diff options
| -rw-r--r-- | qwen_asr_kernels.c | 21 |
1 files changed, 21 insertions, 0 deletions
diff --git a/qwen_asr_kernels.c b/qwen_asr_kernels.c index efa606c..db083d4 100644 --- a/qwen_asr_kernels.c +++ b/qwen_asr_kernels.c @@ -8,6 +8,9 @@ #if (defined(__AVX512F__) || defined(__AVX2__)) && (defined(__x86_64__) || defined(__i386__) || defined(_M_X64) || defined(_M_IX86)) #include <immintrin.h> #endif +#if defined(__ARM_NEON) || defined(__ARM_NEON__) +#include <arm_neon.h> +#endif #ifdef __APPLE__ #include <sys/sysctl.h> #else @@ -1616,6 +1619,24 @@ void qwen_apply_rope_neox(float *x, const float *cos_vals, const float *sin_vals vec[d] = x1 * c[d] + (-x2) * sn[d]; vec[half + d] = x2 * c[half + d] + x1 * sn[half + d]; } +#elif defined(__ARM_NEON) || defined(__ARM_NEON__) + int d = 0; + for (; d + 4 <= half; d += 4) { + float32x4_t x1 = vld1q_f32(vec + d); + float32x4_t x2 = vld1q_f32(vec + half + d); + float32x4_t cc = vld1q_f32(c + d); + float32x4_t ss = vld1q_f32(sn + d); + float32x4_t new1 = vmlsq_f32(vmulq_f32(x1, cc), x2, ss); + float32x4_t new2 = vfmaq_f32(vmulq_f32(x2, cc), x1, ss); + vst1q_f32(vec + d, new1); + vst1q_f32(vec + half + d, new2); + } + for (; d < half; d++) { + float x1 = vec[d]; + float x2 = vec[half + d]; + vec[d] = x1 * c[d] + (-x2) * sn[d]; + vec[half + d] = x2 * c[half + d] + x1 * sn[half + d]; + } #else for (int d = 0; d < half; d++) { float x1 = vec[d]; /* first half */ |
