[FFmpeg-devel] Resubmit patch02 - [patch 4/4] Fix bug forPOWERLE: libswscale/ppc/swscale_altivec.c
Michael Niedermayer
michaelni at gmx.at
Tue Nov 11 19:31:29 CET 2014
[...]
> @@ -140,31 +180,15 @@ static void hScale_altivec_real(SwsContext *c, int16_t *dst, int dstW,
> for (i = 0; i < dstW; i++) {
> register int srcPos = filterPos[i];
>
> - vector unsigned char src_v0 = vec_ld(srcPos, src);
> - vector unsigned char src_v1, src_vF;
> + vector unsigned char src_vF = unaligned_load(srcPos, src);
> vector signed short src_v, filter_v;
> vector signed int val_vEven, val_s;
> - if ((((uintptr_t)src + srcPos) % 16) > 12) {
> - src_v1 = vec_ld(srcPos + 16, src);
> - }
> - src_vF = vec_perm(src_v0, src_v1, vec_lvsl(srcPos, src));
> -
> src_v = // vec_unpackh sign-extends...
> - (vector signed short)(vec_mergeh((vector unsigned char)vzero, src_vF));
> + (vector signed short)(VEC_MERGEH((vector unsigned char)vzero, src_vF));
> // now put our elements in the even slots
> src_v = vec_mergeh(src_v, (vector signed short)vzero);
> -
> - filter_v = vec_ld(i << 3, filter);
> - // The 3 above is 2 (filterSize == 4) + 1 (sizeof(short) == 2).
> -
> - // The neat trick: We only care for half the elements,
> - // high or low depending on (i<<3)%16 (it's 0 or 8 here),
> - // and we're going to use vec_mule, so we choose
> - // carefully how to "unpack" the elements into the even slots.
> - if ((i << 3) % 16)
> - filter_v = vec_mergel(filter_v, (vector signed short)vzero);
> - else
> - filter_v = vec_mergeh(filter_v, (vector signed short)vzero);
> + filter_v = (vector signed short)unaligned_load(i << 3, filter);
> + filter_v = vec_mergeh(filter_v, (vector signed short)vzero);
>
> val_vEven = vec_mule(src_v, filter_v);
> val_s = vec_sums(val_vEven, vzero);
i see 3 vec_ld() before and 2 unaligned_load() each 2 vec_ld()
afterwards on big endian
> @@ -172,25 +196,16 @@ static void hScale_altivec_real(SwsContext *c, int16_t *dst, int dstW,
> dst[i] = FFMIN(tempo[3] >> 7, (1 << 15) - 1);
> }
> break;
> -
> case 8:
> for (i = 0; i < dstW; i++) {
> register int srcPos = filterPos[i];
>
> - vector unsigned char src_v0 = vec_ld(srcPos, src);
> - vector unsigned char src_v1, src_vF;
> + vector unsigned char src_vF = unaligned_load(srcPos, src);
> vector signed short src_v, filter_v;
> vector signed int val_v, val_s;
> - if ((((uintptr_t)src + srcPos) % 16) > 8) {
> - src_v1 = vec_ld(srcPos + 16, src);
> - }
> - src_vF = vec_perm(src_v0, src_v1, vec_lvsl(srcPos, src));
> -
> src_v = // vec_unpackh sign-extends...
> - (vector signed short)(vec_mergeh((vector unsigned char)vzero, src_vF));
> + (vector signed short)(VEC_MERGEH((vector unsigned char)vzero, src_vF));
> filter_v = vec_ld(i << 4, filter);
> - // the 4 above is 3 (filterSize == 8) + 1 (sizeof(short) == 2)
> -
> val_v = vec_msums(src_v, filter_v, (vector signed int)vzero);
> val_s = vec_sums(val_v, vzero);
> vec_st(val_s, 0, tempo);
> @@ -202,25 +217,20 @@ static void hScale_altivec_real(SwsContext *c, int16_t *dst, int dstW,
> for (i = 0; i < dstW; i++) {
> register int srcPos = filterPos[i];
>
> - vector unsigned char src_v0 = vec_ld(srcPos, src);
> - vector unsigned char src_v1 = vec_ld(srcPos + 16, src);
> - vector unsigned char src_vF = vec_perm(src_v0, src_v1, vec_lvsl(srcPos, src));
> -
> + vector unsigned char src_vF = unaligned_load(srcPos, src);
> vector signed short src_vA = // vec_unpackh sign-extends...
> - (vector signed short)(vec_mergeh((vector unsigned char)vzero, src_vF));
> + (vector signed short)(VEC_MERGEH((vector unsigned char)vzero, src_vF));
> vector signed short src_vB = // vec_unpackh sign-extends...
> - (vector signed short)(vec_mergel((vector unsigned char)vzero, src_vF));
> -
> + (vector signed short)(VEC_MERGEL((vector unsigned char)vzero, src_vF));
> vector signed short filter_v0 = vec_ld(i << 5, filter);
> vector signed short filter_v1 = vec_ld((i << 5) + 16, filter);
> - // the 5 above are 4 (filterSize == 16) + 1 (sizeof(short) == 2)
>
> vector signed int val_acc = vec_msums(src_vA, filter_v0, (vector signed int)vzero);
> vector signed int val_v = vec_msums(src_vB, filter_v1, val_acc);
>
> vector signed int val_s = vec_sums(val_v, vzero);
>
> - vec_st(val_s, 0, tempo);
> + VEC_ST(val_s, 0, tempo);
> dst[i] = FFMIN(tempo[3] >> 7, (1 << 15) - 1);
> }
> break;
> @@ -231,56 +241,34 @@ static void hScale_altivec_real(SwsContext *c, int16_t *dst, int dstW,
> register int srcPos = filterPos[i];
>
> vector signed int val_s, val_v = (vector signed int)vzero;
> - vector signed short filter_v0R = vec_ld(i * 2 * filterSize, filter);
> - vector unsigned char permF = vec_lvsl((i * 2 * filterSize), filter);
> -
> - vector unsigned char src_v0 = vec_ld(srcPos, src);
> - vector unsigned char permS = vec_lvsl(srcPos, src);
>
> for (j = 0; j < filterSize - 15; j += 16) {
> - vector unsigned char src_v1 = vec_ld(srcPos + j + 16, src);
> - vector unsigned char src_vF = vec_perm(src_v0, src_v1, permS);
> -
> + vector unsigned char src_vF = unaligned_load(srcPos + j, src);
> vector signed short src_vA = // vec_unpackh sign-extends...
> - (vector signed short)(vec_mergeh((vector unsigned char)vzero, src_vF));
> + (vector signed short)(VEC_MERGEH((vector unsigned char)vzero, src_vF));
> vector signed short src_vB = // vec_unpackh sign-extends...
> - (vector signed short)(vec_mergel((vector unsigned char)vzero, src_vF));
> -
> - vector signed short filter_v1R = vec_ld((i * 2 * filterSize) + (j * 2) + 16, filter);
> - vector signed short filter_v2R = vec_ld((i * 2 * filterSize) + (j * 2) + 32, filter);
> - vector signed short filter_v0 = vec_perm(filter_v0R, filter_v1R, permF);
> - vector signed short filter_v1 = vec_perm(filter_v1R, filter_v2R, permF);
> + (vector signed short)(VEC_MERGEL((vector unsigned char)vzero, src_vF));
> + vector signed short filter_v0 = (vector signed short)unaligned_load((i * 2 * filterSize) + (j * 2), filter);
> + vector signed short filter_v1 = (vector signed short)unaligned_load((i * 2 * filterSize) + (j * 2) + 16, filter);
>
> vector signed int val_acc = vec_msums(src_vA, filter_v0, val_v);
> val_v = vec_msums(src_vB, filter_v1, val_acc);
> -
> - filter_v0R = filter_v2R;
> - src_v0 = src_v1;
> }
i count 3 vec_ld() before and 6 (2 for each unaligned_load())
after the change for big endian
[...]
--
Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB
Awnsering whenever a program halts or runs forever is
On a turing machine, in general impossible (turings halting problem).
On any real computer, always possible as a real computer has a finite number
of states N, and will either halt in less than N cycles or never halt.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 181 bytes
Desc: Digital signature
URL: <https://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20141111/b5ac9e7c/attachment.asc>
More information about the ffmpeg-devel
mailing list