avcodec: loongson move simple idct functions to a separate file
author周晓勇 <zhouxiaoyong@loongson.cn>
Tue, 21 Jul 2015 05:53:44 +0000 (13:53 +0800)
committerMichael Niedermayer <michael@niedermayer.cc>
Tue, 21 Jul 2015 12:48:18 +0000 (14:48 +0200)
Signed-off-by: ZhouXiaoyong <zhouxiaoyong@loongson.cn>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
libavcodec/mips/Makefile
libavcodec/mips/idctdsp_init_mips.c
libavcodec/mips/idctdsp_mmi.c
libavcodec/mips/simple_idct_mmi.c [new file with mode: 0644]

index c2996c2..1d27edd 100644 (file)
@@ -63,4 +63,5 @@ MMI-OBJS-$(CONFIG_H264DSP)                += mips/h264dsp_mmi.o
 MMI-OBJS-$(CONFIG_H264CHROMA)             += mips/h264chroma_mmi.o
 MMI-OBJS-$(CONFIG_H264PRED)               += mips/h264pred_mmi.o
 MMI-OBJS-$(CONFIG_MPEGVIDEO)              += mips/mpegvideo_mmi.o
-MMI-OBJS-$(CONFIG_IDCTDSP)                += mips/idctdsp_mmi.o
+MMI-OBJS-$(CONFIG_IDCTDSP)                += mips/idctdsp_mmi.o           \
+                                             mips/simple_idct_mmi.o
index 3d2192e..ac21669 100644 (file)
@@ -43,7 +43,7 @@ static av_cold void idctdsp_init_msa(IDCTDSPContext *c, AVCodecContext *avctx,
 
 #if HAVE_MMI
 static av_cold void idctdsp_init_mmi(IDCTDSPContext *c, AVCodecContext *avctx,
-                                     unsigned high_bit_depth)
+        unsigned high_bit_depth)
 {
     if ((avctx->lowres != 1) && (avctx->lowres != 2) && (avctx->lowres != 3) &&
         (avctx->bits_per_raw_sample != 10) &&
@@ -61,7 +61,7 @@ static av_cold void idctdsp_init_mmi(IDCTDSPContext *c, AVCodecContext *avctx,
 }
 #endif /* HAVE_MMI */
 
-void ff_idctdsp_init_mips(IDCTDSPContext *c, AVCodecContext *avctx,
+av_cold void ff_idctdsp_init_mips(IDCTDSPContext *c, AVCodecContext *avctx,
                           unsigned high_bit_depth)
 {
 #if HAVE_MSA
index 83afb8a..25476f3 100644 (file)
 #include "idctdsp_mips.h"
 #include "constants.h"
 
-#define C0 23170 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
-#define C1 22725 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
-#define C2 21407 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
-#define C3 19266 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
-#define C4 16383 //cos(i*M_PI/16)*sqrt(2)*(1<<14) - 0.5
-#define C5 12873 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
-#define C6 8867  //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
-#define C7 4520  //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
-
-#define ROW_SHIFT 11
-#define COL_SHIFT 20
-
-DECLARE_ALIGNED(8, static const int16_t, coeffs)[]= {
-    1<<(ROW_SHIFT-1),   0, 1<<(ROW_SHIFT-1),   0,
-    1<<(ROW_SHIFT-1),   1, 1<<(ROW_SHIFT-1),   0,
-                  C4,  C4,               C4,  C4,
-                  C4, -C4,               C4, -C4,
-                  C2,  C6,               C2,  C6,
-                  C6, -C2,               C6, -C2,
-                  C1,  C3,               C1,  C3,
-                  C5,  C7,               C5,  C7,
-                  C3, -C7,               C3, -C7,
-                 -C1, -C5,              -C1, -C5,
-                  C5, -C1,               C5, -C1,
-                  C7,  C3,               C7,  C3,
-                  C7, -C5,               C7, -C5,
-                  C3, -C1,               C3, -C1
-};
-
-static void simple_idct_mmi(int16_t *block)
-{
-        DECLARE_ALIGNED(8, int64_t, align_tmp)[16];
-        int16_t * const temp= (int16_t*)align_tmp;
-
-        __asm__ volatile(
-#undef  DC_COND_IDCT
-#define DC_COND_IDCT(src0, src4, src1, src5, dst, rounder, rarg, shift) \
-        "ldc1 $f0, " #src0 "            \n\t" /* R4     R0      r4      r0 */\
-        "ldc1 $f2, " #src4 "            \n\t" /* R6     R2      r6      r2 */\
-        "ldc1 $f4, " #src1 "            \n\t" /* R3     R1      r3      r1 */\
-        "ldc1 $f6, " #src5 "            \n\t" /* R7     R5      r7      r5 */\
-        "ldc1 $f8, %3                   \n\t"\
-        "and  $f8, $f8, $f0             \n\t"\
-        "or $f8, $f8, $f2               \n\t"\
-        "or $f8, $f8, $f4               \n\t"\
-        "or $f8, $f8, $f6               \n\t"\
-        "packsswh $f8, $f8, $f8         \n\t"\
-        "li $11, " #shift "             \n\t"\
-        "mfc1 $10, $f8                  \n\t"\
-        "mtc1 $11, $f18                 \n\t"\
-        "beqz $10, 1f                   \n\t"\
-        "ldc1 $f8, 16(%2)               \n\t" /* C4     C4      C4      C4 */\
-        "pmaddhw $f8, $f8, $f0          \n\t" /* C4R4+C4R0      C4r4+C4r0 */\
-        "ldc1 $f10, 24(%2)              \n\t" /* -C4    C4      -C4     C4 */\
-        "pmaddhw $f0, $f0, $f10         \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
-        "ldc1 $f10, 32(%2)              \n\t" /* C6     C2      C6      C2 */\
-        "pmaddhw $f10, $f10, $f2        \n\t" /* C6R6+C2R2      C6r6+C2r2 */\
-        "ldc1 $f12, 40(%2)              \n\t" /* -C2    C6      -C2     C6 */\
-        "pmaddhw $f2, $f2, $f12         \n\t" /* -C2R6+C6R2     -C2r6+C6r2 */\
-        "ldc1 $f14, 48(%2)              \n\t" /* C3     C1      C3      C1 */\
-        "ldc1 $f16, " #rarg "           \n\t"\
-        "pmaddhw $f14, $f14, $f4        \n\t" /* C3R3+C1R1      C3r3+C1r1 */\
-        #rounder " $f8, $f8, $f16       \n\t"\
-        "mov.d $f12, $f8                \n\t" /* C4R4+C4R0      C4r4+C4r0 */\
-        "paddw $f8, $f8, $f10           \n\t" /* A0             a0 */\
-        "psubw $f12, $f12, $f10         \n\t" /* A3             a3 */\
-        "ldc1 $f10, 56(%2)              \n\t" /* C7     C5      C7      C5 */\
-        "ldc1 $f16, " #rarg "           \n\t"\
-        "pmaddhw $f10, $f10, $f6        \n\t" /* C7R7+C5R5      C7r7+C5r5 */\
-        #rounder " $f0, $f0, $f16       \n\t"\
-        "paddw $f2, $f2, $f0            \n\t" /* A1             a1 */\
-        "ldc1 $f16, 64(%2)              \n\t"\
-        "paddw $f0, $f0, $f0            \n\t" \
-        "psubw $f0, $f0, $f2            \n\t" /* A2             a2 */\
-        "pmaddhw $f4, $f4, $f16         \n\t" /* -C7R3+C3R1     -C7r3+C3r1 */\
-        "paddw $f14, $f14, $f10         \n\t" /* B0             b0 */\
-        "ldc1 $f10, 72(%2)              \n\t" /* -C5    -C1     -C5     -C1 */\
-        "pmaddhw $f10, $f10, $f6        \n\t" /* -C5R7-C1R5     -C5r7-C1r5 */\
-        "paddw $f14, $f14, $f8          \n\t" /* A0+B0          a0+b0 */\
-        "paddw $f8, $f8, $f8            \n\t" /* 2A0            2a0 */\
-        "psubw $f8, $f8, $f14           \n\t" /* A0-B0          a0-b0 */\
-        "paddw $f10, $f10, $f4          \n\t" /* B1             b1 */\
-        "psraw $f14, $f14, $f18         \n\t"\
-        "psraw $f8, $f8, $f18           \n\t"\
-        "mov.d $f4, $f2                 \n\t" /* A1             a1 */\
-        "paddw $f2, $f2, $f10           \n\t" /* A1+B1          a1+b1 */\
-        "psubw $f4, $f4, $f10           \n\t" /* A1-B1          a1-b1 */\
-        "psraw $f2, $f2, $f18           \n\t"\
-        "psraw $f4, $f4, $f18           \n\t"\
-        "packsswh $f14, $f14, $f2       \n\t" /* A1+B1  a1+b1   A0+B0   a0+b0 */\
-        "packsswh $f4, $f4, $f8         \n\t" /* A0-B0  a0-b0   A1-B1   a1-b1 */\
-        "sdc1 $f14, " #dst "            \n\t"\
-        "ldc1 $f2, " #src1 "            \n\t" /* R3     R1      r3      r1 */\
-        "ldc1 $f8, 80(%2)               \n\t" /* -C1    C5      -C1     C5 */\
-        "sdc1 $f4, 24+" #dst "          \n\t"\
-        "pmaddhw $f8, $f8, $f2          \n\t" /* -C1R3+C5R1     -C1r3+C5r1 */\
-        "ldc1 $f16, 96(%2)              \n\t"\
-        "ldc1 $f14, 88(%2)              \n\t" /* C3     C7      C3      C7 */\
-        "pmaddhw $f2, $f2, $f16         \n\t" /* -C5R3+C7R1     -C5r3+C7r1 */\
-        "pmaddhw $f14, $f14, $f6        \n\t" /* C3R7+C7R5      C3r7+C7r5 */\
-        "ldc1 $f16, 104(%2)             \n\t"\
-        "mov.d $f4, $f0                 \n\t" /* A2             a2 */\
-        "pmaddhw $f6, $f6, $f16         \n\t" /* -C1R7+C3R5     -C1r7+C3r5 */\
-        "paddw $f8, $f8, $f14           \n\t" /* B2             b2 */\
-        "paddw $f4, $f4, $f8            \n\t" /* A2+B2          a2+b2 */\
-        "psubw $f0, $f0, $f8            \n\t" /* a2-B2          a2-b2 */\
-        "psraw $f4, $f4, $f18           \n\t"\
-        "psraw $f0, $f0, $f18           \n\t"\
-        "mov.d $f8, $f12                \n\t" /* A3             a3 */\
-        "paddw $f6, $f6, $f2            \n\t" /* B3             b3 */\
-        "paddw $f12, $f12, $f6          \n\t" /* A3+B3          a3+b3 */\
-        "psubw $f8, $f8, $f6            \n\t" /* a3-B3          a3-b3 */\
-        "psraw $f12, $f12, $f18         \n\t"\
-        "packsswh $f4, $f4, $f12        \n\t" /* A3+B3  a3+b3   A2+B2   a2+b2 */\
-        "sdc1 $f4, 8+" #dst "           \n\t"\
-        "psraw $f8, $f8, $f18           \n\t"\
-        "packsswh $f8, $f8, $f0         \n\t" /* A2-B2  a2-b2   A3-B3   a3-b3 */\
-        "sdc1 $f8, 16+" #dst "          \n\t"\
-        "b 2f                           \n\t"\
-        "1:                             \n\t"\
-        "li $10, 16                     \n\t"\
-        "mtc1 $10, $f16                 \n\t"\
-        "psllw $f0, $f0, $f16           \n\t"\
-        "ldc1 $f16, %4                  \n\t"\
-        "paddw $f0, $f0, $f16           \n\t"\
-        "li $10, 13                     \n\t"\
-        "mtc1 $10, $f16                 \n\t"\
-        "psraw $f0, $f0, $f16           \n\t"\
-        "packsswh $f0, $f0, $f0         \n\t"\
-        "sdc1 $f0, " #dst "             \n\t"\
-        "sdc1 $f0, 8+" #dst "           \n\t"\
-        "sdc1 $f0, 16+" #dst "          \n\t"\
-        "sdc1 $f0, 24+" #dst "          \n\t"\
-        "2:                             \n\t"
-
-#undef  Z_COND_IDCT
-#define Z_COND_IDCT(src0, src4, src1, src5, dst, rounder, rarg, shift, bt) \
-        "ldc1 $f0, " #src0 "            \n\t" /* R4     R0      r4      r0 */\
-        "ldc1 $f2, " #src4 "            \n\t" /* R6     R2      r6      r2 */\
-        "ldc1 $f4, " #src1 "            \n\t" /* R3     R1      r3      r1 */\
-        "ldc1 $f6, " #src5 "            \n\t" /* R7     R5      r7      r5 */\
-        "mov.d $f8, $f0                 \n\t"\
-        "or $f8, $f8, $f2               \n\t"\
-        "or $f8, $f8, $f4               \n\t"\
-        "or $f8, $f8, $f6               \n\t"\
-        "packsswh $f8, $f8, $f8         \n\t"\
-        "mfc1 $10, $f8                  \n\t"\
-        "beqz $10, " #bt "              \n\t"\
-        "ldc1 $f8, 16(%2)               \n\t" /* C4     C4      C4      C4 */\
-        "pmaddhw $f8, $f8, $f0          \n\t" /* C4R4+C4R0      C4r4+C4r0 */\
-        "ldc1 $f10, 24(%2)              \n\t" /* -C4    C4      -C4     C4 */\
-        "pmaddhw $f0, $f0, $f10         \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
-        "ldc1 $f10, 32(%2)              \n\t" /* C6     C2      C6      C2 */\
-        "pmaddhw $f10, $f10, $f2        \n\t" /* C6R6+C2R2      C6r6+C2r2 */\
-        "ldc1 $f12, 40(%2)              \n\t" /* -C2    C6      -C2     C6 */\
-        "pmaddhw $f2, $f2, $f12         \n\t" /* -C2R6+C6R2     -C2r6+C6r2 */\
-        "ldc1 $f14, 48(%2)              \n\t" /* C3     C1      C3      C1 */\
-        "ldc1 $f16, " #rarg "           \n\t"\
-        "pmaddhw $f14, $f14, $f4        \n\t" /* C3R3+C1R1      C3r3+C1r1 */\
-        #rounder " $f8, $f8, $f16       \n\t"\
-        "mov.d $f12, $f8                \n\t" /* C4R4+C4R0      C4r4+C4r0 */\
-        "paddw $f8, $f8, $f10           \n\t" /* A0             a0 */\
-        "psubw $f12, $f12, $f10         \n\t" /* A3             a3 */\
-        "ldc1 $f10, 56(%2)              \n\t" /* C7     C5      C7      C5 */\
-        "ldc1 $f16, " #rarg "           \n\t"\
-        "pmaddhw $f10, $f10, $f6        \n\t" /* C7R7+C5R5      C7r7+C5r5 */\
-        #rounder " $f0, $f0, $f16       \n\t"\
-        "paddw $f2, $f2, $f0            \n\t" /* A1             a1 */\
-        "paddw $f0, $f0, $f0            \n\t" \
-        "ldc1 $f16, 64(%2)              \n\t"\
-        "psubw $f0, $f0, $f2            \n\t" /* A2             a2 */\
-        "pmaddhw $f4, $f4, $f16         \n\t" /* -C7R3+C3R1     -C7r3+C3r1 */\
-        "paddw $f14, $f14, $f10         \n\t" /* B0             b0 */\
-        "ldc1 $f10, 72(%2)              \n\t" /* -C5    -C1     -C5     -C1 */\
-        "pmaddhw $f10, $f10, $f6        \n\t" /* -C5R7-C1R5     -C5r7-C1r5 */\
-        "paddw $f14, $f14, $f8          \n\t" /* A0+B0          a0+b0 */\
-        "paddw $f8, $f8, $f8            \n\t" /* 2A0            2a0 */\
-        "li $10, " #shift "             \n\t"\
-        "psubw $f8, $f8, $f14           \n\t" /* A0-B0          a0-b0 */\
-        "mtc1 $10, $f18                 \n\t"\
-        "paddw $f10, $f10, $f4          \n\t" /* B1             b1 */\
-        "psraw $f14, $f14, $f18         \n\t"\
-        "psraw $f8, $f8, $f18           \n\t"\
-        "mov.d $f4, $f2                 \n\t" /* A1             a1 */\
-        "paddw $f2, $f2, $f10           \n\t" /* A1+B1          a1+b1 */\
-        "psubw $f4, $f4, $f10           \n\t" /* A1-B1          a1-b1 */\
-        "psraw $f2, $f2, $f18           \n\t"\
-        "psraw $f4, $f4, $f18           \n\t"\
-        "packsswh $f14, $f14, $f2       \n\t" /* A1+B1  a1+b1   A0+B0   a0+b0 */\
-        "packsswh $f4, $f4, $f8         \n\t" /* A0-B0  a0-b0   A1-B1   a1-b1 */\
-        "sdc1 $f14, " #dst "            \n\t"\
-        "ldc1 $f2, " #src1 "            \n\t" /* R3     R1      r3      r1 */\
-        "ldc1 $f8, 80(%2)               \n\t" /* -C1    C5      -C1     C5 */\
-        "sdc1 $f4, 24+" #dst "          \n\t"\
-        "pmaddhw $f8, $f8, $f2          \n\t" /* -C1R3+C5R1     -C1r3+C5r1 */\
-        "ldc1 $f16, 96(%2)              \n\t"\
-        "ldc1 $f14, 88(%2)              \n\t" /* C3     C7      C3      C7 */\
-        "pmaddhw $f2, $f2, $f16         \n\t" /* -C5R3+C7R1     -C5r3+C7r1 */\
-        "pmaddhw $f14, $f14, $f6        \n\t" /* C3R7+C7R5      C3r7+C7r5 */\
-        "ldc1 $f16, 104(%2)             \n\t"\
-        "mov.d $f4, $f0                 \n\t" /* A2             a2 */\
-        "pmaddhw $f6, $f6, $f16         \n\t" /* -C1R7+C3R5     -C1r7+C3r5 */\
-        "paddw $f8, $f8, $f14           \n\t" /* B2             b2 */\
-        "paddw $f4, $f4, $f8            \n\t" /* A2+B2          a2+b2 */\
-        "psubw $f0, $f0, $f8            \n\t" /* a2-B2          a2-b2 */\
-        "psraw $f4, $f4, $f18           \n\t"\
-        "psraw $f0, $f0, $f18           \n\t"\
-        "mov.d $f8, $f12                \n\t" /* A3             a3 */\
-        "paddw $f6, $f6, $f2            \n\t" /* B3             b3 */\
-        "paddw $f12, $f12, $f6          \n\t" /* A3+B3          a3+b3 */\
-        "psubw $f8, $f8, $f6            \n\t" /* a3-B3          a3-b3 */\
-        "psraw $f12, $f12, $f18         \n\t"\
-        "packsswh $f4, $f4, $f12        \n\t" /* A3+B3  a3+b3   A2+B2   a2+b2 */\
-        "sdc1 $f4, 8+" #dst "           \n\t"\
-        "psraw $f8, $f8, $f18           \n\t"\
-        "packsswh $f8, $f8, $f0         \n\t" /* A2-B2  a2-b2   A3-B3   a3-b3 */\
-        "sdc1 $f8, 16+" #dst "          \n\t"\
-
-        //IDCT(         src0,   src4,   src1,   src5,    dst,   rounder, shift)
-        DC_COND_IDCT(  0(%0),  8(%0), 16(%0), 24(%0),  0(%1),paddw,8(%2), 11)
-        Z_COND_IDCT(  32(%0), 40(%0), 48(%0), 56(%0), 32(%1),paddw,(%2), 11, 4f)
-        Z_COND_IDCT(  64(%0), 72(%0), 80(%0), 88(%0), 64(%1),paddw,(%2), 11, 2f)
-        Z_COND_IDCT(  96(%0),104(%0),112(%0),120(%0), 96(%1),paddw,(%2), 11, 1f)
-
-#undef  IDCT
-#define IDCT(src0, src4, src1, src5, dst, shift) \
-        "ldc1 $f0, " #src0 "            \n\t" /* R4     R0      r4      r0 */\
-        "ldc1 $f2, " #src4 "            \n\t" /* R6     R2      r6      r2 */\
-        "ldc1 $f4, " #src1 "            \n\t" /* R3     R1      r3      r1 */\
-        "ldc1 $f6, " #src5 "            \n\t" /* R7     R5      r7      r5 */\
-        "ldc1 $f8, 16(%2)               \n\t" /* C4     C4      C4      C4 */\
-        "pmaddhw $f8, $f8, $f0          \n\t" /* C4R4+C4R0      C4r4+C4r0 */\
-        "ldc1 $f10, 24(%2)              \n\t" /* -C4    C4      -C4     C4 */\
-        "pmaddhw $f0, $f0, $f10         \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
-        "ldc1 $f10, 32(%2)              \n\t" /* C6     C2      C6      C2 */\
-        "pmaddhw $f10, $f10, $f2        \n\t" /* C6R6+C2R2      C6r6+C2r2 */\
-        "ldc1 $f12, 40(%2)              \n\t" /* -C2    C6      -C2     C6 */\
-        "pmaddhw $f2, $f2, $f12         \n\t" /* -C2R6+C6R2     -C2r6+C6r2 */\
-        "mov.d $f12, $f8                \n\t" /* C4R4+C4R0      C4r4+C4r0 */\
-        "ldc1 $f14, 48(%2)              \n\t" /* C3     C1      C3      C1 */\
-        "pmaddhw $f14, $f14, $f4        \n\t" /* C3R3+C1R1      C3r3+C1r1 */\
-        "paddw $f8, $f8, $f10           \n\t" /* A0             a0 */\
-        "psubw $f12, $f12, $f10         \n\t" /* A3             a3 */\
-        "mov.d $f10, $f0                \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
-        "paddw $f0, $f0, $f2            \n\t" /* A1             a1 */\
-        "psubw $f10, $f10, $f2          \n\t" /* A2             a2 */\
-        "ldc1 $f2, 56(%2)               \n\t" /* C7     C5      C7      C5 */\
-        "ldc1 $f16, 64(%2)              \n\t"\
-        "pmaddhw $f2, $f2, $f6          \n\t" /* C7R7+C5R5      C7r7+C5r5 */\
-        "pmaddhw $f4, $f4, $f16         \n\t" /* -C7R3+C3R1     -C7r3+C3r1 */\
-        "li $10, " #shift "             \n\t"\
-        "paddw $f14, $f14, $f2          \n\t" /* B0             b0 */\
-        "ldc1 $f2, 72(%2)               \n\t" /* -C5    -C1     -C5     -C1 */\
-        "mtc1 $10, $f18                 \n\t"\
-        "pmaddhw $f2, $f2, $f6          \n\t" /* -C5R7-C1R5     -C5r7-C1r5 */\
-        "paddw $f14, $f14, $f8          \n\t" /* A0+B0          a0+b0 */\
-        "paddw $f8, $f8, $f8            \n\t" /* 2A0            2a0 */\
-        "psubw $f8, $f8, $f14           \n\t" /* A0-B0          a0-b0 */\
-        "paddw $f2, $f2, $f4            \n\t" /* B1             b1 */\
-        "psraw $f14, $f14, $f18         \n\t"\
-        "psraw $f8, $f8, $f18           \n\t"\
-        "mov.d $f4, $f0                 \n\t" /* A1             a1 */\
-        "paddw $f0, $f0, $f2            \n\t" /* A1+B1          a1+b1 */\
-        "psubw $f4, $f4, $f2            \n\t" /* A1-B1          a1-b1 */\
-        "psraw $f0, $f0, $f18           \n\t"\
-        "psraw $f4, $f4, $f18           \n\t"\
-        "packsswh $f14, $f14, $f14      \n\t" /* A0+B0  a0+b0 */\
-        "swc1 $f14, " #dst "            \n\t"\
-        "packsswh $f0, $f0, $f0         \n\t" /* A1+B1  a1+b1 */\
-        "swc1 $f0, 16+" #dst "          \n\t"\
-        "packsswh $f4, $f4, $f4         \n\t" /* A1-B1  a1-b1 */\
-        "swc1 $f4, 96+" #dst "          \n\t"\
-        "packsswh $f8, $f8, $f8         \n\t" /* A0-B0  a0-b0 */\
-        "swc1 $f8, 112+" #dst "         \n\t"\
-        "ldc1 $f0, " #src1 "            \n\t" /* R3     R1      r3      r1 */\
-        "ldc1 $f8, 80(%2)               \n\t" /* -C1    C5      -C1     C5 */\
-        "pmaddhw $f8, $f8, $f0          \n\t" /* -C1R3+C5R1     -C1r3+C5r1 */\
-        "ldc1 $f16, 96(%2)              \n\t"\
-        "ldc1 $f14, 88(%2)              \n\t" /* C3     C7      C3      C7 */\
-        "pmaddhw $f0, $f0, $f16         \n\t" /* -C5R3+C7R1     -C5r3+C7r1 */\
-        "pmaddhw $f14, $f14, $f6        \n\t" /* C3R7+C7R5      C3r7+C7r5 */\
-        "ldc1 $f16, 104(%2)             \n\t"\
-        "mov.d $f4, $f10                \n\t" /* A2             a2 */\
-        "pmaddhw $f6, $f6, $f16         \n\t" /* -C1R7+C3R5     -C1r7+C3r5 */\
-        "paddw $f8, $f8, $f14           \n\t" /* B2             b2 */\
-        "paddw $f4, $f4, $f8            \n\t" /* A2+B2          a2+b2 */\
-        "psubw $f10, $f10, $f8          \n\t" /* a2-B2          a2-b2 */\
-        "psraw $f4, $f4, $f18           \n\t"\
-        "psraw $f10, $f10, $f18         \n\t"\
-        "mov.d $f8, $f12                \n\t" /* A3             a3 */\
-        "paddw $f6, $f6, $f0            \n\t" /* B3             b3 */\
-        "paddw $f12, $f12, $f6          \n\t" /* A3+B3          a3+b3 */\
-        "psubw $f8, $f8, $f6            \n\t" /* a3-B3          a3-b3 */\
-        "psraw $f12, $f12, $f18         \n\t"\
-        "psraw $f8, $f8, $f18           \n\t"\
-        "packsswh $f4, $f4, $f4         \n\t" /* A2+B2  a2+b2 */\
-        "packsswh $f12, $f12, $f12      \n\t" /* A3+B3  a3+b3 */\
-        "swc1 $f4, 32+" #dst "          \n\t"\
-        "packsswh $f8, $f8, $f8         \n\t" /* A3-B3  a3-b3 */\
-        "packsswh $f10, $f10, $f10      \n\t" /* A2-B2  a2-b2 */\
-        "swc1 $f12, 48+" #dst "         \n\t"\
-        "swc1 $f8, 64+" #dst "          \n\t"\
-        "swc1 $f10, 80+" #dst "         \n\t"
-
-        //IDCT(  src0,   src4,   src1,    src5,    dst, shift)
-        IDCT(    (%1), 64(%1), 32(%1),  96(%1),  0(%0), 20)
-        IDCT(   8(%1), 72(%1), 40(%1), 104(%1),  4(%0), 20)
-        IDCT(  16(%1), 80(%1), 48(%1), 112(%1),  8(%0), 20)
-        IDCT(  24(%1), 88(%1), 56(%1), 120(%1), 12(%0), 20)
-        "b 9f                           \n\t"
-
-        "# .p2align 4                   \n\t"\
-        "4:                             \n\t"
-        Z_COND_IDCT(  64(%0), 72(%0), 80(%0), 88(%0), 64(%1),paddw,(%2), 11, 6f)
-        Z_COND_IDCT(  96(%0),104(%0),112(%0),120(%0), 96(%1),paddw,(%2), 11, 5f)
-
-#undef  IDCT
-#define IDCT(src0, src4, src1, src5, dst, shift) \
-        "ldc1 $f0, " #src0 "            \n\t" /* R4     R0      r4      r0 */\
-        "ldc1 $f2, " #src4 "            \n\t" /* R6     R2      r6      r2 */\
-        "ldc1 $f6, " #src5 "            \n\t" /* R7     R5      r7      r5 */\
-        "ldc1 $f8, 16(%2)               \n\t" /* C4     C4      C4      C4 */\
-        "pmaddhw $f8, $f8, $f0          \n\t" /* C4R4+C4R0      C4r4+C4r0 */\
-        "ldc1 $f10, 24(%2)              \n\t" /* -C4    C4      -C4     C4 */\
-        "pmaddhw $f0, $f0, $f10         \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
-        "ldc1 $f10, 32(%2)              \n\t" /* C6     C2      C6      C2 */\
-        "pmaddhw $f10, $f10, $f2        \n\t" /* C6R6+C2R2      C6r6+C2r2 */\
-        "ldc1 $f12, 40(%2)              \n\t" /* -C2    C6      -C2     C6 */\
-        "pmaddhw $f2, $f2, $f12         \n\t" /* -C2R6+C6R2     -C2r6+C6r2 */\
-        "mov.d $f12, $f8                \n\t" /* C4R4+C4R0      C4r4+C4r0 */\
-        "paddw $f8, $f8, $f10           \n\t" /* A0             a0 */\
-        "psubw $f12, $f12, $f10         \n\t" /* A3             a3 */\
-        "mov.d $f10, $f0                \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
-        "paddw $f0, $f0, $f2            \n\t" /* A1             a1 */\
-        "psubw $f10, $f10, $f2          \n\t" /* A2             a2 */\
-        "ldc1 $f2, 56(%2)               \n\t" /* C7     C5      C7      C5 */\
-        "li $10, " #shift "             \n\t"\
-        "pmaddhw $f2, $f2, $f6          \n\t" /* C7R7+C5R5      C7r7+C5r5 */\
-        "ldc1 $f14, 72(%2)              \n\t" /* -C5    -C1     -C5     -C1 */\
-        "mtc1 $10, $f18                 \n\t"\
-        "pmaddhw $f14, $f14, $f6        \n\t" /* -C5R7-C1R5     -C5r7-C1r5 */\
-        "paddw $f2, $f2, $f8            \n\t" /* A0+B0          a0+b0 */\
-        "paddw $f8, $f8, $f8            \n\t" /* 2A0            2a0 */\
-        "psubw $f8, $f8, $f2            \n\t" /* A0-B0          a0-b0 */\
-        "psraw $f2, $f2, $f18           \n\t"\
-        "psraw $f8, $f8, $f18           \n\t"\
-        "mov.d $f4, $f0                 \n\t" /* A1             a1 */\
-        "paddw $f0, $f0, $f14           \n\t" /* A1+B1          a1+b1 */\
-        "psubw $f4, $f4, $f14           \n\t" /* A1-B1          a1-b1 */\
-        "psraw $f0, $f0, $f18           \n\t"\
-        "psraw $f4, $f4, $f18           \n\t"\
-        "packsswh $f2, $f2, $f2         \n\t" /* A0+B0  a0+b0 */\
-        "swc1 $f2, " #dst "             \n\t"\
-        "packsswh $f0, $f0, $f0         \n\t" /* A1+B1  a1+b1 */\
-        "swc1 $f0, 16+" #dst "          \n\t"\
-        "packsswh $f4, $f4, $f4         \n\t" /* A1-B1  a1-b1 */\
-        "swc1 $f4, 96+" #dst "          \n\t"\
-        "packsswh $f8, $f8, $f8         \n\t" /* A0-B0  a0-b0 */\
-        "swc1 $f8, 112+" #dst "         \n\t"\
-        "ldc1 $f2, 88(%2)               \n\t" /* C3     C7      C3      C7 */\
-        "ldc1 $f16, 104(%2)             \n\t"\
-        "pmaddhw $f2, $f2, $f6          \n\t" /* C3R7+C7R5      C3r7+C7r5 */\
-        "mov.d $f4, $f10                \n\t" /* A2             a2 */\
-        "pmaddhw $f6, $f6, $f16         \n\t" /* -C1R7+C3R5     -C1r7+C3r5 */\
-        "paddw $f4, $f4, $f2            \n\t" /* A2+B2          a2+b2 */\
-        "psubw $f10, $f10, $f2          \n\t" /* a2-B2          a2-b2 */\
-        "psraw $f4, $f4, $f18           \n\t"\
-        "psraw $f10, $f10, $f18         \n\t"\
-        "mov.d $f2, $f12                \n\t" /* A3             a3 */\
-        "paddw $f12, $f12, $f6          \n\t" /* A3+B3          a3+b3 */\
-        "psubw $f2, $f2, $f6            \n\t" /* a3-B3          a3-b3 */\
-        "psraw $f12, $f12, $f18         \n\t"\
-        "psraw $f2, $f2, $f18           \n\t"\
-        "packsswh $f4, $f4, $f4         \n\t" /* A2+B2  a2+b2 */\
-        "packsswh $f12, $f12, $f12      \n\t" /* A3+B3  a3+b3 */\
-        "swc1 $f4, 32+" #dst "          \n\t"\
-        "packsswh $f2, $f2, $f2         \n\t" /* A3-B3  a3-b3 */\
-        "packsswh $f10, $f10, $f10      \n\t" /* A2-B2  a2-b2 */\
-        "swc1 $f12, 48+" #dst "         \n\t"\
-        "swc1 $f2, 64+" #dst "          \n\t"\
-        "swc1 $f10, 80+" #dst "         \n\t"
-
-        //IDCT(  src0,   src4,   src1,    src5,    dst, shift)
-        IDCT(    (%1), 64(%1), 32(%1),  96(%1),  0(%0), 20)
-        IDCT(   8(%1), 72(%1), 40(%1), 104(%1),  4(%0), 20)
-        IDCT(  16(%1), 80(%1), 48(%1), 112(%1),  8(%0), 20)
-        IDCT(  24(%1), 88(%1), 56(%1), 120(%1), 12(%0), 20)
-        "b 9f                           \n\t"
-
-        "# .p2align 4                   \n\t"
-        "6:                             \n\t"
-        Z_COND_IDCT(  96(%0),104(%0),112(%0),120(%0), 96(%1),paddw,(%2), 11, 7f)
-
-#undef  IDCT
-#define IDCT(src0, src4, src1, src5, dst, shift) \
-        "ldc1 $f0, " #src0 "            \n\t" /* R4     R0      r4      r0 */\
-        "ldc1 $f6, " #src5 "            \n\t" /* R7     R5      r7      r5 */\
-        "ldc1 $f8, 16(%2)               \n\t" /* C4     C4      C4      C4 */\
-        "pmaddhw $f8, $f8, $f0          \n\t" /* C4R4+C4R0      C4r4+C4r0 */\
-        "ldc1 $f10, 24(%2)              \n\t" /* -C4    C4      -C4     C4 */\
-        "pmaddhw $f0, $f0, $f10         \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
-        "mov.d $f12, $f8                \n\t" /* C4R4+C4R0      C4r4+C4r0 */\
-        "mov.d $f10, $f0                \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
-        "ldc1 $f2, 56(%2)               \n\t" /* C7     C5      C7      C5 */\
-        "pmaddhw $f2, $f2, $f6          \n\t" /* C7R7+C5R5      C7r7+C5r5 */\
-        "ldc1 $f14, 72(%2)              \n\t" /* -C5    -C1     -C5     -C1 */\
-        "li $10, " #shift "             \n\t"\
-        "pmaddhw $f14, $f14, $f6        \n\t" /* -C5R7-C1R5     -C5r7-C1r5 */\
-        "paddw $f2, $f2, $f8            \n\t" /* A0+B0          a0+b0 */\
-        "mtc1 $10, $f18                 \n\t"\
-        "paddw $f8, $f8, $f8            \n\t" /* 2A0            2a0 */\
-        "psubw $f8, $f8, $f2            \n\t" /* A0-B0          a0-b0 */\
-        "psraw $f2, $f2, $f18           \n\t"\
-        "psraw $f8, $f8, $f18           \n\t"\
-        "mov.d $f4, $f0                 \n\t" /* A1             a1 */\
-        "paddw $f0, $f0, $f14           \n\t" /* A1+B1          a1+b1 */\
-        "psubw $f4, $f4, $f14           \n\t" /* A1-B1          a1-b1 */\
-        "psraw $f0, $f0, $f18           \n\t"\
-        "psraw $f4, $f4, $f18           \n\t"\
-        "packsswh $f2, $f2, $f2         \n\t" /* A0+B0  a0+b0 */\
-        "swc1 $f2, " #dst "             \n\t"\
-        "packsswh $f0, $f0, $f0         \n\t" /* A1+B1  a1+b1 */\
-        "swc1 $f0, 16+" #dst "          \n\t"\
-        "packsswh $f4, $f4, $f4         \n\t" /* A1-B1  a1-b1 */\
-        "swc1 $f4, 96+" #dst "          \n\t"\
-        "packsswh $f8, $f8, $f8         \n\t" /* A0-B0  a0-b0 */\
-        "swc1 $f8, 112+" #dst "         \n\t"\
-        "ldc1 $f2, 88(%2)               \n\t" /* C3     C7      C3      C7 */\
-        "ldc1 $f16, 104(%2)             \n\t"\
-        "pmaddhw $f2, $f2, $f6          \n\t" /* C3R7+C7R5      C3r7+C7r5 */\
-        "mov.d $f4, $f10                 \n\t" /* A2             a2 */\
-        "pmaddhw $f6, $f6, $f16         \n\t" /* -C1R7+C3R5     -C1r7+C3r5 */\
-        "paddw $f4, $f4, $f2            \n\t" /* A2+B2          a2+b2 */\
-        "psubw $f10, $f10, $f2          \n\t" /* a2-B2          a2-b2 */\
-        "psraw $f4, $f4, $f18           \n\t"\
-        "psraw $f10, $f10, $f18         \n\t"\
-        "mov.d $f2, $f12                \n\t" /* A3             a3 */\
-        "paddw $f12, $f12, $f6          \n\t" /* A3+B3          a3+b3 */\
-        "psubw $f2, $f2, $f6            \n\t" /* a3-B3          a3-b3 */\
-        "psraw $f12, $f12, $f18         \n\t"\
-        "psraw $f2, $f2, $f18           \n\t"\
-        "packsswh $f4, $f4, $f4         \n\t" /* A2+B2  a2+b2 */\
-        "packsswh $f12, $f12, $f12      \n\t" /* A3+B3  a3+b3 */\
-        "swc1 $f4, 32+" #dst "          \n\t"\
-        "packsswh $f2, $f2, $f2         \n\t" /* A3-B3  a3-b3 */\
-        "packsswh $f10, $f10, $f10      \n\t" /* A2-B2  a2-b2 */\
-        "swc1 $f12, 48+" #dst "         \n\t"\
-        "swc1 $f2, 64+" #dst "          \n\t"\
-        "swc1 $f10, 80+" #dst "         \n\t"
-
-        //IDCT(  src0,   src4,   src1,    src5,    dst, shift)
-        IDCT(    (%1), 64(%1), 32(%1),  96(%1),  0(%0), 20)
-        IDCT(   8(%1), 72(%1), 40(%1), 104(%1),  4(%0), 20)
-        IDCT(  16(%1), 80(%1), 48(%1), 112(%1),  8(%0), 20)
-        IDCT(  24(%1), 88(%1), 56(%1), 120(%1), 12(%0), 20)
-        "b 9f                           \n\t"
-
-        "# .p2align 4                   \n\t"
-        "2:                             \n\t"
-        Z_COND_IDCT(  96(%0),104(%0),112(%0),120(%0), 96(%1),paddw,(%2), 11, 3f)
-
-#undef  IDCT
-#define IDCT(src0, src4, src1, src5, dst, shift) \
-        "ldc1 $f0, " #src0 "            \n\t" /* R4     R0      r4      r0 */\
-        "ldc1 $f4, " #src1 "            \n\t" /* R3     R1      r3      r1 */\
-        "ldc1 $f6, " #src5 "            \n\t" /* R7     R5      r7      r5 */\
-        "ldc1 $f8, 16(%2)               \n\t" /* C4     C4      C4      C4 */\
-        "pmaddhw $f8, $f8, $f0          \n\t" /* C4R4+C4R0      C4r4+C4r0 */\
-        "ldc1 $f10, 24(%2)              \n\t" /* -C4    C4      -C4     C4 */\
-        "pmaddhw $f0, $f0, $f10         \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
-        "mov.d $f12, $f8                \n\t" /* C4R4+C4R0      C4r4+C4r0 */\
-        "ldc1 $f14, 48(%2)              \n\t" /* C3     C1      C3      C1 */\
-        "pmaddhw $f14, $f14, $f4        \n\t" /* C3R3+C1R1      C3r3+C1r1 */\
-        "mov.d $f10, $f0                \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
-        "ldc1 $f2, 56(%2)               \n\t" /* C7     C5      C7      C5 */\
-        "pmaddhw $f2, $f2, $f6          \n\t" /* C7R7+C5R5      C7r7+C5r5 */\
-        "ldc1 $f16, 64(%2)              \n\t"\
-        "pmaddhw $f4, $f4, $f16         \n\t" /* -C7R3+C3R1     -C7r3+C3r1 */\
-        "paddw $f14, $f14, $f2          \n\t" /* B0             b0 */\
-        "ldc1 $f2, 72(%2)               \n\t" /* -C5    -C1     -C5     -C1 */\
-        "li $10, " #shift "             \n\t"\
-        "pmaddhw $f2, $f2, $f6          \n\t" /* -C5R7-C1R5     -C5r7-C1r5 */\
-        "paddw $f14, $f14, $f8          \n\t" /* A0+B0          a0+b0 */\
-        "mtc1 $10, $f18                 \n\t"\
-        "paddw $f8, $f8, $f8            \n\t" /* 2A0            2a0 */\
-        "psubw $f8, $f8, $f14           \n\t" /* A0-B0          a0-b0 */\
-        "paddw $f2, $f2, $f4            \n\t" /* B1             b1 */\
-        "psraw $f14, $f14, $f18         \n\t"\
-        "psraw $f8, $f8, $f18           \n\t"\
-        "mov.d $f4, $f0                 \n\t" /* A1             a1 */\
-        "paddw $f0, $f0, $f2            \n\t" /* A1+B1          a1+b1 */\
-        "psubw $f4, $f4, $f2            \n\t" /* A1-B1          a1-b1 */\
-        "psraw $f0, $f0, $f18           \n\t"\
-        "psraw $f4, $f4, $f18           \n\t"\
-        "packsswh $f14, $f14, $f14      \n\t" /* A0+B0  a0+b0 */\
-        "swc1 $f14, " #dst "            \n\t"\
-        "packsswh $f0, $f0, $f0         \n\t" /* A1+B1  a1+b1 */\
-        "swc1 $f0, 16+" #dst "          \n\t"\
-        "packsswh $f4, $f4, $f4         \n\t" /* A1-B1  a1-b1 */\
-        "swc1 $f4, 96+" #dst "          \n\t"\
-        "packsswh $f8, $f8, $f8         \n\t" /* A0-B0  a0-b0 */\
-        "swc1 $f8, 112+" #dst "         \n\t"\
-        "ldc1 $f0, " #src1 "            \n\t" /* R3     R1      r3      r1 */\
-        "ldc1 $f8, 80(%2)               \n\t" /* -C1    C5      -C1     C5 */\
-        "pmaddhw $f8, $f8, $f0          \n\t" /* -C1R3+C5R1     -C1r3+C5r1 */\
-        "ldc1 $f14, 88(%2)              \n\t" /* C3     C7      C3      C7 */\
-        "ldc1 $f16, 96(%2)              \n\t"\
-        "pmaddhw $f0, $f0, $f16         \n\t" /* -C5R3+C7R1     -C5r3+C7r1 */\
-        "pmaddhw $f14, $f14, $f6        \n\t" /* C3R7+C7R5      C3r7+C7r5 */\
-        "mov.d $f4, $f10                \n\t" /* A2             a2 */\
-        "ldc1 $f16, 104(%2)             \n\t"\
-        "pmaddhw $f6, $f6, $f16         \n\t" /* -C1R7+C3R5     -C1r7+C3r5 */\
-        "paddw $f8, $f8, $f14           \n\t" /* B2             b2 */\
-        "paddw $f4, $f4, $f8            \n\t" /* A2+B2          a2+b2 */\
-        "psubw $f10, $f10, $f8          \n\t" /* a2-B2          a2-b2 */\
-        "psraw $f4, $f4, $f18           \n\t"\
-        "psraw $f10, $f10, $f18         \n\t"\
-        "mov.d $f8, $f12                \n\t" /* A3             a3 */\
-        "paddw $f6, $f6, $f0            \n\t" /* B3             b3 */\
-        "paddw $f12, $f12, $f6          \n\t" /* A3+B3          a3+b3 */\
-        "psubw $f8, $f8, $f6            \n\t" /* a3-B3          a3-b3 */\
-        "psraw $f12, $f12, $f18         \n\t"\
-        "psraw $f8, $f8, $f18           \n\t"\
-        "packsswh $f4, $f4, $f4         \n\t" /* A2+B2  a2+b2 */\
-        "packsswh $f12, $f12, $f12      \n\t" /* A3+B3  a3+b3 */\
-        "swc1 $f4, 32+" #dst "          \n\t"\
-        "packsswh $f8, $f8, $f8         \n\t" /* A3-B3  a3-b3 */\
-        "packsswh $f10, $f10, $f10      \n\t" /* A2-B2  a2-b2 */\
-        "swc1 $f12, 48+" #dst "         \n\t"\
-        "swc1 $f8, 64+" #dst "          \n\t"\
-        "swc1 $f10, 80+" #dst "         \n\t"
-
-        //IDCT(  src0,   src4,   src1,    src5,    dst, shift)
-        IDCT(    (%1), 64(%1), 32(%1),  96(%1),  0(%0), 20)
-        IDCT(   8(%1), 72(%1), 40(%1), 104(%1),  4(%0), 20)
-        IDCT(  16(%1), 80(%1), 48(%1), 112(%1),  8(%0), 20)
-        IDCT(  24(%1), 88(%1), 56(%1), 120(%1), 12(%0), 20)
-        "b 9f                           \n\t"
-
-        "# .p2align 4                   \n\t"
-        "3:                             \n\t"
-
-#undef  IDCT
-#define IDCT(src0, src4, src1, src5, dst, shift) \
-        "ldc1 $f0, " #src0 "            \n\t" /* R4     R0      r4      r0 */\
-        "ldc1 $f4, " #src1 "            \n\t" /* R3     R1      r3      r1 */\
-        "ldc1 $f8, 16(%2)               \n\t" /* C4     C4      C4      C4 */\
-        "pmaddhw $f8, $f8, $f0          \n\t" /* C4R4+C4R0      C4r4+C4r0 */\
-        "ldc1 $f10, 24(%2)              \n\t" /* -C4    C4      -C4     C4 */\
-        "pmaddhw $f0, $f0, $f10         \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
-        "mov.d $f12, $f8                \n\t" /* C4R4+C4R0      C4r4+C4r0 */\
-        "ldc1 $f14, 48(%2)              \n\t" /* C3     C1      C3      C1 */\
-        "pmaddhw $f14, $f14, $f4        \n\t" /* C3R3+C1R1      C3r3+C1r1 */\
-        "mov.d $f10, $f0                \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
-        "ldc1 $f6, 64(%2)               \n\t"\
-        "pmaddhw $f6, $f6, $f4          \n\t" /* -C7R3+C3R1     -C7r3+C3r1 */\
-        "li $10, " #shift "             \n\t"\
-        "paddw $f14, $f14, $f8          \n\t" /* A0+B0          a0+b0 */\
-        "mtc1 $10, $f18                 \n\t"\
-        "paddw $f8, $f8, $f8            \n\t" /* 2A0            2a0 */\
-        "psubw $f8, $f8, $f14           \n\t" /* A0-B0          a0-b0 */\
-        "psraw $f14, $f14, $f18         \n\t"\
-        "psraw $f8, $f8, $f18           \n\t"\
-        "mov.d $f2, $f0                 \n\t" /* A1             a1 */\
-        "paddw $f0, $f0, $f6            \n\t" /* A1+B1          a1+b1 */\
-        "psubw $f2, $f2, $f6            \n\t" /* A1-B1          a1-b1 */\
-        "psraw $f0, $f0, $f18           \n\t"\
-        "psraw $f2, $f2, $f18           \n\t"\
-        "packsswh $f14, $f14, $f14      \n\t" /* A0+B0  a0+b0 */\
-        "swc1 $f14, " #dst "            \n\t"\
-        "packsswh $f0, $f0, $f0         \n\t" /* A1+B1  a1+b1 */\
-        "swc1 $f0, 16+" #dst "          \n\t"\
-        "packsswh $f2, $f2, $f2         \n\t" /* A1-B1  a1-b1 */\
-        "swc1 $f2, 96+" #dst "          \n\t"\
-        "packsswh $f8, $f8, $f8         \n\t" /* A0-B0  a0-b0 */\
-        "swc1 $f8, 112+" #dst "         \n\t"\
-        "ldc1 $f8, 80(%2)               \n\t" /* -C1    C5      -C1     C5 */\
-        "ldc1 $f16, 96(%2)              \n\t"\
-        "pmaddhw $f8, $f8, $f4          \n\t" /* -C1R3+C5R1     -C1r3+C5r1 */\
-        "pmaddhw $f4, $f4, $f16         \n\t" /* -C5R3+C7R1     -C5r3+C7r1 */\
-        "mov.d $f2, $f10                \n\t" /* A2             a2 */\
-        "paddw $f2, $f2, $f8            \n\t" /* A2+B2          a2+b2 */\
-        "psubw $f10, $f10, $f8          \n\t" /* a2-B2          a2-b2 */\
-        "psraw $f2, $f2, $f18           \n\t"\
-        "psraw $f10, $f10, $f18         \n\t"\
-        "mov.d $f8, $f12                \n\t" /* A3             a3 */\
-        "paddw $f12, $f12, $f4          \n\t" /* A3+B3          a3+b3 */\
-        "psubw $f8, $f8, $f4            \n\t" /* a3-B3          a3-b3 */\
-        "psraw $f12, $f12, $f18         \n\t"\
-        "psraw $f8, $f8, $f18           \n\t"\
-        "packsswh $f2, $f2, $f2         \n\t" /* A2+B2  a2+b2 */\
-        "packsswh $f12, $f12, $f12      \n\t" /* A3+B3  a3+b3 */\
-        "swc1 $f2, 32+" #dst "          \n\t"\
-        "packsswh $f8, $f8, $f8         \n\t" /* A3-B3  a3-b3 */\
-        "packsswh $f10, $f10, $f10      \n\t" /* A2-B2  a2-b2 */\
-        "swc1 $f12, 48+" #dst "         \n\t"\
-        "swc1 $f8, 64+" #dst "          \n\t"\
-        "swc1 $f10, 80+" #dst "         \n\t"
-
-        //IDCT(  src0,   src4,   src1,    src5,    dst, shift)
-        IDCT(    (%1), 64(%1), 32(%1),  96(%1),  0(%0), 20)
-        IDCT(   8(%1), 72(%1), 40(%1), 104(%1),  4(%0), 20)
-        IDCT(  16(%1), 80(%1), 48(%1), 112(%1),  8(%0), 20)
-        IDCT(  24(%1), 88(%1), 56(%1), 120(%1), 12(%0), 20)
-        "b 9f                           \n\t"
-
-        "# .p2align 4                   \n\t"
-        "5:                             \n\t"
-
-#undef  IDCT
-#define IDCT(src0, src4, src1, src5, dst, shift) \
-        "ldc1 $f0, " #src0 "            \n\t" /* R4     R0      r4      r0 */\
-        "ldc1 $f2, " #src4 "            \n\t" /* R6     R2      r6      r2 */\
-        "ldc1 $f8, 16(%2)               \n\t" /* C4     C4      C4      C4 */\
-        "pmaddhw $f8, $f8, $f0          \n\t" /* C4R4+C4R0      C4r4+C4r0 */\
-        "ldc1 $f10, 24(%2)              \n\t" /* -C4    C4      -C4     C4 */\
-        "pmaddhw $f0, $f0, $f10         \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
-        "ldc1 $f10, 32(%2)              \n\t" /* C6     C2      C6      C2 */\
-        "pmaddhw $f10, $f10, $f2        \n\t" /* C6R6+C2R2      C6r6+C2r2 */\
-        "ldc1 $f12, 40(%2)              \n\t" /* -C2    C6      -C2     C6 */\
-        "pmaddhw $f2, $f2, $f12         \n\t" /* -C2R6+C6R2     -C2r6+C6r2 */\
-        "mov.d $f12, $f8                \n\t" /* C4R4+C4R0      C4r4+C4r0 */\
-        "paddw $f8, $f8, $f10           \n\t" /* A0             a0 */\
-        "psubw $f12, $f12, $f10         \n\t" /* A3             a3 */\
-        "mov.d $f10, $f0                \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
-        "paddw $f0, $f0, $f2            \n\t" /* A1             a1 */\
-        "psubw $f10, $f10, $f2          \n\t" /* A2             a2 */\
-        "ldc1 $f4, 8+" #src0 "          \n\t" /* R4     R0      r4      r0 */\
-        "ldc1 $f6, 8+" #src4 "          \n\t" /* R6     R2      r6      r2 */\
-        "ldc1 $f2, 16(%2)               \n\t" /* C4     C4      C4      C4 */\
-        "pmaddhw $f2, $f2, $f4          \n\t" /* C4R4+C4R0      C4r4+C4r0 */\
-        "ldc1 $f14, 24(%2)              \n\t" /* -C4    C4      -C4     C4 */\
-        "pmaddhw $f4, $f4, $f14         \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
-        "ldc1 $f14, 32(%2)              \n\t" /* C6     C2      C6      C2 */\
-        "ldc1 $f16, 40(%2)              \n\t"\
-        "pmaddhw $f14, $f14, $f6        \n\t" /* C6R6+C2R2      C6r6+C2r2 */\
-        "pmaddhw $f6, $f6, $f16         \n\t" /* -C2R6+C6R2     -C2r6+C6r2 */\
-        "paddw $f14, $f14, $f2          \n\t" /* A0             a0 */\
-        "paddw $f2, $f2, $f2            \n\t" /* 2C0            2c0 */\
-        "psubw $f2, $f2, $f14           \n\t" /* A3             a3 */\
-        "li $10, " #shift "             \n\t"\
-        "paddw $f6, $f6, $f4            \n\t" /* A1             a1 */\
-        "mtc1 $10, $f18                 \n\t"\
-        "paddw $f4, $f4, $f4            \n\t" /* 2C1            2c1 */\
-        "psubw $f4, $f4, $f6            \n\t" /* A2             a2 */\
-        "psraw $f8, $f8, $f18           \n\t"\
-        "psraw $f14, $f14, $f18         \n\t"\
-        "psraw $f6, $f6, $f18           \n\t"\
-        "packsswh $f8, $f8, $f14        \n\t" /* A0     a0 */\
-        "sdc1 $f8, " #dst "             \n\t"\
-        "psraw $f0, $f0, $f18           \n\t"\
-        "packsswh $f0, $f0, $f6         \n\t" /* A1     a1 */\
-        "sdc1 $f0, 16+" #dst "          \n\t"\
-        "sdc1 $f0, 96+" #dst "          \n\t"\
-        "sdc1 $f8, 112+" #dst "         \n\t"\
-        "psraw $f10, $f10, $f18         \n\t"\
-        "psraw $f12, $f12, $f18         \n\t"\
-        "psraw $f4, $f4, $f18           \n\t"\
-        "packsswh $f10, $f10, $f4       \n\t" /* A2-B2  a2-b2 */\
-        "sdc1 $f10, 32+" #dst "         \n\t"\
-        "psraw $f2, $f2, $f18           \n\t"\
-        "packsswh $f12, $f12, $f2       \n\t" /* A3+B3  a3+b3 */\
-        "sdc1 $f12, 48+" #dst "         \n\t"\
-        "sdc1 $f12, 64+" #dst "         \n\t"\
-        "sdc1 $f10, 80+" #dst "         \n\t"
-
-        //IDCT(  src0,   src4,   src1,    src5,    dst, shift)
-        IDCT(    0(%1), 64(%1), 32(%1),  96(%1),  0(%0), 20)
-        //IDCT(   8(%1), 72(%1), 40(%1), 104(%1),  4(%0), 20)
-        IDCT(  16(%1), 80(%1), 48(%1), 112(%1),  8(%0), 20)
-        //IDCT(  24(%1), 88(%1), 56(%1), 120(%1), 12(%0), 20)
-        "b 9f                           \n\t"
-
-        "# .p2align 4                   \n\t"
-        "1:                             \n\t"
-
-#undef  IDCT
-#define IDCT(src0, src4, src1, src5, dst, shift) \
-        "ldc1 $f0, " #src0 "            \n\t" /* R4     R0      r4      r0 */\
-        "ldc1 $f2, " #src4 "            \n\t" /* R6     R2      r6      r2 */\
-        "ldc1 $f4, " #src1 "            \n\t" /* R3     R1      r3      r1 */\
-        "ldc1 $f8, 16(%2)               \n\t" /* C4     C4      C4      C4 */\
-        "li $10, " #shift "             \n\t"\
-        "pmaddhw $f8, $f8, $f0          \n\t" /* C4R4+C4R0      C4r4+C4r0 */\
-        "ldc1 $f10, 24(%2)              \n\t" /* -C4    C4      -C4     C4 */\
-        "mtc1 $10, $f18                 \n\t"\
-        "pmaddhw $f0, $f0, $f10         \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
-        "ldc1 $f10, 32(%2)              \n\t" /* C6     C2      C6      C2 */\
-        "pmaddhw $f10, $f10, $f2        \n\t" /* C6R6+C2R2      C6r6+C2r2 */\
-        "ldc1 $f12, 40(%2)              \n\t" /* -C2    C6      -C2     C6 */\
-        "pmaddhw $f2, $f2, $f12         \n\t" /* -C2R6+C6R2     -C2r6+C6r2 */\
-        "mov.d $f12, $f8                \n\t" /* C4R4+C4R0      C4r4+C4r0 */\
-        "ldc1 $f14, 48(%2)              \n\t" /* C3     C1      C3      C1 */\
-        "pmaddhw $f14, $f14, $f4        \n\t" /* C3R3+C1R1      C3r3+C1r1 */\
-        "paddw $f8, $f8, $f10           \n\t" /* A0             a0 */\
-        "psubw $f12, $f12, $f10         \n\t" /* A3             a3 */\
-        "mov.d $f10, $f0                \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
-        "paddw $f0, $f0, $f2            \n\t" /* A1             a1 */\
-        "psubw $f10, $f10, $f2          \n\t" /* A2             a2 */\
-        "ldc1 $f2, 64(%2)               \n\t"\
-        "pmaddhw $f2, $f2, $f4          \n\t" /* -C7R3+C3R1     -C7r3+C3r1 */\
-        "paddw $f14, $f14, $f8          \n\t" /* A0+B0          a0+b0 */\
-        "paddw $f8, $f8, $f8            \n\t" /* 2A0            2a0 */\
-        "psubw $f8, $f8, $f14           \n\t" /* A0-B0          a0-b0 */\
-        "psraw $f14, $f14, $f18         \n\t"\
-        "psraw $f8, $f8, $f18           \n\t"\
-        "mov.d $f6, $f0                 \n\t" /* A1             a1 */\
-        "paddw $f0, $f0, $f2            \n\t" /* A1+B1          a1+b1 */\
-        "psubw $f6, $f6, $f2            \n\t" /* A1-B1          a1-b1 */\
-        "psraw $f0, $f0, $f18           \n\t"\
-        "psraw $f6, $f6, $f18           \n\t"\
-        "packsswh $f14, $f14, $f14      \n\t" /* A0+B0  a0+b0 */\
-        "swc1 $f14, " #dst "            \n\t"\
-        "packsswh $f0, $f0, $f0         \n\t" /* A1+B1  a1+b1 */\
-        "swc1 $f0, 16+" #dst "          \n\t"\
-        "packsswh $f6, $f6, $f6         \n\t" /* A1-B1  a1-b1 */\
-        "swc1 $f6, 96+" #dst "          \n\t"\
-        "packsswh $f8, $f8, $f8         \n\t" /* A0-B0  a0-b0 */\
-        "swc1 $f8, 112+" #dst "         \n\t"\
-        "ldc1 $f8, 80(%2)               \n\t" /* -C1    C5      -C1     C5 */\
-        "ldc1 $f16, 96(%2)              \n\t"\
-        "pmaddhw $f8, $f8, $f4          \n\t" /* -C1R3+C5R1     -C1r3+C5r1 */\
-        "pmaddhw $f4, $f4, $f16         \n\t" /* -C5R3+C7R1     -C5r3+C7r1 */\
-        "mov.d $f6, $f10                \n\t" /* A2             a2 */\
-        "paddw $f6, $f6, $f8            \n\t" /* A2+B2          a2+b2 */\
-        "psubw $f10, $f10, $f8          \n\t" /* a2-B2          a2-b2 */\
-        "psraw $f6, $f6, $f18           \n\t"\
-        "psraw $f10, $f10, $f18         \n\t"\
-        "mov.d $f8, $f12                \n\t" /* A3             a3 */\
-        "paddw $f12, $f12, $f4          \n\t" /* A3+B3          a3+b3 */\
-        "psubw $f8, $f8, $f4            \n\t" /* a3-B3          a3-b3 */\
-        "psraw $f12, $f12, $f18         \n\t"\
-        "packsswh $f6, $f6, $f6         \n\t" /* A2+B2  a2+b2 */\
-        "swc1 $f6, 32+" #dst "          \n\t"\
-        "psraw $f8, $f8, $f18           \n\t"\
-        "packsswh $f12, $f12, $f12      \n\t" /* A3+B3  a3+b3 */\
-        "swc1 $f12, 48+" #dst "         \n\t"\
-        "packsswh $f8, $f8, $f8         \n\t" /* A3-B3  a3-b3 */\
-        "packsswh $f10, $f10, $f10      \n\t" /* A2-B2  a2-b2 */\
-        "swc1 $f8, 64+" #dst "          \n\t"\
-        "swc1 $f10, 80+" #dst "         \n\t"
-
-        //IDCT(  src0,   src4,   src1,    src5,    dst, shift)
-        IDCT(    (%1), 64(%1), 32(%1),  96(%1),  0(%0), 20)
-        IDCT(   8(%1), 72(%1), 40(%1), 104(%1),  4(%0), 20)
-        IDCT(  16(%1), 80(%1), 48(%1), 112(%1),  8(%0), 20)
-        IDCT(  24(%1), 88(%1), 56(%1), 120(%1), 12(%0), 20)
-        "b 9f                           \n\t"
-
-        "# .p2align 4                   \n\t"
-        "7:                             \n\t"
-
-#undef  IDCT
-#define IDCT(src0, src4, src1, src5, dst, shift) \
-        "ldc1 $f0, " #src0 "            \n\t" /* R4     R0      r4      r0 */\
-        "ldc1 $f8, 16(%2)               \n\t" /* C4     C4      C4      C4 */\
-        "li $10, " #shift "             \n\t"\
-        "pmaddhw $f8, $f8, $f0          \n\t" /* C4R4+C4R0      C4r4+C4r0 */\
-        "mtc1 $10, $f18                 \n\t"\
-        "ldc1 $f10, 24(%2)              \n\t" /* -C4    C4      -C4     C4 */\
-        "pmaddhw $f0, $f0, $f10         \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
-        "psraw $f8, $f8, $f18           \n\t"\
-        "psraw $f0, $f0, $f18           \n\t"\
-        "ldc1 $f4, 8+" #src0 "          \n\t" /* R4     R0      r4      r0 */\
-        "ldc1 $f2, 16(%2)               \n\t" /* C4     C4      C4      C4 */\
-        "pmaddhw $f2, $f2, $f4          \n\t" /* C4R4+C4R0      C4r4+C4r0 */\
-        "ldc1 $f14, 24(%2)              \n\t" /* -C4    C4      -C4     C4 */\
-        "pmaddhw $f4, $f4, $f14         \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
-        "ldc1 $f14, 32(%2)              \n\t" /* C6     C2      C6      C2 */\
-        "psraw $f2, $f2, $f18           \n\t"\
-        "packsswh $f8, $f8, $f2         \n\t" /* A0     a0 */\
-        "sdc1 $f8, " #dst "             \n\t"\
-        "psraw $f4, $f4, $f18           \n\t"\
-        "packsswh $f0, $f0, $f4         \n\t" /* A1     a1 */\
-        "sdc1 $f0, 16+" #dst "          \n\t"\
-        "sdc1 $f0, 96+" #dst "          \n\t"\
-        "sdc1 $f8, 112+" #dst "         \n\t"\
-        "sdc1 $f0, 32+" #dst "          \n\t"\
-        "sdc1 $f8, 48+" #dst "          \n\t"\
-        "sdc1 $f8, 64+" #dst "          \n\t"\
-        "sdc1 $f0, 80+" #dst "          \n\t"
-
-        //IDCT(  src0,   src4,   src1,    src5,    dst, shift)
-        IDCT(   0(%1), 64(%1), 32(%1),  96(%1),  0(%0), 20)
-        //IDCT(   8(%1), 72(%1), 40(%1), 104(%1),  4(%0), 20)
-        IDCT(  16(%1), 80(%1), 48(%1), 112(%1),  8(%0), 20)
-        //IDCT(  24(%1), 88(%1), 56(%1), 120(%1), 12(%0), 20)
-
-        "9:                             \n\t"
-        ::"r"(block),"r"(temp),"r"(coeffs),"m"(ff_wm1010),"m"(ff_d40000)
-        : "$10","$11"
-    );
-}
-
 void ff_put_pixels_clamped_mmi(const int16_t *block,
         uint8_t *av_restrict pixels, ptrdiff_t line_size)
 {
@@ -982,20 +188,3 @@ void ff_add_pixels_clamped_mmi(const int16_t *block,
         p += 16;
     } while (--i);
 }
-
-void ff_simple_idct_mmi(int16_t *block)
-{
-    simple_idct_mmi(block);
-}
-
-void ff_simple_idct_put_mmi(uint8_t *dest, int32_t line_size, int16_t *block)
-{
-    simple_idct_mmi(block);
-    ff_put_pixels_clamped_mmi(block, dest, line_size);
-}
-
-void ff_simple_idct_add_mmi(uint8_t *dest, int32_t line_size, int16_t *block)
-{
-    simple_idct_mmi(block);
-    ff_add_pixels_clamped_mmi(block, dest, line_size);
-}
diff --git a/libavcodec/mips/simple_idct_mmi.c b/libavcodec/mips/simple_idct_mmi.c
new file mode 100644 (file)
index 0000000..3c1a4f7
--- /dev/null
@@ -0,0 +1,833 @@
+/*
+ * Loongson SIMD optimized simple idct
+ *
+ * Copyright (c) 2015 Loongson Technology Corporation Limited
+ * Copyright (c) 2015 Zhou Xiaoyong <zhouxiaoyong@loongson.cn>
+ *                    Zhang Shuangshuang <zhangshuangshuang@ict.ac.cn>
+ *
+ * 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 "idctdsp_mips.h"
+#include "constants.h"
+
+#define C0 23170 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
+#define C1 22725 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
+#define C2 21407 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
+#define C3 19266 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
+#define C4 16383 //cos(i*M_PI/16)*sqrt(2)*(1<<14) - 0.5
+#define C5 12873 //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
+#define C6 8867  //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
+#define C7 4520  //cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5
+
+#define ROW_SHIFT 11
+#define COL_SHIFT 20
+
+DECLARE_ALIGNED(8, static const int16_t, coeffs)[]= {
+    1<<(ROW_SHIFT-1),   0, 1<<(ROW_SHIFT-1),   0,
+    1<<(ROW_SHIFT-1),   1, 1<<(ROW_SHIFT-1),   0,
+                  C4,  C4,               C4,  C4,
+                  C4, -C4,               C4, -C4,
+                  C2,  C6,               C2,  C6,
+                  C6, -C2,               C6, -C2,
+                  C1,  C3,               C1,  C3,
+                  C5,  C7,               C5,  C7,
+                  C3, -C7,               C3, -C7,
+                 -C1, -C5,              -C1, -C5,
+                  C5, -C1,               C5, -C1,
+                  C7,  C3,               C7,  C3,
+                  C7, -C5,               C7, -C5,
+                  C3, -C1,               C3, -C1
+};
+
+static void simple_idct_mmi(int16_t *block)
+{
+        DECLARE_ALIGNED(8, int64_t, align_tmp)[16];
+        int16_t * const temp= (int16_t*)align_tmp;
+
+        __asm__ volatile (
+#undef  DC_COND_IDCT
+#define DC_COND_IDCT(src0, src4, src1, src5, dst, rounder, rarg, shift)      \
+        "ldc1 $f0, " #src0 "            \n\t" /* R4     R0      r4      r0 */\
+        "ldc1 $f2, " #src4 "            \n\t" /* R6     R2      r6      r2 */\
+        "ldc1 $f4, " #src1 "            \n\t" /* R3     R1      r3      r1 */\
+        "ldc1 $f6, " #src5 "            \n\t" /* R7     R5      r7      r5 */\
+        "ldc1 $f8, %3                   \n\t"                                \
+        "and  $f8, $f8, $f0             \n\t"                                \
+        "or $f8, $f8, $f2               \n\t"                                \
+        "or $f8, $f8, $f4               \n\t"                                \
+        "or $f8, $f8, $f6               \n\t"                                \
+        "packsswh $f8, $f8, $f8         \n\t"                                \
+        "li $11, " #shift "             \n\t"                                \
+        "mfc1 $10, $f8                  \n\t"                                \
+        "mtc1 $11, $f18                 \n\t"                                \
+        "beqz $10, 1f                   \n\t"                                \
+        "ldc1 $f8, 16(%2)               \n\t" /* C4     C4      C4      C4 */\
+        "pmaddhw $f8, $f8, $f0          \n\t" /* C4R4+C4R0      C4r4+C4r0  */\
+        "ldc1 $f10, 24(%2)              \n\t" /* -C4    C4      -C4     C4 */\
+        "pmaddhw $f0, $f0, $f10         \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
+        "ldc1 $f10, 32(%2)              \n\t" /* C6     C2      C6      C2 */\
+        "pmaddhw $f10, $f10, $f2        \n\t" /* C6R6+C2R2      C6r6+C2r2  */\
+        "ldc1 $f12, 40(%2)              \n\t" /* -C2    C6      -C2     C6 */\
+        "pmaddhw $f2, $f2, $f12         \n\t" /* -C2R6+C6R2     -C2r6+C6r2 */\
+        "ldc1 $f14, 48(%2)              \n\t" /* C3     C1      C3      C1 */\
+        "ldc1 $f16, " #rarg "           \n\t"                                \
+        "pmaddhw $f14, $f14, $f4        \n\t" /* C3R3+C1R1      C3r3+C1r1  */\
+        #rounder " $f8, $f8, $f16       \n\t"                                \
+        "mov.d $f12, $f8                \n\t" /* C4R4+C4R0      C4r4+C4r0  */\
+        "paddw $f8, $f8, $f10           \n\t" /* A0             a0         */\
+        "psubw $f12, $f12, $f10         \n\t" /* A3             a3         */\
+        "ldc1 $f10, 56(%2)              \n\t" /* C7     C5      C7      C5 */\
+        "ldc1 $f16, " #rarg "           \n\t"                                \
+        "pmaddhw $f10, $f10, $f6        \n\t" /* C7R7+C5R5      C7r7+C5r5  */\
+        #rounder " $f0, $f0, $f16       \n\t"                                \
+        "paddw $f2, $f2, $f0            \n\t" /* A1             a1         */\
+        "ldc1 $f16, 64(%2)              \n\t"                                \
+        "paddw $f0, $f0, $f0            \n\t"                                \
+        "psubw $f0, $f0, $f2            \n\t" /* A2             a2         */\
+        "pmaddhw $f4, $f4, $f16         \n\t" /* -C7R3+C3R1     -C7r3+C3r1 */\
+        "paddw $f14, $f14, $f10         \n\t" /* B0             b0         */\
+        "ldc1 $f10, 72(%2)              \n\t" /* -C5    -C1     -C5    -C1 */\
+        "pmaddhw $f10, $f10, $f6        \n\t" /* -C5R7-C1R5     -C5r7-C1r5 */\
+        "paddw $f14, $f14, $f8          \n\t" /* A0+B0          a0+b0      */\
+        "paddw $f8, $f8, $f8            \n\t" /* 2A0            2a0        */\
+        "psubw $f8, $f8, $f14           \n\t" /* A0-B0          a0-b0      */\
+        "paddw $f10, $f10, $f4          \n\t" /* B1             b1         */\
+        "psraw $f14, $f14, $f18         \n\t"                                \
+        "psraw $f8, $f8, $f18           \n\t"                                \
+        "mov.d $f4, $f2                 \n\t" /* A1             a1         */\
+        "paddw $f2, $f2, $f10           \n\t" /* A1+B1          a1+b1      */\
+        "psubw $f4, $f4, $f10           \n\t" /* A1-B1          a1-b1      */\
+        "psraw $f2, $f2, $f18           \n\t"                                \
+        "psraw $f4, $f4, $f18           \n\t"                                \
+        "packsswh $f14, $f14, $f2       \n\t" /* A1+B1 a1+b1 A0+B0 a0+b0   */\
+        "packsswh $f4, $f4, $f8         \n\t" /* A0-B0 a0-b0 A1-B1 a1-b1   */\
+        "sdc1 $f14, " #dst "            \n\t"                                \
+        "ldc1 $f2, " #src1 "            \n\t" /* R3     R1      r3      r1 */\
+        "ldc1 $f8, 80(%2)               \n\t" /* -C1    C5      -C1     C5 */\
+        "sdc1 $f4, 24+" #dst "          \n\t"                                \
+        "pmaddhw $f8, $f8, $f2          \n\t" /* -C1R3+C5R1     -C1r3+C5r1 */\
+        "ldc1 $f16, 96(%2)              \n\t"                                \
+        "ldc1 $f14, 88(%2)              \n\t" /* C3     C7      C3      C7 */\
+        "pmaddhw $f2, $f2, $f16         \n\t" /* -C5R3+C7R1     -C5r3+C7r1 */\
+        "pmaddhw $f14, $f14, $f6        \n\t" /* C3R7+C7R5      C3r7+C7r5  */\
+        "ldc1 $f16, 104(%2)             \n\t"                                \
+        "mov.d $f4, $f0                 \n\t" /* A2             a2         */\
+        "pmaddhw $f6, $f6, $f16         \n\t" /* -C1R7+C3R5     -C1r7+C3r5 */\
+        "paddw $f8, $f8, $f14           \n\t" /* B2             b2         */\
+        "paddw $f4, $f4, $f8            \n\t" /* A2+B2          a2+b2      */\
+        "psubw $f0, $f0, $f8            \n\t" /* a2-B2          a2-b2      */\
+        "psraw $f4, $f4, $f18           \n\t"                                \
+        "psraw $f0, $f0, $f18           \n\t"                                \
+        "mov.d $f8, $f12                \n\t" /* A3             a3         */\
+        "paddw $f6, $f6, $f2            \n\t" /* B3             b3         */\
+        "paddw $f12, $f12, $f6          \n\t" /* A3+B3          a3+b3      */\
+        "psubw $f8, $f8, $f6            \n\t" /* a3-B3          a3-b3      */\
+        "psraw $f12, $f12, $f18         \n\t"                                \
+        "packsswh $f4, $f4, $f12        \n\t" /* A3+B3 a3+b3 A2+B2 a2+b2   */\
+        "sdc1 $f4, 8+" #dst "           \n\t"                                \
+        "psraw $f8, $f8, $f18           \n\t"                                \
+        "packsswh $f8, $f8, $f0         \n\t" /* A2-B2 a2-b2 A3-B3 a3-b3   */\
+        "sdc1 $f8, 16+" #dst "          \n\t"                                \
+        "b 2f                           \n\t"                                \
+        "1:                             \n\t"                                \
+        "li $10, 16                     \n\t"                                \
+        "mtc1 $10, $f16                 \n\t"                                \
+        "psllw $f0, $f0, $f16           \n\t"                                \
+        "ldc1 $f16, %4                  \n\t"                                \
+        "paddw $f0, $f0, $f16           \n\t"                                \
+        "li $10, 13                     \n\t"                                \
+        "mtc1 $10, $f16                 \n\t"                                \
+        "psraw $f0, $f0, $f16           \n\t"                                \
+        "packsswh $f0, $f0, $f0         \n\t"                                \
+        "sdc1 $f0, " #dst "             \n\t"                                \
+        "sdc1 $f0, 8+" #dst "           \n\t"                                \
+        "sdc1 $f0, 16+" #dst "          \n\t"                                \
+        "sdc1 $f0, 24+" #dst "          \n\t"                                \
+        "2:                             \n\t"
+
+#undef  Z_COND_IDCT
+#define Z_COND_IDCT(src0, src4, src1, src5, dst, rounder, rarg, shift, bt)   \
+        "ldc1 $f0, " #src0 "            \n\t" /* R4     R0      r4      r0 */\
+        "ldc1 $f2, " #src4 "            \n\t" /* R6     R2      r6      r2 */\
+        "ldc1 $f4, " #src1 "            \n\t" /* R3     R1      r3      r1 */\
+        "ldc1 $f6, " #src5 "            \n\t" /* R7     R5      r7      r5 */\
+        "mov.d $f8, $f0                 \n\t"                                \
+        "or $f8, $f8, $f2               \n\t"                                \
+        "or $f8, $f8, $f4               \n\t"                                \
+        "or $f8, $f8, $f6               \n\t"                                \
+        "packsswh $f8, $f8, $f8         \n\t"                                \
+        "mfc1 $10, $f8                  \n\t"                                \
+        "beqz $10, " #bt "              \n\t"                                \
+        "ldc1 $f8, 16(%2)               \n\t" /* C4     C4      C4      C4 */\
+        "pmaddhw $f8, $f8, $f0          \n\t" /* C4R4+C4R0      C4r4+C4r0  */\
+        "ldc1 $f10, 24(%2)              \n\t" /* -C4    C4      -C4     C4 */\
+        "pmaddhw $f0, $f0, $f10         \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
+        "ldc1 $f10, 32(%2)              \n\t" /* C6     C2      C6      C2 */\
+        "pmaddhw $f10, $f10, $f2        \n\t" /* C6R6+C2R2      C6r6+C2r2  */\
+        "ldc1 $f12, 40(%2)              \n\t" /* -C2    C6      -C2     C6 */\
+        "pmaddhw $f2, $f2, $f12         \n\t" /* -C2R6+C6R2     -C2r6+C6r2 */\
+        "ldc1 $f14, 48(%2)              \n\t" /* C3     C1      C3      C1 */\
+        "ldc1 $f16, " #rarg "           \n\t"                                \
+        "pmaddhw $f14, $f14, $f4        \n\t" /* C3R3+C1R1      C3r3+C1r1  */\
+        #rounder " $f8, $f8, $f16       \n\t"                                \
+        "mov.d $f12, $f8                \n\t" /* C4R4+C4R0      C4r4+C4r0  */\
+        "paddw $f8, $f8, $f10           \n\t" /* A0             a0         */\
+        "psubw $f12, $f12, $f10         \n\t" /* A3             a3         */\
+        "ldc1 $f10, 56(%2)              \n\t" /* C7     C5      C7      C5 */\
+        "ldc1 $f16, " #rarg "           \n\t"                                \
+        "pmaddhw $f10, $f10, $f6        \n\t" /* C7R7+C5R5      C7r7+C5r5  */\
+        #rounder " $f0, $f0, $f16       \n\t"                                \
+        "paddw $f2, $f2, $f0            \n\t" /* A1             a1         */\
+        "paddw $f0, $f0, $f0            \n\t"                                \
+        "ldc1 $f16, 64(%2)              \n\t"                                \
+        "psubw $f0, $f0, $f2            \n\t" /* A2             a2         */\
+        "pmaddhw $f4, $f4, $f16         \n\t" /* -C7R3+C3R1     -C7r3+C3r1 */\
+        "paddw $f14, $f14, $f10         \n\t" /* B0             b0         */\
+        "ldc1 $f10, 72(%2)              \n\t" /* -C5    -C1     -C5    -C1 */\
+        "pmaddhw $f10, $f10, $f6        \n\t" /* -C5R7-C1R5     -C5r7-C1r5 */\
+        "paddw $f14, $f14, $f8          \n\t" /* A0+B0          a0+b0      */\
+        "paddw $f8, $f8, $f8            \n\t" /* 2A0            2a0        */\
+        "li $10, " #shift "             \n\t"                                \
+        "psubw $f8, $f8, $f14           \n\t" /* A0-B0          a0-b0      */\
+        "mtc1 $10, $f18                 \n\t"                                \
+        "paddw $f10, $f10, $f4          \n\t" /* B1             b1         */\
+        "psraw $f14, $f14, $f18         \n\t"                                \
+        "psraw $f8, $f8, $f18           \n\t"                                \
+        "mov.d $f4, $f2                 \n\t" /* A1             a1         */\
+        "paddw $f2, $f2, $f10           \n\t" /* A1+B1          a1+b1      */\
+        "psubw $f4, $f4, $f10           \n\t" /* A1-B1          a1-b1      */\
+        "psraw $f2, $f2, $f18           \n\t"                                \
+        "psraw $f4, $f4, $f18           \n\t"                                \
+        "packsswh $f14, $f14, $f2       \n\t" /* A1+B1 a1+b1 A0+B0 a0+b0   */\
+        "packsswh $f4, $f4, $f8         \n\t" /* A0-B0 a0-b0 A1-B1 a1-b1   */\
+        "sdc1 $f14, " #dst "            \n\t"                                \
+        "ldc1 $f2, " #src1 "            \n\t" /* R3     R1      r3      r1 */\
+        "ldc1 $f8, 80(%2)               \n\t" /* -C1    C5      -C1     C5 */\
+        "sdc1 $f4, 24+" #dst "          \n\t"                                \
+        "pmaddhw $f8, $f8, $f2          \n\t" /* -C1R3+C5R1     -C1r3+C5r1 */\
+        "ldc1 $f16, 96(%2)              \n\t"                                \
+        "ldc1 $f14, 88(%2)              \n\t" /* C3     C7      C3      C7 */\
+        "pmaddhw $f2, $f2, $f16         \n\t" /* -C5R3+C7R1     -C5r3+C7r1 */\
+        "pmaddhw $f14, $f14, $f6        \n\t" /* C3R7+C7R5      C3r7+C7r5  */\
+        "ldc1 $f16, 104(%2)             \n\t"                                \
+        "mov.d $f4, $f0                 \n\t" /* A2             a2         */\
+        "pmaddhw $f6, $f6, $f16         \n\t" /* -C1R7+C3R5     -C1r7+C3r5 */\
+        "paddw $f8, $f8, $f14           \n\t" /* B2             b2         */\
+        "paddw $f4, $f4, $f8            \n\t" /* A2+B2          a2+b2      */\
+        "psubw $f0, $f0, $f8            \n\t" /* a2-B2          a2-b2      */\
+        "psraw $f4, $f4, $f18           \n\t"                                \
+        "psraw $f0, $f0, $f18           \n\t"                                \
+        "mov.d $f8, $f12                \n\t" /* A3             a3         */\
+        "paddw $f6, $f6, $f2            \n\t" /* B3             b3         */\
+        "paddw $f12, $f12, $f6          \n\t" /* A3+B3          a3+b3      */\
+        "psubw $f8, $f8, $f6            \n\t" /* a3-B3          a3-b3      */\
+        "psraw $f12, $f12, $f18         \n\t"                                \
+        "packsswh $f4, $f4, $f12        \n\t" /* A3+B3 a3+b3 A2+B2 a2+b2   */\
+        "sdc1 $f4, 8+" #dst "           \n\t"                                \
+        "psraw $f8, $f8, $f18           \n\t"                                \
+        "packsswh $f8, $f8, $f0         \n\t" /* A2-B2 a2-b2 A3-B3 a3-b3   */\
+        "sdc1 $f8, 16+" #dst "          \n\t"                                \
+
+        //IDCT(       src0,   src4,   src1,   src5,    dst,     rounder, shift)
+        DC_COND_IDCT(0(%0),  8(%0), 16(%0), 24(%0),  0(%1), paddw,8(%2), 11)
+        Z_COND_IDCT(32(%0), 40(%0), 48(%0), 56(%0), 32(%1), paddw,(%2), 11, 4f)
+        Z_COND_IDCT(64(%0), 72(%0), 80(%0), 88(%0), 64(%1), paddw,(%2), 11, 2f)
+        Z_COND_IDCT(96(%0),104(%0),112(%0),120(%0), 96(%1), paddw,(%2), 11, 1f)
+
+#undef  IDCT
+#define IDCT(src0, src4, src1, src5, dst, shift)                             \
+        "ldc1 $f0, " #src0 "            \n\t" /* R4     R0      r4      r0 */\
+        "ldc1 $f2, " #src4 "            \n\t" /* R6     R2      r6      r2 */\
+        "ldc1 $f4, " #src1 "            \n\t" /* R3     R1      r3      r1 */\
+        "ldc1 $f6, " #src5 "            \n\t" /* R7     R5      r7      r5 */\
+        "ldc1 $f8, 16(%2)               \n\t" /* C4     C4      C4      C4 */\
+        "pmaddhw $f8, $f8, $f0          \n\t" /* C4R4+C4R0      C4r4+C4r0  */\
+        "ldc1 $f10, 24(%2)              \n\t" /* -C4    C4      -C4     C4 */\
+        "pmaddhw $f0, $f0, $f10         \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
+        "ldc1 $f10, 32(%2)              \n\t" /* C6     C2      C6      C2 */\
+        "pmaddhw $f10, $f10, $f2        \n\t" /* C6R6+C2R2      C6r6+C2r2  */\
+        "ldc1 $f12, 40(%2)              \n\t" /* -C2    C6      -C2     C6 */\
+        "pmaddhw $f2, $f2, $f12         \n\t" /* -C2R6+C6R2     -C2r6+C6r2 */\
+        "mov.d $f12, $f8                \n\t" /* C4R4+C4R0      C4r4+C4r0  */\
+        "ldc1 $f14, 48(%2)              \n\t" /* C3     C1      C3      C1 */\
+        "pmaddhw $f14, $f14, $f4        \n\t" /* C3R3+C1R1      C3r3+C1r1  */\
+        "paddw $f8, $f8, $f10           \n\t" /* A0             a0         */\
+        "psubw $f12, $f12, $f10         \n\t" /* A3             a3         */\
+        "mov.d $f10, $f0                \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
+        "paddw $f0, $f0, $f2            \n\t" /* A1             a1         */\
+        "psubw $f10, $f10, $f2          \n\t" /* A2             a2         */\
+        "ldc1 $f2, 56(%2)               \n\t" /* C7     C5      C7      C5 */\
+        "ldc1 $f16, 64(%2)              \n\t"                                \
+        "pmaddhw $f2, $f2, $f6          \n\t" /* C7R7+C5R5      C7r7+C5r5  */\
+        "pmaddhw $f4, $f4, $f16         \n\t" /* -C7R3+C3R1     -C7r3+C3r1 */\
+        "li $10, " #shift "             \n\t"                                \
+        "paddw $f14, $f14, $f2          \n\t" /* B0             b0         */\
+        "ldc1 $f2, 72(%2)               \n\t" /* -C5    -C1     -C5    -C1 */\
+        "mtc1 $10, $f18                 \n\t"                                \
+        "pmaddhw $f2, $f2, $f6          \n\t" /* -C5R7-C1R5     -C5r7-C1r5 */\
+        "paddw $f14, $f14, $f8          \n\t" /* A0+B0          a0+b0      */\
+        "paddw $f8, $f8, $f8            \n\t" /* 2A0            2a0        */\
+        "psubw $f8, $f8, $f14           \n\t" /* A0-B0          a0-b0      */\
+        "paddw $f2, $f2, $f4            \n\t" /* B1             b1         */\
+        "psraw $f14, $f14, $f18         \n\t"                                \
+        "psraw $f8, $f8, $f18           \n\t"                                \
+        "mov.d $f4, $f0                 \n\t" /* A1             a1         */\
+        "paddw $f0, $f0, $f2            \n\t" /* A1+B1          a1+b1      */\
+        "psubw $f4, $f4, $f2            \n\t" /* A1-B1          a1-b1      */\
+        "psraw $f0, $f0, $f18           \n\t"                                \
+        "psraw $f4, $f4, $f18           \n\t"                                \
+        "packsswh $f14, $f14, $f14      \n\t" /* A0+B0          a0+b0      */\
+        "swc1 $f14, " #dst "            \n\t"                                \
+        "packsswh $f0, $f0, $f0         \n\t" /* A1+B1          a1+b1      */\
+        "swc1 $f0, 16+" #dst "          \n\t"                                \
+        "packsswh $f4, $f4, $f4         \n\t" /* A1-B1          a1-b1      */\
+        "swc1 $f4, 96+" #dst "          \n\t"                                \
+        "packsswh $f8, $f8, $f8         \n\t" /* A0-B0          a0-b0      */\
+        "swc1 $f8, 112+" #dst "         \n\t"                                \
+        "ldc1 $f0, " #src1 "            \n\t" /* R3     R1      r3      r1 */\
+        "ldc1 $f8, 80(%2)               \n\t" /* -C1    C5      -C1     C5 */\
+        "pmaddhw $f8, $f8, $f0          \n\t" /* -C1R3+C5R1     -C1r3+C5r1 */\
+        "ldc1 $f16, 96(%2)              \n\t"                                \
+        "ldc1 $f14, 88(%2)              \n\t" /* C3     C7      C3      C7 */\
+        "pmaddhw $f0, $f0, $f16         \n\t" /* -C5R3+C7R1     -C5r3+C7r1 */\
+        "pmaddhw $f14, $f14, $f6        \n\t" /* C3R7+C7R5      C3r7+C7r5  */\
+        "ldc1 $f16, 104(%2)             \n\t"                                \
+        "mov.d $f4, $f10                \n\t" /* A2             a2         */\
+        "pmaddhw $f6, $f6, $f16         \n\t" /* -C1R7+C3R5     -C1r7+C3r5 */\
+        "paddw $f8, $f8, $f14           \n\t" /* B2             b2         */\
+        "paddw $f4, $f4, $f8            \n\t" /* A2+B2          a2+b2      */\
+        "psubw $f10, $f10, $f8          \n\t" /* a2-B2          a2-b2      */\
+        "psraw $f4, $f4, $f18           \n\t"                                \
+        "psraw $f10, $f10, $f18         \n\t"                                \
+        "mov.d $f8, $f12                \n\t" /* A3             a3         */\
+        "paddw $f6, $f6, $f0            \n\t" /* B3             b3         */\
+        "paddw $f12, $f12, $f6          \n\t" /* A3+B3          a3+b3      */\
+        "psubw $f8, $f8, $f6            \n\t" /* a3-B3          a3-b3      */\
+        "psraw $f12, $f12, $f18         \n\t"                                \
+        "psraw $f8, $f8, $f18           \n\t"                                \
+        "packsswh $f4, $f4, $f4         \n\t" /* A2+B2          a2+b2      */\
+        "packsswh $f12, $f12, $f12      \n\t" /* A3+B3          a3+b3      */\
+        "swc1 $f4, 32+" #dst "          \n\t"                                \
+        "packsswh $f8, $f8, $f8         \n\t" /* A3-B3          a3-b3      */\
+        "packsswh $f10, $f10, $f10      \n\t" /* A2-B2          a2-b2      */\
+        "swc1 $f12, 48+" #dst "         \n\t"                                \
+        "swc1 $f8, 64+" #dst "          \n\t"                                \
+        "swc1 $f10, 80+" #dst "         \n\t"
+
+        //IDCT(  src0,   src4,   src1,    src5,    dst, shift)
+        IDCT(    (%1), 64(%1), 32(%1),  96(%1),  0(%0),    20)
+        IDCT(   8(%1), 72(%1), 40(%1), 104(%1),  4(%0),    20)
+        IDCT(  16(%1), 80(%1), 48(%1), 112(%1),  8(%0),    20)
+        IDCT(  24(%1), 88(%1), 56(%1), 120(%1), 12(%0),    20)
+        "b 9f                           \n\t"
+
+        "# .p2align 4                   \n\t"
+        "4:                             \n\t"
+        Z_COND_IDCT(64(%0), 72(%0), 80(%0), 88(%0), 64(%1),paddw,(%2), 11, 6f)
+        Z_COND_IDCT(96(%0),104(%0),112(%0),120(%0), 96(%1),paddw,(%2), 11, 5f)
+
+#undef  IDCT
+#define IDCT(src0, src4, src1, src5, dst, shift)                             \
+        "ldc1 $f0, " #src0 "            \n\t" /* R4     R0      r4      r0 */\
+        "ldc1 $f2, " #src4 "            \n\t" /* R6     R2      r6      r2 */\
+        "ldc1 $f6, " #src5 "            \n\t" /* R7     R5      r7      r5 */\
+        "ldc1 $f8, 16(%2)               \n\t" /* C4     C4      C4      C4 */\
+        "pmaddhw $f8, $f8, $f0          \n\t" /* C4R4+C4R0      C4r4+C4r0  */\
+        "ldc1 $f10, 24(%2)              \n\t" /* -C4    C4      -C4     C4 */\
+        "pmaddhw $f0, $f0, $f10         \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
+        "ldc1 $f10, 32(%2)              \n\t" /* C6     C2      C6      C2 */\
+        "pmaddhw $f10, $f10, $f2        \n\t" /* C6R6+C2R2      C6r6+C2r2  */\
+        "ldc1 $f12, 40(%2)              \n\t" /* -C2    C6      -C2     C6 */\
+        "pmaddhw $f2, $f2, $f12         \n\t" /* -C2R6+C6R2     -C2r6+C6r2 */\
+        "mov.d $f12, $f8                \n\t" /* C4R4+C4R0      C4r4+C4r0  */\
+        "paddw $f8, $f8, $f10           \n\t" /* A0             a0         */\
+        "psubw $f12, $f12, $f10         \n\t" /* A3             a3         */\
+        "mov.d $f10, $f0                \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
+        "paddw $f0, $f0, $f2            \n\t" /* A1             a1         */\
+        "psubw $f10, $f10, $f2          \n\t" /* A2             a2         */\
+        "ldc1 $f2, 56(%2)               \n\t" /* C7     C5      C7      C5 */\
+        "li $10, " #shift "             \n\t"                                \
+        "pmaddhw $f2, $f2, $f6          \n\t" /* C7R7+C5R5      C7r7+C5r5  */\
+        "ldc1 $f14, 72(%2)              \n\t" /* -C5    -C1     -C5    -C1 */\
+        "mtc1 $10, $f18                 \n\t"                                \
+        "pmaddhw $f14, $f14, $f6        \n\t" /* -C5R7-C1R5     -C5r7-C1r5 */\
+        "paddw $f2, $f2, $f8            \n\t" /* A0+B0          a0+b0      */\
+        "paddw $f8, $f8, $f8            \n\t" /* 2A0            2a0        */\
+        "psubw $f8, $f8, $f2            \n\t" /* A0-B0          a0-b0      */\
+        "psraw $f2, $f2, $f18           \n\t"                                \
+        "psraw $f8, $f8, $f18           \n\t"                                \
+        "mov.d $f4, $f0                 \n\t" /* A1             a1         */\
+        "paddw $f0, $f0, $f14           \n\t" /* A1+B1          a1+b1      */\
+        "psubw $f4, $f4, $f14           \n\t" /* A1-B1          a1-b1      */\
+        "psraw $f0, $f0, $f18           \n\t"                                \
+        "psraw $f4, $f4, $f18           \n\t"                                \
+        "packsswh $f2, $f2, $f2         \n\t" /* A0+B0          a0+b0      */\
+        "swc1 $f2, " #dst "             \n\t"                                \
+        "packsswh $f0, $f0, $f0         \n\t" /* A1+B1          a1+b1      */\
+        "swc1 $f0, 16+" #dst "          \n\t"                                \
+        "packsswh $f4, $f4, $f4         \n\t" /* A1-B1          a1-b1      */\
+        "swc1 $f4, 96+" #dst "          \n\t"                                \
+        "packsswh $f8, $f8, $f8         \n\t" /* A0-B0          a0-b0      */\
+        "swc1 $f8, 112+" #dst "         \n\t"                                \
+        "ldc1 $f2, 88(%2)               \n\t" /* C3     C7      C3      C7 */\
+        "ldc1 $f16, 104(%2)             \n\t"                                \
+        "pmaddhw $f2, $f2, $f6          \n\t" /* C3R7+C7R5      C3r7+C7r5  */\
+        "mov.d $f4, $f10                \n\t" /* A2             a2         */\
+        "pmaddhw $f6, $f6, $f16         \n\t" /* -C1R7+C3R5     -C1r7+C3r5 */\
+        "paddw $f4, $f4, $f2            \n\t" /* A2+B2          a2+b2      */\
+        "psubw $f10, $f10, $f2          \n\t" /* a2-B2          a2-b2      */\
+        "psraw $f4, $f4, $f18           \n\t"                                \
+        "psraw $f10, $f10, $f18         \n\t"                                \
+        "mov.d $f2, $f12                \n\t" /* A3             a3         */\
+        "paddw $f12, $f12, $f6          \n\t" /* A3+B3          a3+b3      */\
+        "psubw $f2, $f2, $f6            \n\t" /* a3-B3          a3-b3      */\
+        "psraw $f12, $f12, $f18         \n\t"                                \
+        "psraw $f2, $f2, $f18           \n\t"                                \
+        "packsswh $f4, $f4, $f4         \n\t" /* A2+B2          a2+b2      */\
+        "packsswh $f12, $f12, $f12      \n\t" /* A3+B3          a3+b3      */\
+        "swc1 $f4, 32+" #dst "          \n\t"                                \
+        "packsswh $f2, $f2, $f2         \n\t" /* A3-B3          a3-b3      */\
+        "packsswh $f10, $f10, $f10      \n\t" /* A2-B2          a2-b2      */\
+        "swc1 $f12, 48+" #dst "         \n\t"                                \
+        "swc1 $f2, 64+" #dst "          \n\t"                                \
+        "swc1 $f10, 80+" #dst "         \n\t"
+
+        //IDCT(  src0,   src4,   src1,    src5,    dst, shift)
+        IDCT(    (%1), 64(%1), 32(%1),  96(%1),  0(%0),    20)
+        IDCT(   8(%1), 72(%1), 40(%1), 104(%1),  4(%0),    20)
+        IDCT(  16(%1), 80(%1), 48(%1), 112(%1),  8(%0),    20)
+        IDCT(  24(%1), 88(%1), 56(%1), 120(%1), 12(%0),    20)
+        "b 9f                           \n\t"
+
+        "# .p2align 4                   \n\t"
+        "6:                             \n\t"
+        Z_COND_IDCT(96(%0),104(%0),112(%0),120(%0), 96(%1),paddw,(%2), 11, 7f)
+
+#undef  IDCT
+#define IDCT(src0, src4, src1, src5, dst, shift)                             \
+        "ldc1 $f0, " #src0 "            \n\t" /* R4     R0      r4      r0 */\
+        "ldc1 $f6, " #src5 "            \n\t" /* R7     R5      r7      r5 */\
+        "ldc1 $f8, 16(%2)               \n\t" /* C4     C4      C4      C4 */\
+        "pmaddhw $f8, $f8, $f0          \n\t" /* C4R4+C4R0      C4r4+C4r0  */\
+        "ldc1 $f10, 24(%2)              \n\t" /* -C4    C4      -C4     C4 */\
+        "pmaddhw $f0, $f0, $f10         \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
+        "mov.d $f12, $f8                \n\t" /* C4R4+C4R0      C4r4+C4r0  */\
+        "mov.d $f10, $f0                \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
+        "ldc1 $f2, 56(%2)               \n\t" /* C7     C5      C7      C5 */\
+        "pmaddhw $f2, $f2, $f6          \n\t" /* C7R7+C5R5      C7r7+C5r5  */\
+        "ldc1 $f14, 72(%2)              \n\t" /* -C5    -C1     -C5    -C1 */\
+        "li $10, " #shift "             \n\t"                                \
+        "pmaddhw $f14, $f14, $f6        \n\t" /* -C5R7-C1R5     -C5r7-C1r5 */\
+        "paddw $f2, $f2, $f8            \n\t" /* A0+B0          a0+b0      */\
+        "mtc1 $10, $f18                 \n\t"                                \
+        "paddw $f8, $f8, $f8            \n\t" /* 2A0            2a0        */\
+        "psubw $f8, $f8, $f2            \n\t" /* A0-B0          a0-b0      */\
+        "psraw $f2, $f2, $f18           \n\t"                                \
+        "psraw $f8, $f8, $f18           \n\t"                                \
+        "mov.d $f4, $f0                 \n\t" /* A1             a1         */\
+        "paddw $f0, $f0, $f14           \n\t" /* A1+B1          a1+b1      */\
+        "psubw $f4, $f4, $f14           \n\t" /* A1-B1          a1-b1      */\
+        "psraw $f0, $f0, $f18           \n\t"                                \
+        "psraw $f4, $f4, $f18           \n\t"                                \
+        "packsswh $f2, $f2, $f2         \n\t" /* A0+B0          a0+b0      */\
+        "swc1 $f2, " #dst "             \n\t"                                \
+        "packsswh $f0, $f0, $f0         \n\t" /* A1+B1          a1+b1      */\
+        "swc1 $f0, 16+" #dst "          \n\t"                                \
+        "packsswh $f4, $f4, $f4         \n\t" /* A1-B1          a1-b1      */\
+        "swc1 $f4, 96+" #dst "          \n\t"                                \
+        "packsswh $f8, $f8, $f8         \n\t" /* A0-B0          a0-b0      */\
+        "swc1 $f8, 112+" #dst "         \n\t"                                \
+        "ldc1 $f2, 88(%2)               \n\t" /* C3     C7      C3      C7 */\
+        "ldc1 $f16, 104(%2)             \n\t"                                \
+        "pmaddhw $f2, $f2, $f6          \n\t" /* C3R7+C7R5      C3r7+C7r5  */\
+        "mov.d $f4, $f10                \n\t" /* A2             a2         */\
+        "pmaddhw $f6, $f6, $f16         \n\t" /* -C1R7+C3R5     -C1r7+C3r5 */\
+        "paddw $f4, $f4, $f2            \n\t" /* A2+B2          a2+b2      */\
+        "psubw $f10, $f10, $f2          \n\t" /* a2-B2          a2-b2      */\
+        "psraw $f4, $f4, $f18           \n\t"                                \
+        "psraw $f10, $f10, $f18         \n\t"                                \
+        "mov.d $f2, $f12                \n\t" /* A3             a3         */\
+        "paddw $f12, $f12, $f6          \n\t" /* A3+B3          a3+b3      */\
+        "psubw $f2, $f2, $f6            \n\t" /* a3-B3          a3-b3      */\
+        "psraw $f12, $f12, $f18         \n\t"                                \
+        "psraw $f2, $f2, $f18           \n\t"                                \
+        "packsswh $f4, $f4, $f4         \n\t" /* A2+B2          a2+b2      */\
+        "packsswh $f12, $f12, $f12      \n\t" /* A3+B3          a3+b3      */\
+        "swc1 $f4, 32+" #dst "          \n\t"                                \
+        "packsswh $f2, $f2, $f2         \n\t" /* A3-B3          a3-b3      */\
+        "packsswh $f10, $f10, $f10      \n\t" /* A2-B2          a2-b2      */\
+        "swc1 $f12, 48+" #dst "         \n\t"                                \
+        "swc1 $f2, 64+" #dst "          \n\t"                                \
+        "swc1 $f10, 80+" #dst "         \n\t"
+
+        //IDCT(  src0,   src4,   src1,    src5,    dst, shift)
+        IDCT(    (%1), 64(%1), 32(%1),  96(%1),  0(%0),    20)
+        IDCT(   8(%1), 72(%1), 40(%1), 104(%1),  4(%0),    20)
+        IDCT(  16(%1), 80(%1), 48(%1), 112(%1),  8(%0),    20)
+        IDCT(  24(%1), 88(%1), 56(%1), 120(%1), 12(%0),    20)
+        "b 9f                           \n\t"
+
+        "# .p2align 4                   \n\t"
+        "2:                             \n\t"
+        Z_COND_IDCT(96(%0),104(%0),112(%0),120(%0), 96(%1),paddw,(%2), 11, 3f)
+
+#undef  IDCT
+#define IDCT(src0, src4, src1, src5, dst, shift)                             \
+        "ldc1 $f0, " #src0 "            \n\t" /* R4     R0      r4      r0 */\
+        "ldc1 $f4, " #src1 "            \n\t" /* R3     R1      r3      r1 */\
+        "ldc1 $f6, " #src5 "            \n\t" /* R7     R5      r7      r5 */\
+        "ldc1 $f8, 16(%2)               \n\t" /* C4     C4      C4      C4 */\
+        "pmaddhw $f8, $f8, $f0          \n\t" /* C4R4+C4R0      C4r4+C4r0  */\
+        "ldc1 $f10, 24(%2)              \n\t" /* -C4    C4      -C4     C4 */\
+        "pmaddhw $f0, $f0, $f10         \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
+        "mov.d $f12, $f8                \n\t" /* C4R4+C4R0      C4r4+C4r0  */\
+        "ldc1 $f14, 48(%2)              \n\t" /* C3     C1      C3      C1 */\
+        "pmaddhw $f14, $f14, $f4        \n\t" /* C3R3+C1R1      C3r3+C1r1  */\
+        "mov.d $f10, $f0                \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
+        "ldc1 $f2, 56(%2)               \n\t" /* C7     C5      C7      C5 */\
+        "pmaddhw $f2, $f2, $f6          \n\t" /* C7R7+C5R5      C7r7+C5r5  */\
+        "ldc1 $f16, 64(%2)              \n\t"                                \
+        "pmaddhw $f4, $f4, $f16         \n\t" /* -C7R3+C3R1     -C7r3+C3r1 */\
+        "paddw $f14, $f14, $f2          \n\t" /* B0             b0         */\
+        "ldc1 $f2, 72(%2)               \n\t" /* -C5    -C1     -C5    -C1 */\
+        "li $10, " #shift "             \n\t"                                \
+        "pmaddhw $f2, $f2, $f6          \n\t" /* -C5R7-C1R5     -C5r7-C1r5 */\
+        "paddw $f14, $f14, $f8          \n\t" /* A0+B0          a0+b0      */\
+        "mtc1 $10, $f18                 \n\t"                                \
+        "paddw $f8, $f8, $f8            \n\t" /* 2A0            2a0        */\
+        "psubw $f8, $f8, $f14           \n\t" /* A0-B0          a0-b0      */\
+        "paddw $f2, $f2, $f4            \n\t" /* B1             b1         */\
+        "psraw $f14, $f14, $f18         \n\t"                                \
+        "psraw $f8, $f8, $f18           \n\t"                                \
+        "mov.d $f4, $f0                 \n\t" /* A1             a1         */\
+        "paddw $f0, $f0, $f2            \n\t" /* A1+B1          a1+b1      */\
+        "psubw $f4, $f4, $f2            \n\t" /* A1-B1          a1-b1      */\
+        "psraw $f0, $f0, $f18           \n\t"                                \
+        "psraw $f4, $f4, $f18           \n\t"                                \
+        "packsswh $f14, $f14, $f14      \n\t" /* A0+B0          a0+b0      */\
+        "swc1 $f14, " #dst "            \n\t"                                \
+        "packsswh $f0, $f0, $f0         \n\t" /* A1+B1          a1+b1      */\
+        "swc1 $f0, 16+" #dst "          \n\t"                                \
+        "packsswh $f4, $f4, $f4         \n\t" /* A1-B1          a1-b1      */\
+        "swc1 $f4, 96+" #dst "          \n\t"                                \
+        "packsswh $f8, $f8, $f8         \n\t" /* A0-B0          a0-b0      */\
+        "swc1 $f8, 112+" #dst "         \n\t"                                \
+        "ldc1 $f0, " #src1 "            \n\t" /* R3     R1      r3      r1 */\
+        "ldc1 $f8, 80(%2)               \n\t" /* -C1    C5      -C1     C5 */\
+        "pmaddhw $f8, $f8, $f0          \n\t" /* -C1R3+C5R1     -C1r3+C5r1 */\
+        "ldc1 $f14, 88(%2)              \n\t" /* C3     C7      C3      C7 */\
+        "ldc1 $f16, 96(%2)              \n\t"                                \
+        "pmaddhw $f0, $f0, $f16         \n\t" /* -C5R3+C7R1     -C5r3+C7r1 */\
+        "pmaddhw $f14, $f14, $f6        \n\t" /* C3R7+C7R5      C3r7+C7r5  */\
+        "mov.d $f4, $f10                \n\t" /* A2             a2         */\
+        "ldc1 $f16, 104(%2)             \n\t"                                \
+        "pmaddhw $f6, $f6, $f16         \n\t" /* -C1R7+C3R5     -C1r7+C3r5 */\
+        "paddw $f8, $f8, $f14           \n\t" /* B2             b2         */\
+        "paddw $f4, $f4, $f8            \n\t" /* A2+B2          a2+b2      */\
+        "psubw $f10, $f10, $f8          \n\t" /* a2-B2          a2-b2      */\
+        "psraw $f4, $f4, $f18           \n\t"                                \
+        "psraw $f10, $f10, $f18         \n\t"                                \
+        "mov.d $f8, $f12                \n\t" /* A3             a3         */\
+        "paddw $f6, $f6, $f0            \n\t" /* B3             b3         */\
+        "paddw $f12, $f12, $f6          \n\t" /* A3+B3          a3+b3      */\
+        "psubw $f8, $f8, $f6            \n\t" /* a3-B3          a3-b3      */\
+        "psraw $f12, $f12, $f18         \n\t"                                \
+        "psraw $f8, $f8, $f18           \n\t"                                \
+        "packsswh $f4, $f4, $f4         \n\t" /* A2+B2          a2+b2      */\
+        "packsswh $f12, $f12, $f12      \n\t" /* A3+B3          a3+b3      */\
+        "swc1 $f4, 32+" #dst "          \n\t"                                \
+        "packsswh $f8, $f8, $f8         \n\t" /* A3-B3          a3-b3      */\
+        "packsswh $f10, $f10, $f10      \n\t" /* A2-B2          a2-b2      */\
+        "swc1 $f12, 48+" #dst "         \n\t"                                \
+        "swc1 $f8, 64+" #dst "          \n\t"                                \
+        "swc1 $f10, 80+" #dst "         \n\t"
+
+        //IDCT(  src0,   src4,   src1,    src5,    dst, shift)
+        IDCT(    (%1), 64(%1), 32(%1),  96(%1),  0(%0),    20)
+        IDCT(   8(%1), 72(%1), 40(%1), 104(%1),  4(%0),    20)
+        IDCT(  16(%1), 80(%1), 48(%1), 112(%1),  8(%0),    20)
+        IDCT(  24(%1), 88(%1), 56(%1), 120(%1), 12(%0),    20)
+        "b 9f                           \n\t"
+
+        "# .p2align 4                   \n\t"
+        "3:                             \n\t"
+
+#undef  IDCT
+#define IDCT(src0, src4, src1, src5, dst, shift)                             \
+        "ldc1 $f0, " #src0 "            \n\t" /* R4     R0      r4      r0 */\
+        "ldc1 $f4, " #src1 "            \n\t" /* R3     R1      r3      r1 */\
+        "ldc1 $f8, 16(%2)               \n\t" /* C4     C4      C4      C4 */\
+        "pmaddhw $f8, $f8, $f0          \n\t" /* C4R4+C4R0      C4r4+C4r0  */\
+        "ldc1 $f10, 24(%2)              \n\t" /* -C4    C4      -C4     C4 */\
+        "pmaddhw $f0, $f0, $f10         \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
+        "mov.d $f12, $f8                \n\t" /* C4R4+C4R0      C4r4+C4r0  */\
+        "ldc1 $f14, 48(%2)              \n\t" /* C3     C1      C3      C1 */\
+        "pmaddhw $f14, $f14, $f4        \n\t" /* C3R3+C1R1      C3r3+C1r1  */\
+        "mov.d $f10, $f0                \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
+        "ldc1 $f6, 64(%2)               \n\t"                                \
+        "pmaddhw $f6, $f6, $f4          \n\t" /* -C7R3+C3R1     -C7r3+C3r1 */\
+        "li $10, " #shift "             \n\t"                                \
+        "paddw $f14, $f14, $f8          \n\t" /* A0+B0          a0+b0      */\
+        "mtc1 $10, $f18                 \n\t"                                \
+        "paddw $f8, $f8, $f8            \n\t" /* 2A0            2a0        */\
+        "psubw $f8, $f8, $f14           \n\t" /* A0-B0          a0-b0      */\
+        "psraw $f14, $f14, $f18         \n\t"                                \
+        "psraw $f8, $f8, $f18           \n\t"                                \
+        "mov.d $f2, $f0                 \n\t" /* A1             a1         */\
+        "paddw $f0, $f0, $f6            \n\t" /* A1+B1          a1+b1      */\
+        "psubw $f2, $f2, $f6            \n\t" /* A1-B1          a1-b1      */\
+        "psraw $f0, $f0, $f18           \n\t"                                \
+        "psraw $f2, $f2, $f18           \n\t"                                \
+        "packsswh $f14, $f14, $f14      \n\t" /* A0+B0  a0+b0              */\
+        "swc1 $f14, " #dst "            \n\t"                                \
+        "packsswh $f0, $f0, $f0         \n\t" /* A1+B1  a1+b1              */\
+        "swc1 $f0, 16+" #dst "          \n\t"                                \
+        "packsswh $f2, $f2, $f2         \n\t" /* A1-B1  a1-b1              */\
+        "swc1 $f2, 96+" #dst "          \n\t"                                \
+        "packsswh $f8, $f8, $f8         \n\t" /* A0-B0  a0-b0              */\
+        "swc1 $f8, 112+" #dst "         \n\t"                                \
+        "ldc1 $f8, 80(%2)               \n\t" /* -C1    C5      -C1     C5 */\
+        "ldc1 $f16, 96(%2)              \n\t"                                \
+        "pmaddhw $f8, $f8, $f4          \n\t" /* -C1R3+C5R1     -C1r3+C5r1 */\
+        "pmaddhw $f4, $f4, $f16         \n\t" /* -C5R3+C7R1     -C5r3+C7r1 */\
+        "mov.d $f2, $f10                \n\t" /* A2             a2         */\
+        "paddw $f2, $f2, $f8            \n\t" /* A2+B2          a2+b2      */\
+        "psubw $f10, $f10, $f8          \n\t" /* a2-B2          a2-b2      */\
+        "psraw $f2, $f2, $f18           \n\t"                                \
+        "psraw $f10, $f10, $f18         \n\t"                                \
+        "mov.d $f8, $f12                \n\t" /* A3             a3         */\
+        "paddw $f12, $f12, $f4          \n\t" /* A3+B3          a3+b3      */\
+        "psubw $f8, $f8, $f4            \n\t" /* a3-B3          a3-b3      */\
+        "psraw $f12, $f12, $f18         \n\t"                                \
+        "psraw $f8, $f8, $f18           \n\t"                                \
+        "packsswh $f2, $f2, $f2         \n\t" /* A2+B2  a2+b2              */\
+        "packsswh $f12, $f12, $f12      \n\t" /* A3+B3  a3+b3              */\
+        "swc1 $f2, 32+" #dst "          \n\t"                                \
+        "packsswh $f8, $f8, $f8         \n\t" /* A3-B3  a3-b3              */\
+        "packsswh $f10, $f10, $f10      \n\t" /* A2-B2  a2-b2              */\
+        "swc1 $f12, 48+" #dst "         \n\t"                                \
+        "swc1 $f8, 64+" #dst "          \n\t"                                \
+        "swc1 $f10, 80+" #dst "         \n\t"
+
+        //IDCT(  src0,   src4,   src1,    src5,    dst, shift)
+        IDCT(    (%1), 64(%1), 32(%1),  96(%1),  0(%0),    20)
+        IDCT(   8(%1), 72(%1), 40(%1), 104(%1),  4(%0),    20)
+        IDCT(  16(%1), 80(%1), 48(%1), 112(%1),  8(%0),    20)
+        IDCT(  24(%1), 88(%1), 56(%1), 120(%1), 12(%0),    20)
+        "b 9f                           \n\t"
+
+        "# .p2align 4                   \n\t"
+        "5:                             \n\t"
+
+#undef  IDCT
+#define IDCT(src0, src4, src1, src5, dst, shift)                             \
+        "ldc1 $f0, " #src0 "            \n\t" /* R4     R0      r4      r0 */\
+        "ldc1 $f2, " #src4 "            \n\t" /* R6     R2      r6      r2 */\
+        "ldc1 $f8, 16(%2)               \n\t" /* C4     C4      C4      C4 */\
+        "pmaddhw $f8, $f8, $f0          \n\t" /* C4R4+C4R0      C4r4+C4r0  */\
+        "ldc1 $f10, 24(%2)              \n\t" /* -C4    C4      -C4     C4 */\
+        "pmaddhw $f0, $f0, $f10         \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
+        "ldc1 $f10, 32(%2)              \n\t" /* C6     C2      C6      C2 */\
+        "pmaddhw $f10, $f10, $f2        \n\t" /* C6R6+C2R2      C6r6+C2r2  */\
+        "ldc1 $f12, 40(%2)              \n\t" /* -C2    C6      -C2     C6 */\
+        "pmaddhw $f2, $f2, $f12         \n\t" /* -C2R6+C6R2     -C2r6+C6r2 */\
+        "mov.d $f12, $f8                \n\t" /* C4R4+C4R0      C4r4+C4r0  */\
+        "paddw $f8, $f8, $f10           \n\t" /* A0             a0         */\
+        "psubw $f12, $f12, $f10         \n\t" /* A3             a3         */\
+        "mov.d $f10, $f0                \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
+        "paddw $f0, $f0, $f2            \n\t" /* A1             a1         */\
+        "psubw $f10, $f10, $f2          \n\t" /* A2             a2         */\
+        "ldc1 $f4, 8+" #src0 "          \n\t" /* R4     R0      r4      r0 */\
+        "ldc1 $f6, 8+" #src4 "          \n\t" /* R6     R2      r6      r2 */\
+        "ldc1 $f2, 16(%2)               \n\t" /* C4     C4      C4      C4 */\
+        "pmaddhw $f2, $f2, $f4          \n\t" /* C4R4+C4R0      C4r4+C4r0  */\
+        "ldc1 $f14, 24(%2)              \n\t" /* -C4    C4      -C4     C4 */\
+        "pmaddhw $f4, $f4, $f14         \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
+        "ldc1 $f14, 32(%2)              \n\t" /* C6     C2      C6      C2 */\
+        "ldc1 $f16, 40(%2)              \n\t"                                \
+        "pmaddhw $f14, $f14, $f6        \n\t" /* C6R6+C2R2      C6r6+C2r2  */\
+        "pmaddhw $f6, $f6, $f16         \n\t" /* -C2R6+C6R2     -C2r6+C6r2 */\
+        "paddw $f14, $f14, $f2          \n\t" /* A0             a0         */\
+        "paddw $f2, $f2, $f2            \n\t" /* 2C0            2c0        */\
+        "psubw $f2, $f2, $f14           \n\t" /* A3             a3         */\
+        "li $10, " #shift "             \n\t"                                \
+        "paddw $f6, $f6, $f4            \n\t" /* A1             a1         */\
+        "mtc1 $10, $f18                 \n\t"                                \
+        "paddw $f4, $f4, $f4            \n\t" /* 2C1            2c1        */\
+        "psubw $f4, $f4, $f6            \n\t" /* A2             a2         */\
+        "psraw $f8, $f8, $f18           \n\t"                                \
+        "psraw $f14, $f14, $f18         \n\t"                                \
+        "psraw $f6, $f6, $f18           \n\t"                                \
+        "packsswh $f8, $f8, $f14        \n\t" /* A0             a0         */\
+        "sdc1 $f8, " #dst "             \n\t"                                \
+        "psraw $f0, $f0, $f18           \n\t"                                \
+        "packsswh $f0, $f0, $f6         \n\t" /* A1             a1         */\
+        "sdc1 $f0, 16+" #dst "          \n\t"                                \
+        "sdc1 $f0, 96+" #dst "          \n\t"                                \
+        "sdc1 $f8, 112+" #dst "         \n\t"                                \
+        "psraw $f10, $f10, $f18         \n\t"                                \
+        "psraw $f12, $f12, $f18         \n\t"                                \
+        "psraw $f4, $f4, $f18           \n\t"                                \
+        "packsswh $f10, $f10, $f4       \n\t" /* A2-B2          a2-b2      */\
+        "sdc1 $f10, 32+" #dst "         \n\t"                                \
+        "psraw $f2, $f2, $f18           \n\t"                                \
+        "packsswh $f12, $f12, $f2       \n\t" /* A3+B3          a3+b3      */\
+        "sdc1 $f12, 48+" #dst "         \n\t"                                \
+        "sdc1 $f12, 64+" #dst "         \n\t"                                \
+        "sdc1 $f10, 80+" #dst "         \n\t"
+
+        //IDCT(  src0,   src4,   src1,    src5,    dst, shift)
+        IDCT(   0(%1), 64(%1), 32(%1),  96(%1),  0(%0),    20)
+        IDCT(  16(%1), 80(%1), 48(%1), 112(%1),  8(%0),    20)
+        "b 9f                           \n\t"
+
+        "# .p2align 4                   \n\t"
+        "1:                             \n\t"
+
+#undef  IDCT
+#define IDCT(src0, src4, src1, src5, dst, shift)                             \
+        "ldc1 $f0, " #src0 "            \n\t" /* R4     R0      r4      r0 */\
+        "ldc1 $f2, " #src4 "            \n\t" /* R6     R2      r6      r2 */\
+        "ldc1 $f4, " #src1 "            \n\t" /* R3     R1      r3      r1 */\
+        "ldc1 $f8, 16(%2)               \n\t" /* C4     C4      C4      C4 */\
+        "li $10, " #shift "             \n\t"                                \
+        "pmaddhw $f8, $f8, $f0          \n\t" /* C4R4+C4R0      C4r4+C4r0  */\
+        "ldc1 $f10, 24(%2)              \n\t" /* -C4    C4      -C4     C4 */\
+        "mtc1 $10, $f18                 \n\t"                                \
+        "pmaddhw $f0, $f0, $f10         \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
+        "ldc1 $f10, 32(%2)              \n\t" /* C6     C2      C6      C2 */\
+        "pmaddhw $f10, $f10, $f2        \n\t" /* C6R6+C2R2      C6r6+C2r2  */\
+        "ldc1 $f12, 40(%2)              \n\t" /* -C2    C6      -C2     C6 */\
+        "pmaddhw $f2, $f2, $f12         \n\t" /* -C2R6+C6R2     -C2r6+C6r2 */\
+        "mov.d $f12, $f8                \n\t" /* C4R4+C4R0      C4r4+C4r0  */\
+        "ldc1 $f14, 48(%2)              \n\t" /* C3     C1      C3      C1 */\
+        "pmaddhw $f14, $f14, $f4        \n\t" /* C3R3+C1R1      C3r3+C1r1  */\
+        "paddw $f8, $f8, $f10           \n\t" /* A0             a0         */\
+        "psubw $f12, $f12, $f10         \n\t" /* A3             a3         */\
+        "mov.d $f10, $f0                \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
+        "paddw $f0, $f0, $f2            \n\t" /* A1             a1         */\
+        "psubw $f10, $f10, $f2          \n\t" /* A2             a2         */\
+        "ldc1 $f2, 64(%2)               \n\t"                                \
+        "pmaddhw $f2, $f2, $f4          \n\t" /* -C7R3+C3R1     -C7r3+C3r1 */\
+        "paddw $f14, $f14, $f8          \n\t" /* A0+B0          a0+b0      */\
+        "paddw $f8, $f8, $f8            \n\t" /* 2A0            2a0        */\
+        "psubw $f8, $f8, $f14           \n\t" /* A0-B0          a0-b0      */\
+        "psraw $f14, $f14, $f18         \n\t"                                \
+        "psraw $f8, $f8, $f18           \n\t"                                \
+        "mov.d $f6, $f0                 \n\t" /* A1             a1         */\
+        "paddw $f0, $f0, $f2            \n\t" /* A1+B1          a1+b1      */\
+        "psubw $f6, $f6, $f2            \n\t" /* A1-B1          a1-b1      */\
+        "psraw $f0, $f0, $f18           \n\t"                                \
+        "psraw $f6, $f6, $f18           \n\t"                                \
+        "packsswh $f14, $f14, $f14      \n\t" /* A0+B0  a0+b0              */\
+        "swc1 $f14, " #dst "            \n\t"                                \
+        "packsswh $f0, $f0, $f0         \n\t" /* A1+B1  a1+b1              */\
+        "swc1 $f0, 16+" #dst "          \n\t"                                \
+        "packsswh $f6, $f6, $f6         \n\t" /* A1-B1  a1-b1              */\
+        "swc1 $f6, 96+" #dst "          \n\t"                                \
+        "packsswh $f8, $f8, $f8         \n\t" /* A0-B0  a0-b0              */\
+        "swc1 $f8, 112+" #dst "         \n\t"                                \
+        "ldc1 $f8, 80(%2)               \n\t" /* -C1    C5      -C1     C5 */\
+        "ldc1 $f16, 96(%2)              \n\t"                                \
+        "pmaddhw $f8, $f8, $f4          \n\t" /* -C1R3+C5R1     -C1r3+C5r1 */\
+        "pmaddhw $f4, $f4, $f16         \n\t" /* -C5R3+C7R1     -C5r3+C7r1 */\
+        "mov.d $f6, $f10                \n\t" /* A2             a2         */\
+        "paddw $f6, $f6, $f8            \n\t" /* A2+B2          a2+b2      */\
+        "psubw $f10, $f10, $f8          \n\t" /* a2-B2          a2-b2      */\
+        "psraw $f6, $f6, $f18           \n\t"                                \
+        "psraw $f10, $f10, $f18         \n\t"                                \
+        "mov.d $f8, $f12                \n\t" /* A3             a3         */\
+        "paddw $f12, $f12, $f4          \n\t" /* A3+B3          a3+b3      */\
+        "psubw $f8, $f8, $f4            \n\t" /* a3-B3          a3-b3      */\
+        "psraw $f12, $f12, $f18         \n\t"                                \
+        "packsswh $f6, $f6, $f6         \n\t" /* A2+B2          a2+b2      */\
+        "swc1 $f6, 32+" #dst "          \n\t"                                \
+        "psraw $f8, $f8, $f18           \n\t"                                \
+        "packsswh $f12, $f12, $f12      \n\t" /* A3+B3          a3+b3      */\
+        "swc1 $f12, 48+" #dst "         \n\t"                                \
+        "packsswh $f8, $f8, $f8         \n\t" /* A3-B3          a3-b3      */\
+        "packsswh $f10, $f10, $f10      \n\t" /* A2-B2          a2-b2      */\
+        "swc1 $f8, 64+" #dst "          \n\t"                                \
+        "swc1 $f10, 80+" #dst "         \n\t"
+
+        //IDCT(  src0,   src4,   src1,    src5,    dst, shift)
+        IDCT(    (%1), 64(%1), 32(%1),  96(%1),  0(%0),    20)
+        IDCT(   8(%1), 72(%1), 40(%1), 104(%1),  4(%0),    20)
+        IDCT(  16(%1), 80(%1), 48(%1), 112(%1),  8(%0),    20)
+        IDCT(  24(%1), 88(%1), 56(%1), 120(%1), 12(%0),    20)
+        "b 9f                           \n\t"
+
+        "# .p2align 4                   \n\t"
+        "7:                             \n\t"
+
+#undef  IDCT
+#define IDCT(src0, src4, src1, src5, dst, shift)                             \
+        "ldc1 $f0, " #src0 "            \n\t" /* R4     R0      r4      r0 */\
+        "ldc1 $f8, 16(%2)               \n\t" /* C4     C4      C4      C4 */\
+        "li $10, " #shift "             \n\t"                                \
+        "pmaddhw $f8, $f8, $f0          \n\t" /* C4R4+C4R0      C4r4+C4r0  */\
+        "mtc1 $10, $f18                 \n\t"                                \
+        "ldc1 $f10, 24(%2)              \n\t" /* -C4    C4      -C4     C4 */\
+        "pmaddhw $f0, $f0, $f10         \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
+        "psraw $f8, $f8, $f18           \n\t"                                \
+        "psraw $f0, $f0, $f18           \n\t"                                \
+        "ldc1 $f4, 8+" #src0 "          \n\t" /* R4     R0      r4      r0 */\
+        "ldc1 $f2, 16(%2)               \n\t" /* C4     C4      C4      C4 */\
+        "pmaddhw $f2, $f2, $f4          \n\t" /* C4R4+C4R0      C4r4+C4r0  */\
+        "ldc1 $f14, 24(%2)              \n\t" /* -C4    C4      -C4     C4 */\
+        "pmaddhw $f4, $f4, $f14         \n\t" /* -C4R4+C4R0     -C4r4+C4r0 */\
+        "ldc1 $f14, 32(%2)              \n\t" /* C6     C2      C6      C2 */\
+        "psraw $f2, $f2, $f18           \n\t"                                \
+        "packsswh $f8, $f8, $f2         \n\t" /* A0             a0         */\
+        "sdc1 $f8, " #dst "             \n\t"                                \
+        "psraw $f4, $f4, $f18           \n\t"                                \
+        "packsswh $f0, $f0, $f4         \n\t" /* A1             a1         */\
+        "sdc1 $f0, 16+" #dst "          \n\t"                                \
+        "sdc1 $f0, 96+" #dst "          \n\t"                                \
+        "sdc1 $f8, 112+" #dst "         \n\t"                                \
+        "sdc1 $f0, 32+" #dst "          \n\t"                                \
+        "sdc1 $f8, 48+" #dst "          \n\t"                                \
+        "sdc1 $f8, 64+" #dst "          \n\t"                                \
+        "sdc1 $f0, 80+" #dst "          \n\t"
+
+        //IDCT(  src0,   src4,   src1,    src5,    dst, shift)
+        IDCT(   0(%1), 64(%1), 32(%1),  96(%1),  0(%0),    20)
+        IDCT(  16(%1), 80(%1), 48(%1), 112(%1),  8(%0),    20)
+
+        "9:                             \n\t"
+        ::"r"(block),"r"(temp),"r"(coeffs),"m"(ff_wm1010),"m"(ff_d40000)
+        : "$10","$11"
+    );
+}
+
+void ff_simple_idct_mmi(int16_t *block)
+{
+    simple_idct_mmi(block);
+}
+
+void ff_simple_idct_put_mmi(uint8_t *dest, int32_t line_size, int16_t *block)
+{
+    simple_idct_mmi(block);
+    ff_put_pixels_clamped_mmi(block, dest, line_size);
+}
+
+void ff_simple_idct_add_mmi(uint8_t *dest, int32_t line_size, int16_t *block)
+{
+    simple_idct_mmi(block);
+    ff_add_pixels_clamped_mmi(block, dest, line_size);
+}