On Fri, 23 Oct 2015 15:17:11 +0100
David Hunt <david.hunt at intel.com> wrote:
> +
> +typedef float32x4_t __m128;
> +
> +typedef int32x4_t __m128i;
> +
> +typedef __m128i xmm_t;
> +
> +#define XMM_SIZE (sizeof(xmm_t))
> +#define XMM_MASK (XMM_SIZE - 1)
> +
> +typedef union rte_xmm {
> + xmm_t x;
> + uint8_t u8[XMM_SIZE / sizeof(uint8_t)];
> + uint16_t u16[XMM_SIZE / sizeof(uint16_t)];
> + uint32_t u32[XMM_SIZE / sizeof(uint32_t)];
> + uint64_t u64[XMM_SIZE / sizeof(uint64_t)];
> + double pd[XMM_SIZE / sizeof(double)];
> +} rte_xmm_t __aligned(16);
> +
> +#define _mm_srli_epi32(a, imm) { (__m128i)vshrq_n_u32((uint32x4_t)a, imm) }
> +
> +#define _mm_srli_si128(a, imm) { (__m128i)vextq_s8((int8x16_t)a, \
> + vdupq_n_s8(0), (imm)) }
> +
> +static inline __m128i
> +_mm_set_epi32(int i3, int i2, int i1, int i0);
> +static inline int
> +_mm_cvtsi128_si64(__m128i a);
> +
> +static inline __m128i
> +_mm_set_epi32(int i3, int i2, int i1, int i0)
> +{
> + int32_t __aligned(16) data[4] = { i0, i1, i2, i3 };
> + return vld1q_s32(data);
> +}
> +
> +static inline int
> +_mm_cvtsi128_si64(__m128i a)
> +{
> + return vgetq_lane_s64(a, 0);
> +}
> +
> +static inline __m128i
> +_mm_and_si128(__m128i a, __m128i b)
> +{
> + return (__m128i)vandq_s32(a, b);
> +}
> +
What is the purpose of those wrappers? I mean, does it help to compile
some SSE-dependent libraries in DPDK for NEON?
Jan
--
Jan Viktorin E-mail: Viktorin at RehiveTech.com
System Architect Web: www.RehiveTech.com
RehiveTech
Brno, Czech Republic
