search

lattice / quda

QUDA is a library for performing calculations in lattice QCD on GPUs.
https://lattice.github.io/quda
Other
286 stars 94 forks source link

Restore support for BLAS tuning on pre-Fermi cards #58

Closed rbabich closed 12 years ago

rbabich commented 12 years ago

Assigning Mike...

maddyscientist commented 12 years ago

Thanks for the reminder. I'd forgotten about that.

On Mar 27, 2012, at 22:42, Ron Babichreply@reply.github.com wrote:

Assigning Mike...

Reply to this email directly or view it on GitHub: https://github.com/lattice/quda/issues/58

rbabich commented 12 years ago

Mac OS X 10.6.8, CUDA 4.1, single-GPU build for sm_12 with current master, I get:

/usr/local/cuda/bin/nvcc  -O3 -m64 -D__COMPUTE_CAPABILITY__=120 -DOVERLAP_COMMS -DGPU_WILSON_DIRAC -DGPU_CLOVER_DIRAC -DGPU_WILSON_DIRAC -DFERMI_NO_DBLE_TEX -arch=sm_12       -I/usr/local/cuda/include -I../include -Idslash_core -I. --maxrregcount=80 blas_quda.cu -c -o blas_quda.o -Xptxas -abi=no
./blas_core.h(8): Error: Formal parameter space overflowed (256 bytes max) in function _Z10blasKernelI7double2Li1ELi0ELi1ELi0ELi0E13SpinorTextureIS0_S0_S0_Li1ELi0EES1_IS0_S0_S0_Li1ELi1EES1_IS0_S0_S0_Li1ELi2EES1_IS0_S0_S0_Li1ELi3EE6SpinorIS0_S0_S0_Li1EES7_S7_S7_5axpbyIS0_S0_EEvT5_T6_T7_T8_T13_T9_T10_T11_T12_i
make[1]: *** [blas_quda.o] Error 2
make: *** [lib] Error 2
rbabich commented 12 years ago

Nevermind! I was in the wrong branch...