/external/XNNPACK/src/bf16-gemm/gen/ |
D | 1x4c8-minmax-neonfma-zip.c | 62 const float32x4_t vb1e = vreinterpretq_f32_u16(vzip1q_u16(vzero, vb1)); in xnn_bf16_gemm_minmax_ukernel_1x4c8__neonfma_zip() local 97 const float32x4_t vb1e = vreinterpretq_f32_u16(vzip1q_u16(vzero, vb1)); in xnn_bf16_gemm_minmax_ukernel_1x4c8__neonfma_zip() local
|
D | 1x4c8-minmax-neonfma-shland.c | 62 const float32x4_t vb1e = vreinterpretq_f32_u32(vshlq_n_u32(vreinterpretq_u32_u16(vb1), 16)); in xnn_bf16_gemm_minmax_ukernel_1x4c8__neonfma_shland() local 97 const float32x4_t vb1e = vreinterpretq_f32_u32(vshlq_n_u32(vreinterpretq_u32_u16(vb1), 16)); in xnn_bf16_gemm_minmax_ukernel_1x4c8__neonfma_shland() local
|
D | 2x4c8-minmax-neonfma-shland.c | 74 const float32x4_t vb1e = vreinterpretq_f32_u32(vshlq_n_u32(vreinterpretq_u32_u16(vb1), 16)); in xnn_bf16_gemm_minmax_ukernel_2x4c8__neonfma_shland() local 119 const float32x4_t vb1e = vreinterpretq_f32_u32(vshlq_n_u32(vreinterpretq_u32_u16(vb1), 16)); in xnn_bf16_gemm_minmax_ukernel_2x4c8__neonfma_shland() local
|
D | 2x4c8-minmax-neonfma-zip.c | 74 const float32x4_t vb1e = vreinterpretq_f32_u16(vzip1q_u16(vzero, vb1)); in xnn_bf16_gemm_minmax_ukernel_2x4c8__neonfma_zip() local 119 const float32x4_t vb1e = vreinterpretq_f32_u16(vzip1q_u16(vzero, vb1)); in xnn_bf16_gemm_minmax_ukernel_2x4c8__neonfma_zip() local
|
D | 3x4c8-minmax-neonfma-shland.c | 86 const float32x4_t vb1e = vreinterpretq_f32_u32(vshlq_n_u32(vreinterpretq_u32_u16(vb1), 16)); in xnn_bf16_gemm_minmax_ukernel_3x4c8__neonfma_shland() local 141 const float32x4_t vb1e = vreinterpretq_f32_u32(vshlq_n_u32(vreinterpretq_u32_u16(vb1), 16)); in xnn_bf16_gemm_minmax_ukernel_3x4c8__neonfma_shland() local
|
D | 3x4c8-minmax-neonfma-zip.c | 86 const float32x4_t vb1e = vreinterpretq_f32_u16(vzip1q_u16(vzero, vb1)); in xnn_bf16_gemm_minmax_ukernel_3x4c8__neonfma_zip() local 141 const float32x4_t vb1e = vreinterpretq_f32_u16(vzip1q_u16(vzero, vb1)); in xnn_bf16_gemm_minmax_ukernel_3x4c8__neonfma_zip() local
|
D | 4x4c8-minmax-neonfma-zip.c | 98 const float32x4_t vb1e = vreinterpretq_f32_u16(vzip1q_u16(vzero, vb1)); in xnn_bf16_gemm_minmax_ukernel_4x4c8__neonfma_zip() local 163 const float32x4_t vb1e = vreinterpretq_f32_u16(vzip1q_u16(vzero, vb1)); in xnn_bf16_gemm_minmax_ukernel_4x4c8__neonfma_zip() local
|
D | 4x4c8-minmax-neonfma-shland.c | 98 const float32x4_t vb1e = vreinterpretq_f32_u32(vshlq_n_u32(vreinterpretq_u32_u16(vb1), 16)); in xnn_bf16_gemm_minmax_ukernel_4x4c8__neonfma_shland() local 163 const float32x4_t vb1e = vreinterpretq_f32_u32(vshlq_n_u32(vreinterpretq_u32_u16(vb1), 16)); in xnn_bf16_gemm_minmax_ukernel_4x4c8__neonfma_shland() local
|
D | 5x4c8-minmax-neonfma-zip.c | 110 const float32x4_t vb1e = vreinterpretq_f32_u16(vzip1q_u16(vzero, vb1)); in xnn_bf16_gemm_minmax_ukernel_5x4c8__neonfma_zip() local 185 const float32x4_t vb1e = vreinterpretq_f32_u16(vzip1q_u16(vzero, vb1)); in xnn_bf16_gemm_minmax_ukernel_5x4c8__neonfma_zip() local
|
D | 5x4c8-minmax-neonfma-shland.c | 110 const float32x4_t vb1e = vreinterpretq_f32_u32(vshlq_n_u32(vreinterpretq_u32_u16(vb1), 16)); in xnn_bf16_gemm_minmax_ukernel_5x4c8__neonfma_shland() local 185 const float32x4_t vb1e = vreinterpretq_f32_u32(vshlq_n_u32(vreinterpretq_u32_u16(vb1), 16)); in xnn_bf16_gemm_minmax_ukernel_5x4c8__neonfma_shland() local
|