Inconsistency between OpenGL and CUDA maximum number of threads -


my gpu nvidia geforce gt440, compute capability version 2.x. nvidia's official cuda_c_programming_guide points out

limit 1. maximum number of threads per block = 1024
limit 2. maximum number of resident threads per multiprocessor = 1536

however, 2 of opengl computer shader implementation limits are

limit 3. gl_max_compute_work_group_invocations = 1536

my questions are
1. why limit 1 not equal limit 2 , limit 3?
2. should real threads/block (invocations/workgroup) 1024 or 1536?

why limit 1 not equal limit 2 , limit 3?

because isn't same thing. blocks logical construct in cuda , limited maximum of 1024 threads. multiprocessor can run multiple blocks concurrently (up 8 in case of hardware). sm can have 1536 concurrent threads in hardware, not of threads can come single block.

should real threads/block 1024 or 1536?

1024 reasons outlined above. can see complete summary of capabilities of supported hardware here.


Comments

Post a Comment

Popular posts from this blog

PHP and MySQL WP -

i trying accomplish 2 things here code. writing php directly wp page using insert php plugin. 1) query mysql database , return results 1 column 4 columns. 2) make each result own hyperlink directs user new wp page runs new query on page. i able query show results , turn results dynamic hyperlink, cannot formatting down. right returns result 1 column. here have: [insert_php] $servername = "localhost"; $username = "login"; //edited privacy $password = "password"; //edited privacy $database = "database"; //edited privacy // create connection $conn = new mysqli($servername, $username, $password, $database); // check connection if ($conn->connect_error) { die("connection failed: " . $conn->connect_error); } echo "connected successfully"; //get results database $result = mysqli_query($conn,"select mine mines order mine"); $all_property = array(); //declare array saving property //showing proper...
Read more

android - InAppBilling registering BroadcastReceiver in AndroidManifest -

i'm implementing inapp billing in android app. works well, however, i'm trying decouple broadcast receiver activity manifest. suggestion in android's trivialdrive sample: // important: dynamically register broadcast messages updated purchases. // register receiver here instead of <receiver> in manifest // because call getpurchases() @ startup, therefore can ignore // broadcasts sent while app isn't running. // note: registering listener in activity bad idea, done here // because sample. regardless, receiver must registered after // iabhelper setup, before first call getpurchases(). currently there's class extends broadcastreceiver : public class iabbroadcastreceiver extends broadcastreceiver { /** * intent action receiver should filter for. */ public static final string action = "com.android.vending.billing.purchases_updated"; private final iabbroadcastlistener mlistener; public iabbroadcastreceiver(iabbroadcastlistener listener) { m...
Read more

go - golang pprof for c library code -

i have been trying profile golang application pprof, seems details of c code(linked static library) missed in output result. flame graph from frame graph can see c code time consuming reported within "runtime._externalcode", no details @ all. , sure static library compiled debug enable. so there way make pprof details(like stack trace) of 3rd party c code? or there other golang profiling tool this?
Read more