Home
last modified time | relevance | path

Searched refs:vout0x01234567_1x01234567 (Results 1 – 4 of 4) sorted by relevance

/external/XNNPACK/src/q8-igemm/
D4x8-neon.c357 …uint8x16_t vout0x01234567_1x01234567 = vqmovun_high_s16(vqmovun_s16(vacc0x01234567), vacc1x0123456… in xnn_q8_igemm_ukernel_4x8__neon() local
365 …uint8x16_t vout0x01234567_1x01234567 = vcombine_u8(vqmovun_s16(vacc0x01234567), vqmovun_s16(vacc1x… in xnn_q8_igemm_ukernel_4x8__neon() local
371 vout0x01234567_1x01234567 = vmaxq_u8(vout0x01234567_1x01234567, voutput_min); in xnn_q8_igemm_ukernel_4x8__neon()
373 vout0x01234567_1x01234567 = vminq_u8(vout0x01234567_1x01234567, voutput_max); in xnn_q8_igemm_ukernel_4x8__neon()
379 vst1_u8(c1, vget_high_u8(vout0x01234567_1x01234567)); c1 += cn_stride; in xnn_q8_igemm_ukernel_4x8__neon()
380 vst1_u8(c0, vget_low_u8(vout0x01234567_1x01234567)); c0 += cn_stride; in xnn_q8_igemm_ukernel_4x8__neon()
388 …_lane_u32(__builtin_assume_aligned(c1, 1), vreinterpretq_u32_u8(vout0x01234567_1x01234567), 2); c1… in xnn_q8_igemm_ukernel_4x8__neon()
389 …_lane_u32(__builtin_assume_aligned(c0, 1), vreinterpretq_u32_u8(vout0x01234567_1x01234567), 0); c0… in xnn_q8_igemm_ukernel_4x8__neon()
390 … vout0x01234567_1x01234567 = vextq_u8(vout0x01234567_1x01234567, vout0x01234567_1x01234567, 4); in xnn_q8_igemm_ukernel_4x8__neon()
396 …_lane_u16(__builtin_assume_aligned(c1, 1), vreinterpretq_u16_u8(vout0x01234567_1x01234567), 4); c1… in xnn_q8_igemm_ukernel_4x8__neon()
[all …]
D8x8-neon.c561 …uint8x16_t vout0x01234567_1x01234567 = vqmovun_high_s16(vqmovun_s16(vacc0x01234567), vacc1x0123456… in xnn_q8_igemm_ukernel_8x8__neon() local
583 …uint8x16_t vout0x01234567_1x01234567 = vcombine_u8(vqmovun_s16(vacc0x01234567), vqmovun_s16(vacc1x… in xnn_q8_igemm_ukernel_8x8__neon() local
591 vout0x01234567_1x01234567 = vmaxq_u8(vout0x01234567_1x01234567, voutput_min); in xnn_q8_igemm_ukernel_8x8__neon()
595 vout0x01234567_1x01234567 = vminq_u8(vout0x01234567_1x01234567, voutput_max); in xnn_q8_igemm_ukernel_8x8__neon()
607 vst1_u8(c1, vget_high_u8(vout0x01234567_1x01234567)); c1 += cn_stride; in xnn_q8_igemm_ukernel_8x8__neon()
608 vst1_u8(c0, vget_low_u8(vout0x01234567_1x01234567)); c0 += cn_stride; in xnn_q8_igemm_ukernel_8x8__neon()
621 …_lane_u32(__builtin_assume_aligned(c1, 1), vreinterpretq_u32_u8(vout0x01234567_1x01234567), 2); c1… in xnn_q8_igemm_ukernel_8x8__neon()
622 …_lane_u32(__builtin_assume_aligned(c0, 1), vreinterpretq_u32_u8(vout0x01234567_1x01234567), 0); c0… in xnn_q8_igemm_ukernel_8x8__neon()
626 … vout0x01234567_1x01234567 = vextq_u8(vout0x01234567_1x01234567, vout0x01234567_1x01234567, 4); in xnn_q8_igemm_ukernel_8x8__neon()
635 …_lane_u16(__builtin_assume_aligned(c1, 1), vreinterpretq_u16_u8(vout0x01234567_1x01234567), 4); c1… in xnn_q8_igemm_ukernel_8x8__neon()
[all …]
/external/XNNPACK/src/q8-gemm/
D4x8-neon.c320 …uint8x16_t vout0x01234567_1x01234567 = vqmovun_high_s16(vqmovun_s16(vacc0x01234567), vacc1x0123456… in xnn_q8_gemm_ukernel_4x8__neon() local
328 …uint8x16_t vout0x01234567_1x01234567 = vcombine_u8(vqmovun_s16(vacc0x01234567), vqmovun_s16(vacc1x… in xnn_q8_gemm_ukernel_4x8__neon() local
334 vout0x01234567_1x01234567 = vmaxq_u8(vout0x01234567_1x01234567, voutput_min); in xnn_q8_gemm_ukernel_4x8__neon()
336 vout0x01234567_1x01234567 = vminq_u8(vout0x01234567_1x01234567, voutput_max); in xnn_q8_gemm_ukernel_4x8__neon()
340 … vst1_u8(c0, vget_low_u8(vout0x01234567_1x01234567)); c0 = (uint8_t*) ((uintptr_t) c0 + cn_stride); in xnn_q8_gemm_ukernel_4x8__neon()
341 …vst1_u8(c1, vget_high_u8(vout0x01234567_1x01234567)); c1 = (uint8_t*) ((uintptr_t) c1 + cn_stride); in xnn_q8_gemm_ukernel_4x8__neon()
353 …_lane_u32(__builtin_assume_aligned(c0, 1), vreinterpretq_u32_u8(vout0x01234567_1x01234567), 0); c0… in xnn_q8_gemm_ukernel_4x8__neon()
354 …_lane_u32(__builtin_assume_aligned(c1, 1), vreinterpretq_u32_u8(vout0x01234567_1x01234567), 2); c1… in xnn_q8_gemm_ukernel_4x8__neon()
357 … vout0x01234567_1x01234567 = vextq_u8(vout0x01234567_1x01234567, vout0x01234567_1x01234567, 4); in xnn_q8_gemm_ukernel_4x8__neon()
361 …_lane_u16(__builtin_assume_aligned(c0, 1), vreinterpretq_u16_u8(vout0x01234567_1x01234567), 0); c0… in xnn_q8_gemm_ukernel_4x8__neon()
[all …]
D8x8-neon.c516 …uint8x16_t vout0x01234567_1x01234567 = vqmovun_high_s16(vqmovun_s16(vacc0x01234567), vacc1x0123456… in xnn_q8_gemm_ukernel_8x8__neon() local
538 …uint8x16_t vout0x01234567_1x01234567 = vcombine_u8(vqmovun_s16(vacc0x01234567), vqmovun_s16(vacc1x… in xnn_q8_gemm_ukernel_8x8__neon() local
546 vout0x01234567_1x01234567 = vmaxq_u8(vout0x01234567_1x01234567, voutput_min); in xnn_q8_gemm_ukernel_8x8__neon()
550 vout0x01234567_1x01234567 = vminq_u8(vout0x01234567_1x01234567, voutput_max); in xnn_q8_gemm_ukernel_8x8__neon()
556 … vst1_u8(c0, vget_low_u8(vout0x01234567_1x01234567)); c0 = (uint8_t*) ((uintptr_t) c0 + cn_stride); in xnn_q8_gemm_ukernel_8x8__neon()
557 …vst1_u8(c1, vget_high_u8(vout0x01234567_1x01234567)); c1 = (uint8_t*) ((uintptr_t) c1 + cn_stride); in xnn_q8_gemm_ukernel_8x8__neon()
577 …_lane_u32(__builtin_assume_aligned(c0, 1), vreinterpretq_u32_u8(vout0x01234567_1x01234567), 0); c0… in xnn_q8_gemm_ukernel_8x8__neon()
578 …_lane_u32(__builtin_assume_aligned(c1, 1), vreinterpretq_u32_u8(vout0x01234567_1x01234567), 2); c1… in xnn_q8_gemm_ukernel_8x8__neon()
585 … vout0x01234567_1x01234567 = vextq_u8(vout0x01234567_1x01234567, vout0x01234567_1x01234567, 4); in xnn_q8_gemm_ukernel_8x8__neon()
591 …_lane_u16(__builtin_assume_aligned(c0, 1), vreinterpretq_u16_u8(vout0x01234567_1x01234567), 0); c0… in xnn_q8_gemm_ukernel_8x8__neon()
[all …]