cudart_static - when is it necessary?


cudart_static - when is it necessary?



Since newer drivers ship with the CUDA runtime (I can choose 9.1 or 9.2 in the drivers download page) my question is: should my library (which uses a CUDA kernel internally) be shipped with -lcudart_static?


-lcudart_static



I had issues launching kernels compiled with 9.2 on systems which used 9.1 CUDA drivers. What's the most 'compatible' way of ensuring my library will run everywhere a recent CUDA driver is installed? (I'm already compiling for a virtual architecture)




1 Answer
1



Since newer drivers ship with the CUDA runtime (I can choose 9.1 or 9.2 in the drivers download page)



No, that's incorrect. That choice in the drivers download page is related to the fact that each CUDA version has a minimum required driver version associated with it. It does not mean that the driver ships with the CUDA runtime (stated another way, the driver does not install libcudart.so on linux and never has - with some careful experimentation on a clean install, you can prove this to yourself.)



Some additional comments:



-lcudart_static is actually the default for current/recent versions of nvcc. You can discover this by reading the nvcc manual. Therefore, by default, your executable, when compiled/built with nvcc should already be statically linked to the CUDA runtime library corresponding to the version of nvcc that you are using for compilation. The reason you might need to specify this or something like this is if you are building an application with e.g. the gnu toolchain (on linux) rather than nvcc.


-lcudart_static


nvcc


nvcc


nvcc


nvcc



The purpose of static linking to the CUDA runtime library is, as you surmise, so that an application can be built in such a way that it does not need an installation of the CUDA toolkit to run properly. It only needs a machine with a proper GPU driver install.



The most compatible way to ensure that an application will run on a range of machines with a range of GPU driver installs is to compile your application using the oldest CUDA toolkit required to meet the needs of the earliest GPU driver in the range you intend to cover. Again, you can refer to the table here.





Thanks, how can I then address issues arising when kernel_launching<<<..>>> an application compiled with CUDA 9.2 on drivers below 396.xx (which as you link is the minimum to support 9.2) ?
– Dean
Jul 2 at 14:04


kernel_launching<<<..>>>





I've edited my answer. Quite simply, you need to build with an older CUDA toolkit version, as old as is necessary to support the oldest driver in the range you wish to support. CUDA has forward compatibility so that such an application (with static cudart linking) should run correctly on newer installs, but it does not have backward compatibility at this time for applications compiled against newer toolkits running on older drivers.
– Robert Crovella
Jul 2 at 14:06






By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

-

Popular posts from this blog

api-platform.com Unable to generate an IRI for the item of type

api-platform.com Unable to generate an IRI for the item of type I'm trying to setup my first api-platform instance and have run into some presumably noob questions around IDE. What i'm trying to do is setting up an register and reset entity in order to deal with registrations and password reminders from an app. And then have the entity's passed to a custom controller to do some checks, balances and mails afterwards (haven’t gotten around to that yet). But I keep running into this error when posting to my custom entities: { "@context": "/api/contexts/Error", "@type": "hydra:Error", "hydra:title": "An error occurred", "hydra:description": "Unable to generate an IRI for the item of type "AppXXXUserBundleEntityRegister"", "trace": [ { "namespace": "", "short_class": "", "c
Read more

PHP contact form sending but not receiving emails

PHP contact form sending but not receiving emails I have created a contact form using PHP to send the data. The form seems to send fine as there are no errors and success message appears, however I am not receving the emails into the designated inbox. I am using PHPMailer as I originally tried just using the php 'mail' command which I now understand is a bit hit and miss. I cannot see why I would not be receving the emails so I would be very grateful for any help that could be given. I'm quite new to PHP so please be patient with me :) <?php use PHPMailerPHPMailerPHPMailer; use PHPMailerPHPMailerException; $msg = ""; if (isset($_POST['submit'])) { require 'phpmailer/src/Exception.php'; require 'phpmailer/src/PHPMailer.php'; require 'phpmailer/src/SMTP.php'; function sendemail ($to, $from, $body) { $mail = new PHPMailer(true); $mail->setFrom($from); $mail->addAddress($to); $ma
Read more

Do graphics cards have individual ID by which single devices can be distinguished?

Do graphics cards have individual ID by which single devices can be distinguished? I have several graphics cards of the same manufacturer and same model and I want to distinguish them not only by the pcie socket they reside in but by some individual value I can read from the hardware. Is there such value and if so what would it be called and how could I access it? Right now I'm using modern nvidia cards (10xx series) and nvidia inspector.but I'd prefer if there was any value that was also applicable to amd. I have programming and scripting knowledge. I am running Windows 10 1 Answer 1 This is a powershell example to pull information from video cards. The DeviceID is supposed to be unique but only if the manufacturer supports Windows Display Driver Model (WDDM). (I copied this directly from docs.microsoft.com) $strComputer = "." $colItems = get-wmiobject -class "Win32_VideoC
Read more