From 00e4195fec8e0bd41a75af927ca9569e52bbccd9 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Sebastian=20Dr=C3=B6ge?= Date: Tue, 24 Jun 2008 12:08:47 +0000 Subject: gst/deinterlace2/tvtime/greedy.c: Optimize MMX/MMXEXT implementations a bit by requiring two less memory accesses and... Original commit message from CVS: * gst/deinterlace2/tvtime/greedy.c: (deinterlace_greedy_packed422_scanline_mmx), (deinterlace_greedy_packed422_scanline_mmxext): Optimize MMX/MMXEXT implementations a bit by requiring two less memory accesses and fix the workaround for the missing right shift on bytes to unset the highest bit of every byte. --- ChangeLog | 9 +++++++++ gst/deinterlace2/tvtime/greedy.c | 28 ++++++++++++++++++++++------ 2 files changed, 31 insertions(+), 6 deletions(-) diff --git a/ChangeLog b/ChangeLog index 42c37a61..750f7620 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,12 @@ +2008-06-24 Sebastian Dröge + + * gst/deinterlace2/tvtime/greedy.c: + (deinterlace_greedy_packed422_scanline_mmx), + (deinterlace_greedy_packed422_scanline_mmxext): + Optimize MMX/MMXEXT implementations a bit by requiring two less + memory accesses and fix the workaround for the missing right shift + on bytes to unset the highest bit of every byte. + 2008-06-24 Sebastian Dröge * gst/deinterlace2/tvtime/greedy.c: diff --git a/gst/deinterlace2/tvtime/greedy.c b/gst/deinterlace2/tvtime/greedy.c index 500e07fd..16b98312 100644 --- a/gst/deinterlace2/tvtime/greedy.c +++ b/gst/deinterlace2/tvtime/greedy.c @@ -118,6 +118,8 @@ deinterlace_greedy_packed422_scanline_mmx (GstDeinterlace2 * object, { mmx_t MaxComb; + mmx_t ShiftMask; + // How badly do we let it weave? 0-255 MaxComb.ub[0] = GreedyMaxComb; MaxComb.ub[1] = GreedyMaxComb; @@ -128,10 +130,21 @@ deinterlace_greedy_packed422_scanline_mmx (GstDeinterlace2 * object, MaxComb.ub[6] = GreedyMaxComb; MaxComb.ub[7] = GreedyMaxComb; + ShiftMask.ub[0] = 0x7f; + ShiftMask.ub[1] = 0x7f; + ShiftMask.ub[2] = 0x7f; + ShiftMask.ub[3] = 0x7f; + ShiftMask.ub[4] = 0x7f; + ShiftMask.ub[5] = 0x7f; + ShiftMask.ub[6] = 0x7f; + ShiftMask.ub[7] = 0x7f; + // L2 == m0 // L1 == t1 // L3 == b1 - // LP2 == m2 + // LP2 == m2 + + movq_m2r (MaxComb, mm6); for (; width > 7; width -= 8) { movq_m2r (*t1, mm1); // L1 @@ -143,7 +156,9 @@ deinterlace_greedy_packed422_scanline_mmx (GstDeinterlace2 * object, movq_r2r (mm1, mm4); // L1 movq_r2r (mm3, mm5); // L3 psrlw_i2r (1, mm4); // L1/2 + pand_m2r (ShiftMask, mm4); psrlw_i2r (1, mm5); // L3/2 + pand_m2r (ShiftMask, mm5); paddusb_r2r (mm5, mm4); // (L1 + L3) / 2 // get abs value of possible L2 comb @@ -153,7 +168,6 @@ deinterlace_greedy_packed422_scanline_mmx (GstDeinterlace2 * object, psubusb_r2r (mm2, mm5); // avg - L2 por_r2r (mm7, mm5); // abs(avg-L2) - // get abs value of possible LP2 comb movq_r2r (mm0, mm7); // LP2 psubusb_r2r (mm4, mm7); // LP2 - avg @@ -186,8 +200,8 @@ deinterlace_greedy_packed422_scanline_mmx (GstDeinterlace2 * object, psubusb_r2r (mm7, mm3); // now = Min(L1,L3) // allow the value to be above the high or below the low by amt of MaxComb - paddusb_m2r (MaxComb, mm2); // increase max by diff - psubusb_m2r (MaxComb, mm3); // lower min by diff + paddusb_r2r (mm6, mm2); // increase max by diff + psubusb_r2r (mm6, mm3); // lower min by diff psubusb_r2r (mm3, mm4); // best - Min paddusb_r2r (mm3, mm4); // now = Max(best,Min(L1,L3) @@ -236,6 +250,8 @@ deinterlace_greedy_packed422_scanline_mmxext (GstDeinterlace2 * object, // L3 == b1 // LP2 == m2 + movq_m2r (MaxComb, mm6); + for (; width > 7; width -= 8) { movq_m2r (*t1, mm1); // L1 movq_m2r (*m0, mm2); // L2 @@ -281,8 +297,8 @@ deinterlace_greedy_packed422_scanline_mmxext (GstDeinterlace2 * object, pminub_r2r (mm1, mm3); // now = Min(L1,L3) // allow the value to be above the high or below the low by amt of MaxComb - paddusb_m2r (MaxComb, mm2); // increase max by diff - psubusb_m2r (MaxComb, mm3); // lower min by diff + paddusb_r2r (mm6, mm2); // increase max by diff + psubusb_r2r (mm6, mm3); // lower min by diff pmaxub_r2r (mm3, mm4); // now = Max(best,Min(L1,L3) -- cgit v1.2.1