Hi all, I’m encountering different behavior of this test code, depending on the platform that it runs: #include <cub/device/device_select.cuh> #include <cub/iterator/counting_input_iterator.cuh> struct LessThan { int compare; __host__ __device__ __forceinline__ LessThan(int compare): c…

The solution is that CUB failed to work due to a deprecated driver version. compute-sanitizer revealed the provided PTX was compiled with an unsupported toolchain With -sm=arch_90, the output is in agreement with the A100 system. I have checked for the CUDA error code in the test program, but cud…

Different results for cub::DeviceSelect::If

Accelerated Computing GPU-Accelerated Libraries

Robert_Crovella April 5, 2024, 2:34pm 8

No I don’t think that should happen. My suggestion is to file a bug.

1 Like

Topic		Replies	Views
CUB complex number reduction iterator_traits problem GPU-Accelerated Libraries thrust , cub	4	792	September 29, 2022
Debugging device code does not work CUDA Programming and Performance	7	2897	July 11, 2013
float reduction, cpu and cuda answers differ CUDA Programming and Performance	4	3331	April 1, 2008
Performance drop after specifying CUDA_VISIBLE_DEVICES=0 CUDA Programming and Performance cuda	6	337	April 5, 2024
CPU hangs when calling thrust::copy_if CUDA Programming and Performance	14	2591	August 10, 2015
Certain samples fail with 'no device supporting CUDA' CUDA Programming and Performance	4	12476	January 26, 2009
Slow CUDA SGEMM CUDA Programming and Performance	5	663	September 15, 2022
Correct output with emulation mode, wrong with GPU/Execution CUDA Programming and Performance	6	3342	March 25, 2010
Problems with CUDA CUDA Programming and Performance	8	2840	December 3, 2012
cusparseSpMM fails with complex data GPU-Accelerated Libraries cusparse	3	445	June 6, 2023

Different results for cub::DeviceSelect::If

Related topics