@@ -34,6 +34,8 @@ int ff_pix_abs16_y2_neon(MpegEncContext *v, uint8_t *pix1, uint8_t *pix2,
int sse16_neon(MpegEncContext *v, uint8_t *pix1, uint8_t *pix2,
ptrdiff_t stride, int h);
+int sse8_neon(MpegEncContext *v, uint8_t *pix1, uint8_t *pix2,
+ ptrdiff_t stride, int h);
int sse4_neon(MpegEncContext *v, uint8_t *pix1, uint8_t *pix2,
ptrdiff_t stride, int h);
@@ -49,6 +51,7 @@ av_cold void ff_me_cmp_init_aarch64(MECmpContext *c, AVCodecContext *avctx)
c->sad[0] = ff_pix_abs16_neon;
c->sse[0] = sse16_neon;
+ c->sse[1] = sse8_neon;
c->sse[2] = sse4_neon;
}
}
@@ -353,6 +353,78 @@ function sse16_neon, export=1
endfunc
+function sse8_neon, export=1
+ // x0 - unused
+ // x1 - pix1
+ // x2 - pix2
+ // x3 - stride
+ // w4 - h
+
+ movi d18, #0
+ cmp w4, #4
+ b.le 2f
+
+// make 4 iterations at once
+1:
+
+ // res = abs(pix1[0] - pix2[0])
+ // res * res
+
+ ld1 {v0.8b}, [x1], x3 // Load pix1 for first iteration
+ ld1 {v1.8b}, [x2], x3 // Load pix2 for second iteration
+ uabdl v30.8h, v0.8b, v1.8b // Absolute difference, first iteration
+ umull v21.4s, v30.4h, v30.4h // Multiply lower half, first iteration
+ ld1 {v2.8b}, [x1], x3 // Load pix1 for second iteration
+ umull2 v20.4s, v30.8h, v30.8h // Multiply upper half, second iteration
+ ld1 {v3.8b}, [x2], x3 // Load pix2 for second iteration
+ uabdl v29.8h, v2.8b, v3.8b // Absolute difference, second iteration
+ ld1 {v4.8b}, [x1], x3 // Load pix1 for third iteration
+ umlal v21.4s, v29.4h, v29.4h // Multiply lower half, second iteration
+ ld1 {v5.8b}, [x2], x3 // Load pix2 for third iteration
+ umlal2 v20.4s, v29.8h, v29.8h // Multiply upper half, second iteration
+ uabdl v28.8h, v4.8b, v5.8b // Absolute difference, third iteration
+ ld1 {v6.8b}, [x1], x3 // Load pix1 for fourth iteration
+ umlal v21.4s, v28.4h, v28.4h // Multiply lower half, third iteration
+ ld1 {v7.8b}, [x2], x3 // Load pix2 for fourth iteration
+ umlal2 v20.4s, v28.8h, v28.8h // Multiply upper half, third iteration
+ uabdl v27.8h, v6.8b, v7.8b // Absolute difference, fourth iteration
+ umlal v21.4s, v27.4h, v27.4h // Multiply lower half, fourth iteration
+ umlal2 v20.4s, v27.8h, v27.8h // Multiply upper ha;f, fourth iteration
+
+ add v21.4s, v21.4s, v20.4s // Add accumulator vectors together
+ sub w4, w4, #4 // h -= 4
+ uaddlv d17, v21.4s // Add up vector
+ add d18, d18, d17
+ cmp w4, #4
+ b.ge 1b
+
+ cbnz w4, 2f
+ fmov w0, s18
+
+ ret
+
+// iterate by one
+2:
+ ld1 {v0.8b}, [x1], x3 // Load pix1
+ ld1 {v1.8b}, [x2], x3 // Load pix2
+
+ uabdl v30.8h, v0.8b, v1.8b
+ umull v21.4s, v30.4h, v30.4h
+ umull2 v20.4s, v30.8h, v30.8h
+
+ subs w4, w4, #1
+
+ uaddlv d17, v21.4s
+ add d18, d18, d17
+ uaddlv d17, v20.4S
+ add d18, d18, d17
+
+ b.ne 2b
+ fmov w0, s18
+
+ ret
+endfunc
+
function sse4_neon, export=1
// x0 - unused
// x1 - pix1
Provide optimized implementation of sse8 function for arm64. Performance comparison tests are shown below. - sse_1_c: 133.0 - sse_1_neon: 36.7 Benchmarks and tests run with checkasm tool on AWS Graviton 3. Signed-off-by: Hubert Mazur <hum@semihalf.com> --- libavcodec/aarch64/me_cmp_init_aarch64.c | 3 + libavcodec/aarch64/me_cmp_neon.S | 72 ++++++++++++++++++++++++ 2 files changed, 75 insertions(+)