diff options
Diffstat (limited to 'vpx_dsp/x86/highbd_quantize_intrin_sse2.c')
| -rw-r--r-- | vpx_dsp/x86/highbd_quantize_intrin_sse2.c | 40 |
1 files changed, 20 insertions, 20 deletions
diff --git a/vpx_dsp/x86/highbd_quantize_intrin_sse2.c b/vpx_dsp/x86/highbd_quantize_intrin_sse2.c index ae1981a83..adae60756 100644 --- a/vpx_dsp/x86/highbd_quantize_intrin_sse2.c +++ b/vpx_dsp/x86/highbd_quantize_intrin_sse2.c @@ -15,19 +15,22 @@ #include "vpx_dsp/vpx_dsp_common.h" #include "vpx_mem/vpx_mem.h" #include "vpx_ports/mem.h" +#include "vp9/common/vp9_scan.h" +#include "vp9/encoder/vp9_block.h" -#if CONFIG_VP9_HIGHBITDEPTH void vpx_highbd_quantize_b_sse2(const tran_low_t *coeff_ptr, intptr_t count, - const int16_t *zbin_ptr, - const int16_t *round_ptr, - const int16_t *quant_ptr, - const int16_t *quant_shift_ptr, + const struct macroblock_plane *mb_plane, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr, - const int16_t *scan, const int16_t *iscan) { + const struct ScanOrder *const scan_order) { int i, j, non_zero_regs = (int)count / 4, eob_i = 0; __m128i zbins[2]; __m128i nzbins[2]; + const int16_t *iscan = scan_order->iscan; + const int16_t *zbin_ptr = mb_plane->zbin; + const int16_t *round_ptr = mb_plane->round; + const int16_t *quant_ptr = mb_plane->quant; + const int16_t *quant_shift_ptr = mb_plane->quant_shift; zbins[0] = _mm_set_epi32((int)zbin_ptr[1], (int)zbin_ptr[1], (int)zbin_ptr[1], (int)zbin_ptr[0]); @@ -38,8 +41,6 @@ void vpx_highbd_quantize_b_sse2(const tran_low_t *coeff_ptr, intptr_t count, nzbins[0] = _mm_sub_epi32(nzbins[0], zbins[0]); nzbins[1] = _mm_sub_epi32(nzbins[1], zbins[1]); - (void)scan; - memset(qcoeff_ptr, 0, count * sizeof(*qcoeff_ptr)); memset(dqcoeff_ptr, 0, count * sizeof(*dqcoeff_ptr)); @@ -93,19 +94,18 @@ void vpx_highbd_quantize_b_sse2(const tran_low_t *coeff_ptr, intptr_t count, } void vpx_highbd_quantize_b_32x32_sse2( - const tran_low_t *coeff_ptr, intptr_t n_coeffs, const int16_t *zbin_ptr, - const int16_t *round_ptr, const int16_t *quant_ptr, - const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, - tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr, - const int16_t *scan, const int16_t *iscan) { + const tran_low_t *coeff_ptr, const struct macroblock_plane *const mb_plane, + tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, + uint16_t *eob_ptr, const struct ScanOrder *const scan_order) { __m128i zbins[2]; __m128i nzbins[2]; int idx = 0; int idx_arr[1024]; int i, eob = 0; - const int zbin0_tmp = ROUND_POWER_OF_TWO(zbin_ptr[0], 1); - const int zbin1_tmp = ROUND_POWER_OF_TWO(zbin_ptr[1], 1); - (void)scan; + const intptr_t n_coeffs = 32 * 32; + const int16_t *iscan = scan_order->iscan; + const int zbin0_tmp = ROUND_POWER_OF_TWO(mb_plane->zbin[0], 1); + const int zbin1_tmp = ROUND_POWER_OF_TWO(mb_plane->zbin[1], 1); zbins[0] = _mm_set_epi32(zbin1_tmp, zbin1_tmp, zbin1_tmp, zbin0_tmp); zbins[1] = _mm_set1_epi32(zbin1_tmp); @@ -140,14 +140,14 @@ void vpx_highbd_quantize_b_32x32_sse2( const int coeff = coeff_ptr[rc]; const int coeff_sign = (coeff >> 31); const int abs_coeff = (coeff ^ coeff_sign) - coeff_sign; - const int64_t tmp1 = abs_coeff + ROUND_POWER_OF_TWO(round_ptr[rc != 0], 1); - const int64_t tmp2 = ((tmp1 * quant_ptr[rc != 0]) >> 16) + tmp1; + const int64_t tmp1 = + abs_coeff + ROUND_POWER_OF_TWO(mb_plane->round[rc != 0], 1); + const int64_t tmp2 = ((tmp1 * mb_plane->quant[rc != 0]) >> 16) + tmp1; const uint32_t abs_qcoeff = - (uint32_t)((tmp2 * quant_shift_ptr[rc != 0]) >> 15); + (uint32_t)((tmp2 * mb_plane->quant_shift[rc != 0]) >> 15); qcoeff_ptr[rc] = (int)(abs_qcoeff ^ (uint32_t)coeff_sign) - coeff_sign; dqcoeff_ptr[rc] = qcoeff_ptr[rc] * dequant_ptr[rc != 0] / 2; if (abs_qcoeff) eob = iscan[idx_arr[i]] > eob ? iscan[idx_arr[i]] : eob; } *eob_ptr = eob; } -#endif |