вот тебе лично собранный на сраном CL с рокма, а значешь почему там трешь? потому что м***ки из амд решили что у них совой прекрасно кастрированный ЦЛ
user@p51n67:~/tmp/clmagma-1.3.0/testing$ LD_LIBRARY_PATH=/home/suser/tmp/clmagma-1.3.0/lib:/opt/rocm/opencl/lib/x86_64 ./testing_sgetrf
% clMAGMA 1.3.0
% OpenCL platform . MAGMA not compiled with OpenMP.
% Device: gfx900, 8176.0 MiB memory, max allocation 6949.6 MiB, driver 2545.0 (HSA1.1,LC)
% Device: gfx900, 8176.0 MiB memory, max allocation 6949.6 MiB, driver 2545.0 (HSA1.1,LC)
% Device: gfx900, 8176.0 MiB memory, max allocation 6949.6 MiB, driver 2545.0 (HSA1.1,LC)
% Device: gfx900, 8176.0 MiB memory, max allocation 6949.6 MiB, driver 2545.0 (HSA1.1,LC)
Usage: ./testing_sgetrf [options] [-h|--help]
ngpu 1
M N CPU GFlop/s (sec) GPU GFlop/s (sec) |PA-LU|/(N*|A|)
=========================================================================
/tmp/AMD_27820_73/t_27820_75.cl:1:26: warning: unknown OpenCL extension 'cl_amd_printf' - ignoring [-Wignored-pragmas]
#pragma OPENCL EXTENSION cl_amd_printf : enable
^
/tmp/AMD_27820_73/t_27820_75.cl:85:14: warning: assigning to '__global float *' from 'const __global float *' discards qualifiers [-Wincompatible-pointer-types-discards-qualifiers]
uA.f = A + k0;
^ ~~~~~~
/tmp/AMD_27820_73/t_27820_75.cl:131:14: warning: assigning to '__global float *' from 'const __global float *restrict' discards qualifiers [-Wincompatible-pointer-types-discards-qualifiers]
uA.f = A;
^ ~
/tmp/AMD_27820_73/t_27820_75.cl:153:38: warning: incompatible pointer types initializing '__global float4 *' with an expression of type '__global float *' [-Wincompatible-pointer-types]
__global float4 *pC = uC.f;
^ ~~~~
/tmp/AMD_27820_73/t_27820_75.cl:270:34: warning: incompatible pointer types initializing '__global float4 *' with an expression of type '__global float *' [-Wincompatible-pointer-types]
__global float4 *pC = uC.f;
^ ~~~~
5 warnings generated.
1088 1088 --- ( --- ) 1.13 ( 0.76) ---
2112 2112 --- ( --- ) 139.10 ( 0.05) ---
3136 3136 --- ( --- ) 281.23 ( 0.07) ---
4160 4160 --- ( --- ) 27.15 ( 1.77) ---
5184 5184 --- ( --- ) 652.67 ( 0.14) ---
6208 6208 --- ( --- ) 804.78 ( 0.20) ---
7232 7232 --- ( --- ) 726.76 ( 0.35) ---
8256 8256 --- ( --- ) 1031.88 ( 0.36) ---
9280 9280 --- ( --- ) 688.00 ( 0.77) ---
non-GPU-resident version not implemented
10304 10304 --- ( --- ) 18651370.12 ( 0.00) ---
suser@p51n67:~/tmp/clmagma-1.3.0/testing$
это я должен показывать коммисси?
этот треш не КОМПИЛЯЦИЯ, ЭТО СРАНОЕ ИСПОЛНЕНИЕ. а то вдруг, не сообразишь.