Groups

cub-users

Contact owners and managers

1–30 of 145

Welcome to the cub-users discussion forum!

CUB is a library of high-performance parallel primitives and other utilities for building CUDA software. CUB targets all layers of the CUDA programming model, greatly simplifying the development and maintenance of complex device-level, block-level, warp-level, and thread-level operations.

CUB is developed as an open-source project by NVIDIA Research. Our primary release v1.3.2 is now available!

Cheers,
Duane

0 selected

3/26/14

I'll be giving a CUB talk at GTC'14 this week in San Jose (session link), and I invite anyone

unread,

I'll be giving a CUB talk at GTC'14 this week in San Jose (session link), and I invite anyone

3/26/14

3/10/13

What primitives would you like to see implemented in CUB? Post in this discussion category!

CUB is a growing, evolving software library whose goal is to simplify CUDA programming by providing

unread,locked,

What primitives would you like to see implemented in CUB? Post in this discussion category!

CUB is a growing, evolving software library whose goal is to simplify CUDA programming by providing

3/10/13

12/21/20

Large operator in scan

In all the CUB examples, the ScanOps are elementary operations on scalar types T. In my problem, T is

unread,

Large operator in scan

In all the CUB examples, the ScanOps are elementary operations on scalar types T. In my problem, T is

12/21/20

Peng Ding, walden tang2

1/14/20

how to compile cub function with cmake.

just set the include path of CUB in the cmakelists.txt 在 2018年6月2日星期六 UTC+8下午5:54:42，Peng Ding写道：

unread,

how to compile cub function with cmake.

just set the include path of CUB in the cmakelists.txt 在 2018年6月2日星期六 UTC+8下午5:54:42，Peng Ding写道：

1/14/20

rhaney, Robert Crovella4

7/9/19

embedded call to cub in static library

for nvcc it's very simple nvcc -c a.cu b.cu nvcc --lib ao bo -o myLib.a This assumes you don'

unread,

embedded call to cub in static library

for nvcc it's very simple nvcc -c a.cu b.cu nvcc --lib ao bo -o myLib.a This assumes you don'

7/9/19

Shahar, … Vincent Roberge4

6/12/19

In place operations

This addition to the documentation has not been done. It would be important to do it as in-place

unread,

In place operations

This addition to the documentation has not been done. It would be important to do it as in-place

6/12/19

rhaney, Robert Crovella5

4/25/19

There is no reason to be using ::SortPairs if you are not doing key-value sorting. ::SortKeys may be

unread,

There is no reason to be using ::SortPairs if you are not doing key-value sorting. ::SortKeys may be

4/25/19

4/15/19

cuFloatComplex cub::BlockReduce from within CUDA kernel

I figured it out - nothing special but I will post the code below for anyone who may be looking for

unread,

cuFloatComplex cub::BlockReduce from within CUDA kernel

I figured it out - nothing special but I will post the code below for anyone who may be looking for

4/15/19

4/8/19

call CUB DeviceReduce from Host

Hello all, This may be a simple question, but is it possible to call a CUB DeviceReduce function from

unread,

call CUB DeviceReduce from Host

Hello all, This may be a simple question, but is it possible to call a CUB DeviceReduce function from

4/8/19

rhaney, Robert Crovella5

4/5/19

CUB with cuComplex

Thank you for the help. The code sample you posted looks exactly like what I was looking for. I agree

unread,

CUB with cuComplex

Thank you for the help. The code sample you posted looks exactly like what I was looking for. I agree

4/5/19

6/2/18

how to compile cub function with cmake.

Dear friends: I want to use CUB to conduct reduction in my cuda project. My project is compiled with

unread,

how to compile cub function with cmake.

Dear friends: I want to use CUB to conduct reduction in my cuda project. My project is compiled with

6/2/18

5/8/18

CachingDeviceAllocator for pinned memory

Hi, Posted this mistakenly in the general group... I saw some reference in a GTC2018 lecture that

unread,

CachingDeviceAllocator for pinned memory

Hi, Posted this mistakenly in the general group... I saw some reference in a GTC2018 lecture that

5/8/18

5/8/18

CachingDeviceAllocator for pinned memory

Hi, I saw some reference in a GTC2018 lecture that nvidia supplied some version of

unread,

CachingDeviceAllocator for pinned memory

Hi, I saw some reference in a GTC2018 lecture that nvidia supplied some version of

5/8/18

1/21/18

Radix sort partition boundaries

As I understand it, each iteration of radix sort is a partition (to more than 2 bins, depending on

unread,

Radix sort partition boundaries

As I understand it, each iteration of radix sort is a partition (to more than 2 bins, depending on

1/21/18

1/13/18

Hi, What is CUB's equivalent to thrust's adjacent_difference ?? Thanks Eyal

unread,

Hi, What is CUB's equivalent to thrust's adjacent_difference ?? Thanks Eyal

1/13/18

Tarik Saidani, nirvedh meshram2

11/10/17

How do I do block-wide ArgMax

I have the same question. Could someone please answer this. I see this implemented for device-wide

unread,

How do I do block-wide ArgMax

I have the same question. Could someone please answer this. I see this implemented for device-wide

11/10/17

RaulPPelaez, … Robert Crovella6

11/8/17

cub::DeviceRadixSort::SortPairs end_bit usage?

On Wednesday, November 8, 2017 at 5:13:35 PM UTC+1, txbob wrote: it has nothing to do with the number

unread,

cub::DeviceRadixSort::SortPairs end_bit usage?

On Wednesday, November 8, 2017 at 5:13:35 PM UTC+1, txbob wrote: it has nothing to do with the number

11/8/17

David Steinhauer, Erich Elsen2

9/29/17

Reduction where output uses larger data type than input data type (e.g., input array is unsigned short, output is int64)

You can do this by wrapping the input with a transform iterator that does the up conversion. Then the

unread,

Reduction where output uses larger data type than input data type (e.g., input array is unsigned short, output is int64)

You can do this by wrapping the input with a transform iterator that does the up conversion. Then the

9/29/17

Jeremy Buhler, RaulPPelaez2

9/15/17

Compile failure with CUB 1.6.4 WarpScan

I noticed this behavior, it appears hat it is not meant to be used with a logical warp size of 1 for

unread,

Compile failure with CUB 1.6.4 WarpScan

I noticed this behavior, it appears hat it is not meant to be used with a logical warp size of 1 for

9/15/17

Robert Crovella, Duane Merrill III2

8/16/17

Re: [cub-users: 348] I need version CUB 1.1.1

Should be available here: https://github.com/NVlabs/cub/tree/1.1.1 On Mon, Aug 14, 2017 at 11:27 PM,

unread,

Re: [cub-users: 348] I need version CUB 1.1.1

Should be available here: https://github.com/NVlabs/cub/tree/1.1.1 On Mon, Aug 14, 2017 at 11:27 PM,

8/16/17

Martin Dušek, … Eyal Hirsch5

8/13/17

Pitched memory support

Hi, Was this feature ever implemented? I need a pitched input for the cub::DeviceReduce::ArgMin

unread,

Pitched memory support

Hi, Was this feature ever implemented? I need a pitched input for the cub::DeviceReduce::ArgMin

8/13/17

7/30/17

calling DeviceScan from kernel

Hi, I'm calling DeviceScan::ExclusiveSum from a kernel function. Although I'm getting

unread,

calling DeviceScan from kernel

Hi, I'm calling DeviceScan::ExclusiveSum from a kernel function. Although I'm getting

7/30/17

Kar, … Robert Crovella5

7/27/17

CUB InclusiveSum() fails when debugging with VS2015 Nsight Memcheck enabled

Yeah, I did it yesterday and got confirmation it's a bug. They will give me news when it is

unread,

CUB InclusiveSum() fails when debugging with VS2015 Nsight Memcheck enabled

Yeah, I did it yesterday and got confirmation it's a bug. They will give me news when it is

7/27/17

Oren Tropp, Duane Merrill III2

7/23/17

Is Cub sort stable?

Hi Oren, It is a stable sort. (Stability is a necessary property of for the proper operation of the

unread,

Is Cub sort stable?

Hi Oren, It is a stable sort. (Stability is a necessary property of for the proper operation of the

7/23/17

14154...@qq.com, Jacob2

7/12/17

all .cuh compile error visual studio 2010+cuda 5.5+cub-1.6.3 win32

Did you solve this? I met similar problem. On Wednesday, December 21, 2016 at 1:18:20 PM UTC+8, 14154

unread,

all .cuh compile error visual studio 2010+cuda 5.5+cub-1.6.3 win32

Did you solve this? I met similar problem. On Wednesday, December 21, 2016 at 1:18:20 PM UTC+8, 14154

7/12/17

7/8/17

Thousands of errors returned by visual studio intellisense

Whenever I include "cub/cub.cuh" file to my app source code, I get the following error:

unread,

Thousands of errors returned by visual studio intellisense

Whenever I include "cub/cub.cuh" file to my app source code, I get the following error:

7/8/17

7/7/17

Small erratas in documentation

I sometimes see small incoherences in the docs and I thought it would be a good idea to post them

unread,

Small erratas in documentation

I sometimes see small incoherences in the docs and I thought it would be a good idea to post them

7/7/17

Vincent Roberge, cukier9a7b52

7/6/17

Excellent library which needs to be more advertised by NVIDIA

Hi I am a student of the University of Technology, I learnt (basic) cuda a year or two ago. Now I

unread,

Excellent library which needs to be more advertised by NVIDIA

Hi I am a student of the University of Technology, I learnt (basic) cuda a year or two ago. Now I

7/6/17

7/5/17

CUB documentation

I know that there is a CUB documentation at http://nvlabs.github.io/cub/ but is there any possibility

unread,

CUB documentation

I know that there is a CUB documentation at http://nvlabs.github.io/cub/ but is there any possibility

7/5/17

7/5/17

Partitioning an array of n-tuples

I'd like to partition an array of n-tuples based on whether the one of the elements of a tuple is

unread,

Partitioning an array of n-tuples

I'd like to partition an array of n-tuples based on whether the one of the elements of a tuple is

7/5/17

Eugene Brevdo, Duane Merrill III2

6/30/17

Change in sha256 sum of cub-1.7.0.zip?

Hi, sorry for the inconvenience. Although we don't normally change release archives once forked,

unread,

Change in sha256 sum of cub-1.7.0.zip?

Hi, sorry for the inconvenience. Although we don't normally change release archives once forked,

6/30/17

5/31/17

(Segmented) Radix Select

Hey Duane, Long time! I remember you mentioned you were working on a MSB radix sort. Any chance

unread,

(Segmented) Radix Select

Hey Duane, Long time! I remember you mentioned you were working on a MSB radix sort. Any chance

5/31/17

Search

Clear search

Close search

Google apps

Main menu