enum dnnl_fpmath_mode_t¶
Overview¶
Floating-point math mode. More…
#include <dnnl_common_types.h> enum dnnl_fpmath_mode_t { dnnl_fpmath_mode_strict, dnnl_fpmath_mode_bf16, dnnl_fpmath_mode_f16, dnnl_fpmath_mode_any, dnnl_fpmath_mode_tf32, };
Detailed Documentation¶
Floating-point math mode.
Enum Values¶
dnnl_fpmath_mode_strict
Default behavior, no downconversions allowed.
dnnl_fpmath_mode_bf16
Implicit f32->bf16 conversions allowed.
dnnl_fpmath_mode_f16
Implicit f32->f16 conversions allowed.
dnnl_fpmath_mode_any
Implicit f32->f16, f32->tf32 or f32->bf16 conversions allowed.
dnnl_fpmath_mode_tf32
Implicit f32->tf32 conversions allowed.