I finally got the compiler working, thanks Dave for the license key, but now I’m getting a few errors when targetting compute capability 1.1 (sadly, my laptop’s card fx770m only supports cc11)
$ pgcc -o pgpiacc pi.c -Minfo=all -acc -ta=nvidia:cc11
PGC-S-0155-No valid compute capability generated (pi.c)
PGC-W-0155-Compiler failed to translate accelerator region (see -Minfo messages): Could not find GPU binary file (pi.c: 7)
main:
7, Double precision operations disable compute capability 1.1 kernel
8, Loop is parallelizable
#pragma acc loop gang, vector(256) /* blockIdx.x threadIdx.x */
9, Double precision operations disable compute capability 1.1 kernel
10, Sum reduction generated for pi
Double precision operations disable compute capability 1.1 kernel
PGC/x86 Linux 12.4-0: compilation completed with severe errors
If I set compute capability 1.3 or higher all is fine at compile time
$ pgcc -o pgpiacc pi.c -Minfo=all -acc -ta=nvidia:cc13
main:
7, Generating compute capability 1.3 binary
8, Loop is parallelizable
#pragma acc loop gang, vector(256) /* blockIdx.x threadIdx.x */
CC 1.3 : 21 registers; 2072 shared, 36 constant, 0 local memory bytes; 50% occupancy
10, Sum reduction generated for pi
Not so at run time (as expected)
$ ./pgpiacc
Compute capability mismatch
file: /home/k/workspace/pi.c
routine: main
line: 7
device: 0 compute capability 1.1
driver: 4000
Available compute capability: 1.3(elf) 1.3(ptx)
Any ideas?
TIA