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

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

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

viii .

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

NVIDIA. • NVIDIA CUDA Programming Guide. 7/21/2010.pgroup. Version 3.pdf. Available online at http:// www.nvidia. 2009. • PGI Users Guide.Terminology Constant Width is used for filenames.htm Related Publications The following documents contain additional information related to CUDA Fortran programming. and for language statements in the text.com/cuda. 1997 (Fortran 95). There are a wide variety of releases and distributions of each of these types of operating systems.com/support/definitions. NVIDIA.nvidia. • ISO/IEC 1539-1:1997.1. PGI provides a glossary of terms which you can access at www. Available online at http:// www.1. C/C++ C/C++ language statements are shown in the test of this guide using a reduced fixed point size.com/doc/pgiug. x . September.com/cuda. { item2 | item 3} braces indicate that a selection is required. The Portland Group.1. In this case. options. In the context of p/tsets. Zero or more of the preceding item may occur. you must select either item2 or item3. Information Technology – Programming Languages – FORTRAN.. Release 10. In this case item1 is optional. In this example. Bold is used for commands. Geneva. 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.9. ellipsis indicate a repetition. Terminology If there are terms in this guide with which you are unfamiliar. • NVIDIA CUDA Compute Unified Device Architecture Reference Manual. [ item1 ] in general. June 2010. including assembly language statements. multiple filenames are allowed. directories. Version 3.pgroup. arguments. FORTRAN Fortran language statements are shown in the text of this guide using a reduced fixed point size. square brackets are required to specify a p/t-set. square brackets indicate optional items.. Available online at http://www. examples. filename .

highly parallel computational units with very high memory bandwidth. A CUDA programmer must partition the program into coarse grain blocks that can be executed in parallel. The programming model supports four key abstractions: cooperating threads organized into thread groups. Each block is partitioned into fine grain threads. 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 . but are general enough to be useful in many data-parallel. a general purpose parallel programming architecture. Mac OS X and Windows. allowing direct programming of the GPU from a high level language. CUDA Fortran is an analog to NVIDIA's CUDA C compiler. with compilers and libraries to support the programming of NVIDIA GPUs. Introduction Welcome to Release 2011 of PGI CUDA Fortran. Compared to the PGI Accelerator directives-based model and compilers. here called CUDA C. regardless of the number of available processor cores. NVIDIA introduced CUDA™. 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. and coordinated independent thread groups organized into a grid. PGI 2011 includes support for CUDA Fortran on Linux. and tremendous potential for many applications. which can cooperate using shared memory and barrier synchronization. A properly designed CUDA program will run on any CUDA-enabled GPU. a small set of extensions to Fortran that supports and is built upon the CUDA computing architecture. 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. shared memory and barrier synchronization within thread groups. Graphic processing units or GPUs have evolved into programmable. CUDA comes with an extended C compiler.Chapter 1. CUDA Fortran includes a Fortran 2003 compiler and tool chain for programming NVIDIA GPUs using Fortran. compute-intensive programs. GPU designs are optimized for the computations found in graphics rendering.

2 .

each with 64 threads. y real :: a integer :: n ! call the kernel call ksaxpy<<<n/64. each instance will be executed by a different CUDA thread. the call to the kernel ksaxpy specifies n/64 thread blocks. each thread performs one iteration of the common SAXPY loop operation. A call to a kernel specifies how many parallel instances of the kernel must be executed. The CUDA threads are organized into thread blocks. Each thread is assigned a thread block index accessed through the built-in blockidx variable. a.” on page 13. Such a subroutine is called a device kernel or kernel. y ) end subroutine In this case.y real. Programming Guide This chapter introduces the CUDA programming model through examples written in CUDA Fortran. and a local thread index within its thread block. device. A kernel is defined using the attributes(global) specifier on the subroutine statement. x. value :: a integer. value :: n. refer to Chapter 3. a. 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. dimension(*) :: x.Chapter 2. y ) real. and a thread index accessed through threadidx. 3 . 64>>>( n. 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. “Reference. x. For a reference for CUDA Fortran. y ) real. 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. a. x. and each thread has a global thread block index. dimension(*) :: x. In this example.

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

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

8 4. Other device variables have a lifetime of the entire application.Datatypes in Device Subprograms • If declared in a host subprogram. A shared variable may not be data initialized. 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. A pinned variable must be an allocatable array. from when it is allocated until it is deallocated. It can be read or written by all threads in the block. A device array may be an explicit-shape array. When a pinned variable is allocated. or an assumed-shape dummy array. and has a lifetime of the thread block. if the allocation in page-locked memory fails. it will be allocated in host pagelocked memory. the variable may be accessed by any subprogram in that module and by any subprogram that uses the module. but may be modified in host subprograms. An allocatable device variable has a dynamic lifetime. Attributes(pinned) A variable with the pinned attributes is called a pinned variable. Intrinsic Datatypes Type integer logical real double precision Type Kind 1.8 equivalent to real(kind=8) 6 . or size of page-locked memory.4. and have a lifetime of the entire application. the variable will be allocated in the normal host paged memory. an allocatable array. Device constant variables are allocated in the device constant memory space. or by other device subprograms to which it is passed as an argument. Device constant variables may not be allocatable.4. and may only be accessed within that subprogram. Attributes(shared) A variable with the shared attributed is called a device shared variable or a shared variable. Datatypes in Device Subprograms The following intrinsic datatypes are allowed in device subprograms and device data: Table 2. A shared variable is allocated in the device shared memory for a thread block. Attributes(constant) A variable with the constant attributes is called a device constant variable. though a write in one thread is only guaranteed to be visible to other threads after the next call to the SYNCTHREADS() intrinsic.2.1. availability. Device constant data may not be assigned or modified in any device subprogram.2.8 1. the variable may be accessed by that subprogram or subprograms contained in that subprogram. Some operating systems or installations may restrict the use. If declared in a module. A shared variable may only be declared in a device subprogram.

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

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

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

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

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

12 .

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

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

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

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

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

Allocating Pinned Memory Allocatable arrays with the pinned attribute are dynamically allocated using the Allocate statement. device :: adev(n) real. Mixing standard Fortran allocate/deallocate with the runtime Malloc/Free for a given array is not supported. or if it is exhausted. Otherwise. device :: b(:) allocate(b(5024). declare and initialize the scalar on the host as: integer. allocatable. if available. pinned allocatable arrays work and act like any other allocatable array on the host. allocatable. The compiler will generate code to allocate the array in host page-locked memory.” on page 32.Allocating Device and Pinned Arrays real.n) real. and cudaFree can be used to free it: real. pinned :: p(:) allocate(p(5000). If no such memory space is available.stat=istat) … if(allocated(b)) deallocate(b) Scalar variables can be allocated on the device using the Fortran 2003 allocatable scalar feature.c. 100) … istat = cudaFree(v) For a complete list of the memory management runtime routines. 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. device :: v(:) istat = cudaMalloc(v. and thus the arrays are treated more like C malloc’ed areas. real. The cudaMalloc function can be used to allocate single-dimensional arrays of the supported intrinsic datatypes. Fortran interfaces to the CUDA memory management runtime routines are provided.stat=istat) … if(allocated(p)) deallocate(p) 18 . To use these. 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. refer to“Memory Management. allocatable. allocatable. These arrays will also have a lifetime of the subprogram as is usual with the Fortran language: subroutine vfunc(a. the compiler allocates the array in normal paged host memory. These functions return memory which will bypass certain Fortran allocatable properties such as automatic deallocation.

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

stream >>> • grid is an integer. Invoking a kernel subroutine A call to a kernel subroutine must give the execution configuration for the call. as well as the amount of shared memory to use for assumed-size shared memory arrays.Invoking a kernel subroutine The following expressions are not legal as they either promote a false impression of where the actual computation occurs. 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. Array slices are also supported. or from one device array to another. which must be less than the maximum supported by the device. For a complete list of memory management runtime routines. or of type(dim3). Counts expressed in arguments to the Fortran runtime routines are expressed in terms of data type elements. refer to “Memory Management. not bytes. These functions can transfer data either from the host to device. the number of threads per thread block is block%x*block%y*block%z. or would be more efficient written in another way. block.1).” on page 32. Data Transfer Using Runtime Routines For programmers comfortable with the CUDA C programming environment. and the associated stream. device to host. cur. the value of grid%z must be one. The execution configuration gives the size and shape of the grid and thread blocks that execute the function. it is converted to dim3(grid. The execution configuration is specified after the subroutine name in the call statement. it has the form: <<< grid. device :: wrk(1024) real cur(512) istat = cudaMemcpy(wrk. If grid is an integer. If it is type(dim3). 512) For those familiar with the CUDA C routines. the kind parameter to the Memcpy routines is optional in Fortran since the attributes of the arrays are explicitly declared.1. The product grid%x*grid%y gives the number of thread blocks to launch.1). it is converted to dim3(block. bytes. If it is type(dim3). and the implementation. 20 . or of type(dim3).1. bl • block is an integer. The cudaMemcpy function can be used to copy data between the host and the GPU: real. If block is an integer. amount of contiguous data in the slices. and their performance is dependent on the size of the slice.

8 equivalent to real(kind=8) 4(default).8 4(default). the value zero is used. allowing for static shared memory declared in the kernel and for other dedicated uses.8 1 (default) Additionally.1.Chapter 3. DB >>> ( A ) The function call will fail if the grid or block arguments are greater than the maximum sizes allowed.5 on page 41. or another allowed derived type. Device Code Intrinsic Datatypes Type integer logical real double precision complex character(len=1) Type Kind 1.z end type 21 . if present. and have a value of zero. Table 3.” on page 16. where the members of the derived type have one of the allowed intrinsic datatypes. constant. It specifies the stream to which this call is enqueued. or shared attributes. See Section 4. Device code Datatypes allowed Variables and arrays with the device. such as the function arguments and execution configuration arguments. Reference • bytes is optional.4(default). They may have any of the intrinsic datatypes in the following table. refer to “Shared data. a kernel subroutine attributes(global) subroutine sub( a ) can be called like: call sub <<< DG. if present. For instance. they may be of derived type.y. are limited to the types described in this section.8 1. or a value returned by a call to cudaStreamCreate. it must be a scalar integer.4(default). it must be an integer. The system module cudafor includes definitions of the derived type dim3. or if bytes is greater than the shared memory available.2. or declared in device subprograms. For more information. If not specified. defined as type(dim3) integer(kind=4) :: x. • stream is optional. and specifies the number of bytes of shared memory to be allocated for each thread block to use for assumed-size shared memory arrays.2.

for one. the value of griddim%z is always one. real. The value of blockidx%z is always one. as with threadidx. real integer. real integer. • The variable blockidx contains the block index within the grid. blockdim. Fortran Numeric and Logical Intrinsics Table 3. complex logical integer. These variables are read-only.real) complex integer. and griddim are available only in device subprograms. the value of griddim%z is always one.or two-dimensional thread blocks. The value of griddim is the same for all threads in the same grid. real real Name int logical max min mod modulo nint real sign Argument Datatypes integer. Fortran Numeric and Logical Intrinsics Name abs aimag aint anint ceiling cmplx conjg dim floor Argument Datatypes integer. griddim integer(4) :: warpsize • The variable threadidx contains the thread index within its thread block. Fortran intrinsics This section lists the Fortran intrinsic functions allowed in device subprograms. for onedimensional grids. • The variables threadidx. blockidx%y has the value one. • The variable blockdim contains the dimensions of the thread block. real. real integer.or two-dimensional thread blocks. complex complex real real real real or (real. blockdim. real 22 . • The variable warpsize contains the number of threads in a warp. the threadidx%y and/or threadidx%z components have the value one. It has constant value. the blockdim%y and/or blockdim %z components have the value one. blockidx. complex integer. real. currently defined to be 32.2. • The variable griddim contains the dimensions of the grid. real real integer. for one. the value of griddim%y is one for one-dimensional grids. blockidx. They are declared as follows: type(dim3) :: threadidx.Device code Builtin variables The system module cudafor declares several predefined variables. blockdim has the same value for all threads in the same grid. The value of blockidx is the same for all threads in the same thread block.

5.Chapter 3. real integer. complex integer. Reference Fortran Mathematical Intrinsics Table 3. complex real real. complex real real Fortran Numeric Inquiry Intrinsics Table 3. real real integer. complex real real.integer) real Fortran Bit Manipulation Intrinsics Table 3. complex real real. complex Name log log10 sin sinh sqrt tan tanh Argument Datatypes real. Fortran Numeric Inquiry Intrinsics Name bit_size digits epsilon huge maxexponent minexponent Argument Datatypes integer integer.real) real. complex integer (integer. Fortran Mathematical Intrinsics Name acos asin atan atan2 cos cosh exp Argument Datatypes real real real (real.3. real real real Name precision radix range selected_int_kind selected_real_kind tiny Argument Datatypes real. real.4. 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 .

refer to “Thread Management. SYNCTHREADS The syncthreads intrinsic subroutine acts as a barrier synchronization for all threads in a single thread block. complex Fortran Random Number Intrinsics Table 3. it has no arguments: void syncthreads() Sometimes threads within a block access the same addresses in shared or global memory. real integer. real integer. Fortran Reduction Intrinsics Name all any count maxloc maxval Argument Datatypes logical logical logical integer. syncthreads syncthreads_count syncthreads_and syncthread_or threadfence threadfence_block threadfence_system For detailed information on these functions.7.Device code Fortran Reduction Intrinsics Table 3. Synchronization Functions The synchronization functions control the synchronization of various threads during execution of thread blocks. thus creating potential read-after-write. This 24 . complex integer. . or write-after-write hazards for some of these memory accesses.” on page 32.6. real. real integer. real. write-after-read. To avoid these potential issues. use syncthreads()to specify synchronization points in the kernel. 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. real Name minloc minval product sum Argument Datatypes integer.

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

The function allthreads evaluates its argument for all threads in the current warp.true. Each of these functions has a single argument.” on page 32.0 or higher. ALLTHREADS The allthreads function is a warp-vote operation with a single scalar logical argument: if( allthreads(a(i)<0.0) ) allneg = . 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. GPU_TIME The gpu_time intrinsic returns the value of the clock cycle counter on the GPU. THREADFENCE_SYSTEM void threadfence_system() threadfence_system acts as a memory fence. only if the value of the argument is .true. 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. 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. 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.true. The value of the function is . for all threads in the warp.Device code threadfence() is one method to enforce a specific order. as described in “cudaGetDeviceProperties. The clock frequency can be determined by calling cudaGetDeviceProperties. 26 .2 and higher. Warp-Vote Operations Warp-vote operations are only supported by devices with compute capability 1.

depending on the function. Both arguments must be of type integer(kind=4). mem.1 and higher. The function anythread evaluates its argument for all threads in the current warp. 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.2 or higher is required if the first argument has the shared attribute. value ) atomicsub( mem. only if the value of the argument is . Compute capability 1.false.true.false. This function is only supported by devices of compute capability 2.0) ) allneg = . and store the combined value back to the first argument location.8. which must be a variable or array element in shared memory (with the shared attribute) or in device global memory (with the device attribute). for all threads in the warp. They also combine that value with the value of the second argument. The value of the function is . Atomic functions are only supported by devices with compute capability 1.Chapter 3.0. Example: if( ballot(int_value) ) allneg = . Note The return value for each of these functions is the first argument. Arithmetic and Bitwise Atomic Functions Function atomicadd( mem. These functions are: Table 3. Atomic Functions The atomic functions read and write the value of their first operand. value ) Additional Atomic Update mem = mem + value mem = mem – value 27 . Reference ANYTHREAD The anythread function is a warp-vote operation with a single scalar logical argument: if( anythread(a(i)<0. 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.true.

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

OPEN. • Optional arguments are not allowed. FORMAT. SIZEOF Intrinsic A call to sizeof(A). Reference Table 3. • ENTRY statements are not allowed. • Objects with the Pointer and Allocatable attribute are not allowed. • Cray pointers are not supported. Host code Host subprograms may use intrinsic functions. • STOP and PAUSE statements are not allowed. where A is a variable or expression. • Input/Output statements are not allowed: READ. INQUIRE.6. • Fortran intrinsic functions not listed in Section 3.Chapter 3. NAMELIST.10.3 are not supported. such as the new sizeof intrinsic function. j j = sizeof(i) ! this assigns the value 4 to j 29 . • Floating point exception handling is not supported. CLOSE. ENDFILE. REWIND. • Recursive subroutines and functions are not allowed. WRITE.comp. returns the number of bytes required to hold the value of A. character substrings are not supported. • Subroutine and function calls are supported only if they can be inlined.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. • Automatic arrays must be fixed size. Compare and Swap Atomic Function Function atomiccas(mem. PRINT. • Alternate return specifications are not allowed. integer(kind=4) :: i. • Objects with character type must have LEN=1. BACKSPACE.

30 .

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

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

1. Similarly. cudaMallocPitch may pad the data.” on page 21. Or. one-dimensional device array of a supported type specified in Table 3. Currently.” on page 21. devptr may be any allocatable device array of a supported type specified in Table 3. devptr may also be of TYPE(C_DEVPTR).1. The width is in terms of number of elements. devptr may be of TYPE(C_DEVPTR). cudaFreeArray integer function cudaFreeArray(carray) type(cudaArrayPtr) :: carray 33 . cudaMalloc integer function cudaMalloc(devptr. devptr may be any allocatable.1. pitch. Use the functions in this section for memory management. in the future. devptr may be of TYPE(C_DEVPTR). in which case the integer values are expressed in bytes. as well as consistency checks between Fortran device arrays and host arrays. 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. width. The height is an integer. width. Or. a shape as the third argument. count) cudaMalloc allocates data on the device. cudaMallocPitch integer function cudaMallocPitch(devptr. an allocatable device array as the second argument. Both of these features are subject to change when. cudaFree integer function cudaFree(devptr) cudaFree deallocates data on the device. should be of benefit. “Device Code Intrinsic Datatypes. Under CUDA Fortran. in which case the count is in bytes. proper Fortran pointers for device data are supported. “Device Code Intrinsic Datatypes. cudaMallocArray integer function cudaMallocArray(carray. Runtime APIs TYPE(C_DEVPTR).” on page 21. two-dimensional device array of a supported type specified in Table 3. c_f_pointer will take a TYPE(C_DEVPTR) as the first argument. height) cudaMallocPitch allocates data on the device. The count is in terms of elements. “Device Code Intrinsic Datatypes. and in effect transfer the allocation to the Fortran array. 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. height) type(cudaArrayPtr) :: carray type(cudaChannelFormatDesc) :: cdesc integer :: width. cdesc.Chapter 4. height cudaMallocArray allocates a data array on the device. devptr may be any allocatable. and the padded width is returned in the variable pitch.

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

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

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

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

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

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

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

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

42 .

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

1:L) ! Create the grid and block dimensions dimGrid = dim3( N/16.j)) i = (blockidx%x-1) * 16 + tx j = (blockidx%y-1) * 16 + ty Cij = 0.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. 1 ) call mmul_kernel<<<dimGrid. C ) real. dimension(:.:) :: Adev. device.j) = Cij end subroutine mmul_kernel ! The host routine to drive the matrix multiplication subroutine mmul( A. Bdev. Cdev ) end subroutine mmul end module mmul_mod N 44 .1:M) Bdev(:. 2 ) ! Allocate the device arrays allocate( Adev(N. 2 ) L = size( B.Bdev.dimBlock>>>( Adev.ty) = A(i. 16. Bdev. ! Copy the results back and free up memory C(1:N. M. 1 ) M = size( A.kb+ty-1) Bsub(tx.0 ! Do the k loop in chunks of 16.:) :: A. 1 ) dimBlock = dim3( 16.j) do k = 1. 16 ! Fill the submatrices ! Each of the 16x16 threads in the thread block ! loads one element of Asub and Bsub Asub(tx. C ! allocatable device arrays real.k) * Bsub(k.L). Bdev(M.16 Cij = Cij + Asub(tx. the block size do kb = 1.ty) = B(kb+tx-1.j) = sum(A(i.:) * B(:.L) ) ! Copy A and B to the device Adev = A(1:N. L/16. dimBlock ! Get the array sizes N = size( A.Matrix Multiplication Example real :: Cij ! Get the thread indices tx = threadidx%x ty = threadidx%y ! This thread computes C(i. Cdev.:) = B(1:M.1:L) = Cdev deallocate( Adev. B. dimension(:.Cdev ! dim3 variables to define the grid and block shapes type(dim3) :: dimGrid.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. B.M). allocatable. Cdev(N.

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

46 .

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

48 .

Sign up to vote on this title
UsefulNot useful