQuery tests whether an event has been recorded. intent() :: start. stream ) type(cudaEvent). This function is only valid with events recorded on stream zero. It returns a value of cudaErrorInvalidValue if cudaEventRecord has not been called for this event. cudaEventSynchronize integer function cudaEventSynchronize( event ) type(cudaEvent). the event is recorded after all preceding operations in all streams are complete. If stream is zero. cudaEventDestroy integer function cudaEventDestroy( event ) type(cudaEvent). It returns cudaErrorInvalidValue if cudaEventRecord has not been called for this event. It returns cudaErrorInvalidValue if either event has not yet been recorded. cudaEventCreate integer function cudaEventCreate( event ) type(cudaEvent). intent(in) :: event cudaEventDestroy destroys the resources associated with an event object. end) float :: time type(cudaEvent).Event Management Event Management Use the functions in this section to manage events. intent(in) :: event integer. intent(out) :: event cudaEventCreate creates an event object and assigns the event identifier to its first argument cudaEventRecord integer function cudaEventRecord( event.

This function automatically returns cudaErrorInvalidValue if the iversion argument is NULL. cudaRuntimeGetVersion integer function cudaRuntimeGetVersion(iversion) integer :: iversion cudaRuntimeGetVersion returns the version number of the installed CUDA Runtime as iversion.Chapter 4. cudaGetErrorString function cudaGetErrorString( errcode ) integer. 41 . Version Management Use the functions in this section for version management. This function automatically returns cudaErrorInvalidValue if the iversion argument is NULL. cudaDriverGetVersion integer function cudaDriverGetVersion(iversion) integer :: iversion cudaDriverGetVersion returns the version number of the installed CUDA driver as iversion. then it returns 0 as iversion. cudaGetLastError integer function cudaGetLastError() cudaGetLastError returns the error code that was most recently returned from any runtime call in this host thread. Runtime APIs Error Handling Use the functions in this section for error handling. If no driver is installed. intent(in) :: errcode character*(*) :: cudaGetErrorString cudaGetErrorString returns the message string associated with the given error code.

42 .

The program computes the products by accumulating submatrix products. the program assumes the matrix sizes are a multiple of 16. Each element of the result is the product of one row of A by one column of B. j.j) being computed 43 . and has not been highly optimized for execution time. ty ! submatrices stored in shared memory real. The program caches the submatrices of A and B in the fast shared memory. Bsub(16.16) ! the value of C(i.M). then moves to the next submatrix of A rowwise and of B columnwise. M. B. N. visit the frequently asked questions (FAQ) section of the pgroup. tx.htm Matrix Multiplication Example This example shows a program to compute the product C of two matrices A and B.pgroup.16). • Each thread within the block computes one element of the submatrix.Chapter 5. B is MxL. L integer :: i. B(M. Source Code Listing ! start the module containing the matmul kernel module mmul_mod use cudafor contains ! mmul_kernel computes A*B into C where ! A is NxM. Examples This chapter contains For up-to-date information about the state of the current release. accumulates the submatrix product. C.L). shared :: Asub(16. M. kb. L ) real :: A(N. For simplicity. The submatrix size is chosen so the number of threads in a block is a multiple of the warp size (32) and is less than the maximum number of threads per thread block (512). C is then NxL attributes(global) subroutine mmul_kernel( A.com/support/index. k. C(N.L) integer. value :: N. as follows: • Each thread block computes one 16x16 submatrix of C.com web page at: www. it reads a block submatrix of A and a block submatrix of B.

B.j) ! Wait until all elements are filled call syncthreads() ! Multiply the two submatrices ! Each of the 16x16 threads accumulates the ! dot product for its element of C(i.Bdev. 16. Cdev(N. dimension(:.ty) = B(kb+tx-1. L/16. B. M. device.:) = B(1:M. 16 ! Fill the submatrices ! Each of the 16x16 threads in the thread block ! loads one element of Asub and Bsub Asub(tx. dimBlock ! Get the array sizes N = size( A. Bdev(M.dimBlock>>>( Adev. C ) real.1:M) Bdev(:.:) * B(:.ty) = A(i.1:L) ! Create the grid and block dimensions dimGrid = dim3( N/16. the block size do kb = 1. Bdev.:) :: A. 1 ) M = size( A.:) :: Adev.L) ) ! Copy A and B to the device Adev = A(1:N. 2 ) L = size( B. Bdev. dimension(:.k) * Bsub(k.0 ! Do the k loop in chunks of 16. 1 ) call mmul_kernel<<<dimGrid. C ! allocatable device arrays real. 2 ) ! Allocate the device arrays allocate( Adev(N. allocatable.1:L) = Cdev deallocate( Adev.ty) enddo ! Synchronize to make sure all threads are done ! reading the submatrices before overwriting them ! in the next iteration of the kb loop call syncthreads() enddo ! Each of the 16x16 threads stores its element ! to the global C array C(i.kb+ty-1) Bsub(tx.M). Cdev ) end subroutine mmul end module mmul_mod N 44 . Cdev. ! Copy the results back and free up memory C(1:N.j) do k = 1.Matrix Multiplication Example real :: Cij ! Get the thread indices tx = threadidx%x ty = threadidx%y ! This thread computes C(i.16 Cij = Cij + Asub(tx.j) = Cij end subroutine mmul_kernel ! The host routine to drive the matrix multiplication subroutine mmul( A.Cdev ! dim3 variables to define the grid and block shapes type(dim3) :: dimGrid.L).j)) i = (blockidx%x-1) * 16 + tx j = (blockidx%y-1) * 16 + ty Cij = 0. 1 ) dimBlock = dim3( 16.j) = sum(A(i.

• It initializes a scalar in which it will accumulate the dot product. depending on the block and thread index. • It frees the device memory arrays. C. one device memory output array. • It synchronizes again to make sure all threads are done reading the submatrices before starting the next block. M. it does the following steps: • It loads one element of the submatrices of A and B into shared memory. • It allocates device memory arrays Adev. A and B. and three scalars giving the array sizes. and L. no extra synchronization is needed between the copy operations and the kernel launch. and Cdev. • It calls mmul_kernel to compute Cdev on the device. The host subroutine mmul is a wrapper for the kernel routine mmul_kernel. C. passed as assumed-shape arrays. • Finally. • It determines the i and j indices. It performs the following operations: • It determines the thread indices for this thread. • It accumulates the dot product of its row and column of the submatrices. Examples Source Code Description This source code module mmul_mod has two subroutines. • For each block. • It copies Cdev back from device memory to C. The routine performs the following operations: • It determines the size of the matrices in N. Because the data copy operations are synchronous. MMUL This host subroutine has two input arrays. This routine computes the dot product of A(i.j) it is computing. The thread executing this routine is one of 16x16 threads cooperating in a thread block. • It copies the arrays A and B to Adev and Bdev using array assignments.j) for a particular value of i and j. MMUL_KERNEL This kernel subroutine has two device memory input arrays. A and B. • It synchronizes to make sure both submatrices are loaded by all threads in the block.:)*B(:.Chapter 5. • It steps through the arrays A and B in blocks of size 16. Bdev. • It fills dimGrid and dimBlock to hold the grid and thread block sizes. 45 . for which element of C(i. it stores the computed value into the correct element of C. and one output array.

46 .

pgroup.htm.com All technical support is by email or submissions using an online form at www. Many questions and problems can be resolved at our frequently asked questions (FAQ) site at www.Chapter 6. Log in to the PGI website to access the forum: www.pgroup.pgroup.com/support/faq. Two Centerpointe Drive Lake Oswego. OR 97035 USA The PGI User Forum is monitored by members of the PGI engineering and support teams as well as other PGI customers.com trs@pgroup. PGI documentation is available at www.pgroup. Inc.php Or contact us electronically using any of the following means: Fax Sales Support WWW +1-503-682-2637 sales@pgroup.com/resources/docs.pgroup.com www. Phone support is not currently available.htm or in your local copy of the documentation in the release directory doc/index. Contact Information You can contact The Portland Group at: The Portland Group STMicroelectronics.com/userforum/index. 47 . The forum newsgroups may contain answers to commonly asked questions.com/support.htm.

48 .

Sign up to vote on this title
UsefulNot useful