[x265] [PATCH] asm: 10bpp code for blockcopy_ps_48x64

murugan at multicorewareinc.com murugan at multicorewareinc.com
Tue Dec 10 09:32:23 CET 2013


# HG changeset patch
# User Murugan Vairavel <murugan at multicorewareinc.com>
# Date 1386664204 -19800
#      Tue Dec 10 14:00:04 2013 +0530
# Node ID 674a87448095d5f907454ed3fe2e1b1fe411a480
# Parent  c86148cae36d031891801911fbe29f7390f01da3
asm: 10bpp code for blockcopy_ps_48x64

diff -r c86148cae36d -r 674a87448095 source/common/x86/asm-primitives.cpp
--- a/source/common/x86/asm-primitives.cpp	Tue Dec 10 13:41:27 2013 +0530
+++ b/source/common/x86/asm-primitives.cpp	Tue Dec 10 14:00:04 2013 +0530
@@ -690,6 +690,7 @@
         p.chroma[X265_CSP_I420].copy_pp[LUMA_64x48] = x265_blockcopy_pp_32x24_sse2;
         p.chroma[X265_CSP_I420].copy_pp[LUMA_64x64] = x265_blockcopy_pp_32x32_sse2;
         p.luma_copy_pp[LUMA_32x64] = x265_blockcopy_pp_32x64_sse2;
+        p.luma_copy_pp[LUMA_48x64] = x265_blockcopy_pp_48x64_sse2;
     }
     if (cpuMask & X265_CPU_SSSE3)
     {
diff -r c86148cae36d -r 674a87448095 source/common/x86/blockcopy8.asm
--- a/source/common/x86/blockcopy8.asm	Tue Dec 10 13:41:27 2013 +0530
+++ b/source/common/x86/blockcopy8.asm	Tue Dec 10 14:00:04 2013 +0530
@@ -984,33 +984,111 @@
 ;-----------------------------------------------------------------------------
 %macro BLOCKCOPY_PP_W48_H2 2
 INIT_XMM sse2
-cglobal blockcopy_pp_%1x%2, 4, 5, 8, dest, deststride, src, srcstride
-
-mov         r4d,       %2
-
+cglobal blockcopy_pp_%1x%2, 4, 5, 6, dest, deststride, src, srcstride
+    mov    r4d,    %2/4
+%if HIGH_BIT_DEPTH
+    add     r1,    r1
+    add     r3,    r3
 .loop
-     movu     m0,     [r2]
-     movu     m1,     [r2 + 16]
-     movu     m2,     [r2 + 32]
-
-     movu     m3,     [r2 + r3]
-     movu     m4,     [r2 + r3 + 16]
-     movu     m5,     [r2 + r3 + 32]
-
-     movu     [r0],                 m0
-     movu     [r0 + 16],            m1
-     movu     [r0 + 32],            m2
-
-     movu     [r0 + r1],            m3
-     movu     [r0 + r1 + 16],       m4
-     movu     [r0 + r1 + 32],       m5
-
-     lea      r0,                   [r0 + 2 * r1]
-     lea      r2,                   [r2 + 2 * r3]
-
-     sub      r4d,                  2
-     jnz      .loop
-
+    movu    m0,    [r2]
+    movu    m1,    [r2 + 16]
+    movu    m2,    [r2 + 32]
+    movu    m3,    [r2 + 48]
+    movu    m4,    [r2 + 64]
+    movu    m5,    [r2 + 80]
+
+    movu    [r0],         m0
+    movu    [r0 + 16],    m1
+    movu    [r0 + 32],    m2
+    movu    [r0 + 48],    m3
+    movu    [r0 + 64],    m4
+    movu    [r0 + 80],    m5
+
+    movu    m0,    [r2 + r3]
+    movu    m1,    [r2 + r3 + 16]
+    movu    m2,    [r2 + r3 + 32]
+    movu    m3,    [r2 + r3 + 48]
+    movu    m4,    [r2 + r3 + 64]
+    movu    m5,    [r2 + r3 + 80]
+    lea    r2,    [r2 + 2 * r3]
+
+    movu    [r0 + r1],         m0
+    movu    [r0 + r1 + 16],    m1
+    movu    [r0 + r1 + 32],    m2
+    movu    [r0 + r1 + 48],    m3
+    movu    [r0 + r1 + 64],    m4
+    movu    [r0 + r1 + 80],    m5
+    lea     r0,    [r0 + 2 * r1]
+
+    movu    m0,    [r2]
+    movu    m1,    [r2 + 16]
+    movu    m2,    [r2 + 32]
+    movu    m3,    [r2 + 48]
+    movu    m4,    [r2 + 64]
+    movu    m5,    [r2 + 80]
+
+    movu    [r0],         m0
+    movu    [r0 + 16],    m1
+    movu    [r0 + 32],    m2
+    movu    [r0 + 48],    m3
+    movu    [r0 + 64],    m4
+    movu    [r0 + 80],    m5
+
+    movu    m0,    [r2 + r3]
+    movu    m1,    [r2 + r3 + 16]
+    movu    m2,    [r2 + r3 + 32]
+    movu    m3,    [r2 + r3 + 48]
+    movu    m4,    [r2 + r3 + 64]
+    movu    m5,    [r2 + r3 + 80]
+
+    movu    [r0 + r1],         m0
+    movu    [r0 + r1 + 16],    m1
+    movu    [r0 + r1 + 32],    m2
+    movu    [r0 + r1 + 48],    m3
+    movu    [r0 + r1 + 64],    m4
+    movu    [r0 + r1 + 80],    m5
+
+    dec     r4d
+    lea     r0,    [r0 + 2 * r1]
+    lea     r2,    [r2 + 2 * r3]
+    jnz     .loop
+%else
+.loop
+    movu    m0,    [r2]
+    movu    m1,    [r2 + 16]
+    movu    m2,    [r2 + 32]
+    movu    m3,    [r2 + r3]
+    movu    m4,    [r2 + r3 + 16]
+    movu    m5,    [r2 + r3 + 32]
+    lea     r2,    [r2 + 2 * r3]
+
+    movu    [r0],              m0
+    movu    [r0 + 16],         m1
+    movu    [r0 + 32],         m2
+    movu    [r0 + r1],         m3
+    movu    [r0 + r1 + 16],    m4
+    movu    [r0 + r1 + 32],    m5
+    lea     r0,    [r0 + 2 * r1]
+
+    movu    m0,    [r2]
+    movu    m1,    [r2 + 16]
+    movu    m2,    [r2 + 32]
+    movu    m3,    [r2 + r3]
+    movu    m4,    [r2 + r3 + 16]
+    movu    m5,    [r2 + r3 + 32]
+
+    movu    [r0],              m0
+    movu    [r0 + 16],         m1
+    movu    [r0 + 32],         m2
+    movu    [r0 + r1],         m3
+    movu    [r0 + r1 + 16],    m4
+    movu    [r0 + r1 + 32],    m5
+
+    dec    r4d
+    lea    r0,    [r0 + 2 * r1]
+    lea    r2,    [r2 + 2 * r3]
+    jnz    .loop
+%endif
 RET
 %endmacro
 


More information about the x265-devel mailing list