Cannot run OpenACC on the AMD Radeon R9 380 GPU


I tried another AMD Radeon GPU. This time, the R9 380 (Tonga), which is similar to the R9 285 (Tonga) GPU. The R9 285, R9 380, and R9 Fury Nano feature AMD’s GCN 1.2 micro-architecture. Could this be the reason why I’m not having any luck?

$ pgcc -acc -Minfo -ta=radeon:hawaii acc_c1.c 
     36, Generating copyout(r[:n])
         Generating copyin(a[:n])
     37, Loop is parallelizable
         Accelerator kernel generated
         Generating Radeon code
         37, #pragma acc loop gang, vector(128) /* global dim(0) local dim(0) */

$ ./a.out 
r=-1, AMD GPU not found or not available
Current file:     /opt/pgi/linux86-64/2015/examples/OpenACC/samples/acc_c1/acc_c1.c
        function: main
        line:     36
Current region was compiled for:
  AMD Radeon GPU
Available accelerators:
  device[1]: Native X86 (CURRENT DEVICE)
The accelerator does not match the profile for which this program was compiled

$ pgaccelinfo

OpenCL Platform:               AMD Accelerated Parallel Processing
OpenCL Vendor:                 Advanced Micro Devices, Inc.

Device Number:                 0
Device Name:                   Tonga
Available:                     Yes
Compiler Available:            Yes
Board Name:                    AMD Radeon (TM) R9 380 Series 
Device Version:                OpenCL 2.0 AMD-APP (1800.11)
Global Memory Size:            1861222400
Maximum Object Size:           1345585152
Global Cache Size:             16384
Free Memory:                   1811192000
Max Clock (MHz):               980
Compute Units:                 28
SIMD Units:                    4
SIMD Width:                    16
GPU Cores:                     1792
Wavefront Width:               64
Constant Memory Size:          65536
Local Memory Size:             32768
Workgroup Size:                256
Address Bits:                  64
ECC Support:                   No

The R9 380 is the Sapphire R9 380 2G D5 (ITX Compact), which is the the small form factor of the bunch, similar in size to the R9 Fury Nano.

Will support for (-ta=radeon:tonga for R9 285/380) and/or (-ta=radeon:fiji for R9 Fury) be added in a later OpenACC release?

Thank you,