FFmpeg: libavcodec/x86/vp3dsp

00001 /*
00002  * Copyright (C) 2004 the ffmpeg project
00003  *
00004  * This file is part of FFmpeg.
00005  *
00006  * FFmpeg is free software; you can redistribute it and/or
00007  * modify it under the terms of the GNU Lesser General Public
00008  * License as published by the Free Software Foundation; either
00009  * version 2.1 of the License, or (at your option) any later version.
00010  *
00011  * FFmpeg is distributed in the hope that it will be useful,
00012  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00013  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
00014  * Lesser General Public License for more details.
00015  *
00016  * You should have received a copy of the GNU Lesser General Public
00017  * License along with FFmpeg; if not, write to the Free Software
00018  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
00019  */
00020 
00026 #include "libavcodec/dsputil.h"
00027 #include "dsputil_mmx.h"
00028 #include "vp3dsp_sse2.h"
00029 
00030 DECLARE_ALIGNED(16, const uint16_t, ff_vp3_idct_data)[7 * 8] =
00031 {
00032     64277,64277,64277,64277,64277,64277,64277,64277,
00033     60547,60547,60547,60547,60547,60547,60547,60547,
00034     54491,54491,54491,54491,54491,54491,54491,54491,
00035     46341,46341,46341,46341,46341,46341,46341,46341,
00036     36410,36410,36410,36410,36410,36410,36410,36410,
00037     25080,25080,25080,25080,25080,25080,25080,25080,
00038     12785,12785,12785,12785,12785,12785,12785,12785
00039 };
00040 
00041 
00042 #define VP3_1D_IDCT_SSE2(ADD, SHIFT) \
00043     "movdqa "I(3)", %%xmm2 \n\t"     /* xmm2 = i3 */ \
00044     "movdqa "C(3)", %%xmm6 \n\t"     /* xmm6 = c3 */ \
00045     "movdqa %%xmm2, %%xmm4 \n\t"     /* xmm4 = i3 */ \
00046     "movdqa "I(5)", %%xmm7 \n\t"     /* xmm7 = i5 */ \
00047     "pmulhw %%xmm6, %%xmm4 \n\t"     /* xmm4 = c3 * i3 - i3 */ \
00048     "movdqa "C(5)", %%xmm1 \n\t"     /* xmm1 = c5 */ \
00049     "pmulhw %%xmm7, %%xmm6 \n\t"     /* xmm6 = c3 * i5 - i5 */ \
00050     "movdqa %%xmm1, %%xmm5 \n\t"     /* xmm5 = c5 */ \
00051     "pmulhw %%xmm2, %%xmm1 \n\t"     /* xmm1 = c5 * i3 - i3 */ \
00052     "movdqa "I(1)", %%xmm3 \n\t"     /* xmm3 = i1 */ \
00053     "pmulhw %%xmm7, %%xmm5 \n\t"     /* xmm5 = c5 * i5 - i5 */ \
00054     "movdqa "C(1)", %%xmm0 \n\t"     /* xmm0 = c1 */ \
00055     "paddw  %%xmm2, %%xmm4 \n\t"     /* xmm4 = c3 * i3 */ \
00056     "paddw  %%xmm7, %%xmm6 \n\t"     /* xmm6 = c3 * i5 */ \
00057     "paddw  %%xmm1, %%xmm2 \n\t"     /* xmm2 = c5 * i3 */ \
00058     "movdqa "I(7)", %%xmm1 \n\t"     /* xmm1 = i7 */ \
00059     "paddw  %%xmm5, %%xmm7 \n\t"     /* xmm7 = c5 * i5 */ \
00060     "movdqa %%xmm0, %%xmm5 \n\t"     /* xmm5 = c1 */ \
00061     "pmulhw %%xmm3, %%xmm0 \n\t"     /* xmm0 = c1 * i1 - i1 */ \
00062     "paddsw %%xmm7, %%xmm4 \n\t"     /* xmm4 = c3 * i3 + c5 * i5 = C */ \
00063     "pmulhw %%xmm1, %%xmm5 \n\t"     /* xmm5 = c1 * i7 - i7 */ \
00064     "movdqa "C(7)", %%xmm7 \n\t"     /* xmm7 = c7 */ \
00065     "psubsw %%xmm2, %%xmm6 \n\t"     /* xmm6 = c3 * i5 - c5 * i3 = D */ \
00066     "paddw  %%xmm3, %%xmm0 \n\t"     /* xmm0 = c1 * i1 */ \
00067     "pmulhw %%xmm7, %%xmm3 \n\t"     /* xmm3 = c7 * i1 */ \
00068     "movdqa "I(2)", %%xmm2 \n\t"     /* xmm2 = i2 */ \
00069     "pmulhw %%xmm1, %%xmm7 \n\t"     /* xmm7 = c7 * i7 */ \
00070     "paddw  %%xmm1, %%xmm5 \n\t"     /* xmm5 = c1 * i7 */ \
00071     "movdqa %%xmm2, %%xmm1 \n\t"     /* xmm1 = i2 */ \
00072     "pmulhw "C(2)", %%xmm2 \n\t"     /* xmm2 = i2 * c2 -i2 */ \
00073     "psubsw %%xmm5, %%xmm3 \n\t"     /* xmm3 = c7 * i1 - c1 * i7 = B */ \
00074     "movdqa "I(6)", %%xmm5 \n\t"     /* xmm5 = i6 */ \
00075     "paddsw %%xmm7, %%xmm0 \n\t"     /* xmm0 = c1 * i1 + c7 * i7 = A */ \
00076     "movdqa %%xmm5, %%xmm7 \n\t"     /* xmm7 = i6 */ \
00077     "psubsw %%xmm4, %%xmm0 \n\t"     /* xmm0 = A - C */ \
00078     "pmulhw "C(2)", %%xmm5 \n\t"     /* xmm5 = c2 * i6 - i6 */ \
00079     "paddw  %%xmm1, %%xmm2 \n\t"     /* xmm2 = i2 * c2 */ \
00080     "pmulhw "C(6)", %%xmm1 \n\t"     /* xmm1 = c6 * i2 */ \
00081     "paddsw %%xmm4, %%xmm4 \n\t"     /* xmm4 = C + C */ \
00082     "paddsw %%xmm0, %%xmm4 \n\t"     /* xmm4 = A + C = C. */ \
00083     "psubsw %%xmm6, %%xmm3 \n\t"     /* xmm3 = B - D */ \
00084     "paddw  %%xmm7, %%xmm5 \n\t"     /* xmm5 = c2 * i6 */ \
00085     "paddsw %%xmm6, %%xmm6 \n\t"     /* xmm6 = D + D */ \
00086     "pmulhw "C(6)", %%xmm7 \n\t"     /* xmm7 = c6 * i6 */ \
00087     "paddsw %%xmm3, %%xmm6 \n\t"     /* xmm6 = B + D = D. */ \
00088     "movdqa %%xmm4, "I(1)" \n\t"     /* Save C. at I(1) */ \
00089     "psubsw %%xmm5, %%xmm1 \n\t"     /* xmm1 = c6 * i2 - c2 * i6 = H */ \
00090     "movdqa "C(4)", %%xmm4 \n\t"     /* xmm4 = c4 */ \
00091     "movdqa %%xmm3, %%xmm5 \n\t"     /* xmm5 = B - D */ \
00092     "pmulhw %%xmm4, %%xmm3 \n\t"     /* xmm3 = ( c4 -1 ) * ( B - D ) */ \
00093     "paddsw %%xmm2, %%xmm7 \n\t"     /* xmm7 = c2 * i2 + c6 * i6 = G */ \
00094     "movdqa %%xmm6, "I(2)" \n\t"     /* Save D. at I(2) */ \
00095     "movdqa %%xmm0, %%xmm2 \n\t"     /* xmm2 = A - C */ \
00096     "movdqa "I(0)", %%xmm6 \n\t"     /* xmm6 = i0 */ \
00097     "pmulhw %%xmm4, %%xmm0 \n\t"     /* xmm0 = ( c4 - 1 ) * ( A - C ) = A. */ \
00098     "paddw  %%xmm3, %%xmm5 \n\t"     /* xmm5 = c4 * ( B - D ) = B. */ \
00099     "movdqa "I(4)", %%xmm3 \n\t"     /* xmm3 = i4 */ \
00100     "psubsw %%xmm1, %%xmm5 \n\t"     /* xmm5 = B. - H = B.. */ \
00101     "paddw  %%xmm0, %%xmm2 \n\t"     /* xmm2 = c4 * ( A - C) = A. */ \
00102     "psubsw %%xmm3, %%xmm6 \n\t"     /* xmm6 = i0 - i4 */ \
00103     "movdqa %%xmm6, %%xmm0 \n\t"     /* xmm0 = i0 - i4 */ \
00104     "pmulhw %%xmm4, %%xmm6 \n\t"     /* xmm6 = (c4 - 1) * (i0 - i4) = F */ \
00105     "paddsw %%xmm3, %%xmm3 \n\t"     /* xmm3 = i4 + i4 */ \
00106     "paddsw %%xmm1, %%xmm1 \n\t"     /* xmm1 = H + H */ \
00107     "paddsw %%xmm0, %%xmm3 \n\t"     /* xmm3 = i0 + i4 */ \
00108     "paddsw %%xmm5, %%xmm1 \n\t"     /* xmm1 = B. + H = H. */ \
00109     "pmulhw %%xmm3, %%xmm4 \n\t"     /* xmm4 = ( c4 - 1 ) * ( i0 + i4 )  */ \
00110     "paddw  %%xmm0, %%xmm6 \n\t"     /* xmm6 = c4 * ( i0 - i4 ) */ \
00111     "psubsw %%xmm2, %%xmm6 \n\t"     /* xmm6 = F - A. = F. */ \
00112     "paddsw %%xmm2, %%xmm2 \n\t"     /* xmm2 = A. + A. */ \
00113     "movdqa "I(1)", %%xmm0 \n\t"     /* Load        C. from I(1) */ \
00114     "paddsw %%xmm6, %%xmm2 \n\t"     /* xmm2 = F + A. = A.. */ \
00115     "paddw  %%xmm3, %%xmm4 \n\t"     /* xmm4 = c4 * ( i0 + i4 ) = 3 */ \
00116     "psubsw %%xmm1, %%xmm2 \n\t"     /* xmm2 = A.. - H. = R2 */ \
00117     ADD(%%xmm2)                      /* Adjust R2 and R1 before shifting */ \
00118     "paddsw %%xmm1, %%xmm1 \n\t"     /* xmm1 = H. + H. */ \
00119     "paddsw %%xmm2, %%xmm1 \n\t"     /* xmm1 = A.. + H. = R1 */ \
00120     SHIFT(%%xmm2)                    /* xmm2 = op2 */ \
00121     "psubsw %%xmm7, %%xmm4 \n\t"     /* xmm4 = E - G = E. */ \
00122     SHIFT(%%xmm1)                    /* xmm1 = op1 */ \
00123     "movdqa "I(2)", %%xmm3 \n\t"     /* Load D. from I(2) */ \
00124     "paddsw %%xmm7, %%xmm7 \n\t"     /* xmm7 = G + G */ \
00125     "paddsw %%xmm4, %%xmm7 \n\t"     /* xmm7 = E + G = G. */ \
00126     "psubsw %%xmm3, %%xmm4 \n\t"     /* xmm4 = E. - D. = R4 */ \
00127     ADD(%%xmm4)                      /* Adjust R4 and R3 before shifting */ \
00128     "paddsw %%xmm3, %%xmm3 \n\t"     /* xmm3 = D. + D. */ \
00129     "paddsw %%xmm4, %%xmm3 \n\t"     /* xmm3 = E. + D. = R3 */ \
00130     SHIFT(%%xmm4)                    /* xmm4 = op4 */ \
00131     "psubsw %%xmm5, %%xmm6 \n\t"     /* xmm6 = F. - B..= R6 */ \
00132     SHIFT(%%xmm3)                    /* xmm3 = op3 */ \
00133     ADD(%%xmm6)                      /* Adjust R6 and R5 before shifting */ \
00134     "paddsw %%xmm5, %%xmm5 \n\t"     /* xmm5 = B.. + B.. */ \
00135     "paddsw %%xmm6, %%xmm5 \n\t"     /* xmm5 = F. + B.. = R5 */ \
00136     SHIFT(%%xmm6)                    /* xmm6 = op6 */ \
00137     SHIFT(%%xmm5)                    /* xmm5 = op5 */ \
00138     "psubsw %%xmm0, %%xmm7 \n\t"     /* xmm7 = G. - C. = R7 */ \
00139     ADD(%%xmm7)                      /* Adjust R7 and R0 before shifting */ \
00140     "paddsw %%xmm0, %%xmm0 \n\t"     /* xmm0 = C. + C. */ \
00141     "paddsw %%xmm7, %%xmm0 \n\t"     /* xmm0 = G. + C. */ \
00142     SHIFT(%%xmm7)                    /* xmm7 = op7 */ \
00143     SHIFT(%%xmm0)                    /* xmm0 = op0 */
00144 
00145 #define PUT_BLOCK(r0, r1, r2, r3, r4, r5, r6, r7) \
00146     "movdqa " #r0 ", " O(0) "\n\t" \
00147     "movdqa " #r1 ", " O(1) "\n\t" \
00148     "movdqa " #r2 ", " O(2) "\n\t" \
00149     "movdqa " #r3 ", " O(3) "\n\t" \
00150     "movdqa " #r4 ", " O(4) "\n\t" \
00151     "movdqa " #r5 ", " O(5) "\n\t" \
00152     "movdqa " #r6 ", " O(6) "\n\t" \
00153     "movdqa " #r7 ", " O(7) "\n\t"
00154 
00155 #define NOP(xmm)
00156 #define SHIFT4(xmm) "psraw  $4, "#xmm"\n\t"
00157 #define ADD8(xmm)   "paddsw %2, "#xmm"\n\t"
00158 
00159 void ff_vp3_idct_sse2(int16_t *input_data)
00160 {
00161 #define I(x) AV_STRINGIFY(16*x)"(%0)"
00162 #define O(x) I(x)
00163 #define C(x) AV_STRINGIFY(16*(x-1))"(%1)"
00164 
00165     __asm__ volatile (
00166         VP3_1D_IDCT_SSE2(NOP, NOP)
00167 
00168         TRANSPOSE8(%%xmm0, %%xmm1, %%xmm2, %%xmm3, %%xmm4, %%xmm5, %%xmm6, %%xmm7, (%0))
00169         PUT_BLOCK(%%xmm0, %%xmm5, %%xmm7, %%xmm3, %%xmm6, %%xmm4, %%xmm2, %%xmm1)
00170 
00171         VP3_1D_IDCT_SSE2(ADD8, SHIFT4)
00172         PUT_BLOCK(%%xmm0, %%xmm1, %%xmm2, %%xmm3, %%xmm4, %%xmm5, %%xmm6, %%xmm7)
00173         :: "r"(input_data), "r"(ff_vp3_idct_data), "m"(ff_pw_8)
00174     );
00175 }
00176 
00177 void ff_vp3_idct_put_sse2(uint8_t *dest, int line_size, DCTELEM *block)
00178 {
00179     ff_vp3_idct_sse2(block);
00180     put_signed_pixels_clamped_mmx(block, dest, line_size);
00181 }
00182 
00183 void ff_vp3_idct_add_sse2(uint8_t *dest, int line_size, DCTELEM *block)
00184 {
00185     ff_vp3_idct_sse2(block);
00186     add_pixels_clamped_mmx(block, dest, line_size);
00187 }
libavcodec/x86/vp3dsp_sse2.c
