OpenCL HostApp fails with acl_bind_buffer_to_device: Assertion `mem' failed

[Altera_Forum]

Hey, 

 first some infos about the environment:  Board: Nallatech 510T (2x Arria 10) Ubuntu 16.04. Quartus 17.1.0  Flashing the .aocx and compiling of the HostApp was successfull.  Now when i start the HostApp, the device was found and the HostApp tried to start.  But then the following error is shown: acl_mem.c:398: acl_bind_buffer_to_device: Assertion `mem' failed.   See full result: Listing OpenCL devices (OCLMiner). ℹ 13:00:38|ethminer Found suitable OpenCL device with 4294967296 bytes of GPU memory ℹ 13:00:38|stratum Connecting to stratumV2 server eth-eu1.nanopool.org:9999 ℹ 13:00:38|stratum Connected! ℹ 13:00:38|stratum Starting farm ocl 13:00:38|ocl-0 No work. Pause for 3 s. ocl 13:00:38|ocl-1 No work. Pause for 3 s. ℹ 13:00:38|stratum Received new job# 0x7c1a12 seed:# 8308d376eeb469b7ff84bd59c51988d9 target:# 000000006df37f675ef6eadf ℹ 13:00:38|stratum Received new job# 0x7c1a12 seed:# 8308d376eeb469b7ff84bd59c51988d9 target:# 000000006df37f675ef6eadf ocl 13:00:41|ocl-0 New work: header# 7c1a1211… target 000000006df37f675ef6eadf5ab9a2072d44268d97df837e6748956e5c6c2116 ocl 13:00:41|ocl-0 New seed# 8308d376… ocl 13:00:41|ocl-1 New work: header# 7c1a1211… target 000000006df37f675ef6eadf5ab9a2072d44268d97df837e6748956e5c6c2116 ocl 13:00:41|ocl-1 New seed# 8308d376… ocl 13:00:43|ocl-0 Platform: Intel(R) FPGA SDK for OpenCL(TM) ocl 13:00:43|ocl-1 Platform: Intel(R) FPGA SDK for OpenCL(TM) ocl 13:00:43|ocl-0 Device: p510t_min_ax115 : nalla_pcie (aclnalla_pcie0) 1CU 4096MB / OpenCL 1.0 Intel(R) FPGA SDK for OpenCL(TM), Version 17.1 ocl 13:00:43|ocl-0 OpenCL kernel: Custom 'p510t_min_ax115 : nalla_pcie (aclnalla_pcie0).aocx' ocl 13:00:43|ocl-1 Device: p510t_min_ax115 : nalla_pcie (aclnalla_pcie1) 1CU 4096MB / OpenCL 1.0 Intel(R) FPGA SDK for OpenCL(TM), Version 17.1 ocl 13:00:43|ocl-1 OpenCL kernel: Custom 'p510t_min_ax115 : nalla_pcie (aclnalla_pcie1).aocx' m 13:00:43|ethminer Speed 0.00Mh/s No-Fee (295s) p510t_min_ax115 : nalla_pcie (aclnalla_pcie0) 1CU 4096MB - 0C 0% - 0.00Mh/s p510t_min_ax115 : nalla_pcie (aclnalla_pcie1) 1CU 4096MB - 0C 0% - 0.00Mh/s ocl 13:00:43|ocl-1 OpenCL kernel: Custom 'p510t_min_ax115 : nalla_pcie (aclnalla_pcie1).aocx' ocl 13:00:43|ocl-1 OpenCL kernel: GROUP_SIZE 128 ocl 13:00:43|ocl-0 OpenCL kernel: Custom 'p510t_min_ax115 : nalla_pcie (aclnalla_pcie0).aocx' ocl 13:00:43|ocl-0 OpenCL kernel: GROUP_SIZE 128 ocl 13:00:43|ocl-1 OpenCL kernel: DAG_SIZE 21495797 ocl 13:00:43|ocl-0 OpenCL kernel: DAG_SIZE 21495797 ocl 13:00:43|ocl-1 OpenCL kernel: LIGHT_SIZE 671743 ocl 13:00:43|ocl-0 OpenCL kernel: LIGHT_SIZE 671743 ocl 13:00:43|ocl-1 OpenCL kernel: ACCESSES 64 ocl 13:00:43|ocl-0 OpenCL kernel: ACCESSES 64 ocl 13:00:43|ocl-1 OpenCL kernel: MAX_OUTPUTS 1 ocl 13:00:43|ocl-0 OpenCL kernel: MAX_OUTPUTS 1 ocl 13:00:43|ocl-1 OpenCL kernel: PLATFORM 0 ocl 13:00:43|ocl-0 OpenCL kernel: PLATFORM 0 ocl 13:00:43|ocl-1 OpenCL kernel: COMPUTE 0 ocl 13:00:43|ocl-0 OpenCL kernel: COMPUTE 0 ocl 13:00:43|ocl-1 OpenCL kernel: THREADS_PER_HASH 4 ocl 13:00:43|ocl-0 OpenCL kernel: THREADS_PER_HASH 4 ✘ 13:00:44|ocl-0 Build info: ✘ 13:00:44|ocl-1 Build info: ethminer: acl_mem.c:398: acl_bind_buffer_to_device: Assertion `mem' failed.   I used the ethminer from https://github.com/maetti79/ethminer  Hope you can help.  If you need more information, just ask.  thank you!

[Altera_Forum]

You should probably ask the author of that repository. I would be surprised if that code even works.

[Altera_Forum]

We talked to the author already and he sent us a newer version, but with the same issue and nearly the same code.  

We added a debugging and error helpers, but can't get past the initialisation stage. The author does not respond anymore, but he said, he has a working Nallatech 510S with this code. That was the last information we received from him.   I think the error has something to do with the memory release. The verbose debugging says "double free or corruption (!pref): 0x00000...124c6a0...)". The host has 32GB physical ram and our FPGA has 2 x 8GB Global Ram per FPGA Accelerator (Nallatech 510T with 2x Arria10).   Maybe it has something to do with host memory. Do you know the error or could you tell us where we should look?  Thank you!  PS: hello_world and stuff like that is working without any issue.