[x265] [PATCH] asm: 10bpp code for bolckcopy_ps_12x16

[murugan at multicorewareinc.com]

# HG changeset patch
# User Murugan Vairavel <murugan at multicorewareinc.com>
# Date 1386657211 -19800
#      Tue Dec 10 12:03:31 2013 +0530
# Node ID cbeac6eb3e8a26e6fcf35da5a65ff3e2157e46a9
# Parent  d6bb400f0733701f02f65e0f060284df3d77d9b0
asm: 10bpp code for bolckcopy_ps_12x16

diff -r d6bb400f0733 -r cbeac6eb3e8a source/common/x86/asm-primitives.cpp
--- a/source/common/x86/asm-primitives.cpp	Tue Dec 10 11:50:44 2013 +0530
+++ b/source/common/x86/asm-primitives.cpp	Tue Dec 10 12:03:31 2013 +0530
@@ -677,6 +677,7 @@
         p.chroma[X265_CSP_I420].copy_pp[LUMA_16x16] = x265_blockcopy_pp_8x8_sse2;
         p.chroma[X265_CSP_I420].copy_pp[LUMA_16x32] = x265_blockcopy_pp_8x16_sse2;
         p.chroma[X265_CSP_I420].copy_pp[LUMA_16x64] = x265_blockcopy_pp_8x32_sse2;
+        p.chroma[X265_CSP_I420].copy_pp[LUMA_24x32] = x265_blockcopy_pp_12x16_sse2;
     }
     if (cpuMask & X265_CPU_SSSE3)
     {
diff -r d6bb400f0733 -r cbeac6eb3e8a source/common/x86/blockcopy8.asm
--- a/source/common/x86/blockcopy8.asm	Tue Dec 10 11:50:44 2013 +0530
+++ b/source/common/x86/blockcopy8.asm	Tue Dec 10 12:03:31 2013 +0530
@@ -578,46 +578,69 @@
 ;-----------------------------------------------------------------------------
 %macro BLOCKCOPY_PP_W12_H4 2
 INIT_XMM sse2
-cglobal blockcopy_pp_%1x%2, 4, 7, 8, dest, deststride, src, srcstride
-
-mov         r4d,       %2
-
+cglobal blockcopy_pp_%1x%2, 4, 5, 4, dest, deststride, src, srcstride
+
+    mov         r4d,       %2/4
+%if HIGH_BIT_DEPTH
+    add     r1,    r1
+    add     r3,    r3
 .loop
-      movh     m0,     [r2]
-      movd     m1,     [r2 + 8]
-
-      movh     m2,     [r2 + r3]
-      movd     m3,     [r2 + r3 + 8]
-
-      movh     m4,     [r2 + 2 * r3]
-      movd     m5,     [r2 + 2 * r3 + 8]
-
-      lea      r5,     [r2 + 2 * r3]
-
-      movh     m6,     [r5 + r3]
-      movd     m7,     [r5 + r3 + 8]
-
-      movh     [r0],                 m0
-      movd     [r0 + 8],             m1
-
-      movh     [r0 + r1],            m2
-      movd     [r0 + r1 + 8],        m3
-
-      movh     [r0 + 2 * r1],        m4
-      movd     [r0 + 2 * r1 + 8],    m5
-
-      lea      r6,                   [r0 + 2 * r1]
-
-      movh     [r6 + r1],            m6
-      movd     [r6 + r1 + 8],        m7
-
-      lea      r0,                   [r0 + 4 * r1]
-      lea      r2,                   [r2 + 4 * r3]
-
-      sub      r4d,                   4
-      jnz      .loop
-
-RET
+    movu    m0,    [r2]
+    movh    m1,    [r2 + 16]
+    movu    m2,    [r2 + r3]
+    movh    m3,    [r2 + r3 + 16]
+    lea     r2,    [r2 + 2 * r3]
+
+    movu    [r0],              m0
+    movh    [r0 + 16],         m1
+    movu    [r0 + r1],         m2
+    movh    [r0 + r1 + 16],    m3
+
+    lea     r0,    [r0 + 2 * r1]
+    movu    m0,    [r2]
+    movh    m1,    [r2 + 16]
+    movu    m2,    [r2 + r3]
+    movh    m3,    [r2 + r3 + 16]
+
+    movu    [r0],              m0
+    movh    [r0 + 16],         m1
+    movu    [r0 + r1],         m2
+    movh    [r0 + r1 + 16],    m3
+
+    dec     r4d
+    lea     r0,    [r0 + 2 * r1]
+    lea     r2,    [r2 + 2 * r3]
+    jnz     .loop
+%else
+.loop
+    movh    m0,     [r2]
+    movd    m1,     [r2 + 8]
+    movh    m2,     [r2 + r3]
+    movd    m3,     [r2 + r3 + 8]
+    lea     r2,     [r2 + 2 * r3]
+
+    movh    [r0],             m0
+    movd    [r0 + 8],         m1
+    movh    [r0 + r1],        m2
+    movd    [r0 + r1 + 8],    m3
+    lea     r0,               [r0 + 2 * r1]
+
+    movh    m0,     [r2]
+    movd    m1,     [r2 + 8]
+    movh    m2,     [r2 + r3]
+    movd    m3,     [r2 + r3 + 8]
+
+    movh    [r0],             m0
+    movd    [r0 + 8],         m1
+    movh    [r0 + r1],        m2
+    movd    [r0 + r1 + 8],    m3
+
+    dec     r4d
+    lea     r0,               [r0 + 4 * r1]
+    lea     r2,               [r2 + 4 * r3]
+    jnz     .loop
+%endif
+    RET
 %endmacro
 
 BLOCKCOPY_PP_W12_H4 12, 16