| /external/XNNPACK/src/bf16-gemm/gen/ |
| D | 1x4c8-minmax-neonfma-zip.c | 64 const float32x4_t vb3e = vreinterpretq_f32_u16(vzip1q_u16(vzero, vb3)); in xnn_bf16_gemm_minmax_ukernel_1x4c8__neonfma_zip() local 99 const float32x4_t vb3e = vreinterpretq_f32_u16(vzip1q_u16(vzero, vb3)); in xnn_bf16_gemm_minmax_ukernel_1x4c8__neonfma_zip() local
|
| D | 1x4c8-minmax-neonfma-shland.c | 64 const float32x4_t vb3e = vreinterpretq_f32_u32(vshlq_n_u32(vreinterpretq_u32_u16(vb3), 16)); in xnn_bf16_gemm_minmax_ukernel_1x4c8__neonfma_shland() local 99 const float32x4_t vb3e = vreinterpretq_f32_u32(vshlq_n_u32(vreinterpretq_u32_u16(vb3), 16)); in xnn_bf16_gemm_minmax_ukernel_1x4c8__neonfma_shland() local
|
| D | 2x4c8-minmax-neonfma-shland.c | 76 const float32x4_t vb3e = vreinterpretq_f32_u32(vshlq_n_u32(vreinterpretq_u32_u16(vb3), 16)); in xnn_bf16_gemm_minmax_ukernel_2x4c8__neonfma_shland() local 121 const float32x4_t vb3e = vreinterpretq_f32_u32(vshlq_n_u32(vreinterpretq_u32_u16(vb3), 16)); in xnn_bf16_gemm_minmax_ukernel_2x4c8__neonfma_shland() local
|
| D | 2x4c8-minmax-neonfma-zip.c | 76 const float32x4_t vb3e = vreinterpretq_f32_u16(vzip1q_u16(vzero, vb3)); in xnn_bf16_gemm_minmax_ukernel_2x4c8__neonfma_zip() local 121 const float32x4_t vb3e = vreinterpretq_f32_u16(vzip1q_u16(vzero, vb3)); in xnn_bf16_gemm_minmax_ukernel_2x4c8__neonfma_zip() local
|
| D | 3x4c8-minmax-neonfma-zip.c | 88 const float32x4_t vb3e = vreinterpretq_f32_u16(vzip1q_u16(vzero, vb3)); in xnn_bf16_gemm_minmax_ukernel_3x4c8__neonfma_zip() local 143 const float32x4_t vb3e = vreinterpretq_f32_u16(vzip1q_u16(vzero, vb3)); in xnn_bf16_gemm_minmax_ukernel_3x4c8__neonfma_zip() local
|
| D | 3x4c8-minmax-neonfma-shland.c | 88 const float32x4_t vb3e = vreinterpretq_f32_u32(vshlq_n_u32(vreinterpretq_u32_u16(vb3), 16)); in xnn_bf16_gemm_minmax_ukernel_3x4c8__neonfma_shland() local 143 const float32x4_t vb3e = vreinterpretq_f32_u32(vshlq_n_u32(vreinterpretq_u32_u16(vb3), 16)); in xnn_bf16_gemm_minmax_ukernel_3x4c8__neonfma_shland() local
|
| D | 4x4c8-minmax-neonfma-shland.c | 100 const float32x4_t vb3e = vreinterpretq_f32_u32(vshlq_n_u32(vreinterpretq_u32_u16(vb3), 16)); in xnn_bf16_gemm_minmax_ukernel_4x4c8__neonfma_shland() local 165 const float32x4_t vb3e = vreinterpretq_f32_u32(vshlq_n_u32(vreinterpretq_u32_u16(vb3), 16)); in xnn_bf16_gemm_minmax_ukernel_4x4c8__neonfma_shland() local
|
| D | 4x4c8-minmax-neonfma-zip.c | 100 const float32x4_t vb3e = vreinterpretq_f32_u16(vzip1q_u16(vzero, vb3)); in xnn_bf16_gemm_minmax_ukernel_4x4c8__neonfma_zip() local 165 const float32x4_t vb3e = vreinterpretq_f32_u16(vzip1q_u16(vzero, vb3)); in xnn_bf16_gemm_minmax_ukernel_4x4c8__neonfma_zip() local
|
| D | 5x4c8-minmax-neonfma-zip.c | 112 const float32x4_t vb3e = vreinterpretq_f32_u16(vzip1q_u16(vzero, vb3)); in xnn_bf16_gemm_minmax_ukernel_5x4c8__neonfma_zip() local 187 const float32x4_t vb3e = vreinterpretq_f32_u16(vzip1q_u16(vzero, vb3)); in xnn_bf16_gemm_minmax_ukernel_5x4c8__neonfma_zip() local
|
| D | 5x4c8-minmax-neonfma-shland.c | 112 const float32x4_t vb3e = vreinterpretq_f32_u32(vshlq_n_u32(vreinterpretq_u32_u16(vb3), 16)); in xnn_bf16_gemm_minmax_ukernel_5x4c8__neonfma_shland() local 187 const float32x4_t vb3e = vreinterpretq_f32_u32(vshlq_n_u32(vreinterpretq_u32_u16(vb3), 16)); in xnn_bf16_gemm_minmax_ukernel_5x4c8__neonfma_shland() local
|