Welcome to Scribd, the world's digital library. Read, publish, and share books and documents. See more
Download
Standard view
Full view
of .
Look up keyword or section
Like this
3Activity

Table Of Contents

0 of .
Results for:
No results containing your search query
P. 1
CUDA C Programming Guide

CUDA C Programming Guide

Ratings: (0)|Views: 332 |Likes:
Published by Rifat Aras

More info:

Published by: Rifat Aras on Nov 02, 2010
Copyright:Attribution Non-commercial

Availability:

Read on Scribd mobile: iPhone, iPad and Android.
download as PDF, TXT or read online from Scribd
See more
See less

11/05/2011

pdf

text

original

 
 
 Version 3.2
9/8/2010
NVIDIA CUDA 
 
NVIDIA CUDA CProgramming Guide
 
 
ii
CUDA C Programming Guide Version 3.2
 
Changes from Version 3.1.1
 
Simplified all the code samples that use
cuParamSetv()
to set a kernelparameter of type
CUdeviceptr
since
CUdeviceptr
is now of same size andalignment as
 void*
, so there is no longer any need to go through aninterneditate
 void*
variable.
 
 Added Section3.2.4.4on 16-bit floating-point textures.
 
 Added Section3.2.4.5on texture cache coherency.
 
 Added more details about surface memory access to Section3.2.5. 
 
Mentioned new stream synchronization function
cudaStreamSynchronize()
 in Section3.2.7.5.2. 
 
Mentioned in Sections3.2.8.2,3.3.11.2,and4.3the new API calls to deal with devices using NVIDIA SLI in AFR mode.
 
 Added Sections3.2.10and3.3.13about the call stack.
 
Changed the type of the
 pitch
variable in the second code sample of Section
3.3.4from
unsigned int
to
size_t
following the functionsignature change of 
cuMemAllocPitch()
.
 
Changed the type of the
 bytes
variable in the last code sample of Section3.3.4from
unsigned int
to
size_t
following the function signature change of 
cuModuleGetGlobal()
.
 
Removed
cuParamSetTexRef()
from Section3.3.7as it is no longernecessary.
 
Updated Section5.2.3,Table 5-1,and SectionG.4.1for devices of compute capability 2.1.
 
 Added GeForce GTX 460, GeForce GTX 465, Quadro 4000, Quadro 5000,Quadro 5000M, and Quadro 6000 toTable A-1. 
 
Fixed sample code in SectionB.2.3: 
array[]
was declared as an array of 
char
 
causing a compiler error (“
Unaligne
d memory accesses not supported”) when
casting 
array
to a pointer of higher alignment requirement; declaring 
array[]
as an array of 
float
fixes it.
 
 Added SectionB.15on the new 
 malloc()
and
free()
device functions.
 
Moved the type casting functions to a separate sectionC.2.4. 
 
Fixed the maximum height of a 2D texture reference for devices of computecapability 2.x (65535 instead of 65536) in SectionG.1. 
 
Fixed the maximum dimensions for surface references in SectionG.1. 
 
Mentioned the new 
cudaThreadSetCacheConfig()
/
cuCtxSetCacheConfig()
API calls inSectionG.4.1. 
 
Mentioned in SectionG.4.2that global memory accesses that are cached in L2only are serviced with 32-byte memory transactions.
 
 
CUDA C Programming Guide Version 3.2
 
iii
 
Table of Contents

Activity (3)

You've already reviewed this. Edit your review.
1 thousand reads
1 hundred reads
mr.girishk liked this

You're Reading a Free Preview

Download
scribd
/*********** DO NOT ALTER ANYTHING BELOW THIS LINE ! ************/ var s_code=s.t();if(s_code)document.write(s_code)//-->