mirror of https://git.ffmpeg.org/ffmpeg.git
aarch64: me_cmp: Avoid using the non-unrolled codepath for the minimum unroll size
Signed-off-by: Martin Storsjö <martin@martin.st>
This commit is contained in:
parent
6f2ad7f951
commit
8089fe072e
|
@ -471,7 +471,7 @@ function sse8_neon, export=1
|
||||||
movi v21.4s, #0
|
movi v21.4s, #0
|
||||||
movi v20.4s, #0
|
movi v20.4s, #0
|
||||||
cmp w4, #4
|
cmp w4, #4
|
||||||
b.le 2f
|
b.lt 2f
|
||||||
|
|
||||||
// make 4 iterations at once
|
// make 4 iterations at once
|
||||||
1:
|
1:
|
||||||
|
@ -534,7 +534,7 @@ function sse4_neon, export=1
|
||||||
|
|
||||||
movi v16.4s, #0 // clear the result accumulator
|
movi v16.4s, #0 // clear the result accumulator
|
||||||
cmp w4, #4
|
cmp w4, #4
|
||||||
b.le 2f
|
b.lt 2f
|
||||||
|
|
||||||
// make 4 iterations at once
|
// make 4 iterations at once
|
||||||
1:
|
1:
|
||||||
|
@ -663,7 +663,7 @@ function vsse16_neon, export=1
|
||||||
cmp w4, #3 // check if we can make 3 iterations at once
|
cmp w4, #3 // check if we can make 3 iterations at once
|
||||||
usubl v31.8h, v0.8b, v1.8b // Signed difference of pix1[0] - pix2[0], first iteration
|
usubl v31.8h, v0.8b, v1.8b // Signed difference of pix1[0] - pix2[0], first iteration
|
||||||
usubl2 v30.8h, v0.16b, v1.16b // Signed difference of pix1[0] - pix2[0], first iteration
|
usubl2 v30.8h, v0.16b, v1.16b // Signed difference of pix1[0] - pix2[0], first iteration
|
||||||
b.le 2f
|
b.lt 2f
|
||||||
|
|
||||||
|
|
||||||
1:
|
1:
|
||||||
|
|
Loading…
Reference in New Issue