__device__ void single_precision_intrinsics() { float fX, fY; __cosf(0.0f); __exp10f(0.0f); __expf(0.0f); __fadd_rd(0.0f, 1.0f); __fadd_rn(0.0f, 1.0f); __fadd_ru(0.0f, 1.0f); __fadd_rz(0.0f, 1.0f); __fdiv_rd(4.0f, 2.0f); __fdiv_rn(4.0f, 2.0f); __fdiv_ru(4.0f, 2.0f); __fdiv_rz(4.0f, 2.0f); __fdividef(4.0f, 2.0f); __fmaf_rd(1.0f, 2.0f, 3.0f); __fmaf_rn(1.0f, 2.0f, 3.0f); __fmaf_ru(1.0f, 2.0f, 3.0f); __fmaf_rz(1.0f, 2.0f, 3.0f); __fmul_rd(1.0f, 2.0f); __fmul_rn(1.0f, 2.0f); __fmul_ru(1.0f, 2.0f); __fmul_rz(1.0f, 2.0f); __frcp_rd(2.0f); __frcp_rn(2.0f); __frcp_ru(2.0f); __frcp_rz(2.0f); __frsqrt_rn(4.0f); __fsqrt_rd(4.0f); __fsqrt_rn(4.0f); __fsqrt_ru(4.0f); __fsqrt_rz(4.0f); __fsub_rd(2.0f, 1.0f); __fsub_rn(2.0f, 1.0f); __fsub_ru(2.0f, 1.0f); __fsub_rz(2.0f, 1.0f); __log10f(1.0f); __log2f(1.0f); __logf(1.0f); __powf(1.0f, 0.0f); __saturatef(0.1f); __sincosf(0.0f, &fX, &fY); __sinf(0.0f); __tanf(0.0f); }
__device__ inline float occaCuda_fastLog2(const float x){ return __log2f(x); }