In h264_deblock_q1, the result of the deblock needs to be kept to be used in future deblocks, so return this value now. Also change the sign of tc0 vector: It is really a signed value, so treat it as such until after the >=0 check; then, at that point, after being masked, it can be treated as unsigned. Patch by Graham Booker % gbooker A tamu P edu% Originally committed as revision 9349 to svn://svn.ffmpeg.org/ffmpeg/trunk
Unified · Split
| @@ -740,8 +740,8 @@ static inline vector unsigned char h264_deblock_mask ( register vector unsigned · 739 unmodified lines | ||
| 740 | 740 | return mask; |
| 741 | 741 | } |
| 742 | 742 | |
| 743 | −// out: p1 = clip((p2 + ((p0 + q0 + 1) >> 1)) >> 1, p1-tc0, p1+tc0) | |
| 744 | −static inline void h264_deblock_q1(register vector unsigned char p0, | |
| 743 | +// out: newp1 = clip((p2 + ((p0 + q0 + 1) >> 1)) >> 1, p1-tc0, p1+tc0) | |
| 744 | +static inline vector unsigned char h264_deblock_q1(register vector unsigned char p0, | |
| 745 | 745 | register vector unsigned char p1, |
| 746 | 746 | register vector unsigned char p2, |
| 747 | 747 | register vector unsigned char q0, |
| @@ -753,6 +753,7 @@ static inline void h264_deblock_q1(register vector unsigned char p0, · 5 unmodified lines | ||
| 753 | 753 | register vector unsigned char ones; |
| 754 | 754 | register vector unsigned char max; |
| 755 | 755 | register vector unsigned char min; |
| 756 | + register vector unsigned char newp1; | |
| 756 | 757 | |
| 757 | 758 | temp = vec_xor(average, p2); |
| 758 | 759 | average = vec_avg(average, p2); /*avg(p2, avg(p0, q0)) */ |
| @@ -761,8 +762,9 @@ static inline void h264_deblock_q1(register vector unsigned char p0, · 2 unmodified lines | ||
| 761 | 762 | uncliped = vec_subs(average, temp); /*(p2+((p0+q0+1)>>1))>>1 */ |
| 762 | 763 | max = vec_adds(p1, tc0); |
| 763 | 764 | min = vec_subs(p1, tc0); |
| 764 | − p1 = vec_max(min, uncliped); | |
| 765 | − p1 = vec_min(max, p1); | |
| 765 | + newp1 = vec_max(min, uncliped); | |
| 766 | + newp1 = vec_min(max, newp1); | |
| 767 | + return newp1; | |
| 766 | 768 | } |
| 767 | 769 | |
| 768 | 770 | #define h264_deblock_p0_q0(p0, p1, q0, q1, tc0masked) { \ |
| @@ -804,9 +806,11 @@ static inline void h264_deblock_q1(register vector unsigned char p0, · 35 unmodified lines | ||
| 804 | 806 | register vector unsigned char mask; \ |
| 805 | 807 | register vector unsigned char p1mask; \ |
| 806 | 808 | register vector unsigned char q1mask; \ |
| 807 | − register vector unsigned char tc0vec; \ | |
| 809 | + register vector char tc0vec; \ | |
| 808 | 810 | register vector unsigned char finaltc0; \ |
| 809 | 811 | register vector unsigned char tc0masked; \ |
| 812 | + register vector unsigned char newp1; \ | |
| 813 | + register vector unsigned char newq1; \ | |
| 810 | 814 | \ |
| 811 | 815 | temp[0] = alpha; \ |
| 812 | 816 | temp[1] = beta; \ |
| @@ -819,24 +823,26 @@ static inline void h264_deblock_q1(register vector unsigned char p0, · 6 unmodified lines | ||
| 819 | 823 | tc0vec = vec_ld(0, temp); \ |
| 820 | 824 | tc0vec = vec_mergeh(tc0vec, tc0vec); \ |
| 821 | 825 | tc0vec = vec_mergeh(tc0vec, tc0vec); \ |
| 822 | − mask = vec_and(mask, vec_cmpgt(tc0vec, vec_splat_u8(-1))); /* if tc0[i] >= 0 */ \ | |
| 823 | − finaltc0 = vec_and(tc0vec, mask); /*tc = tc0[i]*/ \ | |
| 826 | + mask = vec_and(mask, vec_cmpgt(tc0vec, vec_splat_s8(-1))); /* if tc0[i] >= 0 */ \ | |
| 827 | + finaltc0 = vec_and((vector unsigned char)tc0vec, mask); /* tc = tc0 */ \ | |
| 824 | 828 | \ |
| 825 | 829 | p1mask = diff_lt_altivec(p2, p0, betavec); \ |
| 826 | 830 | p1mask = vec_and(p1mask, mask); /* if( |p2 - p0| < beta) */ \ |
| 827 | 831 | tc0masked = vec_and(p1mask, tc0vec); \ |
| 828 | 832 | finaltc0 = vec_sub(finaltc0, p1mask); /* tc++ */ \ |
| 829 | − h264_deblock_q1(p0, p1, p2, q0, tc0masked); \ | |
| 833 | + newp1 = h264_deblock_q1(p0, p1, p2, q0, tc0masked); \ | |
| 830 | 834 | /*end if*/ \ |
| 831 | 835 | \ |
| 832 | 836 | q1mask = diff_lt_altivec(q2, q0, betavec); \ |
| 833 | 837 | q1mask = vec_and(q1mask, mask); /* if ( |q2 - q0| < beta ) */\ |
| 834 | 838 | tc0masked = vec_and(q1mask, tc0vec); \ |
| 835 | 839 | finaltc0 = vec_sub(finaltc0, q1mask); /* tc++ */ \ |
| 836 | − h264_deblock_q1(p0, q1, q2, q0, tc0masked); \ | |
| 840 | + newq1 = h264_deblock_q1(p0, q1, q2, q0, tc0masked); \ | |
| 837 | 841 | /*end if*/ \ |
| 838 | 842 | \ |
| 839 | 843 | h264_deblock_p0_q0(p0, p1, q0, q1, finaltc0); \ |
| 844 | + p1 = newp1; \ | |
| 845 | + q1 = newq1; \ | |
| 840 | 846 | } |
| 841 | 847 | |
| 842 | 848 | static void h264_v_loop_filter_luma_altivec(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0) { |