Could you try to analyze the Dynamic Range, I worry about the intermedia result more than 16 bits.
At 2015-09-28 13:09:33,[email protected] wrote: ># HG changeset patch ># User Dnyaneshwar G <[email protected]> ># Date 1443416892 -19800 ># Mon Sep 28 10:38:12 2015 +0530 ># Node ID 21ab6be50abfb116256093a4d7e2bdb2783bc93d ># Parent 4f524673358748837d281abf28e4382f2ee845d2 >main12: fix intra_pred_planar8x8 overflow > >diff -r 4f5246733587 -r 21ab6be50abf source/common/x86/asm-primitives.cpp >--- a/source/common/x86/asm-primitives.cpp Thu Sep 24 18:53:28 2015 -0500 >+++ b/source/common/x86/asm-primitives.cpp Mon Sep 28 10:38:12 2015 +0530 >@@ -960,8 +960,12 @@ > ALL_LUMA_TU_S(calcresidual, getResidual, sse2); > ALL_LUMA_TU_S(transpose, transpose, sse2); > >+ p.cu[BLOCK_4x4].intra_pred[PLANAR_IDX] = PFX(intra_pred_planar4_sse2); >+ p.cu[BLOCK_8x8].intra_pred[PLANAR_IDX] = PFX(intra_pred_planar8_sse2); >+ > #if X265_DEPTH <= 10 >- ALL_LUMA_TU_S(intra_pred[PLANAR_IDX], intra_pred_planar, sse2); >+ p.cu[BLOCK_16x16].intra_pred[PLANAR_IDX] = >PFX(intra_pred_planar16_sse2); >+ p.cu[BLOCK_32x32].intra_pred[PLANAR_IDX] = >PFX(intra_pred_planar32_sse2); > #endif /* X265_DEPTH <= 10 */ > ALL_LUMA_TU_S(intra_pred[DC_IDX], intra_pred_dc, sse2); > >@@ -1137,8 +1141,12 @@ > p.cu[BLOCK_16x16].intra_filter = PFX(intra_filter_16x16_sse4); > p.cu[BLOCK_32x32].intra_filter = PFX(intra_filter_32x32_sse4); > >+ p.cu[BLOCK_4x4].intra_pred[PLANAR_IDX] = PFX(intra_pred_planar4_sse4); >+ p.cu[BLOCK_8x8].intra_pred[PLANAR_IDX] = PFX(intra_pred_planar8_sse4); >+ > #if X265_DEPTH <= 10 >- ALL_LUMA_TU_S(intra_pred[PLANAR_IDX], intra_pred_planar, sse4); >+ p.cu[BLOCK_16x16].intra_pred[PLANAR_IDX] = >PFX(intra_pred_planar16_sse4); >+ p.cu[BLOCK_32x32].intra_pred[PLANAR_IDX] = >PFX(intra_pred_planar32_sse4); > #endif > ALL_LUMA_TU_S(intra_pred[DC_IDX], intra_pred_dc, sse4); > INTRA_ANG_SSE4_COMMON(sse4); >diff -r 4f5246733587 -r 21ab6be50abf source/common/x86/intrapred16.asm >--- a/source/common/x86/intrapred16.asm Thu Sep 24 18:53:28 2015 -0500 >+++ b/source/common/x86/intrapred16.asm Mon Sep 28 10:38:12 2015 +0530 >@@ -709,7 +709,7 @@ > %endif > pmullw m1, [pw_planar16_mul + mmsize] > paddw m1, m3 >- psraw m1, 4 >+ psrlw m1, 4 > movu [r0], m1 > %if (%1 < 7) > paddw m3, m4 >@@ -2197,7 +2197,7 @@ > pmullw m1, m0 > paddw m1, m3 > paddw m3, m4 >- psraw m1, 4 >+ psrlw m1, 4 > movu [r0], m1 > lea r0, [r0 + r1] > %endmacro >_______________________________________________ >x265-devel mailing list >[email protected] >https://mailman.videolan.org/listinfo/x265-devel
_______________________________________________ x265-devel mailing list [email protected] https://mailman.videolan.org/listinfo/x265-devel
