150         uint32_t const bc_max = msb_idx_u32(pow2_rd_u32(merge->warps));  in hsg_target_cuda()
185                 merge->warps,  in hsg_target_cuda()
186                 msb_idx_u32(pow2_ru_u32(merge->warps)),  in hsg_target_cuda()
188                 config->merge.flip.warps,  in hsg_target_cuda()
191                 config->merge.half.warps,  in hsg_target_cuda()
310         uint32_t const bs  = pow2_ru_u32(m->warps);  in hsg_target_cuda()
317                     m->warps,msb);  in hsg_target_cuda()
323                     m->warps,msb);  in hsg_target_cuda()
332         if (m->warps > 1)  in hsg_target_cuda()
336                     m->warps * config->warp.lanes,  in hsg_target_cuda()
357         uint32_t const msb = msb_idx_u32(m->warps);  in hsg_target_cuda()
361                 m->warps,msb);  in hsg_target_cuda()
369         if (m->warps > 1)  in hsg_target_cuda()
373                     m->warps * config->warp.lanes,  in hsg_target_cuda()
384         uint32_t const span_left  = (merge[0].warps << ops->a) / 2;  in hsg_target_cuda()
407         uint32_t const msb = msb_idx_u32(pow2_ru_u32(merge[0].warps));  in hsg_target_cuda()
551               merge[ops->a].warps,ops->c,ops->b);  in hsg_target_cuda()
557               ops->b,merge[ops->a].warps,ops->c);  in hsg_target_cuda()
636                 m->warps);  in hsg_target_cuda()
646                 m->warps);  in hsg_target_cuda()
660         if (m->warps <= 32)  in hsg_target_cuda()