summaryrefslogtreecommitdiffstats
path: root/third_party/aom/aom_dsp/arm/intrapred_neon.c
diff options
context:
space:
mode:
Diffstat (limited to 'third_party/aom/aom_dsp/arm/intrapred_neon.c')
-rw-r--r--third_party/aom/aom_dsp/arm/intrapred_neon.c225
1 files changed, 0 insertions, 225 deletions
diff --git a/third_party/aom/aom_dsp/arm/intrapred_neon.c b/third_party/aom/aom_dsp/arm/intrapred_neon.c
index 2dc5b2e56..7d5f64004 100644
--- a/third_party/aom/aom_dsp/arm/intrapred_neon.c
+++ b/third_party/aom/aom_dsp/arm/intrapred_neon.c
@@ -529,229 +529,4 @@ void aom_h_predictor_32x32_neon(uint8_t *dst, ptrdiff_t stride,
}
}
}
-
-void aom_tm_predictor_4x4_neon(uint8_t *dst, ptrdiff_t stride,
- const uint8_t *above, const uint8_t *left) {
- int i;
- uint16x8_t q1u16, q3u16;
- int16x8_t q1s16;
- uint8x8_t d0u8 = vdup_n_u8(0);
- uint32x2_t d2u32 = vdup_n_u32(0);
-
- d0u8 = vld1_dup_u8(above - 1);
- d2u32 = vld1_lane_u32((const uint32_t *)above, d2u32, 0);
- q3u16 = vsubl_u8(vreinterpret_u8_u32(d2u32), d0u8);
- for (i = 0; i < 4; i++, dst += stride) {
- q1u16 = vdupq_n_u16((uint16_t)left[i]);
- q1s16 =
- vaddq_s16(vreinterpretq_s16_u16(q1u16), vreinterpretq_s16_u16(q3u16));
- d0u8 = vqmovun_s16(q1s16);
- vst1_lane_u32((uint32_t *)dst, vreinterpret_u32_u8(d0u8), 0);
- }
-}
-
-void aom_tm_predictor_8x8_neon(uint8_t *dst, ptrdiff_t stride,
- const uint8_t *above, const uint8_t *left) {
- int j;
- uint16x8_t q0u16, q3u16, q10u16;
- int16x8_t q0s16;
- uint16x4_t d20u16;
- uint8x8_t d0u8, d2u8, d30u8;
-
- d0u8 = vld1_dup_u8(above - 1);
- d30u8 = vld1_u8(left);
- d2u8 = vld1_u8(above);
- q10u16 = vmovl_u8(d30u8);
- q3u16 = vsubl_u8(d2u8, d0u8);
- d20u16 = vget_low_u16(q10u16);
- for (j = 0; j < 2; j++, d20u16 = vget_high_u16(q10u16)) {
- q0u16 = vdupq_lane_u16(d20u16, 0);
- q0s16 =
- vaddq_s16(vreinterpretq_s16_u16(q3u16), vreinterpretq_s16_u16(q0u16));
- d0u8 = vqmovun_s16(q0s16);
- vst1_u64((uint64_t *)dst, vreinterpret_u64_u8(d0u8));
- dst += stride;
- q0u16 = vdupq_lane_u16(d20u16, 1);
- q0s16 =
- vaddq_s16(vreinterpretq_s16_u16(q3u16), vreinterpretq_s16_u16(q0u16));
- d0u8 = vqmovun_s16(q0s16);
- vst1_u64((uint64_t *)dst, vreinterpret_u64_u8(d0u8));
- dst += stride;
- q0u16 = vdupq_lane_u16(d20u16, 2);
- q0s16 =
- vaddq_s16(vreinterpretq_s16_u16(q3u16), vreinterpretq_s16_u16(q0u16));
- d0u8 = vqmovun_s16(q0s16);
- vst1_u64((uint64_t *)dst, vreinterpret_u64_u8(d0u8));
- dst += stride;
- q0u16 = vdupq_lane_u16(d20u16, 3);
- q0s16 =
- vaddq_s16(vreinterpretq_s16_u16(q3u16), vreinterpretq_s16_u16(q0u16));
- d0u8 = vqmovun_s16(q0s16);
- vst1_u64((uint64_t *)dst, vreinterpret_u64_u8(d0u8));
- dst += stride;
- }
-}
-
-void aom_tm_predictor_16x16_neon(uint8_t *dst, ptrdiff_t stride,
- const uint8_t *above, const uint8_t *left) {
- int j, k;
- uint16x8_t q0u16, q2u16, q3u16, q8u16, q10u16;
- uint8x16_t q0u8, q1u8;
- int16x8_t q0s16, q1s16, q8s16, q11s16;
- uint16x4_t d20u16;
- uint8x8_t d2u8, d3u8, d18u8, d22u8, d23u8;
-
- q0u8 = vld1q_dup_u8(above - 1);
- q1u8 = vld1q_u8(above);
- q2u16 = vsubl_u8(vget_low_u8(q1u8), vget_low_u8(q0u8));
- q3u16 = vsubl_u8(vget_high_u8(q1u8), vget_high_u8(q0u8));
- for (k = 0; k < 2; k++, left += 8) {
- d18u8 = vld1_u8(left);
- q10u16 = vmovl_u8(d18u8);
- d20u16 = vget_low_u16(q10u16);
- for (j = 0; j < 2; j++, d20u16 = vget_high_u16(q10u16)) {
- q0u16 = vdupq_lane_u16(d20u16, 0);
- q8u16 = vdupq_lane_u16(d20u16, 1);
- q1s16 =
- vaddq_s16(vreinterpretq_s16_u16(q0u16), vreinterpretq_s16_u16(q2u16));
- q0s16 =
- vaddq_s16(vreinterpretq_s16_u16(q0u16), vreinterpretq_s16_u16(q3u16));
- q11s16 =
- vaddq_s16(vreinterpretq_s16_u16(q8u16), vreinterpretq_s16_u16(q2u16));
- q8s16 =
- vaddq_s16(vreinterpretq_s16_u16(q8u16), vreinterpretq_s16_u16(q3u16));
- d2u8 = vqmovun_s16(q1s16);
- d3u8 = vqmovun_s16(q0s16);
- d22u8 = vqmovun_s16(q11s16);
- d23u8 = vqmovun_s16(q8s16);
- vst1_u64((uint64_t *)dst, vreinterpret_u64_u8(d2u8));
- vst1_u64((uint64_t *)(dst + 8), vreinterpret_u64_u8(d3u8));
- dst += stride;
- vst1_u64((uint64_t *)dst, vreinterpret_u64_u8(d22u8));
- vst1_u64((uint64_t *)(dst + 8), vreinterpret_u64_u8(d23u8));
- dst += stride;
-
- q0u16 = vdupq_lane_u16(d20u16, 2);
- q8u16 = vdupq_lane_u16(d20u16, 3);
- q1s16 =
- vaddq_s16(vreinterpretq_s16_u16(q0u16), vreinterpretq_s16_u16(q2u16));
- q0s16 =
- vaddq_s16(vreinterpretq_s16_u16(q0u16), vreinterpretq_s16_u16(q3u16));
- q11s16 =
- vaddq_s16(vreinterpretq_s16_u16(q8u16), vreinterpretq_s16_u16(q2u16));
- q8s16 =
- vaddq_s16(vreinterpretq_s16_u16(q8u16), vreinterpretq_s16_u16(q3u16));
- d2u8 = vqmovun_s16(q1s16);
- d3u8 = vqmovun_s16(q0s16);
- d22u8 = vqmovun_s16(q11s16);
- d23u8 = vqmovun_s16(q8s16);
- vst1_u64((uint64_t *)dst, vreinterpret_u64_u8(d2u8));
- vst1_u64((uint64_t *)(dst + 8), vreinterpret_u64_u8(d3u8));
- dst += stride;
- vst1_u64((uint64_t *)dst, vreinterpret_u64_u8(d22u8));
- vst1_u64((uint64_t *)(dst + 8), vreinterpret_u64_u8(d23u8));
- dst += stride;
- }
- }
-}
-
-void aom_tm_predictor_32x32_neon(uint8_t *dst, ptrdiff_t stride,
- const uint8_t *above, const uint8_t *left) {
- int j, k;
- uint16x8_t q0u16, q3u16, q8u16, q9u16, q10u16, q11u16;
- uint8x16_t q0u8, q1u8, q2u8;
- int16x8_t q12s16, q13s16, q14s16, q15s16;
- uint16x4_t d6u16;
- uint8x8_t d0u8, d1u8, d2u8, d3u8, d26u8;
-
- q0u8 = vld1q_dup_u8(above - 1);
- q1u8 = vld1q_u8(above);
- q2u8 = vld1q_u8(above + 16);
- q8u16 = vsubl_u8(vget_low_u8(q1u8), vget_low_u8(q0u8));
- q9u16 = vsubl_u8(vget_high_u8(q1u8), vget_high_u8(q0u8));
- q10u16 = vsubl_u8(vget_low_u8(q2u8), vget_low_u8(q0u8));
- q11u16 = vsubl_u8(vget_high_u8(q2u8), vget_high_u8(q0u8));
- for (k = 0; k < 4; k++, left += 8) {
- d26u8 = vld1_u8(left);
- q3u16 = vmovl_u8(d26u8);
- d6u16 = vget_low_u16(q3u16);
- for (j = 0; j < 2; j++, d6u16 = vget_high_u16(q3u16)) {
- q0u16 = vdupq_lane_u16(d6u16, 0);
- q12s16 =
- vaddq_s16(vreinterpretq_s16_u16(q0u16), vreinterpretq_s16_u16(q8u16));
- q13s16 =
- vaddq_s16(vreinterpretq_s16_u16(q0u16), vreinterpretq_s16_u16(q9u16));
- q14s16 = vaddq_s16(vreinterpretq_s16_u16(q0u16),
- vreinterpretq_s16_u16(q10u16));
- q15s16 = vaddq_s16(vreinterpretq_s16_u16(q0u16),
- vreinterpretq_s16_u16(q11u16));
- d0u8 = vqmovun_s16(q12s16);
- d1u8 = vqmovun_s16(q13s16);
- d2u8 = vqmovun_s16(q14s16);
- d3u8 = vqmovun_s16(q15s16);
- q0u8 = vcombine_u8(d0u8, d1u8);
- q1u8 = vcombine_u8(d2u8, d3u8);
- vst1q_u64((uint64_t *)dst, vreinterpretq_u64_u8(q0u8));
- vst1q_u64((uint64_t *)(dst + 16), vreinterpretq_u64_u8(q1u8));
- dst += stride;
-
- q0u16 = vdupq_lane_u16(d6u16, 1);
- q12s16 =
- vaddq_s16(vreinterpretq_s16_u16(q0u16), vreinterpretq_s16_u16(q8u16));
- q13s16 =
- vaddq_s16(vreinterpretq_s16_u16(q0u16), vreinterpretq_s16_u16(q9u16));
- q14s16 = vaddq_s16(vreinterpretq_s16_u16(q0u16),
- vreinterpretq_s16_u16(q10u16));
- q15s16 = vaddq_s16(vreinterpretq_s16_u16(q0u16),
- vreinterpretq_s16_u16(q11u16));
- d0u8 = vqmovun_s16(q12s16);
- d1u8 = vqmovun_s16(q13s16);
- d2u8 = vqmovun_s16(q14s16);
- d3u8 = vqmovun_s16(q15s16);
- q0u8 = vcombine_u8(d0u8, d1u8);
- q1u8 = vcombine_u8(d2u8, d3u8);
- vst1q_u64((uint64_t *)dst, vreinterpretq_u64_u8(q0u8));
- vst1q_u64((uint64_t *)(dst + 16), vreinterpretq_u64_u8(q1u8));
- dst += stride;
-
- q0u16 = vdupq_lane_u16(d6u16, 2);
- q12s16 =
- vaddq_s16(vreinterpretq_s16_u16(q0u16), vreinterpretq_s16_u16(q8u16));
- q13s16 =
- vaddq_s16(vreinterpretq_s16_u16(q0u16), vreinterpretq_s16_u16(q9u16));
- q14s16 = vaddq_s16(vreinterpretq_s16_u16(q0u16),
- vreinterpretq_s16_u16(q10u16));
- q15s16 = vaddq_s16(vreinterpretq_s16_u16(q0u16),
- vreinterpretq_s16_u16(q11u16));
- d0u8 = vqmovun_s16(q12s16);
- d1u8 = vqmovun_s16(q13s16);
- d2u8 = vqmovun_s16(q14s16);
- d3u8 = vqmovun_s16(q15s16);
- q0u8 = vcombine_u8(d0u8, d1u8);
- q1u8 = vcombine_u8(d2u8, d3u8);
- vst1q_u64((uint64_t *)dst, vreinterpretq_u64_u8(q0u8));
- vst1q_u64((uint64_t *)(dst + 16), vreinterpretq_u64_u8(q1u8));
- dst += stride;
-
- q0u16 = vdupq_lane_u16(d6u16, 3);
- q12s16 =
- vaddq_s16(vreinterpretq_s16_u16(q0u16), vreinterpretq_s16_u16(q8u16));
- q13s16 =
- vaddq_s16(vreinterpretq_s16_u16(q0u16), vreinterpretq_s16_u16(q9u16));
- q14s16 = vaddq_s16(vreinterpretq_s16_u16(q0u16),
- vreinterpretq_s16_u16(q10u16));
- q15s16 = vaddq_s16(vreinterpretq_s16_u16(q0u16),
- vreinterpretq_s16_u16(q11u16));
- d0u8 = vqmovun_s16(q12s16);
- d1u8 = vqmovun_s16(q13s16);
- d2u8 = vqmovun_s16(q14s16);
- d3u8 = vqmovun_s16(q15s16);
- q0u8 = vcombine_u8(d0u8, d1u8);
- q1u8 = vcombine_u8(d2u8, d3u8);
- vst1q_u64((uint64_t *)dst, vreinterpretq_u64_u8(q0u8));
- vst1q_u64((uint64_t *)(dst + 16), vreinterpretq_u64_u8(q1u8));
- dst += stride;
- }
- }
-}
#endif // !HAVE_NEON_ASM
generated by cgit v1.2.3 (git 2.25.1) at 2024-11-23 16:47:28 +0000