def _execute_and_check(self): prg = cl.Program(cfg.OPENCL.ctx, gpu_util.get_source([self.kernel_fn])).build() mem, ar = self._create_mem_objs(self.n) prg.float_test(cfg.OPENCL.queue, (self.n, ), None, mem) cl.enqueue_copy(cfg.OPENCL.queue, ar, mem) res = ar == np.array([0, 1], dtype=cfg.PRECISION.np_float) self.assertTrue(res.all())
def setUp(self): syris.init(device_index=0) self.prg = g_util.get_program( g_util.get_source(["polyobject.cl", "heapsort.cl"], precision_sensitive=True)) self.num = 10 self.data = np.array([1, 8, np.nan, -1, np.nan, 8, 680, 74, 2, 0]).\ astype(cfg.PRECISION.np_float) self.mem = cl.Buffer(cfg.OPENCL.ctx, cl.mem_flags.READ_WRITE | cl.mem_flags.COPY_HOST_PTR, hostbuf=self.data)
def setUp(self): syris.init(device_index=0) self.num_0 = np.array([17 - 38j], dtype=cfg.PRECISION.np_cplx) self.num_1 = np.array([-135 + 563j], dtype=cfg.PRECISION.np_cplx) self.mem_0 = cl.Buffer(cfg.OPENCL.ctx, cl.mem_flags.READ_WRITE | cl.mem_flags.COPY_HOST_PTR, hostbuf=self.num_0) self.mem_1 = cl.Buffer(cfg.OPENCL.ctx, cl.mem_flags.READ_WRITE | cl.mem_flags.COPY_HOST_PTR, hostbuf=self.num_1) self.mem_out = cl.Buffer(cfg.OPENCL.ctx, cl.mem_flags.READ_WRITE, cfg.PRECISION.cl_cplx) self.host_array = np.empty(1, cfg.PRECISION.np_cplx) src = gpu_util.get_source(["vcomplex.cl"]) self.prg = cl.Program(cfg.OPENCL.ctx, src).build()
def setUp(self): default_syris_init() self.prg = g_util.get_program( g_util.get_source(["polyobject.cl", "heapsort.cl"], precision_sensitive=True)) self.num = 10 self.data = np.array([1, 8, np.nan, -1, np.nan, 8, 680, 74, 2, 0]).astype(cfg.PRECISION.np_float) self.mem = cl.Buffer(cfg.OPENCL.ctx, cl.mem_flags.READ_WRITE | cl.mem_flags.COPY_HOST_PTR, hostbuf=self.data)
def setUp(self): default_syris_init() self.num_0 = np.array([17 - 38j], dtype=cfg.PRECISION.np_cplx) self.num_1 = np.array([-135 + 563j], dtype=cfg.PRECISION.np_cplx) self.mem_0 = cl.Buffer(cfg.OPENCL.ctx, cl.mem_flags.READ_WRITE | cl.mem_flags.COPY_HOST_PTR, hostbuf=self.num_0) self.mem_1 = cl.Buffer(cfg.OPENCL.ctx, cl.mem_flags.READ_WRITE | cl.mem_flags.COPY_HOST_PTR, hostbuf=self.num_1) self.mem_out = cl.Buffer(cfg.OPENCL.ctx, cl.mem_flags.READ_WRITE, cfg.PRECISION.cl_cplx) self.host_array = np.empty(1, cfg.PRECISION.np_cplx) src = gpu_util.get_source(["vcomplex.cl"]) self.prg = cl.Program(cfg.OPENCL.ctx, src).build()