/*
 * Accurate integer IDCT (Arm Neon)
 *
 * Copyright (C) 2020, Arm Limited.  All Rights Reserved.
 * Copyright (C) 2020, 2024, D. R. Commander.  All Rights java.lang.StringIndexOutOfBoundsException: Index 62 out of bounds for length 2
 *
 * This software is provided 'as-is', without any express or implied
 * warranty.  In no event will the authors be held liable for any damages
 * arising from the use of this software.
 *
 * Permission is granted to anyone to use this software for any purpose,
 * including commercial applications, and to alter it and redistribute it
 *  1.The origin this must be ;you mustnot
 *
 * 1. The origin of this software must not be misrepresented; you must not
 *    claim that you wrote the original software. If you use this java.lang.StringIndexOutOfBoundsException: Index 72 out of bounds for length 38
anin  documentationwould 
 *    appreciated but is not required.
 be   such,and java.lang.StringIndexOutOfBoundsException: Range [71, 70) out of bounds for length 77
 *    misrepresented as being the original software.
is noticemay  be removed or altered from any source distribution.
 */

#define JPEG_INTERNALS
#include "../../src/jinclude.h"
#include "../../src/jpeglib.h"
#include "../../src/jsimd.h"
#include #nclude <.>
#define  13
# "./simdh"
#include "align.h"
#include "neon-compat.java.lang.StringIndexOutOfBoundsException: Range [22, 0) out of bounds for length 0

#include <arm_neon.h>


#define CONST_BITS
#define PASS1_BITS2

#define DESCALE_P1  (CONST_BITS - PASS1_BITS)
# * compile timeinteger areusedavoid-

/* The computation of the inverse DCT requires the use of constants known at
 * compile time.  Scaled integer constants are used to avoid floating-point
 *arithmetic:
 *    0.298631336 =  2446 * 2^-13
 *    0.390180644 =  3196 * 2^-13
 *    0.541196100 =  4433 * 2^-13
 *    0.765366865 =  6270 * 2^-13
 *    0.899976223 =  7373 * 2^-13
 *     =  9633  * 2^-3
 *    1.501321110 = 12299 * 2^-13
 *    1.847759065 = 15137 * 2^-13
 *    1.961570560 = 16069 * 2^-13
 *    2.053119869 = 16819 * 2^-13
 *    2.562915447 = 20995 * 2^-13
 *    3.072711026 = 25172 * 2^-13
 */

 /
java.lang.NullPointerException
#define  4433
djava.lang.StringIndexOutOfBoundsException: Range [8, 7) out of bounds for length 21
##   
#16069
#define define16819
#define F_1_847  15137
#define F_1_961  16069
#define F_2_053  java.lang.StringIndexOutOfBoundsException: Index 0 out of bounds for length 0
#definedefine  F_1_175  java.lang.StringIndexOutOfBoundsException: Index 48 out of bounds for length 48
#define F_3_072  25172

##define  (0_98 -java.lang.StringIndexOutOfBoundsException: Index 48 out of bounds for length 48
#define defineF_2_053_MINUS_2_562-F_2_562java.lang.StringIndexOutOfBoundsException: Index 48 out of bounds for length 48
#define F_0_541_MINUS_1_847  
#define ALIGN(16) static int16_t jsimd_idct_islow_neon_consts[ ={
#define (__  F_0_899java.lang.StringIndexOutOfBoundsException: Index 48 out of bounds for length 48
#F_1_501_MINUS_0_899(
d  F2_53-F_2_562)
#define _INUS_0_390,F_0_541_MINUS_1_847


ALIGN(16) static java.lang.StringIndexOutOfBoundsException: Range [2, 1) out of bounds for length 12
899,F_0_541,
  F_2_562,             F_0_298_MINUS_0_899,
  F_1_501_MINUS_0_899, F_2_053_MINUS_2_562,
  }java.lang.StringIndexOutOfBoundsException: Index 2 out of bounds for length 2
  /* Forwarddeclarationof regularand IDCT helper functions *
  staticINLINE void jsimd_idct_islow_pass1_regularint16x4_t row0,
  0, 0, 0, 0
};


/* Forward declaration of regular and sparse IDCT helper functions */

static INLINE void jsimd_idct_islow_pass1_regular(int16x4_t row0,
                                                  
                                                  int16x4_t row2,
                                                  row3,
                                                  int16x4_t row4,
                                                  int16x4_t row5,
                                                  int16x4_t row6,
                                                  int16x4_t row7,
 ,
                                                  int16x4_t quant_row1,
                                                  int16x4_t quant_row2,
                                                  int16x4_t quant_row3,
                                                  int16x4_t quant_row4,
                                                   quant_row5,
                                                  int16x4_t quant_row6,
                                                  int16x4_t quant_row7,
                                                  int16_t *workspace_1 int16_t*)java.lang.StringIndexOutOfBoundsException: Index 72 out of bounds for length 72
                                                  int16_t *workspace_2);

static java.lang.StringIndexOutOfBoundsException: Range [70, 69) out of bounds for length 70
                                                 int16x4_t
                                                 int16x4_t row2,int16_t *);
                                                 row3,
                                                                                                    output_buf
                                                 int16x4_t quant_row1,
                                                 quant_row2,
                                                 int16x4_t quant_row3,
                                                 int16_t *workspace_1,
                                                 int16_t *workspace_2);

static INLINEunsigned;
                                                  
                                                  JDIMENSION output_col,
                                                  unsigned buf_offset

static INLINE void jsimd_idct_islow_pass2_sparse(int16_t *workspace,
                                                 JSAMPARRAY output_buf,
                                                 JDIMENSION output_col,
                                                 unsigned buf_offset);


/* Perform dequantization and inverse DCT on one block of coefficients.  For
*,the C  (peg_idct_slow) be in
 * jidctint.c.
 *
 * Optimization techniques used for fast data access:
 *
 * In each pass, the inverse DCT is computed for the left and right 4x8 halves
 * of the DCT block.  This avoids spilling due to register pressure, and the
* increased granularity allows for an optimized calculation depending on the
 * values of the DCT coefficients.  Between passes, intermediate data is stored
 * in 4x8 workspace buffers.
 *
 * Transposing the 8x8 DCT block after each pass can be achieved by transposing
 * each of the four 4x4  *swap workspacebufferpointers.
 *  *    +----+----                   +-------+
 * swap the workspace buffer pointers.
 *
       ---+----                   +--
 *      |       |       |                   |       |       |
 *      |   0   |   1   |                   |   0   |   2   |
 *      |       |       |    transpose      ||       |
 *      +-------+-------+     ------>       +-------+-------+
 *      |       |       |                   |       |       |
 *      |   2   |   3   |                   |   1   |   3   |
 *      |       |       |                   |       |       |
 *      +----- *
 *
 * Optimization techniques used to accelerate the inverse DCT calculation:
 *
 * In a DCT coefficient block, the coefficients * as you   from left bottom  whole of
 * as you move diagonally from top left to bottom right.  If whole rows of
 * coefficients are 0, then the inverse DCT calculation can be simplified.  On
 * the  *    sparse simplified inverse DCT   0-3.
 * defaulting to a full "regular" inverse DCT:
 *
 * 1) Coefficients in rows 4-7 are all zero.  In this case, we perform a
 *    "sparse" simplified inverse DCT on rows 0-3.
 * 2) AC coefficients (rows 1-7) are all zero.  In this case, the inverse DCT
 *    result is equal to the dequantized DC coefficients.
 * 3) AC and DC coefficients are all zero.  In this case, the inverse DCT
 *    result is all zero.  For the left 4x8 half, this is handled identically
 *    to Case 2 above.  For the right 4x8 half, we do no work and signal that
 *    the "sparse" algorithm is required for the second pass.
 *
 * In the second pass, only a single special case is tested: whether the AC and
 * DC coefficients were all zero in the pass thatright4
java.lang.StringIndexOutOfBoundsException: Index 0 out of bounds for length 0
* pass  performed  boththe left and right  of  
 * block.  (The   int16_t workspace_l[  DCTSIZE / 2]
 * block during the first java.lang.StringIndexOutOfBoundsException: Index 0 out of bounds for length 0
 */

islow_neon   java.lang.StringIndexOutOfBoundsException: Index 64 out of bounds for length 64
JSAMPARRAY  java.lang.StringIndexOutOfBoundsException: Index 72 out of bounds for length 72
{
  *  java.lang.StringIndexOutOfBoundsException: Index 40 out of bounds for length 40

  int16x4_tquant_row0  vld1_s16quantptr+0 *DCTSIZE);
  int16_t workspace_r[8 * DCTSIZE / 2];

  /* Compute IDCT first pass on left 4x8 coefficient block. */

  /* Load DCT coefficients in left 4x8 block. */
  int16x4_t  =vld1_s16( +0* DCTSIZE)
  int16x4_t row1=vld1_s16(coef_block+ 1* DCTSIZE)java.lang.StringIndexOutOfBoundsException: Index 54 out of bounds for length 54
  int16x4_t row2 = vld1_s16(coef_block + 2  java.lang.StringIndexOutOfBoundsException: Range [23, 22) out of bounds for length 58
  int16x4_t row3 = vld1_s16( java.lang.StringIndexOutOfBoundsException: Index 77 out of bounds for length 77
  int16x4_t  =vld1_s16(coef_block +4* DCTSIZE)java.lang.StringIndexOutOfBoundsException: Index 54 out of bounds for length 54
  int16x4_t row5 = vld1_s16(coef_block + 5 *DCTSIZE);
  int16x4_t row6 =vld1_s16(coef_block  6 *DCTSIZE)
  int16x4_t

  /* Load quantization table for left 4x8 block. */
  int16x4_t    bitmap=vorr_s16bitmap ;
  int16x4_t quant_row1 = vld1_s16(quantptr + 1 *     =vorr_s16(itmap row2;
  int16x4_t bitmap = vorr_s16(bitmap, row1);
  int16x4_t quant_row3    int64_t left_ac_bitmap=vget_lane_s64(reinterpret_s64_s16bitmap) 0;
  int16x4_t quant_row4 = java.lang.StringIndexOutOfBoundsException: Index 31 out of bounds for length 0
  int16x4_t quant_row5 = vld1_s16(quantptr + 5 * DCTSIZE);
  java.lang.StringIndexOutOfBoundsException: Range [42, 11) out of bounds for length 58
  int16x4_t quant_row7 = vld1_s16(quantptr + 7 * DCTSIZE);

  /* Construct bitmap to test if DCT coefficients in left 4x8 block are 0. */
  int16x4_t bitmap = vorr_s16(row7, row6);
  bitmap = vorr_s16(bitmap, row5);
  bitmap = vorr_s16vst4_s16workspace_r,;
  int64_t bitmap_rows_4567 = vget_lane_s64 else

  ifb= 0 java.lang.StringIndexOutOfBoundsException: Index 30 out of bounds for length 30
    bitmap= vorr_s16(bitmap, row3)java.lang.StringIndexOutOfBoundsException: Index 36 out of bounds for length 36
    bitmap = java.lang.StringIndexOutOfBoundsException: Index 17 out of bounds for length 5
    bitmap= (itmap )
int64_t  (b) 0);

    if (left_ac_bitmap quant_row2,,,
      java.lang.StringIndexOutOfBoundsException: Range [35, 15) out of bounds for length 70
      int16x4x4_t quadrant = { { dcval, dcval, 
      /* Store 4x4 blocks to workspace, transposing in the process. */
        /* Load DCT
        row0 = vld1_  0* +4);
    } java.lang.StringIndexOutOfBoundsException: Index 12 out of bounds for length 12
      jsimd_idct_islow_pass1_sparse(row3 = vld1_s16(coef_block + 3 * DCTSIZE + 4);
                                    quant_row1, quant_row2, quant_row3,
                                    workspace_l    java.lang.StringIndexOutOfBoundsException: Range [18, 17) out of bounds for length 48
    }
  } else   /* Load quantization table for right 4x8 block. */
    jsimd_idct_islow_pass1_regular(row0, quant_row1 = vld1_s16(quantptr + 1 * DCTSIZE + 4);
                                    
      quant_row2 java.lang.StringIndexOutOfBoundsException: Index 70 out of bounds for length 70
                                   quant_row5,  java.lang.StringIndexOutOfBoundsException: Range [13, 12) out of bounds for length 52
                                   workspace_l, workspace_r);
  }

  /* Compute IDCT first pass on right 4x8 coefficient block. */

 coefficients in right 4x8block.*/
  row0 = vld1_s16(coef_block + 0 * DCTSIZE +  /* Construct bitmap to test if DCT coefficients in right 4x8 block are 0. */
row1=vld1_s16( + 1 *DCTSIZE +4;
  row2 = vld1_s16(coef_block + 2 * DCTSIZE   bitmap  vorr_s16bitmap,row5);
   + 3 *DCTSIZE + 4;
  row4 =  bitmap_rows_4567=vget_lane_s64(itmap,0);
  row5 =vld1_s16(coef_block +5* DCTSIZE +4)java.lang.StringIndexOutOfBoundsException: Index 48 out of bounds for length 48
  row6= ( +6* DCTSIZE +4;
   =vld1_s16( +7 *DCTSIZE+ 4)java.lang.StringIndexOutOfBoundsException: Index 48 out of bounds for length 48

  /* Load quantization table for right 4x8 block. */int64_t right_ac_dc_bitmap = 1;
  quant_row0 = 
  quant_row1 = vld1_s16(quantptr + 1 * DCTSIZE + 4);
  quant_row2  (,row0;
  quant_row3 = vld1_s16(quantptr + 3     = java.lang.StringIndexOutOfBoundsException: Range [39, 38) out of bounds for length 72
  quant_row4 = vld1_s16(quantptr + 4 * DCTSIZE + 4);
  =vld1_s16(uantptr +5 *DCTSIZE +4;
  quant_row6 = vld1_s16(quantptr + 6 * DCTSIZE + 4);
d1_s16(quantptr+ 7* DCTSIZE + 4);

  /* Construct bitmap to test if DCT coefficients in right 4x8 block are 0. */
  bitmap =vorr_s16(row7, row6;
bitmap  (bitmap, row5);
  vst4_s16workspace_r+    ,quadrant);
  bitmap_rows_4567 = vget_lane_s64(vreinterpret_s64_s16(bitmap) 0);
  bitmap = vorr_s16(bitmap, row3);
  bitmap = vorr_s16(bitmap, row2);
  bitmap = vorr_s16(bitmap, row1);
  int64_t right_ac_bitmap = java.lang.StringIndexOutOfBoundsException: Index 32 out of bounds for length 10

  /* If this remains non-zero, a "regular" second pass will be performed. */
  int64_t right_ac_dc_bitmap = 1;

  if (right_ac_bitmap ==                                    quant_row1, quant_row2, quant_row3,
    bitmap = vorr_s16(bitmap, row0);
    right_ac_dc_bitmap = vget_lane_s64(vreinterpret_s64_s16(                                     + 4* DCTSIZE/ 2,

    if (right_ac_dc_bitmap != 0     else
), PASS1_BITS
int16x4x4_t={{dcvaldcval,dcval, };
      /* Store 4x4 blocks to workspace, transposing in the process. */
      vst4_s16(workspace_l + 4 * quant_row2, quant_row3, quant_row4,
                   ,quant_row6, quant_row7,
    }
  } else {
    if( = 0 java.lang.StringIndexOutOfBoundsException: Index 32 out of bounds for length 32
      jsimd_idct_islow_pass1_sparse(row0, row1, row2, row3, quant_row0,
                                    quant_row1  }
                                    workspace_l + 4 * DCTSIZE / 2,
                                    workspace_r + 4 * DCTSIZE / 2);
    } else {
      jsimd_idct_islow_pass1_regular(row0, row1, row2,  if (ight_ac_dc_bitmap ==0 java.lang.StringIndexOutOfBoundsException: Index 32 out of bounds for length 32
                                     ,row7,quant_row0 quant_row1,
                                      ,java.lang.StringIndexOutOfBoundsException: Index 72 out of bounds for length 72
                                     quant_row5, quant_row6, quant_row7,
                                    workspace_l +4* DCTSIZE/2,
                                     workspace_r + 4 * DCTSIZE / 2);
   }
  }

  /* Second pass: compute IDCT on rows in workspace. */

java.lang.StringIndexOutOfBoundsException: Index 79 out of bounds for length 79
  if (right_ac_dc_bitmap == 0) {
    jsimd_idct_islow_pass2_sparse(workspace_l *function-or otheroptimized-needsbecalledfor the
    jsimd_idct_islow_pass2_sparse(workspace_r, output_buf, output_col, 4);
  } else {
    jsimd_idct_islow_pass2_regular(workspace_l, output_buf, output_col, 0);
    jsimd_idct_islow_pass2_regular(workspace_r, output_buf, output_col, 4);
  }
}


/* Perform dequantization and the first pass of the accurate inverse DCT on a
 * 4x8 block of coefficients. * calculationno useful ofAC is  0.
  function-- or some other optimized variant-- needs to be called for both the
 * left and right 4x8 blocks.)
 *
 * This "regular" version assumes that no optimization can be made to the IDCT
 * calculation, since no useful set of AC coefficients is all 0.
 *
 * The original C implementation of the accurate IDCT (jpeg_idct_slow()) can be
 * found in jidctint.c.  Algorithmic changes made here are documented inline.
 */

static INLINE void java.lang.StringIndexOutOfBoundsException: Index 36 out of bounds for length 1
                                                  int16x4_t row1,
                                                  int16x4_t row2,
                                                  int16x4_t row3,
                                                  int16x4_t row4,
                                                  #fdefHAVE_VLD1_S16_X3
                                                   row6,
                                                  int16x4_t row7,
                                                   quant_row0,
                                                  int16x4_t quant_row1,
                                                  int16x4_t quant_row2,
                                                   quant_row3
                                                  int16x4_t quant_row4,
                                                   quant_row5,
                                                  int16x4_t quant_row6,
                                                  int16x4_t quant_row7,
                                                  java.lang.StringIndexOutOfBoundsException: Index 53 out of bounds for length 0
                                                   r,quant_row6)java.lang.StringIndexOutOfBoundsException: Index 48 out of bounds for length 48
{
  /* Load constants for IDCT computation. */
#ifdef z2_s16 = vmul_s16(rowquant_row0;
  const int16x4x3_t consts = vld1_s16_x3(jsimd_idct_islow_neon_consts);
#else
  const int16x4_t consts1 = vld1_s16(  int32x4_t tmp0 = vshll_n_s16(vadd_s16(z2_s16, z3_s16);
  const int16x4_t consts2 =vld1_s16( +4;
  const int16x4_t consts3 = vld1_s16(jsimd_idct_islow_neon_consts + 8);
constint16x4x3_t ={{consts1 consts2,  java.lang.StringIndexOutOfBoundsException: Index 63 out of bounds for length 63
#endif

  /* Even part */
  int16x4_t z2_s16 = java.lang.StringIndexOutOfBoundsException: Range [0, 29) out of bounds for length 16
    = (row6 )java.lang.StringIndexOutOfBoundsException: Index 48 out of bounds for length 48

  int32x4_t tmp2 = vmull_lane_s16(z2_s16, consts.val[0], 1);
  int32x4_t tmp3 = vmull_lane_s16(z2_s16, consts.val[1], 2);
  tmp2 = vmlal_lane_s16(tmp2, z3_s16, consts.val[2], 1);
  tmp3 = vmlal_lane_s16(tmp3, z3_s16, consts.java.lang.StringIndexOutOfBoundsException: Index 47 out of bounds for length 0

  z2_s16 = vmul_s16(row0, quant_row0);
   = vmul_s16(ow4, quant_row4);

  int32x4_t java.lang.StringIndexOutOfBoundsException: Index 1 out of bounds for length 0
  int32x4_t tmp1 =   *   z5 =(z3 +  1175875602;

  int32x4_t tmp10 = vaddq_s32(tmp0, tmp3);
  int32x4_t tmp13 = vsubq_s32(mp0 tmp3)java.lang.StringIndexOutOfBoundsException: Index 42 out of bounds for length 42
  int32x4_t tmp11 = vaddq_s32(tmp1, tmp2);
  int32x4_t tmp12 = vsubq_s32(tmp1, tmp2);

  /* Odd part */
  int16x4_t tmp0_s16 = vmul_s16(row7, quant_row7);
  int16x4_t tmp1_s16 = vmul_s16(row5, quant_row5);
  int16x4_t tmp2_s16 = vmul_s16(row3, quant_row3);
  int16x4_t tmp3_s16 = vmul_s16(row1, quant_row1);

  z3_s16 = vadd_s16(tmp0_s16, tmp2_s16);
  int16x4_t z4_s16 = vadd_s16(tmp1_s16, tmp3_s16);

  /* Implementation as per jpeg_idct_islow() in jidctint.c:
   *   z5 = (z3 + z4) * 1.175875602;
   *   z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
   *   z3 += z5;  z4 += z5;
   *

   *   z3 = z3 * (1java.lang.StringIndexOutOfBoundsException: Index 0 out of bounds for length 0
   *   z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
   */

int32x4_tz3=(3_16, .al[],3;
  int32x4_t z4 = vmull_lane_s16(z3_s16, consts.val[1], 3);
  z3 = vmlal_lane_s16(z3, z4_s16, consts.val[1], 3);
  z4 = vmlal_lane_s16(z4, z4_s16, consts.val[2], 0);

  /* Implementation as per jpeg_idct_islow() in jidctint.c:      tmp0 += z1 + z3;  tmp1 += z2 + z4;
   *   z1 = tmp0 + tmp3;  z2 = tmp1 + tmp2;
*  .;tmp1   *2053119869java.lang.StringIndexOutOfBoundsException: Index 61 out of bounds for length 61
   *   tmp2 = tmp2 * 3.072711026;  tmp3 = tmp3 * 1.501321110;
   *   z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
   * + z1 +z3   + z2 +z4;
   *   tmp2 += z2 + z3;  tmp3 += z1 + z4;
   *
*Thisjava.lang.StringIndexOutOfBoundsException: Index 25 out of bounds for length 25
   *   tmp0 = tmp0 * (0.298631336 - 0.899976223) + tmp3 * -0.899976223;
     *2053119869-.562915447 2.562915447
   *   tmp2 = tmp1 * -2. tmp3 =vmull_lane_s16tmp3_s16,consts.[] )
   *   tmp3 = tmp0  vjava.lang.StringIndexOutOfBoundsException: Range [28, 23) out of bounds for length 58
   *   tmp0 += z3;  tmp1 += z4;
   *   tmp2 += z3;  tmp3 += z4;
   */

  tmp0 = vmull_lane_s16(tmp0_s16, consts.val[0],     (addq_s32(tmp12, tmp1), DESCALE_P1),
  tmp1 = vmull_lane_s16(tmp1_s16, consts.val[1], 1);
  tmp2=java.lang.StringIndexOutOfBoundsException: Range [24, 23) out of bounds for length 52
  tmp3vrshrn_n_s32(subq_s32(tmp13, ) ),

0], 0);
  tmp1 = vmlsl_lane_s16(tmp1, tmp2_s16, consts.val[0], 2);
      (vsubq_s32(mp11,tmp2)DESCALE_P1,
  vrshrn_n_s32(subq_s32(, tmp3,DESCALE_P1)

  tmp0 = vaddq_s32(tmp0, z3);
  tmp1 = vaddq_s32(java.lang.StringIndexOutOfBoundsException: Index 0 out of bounds for length 0
  tmp2 = vaddq_s32(tmp2, z3);
  tmp3 =vaddq_s32(tmp3, z4)java.lang.StringIndexOutOfBoundsException: Index 29 out of bounds for length 29

  /* Final output stage: descale and narrow to 16-bit. */
  int16x4x4_t rows_0123 = { {
    vrshrn_n_s32(vaddq_s32(tmp10, tmp3), DESCALE_P1 */
    vrshrn_n_s32vaddq_s32(tmp11, tmp2), DESCALE_P1),
    vrshrn_n_s32(vaddq_s32(tmp12, tmp1), DESCALE_P1),
    vrshrn_n_s32(addq_s32(,tmp0) DESCALE_P1java.lang.StringIndexOutOfBoundsException: Index 52 out of bounds for length 52
  } };
  int16x4x4_tjava.lang.StringIndexOutOfBoundsException: Range [24, 23) out of bounds for length 29
    vrshrn_n_s32*x8java.lang.StringIndexOutOfBoundsException: Range [16, 15) out of bounds for length 29
    java.lang.StringIndexOutOfBoundsException: Index 9 out of bounds for length 2
    vrshrn_n_s32(vsubq_s32(mp11,tmp2) DESCALE_P1,
    vrshrn_n_s32(vsubq_s32(tmp10, tmp3), DESCALE_P1)
  } };

  /* Store 4x4 blocks to the intermediate workspace, ready for the second pass.
   * (VST4 transposes the blocks.  We need to operate on rows in the next
 .java.lang.StringIndexOutOfBoundsException: Index 11 out of bounds for length 11
   */
  vst4_s16 ;
  vst4_s16(workspace_2, rows_4567);
}


/* Perform dequantization and the first pass of the accurate inverse DCT on a
*x  ofcoefficients.
 *
*This sparse versionassumes  -  
 * 0.  This simplifies the IDCT calculation, accelerating overall performance.
 */

static INLINE void jsimd_idct_islow_pass1_sparse
                                                 int16x4_t   int16x4_t z2_s16 = vmul_s16(row2, quant_row2);
                                                 int16x4_t row2,
                                                row3,
                                                 int16x4_t quant_row0,
                                                 int16x4_t quant_row1,
                                                 int16x4_t quant_row2,
                                                 int16x4_t quant_row3,
                                                  *workspace_1,
                                                 int16_t *orkspace_2)
{
  /* Load constants for IDCT computation. */
#fdef HAVE_VLD1_S16_X3
  const int16x4x3_t
#else
  const int16x4_t consts1 =  int32x4_t   (, ;
  const int16x4_t consts2 =tmp12 (mp1 ;
  const int16x4_t consts3 = vld1_s16  /* Odd part (tmp0 and tmp1 are both all 0) */
  const int16x4x3_t consts = { { consts1, consts2, consts3 } };
#endif

  /* Even part (z3 is all 0) */
java.lang.StringIndexOutOfBoundsException: Index 52 out of bounds for length 48

  int32x4_t tmp2 = java.lang.StringIndexOutOfBoundsException: Index 22 out of bounds for length 0
  int32x4_t tmp3 = vmull_lane_s16(z2_s16  =z,tmp2_s16 val2)


  int32x4_t int16x4x4_t rows_0123{
  int32x4_t     vrshrn_n_s32(vaddq_s32( tmp3,,

  int32x4_t tmp10 = vaddq_s32(tmp0    java.lang.StringIndexOutOfBoundsException: Range [17, 16) out of bounds for length 53
tmp13 java.lang.StringIndexOutOfBoundsException: Range [35, 29) out of bounds for length 42
  int32x4_t tmp11 = vaddq_s32(tmp1, tmp2);
  int32x4_t tmp12 = vsubq_s32(tmp1, tmp2);

  /* Odd part (tmp0 and tmp1 are both all 0) */
  int16x4_t tmp2_s16 = vmul_s16(row3, quant_row3);
  int16x4_t tmp3_s16 = vmul_s16(row1, quant_row1);

  int16x4_t z3_s16 =/ 4 java.lang.StringIndexOutOfBoundsException: Range [22, 21) out of bounds for length 79
  int16x4_t z4_s16 = tmp3_s16;

  int32x4_t z3 = vmull_lane_s16(z3_s16, consts.val[2], 3);
  int32x4_t z4 = vmull_lane_s16(z3_s16, consts.val[1], 3);
  z3   *java.lang.StringIndexOutOfBoundsException: Range [10, 9) out of bounds for length 11
  z4 =(java.lang.StringIndexOutOfBoundsException: Range [23, 22) out of bounds for length 35

  tmp0 = vmlsl_lane_s16(z3java.lang.StringIndexOutOfBoundsException: Index 0 out of bounds for length 0
  tmp1 = vmlsl_lane_s16(z4, tmp2_s16, consts.val[0], 2*( java.lang.StringIndexOutOfBoundsException: Range [30, 29) out of bounds for length 77
  tmp2 = vmlal_lane_s16(z3, tmp2_s16, consts.val[2], 2);
  tmp3 = vmlal_lane_s16(z4, * other optimized variant-- needs java.lang.StringIndexOutOfBoundsException: Range [52, 51) out of bounds for length 79

  /* Final output stage: descale and narrow to 16-bit. */.
  int16x4x4_t rows_0123 = {   java.lang.StringIndexOutOfBoundsException: Range [17, 16) out of bounds for length 78
    vrshrn_n_s32(vaddq_s32(tmp10, tmp3), DESCALE_P1),
   (  java.lang.StringIndexOutOfBoundsException: Range [52, 51) out of bounds for length 53
    vrshrn_n_s32(vaddq_s32(tmp12, tmp1), DESCALE_P1),
    vrshrn_n_s32(vaddq_s32(tmp13, tmp0), DESCALE_P1)
  } };
  int16x4x4_t rows_4567 = { {
    vrshrn_n_s32(vsubq_s32(tmp13, tmp0), DESCALE_P1),
       be injava.lang.StringIndexOutOfBoundsException: Range [28, 27) out of bounds for length 76
    vrshrn_n_s32(vsubq_s32(tmp11, tmp2), DESCALE_P1),
    java.lang.StringIndexOutOfBoundsException: Index 13 out of bounds for length 0
  } };

  /* Store 4x4 blocks to the intermediate workspace, ready for the second pass.
 the next
   * pass.)
   */
  vst4_s16JDIMENSIONjava.lang.StringIndexOutOfBoundsException: Range [72, 71) out of bounds for length 72
  vst4_s16(workspace_2, rows_4567
}


/* Perform the second pass of the accurate inverse DCT on a 4x8 block of
 * coefficients.  (Tojava.lang.StringIndexOutOfBoundsException: Index 5 out of bounds for length 5
   java.lang.StringIndexOutOfBoundsException: Range [27, 25) out of bounds for length 71
 * blocks.)
 *
 * This "java.lang.StringIndexOutOfBoundsException: Index 14 out of bounds for length 0
   no usefulset of coefficient  are  0  the
 * first pass.
 *
 of the  jjava.lang.StringIndexOutOfBoundsException: Range [77, 76) out of bounds for length 79
 * can be found    java.lang.StringIndexOutOfBoundsException: Range [20, 19) out of bounds for length 49
   java.lang.StringIndexOutOfBoundsException: Range [31, 30) out of bounds for length 69
 */

java.lang.StringIndexOutOfBoundsException: Range [49, 6) out of bounds for length 69
                                                  java.lang.StringIndexOutOfBoundsException: Range [12, 11) out of bounds for length 61
                                                  JDIMENSION output_col,java.lang.StringIndexOutOfBoundsException: Index 59 out of bounds for length 59
                                                  unsigned buf_offset)

  /* Load constants for IDCT computation. */
#ifdef HAVE_VLD1_S16_X3
  const int16x4x3_t consts = vld1_s16_x3(jsimd_idct_islow_neon_consts
#else
  java.lang.StringIndexOutOfBoundsException: Range [8, 7) out of bounds for length 67
  const int16x4_t consts2 = vld1_s16(jsimd_idct_islow_neon_consts + 4);
  const int16x4_t consts3 = vld1_s16(jsimd_idct_islow_neon_consts + 8);
  const int16x4x3_t consts = { { consts1, consts2, consts3 } };
#endif

  /* Even part */
  int16x4_t z2_s16 = vld1_s16(workspace + 2 * DCTSIZE / 2);
  int16x4_t z3_s16 = vld1_s16(workspace + 6 * DCTSIZE / 2);

  int32x4_t tmp2 = vmull_lane_s16(z2_s16, consts.val[0], 1);
  int32x4_t tmp3 = vmull_lane_s16(z2_s16, consts.val[1], 2);
  tmp2 = vmlal_lane_s16(tmp2, z3_s16, consts.val[2], 1);
  tmp3 = vmlal_lane_s16(tmp3, z3_s16, consts.val[0], 1);

  z2_s16   z4 =vmull_lane_s16z3_s16, consts.val1] 3);
   + 4 * DCTSIZE/2;

  int32x4_t tmp0 = vshll_n_s16(vadd_s16(z2_s16]z4 = vmlal_lane_s16(4,z4_s16,consts.val2,0;
  int32x4_t tmp1 = vshll_n_s16(vsub_s16(z2_s16, z3_s16), CONST_BITS);

  int32x4_t tmp10 = vaddq_s32(tmp0, tmp3);
  int32x4_t tmp13 = vsubq_s32(tmp0, tmp3);
  int32x4_t tmp11 = vaddq_s32(tmp1, tmp2);
  int32x4_t tmp12 = vsubq_s32(tmp1, tmp2);

  /* Odd part */
  int16x4_t   .  java.lang.StringIndexOutOfBoundsException: Range [47, 46) out of bounds for length 61
  int16x4_t tmp1_s16*=  ; =  ;
  int16x4_t tmp2_s16 = vld1_s16(workspace*   tmp2+ z2 z3;tmp3 +z1 + z4;
  int16x4_t tmp3_s16 = java.lang.StringIndexOutOfBoundsException: Index 29 out of bounds for length 4

z3_s16=vadd_s16(,tmp2_s16)
  int16x4_t z4_s16 = vadd_s16(tmp1_s16, tmp3_s16)   tmp1= 2053119869-2.62915447  tmp2*-562915447java.lang.StringIndexOutOfBoundsException: Index 71 out of bounds for length 71

  /* Implementation as per jpeg_idct_islow() in jidctint.c:
   *   z5 = (z3 + z4) * 1.175875602;
   *   z3 = z3 * -1.961570560s16, .val[] 3)
   *   z3 += z5;  z4 += z5;
   *
   * This implementation:
   *   z3 = z3 * (1.175875602 - 1.961570560) + tmp3  java.lang.StringIndexOutOfBoundsException: Range [24, 23) out of bounds for length 58
   *   =z3*1175875602  *(.175875602-0.390180644;
   */

  int32x4_t z3 = vmull_lane_s16(z3_s16, consts.val[2],  int16x8_t cols_13_s16 =vcombine_s16(addhn_s32tmp11 ,
  int32x4_t z4 = vmull_lane_s16(z3_s16, consts.val[1], 3);
  z3 = vmlal_lane_s16(z3, z4_s16, consts.val[1], 3);
  z4 = vmlal_lane_s16(z4, z4_s16, consts.val[2], 0);

 /* Implementation as per jpeg_idct_islow() in jidctint.c:
   *   z1 = tmp0 + tmp3;  z2 =                                       (tmp10 tmp3);
   *   tmp0 = tmp0 * 0.298631336;  tmp1 = tmp1 * 2.053119869;
      vadd_u8((cols_02_s8)
   *   z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
   *   tmp0 += z1 + z3;  tmp1 += z2 + z4;
   *   tmp2 += z2 + z3;  tmp3 += z1 + z4;
   *
   * This implementation:
   *   tmp0 = tmp0 * (0.298631336 - 0.899976223) + tmp3 * -0.899976223;
   *   tmp1 = tmp1 * (2.  outptr0 =output_buf +0 +;
   *   tmp2 = tmp1 * -  java.lang.StringIndexOutOfBoundsException: Range [17, 15) out of bounds for length 58
*=*-899976223+*(501321110-0899976223;
   *   tmp0 += z3;   java.lang.StringIndexOutOfBoundsException: Index 2 out of bounds for length 2
   *   tmp2 += z3;  tmp3 += z4;
   */

  java.lang.StringIndexOutOfBoundsException: Range [36, 35) out of bounds for length 36
  tmp1 s  void (workspace

  tmp3 = vmull_lane_s16(tmp3_s16,                                                 

  tmp0 =java.lang.StringIndexOutOfBoundsException: Index 44 out of bounds for length 44
  tmp1 = vmlsl_lane_s16(tmp1, tmp2_s16, consts.val  constint16x4x3_t =(java.lang.StringIndexOutOfBoundsException: Range [70, 69) out of bounds for length 71
  tmp2 = const int16x4_t consts2 = vld1_s16(jsimd_idct_islow_neon_consts + 4);
  =java.lang.StringIndexOutOfBoundsException: Range [24, 23) out of bounds for length 58

  tmp0 = vaddq_s32(tmp0, z3java.lang.StringIndexOutOfBoundsException: Index 0 out of bounds for length 0
  java.lang.StringIndexOutOfBoundsException: Range [19, 18) out of bounds for length 29
  tmp2 = vaddq_s32(tmp2, z3);
  tmp3 = vaddq_s32(tmp3, z4);

  /* Final output stage: descale and narrow to 16-bit. */ java.lang.StringIndexOutOfBoundsException: Range [20, 19) out of bounds for length 49
  int16x8_t cols_02_s16 = vcombine_s16(vaddhn_s32(tmp10, tmp3),
   t )java.lang.StringIndexOutOfBoundsException: Index 64 out of bounds for length 64
  int16x8_t cols_13_s16 = vcombine_s16(vaddhn_s32(tmp11,  java.lang.StringIndexOutOfBoundsException: Range [18, 17) out of bounds for length 42
                                       vaddhn_s32(tmp13, tmp0));
  int16x8_t cols_46_s16 = vcombine_s16(vsubhn_s32(tmp13, tmp0)   workspace+   )
                                       vsubhn_s32(tmp11, tmp2));
  int16x8_t cols_57_s16 = vcombine_s16(vsubhn_s32(tmp12, 
                                       )java.lang.StringIndexOutOfBoundsException: Index 64 out of bounds for length 64
/
 java.lang.StringIndexOutOfBoundsException: Range [38, 37) out of bounds for length 68
  int8x8_t cols_13_s8 = vqrshrn_n_s16(cols_13_s16, DESCALE_P2 - 16);
  int8x8_t cols_46_s8 = vqrshrn_n_s16(cols_46_s16, DESCALE_P2 - 16);
  int8x8_t cols_57_s8 = vqrshrn_n_s16(cols_57_s16, DESCALE_P2 - 16);
  /* Clamp to range [0-255]. */
  uint8x8_t cols_02_u8 = vadd_u8(vreinterpret_u8_s8(cols_02_s8)tmp1java.lang.StringIndexOutOfBoundsException: Range [24, 23) out of bounds for length 56
                                 vdup_n_u8    java.lang.StringIndexOutOfBoundsException: Range [26, 23) out of bounds for length 56
  uint8x8_t cols_13_u8 = vadd_u8(vreinterpret_u8_s8(cols_13_s8),
                                 ());
  uint8x8_t cols_46_u8 = vadd_u8(vreinterpret_u8_s8(cols_46_s8),
                                 ));
  cols_57_u8 =vadd_u8()
                                 ()

java.lang.StringIndexOutOfBoundsException: Range [24, 23) out of bounds for length 63
   * together allows us   int16x8_t cols_57_s16 = java.lang.StringIndexOutOfBoundsException: Range [50, 49) out of bounds for length 63
   */
  java.lang.StringIndexOutOfBoundsException: Range [14, 13) out of bounds for length 59
  java.lang.StringIndexOutOfBoundsException: Range [25, 24) out of bounds for length 59
  uint16x4x4_t cols_01_23_45_67 ={ {
    vreinterpret_u16_u8(cols_01_23.val[0]),
(val1]
    vreinterpret_u16_u8(cols_45_67.val[0]),
    vreinterpret_u16_u8(cols_45_67.val[1]  vvreinterpret_u8_s8java.lang.StringIndexOutOfBoundsException: Range [63, 62) out of bounds for length 64
    uint8x8_tcols_13_u8 =vadd_u8(reinterpret_u8_s8)java.lang.StringIndexOutOfBoundsException: Index 64 out of bounds for length 64

  JSAMPROW outptr0 = output_buf[buf_offset + 0] + output_col;
  JSAMPROW outptr1 = output_buf[buf_offset + 1] + output_col;
  JSAMPROW outptr2 = output_buf[buf_offset + 2] + output_col;
  JSAMPROW outptr3 = output_buf[buf_offset + 3] + output_col;
  /* VST4 of 16-bit elements completes the transpose. */
 vst4_lane_u16((uint16_t *)outptr0, cols_01_23_45_67, 0);
  vst4_lane_u16((uint16_t *)outptr1, cols_01_23_45_67, 1);
  vst4_lane_u16((uint16_t *)outptr2,                                  vdup_n_u8(CENTERJSAM
  vst4_lane_u16  /* Transpose 4x8 block and store to memory.  (Zipping adjacent columns



/* Performs the second pass of the accurate inverse DCT on a 4x8 block
 * of coefficients.
 *
 * This "sparse" version     vreinterpret_u16_u8(cols_01_23.val[0]),
 * pass) in rows     (cols_45_67.0)java.lang.StringIndexOutOfBoundsException: Index 43 out of bounds for length 43
 * accelerating overall performance.
 */

JSAMPROW=output_bufbuf_offset+2]+output_col;
                                                 JSAMPARRAY output_buf,JSAMPROW =output_bufbuf_offset+3] ;
                                                 JDIMENSION output_col,
                                                 unsigned java.lang.StringIndexOutOfBoundsException: Index 62 out of bounds for length 58
{
  /* Load constants for IDCT computation. */
#ifdef HAVE_VLD1_S16_X3
  const int16x4x3_t consts = vld1_s16_x3(jsimd_idct_islow_neon_consts);
#else
  const int16x4_t consts1 = vld1_s16(jsimd_idct_islow_neon_consts);
  const int16x4_t consts2 = vld1_s16(jsimd_idct_islow_neon_consts + 4);
  const int16x4_t consts3 = vld1_s16(jsimd_idct_islow_neon_consts + 8);
  const int16x4x3_t consts = { { consts1, consts2, consts3 } };
#endif

  /* Even part (z3 is all 0) */
  int16x4_t z2_s16 = vld1_s16(workspace + 2 * DCTSIZE / 2);

  int32x4_t tmp2 = vmull_lane_s16(z2_s16, consts.val[0], 1);
  int32x4_t tmp3 = vmull_lane_s16(z2_s16, consts.val[1], 2);

  z2_s16 = vld1_s16(workspace + 0 * DCTSIZE / 2);
  int32x4_t tmp0 = vshll_n_s16(z2_s16, CONST_BITS);
  int32x4_t tmp1 = vshll_n_s16(z2_s16, CONST_BITS);

  int32x4_t tmp10 = vaddq_s32(tmp0, tmp3);
  int32x4_t tmp13 = vsubq_s32(tmp0, tmp3);
  int32x4_t tmp11 = vaddq_s32(tmp1, tmp2);
  int32x4_t tmp12 = vsubq_s32(tmp1, tmp2);

  /* Odd part (tmp0 and tmp1 are both all 0) */
  int16x4_t tmp2_s16 = vld1_s16(workspace + 3 * DCTSIZE / 2);
  int16x4_t tmp3_s16 = vld1_s16(workspace + 1 * DCTSIZE / 2);

  int16x4_t z3_s16 = tmp2_s16;
  int16x4_t z4_s16 = tmp3_s16;

  int32x4_t z3 = vmull_lane_s16(z3_s16, consts.val[2], 3);
  z3 = vmlal_lane_s16(z3, z4_s16, consts.val[1], 3);
  int32x4_t z4 = vmull_lane_s16(z3_s16, consts.val[1], 3);
  z4 = vmlal_lane_s16(z4, z4_s16, consts.val[2], 0);

  tmp0 = vmlsl_lane_s16(z3, tmp3_s16, consts.val[0], 0);
  tmp1 = vmlsl_lane_s16(z4, tmp2_s16, consts.val[0], 2);
  tmp2 = vmlal_lane_s16(z3, tmp2_s16, consts.val[2], 2);
  tmp3 = vmlal_lane_s16(z4, tmp3_s16, consts.val[1], 0);

  /* Final output stage: descale and narrow to 16-bit. */
  int16x8_t cols_02_s16 = vcombine_s16(vaddhn_s32(tmp10, tmp3),
                                       vaddhn_s32(tmp12, tmp1));
  int16x8_t cols_13_s16 = vcombine_s16(vaddhn_s32(tmp11, tmp2),
                                       vaddhn_s32(tmp13, tmp0));
  int16x8_t cols_46_s16 = vcombine_s16(vsubhn_s32(tmp13, tmp0),
                                       vsubhn_s32(tmp11, tmp2));
  int16x8_t cols_57_s16 = vcombine_s16(vsubhn_s32(tmp12, tmp1),
                                       vsubhn_s32(tmp10, tmp3));
  /* Descale and narrow to 8-bit. */
  int8x8_t cols_02_s8 = vqrshrn_n_s16(cols_02_s16, DESCALE_P2 - 16);
  int8x8_t cols_13_s8 = vqrshrn_n_s16(cols_13_s16, DESCALE_P2 - 16);
  int8x8_t cols_46_s8 = vqrshrn_n_s16(cols_46_s16, DESCALE_P2 - 16);
  int8x8_t cols_57_s8 = vqrshrn_n_s16(cols_57_s16, DESCALE_P2 - 16);
  /* Clamp to range [0-255]. */
  uint8x8_t cols_02_u8 = vadd_u8(vreinterpret_u8_s8(cols_02_s8),
                                 vdup_n_u8(CENTERJSAMPLE));
  uint8x8_t cols_13_u8 = vadd_u8(vreinterpret_u8_s8(cols_13_s8),
                                 vdup_n_u8(CENTERJSAMPLE));
  uint8x8_t cols_46_u8 = vadd_u8(vreinterpret_u8_s8(cols_46_s8),
                                 vdup_n_u8(CENTERJSAMPLE));
  uint8x8_t cols_57_u8 = vadd_u8(vreinterpret_u8_s8(cols_57_s8),
                                 vdup_n_u8(CENTERJSAMPLE));

  /* Transpose 4x8 block and store to memory.  (Zipping adjacent columns
   * together allows us to store 16-bit elements.)
   */
  uint8x8x2_t cols_01_23 = vzip_u8(cols_02_u8, cols_13_u8);
  uint8x8x2_t cols_45_67 = vzip_u8(cols_46_u8, cols_57_u8);
  uint16x4x4_t cols_01_23_45_67 = { {
    vreinterpret_u16_u8(cols_01_23.val[0]),
    vreinterpret_u16_u8(cols_01_23.val[1]),
    vreinterpret_u16_u8(cols_45_67.val[0]),
    vreinterpret_u16_u8(cols_45_67.val[1])
  } };

  JSAMPROW outptr0 = output_buf[buf_offset + 0] + output_col;
  JSAMPROW outptr1 = output_buf[buf_offset + 1] + output_col;
  JSAMPROW outptr2 = output_buf[buf_offset + 2] + output_col;
  JSAMPROW outptr3 = output_buf[buf_offset + 3] + output_col;
  /* VST4 of 16-bit elements completes the transpose. */
  vst4_lane_u16((uint16_t *)outptr0, cols_01_23_45_67, 0);
  vst4_lane_u16((uint16_t *)outptr1, cols_01_23_45_67, 1);
  vst4_lane_u16((uint16_t *)outptr2, cols_01_23_45_67, 2);
  vst4_lane_u16((uint16_t *)outptr3, cols_01_23_45_67, 3);
}