swscale: NEON optimized unscaled rgba to nv12 conversion

Signed-off-by: Yu Xiaolei <dreifachstein@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

1 files changed, 119 insertions, 0 deletions

diff --git a/libswscale/arm/rgb2yuv_neon_32.S b/libswscale/arm/rgb2yuv_neon_32.S
new file mode 100644
index 0000000000..f51a5f149f
--- /dev/null
+++ b/libswscale/arm/rgb2yuv_neon_32.S
@@ -0,0 +1,119 @@
+/*
+ * Copyright (C) 2013 Xiaolei Yu <dreifachstein@gmail.com>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "rgb2yuv_neon_common.S"
+
+/* downsampled R16G16B16 x8 */
+alias_qw    r16x8,  q7
+alias_qw    g16x8,  q8
+alias_qw    b16x8,  q9
+
+alias   n16x16_o,   q11
+alias   n16x16_ol,  q11_l
+alias   n16x16_oh,  q11_h
+
+alias   y32x16_el,  q12
+alias   y32x16_eh,  q13
+alias   y32x16_ol,  q14
+alias   y32x16_oh,  q15
+
+alias   y16x16_e,   q12
+alias   y16x16_el,  q12_l
+alias   y16x16_eh,  q12_h
+alias   y16x16_o,   q13
+alias   y16x16_ol,  q13_l
+alias   y16x16_oh,  q13_h
+
+
+alias   y8x16,  y16x16_e
+
+
+.macro init     src
+    // load s32x3x3, narrow to s16x3x3
+    vld3.i32    {q13_l, q14_l, q15_l},          [\src]!
+    vld3.i32    {q13_h[0], q14_h[0], q15_h[0]}, [\src]
+
+    vmovn.i32   CO_R, q13
+    vmovn.i32   CO_G, q14
+    vmovn.i32   CO_B, q15
+
+    vmov.u8     BIAS_Y, #16
+    vmov.u8     BIAS_U, #128
+.endm
+
+
+.macro compute_y_16x1_step  action, s8x16, coeff
+    vmov.u8     n16x16_o,   #0
+    vtrn.u8     \s8x16,     n16x16_o
+
+    \action     y32x16_el,  \s8x16\()_l,    \coeff
+    \action     y32x16_eh,  \s8x16\()_h,    \coeff
+    \action     y32x16_ol,  n16x16_ol,      \coeff
+    \action     y32x16_oh,  n16x16_oh,      \coeff
+.endm
+
+/*
+ * in:      r8x16, g8x16, b8x16
+ * out:     y8x16
+ * clobber: q11-q15, r8x16, g8x16, b8x16
+ */
+.macro compute_y_16x1
+    compute_y_16x1_step vmull, r8x16, CO_RY
+    compute_y_16x1_step vmlal, g8x16, CO_GY
+    compute_y_16x1_step vmlal, b8x16, CO_BY
+
+    vrshrn.i32  y16x16_el,  y32x16_el,  #15
+    vrshrn.i32  y16x16_eh,  y32x16_eh,  #15
+    vrshrn.i32  y16x16_ol,  y32x16_ol,  #15
+    vrshrn.i32  y16x16_oh,  y32x16_oh,  #15
+
+    vtrn.8      y16x16_e,   y16x16_o
+    vadd.u8     y8x16,      y8x16,      BIAS_Y
+.endm
+
+alias   c32x8_l,    q14
+alias   c32x8_h,    q15
+
+alias_qw    c16x8,  q13
+alias_qw    c8x8x2, q10
+
+.macro compute_chroma_8x1_step  action, s16x8, coeff
+    \action     c32x8_l,    \s16x8\()_l,    \coeff
+    \action     c32x8_h,    \s16x8\()_h,    \coeff
+.endm
+
+/*
+ * in:      r16x8, g16x8, b16x8
+ * out:     c8x8
+ * clobber: q14-q15
+ */
+.macro compute_chroma_8x1   c, C
+    compute_chroma_8x1_step vmull, r16x8, CO_R\C
+    compute_chroma_8x1_step vmlal, g16x8, CO_G\C
+    compute_chroma_8x1_step vmlal, b16x8, CO_B\C
+
+    vrshrn.i32  c16x8_l,    c32x8_l,    #15
+    vrshrn.i32  c16x8_h,    c32x8_h,    #15
+    vmovn.i16   \c\()8x8,   c16x8
+    vadd.u8     \c\()8x8,   \c\()8x8,   BIAS_\C
+.endm
+
+
+    loop_420sp  rgbx, nv12, init, kernel_420_16x2, 32