diff options
author | Srujan Vandrangi <srujan.vandrangi@ittiam.com> | 2024-04-24 22:27:04 +0530 |
---|---|---|
committer | Srujan Vandrangi <srujan.vandrangi@ittiam.com> | 2024-04-24 17:01:56 +0000 |
commit | 6e8f8ec91d5b38031bb3aaab304f1364ef2c677b (patch) | |
tree | 6f625d7c45940f579530132d178bec86495c5cfb | |
parent | 21f3eddcb038076220c6da4e4cdd7ecec88a4f1c (diff) | |
parent | 828cdb77ba434eb11ce19f7f0d4482e744b27064 (diff) | |
download | libavc-master.tar.gz |
This project was upgraded with external_updater.
Usage: tools/external_updater/updater.sh update external/libavc
For more info, check https://cs.android.com/android/platform/superproject/+/main:tools/external_updater/README.md
Bug: 325356593
Test: atest CtsMediaV2TestCases
Change-Id: If5cfbadef49d467bb00914988690cab24e80070e
35 files changed, 458 insertions, 1999 deletions
@@ -1,19 +1,19 @@ # This project was upgraded with external_updater. -# Usage: tools/external_updater/updater.sh update libavc +# Usage: tools/external_updater/updater.sh update external/libavc # For more info, check https://cs.android.com/android/platform/superproject/+/main:tools/external_updater/README.md name: "libavc" description: "Android fork of the libavc library." third_party { - url { - type: GIT - value: "https://github.com/ittiam-systems/libavc.git" - } - version: "v1.3.0" license_type: NOTICE last_upgrade_date { - year: 2023 - month: 10 - day: 23 + year: 2024 + month: 4 + day: 24 + } + identifier { + type: "Git" + value: "https://github.com/ittiam-systems/libavc.git" + version: "v1.4.0" } } diff --git a/common/arm/svc/isvc_iquant_itrans_recon_neon.c b/common/arm/svc/isvc_iquant_itrans_recon_neon.c index 270adde..8a97fbc 100644 --- a/common/arm/svc/isvc_iquant_itrans_recon_neon.c +++ b/common/arm/svc/isvc_iquant_itrans_recon_neon.c @@ -587,193 +587,6 @@ void isvc_iquant_itrans_recon_4x4_with_res_accumulate_neon( vreinterpret_u32_u8(pred23_un), 1); } -void isvc_iquant_itrans_recon_chroma_4x4_neon( - buffer_container_t *ps_src, buffer_container_t *ps_pred, buffer_container_t *ps_res_pred, - buffer_container_t *ps_res, buffer_container_t *ps_rec, - iq_it_res_rec_constants_t *ps_iq_it_res_rec_constants, WORD16 *pi2_tmp, WORD16 *pi2_dc_src, - WORD32 i4_iq_start_idx, UWORD8 u1_res_accumulate) -{ - WORD16 *pi2_src = (WORD16 *) ps_src->pv_data; - UWORD8 *pu1_pred = (UWORD8 *) ps_pred->pv_data; - UWORD8 *pu1_out = (UWORD8 *) ps_rec->pv_data; - WORD32 i4_pred_stride = ps_pred->i4_data_stride; - WORD32 i4_out_stride = ps_rec->i4_data_stride; - const UWORD16 *pu2_iscal_mat = ps_iq_it_res_rec_constants->pu2_iscal_mat; - const UWORD16 *pu2_weigh_mat = ps_iq_it_res_rec_constants->pu2_weigh_mat; - UWORD32 u4_qp_div_6 = ps_iq_it_res_rec_constants->u4_qp_div_6; - - WORD16 i2_rnd_factor = (u4_qp_div_6 < 4) ? 1 << (3 - u4_qp_div_6) : 0; - - int16x4x4_t src_16x4x2; - int16x4x4_t iscal_16x4x2; - int16x4x4_t weigh_16x4x2; - - int16x4_t q0_16x4, q1_16x4, q2_16x4, q3_16x4; - int32x4_t q0_32x4, q1_32x4, q2_32x4, q3_32x4; - int16x4_t rq1_16x4, rq3_16x4; - int16x4_t x0_16x4, x1_16x4, x2_16x4, x3_16x4; - int16x8_t x0_16x8, x1_16x8, x2_16x8, x3_16x8; - int16x4_t xx0_16x4, xx1_16x4, xx2_16x4, xx3_16x4; - int16x4x2_t xx0_16x4x2, xx1_16x4x2; - int32x2x2_t x0_32x2x2, x1_32x2x2; - int16x4_t weigh0_16x4, weigh1_16x4, weigh2_16x4, weigh3_16x4; - - uint8x8_t pred0_in, pred1_in, pred2_in, pred3_in; - int16x8_t pred0, pred1, pred2, pred3; - int16x8_t rec0, rec1, rec2, rec3; - uint8x8_t rec0_un, rec1_un, rec2_un, rec3_un; - uint8x8_t out0, out1, out2, out3; - - uint8x8_t chroma_mask_8x8 = vreinterpret_u8_u16(vdup_n_u16(0x00ff)); - - int16x4_t pos_255_16x4 = vdup_n_s16(((WORD16) UINT8_MAX)); - int16x4_t neg_255_16x4 = vdup_n_s16(-((WORD16) UINT8_MAX)); - int32x4_t qp_div_6_32x4 = vdupq_n_s32(u4_qp_div_6); - int32x4_t rnd_fact = vdupq_n_s32(i2_rnd_factor); - - UNUSED(i4_iq_start_idx); - UNUSED(ps_res); - UNUSED(ps_res_pred); - UNUSED(u1_res_accumulate); - - src_16x4x2 = vld4_s16(pi2_src); - iscal_16x4x2 = vld4_s16((const int16_t *) pu2_iscal_mat); - weigh_16x4x2 = vld4_s16((const int16_t *) pu2_weigh_mat); - - weigh0_16x4 = vmul_s16(weigh_16x4x2.val[0], iscal_16x4x2.val[0]); - weigh1_16x4 = vmul_s16(weigh_16x4x2.val[1], iscal_16x4x2.val[1]); - weigh2_16x4 = vmul_s16(weigh_16x4x2.val[2], iscal_16x4x2.val[2]); - weigh3_16x4 = vmul_s16(weigh_16x4x2.val[3], iscal_16x4x2.val[3]); - - q0_32x4 = vmull_s16(weigh0_16x4, src_16x4x2.val[0]); - q1_32x4 = vmull_s16(weigh1_16x4, src_16x4x2.val[1]); - q2_32x4 = vmull_s16(weigh2_16x4, src_16x4x2.val[2]); - q3_32x4 = vmull_s16(weigh3_16x4, src_16x4x2.val[3]); - - q0_32x4 = vaddq_s32(q0_32x4, rnd_fact); - q1_32x4 = vaddq_s32(q1_32x4, rnd_fact); - q2_32x4 = vaddq_s32(q2_32x4, rnd_fact); - q3_32x4 = vaddq_s32(q3_32x4, rnd_fact); - - q0_32x4 = vshlq_s32(q0_32x4, qp_div_6_32x4); - q1_32x4 = vshlq_s32(q1_32x4, qp_div_6_32x4); - q2_32x4 = vshlq_s32(q2_32x4, qp_div_6_32x4); - q3_32x4 = vshlq_s32(q3_32x4, qp_div_6_32x4); - - q0_16x4 = vqshrn_n_s32(q0_32x4, 4); - q1_16x4 = vqshrn_n_s32(q1_32x4, 4); - q2_16x4 = vqshrn_n_s32(q2_32x4, 4); - q3_16x4 = vqshrn_n_s32(q3_32x4, 4); - - q0_16x4 = vset_lane_s16(pi2_dc_src[0], q0_16x4, 0); - - rq1_16x4 = vshr_n_s16(q1_16x4, 1); - rq3_16x4 = vshr_n_s16(q3_16x4, 1); - - x0_16x4 = vadd_s16(q0_16x4, q2_16x4); - x1_16x4 = vsub_s16(q0_16x4, q2_16x4); - x2_16x4 = vsub_s16(rq1_16x4, q3_16x4); - x3_16x4 = vadd_s16(q1_16x4, rq3_16x4); - - xx0_16x4 = vadd_s16(x0_16x4, x3_16x4); - xx1_16x4 = vadd_s16(x1_16x4, x2_16x4); - xx2_16x4 = vsub_s16(x1_16x4, x2_16x4); - xx3_16x4 = vsub_s16(x0_16x4, x3_16x4); - - /* row 0 to row 3 */ - xx0_16x4x2 = vtrn_s16(xx0_16x4, xx1_16x4); - xx1_16x4x2 = vtrn_s16(xx2_16x4, xx3_16x4); - x0_32x2x2 = - vzip_s32(vreinterpret_s32_s16(xx0_16x4x2.val[0]), vreinterpret_s32_s16(xx1_16x4x2.val[0])); - x1_32x2x2 = - vzip_s32(vreinterpret_s32_s16(xx0_16x4x2.val[1]), vreinterpret_s32_s16(xx1_16x4x2.val[1])); - - x0_16x4 = vreinterpret_s16_s32(x0_32x2x2.val[0]); - x1_16x4 = vreinterpret_s16_s32(x1_32x2x2.val[0]); - x2_16x4 = vreinterpret_s16_s32(x0_32x2x2.val[1]); - x3_16x4 = vreinterpret_s16_s32(x1_32x2x2.val[1]); - - /* Store Horz transform output into temp */ - vst1_s16(pi2_tmp, x0_16x4); - vst1_s16(pi2_tmp + 4, x1_16x4); - vst1_s16(pi2_tmp + 8, x2_16x4); - vst1_s16(pi2_tmp + 12, x3_16x4); - - /* vertical inverse transform */ - rq1_16x4 = vshr_n_s16(x1_16x4, 1); - rq3_16x4 = vshr_n_s16(x3_16x4, 1); - - xx0_16x4 = vadd_s16(x0_16x4, x2_16x4); - xx1_16x4 = vsub_s16(x0_16x4, x2_16x4); - xx2_16x4 = vsub_s16(rq1_16x4, x3_16x4); - xx3_16x4 = vadd_s16(x1_16x4, rq3_16x4); - - x0_16x4 = vadd_s16(xx0_16x4, xx3_16x4); - x1_16x4 = vadd_s16(xx1_16x4, xx2_16x4); - x2_16x4 = vsub_s16(xx1_16x4, xx2_16x4); - x3_16x4 = vsub_s16(xx0_16x4, xx3_16x4); - - x0_16x4 = vrshr_n_s16(x0_16x4, 6); - x1_16x4 = vrshr_n_s16(x1_16x4, 6); - x2_16x4 = vrshr_n_s16(x2_16x4, 6); - x3_16x4 = vrshr_n_s16(x3_16x4, 6); - - /* Saturate all values < -255 to -255 and retain the rest as it is */ - x0_16x4 = vmax_s16(x0_16x4, neg_255_16x4); - x1_16x4 = vmax_s16(x1_16x4, neg_255_16x4); - x2_16x4 = vmax_s16(x2_16x4, neg_255_16x4); - x3_16x4 = vmax_s16(x3_16x4, neg_255_16x4); - - /* Saturate all values > 255 to 255 and retain the rest as it is */ - x0_16x4 = vmin_s16(x0_16x4, pos_255_16x4); - x1_16x4 = vmin_s16(x1_16x4, pos_255_16x4); - x2_16x4 = vmin_s16(x2_16x4, pos_255_16x4); - x3_16x4 = vmin_s16(x3_16x4, pos_255_16x4); - - x0_16x8 = vreinterpretq_s16_s32(vmovl_s16(x0_16x4)); - x1_16x8 = vreinterpretq_s16_s32(vmovl_s16(x1_16x4)); - x2_16x8 = vreinterpretq_s16_s32(vmovl_s16(x2_16x4)); - x3_16x8 = vreinterpretq_s16_s32(vmovl_s16(x3_16x4)); - - pred0_in = vld1_u8((uint8_t *) pu1_pred); - pred1_in = vld1_u8((uint8_t *) pu1_pred + (i4_pred_stride)); - pred2_in = vld1_u8((uint8_t *) pu1_pred + (i4_pred_stride << 1)); - pred3_in = vld1_u8((uint8_t *) pu1_pred + (i4_pred_stride * 3)); - - pred0 = vreinterpretq_s16_u16(vmovl_u8(pred0_in)); - pred1 = vreinterpretq_s16_u16(vmovl_u8(pred1_in)); - pred2 = vreinterpretq_s16_u16(vmovl_u8(pred2_in)); - pred3 = vreinterpretq_s16_u16(vmovl_u8(pred3_in)); - - /* Out pixel = pred + res */ - rec0 = vaddq_s16(pred0, x0_16x8); - rec1 = vaddq_s16(pred1, x1_16x8); - rec2 = vaddq_s16(pred2, x2_16x8); - rec3 = vaddq_s16(pred3, x3_16x8); - - out0 = vld1_u8(pu1_out); - out1 = vld1_u8(pu1_out + i4_out_stride); - out2 = vld1_u8(pu1_out + i4_out_stride * 2); - out3 = vld1_u8(pu1_out + i4_out_stride * 3); - - /* Convert to 8 bit unsigned with saturation */ - rec0_un = vqmovun_s16(rec0); - rec1_un = vqmovun_s16(rec1); - rec2_un = vqmovun_s16(rec2); - rec3_un = vqmovun_s16(rec3); - - /* Store in alternate postions */ - out0 = vbsl_u8(chroma_mask_8x8, rec0_un, out0); - out1 = vbsl_u8(chroma_mask_8x8, rec1_un, out1); - out2 = vbsl_u8(chroma_mask_8x8, rec2_un, out2); - out3 = vbsl_u8(chroma_mask_8x8, rec3_un, out3); - - vst1_u8((pu1_out), out0); - vst1_u8((pu1_out + i4_out_stride), out1); - vst1_u8((pu1_out + (i4_out_stride << 1)), out2); - vst1_u8((pu1_out + ((i4_out_stride << 1) + i4_out_stride)), out3); -} - void isvc_iquant_itrans_recon_chroma_4x4_with_res_output_neon( buffer_container_t *ps_src, buffer_container_t *ps_pred, buffer_container_t *ps_res_pred, buffer_container_t *ps_res, buffer_container_t *ps_rec, @@ -1280,271 +1093,6 @@ void isvc_iquant_itrans_recon_4x4_dc_neon(buffer_container_t *ps_src, buffer_con vst1_lane_u32((uint32_t *) (pu1_out + i4_out_stride * 3), vreinterpret_u32_u8(pred3_in), 0); } -void isvc_iquant_itrans_recon_4x4_dc_with_res_output_neon( - buffer_container_t *ps_src, buffer_container_t *ps_pred, buffer_container_t *ps_res_pred, - buffer_container_t *ps_res, buffer_container_t *ps_rec, - iq_it_res_rec_constants_t *ps_iq_it_res_rec_constants, WORD16 *pi2_tmp, WORD16 *pi2_dc_src, - WORD32 i4_iq_start_idx, UWORD8 u1_res_accumulate) -{ - WORD16 *pi2_src = (WORD16 *) ps_src->pv_data; - WORD16 *pi2_res = (WORD16 *) ps_res->pv_data; - UWORD8 *pu1_pred = (UWORD8 *) ps_pred->pv_data; - UWORD8 *pu1_out = (UWORD8 *) ps_rec->pv_data; - WORD32 i4_res_stride = ps_res->i4_data_stride; - WORD32 i4_pred_stride = ps_pred->i4_data_stride; - WORD32 i4_out_stride = ps_rec->i4_data_stride; - const UWORD16 *pu2_iscal_mat = ps_iq_it_res_rec_constants->pu2_iscal_mat; - const UWORD16 *pu2_weigh_mat = ps_iq_it_res_rec_constants->pu2_weigh_mat; - UWORD32 u4_qp_div_6 = ps_iq_it_res_rec_constants->u4_qp_div_6; - WORD16 rnd_fact = (u4_qp_div_6 < 4) ? 1 << (3 - u4_qp_div_6) : 0; - - WORD16 i2_it_out; - WORD32 i4_iq_out_temp; - int16x8_t temp_0; - int16x4_t residue_res; - uint8x8_t pred0_in, pred1_in, pred2_in, pred3_in; - int16x8_t pred0, pred1, pred2, pred3; - - UNUSED(pi2_tmp); - UNUSED(ps_res_pred); - UNUSED(u1_res_accumulate); - - if(i4_iq_start_idx == 0) - { - i4_iq_out_temp = pi2_src[0]; - INV_QUANT(i4_iq_out_temp, pu2_iscal_mat[0], pu2_weigh_mat[0], u4_qp_div_6, rnd_fact, 4); - } - else - { - i4_iq_out_temp = pi2_dc_src[0]; - } - - i2_it_out = ((i4_iq_out_temp + 32) >> 6); - temp_0 = vdupq_n_s16(i2_it_out); - residue_res = vdup_n_s16(isvc_get_residue(i2_it_out, 0, 0)); - - vst1_s16(pi2_res, residue_res); - vst1_s16(pi2_res + i4_res_stride, residue_res); - vst1_s16(pi2_res + (i4_res_stride << 1), residue_res); - vst1_s16(pi2_res + (i4_res_stride << 1) + i4_res_stride, residue_res); - - pred0_in = vld1_u8(pu1_pred); - pu1_pred = pu1_pred + i4_pred_stride; - pred1_in = vld1_u8(pu1_pred); - pu1_pred = pu1_pred + i4_pred_stride; - pred2_in = vld1_u8(pu1_pred); - pu1_pred = pu1_pred + i4_pred_stride; - pred3_in = vld1_u8(pu1_pred); - - pred0 = vreinterpretq_s16_u16(vmovl_u8(pred0_in)); - pred1 = vreinterpretq_s16_u16(vmovl_u8(pred1_in)); - pred2 = vreinterpretq_s16_u16(vmovl_u8(pred2_in)); - pred3 = vreinterpretq_s16_u16(vmovl_u8(pred3_in)); - - /* Out pixel = Res + pred */ - pred0 = vaddq_s16(pred0, temp_0); - pred1 = vaddq_s16(pred1, temp_0); - pred2 = vaddq_s16(pred2, temp_0); - pred3 = vaddq_s16(pred3, temp_0); - - /* Convert to unsigned 8 bit with saturation */ - pred0_in = vqmovun_s16(pred0); - pred1_in = vqmovun_s16(pred1); - pred2_in = vqmovun_s16(pred2); - pred3_in = vqmovun_s16(pred3); - - vst1_lane_u32((uint32_t *) (pu1_out), vreinterpret_u32_u8(pred0_in), 0); - vst1_lane_u32((uint32_t *) (pu1_out + i4_out_stride), vreinterpret_u32_u8(pred1_in), 0); - vst1_lane_u32((uint32_t *) (pu1_out + i4_out_stride * 2), vreinterpret_u32_u8(pred2_in), 0); - vst1_lane_u32((uint32_t *) (pu1_out + i4_out_stride * 3), vreinterpret_u32_u8(pred3_in), 0); -} - -void isvc_iquant_itrans_recon_4x4_dc_with_res_accumulate_neon( - buffer_container_t *ps_src, buffer_container_t *ps_pred, buffer_container_t *ps_res_pred, - buffer_container_t *ps_res, buffer_container_t *ps_rec, - iq_it_res_rec_constants_t *ps_iq_it_res_rec_constants, WORD16 *pi2_tmp, WORD16 *pi2_dc_src, - WORD32 i4_iq_start_idx, UWORD8 u1_res_accumulate) -{ - WORD16 *pi2_src = (WORD16 *) ps_src->pv_data; - WORD16 *pi2_res = (WORD16 *) ps_res->pv_data; - WORD16 *pi2_res_pred = (WORD16 *) ps_res_pred->pv_data; - UWORD8 *pu1_pred = (UWORD8 *) ps_pred->pv_data; - UWORD8 *pu1_out = (UWORD8 *) ps_rec->pv_data; - WORD32 i4_res_stride = ps_res->i4_data_stride; - WORD32 i4_res_pred_stride = ps_res_pred->i4_data_stride; - WORD32 i4_pred_stride = ps_pred->i4_data_stride; - WORD32 i4_out_stride = ps_rec->i4_data_stride; - const UWORD16 *pu2_iscal_mat = ps_iq_it_res_rec_constants->pu2_iscal_mat; - const UWORD16 *pu2_weigh_mat = ps_iq_it_res_rec_constants->pu2_weigh_mat; - UWORD32 u4_qp_div_6 = ps_iq_it_res_rec_constants->u4_qp_div_6; - WORD16 rnd_fact = (u4_qp_div_6 < 4) ? 1 << (3 - u4_qp_div_6) : 0; - - WORD32 i4_iq_out_temp; - int16x4_t temp_0; - uint8x8_t pred0_in, pred1_in, pred2_in, pred3_in; - int16x8_t pred0, pred1, pred2, pred3; - int16x8_t pred01_in, pred23_in; - uint8x8_t pred01_un, pred23_un; - - int16x4_t resd0_in, resd1_in, resd2_in, resd3_in; - int16x8_t resd01_in, resd23_in; - int16x4_t pos_255 = vdup_n_s16(((WORD16) UINT8_MAX)); - int16x4_t neg_255 = vdup_n_s16(-((WORD16) UINT8_MAX)); - - UNUSED(pi2_tmp); - UNUSED(u1_res_accumulate); - - if(i4_iq_start_idx == 0) - { - i4_iq_out_temp = pi2_src[0]; - INV_QUANT(i4_iq_out_temp, pu2_iscal_mat[0], pu2_weigh_mat[0], u4_qp_div_6, rnd_fact, 4); - } - else - { - i4_iq_out_temp = pi2_dc_src[0]; - } - - temp_0 = vdup_n_s16((i4_iq_out_temp + 32) >> 6); - - resd0_in = vld1_s16((int16_t *) pi2_res_pred); - resd1_in = vld1_s16((int16_t *) pi2_res_pred + i4_res_pred_stride); - resd2_in = vld1_s16((int16_t *) pi2_res_pred + (i4_res_pred_stride * 2)); - resd3_in = vld1_s16((int16_t *) pi2_res_pred + (i4_res_pred_stride * 3)); - - /* Add res pred to the res obtained */ - resd0_in = vadd_s16(resd0_in, temp_0); - resd1_in = vadd_s16(resd1_in, temp_0); - resd2_in = vadd_s16(resd2_in, temp_0); - resd3_in = vadd_s16(resd3_in, temp_0); - - /* Saturate all values < -255 to -255 and retain the rest as it is */ - resd0_in = vmax_s16(resd0_in, neg_255); - resd1_in = vmax_s16(resd1_in, neg_255); - resd2_in = vmax_s16(resd2_in, neg_255); - resd3_in = vmax_s16(resd3_in, neg_255); - - /* Saturate all values > 255 to 255 and retain the rest as it is */ - resd0_in = vmin_s16(resd0_in, pos_255); - resd1_in = vmin_s16(resd1_in, pos_255); - resd2_in = vmin_s16(resd2_in, pos_255); - resd3_in = vmin_s16(resd3_in, pos_255); - - vst1_s16(pi2_res, resd0_in); - vst1_s16(pi2_res + i4_res_stride, resd1_in); - vst1_s16(pi2_res + (i4_res_stride << 1), resd2_in); - vst1_s16(pi2_res + (i4_res_stride << 1) + i4_res_stride, resd3_in); - - resd01_in = vcombine_s16(resd0_in, resd1_in); - resd23_in = vcombine_s16(resd2_in, resd3_in); - - pred0_in = vld1_u8(pu1_pred); - pu1_pred = pu1_pred + i4_pred_stride; - pred1_in = vld1_u8(pu1_pred); - pu1_pred = pu1_pred + i4_pred_stride; - pred2_in = vld1_u8(pu1_pred); - pu1_pred = pu1_pred + i4_pred_stride; - pred3_in = vld1_u8(pu1_pred); - - pred0 = vreinterpretq_s16_u16(vmovl_u8(pred0_in)); - pred1 = vreinterpretq_s16_u16(vmovl_u8(pred1_in)); - pred2 = vreinterpretq_s16_u16(vmovl_u8(pred2_in)); - pred3 = vreinterpretq_s16_u16(vmovl_u8(pred3_in)); - - pred01_in = vcombine_s16(vget_low_s16(pred0), vget_low_s16(pred1)); - pred23_in = vcombine_s16(vget_low_s16(pred2), vget_low_s16(pred3)); - - /* Out pixel = Res + pred */ - pred01_in = vaddq_s16(pred01_in, resd01_in); - pred23_in = vaddq_s16(pred23_in, resd23_in); - - /* Convert to unsigned 8 bit with saturation */ - pred01_un = vqmovun_s16(pred01_in); - pred23_un = vqmovun_s16(pred23_in); - - vst1_lane_u32((uint32_t *) (pu1_out), vreinterpret_u32_u8(pred01_un), 0); - vst1_lane_u32((uint32_t *) (pu1_out + i4_out_stride), vreinterpret_u32_u8(pred01_un), 1); - vst1_lane_u32((uint32_t *) (pu1_out + (i4_out_stride << 1)), vreinterpret_u32_u8(pred23_un), 0); - vst1_lane_u32((uint32_t *) (pu1_out + ((i4_out_stride << 1) + i4_out_stride)), - vreinterpret_u32_u8(pred23_un), 1); -} - -void isvc_iquant_itrans_recon_chroma_4x4_dc_neon( - buffer_container_t *ps_src, buffer_container_t *ps_pred, buffer_container_t *ps_res_pred, - buffer_container_t *ps_res, buffer_container_t *ps_rec, - iq_it_res_rec_constants_t *ps_iq_it_res_rec_constants, WORD16 *pi2_tmp, WORD16 *pi2_dc_src, - WORD32 i4_iq_start_idx, UWORD8 u1_res_accumulate) -{ - WORD16 *pi2_src = (WORD16 *) ps_src->pv_data; - UWORD8 *pu1_pred = (UWORD8 *) ps_pred->pv_data; - UWORD8 *pu1_out = (UWORD8 *) ps_rec->pv_data; - WORD32 i4_pred_stride = ps_pred->i4_data_stride; - WORD32 i4_out_stride = ps_rec->i4_data_stride; - const UWORD16 *pu2_iscal_mat = ps_iq_it_res_rec_constants->pu2_iscal_mat; - const UWORD16 *pu2_weigh_mat = ps_iq_it_res_rec_constants->pu2_weigh_mat; - UWORD32 u4_qp_div_6 = ps_iq_it_res_rec_constants->u4_qp_div_6; - - WORD32 i4_iq_out_temp; - int16x8_t temp_0; - uint8x8_t pred0_in, pred1_in, pred2_in, pred3_in; - int16x8_t pred0, pred1, pred2, pred3; - uint8x8_t i4_out_horz_8x8_r0, i4_out_horz_8x8_r1, i4_out_horz_8x8_r2, i4_out_horz_8x8_r3; - uint8x8_t chroma_mask_8x8 = vreinterpret_u8_u16(vdup_n_u16(0x00ff)); - - UNUSED(pi2_src); - UNUSED(pu2_iscal_mat); - UNUSED(pu2_weigh_mat); - UNUSED(u4_qp_div_6); - UNUSED(pi2_tmp); - UNUSED(i4_iq_start_idx); - UNUSED(ps_res); - UNUSED(ps_res_pred); - UNUSED(u1_res_accumulate); - - i4_iq_out_temp = pi2_dc_src[0]; - temp_0 = vdupq_n_s16((i4_iq_out_temp + 32) >> 6); - - pred0_in = vld1_u8(pu1_pred); - pu1_pred = pu1_pred + i4_pred_stride; - pred1_in = vld1_u8(pu1_pred); - pu1_pred = pu1_pred + i4_pred_stride; - pred2_in = vld1_u8(pu1_pred); - pu1_pred = pu1_pred + i4_pred_stride; - pred3_in = vld1_u8(pu1_pred); - - pred0 = vreinterpretq_s16_u16(vmovl_u8(pred0_in)); - pred1 = vreinterpretq_s16_u16(vmovl_u8(pred1_in)); - pred2 = vreinterpretq_s16_u16(vmovl_u8(pred2_in)); - pred3 = vreinterpretq_s16_u16(vmovl_u8(pred3_in)); - - /* Out pixel = Res + pred */ - pred0 = vaddq_s16(pred0, temp_0); - pred1 = vaddq_s16(pred1, temp_0); - pred2 = vaddq_s16(pred2, temp_0); - pred3 = vaddq_s16(pred3, temp_0); - - /* Convert to unsigned 8 bit with saturation */ - pred0_in = vqmovun_s16(pred0); - pred1_in = vqmovun_s16(pred1); - pred2_in = vqmovun_s16(pred2); - pred3_in = vqmovun_s16(pred3); - - i4_out_horz_8x8_r0 = vld1_u8(pu1_out); - i4_out_horz_8x8_r1 = vld1_u8(pu1_out + i4_out_stride); - i4_out_horz_8x8_r2 = vld1_u8(pu1_out + i4_out_stride * 2); - i4_out_horz_8x8_r3 = vld1_u8(pu1_out + i4_out_stride * 3); - - /* Store out pixels in alternate positions */ - i4_out_horz_8x8_r0 = vbsl_u8(chroma_mask_8x8, pred0_in, i4_out_horz_8x8_r0); - i4_out_horz_8x8_r1 = vbsl_u8(chroma_mask_8x8, pred1_in, i4_out_horz_8x8_r1); - i4_out_horz_8x8_r2 = vbsl_u8(chroma_mask_8x8, pred2_in, i4_out_horz_8x8_r2); - i4_out_horz_8x8_r3 = vbsl_u8(chroma_mask_8x8, pred3_in, i4_out_horz_8x8_r3); - - vst1_u8((uint8_t *) (pu1_out), i4_out_horz_8x8_r0); - vst1_u8((uint8_t *) (pu1_out + i4_out_stride), i4_out_horz_8x8_r1); - vst1_u8((uint8_t *) (pu1_out + i4_out_stride * 2), i4_out_horz_8x8_r2); - vst1_u8((uint8_t *) (pu1_out + i4_out_stride * 3), i4_out_horz_8x8_r3); -} - void isvc_iquant_itrans_recon_chroma_4x4_dc_with_res_output_neon( buffer_container_t *ps_src, buffer_container_t *ps_pred, buffer_container_t *ps_res_pred, buffer_container_t *ps_res, buffer_container_t *ps_rec, diff --git a/common/svc/isvc_mem_fns.c b/common/svc/isvc_mem_fns.c index 345715a..35a4c66 100644 --- a/common/svc/isvc_mem_fns.c +++ b/common/svc/isvc_mem_fns.c @@ -124,164 +124,6 @@ void isvc_memset_2d(UWORD8 *pu1_dst, WORD32 i4_dst_stride, UWORD8 u1_val, WORD32 ******************************************************************************* * * @brief - * Function for copying to an interleaved destination - * - * @par Description: - * Copies the array of width 'wd' and height 'ht' from the location pointed - * by 'src' to the location pointed by 'dst' - * - * @param[in] pu1_src - * UWORD8 pointer to the source - * - * @param[out] pu1_dst - * UWORD8 pointer to the destination - * - * @param[in] src_strd - * integer source stride - * - * @param[in] dst_strd - * integer destination stride - * - * @param[in] ht - * integer height of the array - * - * @param[in] wd - * integer width of the array - * - * @returns - * - * @remarks - * The alternate elements of src will be copied to alternate locations in dsr - * Other locations are not touched - * - ******************************************************************************* - */ -void isvc_interleaved_copy(UWORD8 *pu1_src, UWORD8 *pu1_dst, WORD32 src_strd, WORD32 dst_strd, - WORD32 ht, WORD32 wd) -{ - WORD32 row, col; - wd *= 2; - - for(row = 0; row < ht; row++) - { - for(col = 0; col < wd; col += 2) - { - pu1_dst[col] = pu1_src[col]; - } - - pu1_src += src_strd; - pu1_dst += dst_strd; - } -} - -/** - ******************************************************************************* - * - * @brief - * Function for copying to an interleaved destination - * - * @par Description: - * Copies the array of width 'wd' and height 'ht' from the location pointed - * by 'src' to the location pointed by 'dst' - * - * @param[in] pu1_src - * UWORD8 pointer to the source - * - * @param[out] pu1_dst - * UWORD8 pointer to the destination - * - * @param[in] src_strd - * integer source stride - * - * @param[in] dst_strd - * integer destination stride - * - * @param[in] ht - * integer height of the array - * - * @param[in] wd - * integer width of the array - * - * @returns - * - * @remarks - * The alternate elements of src will be copied to alternate locations in dsr - * Other locations are not touched - * - ******************************************************************************* - */ -void isvc_16bit_interleaved_copy(WORD16 *pi2_src, WORD16 *pi2_dst, WORD32 src_strd, WORD32 dst_strd, - WORD32 ht, WORD32 wd) -{ - WORD32 row, col; - wd *= 2; - - for(row = 0; row < ht; row++) - { - for(col = 0; col < wd; col += 2) - { - pi2_dst[col] = pi2_src[col]; - } - - pi2_src += src_strd; - pi2_dst += dst_strd; - } -} - -/** - ******************************************************************************* - * - * @brief - * Function for memsetting to an interleaved destination - * - * @par Description: - * Memsets the array of width 'wd' and height 'ht' pointed by 'src' - * - * @param[in] pu1_src - * UWORD8 pointer to the source - * - * @param[in] src_strd - * integer source stride - * - * @param[in] value - * Value to set - * - * @param[in] ht - * integer height of the array - * - * @param[in] wd - * integer width of the array - * - * @returns - * - * @remarks - * The alternate elements of src will be copied to alternate locations in dsr - * Other locations are not touched - * - ******************************************************************************* - */ -void isvc_16bit_interleaved_memset(WORD16 *pi2_src, WORD32 i4_src_strd, WORD16 i2_value, - WORD32 i4_wd, WORD32 i4_ht) -{ - WORD32 row, col; - - i4_wd *= 2; - - for(row = 0; row < i4_ht; row++) - { - for(col = 0; col < i4_wd; col += 2) - { - pi2_src[col] = i2_value; - } - - pi2_src += i4_src_strd; - } -} - -/** - ******************************************************************************* - * - * @brief * Checks if any pixel in a block is non-zero * * @param[in] pu1_data diff --git a/common/svc/isvc_mem_fns.h b/common/svc/isvc_mem_fns.h index a4d95f7..581e478 100644 --- a/common/svc/isvc_mem_fns.h +++ b/common/svc/isvc_mem_fns.h @@ -64,32 +64,24 @@ typedef UWORD8 FT_NONZERO_CHECKER(UWORD8 *pu1_data, WORD32 i4_data_strd, UWORD32 UWORD32 u4_ht); /* C function declarations */ -extern FT_MEMCPY ih264_memcpy; extern FT_MEMCPY ih264_memcpy_mul_8; -extern FT_MEMSET ih264_memset; extern FT_MEMSET ih264_memset_mul_8; extern FT_MEMSET_16BIT ih264_memset_16bit; extern FT_MEMSET_16BIT ih264_memset_16bit_mul_8; extern FT_COPY_2D isvc_copy_2d; extern FT_MEMSET_2D isvc_memset_2d; -extern FT_16BIT_INTERLEAVED_COPY isvc_16bit_interleaved_copy; -extern FT_16BIT_INTERLEAVED_MEMSET isvc_16bit_interleaved_memset; extern FT_NONZERO_CHECKER isvc_is_nonzero_blk; extern FT_MEM_ALLOC isvc_memory_alloc; extern FT_MEM_FREE isvc_memory_free; /* A9 Q function declarations */ -extern FT_MEMCPY isvc_memcpy_a9q; extern FT_MEMCPY ih264_memcpy_mul_8_a9q; -extern FT_MEMSET ih264_memset_a9q; extern FT_MEMSET ih264_memset_mul_8_a9q; extern FT_MEMSET_16BIT ih264_memset_16bit_a9q; extern FT_MEMSET_16BIT ih264_memset_16bit_mul_8_a9q; /* AV8 function declarations */ -extern FT_MEMCPY ih264_memcpy_av8; extern FT_MEMCPY ih264_memcpy_mul_8_av8; -extern FT_MEMSET ih264_memset_av8; extern FT_MEMSET ih264_memset_mul_8_av8; extern FT_MEMSET_16BIT ih264_memset_16bit_av8; extern FT_MEMSET_16BIT ih264_memset_16bit_mul_8_av8; diff --git a/common/svc/isvc_trans_quant_itrans_iquant.h b/common/svc/isvc_trans_quant_itrans_iquant.h index fd15dcc..7ded811 100644 --- a/common/svc/isvc_trans_quant_itrans_iquant.h +++ b/common/svc/isvc_trans_quant_itrans_iquant.h @@ -195,15 +195,11 @@ extern FT_IQ_IT_RECON isvc_iquant_itrans_recon_4x4_sse42; extern FT_IQ_IT_RECON isvc_iquant_itrans_recon_res_4x4_sse42; extern FT_IQ_IT_RECON isvc_iquant_itrans_recon_res_4x4_with_res_acc_sse42; -extern FT_IQ_IT_RECON isvc_iquant_itrans_recon_chroma_4x4_sse42; extern FT_IQ_IT_RECON isvc_iquant_itrans_recon_res_chroma_4x4_sse42; extern FT_IQ_IT_RECON isvc_iquant_itrans_recon_res_chroma_4x4_with_res_acc_sse42; extern FT_IQ_IT_RECON isvc_iquant_itrans_recon_dc_4x4_sse42; -extern FT_IQ_IT_RECON isvc_iquant_itrans_recon_res_dc_4x4_sse42; -extern FT_IQ_IT_RECON isvc_iquant_itrans_recon_res_dc_with_res_acc_4x4_sse42; -extern FT_IQ_IT_RECON isvc_iquant_itrans_recon_chroma_4x4_dc_sse42; extern FT_IQ_IT_RECON isvc_iquant_itrans_recon_res_chroma_4x4_dc_sse42; extern FT_IQ_IT_RECON isvc_iquant_itrans_recon_res_chroma_4x4_dc_with_res_acc_sse42; @@ -217,15 +213,11 @@ extern FT_IQ_IT_RECON isvc_iquant_itrans_recon_4x4_neon; extern FT_IQ_IT_RECON isvc_iquant_itrans_recon_4x4_with_res_output_neon; extern FT_IQ_IT_RECON isvc_iquant_itrans_recon_4x4_with_res_accumulate_neon; -extern FT_IQ_IT_RECON isvc_iquant_itrans_recon_chroma_4x4_neon; extern FT_IQ_IT_RECON isvc_iquant_itrans_recon_chroma_4x4_with_res_output_neon; extern FT_IQ_IT_RECON isvc_iquant_itrans_recon_chroma_4x4_with_res_accumulate_neon; extern FT_IQ_IT_RECON isvc_iquant_itrans_recon_4x4_dc_neon; -extern FT_IQ_IT_RECON isvc_iquant_itrans_recon_4x4_dc_with_res_output_neon; -extern FT_IQ_IT_RECON isvc_iquant_itrans_recon_4x4_dc_with_res_accumulate_neon; -extern FT_IQ_IT_RECON isvc_iquant_itrans_recon_chroma_4x4_dc_neon; extern FT_IQ_IT_RECON isvc_iquant_itrans_recon_chroma_4x4_dc_with_res_output_neon; extern FT_IQ_IT_RECON isvc_iquant_itrans_recon_chroma_4x4_dc_with_res_accumulate_neon; diff --git a/common/x86/svc/isvc_iquant_itrans_recon_sse42.c b/common/x86/svc/isvc_iquant_itrans_recon_sse42.c index 829952b..f86ebc7 100644 --- a/common/x86/svc/isvc_iquant_itrans_recon_sse42.c +++ b/common/x86/svc/isvc_iquant_itrans_recon_sse42.c @@ -1034,356 +1034,6 @@ void isvc_iquant_itrans_recon_res_4x4_with_res_acc_sse42( *(pu4_out) = _mm_cvtsi128_si32(resq_r3); } -/* - ******************************************************************************** - * - * @brief This function reconstructs a 4x4 sub block from quantized chroma - *resiude and prediction buffer - * - * @par Description: - * The quantized residue is first inverse quantized, then inverse transformed. - * This inverse transformed content is added to the prediction buffer to recon- - * struct the end output - * - * @param[in] pi2_src - * quantized 4x4 block - * - * @param[in] pu1_pred - * prediction 4x4 block - * - * @param[out] pu1_out - * reconstructed 4x4 block - * - * @param[in] src_strd - * quantization buffer stride - * - * @param[in] i4_pred_stride, - * Prediction buffer stride - * - * @param[in] i4_out_stride - * recon buffer Stride - * - * @param[in] pu2_scaling_list - * pointer to scaling list - * - * @param[in] pu2_norm_adjust - * pointer to inverse scale matrix - * - * @param[in] u4_qp_div_6 - * Floor (qp/6) - * - * @param[in] pi4_tmp - * temporary buffer of size 1*16 - * - * @returns none - * - * @remarks none - * - ******************************************************************************* - */ -void isvc_iquant_itrans_recon_chroma_4x4_sse42( - buffer_container_t *ps_src, buffer_container_t *ps_pred, buffer_container_t *ps_res_pred, - buffer_container_t *ps_res, buffer_container_t *ps_rec, - iq_it_res_rec_constants_t *ps_iq_it_res_rec_constants, WORD16 *pi2_tmp, WORD16 *pi2_dc_src, - WORD32 i4_iq_start_idx, UWORD8 u1_res_accumulate) -{ - WORD16 *pi2_src = (WORD16 *) ps_src->pv_data; - UWORD8 *pu1_pred = (UWORD8 *) ps_pred->pv_data; - UWORD8 *pu1_out = (UWORD8 *) ps_rec->pv_data; - WORD32 i4_src_stride = ps_src->i4_data_stride; - WORD32 i4_pred_stride = ps_pred->i4_data_stride; - WORD32 i4_out_stride = ps_rec->i4_data_stride; - const UWORD16 *pu2_iscal_mat = ps_iq_it_res_rec_constants->pu2_iscal_mat; - const UWORD16 *pu2_weigh_mat = ps_iq_it_res_rec_constants->pu2_weigh_mat; - UWORD32 u4_qp_div_6 = ps_iq_it_res_rec_constants->u4_qp_div_6; - __m128i src_r0_r1, src_r2_r3; - __m128i src_r0, src_r1, src_r2, src_r3; - __m128i scalemat_r0_r1, scalemat_r2_r3; - __m128i pred_r0, pred_r1, pred_r2, pred_r3; - __m128i sign_reg, dequant_r0_r1, dequant_r2_r3; - /* all bits reset to zero */ - __m128i zero_8x16b = _mm_setzero_si128(); - __m128i neg_255_8x16b = _mm_set1_epi16(-((WORD16) UINT8_MAX)); - __m128i pos_255_8x16b = _mm_set1_epi16(((WORD16) UINT8_MAX)); - __m128i temp0, temp1, temp2, temp3, temp4, temp5, temp6, temp7; - __m128i resq_r0, resq_r1, resq_r2, resq_r3; - __m128i add_rshift = _mm_set1_epi32((u4_qp_div_6 < 4) ? (1 << (3 - u4_qp_div_6)) : 0); - __m128i value_32 = _mm_set1_epi32(32); - __m128i chroma_mask = _mm_set1_epi16(0xFF); - __m128i out_r0, out_r1, out_r2, out_r3; - - ASSERT(4 == i4_src_stride); - ASSERT(0 == u1_res_accumulate); - - UNUSED(i4_src_stride); - UNUSED(u1_res_accumulate); - UNUSED(ps_res); - UNUSED(ps_res_pred); - UNUSED(i4_iq_start_idx); - - /*************************************************************/ - /* Dequantization of coefficients. Will be replaced by SIMD */ - /* operations on platform */ - /*************************************************************/ - /* a00 a01 a02 a03 a10 a11 a12 a13 -- the source - matrix 0th,1st row */ - src_r0_r1 = _mm_loadu_si128((__m128i *) (pi2_src)); - - /* a20 a21 a22 a23 a30 a31 a32 a33 -- the - source matrix 2nd,3rd row */ - src_r2_r3 = _mm_loadu_si128((__m128i *) (pi2_src + 8)); - - /* b00 b01 b02 b03 b10 b11 b12 b13 -- the - scaling matrix 0th,1st row */ - scalemat_r0_r1 = _mm_loadu_si128((__m128i *) (pu2_iscal_mat)); - - /* b20 b21 b22 b23 b30 b31 b32 b33 --b12 b13 -- the - the scaling matrix 2nd,3rd row */ - scalemat_r2_r3 = _mm_loadu_si128((__m128i *) (pu2_iscal_mat + 8)); - - /* q00 q01 q02 q03 q10 q11 - q12 q13 -- all 16 bits */ - dequant_r0_r1 = _mm_loadu_si128((__m128i *) (pu2_weigh_mat)); - - /* q20 q21 q22 q23 q30 q31 - q32 q33 -- all 16 bits */ - dequant_r2_r3 = _mm_loadu_si128((__m128i *) (pu2_weigh_mat + 8)); - - temp0 = _mm_mullo_epi16(scalemat_r0_r1, - dequant_r0_r1); // b00*q00 b01*q01 b02*q02 b03*q03 b10*q10 b11*q11 - // b12*q12 b13*q13 -- 16 bit result - - temp1 = _mm_mullo_epi16(scalemat_r2_r3, dequant_r2_r3); - - /* b00*q00 0 b01*q01 0 b02*q02 0 b03*q03 0 -- 16 bit long */ - temp4 = _mm_unpacklo_epi16(temp0, zero_8x16b); - - /* b10*q10 0 b11*q11 0 b12*q12 0 b13*q13 0 -- 16 bit long */ - temp5 = _mm_unpackhi_epi16(temp0, zero_8x16b); - - /* b00*q00 0 b01*q01 0 b02*q02 0 b03*q03 0 -- 16 bit long */ - temp6 = _mm_unpacklo_epi16(temp1, zero_8x16b); - - /* b10*q10 0 b11*q11 0 b12*q12 0 b13*q13 0 -- 16 bit long */ - temp7 = _mm_unpackhi_epi16(temp1, zero_8x16b); - - /* a00 0 a01 0 a02 0 a03 0 -- 16 bit long */ - src_r0 = _mm_unpacklo_epi16(src_r0_r1, zero_8x16b); - /* a10 0 a11 0 a12 0 a13 0 -- 16 bit long */ - src_r1 = _mm_unpackhi_epi16(src_r0_r1, zero_8x16b); - /* a20 0 a21 0 a22 0 a23 0 -- 16 bit long */ - src_r2 = _mm_unpacklo_epi16(src_r2_r3, zero_8x16b); - /* a30 0 a31 0 a32 0 a33 0 -- 16 bit long */ - src_r3 = _mm_unpackhi_epi16(src_r2_r3, zero_8x16b); - - temp4 = _mm_madd_epi16(src_r0, temp4); - temp5 = _mm_madd_epi16(src_r1, temp5); - temp6 = _mm_madd_epi16(src_r2, temp6); - temp7 = _mm_madd_epi16(src_r3, temp7); - - if(u4_qp_div_6 >= 4) - { - resq_r0 = _mm_slli_epi32(temp4, u4_qp_div_6 - 4); - resq_r1 = _mm_slli_epi32(temp5, u4_qp_div_6 - 4); - resq_r2 = _mm_slli_epi32(temp6, u4_qp_div_6 - 4); - resq_r3 = _mm_slli_epi32(temp7, u4_qp_div_6 - 4); - } - else - { - temp4 = _mm_add_epi32(temp4, add_rshift); - temp5 = _mm_add_epi32(temp5, add_rshift); - temp6 = _mm_add_epi32(temp6, add_rshift); - temp7 = _mm_add_epi32(temp7, add_rshift); - resq_r0 = _mm_srai_epi32(temp4, 4 - u4_qp_div_6); - resq_r1 = _mm_srai_epi32(temp5, 4 - u4_qp_div_6); - resq_r2 = _mm_srai_epi32(temp6, 4 - u4_qp_div_6); - resq_r3 = _mm_srai_epi32(temp7, 4 - u4_qp_div_6); - } - - resq_r0 = _mm_insert_epi32(resq_r0, (WORD32) pi2_dc_src[0], 0); - /* Perform Inverse transform */ - /*-------------------------------------------------------------*/ - /* IDCT [ Horizontal transformation ] */ - /*-------------------------------------------------------------*/ - // Matrix transpose - /* - * a0 a1 a2 a3 - * b0 b1 b2 b3 - * c0 c1 c2 c3 - * d0 d1 d2 d3 - */ - /* a0 b0 a1 b1 */ - temp1 = _mm_unpacklo_epi32(resq_r0, resq_r1); - /* c0 d0 c1 d1 */ - temp3 = _mm_unpacklo_epi32(resq_r2, resq_r3); - /* a2 b2 a3 b3 */ - temp2 = _mm_unpackhi_epi32(resq_r0, resq_r1); - /* c2 d2 c3 d3 */ - temp4 = _mm_unpackhi_epi32(resq_r2, resq_r3); - /* a0 b0 c0 d0 */ - resq_r0 = _mm_unpacklo_epi64(temp1, temp3); - /* a1 b1 c1 d1 */ - resq_r1 = _mm_unpackhi_epi64(temp1, temp3); - /* a2 b2 c2 d2 */ - resq_r2 = _mm_unpacklo_epi64(temp2, temp4); - /* a3 b3 c3 d3 */ - resq_r3 = _mm_unpackhi_epi64(temp2, temp4); - /* Transform starts -- horizontal transform */ - - /*------------------------------------------------------------------*/ - /* z0 = w0 + w2 */ - temp0 = _mm_add_epi32(resq_r0, resq_r2); - /* z1 = w0 - w2 */ - temp1 = _mm_sub_epi32(resq_r0, resq_r2); - /* z2 = (w1 >> 1) - w3 */ - temp2 = _mm_srai_epi32(resq_r1, 1); - temp2 = _mm_sub_epi32(temp2, resq_r3); - /* z3 = w1 + (w3 >> 1) */ - temp3 = _mm_srai_epi32(resq_r3, 1); //(w3>>1) + w1 - temp3 = _mm_add_epi32(temp3, resq_r1); - /*----------------------------------------------------------*/ - /* x0 = z0 + z3 */ - resq_r0 = _mm_add_epi32(temp0, temp3); - /* x1 = z1 + z2 */ - resq_r1 = _mm_add_epi32(temp1, temp2); - /* x2 = z1 - z2 */ - resq_r2 = _mm_sub_epi32(temp1, temp2); - /* x3 = z0 - z3 */ - resq_r3 = _mm_sub_epi32(temp0, temp3); - // Matrix transpose - /* - * a0 b0 c0 d0 - * a1 b1 c1 d1 - * a2 b2 c2 d2 - * a3 b3 c3 d3 - */ - /* a0 a1 b0 b1 */ - temp1 = _mm_unpacklo_epi32(resq_r0, resq_r1); - /* a2 a3 b2 b3 */ - temp3 = _mm_unpacklo_epi32(resq_r2, resq_r3); - /* c0 c1 d0 d1 */ - temp2 = _mm_unpackhi_epi32(resq_r0, resq_r1); - /* c2 c3 d2 d3 */ - temp4 = _mm_unpackhi_epi32(resq_r2, resq_r3); - /* a0 a1 a2 a3 */ - resq_r0 = _mm_unpacklo_epi64(temp1, temp3); - /* b0 b1 b2 b3 */ - resq_r1 = _mm_unpackhi_epi64(temp1, temp3); - /* c0 c1 c2 c3 */ - resq_r2 = _mm_unpacklo_epi64(temp2, temp4); - /* d0 d1 d2 d3 */ - resq_r3 = _mm_unpackhi_epi64(temp2, temp4); - /* Transform ends -- horizontal transform */ - - temp0 = _mm_packs_epi32(resq_r0, resq_r1); - temp1 = _mm_packs_epi32(resq_r2, resq_r3); - - _mm_storeu_si128((__m128i *) (&pi2_tmp[0]), temp0); - _mm_storeu_si128((__m128i *) (&pi2_tmp[2 * 4]), temp1); - - /* Load pred buffer */ - pred_r0 = _mm_loadl_epi64((__m128i *) (&pu1_pred[0])); - pred_r1 = _mm_loadl_epi64((__m128i *) (&pu1_pred[i4_pred_stride])); - pred_r2 = _mm_loadl_epi64((__m128i *) (&pu1_pred[2 * i4_pred_stride])); - pred_r3 = _mm_loadl_epi64((__m128i *) (&pu1_pred[3 * i4_pred_stride])); - - pred_r0 = _mm_and_si128(pred_r0, chroma_mask); - pred_r1 = _mm_and_si128(pred_r1, chroma_mask); - pred_r2 = _mm_and_si128(pred_r2, chroma_mask); - pred_r3 = _mm_and_si128(pred_r3, chroma_mask); - - pred_r0 = _mm_unpacklo_epi64(pred_r0, pred_r1); - pred_r1 = _mm_unpacklo_epi64(pred_r2, pred_r3); - - /*--------------------------------------------------------------*/ - /* IDCT [ Vertical transformation] and Xij = (xij + 32)>>6 */ - /* */ - /* Add the prediction and store it back to same buffer */ - /*--------------------------------------------------------------*/ - /* z0j = y0j + y2j */ - temp0 = _mm_add_epi32(resq_r0, resq_r2); - /* z1j = y0j - y2j */ - temp1 = _mm_sub_epi32(resq_r0, resq_r2); - /* z2j = (y1j>>1) - y3j */ - temp2 = _mm_srai_epi32(resq_r1, 1); - temp2 = _mm_sub_epi32(temp2, resq_r3); - /* z3j = y1j + (y3j>>1) */ - temp3 = _mm_srai_epi32(resq_r3, 1); - temp3 = _mm_add_epi32(temp3, resq_r1); - - /* x0j = z0j + z3j */ - temp4 = _mm_add_epi32(temp0, temp3); - temp4 = _mm_add_epi32(temp4, value_32); - temp4 = _mm_srai_epi32(temp4, 6); - /* x1j = z1j + z2j */ - temp5 = _mm_add_epi32(temp1, temp2); - temp5 = _mm_add_epi32(temp5, value_32); - temp5 = _mm_srai_epi32(temp5, 6); - /* x2j = z1j - z2j */ - temp6 = _mm_sub_epi32(temp1, temp2); - temp6 = _mm_add_epi32(temp6, value_32); - temp6 = _mm_srai_epi32(temp6, 6); - /* x3j = z0j - z3j */ - temp7 = _mm_sub_epi32(temp0, temp3); - temp7 = _mm_add_epi32(temp7, value_32); - temp7 = _mm_srai_epi32(temp7, 6); - - /* 32-bit to 16-bit conversion */ - temp0 = _mm_packs_epi32(temp4, temp5); - temp1 = _mm_packs_epi32(temp6, temp7); - - /* Saturate all values < -255 to -255 and retain the rest as it is */ - temp4 = _mm_max_epi16(temp0, neg_255_8x16b); - /* Saturate all values > 255 to 255 and retain the rest as it is */ - temp4 = _mm_min_epi16(temp4, pos_255_8x16b); - - /* Saturate all values < -255 to -255 and retain the rest as it is */ - temp5 = _mm_max_epi16(temp1, neg_255_8x16b); - /* Saturate all values > 255 to 255 and retain the rest as it is */ - temp5 = _mm_min_epi16(temp5, pos_255_8x16b); - - temp0 = _mm_add_epi16(temp4, pred_r0); - temp1 = _mm_add_epi16(temp5, pred_r1); - - /*------------------------------------------------------------------*/ - /* Clipping the results to 8 bits */ - sign_reg = _mm_cmpgt_epi16(temp0, zero_8x16b); - temp0 = _mm_and_si128(temp0, sign_reg); - sign_reg = _mm_cmpgt_epi16(temp1, zero_8x16b); - temp1 = _mm_and_si128(temp1, sign_reg); - - resq_r0 = _mm_packus_epi16(temp0, temp1); - resq_r1 = _mm_srli_si128(resq_r0, 4); - resq_r2 = _mm_srli_si128(resq_r1, 4); - resq_r3 = _mm_srli_si128(resq_r2, 4); - - resq_r0 = _mm_cvtepu8_epi16(resq_r0); - resq_r1 = _mm_cvtepu8_epi16(resq_r1); - resq_r2 = _mm_cvtepu8_epi16(resq_r2); - resq_r3 = _mm_cvtepu8_epi16(resq_r3); - - chroma_mask = _mm_set1_epi16(0xFF00); - out_r0 = _mm_loadl_epi64((__m128i *) (&pu1_out[0])); - out_r1 = _mm_loadl_epi64((__m128i *) (&pu1_out[i4_out_stride])); - out_r2 = _mm_loadl_epi64((__m128i *) (&pu1_out[2 * i4_out_stride])); - out_r3 = _mm_loadl_epi64((__m128i *) (&pu1_out[3 * i4_out_stride])); - - out_r0 = _mm_and_si128(out_r0, chroma_mask); - out_r1 = _mm_and_si128(out_r1, chroma_mask); - out_r2 = _mm_and_si128(out_r2, chroma_mask); - out_r3 = _mm_and_si128(out_r3, chroma_mask); - - out_r0 = _mm_add_epi8(out_r0, resq_r0); - out_r1 = _mm_add_epi8(out_r1, resq_r1); - out_r2 = _mm_add_epi8(out_r2, resq_r2); - out_r3 = _mm_add_epi8(out_r3, resq_r3); - - _mm_storel_epi64((__m128i *) (&pu1_out[0]), out_r0); - _mm_storel_epi64((__m128i *) (&pu1_out[i4_out_stride]), out_r1); - _mm_storel_epi64((__m128i *) (&pu1_out[2 * i4_out_stride]), out_r2); - _mm_storel_epi64((__m128i *) (&pu1_out[3 * i4_out_stride]), out_r3); -} - void isvc_iquant_itrans_recon_res_chroma_4x4_sse42( buffer_container_t *ps_src, buffer_container_t *ps_pred, buffer_container_t *ps_res_pred, buffer_container_t *ps_res, buffer_container_t *ps_rec, @@ -2191,337 +1841,6 @@ void isvc_iquant_itrans_recon_dc_4x4_sse42(buffer_container_t *ps_src, buffer_co *(pu4_out) = _mm_cvtsi128_si32(temp7); } -void isvc_iquant_itrans_recon_res_dc_4x4_sse42( - buffer_container_t *ps_src, buffer_container_t *ps_pred, buffer_container_t *ps_res_pred, - buffer_container_t *ps_res, buffer_container_t *ps_rec, - iq_it_res_rec_constants_t *ps_iq_it_res_rec_constants, WORD16 *pi2_tmp, WORD16 *pi2_dc_src, - WORD32 i4_iq_start_idx, UWORD8 u1_res_accumulate) -{ - WORD16 *pi2_res = (WORD16 *) ps_res->pv_data; - WORD16 *pi2_res_ptr = pi2_res; - UWORD8 *pu1_pred = (UWORD8 *) ps_pred->pv_data; - UWORD8 *pu1_out = (UWORD8 *) ps_rec->pv_data; - WORD32 i4_res_stride = ps_res->i4_data_stride; - WORD32 i4_pred_stride = ps_pred->i4_data_stride; - WORD32 i4_out_stride = ps_rec->i4_data_stride; - const UWORD16 *pu2_iscal_mat = ps_iq_it_res_rec_constants->pu2_iscal_mat; - const UWORD16 *pu2_weigh_mat = ps_iq_it_res_rec_constants->pu2_weigh_mat; - UWORD32 u4_qp_div_6 = ps_iq_it_res_rec_constants->u4_qp_div_6; - UWORD32 *pu4_out = (UWORD32 *) pu1_out; - WORD32 q0 = ((WORD16 *) (ps_src->pv_data))[0]; - WORD16 i_macro, rnd_fact = (u4_qp_div_6 < 4) ? 1 << (3 - u4_qp_div_6) : 0; - - __m128i pred_r0, pred_r1, pred_r2, pred_r3; - __m128i sign_reg; - /* all bits reset to zero */ - __m128i zero_8x16b = _mm_setzero_si128(); - __m128i temp4, temp5, temp6, temp7; - __m128i value_add; - - ASSERT(0 == u1_res_accumulate); - - UNUSED(pi2_tmp); - UNUSED(ps_res_pred); - UNUSED(u1_res_accumulate); - - INV_QUANT(q0, pu2_iscal_mat[0], pu2_weigh_mat[0], u4_qp_div_6, rnd_fact, 4); - - /* Restoring dc value for intra case */ - if(i4_iq_start_idx != 0) q0 = pi2_dc_src[0]; - - i_macro = ((q0 + 32) >> 6); - - value_add = _mm_set1_epi16(isvc_get_residue(i_macro, 0, 0)); - - zero_8x16b = _mm_setzero_si128(); - - /* Load pred buffer */ - - /* p00 p01 p02 p03 0 0 0 0 -- all 8 bits */ - pred_r0 = _mm_loadl_epi64((__m128i *) (&pu1_pred[0])); - - /* p10 p11 p12 p13 0 0 0 0 -- all 8 bits */ - pred_r1 = _mm_loadl_epi64((__m128i *) (&pu1_pred[i4_pred_stride])); - - /* p20 p21 p22 p23 0 0 0 0 -- all 8 bits */ - pred_r2 = _mm_loadl_epi64((__m128i *) (&pu1_pred[2 * i4_pred_stride])); - - /* p30 p31 p32 p33 0 0 0 0 -- all 8 bits */ - pred_r3 = _mm_loadl_epi64((__m128i *) (&pu1_pred[3 * i4_pred_stride])); - - pred_r0 = _mm_cvtepu8_epi16(pred_r0); - pred_r1 = _mm_cvtepu8_epi16(pred_r1); - pred_r2 = _mm_cvtepu8_epi16(pred_r2); - pred_r3 = _mm_cvtepu8_epi16(pred_r3); - - pred_r0 = _mm_unpacklo_epi64(pred_r0, pred_r1); - pred_r2 = _mm_unpacklo_epi64(pred_r2, pred_r3); - - temp4 = _mm_add_epi16(value_add, pred_r0); - temp5 = _mm_add_epi16(value_add, pred_r2); - - _mm_storeu_si128((__m128i *) (&pi2_res_ptr[0]), value_add); - _mm_storeu_si128((__m128i *) (&pi2_res_ptr[i4_res_stride]), value_add); - _mm_storeu_si128((__m128i *) (&pi2_res_ptr[2 * i4_res_stride]), value_add); - _mm_storeu_si128((__m128i *) (&pi2_res_ptr[3 * i4_res_stride]), value_add); - /*------------------------------------------------------------------*/ - /* Clipping the results to 8 bits */ - sign_reg = _mm_cmpgt_epi16(temp4, zero_8x16b); - temp4 = _mm_and_si128(temp4, sign_reg); - sign_reg = _mm_cmpgt_epi16(temp5, zero_8x16b); - temp5 = _mm_and_si128(temp5, sign_reg); - - temp4 = _mm_packus_epi16(temp4, temp5); - temp5 = _mm_srli_si128(temp4, 4); - temp6 = _mm_srli_si128(temp5, 4); - temp7 = _mm_srli_si128(temp6, 4); - - *pu4_out = _mm_cvtsi128_si32(temp4); - pu1_out += i4_out_stride; - pu4_out = (UWORD32 *) (pu1_out); - *(pu4_out) = _mm_cvtsi128_si32(temp5); - pu1_out += i4_out_stride; - pu4_out = (UWORD32 *) (pu1_out); - *(pu4_out) = _mm_cvtsi128_si32(temp6); - pu1_out += i4_out_stride; - pu4_out = (UWORD32 *) (pu1_out); - *(pu4_out) = _mm_cvtsi128_si32(temp7); -} - -void isvc_iquant_itrans_recon_res_dc_with_res_acc_4x4_sse42( - buffer_container_t *ps_src, buffer_container_t *ps_pred, buffer_container_t *ps_res_pred, - buffer_container_t *ps_res, buffer_container_t *ps_rec, - iq_it_res_rec_constants_t *ps_iq_it_res_rec_constants, WORD16 *pi2_tmp, WORD16 *pi2_dc_src, - WORD32 i4_iq_start_idx, UWORD8 u1_res_accumulate) -{ - WORD16 *pi2_res = (WORD16 *) ps_res->pv_data; - WORD16 *pi2_res_ptr = pi2_res; - WORD16 *pi2_res_pred = (WORD16 *) ps_res_pred->pv_data; - WORD16 *pi2_res_pred_ptr = pi2_res_pred; - UWORD8 *pu1_pred = (UWORD8 *) ps_pred->pv_data; - UWORD8 *pu1_out = (UWORD8 *) ps_rec->pv_data; - WORD32 i4_res_stride = ps_res->i4_data_stride; - WORD32 i4_res_pred_stride = ps_res_pred->i4_data_stride; - WORD32 i4_pred_stride = ps_pred->i4_data_stride; - WORD32 i4_out_stride = ps_rec->i4_data_stride; - const UWORD16 *pu2_iscal_mat = ps_iq_it_res_rec_constants->pu2_iscal_mat; - const UWORD16 *pu2_weigh_mat = ps_iq_it_res_rec_constants->pu2_weigh_mat; - UWORD32 u4_qp_div_6 = ps_iq_it_res_rec_constants->u4_qp_div_6; - UWORD32 *pu4_out = (UWORD32 *) pu1_out; - WORD32 q0 = ((WORD16 *) (ps_src->pv_data))[0]; - WORD16 i_macro, rnd_fact = (u4_qp_div_6 < 4) ? 1 << (3 - u4_qp_div_6) : 0; - - __m128i pred_r0, pred_r1, pred_r2, pred_r3; - __m128i sign_reg; - /* all bits reset to zero */ - __m128i zero_8x16b = _mm_setzero_si128(); - __m128i temp4, temp5, temp6, temp7; - __m128i value_add; - __m128i res_pred_r0, res_pred_r1, res_pred_r2, res_pred_r3; - __m128i temp0, temp1; - __m128i neg_255_8x16b = _mm_set1_epi16(-((WORD16) UINT8_MAX)); - __m128i pos_255_8x16b = _mm_set1_epi16(((WORD16) UINT8_MAX)); - - ASSERT(1 == u1_res_accumulate); - - UNUSED(pi2_tmp); - UNUSED(u1_res_accumulate); - - INV_QUANT(q0, pu2_iscal_mat[0], pu2_weigh_mat[0], u4_qp_div_6, rnd_fact, 4); - - /* Restoring dc value for intra case */ - if(i4_iq_start_idx != 0) q0 = pi2_dc_src[0]; - - i_macro = ((q0 + 32) >> 6); - - value_add = _mm_set1_epi16(i_macro); - - zero_8x16b = _mm_setzero_si128(); - - /* Load pred buffer */ - - /* p00 p01 p02 p03 0 0 0 0 -- all 8 bits */ - pred_r0 = _mm_loadl_epi64((__m128i *) (&pu1_pred[0])); - - /* p10 p11 p12 p13 0 0 0 0 -- all 8 bits */ - pred_r1 = _mm_loadl_epi64((__m128i *) (&pu1_pred[i4_pred_stride])); - - /* p20 p21 p22 p23 0 0 0 0 -- all 8 bits */ - pred_r2 = _mm_loadl_epi64((__m128i *) (&pu1_pred[2 * i4_pred_stride])); - - /* p30 p31 p32 p33 0 0 0 0 -- all 8 bits */ - pred_r3 = _mm_loadl_epi64((__m128i *) (&pu1_pred[3 * i4_pred_stride])); - - pred_r0 = _mm_cvtepu8_epi16(pred_r0); - pred_r1 = _mm_cvtepu8_epi16(pred_r1); - pred_r2 = _mm_cvtepu8_epi16(pred_r2); - pred_r3 = _mm_cvtepu8_epi16(pred_r3); - - pred_r0 = _mm_unpacklo_epi64(pred_r0, pred_r1); - pred_r2 = _mm_unpacklo_epi64(pred_r2, pred_r3); - - /* Accumulating res */ - res_pred_r0 = _mm_loadl_epi64((__m128i *) &pi2_res_pred_ptr[0]); - res_pred_r1 = _mm_loadl_epi64((__m128i *) &pi2_res_pred_ptr[i4_res_pred_stride]); - res_pred_r2 = _mm_loadl_epi64((__m128i *) &pi2_res_pred_ptr[2 * i4_res_pred_stride]); - res_pred_r3 = _mm_loadl_epi64((__m128i *) &pi2_res_pred_ptr[3 * i4_res_pred_stride]); - - res_pred_r0 = _mm_unpacklo_epi64(res_pred_r0, res_pred_r1); - res_pred_r1 = _mm_unpacklo_epi64(res_pred_r2, res_pred_r3); - - temp0 = _mm_add_epi16(value_add, res_pred_r0); - temp1 = _mm_add_epi16(value_add, res_pred_r1); - - /* Saturate all values < -255 to -255 and retain the rest as it is */ - temp0 = _mm_max_epi16(temp0, neg_255_8x16b); - /* Saturate all values > 255 to 255 and retain the rest as it is */ - temp0 = _mm_min_epi16(temp0, pos_255_8x16b); - - /* Saturate all values < -255 to -255 and retain the rest as it is */ - temp1 = _mm_max_epi16(temp1, neg_255_8x16b); - /* Saturate all values > 255 to 255 and retain the rest as it is */ - temp1 = _mm_min_epi16(temp1, pos_255_8x16b); - - _mm_storeu_si128((__m128i *) (&pi2_res_ptr[0]), temp0); - _mm_storeu_si128((__m128i *) (&pi2_res_ptr[2 * i4_res_stride]), temp1); - - temp4 = _mm_add_epi16(temp0, pred_r0); - temp5 = _mm_add_epi16(temp1, pred_r2); - - temp0 = _mm_srli_si128(temp0, 8); - temp1 = _mm_srli_si128(temp1, 8); - - _mm_storeu_si128((__m128i *) (&pi2_res_ptr[i4_res_stride]), temp0); - _mm_storeu_si128((__m128i *) (&pi2_res_ptr[3 * i4_res_stride]), temp1); - - /*------------------------------------------------------------------*/ - /* Clipping the results to 8 bits */ - sign_reg = _mm_cmpgt_epi16(temp4, zero_8x16b); - temp4 = _mm_and_si128(temp4, sign_reg); - sign_reg = _mm_cmpgt_epi16(temp5, zero_8x16b); - temp5 = _mm_and_si128(temp5, sign_reg); - - temp4 = _mm_packus_epi16(temp4, temp5); - temp5 = _mm_srli_si128(temp4, 4); - temp6 = _mm_srli_si128(temp5, 4); - temp7 = _mm_srli_si128(temp6, 4); - - *pu4_out = _mm_cvtsi128_si32(temp4); - pu1_out += i4_out_stride; - pu4_out = (UWORD32 *) (pu1_out); - *(pu4_out) = _mm_cvtsi128_si32(temp5); - pu1_out += i4_out_stride; - pu4_out = (UWORD32 *) (pu1_out); - *(pu4_out) = _mm_cvtsi128_si32(temp6); - pu1_out += i4_out_stride; - pu4_out = (UWORD32 *) (pu1_out); - *(pu4_out) = _mm_cvtsi128_si32(temp7); -} - -void isvc_iquant_itrans_recon_chroma_4x4_dc_sse42( - buffer_container_t *ps_src, buffer_container_t *ps_pred, buffer_container_t *ps_res_pred, - buffer_container_t *ps_res, buffer_container_t *ps_rec, - iq_it_res_rec_constants_t *ps_iq_it_res_rec_constants, WORD16 *pi2_tmp, WORD16 *pi2_dc_src, - WORD32 i4_iq_start_idx, UWORD8 u1_res_accumulate) -{ - WORD16 *pi2_src = (WORD16 *) ps_src->pv_data; - UWORD8 *pu1_pred = (UWORD8 *) ps_pred->pv_data; - UWORD8 *pu1_out = (UWORD8 *) ps_rec->pv_data; - WORD32 i4_pred_stride = ps_pred->i4_data_stride; - WORD32 i4_out_stride = ps_rec->i4_data_stride; - const UWORD16 *pu2_iscal_mat = ps_iq_it_res_rec_constants->pu2_iscal_mat; - const UWORD16 *pu2_weigh_mat = ps_iq_it_res_rec_constants->pu2_weigh_mat; - UWORD32 u4_qp_div_6 = ps_iq_it_res_rec_constants->u4_qp_div_6; - /* DC value won't be dequantized for chroma - inverse transform */ - WORD16 q0 = pi2_dc_src[0]; - WORD16 i_macro = ((q0 + 32) >> 6); - - __m128i pred_r0, pred_r1, pred_r2, pred_r3; - /* all bits reset to zero */ - __m128i zero_8x16b = _mm_setzero_si128(); - __m128i chroma_mask = _mm_set1_epi16(0xFF); - __m128i value_add = _mm_set1_epi16(i_macro); - __m128i out_r0, out_r1, out_r2, out_r3; - - ASSERT(0 == u1_res_accumulate); - - UNUSED(pi2_src); - UNUSED(pu2_iscal_mat); - UNUSED(pu2_weigh_mat); - UNUSED(u4_qp_div_6); - UNUSED(pi2_tmp); - UNUSED(ps_res_pred); - UNUSED(ps_res); - UNUSED(i4_iq_start_idx); - UNUSED(u1_res_accumulate); - - /* Load pred buffer */ - pred_r0 = _mm_loadl_epi64((__m128i *) (&pu1_pred[0])); - - pred_r1 = _mm_loadl_epi64((__m128i *) (&pu1_pred[i4_pred_stride])); - - pred_r2 = _mm_loadl_epi64((__m128i *) (&pu1_pred[2 * i4_pred_stride])); - - pred_r3 = _mm_loadl_epi64((__m128i *) (&pu1_pred[3 * i4_pred_stride])); - - /* Mask alternate pred values from the interleaved pred buf */ - pred_r0 = _mm_and_si128(pred_r0, chroma_mask); - pred_r1 = _mm_and_si128(pred_r1, chroma_mask); - pred_r2 = _mm_and_si128(pred_r2, chroma_mask); - pred_r3 = _mm_and_si128(pred_r3, chroma_mask); - - /* Pack the first four 16 bit values of 2 regs into a single reg*/ - pred_r0 = _mm_unpacklo_epi64(pred_r0, pred_r1); - pred_r2 = _mm_unpacklo_epi64(pred_r2, pred_r3); - - /* Compute out pixel by adding res to pred */ - pred_r0 = _mm_add_epi16(value_add, pred_r0); - pred_r2 = _mm_add_epi16(value_add, pred_r2); - /*------------------------------------------------------------------*/ - /* Clipping the results to 8 bits */ - pred_r0 = _mm_packus_epi16(pred_r0, pred_r2); - pred_r1 = _mm_srli_si128(pred_r0, 4); - pred_r2 = _mm_srli_si128(pred_r1, 4); - pred_r3 = _mm_srli_si128(pred_r2, 4); - - /* p00 p01 p02 p03 -- all 16 bits */ - pred_r0 = _mm_unpacklo_epi8(pred_r0, zero_8x16b); - /* p10 p11 p12 p13 -- all 16 bits */ - pred_r1 = _mm_unpacklo_epi8(pred_r1, zero_8x16b); - /* p20 p21 p22 p23 -- all 16 bits */ - pred_r2 = _mm_unpacklo_epi8(pred_r2, zero_8x16b); - /* p30 p31 p32 p33 -- all 16 bits */ - pred_r3 = _mm_unpacklo_epi8(pred_r3, zero_8x16b); - - /* Load interleaved out buffer */ - out_r0 = _mm_loadl_epi64((__m128i *) (&pu1_out[0])); - out_r1 = _mm_loadl_epi64((__m128i *) (&pu1_out[i4_out_stride])); - out_r2 = _mm_loadl_epi64((__m128i *) (&pu1_out[2 * i4_out_stride])); - out_r3 = _mm_loadl_epi64((__m128i *) (&pu1_out[3 * i4_out_stride])); - - /* Mask the interleaved out buf in order to save the U/V out pixel computed in - this function call without thrashing the U/V out pixel that was saved - during an earlier function call */ - chroma_mask = _mm_set1_epi16(0xFF00); - - out_r0 = _mm_and_si128(out_r0, chroma_mask); - out_r1 = _mm_and_si128(out_r1, chroma_mask); - out_r2 = _mm_and_si128(out_r2, chroma_mask); - out_r3 = _mm_and_si128(out_r3, chroma_mask); - - /* Save the out pixels in alternate locations */ - out_r0 = _mm_add_epi8(out_r0, pred_r0); - out_r1 = _mm_add_epi8(out_r1, pred_r1); - out_r2 = _mm_add_epi8(out_r2, pred_r2); - out_r3 = _mm_add_epi8(out_r3, pred_r3); - - _mm_storel_epi64((__m128i *) (&pu1_out[0]), out_r0); - _mm_storel_epi64((__m128i *) (&pu1_out[i4_out_stride]), out_r1); - _mm_storel_epi64((__m128i *) (&pu1_out[2 * i4_out_stride]), out_r2); - _mm_storel_epi64((__m128i *) (&pu1_out[3 * i4_out_stride]), out_r3); -} - void isvc_iquant_itrans_recon_res_chroma_4x4_dc_sse42( buffer_container_t *ps_src, buffer_container_t *ps_pred, buffer_container_t *ps_res_pred, buffer_container_t *ps_res, buffer_container_t *ps_rec, diff --git a/decoder/ih264d_parse_headers.c b/decoder/ih264d_parse_headers.c index b25af69..a92ad8a 100644 --- a/decoder/ih264d_parse_headers.c +++ b/decoder/ih264d_parse_headers.c @@ -1095,6 +1095,20 @@ WORD32 ih264d_parse_sps(dec_struct_t *ps_dec, dec_bit_stream_t *ps_bitstrm) ret = ih264d_parse_vui_parametres(&ps_seq->s_vui, ps_bitstrm); if(ret != OK) return ret; + + if (ps_dec->pu1_bits_buf_dynamic != NULL) { + vui_t *ps_vui = &ps_seq->s_vui; + dec_seq_params_t *ps_sps_old = ps_dec->ps_sps; + vui_t *ps_vui_old = &ps_sps_old->s_vui; + + if (ps_vui->u1_video_full_range_flag != ps_vui_old->u1_video_full_range_flag || + ps_vui->u1_colour_primaries != ps_vui_old->u1_colour_primaries || + ps_vui->u1_tfr_chars != ps_vui_old->u1_tfr_chars || + ps_vui->u1_matrix_coeffs != ps_vui_old->u1_matrix_coeffs) { + ps_dec->u1_res_changed = 1; + return IVD_RES_CHANGED; + } + } } /* Compare older num_reorder_frames with the new one if header is already decoded */ diff --git a/decoder/ih264d_sei.c b/decoder/ih264d_sei.c index 079f036..e7d436c 100644 --- a/decoder/ih264d_sei.c +++ b/decoder/ih264d_sei.c @@ -853,6 +853,14 @@ WORD32 ih264d_parse_fgc(dec_bit_stream_t *ps_bitstrm, dec_struct_t *ps_dec, for(i = 0; i <= ps_sei->s_sei_fgc_params.au1_num_intensity_intervals_minus1[c]; i++) { + /* Although the fag end of both the NALU and the bitstream buffer */ + /* is being parsed, not all FGC SEI symbols would have been */ + /* decoded semantically. The code below detects this condition */ + if((ps_bitstrm->u4_ofst + 8 + 8) >= ps_bitstrm->u4_max_ofst) + { + return ERROR_INV_SEI_FGC_PARAMS; + } + ps_sei->s_sei_fgc_params.au1_intensity_interval_lower_bound[c][i] = (UWORD8) ih264d_get_bits_h264(ps_bitstrm, 8); diff --git a/decoder/ih264d_utils.c b/decoder/ih264d_utils.c index 9d616d3..ffb575a 100644 --- a/decoder/ih264d_utils.c +++ b/decoder/ih264d_utils.c @@ -1599,9 +1599,10 @@ WORD32 ih264d_decode_gaps_in_frame_num(dec_struct_t *ps_dec, if(ps_dec->u2_prev_ref_frame_num == u2_frame_num) return 0; } + ps_pic_params = ps_dec->ps_cur_pps; u4_next_frm_num = ps_dec->u2_prev_ref_frame_num + 1; - u4_max_frm_num = ps_dec->ps_cur_sps->u2_u4_max_pic_num_minus1 + 1; + u4_max_frm_num = ps_pic_params->ps_sps->u2_u4_max_pic_num_minus1 + 1; // check if(u4_next_frm_num >= u4_max_frm_num) @@ -1630,7 +1631,6 @@ WORD32 ih264d_decode_gaps_in_frame_num(dec_struct_t *ps_dec, s_tmp_poc.i4_delta_pic_order_cnt[1] = 0; ps_cur_slice = ps_dec->ps_cur_slice; - ps_pic_params = ps_dec->ps_cur_pps; i4_frame_gaps = 0; ps_dpb_mgr = ps_dec->ps_dpb_mgr; diff --git a/decoder/mvc/imvcd_api.c b/decoder/mvc/imvcd_api.c index 45fffd1..1026fc8 100644 --- a/decoder/mvc/imvcd_api.c +++ b/decoder/mvc/imvcd_api.c @@ -731,17 +731,27 @@ static IV_API_CALL_STATUS_T imvcd_view_decode(iv_obj_t *ps_dec_hdl, imvcd_video_ return IV_FAIL; } - /* Ignore bytes beyond the allocated size of intermediate buffer */ - /* Since 8 bytes are read ahead, ensure 8 bytes are free at the - end of the buffer, which will be memset to 0 after emulation prevention */ - i4_nalu_length = MIN((UWORD32) i4_nalu_length, u4_bitstream_buf_size - 8); - if(i4_nalu_length) { + UWORD32 u4_nalu_buf_size = ((UWORD32) i4_nalu_length) + 8; + + if(u4_nalu_buf_size > u4_bitstream_buf_size) + { + /* 64 extra bytes to account for OOB accesses during SEI parsing in */ + /* some fuzzer bitstreams */ + if(IV_SUCCESS != imvcd_bitstream_buf_realloc(ps_view_ctxt, u4_nalu_buf_size + 64)) + { + return IV_FAIL; + } + + pu1_bitstream_buf = ps_view_ctxt->pu1_bits_buf_dynamic; + u4_bitstream_buf_size = ps_view_ctxt->u4_dynamic_bits_buf_size; + } + memcpy(pu1_bitstream_buf, pu1_input_buffer + u4_length_of_start_code, i4_nalu_length); /* Decoder may read extra 8 bytes near end of the frame */ - if(((UWORD32) (i4_nalu_length + 8)) < u4_bitstream_buf_size) + if(u4_nalu_buf_size < u4_bitstream_buf_size) { memset(pu1_bitstream_buf + i4_nalu_length, 0, 8 * sizeof(pu1_bitstream_buf[0])); } diff --git a/decoder/mvc/imvcd_api_utils.c b/decoder/mvc/imvcd_api_utils.c index 7d1f309..3a40d3b 100644 --- a/decoder/mvc/imvcd_api_utils.c +++ b/decoder/mvc/imvcd_api_utils.c @@ -391,3 +391,19 @@ void imvcd_bitsteam_buf_free(dec_struct_t *ps_view_ctxt) { PS_DEC_ALIGNED_FREE(ps_view_ctxt, ps_view_ctxt->pu1_bits_buf_dynamic); } + +IV_API_CALL_STATUS_T imvcd_bitstream_buf_realloc(dec_struct_t *ps_view_ctxt, UWORD32 u4_size) +{ + imvcd_bitsteam_buf_free(ps_view_ctxt); + + u4_size = MAX(MIN_BITSTREAMS_BUF_SIZE, u4_size); + + ps_view_ctxt->pu1_bits_buf_dynamic = + ps_view_ctxt->pf_aligned_alloc(ps_view_ctxt->pv_mem_ctxt, 128, u4_size); + RETURN_IF((NULL == ps_view_ctxt->pu1_bits_buf_dynamic), IV_FAIL); + + memset(ps_view_ctxt->pu1_bits_buf_dynamic, 0, u4_size); + ps_view_ctxt->u4_dynamic_bits_buf_size = u4_size; + + return IV_SUCCESS; +} diff --git a/decoder/mvc/imvcd_api_utils.h b/decoder/mvc/imvcd_api_utils.h index 0707f45..8b025b2 100644 --- a/decoder/mvc/imvcd_api_utils.h +++ b/decoder/mvc/imvcd_api_utils.h @@ -53,6 +53,9 @@ extern IV_API_CALL_STATUS_T imvcd_bitstream_buf_alloc(dec_struct_t *ps_view_ctxt extern void imvcd_bitsteam_buf_free(dec_struct_t *ps_view_ctxt); +extern IV_API_CALL_STATUS_T imvcd_bitstream_buf_realloc(dec_struct_t *ps_view_ctxt, + UWORD32 u4_size); + extern void imvcd_convert_to_app_disp_buf(mvc_dec_ctxt_t *ps_mvcd_ctxt, iv_yuv_buf_t *ps_view_disp_bufs); #endif diff --git a/decoder/mvc/imvcd_defs.h b/decoder/mvc/imvcd_defs.h index 0b8c976..2bf5e54 100644 --- a/decoder/mvc/imvcd_defs.h +++ b/decoder/mvc/imvcd_defs.h @@ -34,14 +34,6 @@ is still greater than any possible value of u1_pic_buf_id */ #define IVP_PIC_BUF_ID UINT8_MAX -/* In FGC SEI - - Worst-case bits for all elements before 'num_intensity_intervals_minus1' = 47 - - Worst-case bits for all elements before 'film_grain_characteristics_repetition_period', not - including elements from previous line = 3 * (8 + 3 + 256 * (8 + 8 + 8 * 16)) = 110625 - - Worst-case bits for 'film_grain_characteristics_repetition_period' = 30 - Total of (47 + 110625 + 30) = 110702 byte */ -#define MAX_FGC_SEI_SIZE 110702 - -#define MIN_BITSTREAMS_BUF_SIZE (MAX_FGC_SEI_SIZE + 256000) +#define MIN_BITSTREAMS_BUF_SIZE 256000 #endif diff --git a/decoder/svc/isvcd_api.c b/decoder/svc/isvcd_api.c index c449fba..c789785 100644 --- a/decoder/svc/isvcd_api.c +++ b/decoder/svc/isvcd_api.c @@ -2619,7 +2619,7 @@ WORD32 isvcd_allocate_static_bufs(iv_obj_t **dec_hdl, void *pv_api_ip, void *pv_ ps_svc_lyr_dec->u1_dyadic_flag = 1; ps_svc_lyr_dec->u1_restricted_res_change_flag = 1; ps_svc_lyr_dec->u1_base_res_flag = 1; - ps_svc_lyr_dec->u1_ref_layer_id = 0; + ps_svc_lyr_dec->u1_ref_layer_id = u1_layer_id - 1; ps_svc_lyr_dec->ps_dec_svc_ref_layer = &ps_svcd_ctxt->ps_svc_dec_lyr[ps_svc_lyr_dec->u1_ref_layer_id]; ps_svc_lyr_dec->u4_pps_id_for_layer = UINT32_MAX; @@ -4435,6 +4435,15 @@ WORD32 isvcd_dec_non_vcl(void *pv_out_non_vcl, void *pv_seq_params, void *pv_pic { ps_dec->i4_header_decoded |= 0x1; ps_svcd_ctxt->u4_num_sps_ctr++; + + if(ps_svcd_ctxt->pic_width < ps_svc_lyr_dec->pic_width) + { + ps_svcd_ctxt->pic_width = ps_svc_lyr_dec->pic_width; + } + if(ps_svcd_ctxt->pic_height < ps_svc_lyr_dec->pic_height) + { + ps_svcd_ctxt->pic_height = ps_svc_lyr_dec->pic_height; + } } if(i_status) return i_status; @@ -4448,6 +4457,15 @@ WORD32 isvcd_dec_non_vcl(void *pv_out_non_vcl, void *pv_seq_params, void *pv_pic { ps_svcd_ctxt->u4_num_sps_ctr++; ps_dec->i4_header_decoded |= 0x1; + + if(ps_svcd_ctxt->pic_width < ps_svc_lyr_dec->pic_width) + { + ps_svcd_ctxt->pic_width = ps_svc_lyr_dec->pic_width; + } + if(ps_svcd_ctxt->pic_height < ps_svc_lyr_dec->pic_height) + { + ps_svcd_ctxt->pic_height = ps_svc_lyr_dec->pic_height; + } } if(i_status) return i_status; @@ -5183,6 +5201,7 @@ WORD32 isvcd_video_decode(iv_obj_t *dec_hdl, void *pv_api_ip, void *pv_api_op) UWORD8 u1_layer_nal_data_present = 0; ps_svcd_ctxt->u1_cur_layer_id = u1_res_id; ps_svc_lyr_dec = ps_svcd_ctxt->ps_svc_dec_lyr + u1_res_id; + ps_svc_lyr_dec->u1_res_init_done = 0; ps_dec = &ps_svc_lyr_dec->s_dec; ps_dec->i4_decode_header = ps_dec_zero_lyr->i4_decode_header; @@ -5319,7 +5338,7 @@ WORD32 isvcd_video_decode(iv_obj_t *dec_hdl, void *pv_api_ip, void *pv_api_op) ps_dec = &ps_svc_lyr_dec->s_dec; if((0 == ps_svcd_ctxt->u4_num_sps_ctr) || (0 == ps_svcd_ctxt->u4_num_pps_ctr) || - (NULL == ps_dec->ps_cur_pps)) + (NULL == ps_dec->ps_cur_pps) || (ps_svc_lyr_dec->u1_res_init_done == 0)) { ps_svcd_ctxt->u1_exit_till_next_IDR = 1; ps_dec_op->u4_error_code = ERROR_UNKNOWN_NAL; @@ -6085,16 +6104,8 @@ WORD32 isvcd_get_buf_info(iv_obj_t *dec_hdl, void *pv_api_ip, void *pv_api_op) if(ps_dec->i4_header_decoded == 3) { - if(0 == ps_dec->u4_share_disp_buf) - { - pic_wd = ps_dec->u2_disp_width; - pic_ht = ps_dec->u2_disp_height; - } - else - { - pic_wd = ps_dec->u2_frm_wd_y; - pic_ht = ps_dec->u2_frm_ht_y; - } + pic_wd = ps_svcd_ctxt->pic_width; + pic_ht = ps_svcd_ctxt->pic_height; } for(i = 0; i < ps_ctl_op->u4_min_num_in_bufs; i++) diff --git a/decoder/svc/isvcd_intra_resamp.c b/decoder/svc/isvcd_intra_resamp.c index 34e3779..326dd5c 100644 --- a/decoder/svc/isvcd_intra_resamp.c +++ b/decoder/svc/isvcd_intra_resamp.c @@ -4959,6 +4959,45 @@ void isvcd_crop_wnd_flag_res_int(void *pv_svc_dec) /* 26 08 2021 vijayakumar creation */ /* */ /*****************************************************************************/ +void isvcd_intra_resamp_res_init_update_flags(void *pv_svc_dec) +{ + intra_sampling_ctxt_t *ps_ctxt; + intra_samp_lyr_ctxt *ps_lyr_ctxt; + svc_dec_lyr_struct_t *ps_svc_lyr_dec = (svc_dec_lyr_struct_t *) pv_svc_dec; + + ps_ctxt = (intra_sampling_ctxt_t *) ps_svc_lyr_dec->pv_intra_sample_ctxt;; + /* get the current layer ctxt */ + ps_lyr_ctxt = &ps_ctxt->as_res_lyrs[ps_svc_lyr_dec->u1_layer_id - 1]; + + ps_lyr_ctxt->i1_constrained_intra_rsmpl_flag = + ps_svc_lyr_dec->s_svc_slice_params.u1_constrained_intra_resampling_flag; +} + +/*****************************************************************************/ +/* */ +/* Function Name : isvcd_intra_resamp_res_init */ +/* */ +/* Description : this function calculates the scale factors and initialise*/ +/* the context structure */ +/* */ +/* Inputs : pv_intra_samp_ctxt: handle to private structure */ +/* ps_curr_lyr_res_prms: pointer to current resolution */ +/* params */ +/* ps_ref_lyr_res_prms : pointer to ref resolution params */ +/* Globals : none */ +/* Processing : it stores the layer dimensions */ +/* */ +/* Outputs : none */ +/* Returns : none */ +/* */ +/* Issues : none */ +/* */ +/* Revision History: */ +/* */ +/* DD MM YYYY Author(s) Changes (Describe the changes made) */ +/* 26 08 2021 vijayakumar creation */ +/* */ +/*****************************************************************************/ WORD32 isvcd_intra_resamp_res_init(void *pv_svc_dec) { intra_sampling_ctxt_t *ps_ctxt; diff --git a/decoder/svc/isvcd_intra_resamp.h b/decoder/svc/isvcd_intra_resamp.h index 9b15a6d..7f09117 100644 --- a/decoder/svc/isvcd_intra_resamp.h +++ b/decoder/svc/isvcd_intra_resamp.h @@ -628,4 +628,6 @@ void isvcd_crop_wnd_flag_res_int(void *ps_svc_dec); WORD32 isvcd_intra_resamp_res_init(void *ps_svc_dec); +void isvcd_intra_resamp_res_init_update_flags(void *ps_svc_dec); + #endif /* _ISVCD_INTRA_RESAMPLE_H_ */ diff --git a/decoder/svc/isvcd_parse_ebslice.c b/decoder/svc/isvcd_parse_ebslice.c index d805764..b1807ea 100644 --- a/decoder/svc/isvcd_parse_ebslice.c +++ b/decoder/svc/isvcd_parse_ebslice.c @@ -622,6 +622,7 @@ WORD32 isvcd_mv_pred_ref_tfr_nby2_ebmb(dec_struct_t *ps_dec, UWORD8 u1_mb_idx, U } } } + return OK; } @@ -1790,12 +1791,6 @@ WORD32 isvcd_parse_ebslice(svc_dec_lyr_struct_t *ps_svc_lyr_dec, UWORD16 u2_firs return ERROR_INV_SLICE_HDR_T; } - ret = isvcd_parse_interlayer_resamp_func_init(ps_svc_lyr_dec, u2_first_mb_in_slice); - if(ret != OK) - { - return ERROR_CORRUPTED_SLICE; - } - ps_dec->u1_slice_header_done = 2; if(!ps_svc_slice_params->u1_slice_skip_flag) { @@ -2130,11 +2125,6 @@ WORD32 isvcd_parse_bslice(svc_dec_lyr_struct_t *ps_svc_lyr_dec, UWORD16 u2_first ps_slice->i1_slice_beta_offset = 0; } - ret = isvcd_parse_interlayer_resamp_func_init(ps_svc_lyr_dec, u2_first_mb_in_slice); - if(ret != OK) - { - return ERROR_CORRUPTED_SLICE; - } ps_dec->u1_slice_header_done = 2; if(ps_pps->u1_entropy_coding_mode) diff --git a/decoder/svc/isvcd_parse_eislice.c b/decoder/svc/isvcd_parse_eislice.c index 4b3728c..c412fc5 100644 --- a/decoder/svc/isvcd_parse_eislice.c +++ b/decoder/svc/isvcd_parse_eislice.c @@ -1127,11 +1127,6 @@ WORD32 isvcd_parse_eislice(svc_dec_lyr_struct_t *ps_svc_lyr_dec, UWORD16 u2_firs ps_dec->u2_mv_2mb[0] = 0; ps_dec->u2_mv_2mb[1] = 0; - ret = isvcd_parse_interlayer_resamp_func_init(ps_svc_lyr_dec, u2_first_mb_in_slice); - if(ret != OK) - { - return ERROR_CORRUPTED_SLICE; - } /*set slice header cone to 2 ,to indicate correct header*/ ps_dec->u1_slice_header_done = 2; @@ -2064,11 +2059,6 @@ WORD32 isvcd_parse_islice(svc_dec_lyr_struct_t *ps_svc_lyr_dec, UWORD16 u2_first ps_dec->u2_mv_2mb[0] = 0; ps_dec->u2_mv_2mb[1] = 0; - ret = isvcd_parse_interlayer_resamp_func_init(ps_svc_lyr_dec, u2_first_mb_in_slice); - if(ret != OK) - { - return ERROR_CORRUPTED_SLICE; - } /*set slice header cone to 2 ,to indicate correct header*/ ps_dec->u1_slice_header_done = 2; if(ps_pps->u1_entropy_coding_mode) diff --git a/decoder/svc/isvcd_parse_epslice.c b/decoder/svc/isvcd_parse_epslice.c index bfd32bb..36621ba 100644 --- a/decoder/svc/isvcd_parse_epslice.c +++ b/decoder/svc/isvcd_parse_epslice.c @@ -425,11 +425,6 @@ WORD32 isvcd_parse_epslice(svc_dec_lyr_struct_t *ps_svc_lyr_dec, UWORD16 u2_firs return ERROR_INV_SLICE_HDR_T; } - ret = isvcd_parse_interlayer_resamp_func_init(ps_svc_lyr_dec, u2_first_mb_in_slice); - if(ret != OK) - { - return ERROR_CORRUPTED_SLICE; - } ps_dec->u1_slice_header_done = 2; if(!ps_svc_slice_params->u1_slice_skip_flag) @@ -3188,6 +3183,8 @@ WORD32 isvcd_parse_interlayer_resamp_func_init(svc_dec_lyr_struct_t *ps_svc_lyr_ dec_struct_t *ps_dec = &ps_svc_lyr_dec->s_dec; dec_slice_params_t *ps_slice = ps_dec->ps_cur_slice; WORD32 ret = OK; + if(ps_svc_lyr_dec->u1_res_init_done == 1) + return ret; if(TARGET_LAYER != ps_svc_lyr_dec->u1_layer_identifier) { @@ -3209,6 +3206,8 @@ WORD32 isvcd_parse_interlayer_resamp_func_init(svc_dec_lyr_struct_t *ps_svc_lyr_ if(ret != OK) return NOT_OK; ret = isvcd_residual_samp_res_init(ps_svc_lyr_dec); if(ret != OK) return NOT_OK; + + ps_svc_lyr_dec->u1_res_init_done = 1; } return ret; @@ -3461,8 +3460,6 @@ WORD32 isvcd_parse_pslice(svc_dec_lyr_struct_t *ps_svc_lyr_dec, UWORD16 u2_first ps_cur_slice->i1_slice_beta_offset = 0; } - isvcd_parse_interlayer_resamp_func_init(ps_svc_lyr_dec, u2_first_mb_in_slice); - ps_dec->u1_slice_header_done = 2; if(ps_pps->u1_entropy_coding_mode) { diff --git a/decoder/svc/isvcd_parse_headers.c b/decoder/svc/isvcd_parse_headers.c index abf4a1b..d137040 100644 --- a/decoder/svc/isvcd_parse_headers.c +++ b/decoder/svc/isvcd_parse_headers.c @@ -193,14 +193,6 @@ WORD32 isvcd_parse_subset_sps(svc_dec_lyr_struct_t *ps_svc_lyr_dec, dec_bit_stre ps_seq = ps_dec->pv_scratch_sps_pps; memset(ps_seq, 0, sizeof(dec_seq_params_t)); - if(ps_dec->i4_header_decoded & 1) - { - if(NULL != ps_dec->ps_cur_sps) - *ps_seq = *ps_dec->ps_cur_sps; - else - return ERROR_INV_SPS_PPS_T; - } - ps_seq->u1_profile_idc = u1_profile_idc; ps_seq->u1_level_idc = u1_level_idc; ps_seq->u1_seq_parameter_set_id = u1_seq_parameter_set_id; @@ -414,6 +406,15 @@ WORD32 isvcd_parse_subset_sps(svc_dec_lyr_struct_t *ps_svc_lyr_dec, dec_bit_stre u2_pic_wd = (u4_pic_width_in_mbs << 4); u2_pic_ht = (u4_pic_height_in_map_units << 4); + if(ps_svc_lyr_dec->pic_width < u2_pic_wd) + { + ps_svc_lyr_dec->pic_width = u2_pic_wd; + } + if(ps_svc_lyr_dec->pic_height < u2_pic_ht) + { + ps_svc_lyr_dec->pic_height = u2_pic_ht; + } + /*--------------------------------------------------------------------*/ /* Get the value of MaxMbAddress and Number of bits needed for it */ /*--------------------------------------------------------------------*/ @@ -850,35 +851,6 @@ WORD32 isvcd_parse_subset_sps(svc_dec_lyr_struct_t *ps_svc_lyr_dec, dec_bit_stre ps_subset_seq->u2_crop_offset_y = u2_crop_offset_y; ps_subset_seq->u2_crop_offset_uv = u2_crop_offset_uv; - if(((ps_dec->u2_pic_wd * ps_dec->u2_pic_ht) < - (ps_subset_seq->u2_pic_wd * ps_subset_seq->u2_pic_ht)) || - (ps_dec->i4_reorder_depth < ps_subset_seq->i4_reorder_depth)) - { - ps_dec->i4_reorder_depth = ps_subset_seq->i4_reorder_depth; - - ps_dec->u2_disp_height = ps_subset_seq->u2_disp_height; - ps_dec->u2_disp_width = ps_subset_seq->u2_disp_width; - - ps_dec->u2_pic_wd = ps_subset_seq->u2_pic_wd; - ps_dec->u2_pic_ht = ps_subset_seq->u2_pic_ht; - ps_dec->u4_total_mbs = ps_seq->u2_total_num_of_mbs << (1 - ps_seq->u1_frame_mbs_only_flag); - - /* Determining the Width and Height of Frame from that of Picture */ - ps_dec->u2_frm_wd_y = ps_subset_seq->u2_frm_wd_y; - ps_dec->u2_frm_ht_y = ps_subset_seq->u2_frm_ht_y; - ps_dec->u2_frm_wd_uv = ps_subset_seq->u2_frm_wd_uv; - ps_dec->u2_frm_ht_uv = ps_subset_seq->u2_frm_ht_uv; - - ps_dec->s_pad_mgr.u1_pad_len_y_v = ps_subset_seq->u1_pad_len_y_v; - ps_dec->s_pad_mgr.u1_pad_len_cr_v = ps_subset_seq->u1_pad_len_cr_v; - - ps_dec->u2_frm_wd_in_mbs = ps_seq->u2_frm_wd_in_mbs; - ps_dec->u2_frm_ht_in_mbs = ps_seq->u2_frm_ht_in_mbs; - - ps_dec->u2_crop_offset_y = ps_subset_seq->u2_crop_offset_y; - ps_dec->u2_crop_offset_uv = ps_subset_seq->u2_crop_offset_uv; - } - ps_seq->u1_is_valid = TRUE; ps_dec->ps_sps[u1_seq_parameter_set_id] = *ps_seq; if(NULL != ps_svc_lyr_dec->ps_subset_sps[u1_seq_parameter_set_id].s_sps_svc_ext.ps_svc_vui_ext) @@ -887,7 +859,6 @@ WORD32 isvcd_parse_subset_sps(svc_dec_lyr_struct_t *ps_svc_lyr_dec, dec_bit_stre ps_svc_lyr_dec->ps_subset_sps[u1_seq_parameter_set_id].s_sps_svc_ext.ps_svc_vui_ext; } ps_svc_lyr_dec->ps_subset_sps[u1_seq_parameter_set_id] = *ps_subset_seq; - ps_dec->ps_cur_sps = &ps_dec->ps_sps[u1_seq_parameter_set_id]; ps_svc_lyr_dec->ps_cur_subset_sps = &ps_svc_lyr_dec->ps_subset_sps[u1_seq_parameter_set_id]; return OK; @@ -1186,14 +1157,6 @@ WORD32 isvcd_parse_sps(svc_dec_lyr_struct_t *ps_svc_lyr_dec, dec_bit_stream_t *p ps_seq = ps_dec->pv_scratch_sps_pps; memset(ps_seq, 0, sizeof(dec_seq_params_t)); - if(ps_dec->i4_header_decoded & 1) - { - if(ps_dec->ps_cur_sps != NULL) - *ps_seq = *ps_dec->ps_cur_sps; - else - return ERROR_INV_SPS_PPS_T; - } - if((ps_dec->i4_header_decoded & 1) && (1 == ps_dec->ps_sps[u1_seq_parameter_set_id].u1_is_valid) && (ps_dec->ps_sps[u1_seq_parameter_set_id].u1_profile_idc != u1_profile_idc)) @@ -1404,6 +1367,15 @@ WORD32 isvcd_parse_sps(svc_dec_lyr_struct_t *ps_svc_lyr_dec, dec_bit_stream_t *p ps_seq->u2_frm_ht_in_mbs = u4_pic_height_in_map_units; u2_pic_wd = (u4_pic_width_in_mbs << 4); u2_pic_ht = (u4_pic_height_in_map_units << 4); + if(ps_svc_lyr_dec->pic_width < u2_pic_wd) + { + ps_svc_lyr_dec->pic_width = u2_pic_wd; + } + if(ps_svc_lyr_dec->pic_height < u2_pic_ht) + { + ps_svc_lyr_dec->pic_height = u2_pic_ht; + } + /*--------------------------------------------------------------------*/ /* Get the value of MaxMbAddress and Number of bits needed for it */ /*--------------------------------------------------------------------*/ @@ -1639,37 +1611,9 @@ WORD32 isvcd_parse_sps(svc_dec_lyr_struct_t *ps_svc_lyr_dec, dec_bit_stream_t *p ps_subset_seq->u2_crop_offset_y = u2_crop_offset_y; ps_subset_seq->u2_crop_offset_uv = u2_crop_offset_uv; - if(((ps_dec->u2_pic_wd * ps_dec->u2_pic_ht) < - (ps_subset_seq->u2_pic_wd * ps_subset_seq->u2_pic_ht)) || - (ps_dec->i4_reorder_depth < ps_subset_seq->i4_reorder_depth)) - { - ps_dec->i4_reorder_depth = ps_subset_seq->i4_reorder_depth; - - ps_dec->u2_disp_height = ps_subset_seq->u2_disp_height; - ps_dec->u2_disp_width = ps_subset_seq->u2_disp_width; - ps_dec->u2_pic_wd = ps_subset_seq->u2_pic_wd; - ps_dec->u2_pic_ht = ps_subset_seq->u2_pic_ht; - ps_dec->u4_total_mbs = ps_seq->u2_total_num_of_mbs << (1 - ps_seq->u1_frame_mbs_only_flag); - - /* Determining the Width and Height of Frame from that of Picture */ - ps_dec->u2_frm_wd_y = ps_subset_seq->u2_frm_wd_y; - ps_dec->u2_frm_ht_y = ps_subset_seq->u2_frm_ht_y; - ps_dec->u2_frm_wd_uv = ps_subset_seq->u2_frm_wd_uv; - ps_dec->u2_frm_ht_uv = ps_subset_seq->u2_frm_ht_uv; - - ps_dec->s_pad_mgr.u1_pad_len_y_v = ps_subset_seq->u1_pad_len_y_v; - ps_dec->s_pad_mgr.u1_pad_len_cr_v = ps_subset_seq->u1_pad_len_cr_v; - - ps_dec->u2_frm_wd_in_mbs = ps_seq->u2_frm_wd_in_mbs; - ps_dec->u2_frm_ht_in_mbs = ps_seq->u2_frm_ht_in_mbs; - ps_dec->u2_crop_offset_y = ps_subset_seq->u2_crop_offset_y; - ps_dec->u2_crop_offset_uv = ps_subset_seq->u2_crop_offset_uv; - } - ps_seq->u1_is_valid = TRUE; ps_dec->ps_sps[u1_seq_parameter_set_id] = *ps_seq; ps_svc_lyr_dec->ps_subset_sps[u1_seq_parameter_set_id] = *ps_subset_seq; - ps_dec->ps_cur_sps = &ps_dec->ps_sps[u1_seq_parameter_set_id]; ps_svc_lyr_dec->ps_cur_subset_sps = &ps_svc_lyr_dec->ps_subset_sps[u1_seq_parameter_set_id]; return OK; diff --git a/decoder/svc/isvcd_parse_slice.c b/decoder/svc/isvcd_parse_slice.c index f8270b6..226a6e4 100644 --- a/decoder/svc/isvcd_parse_slice.c +++ b/decoder/svc/isvcd_parse_slice.c @@ -703,6 +703,14 @@ WORD32 isvcd_parse_decode_slice_ext_nal(UWORD8 u1_is_idr_slice, UWORD8 u1_nal_re return ERROR_CORRUPTED_SLICE; } + if(ps_dec->u4_first_slice_in_pic == 1) + { + if(u2_first_mb_in_slice != 0) + { + return ERROR_CORRUPTED_SLICE; + } + } + COPYTHECONTEXT("Slice Header SVC ext: first_mb_in_slice", u2_first_mb_in_slice); u4_temp = ih264d_uev(pu4_bitstrm_ofst, pu4_bitstrm_buf); @@ -1202,8 +1210,13 @@ WORD32 isvcd_parse_decode_slice_ext_nal(UWORD8 u1_is_idr_slice, UWORD8 u1_nal_re else ps_dec->u4_output_present = 1; } + ret = isvcd_parse_interlayer_resamp_func_init(ps_svc_lyr_dec, u2_first_mb_in_slice); + if(ret != OK) + { + return ERROR_CORRUPTED_SLICE; + } if((ps_dec->u1_separate_parse == 1) && - (ps_svc_lyr_dec->u1_layer_identifier == TARGET_LAYER)) + (ps_svc_lyr_dec->u1_layer_identifier == TARGET_LAYER) && (ps_svc_lyr_dec->u1_res_init_done == 1)) { if(ps_dec->u4_dec_thread_created == 0) { @@ -1635,11 +1648,14 @@ WORD32 isvcd_parse_slice_header(svc_dec_lyr_struct_t *ps_svc_lyr_dec) { return ERROR_INV_SLICE_HDR_T; } + /* Reference layer id update is taken care during resolution init */ + /* ps_svc_lyr_dec->u1_ref_layer_id = ps_svc_slice_params->u4_ref_layer_dq_id >> 4; if(ps_svc_lyr_dec->u1_ref_layer_id >= ps_svc_lyr_dec->u1_layer_id) { return ERROR_INV_SLICE_HDR_T; } + */ ps_svc_lyr_dec->ps_dec_svc_ref_layer = &ps_svcd_ctxt->ps_svc_dec_lyr[ps_svc_lyr_dec->u1_ref_layer_id]; @@ -1684,6 +1700,11 @@ WORD32 isvcd_parse_slice_header(svc_dec_lyr_struct_t *ps_svc_lyr_dec) ps_svc_slice_params->u1_constrained_intra_resampling_flag = ih264d_get_bit_h264(ps_bitstrm); COPYTHECONTEXT("Slice Header SVC ext: u1_constrained_intra_resampling_flag", ps_svc_slice_params->u1_constrained_intra_resampling_flag); + + ps_svc_lyr_dec->s_res_prms.i1_constrained_intra_rsmpl_flag = + ps_svc_lyr_dec->s_svc_slice_params.u1_constrained_intra_resampling_flag; + isvcd_intra_resamp_res_init_update_flags(ps_svc_lyr_dec); + if(2 == ps_sps_svc_ext->u1_extended_spatial_scalability_idc) { /* ChromaArrayType = i4_chroma_format_idc if separate_colour_plane_flag @@ -1926,6 +1947,14 @@ WORD32 isvcd_parse_decode_slice(UWORD8 u1_is_idr_slice, UWORD8 u1_nal_ref_idc, return ERROR_CORRUPTED_SLICE; } + if(ps_dec->u4_first_slice_in_pic == 1) + { + if(u2_first_mb_in_slice != 0) + { + return ERROR_CORRUPTED_SLICE; + } + } + COPYTHECONTEXT("SH: first_mb_in_slice", u2_first_mb_in_slice); u4_temp = ih264d_uev(pu4_bitstrm_ofst, pu4_bitstrm_buf); @@ -2421,7 +2450,12 @@ WORD32 isvcd_parse_decode_slice(UWORD8 u1_is_idr_slice, UWORD8 u1_nal_ref_idc, else ps_dec->u4_output_present = 1; } - if(ps_dec->u1_separate_parse == 1) + ret = isvcd_parse_interlayer_resamp_func_init(ps_svc_lyr_dec, u2_first_mb_in_slice); + if(ret != OK) + { + return ERROR_CORRUPTED_SLICE; + } + if((ps_dec->u1_separate_parse == 1) && (ps_svc_lyr_dec->u1_res_init_done == 1)) { if(ps_dec->u4_dec_thread_created == 0) { diff --git a/decoder/svc/isvcd_process_epslice.c b/decoder/svc/isvcd_process_epslice.c index cbcce8c..b351319 100644 --- a/decoder/svc/isvcd_process_epslice.c +++ b/decoder/svc/isvcd_process_epslice.c @@ -158,11 +158,13 @@ WORD32 isvcd_interlyr_motion_mode_pred(svc_dec_lyr_struct_t *ps_svc_lyr_dec, /* get the current layer ctxt */ ps_lyr_mem = &ps_ctxt->as_res_lyr_mem[ps_ctxt->i4_res_id]; - ps_ctxt->i4_listx = i4_listx; + { + ps_ctxt->i4_listx = i4_listx; - i4_mb_mode = - ps_lyr_mem->pf_inter_lyr_pred(ps_svc_lyr_dec->pv_mode_mv_sample_ctxt, ps_cur_mb_info, + i4_mb_mode = + ps_lyr_mem->pf_inter_lyr_pred(ps_svc_lyr_dec->pv_mode_mv_sample_ctxt, ps_cur_mb_info, ps_svc_cur_mb_info, ps_dec, ps_mb_part_info, ps_part); + } } return i4_mb_mode; } @@ -485,6 +487,7 @@ WORD32 isvcd_mv_pred_ref_tfr_nby2_epmb(dec_struct_t *ps_dec, UWORD8 u1_mb_idx, U } } } + return OK; } /*! diff --git a/decoder/svc/isvcd_structs.h b/decoder/svc/isvcd_structs.h index 1959202..85f1295 100644 --- a/decoder/svc/isvcd_structs.h +++ b/decoder/svc/isvcd_structs.h @@ -660,6 +660,9 @@ typedef struct _SvcDecLyrStruct WORD32 i4_frm_svc_base_mode_cabac_size; UWORD32 u4_pps_id_for_layer; UWORD8 u1_error_in_cur_frame; + UWORD8 u1_res_init_done; + WORD32 pic_width; + WORD32 pic_height; } svc_dec_lyr_struct_t; typedef struct @@ -719,6 +722,8 @@ typedef struct UWORD8 u1_parse_nal_unit_error; UWORD8 u1_exit_till_next_IDR; UWORD8 u1_pre_parse_in_flush; + WORD32 pic_width; + WORD32 pic_height; } svc_dec_ctxt_t; #endif /*_ISVCD_STRUCTS_H_*/
\ No newline at end of file diff --git a/encoder/arm/svc/isvce_function_selector_a9q.c b/encoder/arm/svc/isvce_function_selector_a9q.c index b5f8ba4..6a1abb1 100644 --- a/encoder/arm/svc/isvce_function_selector_a9q.c +++ b/encoder/arm/svc/isvce_function_selector_a9q.c @@ -161,35 +161,23 @@ void isvce_init_function_ptr_neon_a9q(isvce_codec_t *ps_codec) isvc_resi_trans_quant_chroma_4x4_with_residual_sub_neon; /* Init inverse transform fn ptr */ - ps_enc_loop_fxns->apf_iquant_itrans_recon_8x8[0] = isvc_iquant_itrans_recon_8x8; - ps_enc_loop_fxns->apf_iquant_itrans_recon_8x8[1] = isvc_iquant_itrans_recon_8x8; - ps_enc_loop_fxns->apf_iquant_itrans_recon_8x8[2] = isvc_iquant_itrans_recon_8x8; - ps_enc_loop_fxns->apf_iquant_itrans_recon_4x4[0] = isvc_iquant_itrans_recon_4x4_with_res_output_neon; ps_enc_loop_fxns->apf_iquant_itrans_recon_4x4[1] = isvc_iquant_itrans_recon_4x4_with_res_accumulate_neon; ps_enc_loop_fxns->apf_iquant_itrans_recon_4x4[2] = isvc_iquant_itrans_recon_4x4_neon; - ps_enc_loop_fxns->apf_iquant_itrans_recon_4x4_dc[0] = - isvc_iquant_itrans_recon_4x4_dc_with_res_output_neon; - ps_enc_loop_fxns->apf_iquant_itrans_recon_4x4_dc[1] = - isvc_iquant_itrans_recon_4x4_dc_with_res_accumulate_neon; ps_enc_loop_fxns->apf_iquant_itrans_recon_4x4_dc[2] = isvc_iquant_itrans_recon_4x4_dc_neon; ps_enc_loop_fxns->apf_iquant_itrans_recon_chroma_4x4[0] = isvc_iquant_itrans_recon_chroma_4x4_with_res_output_neon; ps_enc_loop_fxns->apf_iquant_itrans_recon_chroma_4x4[1] = isvc_iquant_itrans_recon_chroma_4x4_with_res_accumulate_neon; - ps_enc_loop_fxns->apf_iquant_itrans_recon_chroma_4x4[2] = - isvc_iquant_itrans_recon_chroma_4x4_neon; ps_enc_loop_fxns->apf_iquant_itrans_recon_chroma_4x4_dc[0] = isvc_iquant_itrans_recon_chroma_4x4_dc_with_res_output_neon; ps_enc_loop_fxns->apf_iquant_itrans_recon_chroma_4x4_dc[1] = isvc_iquant_itrans_recon_chroma_4x4_dc_with_res_accumulate_neon; - ps_enc_loop_fxns->apf_iquant_itrans_recon_chroma_4x4_dc[2] = - isvc_iquant_itrans_recon_chroma_4x4_dc_neon; ps_enc_loop_fxns->pf_ihadamard_scaling_4x4 = ih264_ihadamard_scaling_4x4_a9; ps_enc_loop_fxns->pf_ihadamard_scaling_2x2_uv = ih264_ihadamard_scaling_2x2_uv_a9; @@ -243,9 +231,7 @@ void isvce_init_function_ptr_neon_a9q(isvce_codec_t *ps_codec) ps_codec->pf_compute_sad_16x8 = ime_compute_sad_16x8_a9q; /* memor handling operations */ - ps_mem_fxns->pf_mem_cpy = ih264_memcpy_a9q; ps_mem_fxns->pf_mem_cpy_mul8 = ih264_memcpy_mul_8_a9q; - ps_mem_fxns->pf_mem_set = ih264_memset_a9q; ps_mem_fxns->pf_mem_set_mul8 = ih264_memset_mul_8_a9q; /* sad me level functions */ diff --git a/encoder/arm/svc/isvce_function_selector_av8.c b/encoder/arm/svc/isvce_function_selector_av8.c index 16c08bb..c517c03 100644 --- a/encoder/arm/svc/isvce_function_selector_av8.c +++ b/encoder/arm/svc/isvce_function_selector_av8.c @@ -161,35 +161,23 @@ void isvce_init_function_ptr_neon_av8(isvce_codec_t *ps_codec) isvc_resi_trans_quant_chroma_4x4_with_residual_sub_neon; /* Init inverse transform fn ptr */ - ps_enc_loop_fxns->apf_iquant_itrans_recon_8x8[0] = isvc_iquant_itrans_recon_8x8; - ps_enc_loop_fxns->apf_iquant_itrans_recon_8x8[1] = isvc_iquant_itrans_recon_8x8; - ps_enc_loop_fxns->apf_iquant_itrans_recon_8x8[2] = isvc_iquant_itrans_recon_8x8; - ps_enc_loop_fxns->apf_iquant_itrans_recon_4x4[0] = isvc_iquant_itrans_recon_4x4_with_res_output_neon; ps_enc_loop_fxns->apf_iquant_itrans_recon_4x4[1] = isvc_iquant_itrans_recon_4x4_with_res_accumulate_neon; ps_enc_loop_fxns->apf_iquant_itrans_recon_4x4[2] = isvc_iquant_itrans_recon_4x4_neon; - - ps_enc_loop_fxns->apf_iquant_itrans_recon_4x4_dc[0] = - isvc_iquant_itrans_recon_4x4_dc_with_res_output_neon; - ps_enc_loop_fxns->apf_iquant_itrans_recon_4x4_dc[1] = - isvc_iquant_itrans_recon_4x4_dc_with_res_accumulate_neon; + ; ps_enc_loop_fxns->apf_iquant_itrans_recon_4x4_dc[2] = isvc_iquant_itrans_recon_4x4_dc_neon; ps_enc_loop_fxns->apf_iquant_itrans_recon_chroma_4x4[0] = isvc_iquant_itrans_recon_chroma_4x4_with_res_output_neon; ps_enc_loop_fxns->apf_iquant_itrans_recon_chroma_4x4[1] = isvc_iquant_itrans_recon_chroma_4x4_with_res_accumulate_neon; - ps_enc_loop_fxns->apf_iquant_itrans_recon_chroma_4x4[2] = - isvc_iquant_itrans_recon_chroma_4x4_neon; ps_enc_loop_fxns->apf_iquant_itrans_recon_chroma_4x4_dc[0] = isvc_iquant_itrans_recon_chroma_4x4_dc_with_res_output_neon; ps_enc_loop_fxns->apf_iquant_itrans_recon_chroma_4x4_dc[1] = isvc_iquant_itrans_recon_chroma_4x4_dc_with_res_accumulate_neon; - ps_enc_loop_fxns->apf_iquant_itrans_recon_chroma_4x4_dc[2] = - isvc_iquant_itrans_recon_chroma_4x4_dc_neon; ps_enc_loop_fxns->pf_ihadamard_scaling_4x4 = ih264_ihadamard_scaling_4x4_av8; ps_enc_loop_fxns->pf_ihadamard_scaling_2x2_uv = ih264_ihadamard_scaling_2x2_uv_av8; @@ -243,9 +231,7 @@ void isvce_init_function_ptr_neon_av8(isvce_codec_t *ps_codec) ps_codec->pf_compute_sad_16x8 = ime_compute_sad_16x8_av8; /* memor handling operations */ - ps_mem_fxns->pf_mem_cpy = ih264_memcpy_av8; ps_mem_fxns->pf_mem_cpy_mul8 = ih264_memcpy_mul_8_av8; - ps_mem_fxns->pf_mem_set = ih264_memset_av8; ps_mem_fxns->pf_mem_set_mul8 = ih264_memset_mul_8_av8; /* sad me level functions */ diff --git a/encoder/ih264e_api.c b/encoder/ih264e_api.c index f76fbd1..c952225 100644 --- a/encoder/ih264e_api.c +++ b/encoder/ih264e_api.c @@ -139,101 +139,6 @@ ******************************************************************************* * * @brief -* Used to test validity of input dimensions -* -* @par Description: -* Dimensions of the input buffer passed to encode call are validated -* -* @param[in] ps_codec -* Codec context -* -* @param[in] ps_ip -* Pointer to input structure -* -* @param[out] ps_op -* Pointer to output structure -* -* @returns error status -* -* @remarks none -* -******************************************************************************* -*/ -static IV_STATUS_T api_check_input_dimensions(codec_t *ps_codec, - ih264e_video_encode_ip_t *ps_ip, - ih264e_video_encode_op_t *ps_op) -{ - UWORD32 u4_wd, u4_ht; - cfg_params_t *ps_curr_cfg = &ps_codec->s_cfg; - iv_raw_buf_t *ps_inp_buf = &ps_ip->s_ive_ip.s_inp_buf; - - u4_wd = ps_inp_buf->au4_wd[0]; - u4_ht = ps_inp_buf->au4_ht[0]; - switch (ps_inp_buf->e_color_fmt) - { - case IV_YUV_420P: - if (((ps_inp_buf->au4_wd[0] / 2) != ps_inp_buf->au4_wd[1]) || - ((ps_inp_buf->au4_wd[0] / 2) != ps_inp_buf->au4_wd[2]) || - (ps_inp_buf->au4_wd[1] != ps_inp_buf->au4_wd[2])) - { - ps_op->s_ive_op.u4_error_code |= 1 << IVE_UNSUPPORTEDPARAM; - ps_op->s_ive_op.u4_error_code |= IH264E_WIDTH_NOT_SUPPORTED; - return (IV_FAIL); - } - if (((ps_inp_buf->au4_ht[0] / 2) != ps_inp_buf->au4_ht[1]) || - ((ps_inp_buf->au4_ht[0] / 2) != ps_inp_buf->au4_ht[2]) || - (ps_inp_buf->au4_ht[1] != ps_inp_buf->au4_ht[2])) - { - ps_op->s_ive_op.u4_error_code |= 1 << IVE_UNSUPPORTEDPARAM; - ps_op->s_ive_op.u4_error_code |= IH264E_HEIGHT_NOT_SUPPORTED; - return (IV_FAIL); - } - break; - case IV_YUV_420SP_UV: - case IV_YUV_420SP_VU: - if (ps_inp_buf->au4_wd[0] != ps_inp_buf->au4_wd[1]) - { - ps_op->s_ive_op.u4_error_code |= 1 << IVE_UNSUPPORTEDPARAM; - ps_op->s_ive_op.u4_error_code |= IH264E_WIDTH_NOT_SUPPORTED; - return (IV_FAIL); - } - if ((ps_inp_buf->au4_ht[0] / 2) != ps_inp_buf->au4_ht[1]) - { - ps_op->s_ive_op.u4_error_code |= 1 << IVE_UNSUPPORTEDPARAM; - ps_op->s_ive_op.u4_error_code |= IH264E_HEIGHT_NOT_SUPPORTED; - return (IV_FAIL); - } - break; - case IV_YUV_422ILE: - u4_wd = ps_inp_buf->au4_wd[0] / 2; - break; - default: - ps_op->s_ive_op.u4_error_code |= 1 << IVE_UNSUPPORTEDPARAM; - ps_op->s_ive_op.u4_error_code |= IH264E_INPUT_CHROMA_FORMAT_NOT_SUPPORTED; - return (IV_FAIL); - } - - if (u4_wd != ps_curr_cfg->u4_disp_wd) - { - ps_op->s_ive_op.u4_error_code |= 1 << IVE_UNSUPPORTEDPARAM; - ps_op->s_ive_op.u4_error_code |= IH264E_WIDTH_NOT_SUPPORTED; - return (IV_FAIL); - } - - if (u4_ht != ps_curr_cfg->u4_disp_ht) - { - ps_op->s_ive_op.u4_error_code |= 1 << IVE_UNSUPPORTEDPARAM; - ps_op->s_ive_op.u4_error_code |= IH264E_HEIGHT_NOT_SUPPORTED; - return (IV_FAIL); - } - - return IV_SUCCESS; -} - -/** -******************************************************************************* -* -* @brief * Used to test arguments for corresponding API call * * @par Description: @@ -910,7 +815,6 @@ static IV_STATUS_T api_check_struct_sanity(iv_obj_t *ps_handle, case IVE_CMD_VIDEO_ENCODE: { - codec_t *ps_codec = (codec_t *) (ps_handle->pv_codec_handle); ih264e_video_encode_ip_t *ps_ip = pv_api_ip; ih264e_video_encode_op_t *ps_op = pv_api_op; @@ -930,14 +834,6 @@ static IV_STATUS_T api_check_struct_sanity(iv_obj_t *ps_handle, return (IV_FAIL); } - if (NULL != ps_ip->s_ive_ip.s_inp_buf.apv_bufs[0] && - ps_codec->i4_header_mode != 1 && - IV_SUCCESS != api_check_input_dimensions(ps_codec, ps_ip, ps_op)) - { - ps_op->s_ive_op.u4_error_code |= 1 << IVE_UNSUPPORTEDPARAM; - ps_op->s_ive_op.u4_error_code |= IVE_ERR_OP_ENCODE_API_STRUCT_SIZE_INCORRECT; - return (IV_FAIL); - } break; } diff --git a/encoder/ih264e_encode.c b/encoder/ih264e_encode.c index f5d14d2..1613e77 100644 --- a/encoder/ih264e_encode.c +++ b/encoder/ih264e_encode.c @@ -184,6 +184,101 @@ IH264E_ERROR_T ih264e_wait_for_thread(UWORD32 sleep_us) } /** +******************************************************************************* +* +* @brief +* Used to test validity of input dimensions +* +* @par Description: +* Dimensions of the input buffer passed to encode call are validated +* +* @param[in] ps_codec +* Codec context +* +* @param[in] ps_ip +* Pointer to input structure +* +* @param[out] ps_op +* Pointer to output structure +* +* @returns error status +* +* @remarks none +* +******************************************************************************* +*/ +static IV_STATUS_T api_check_input_dimensions(codec_t *ps_codec, + ih264e_video_encode_ip_t *ps_ip, + ih264e_video_encode_op_t *ps_op) +{ + UWORD32 u4_wd, u4_ht; + cfg_params_t *ps_curr_cfg = &ps_codec->s_cfg; + iv_raw_buf_t *ps_inp_buf = &ps_ip->s_ive_ip.s_inp_buf; + + u4_wd = ps_inp_buf->au4_wd[0]; + u4_ht = ps_inp_buf->au4_ht[0]; + switch (ps_inp_buf->e_color_fmt) + { + case IV_YUV_420P: + if (((ps_inp_buf->au4_wd[0] / 2) != ps_inp_buf->au4_wd[1]) || + ((ps_inp_buf->au4_wd[0] / 2) != ps_inp_buf->au4_wd[2]) || + (ps_inp_buf->au4_wd[1] != ps_inp_buf->au4_wd[2])) + { + ps_op->s_ive_op.u4_error_code |= 1 << IVE_UNSUPPORTEDPARAM; + ps_op->s_ive_op.u4_error_code |= IH264E_WIDTH_NOT_SUPPORTED; + return (IV_FAIL); + } + if (((ps_inp_buf->au4_ht[0] / 2) != ps_inp_buf->au4_ht[1]) || + ((ps_inp_buf->au4_ht[0] / 2) != ps_inp_buf->au4_ht[2]) || + (ps_inp_buf->au4_ht[1] != ps_inp_buf->au4_ht[2])) + { + ps_op->s_ive_op.u4_error_code |= 1 << IVE_UNSUPPORTEDPARAM; + ps_op->s_ive_op.u4_error_code |= IH264E_HEIGHT_NOT_SUPPORTED; + return (IV_FAIL); + } + break; + case IV_YUV_420SP_UV: + case IV_YUV_420SP_VU: + if (ps_inp_buf->au4_wd[0] != ps_inp_buf->au4_wd[1]) + { + ps_op->s_ive_op.u4_error_code |= 1 << IVE_UNSUPPORTEDPARAM; + ps_op->s_ive_op.u4_error_code |= IH264E_WIDTH_NOT_SUPPORTED; + return (IV_FAIL); + } + if ((ps_inp_buf->au4_ht[0] / 2) != ps_inp_buf->au4_ht[1]) + { + ps_op->s_ive_op.u4_error_code |= 1 << IVE_UNSUPPORTEDPARAM; + ps_op->s_ive_op.u4_error_code |= IH264E_HEIGHT_NOT_SUPPORTED; + return (IV_FAIL); + } + break; + case IV_YUV_422ILE: + u4_wd = ps_inp_buf->au4_wd[0] / 2; + break; + default: + ps_op->s_ive_op.u4_error_code |= 1 << IVE_UNSUPPORTEDPARAM; + ps_op->s_ive_op.u4_error_code |= IH264E_INPUT_CHROMA_FORMAT_NOT_SUPPORTED; + return (IV_FAIL); + } + + if (u4_wd != ps_curr_cfg->u4_disp_wd) + { + ps_op->s_ive_op.u4_error_code |= 1 << IVE_UNSUPPORTEDPARAM; + ps_op->s_ive_op.u4_error_code |= IH264E_WIDTH_NOT_SUPPORTED; + return (IV_FAIL); + } + + if (u4_ht != ps_curr_cfg->u4_disp_ht) + { + ps_op->s_ive_op.u4_error_code |= 1 << IVE_UNSUPPORTEDPARAM; + ps_op->s_ive_op.u4_error_code |= IH264E_HEIGHT_NOT_SUPPORTED; + return (IV_FAIL); + } + + return IV_SUCCESS; +} + +/** ****************************************************************************** * * @brief @@ -427,24 +522,33 @@ WORD32 ih264e_encode(iv_obj_t *ps_codec_obj, void *pv_api_ip, void *pv_api_op) } } #endif - /****************************************************************** - * INSERT LOGO - *****************************************************************/ -#ifdef LOGO_EN - if (s_inp_buf.s_raw_buf.apv_bufs[0] != NULL && + + if (ps_video_encode_ip->s_ive_ip.s_inp_buf.apv_bufs[0] != NULL && ps_codec->i4_header_mode != 1) { - ih264e_insert_logo(s_inp_buf.s_raw_buf.apv_bufs[0], - s_inp_buf.s_raw_buf.apv_bufs[1], - s_inp_buf.s_raw_buf.apv_bufs[2], - s_inp_buf.s_raw_buf.au4_strd[0], + if (IV_SUCCESS != api_check_input_dimensions(ps_codec, pv_api_ip, pv_api_op)) + { + error_status = IH264E_FAIL; + SET_ERROR_ON_RETURN(error_status, + IVE_FATALERROR, + ps_video_encode_op->s_ive_op.u4_error_code, + IV_FAIL); + } + /****************************************************************** + * INSERT LOGO + *****************************************************************/ +#ifdef LOGO_EN + ih264e_insert_logo(ps_video_encode_ip->s_ive_ip.s_inp_buf.apv_bufs[0], + ps_video_encode_ip->s_ive_ip.s_inp_buf.apv_bufs[1], + ps_video_encode_ip->s_ive_ip.s_inp_buf.apv_bufs[2], + ps_video_encode_ip->s_ive_ip.s_inp_buf.au4_strd[0], 0, 0, ps_codec->s_cfg.e_inp_color_fmt, ps_codec->s_cfg.u4_disp_wd, ps_codec->s_cfg.u4_disp_ht); - } #endif /*LOGO_EN*/ + } /* In case of alt ref and B pics we will have non reference frame in stream */ if (ps_codec->s_cfg.u4_enable_alt_ref || ps_codec->s_cfg.u4_num_bframes) diff --git a/encoder/svc/isvce_api.c b/encoder/svc/isvce_api.c index 2004dba..d44d7e7 100644 --- a/encoder/svc/isvce_api.c +++ b/encoder/svc/isvce_api.c @@ -3342,10 +3342,14 @@ static WORD32 isvce_init_mem_rec(iv_obj_t *ps_codec_obj, void *pv_api_ip, void * WORD32 max_wd_luma, max_ht_luma; WORD32 max_mb_rows, max_mb_cols, max_mb_cnt; - /* temp var */ WORD32 i, j; WORD32 status = IV_SUCCESS; + if(MAX_CTXT_SETS != 1) + { + return IV_FAIL; + } + /* mem records */ ps_mem_rec_base = ps_ip->s_ive_ip.ps_mem_rec; @@ -3493,117 +3497,59 @@ static WORD32 isvce_init_mem_rec(iv_obj_t *ps_codec_obj, void *pv_api_ip, void * ps_mem_rec = &ps_mem_rec_base[ISVCE_MEM_REC_ENTROPY]; { /* temp var */ - WORD32 size = 0, offset; + WORD32 size = 0; for(i = 0; i < MAX_PROCESS_CTXT; i++) { - if(i < MAX_PROCESS_CTXT / MAX_CTXT_SETS) - { - /* base ptr */ - UWORD8 *pu1_buf = ps_mem_rec->pv_base; - - /* reset size */ - size = 0; - - /* skip mb run */ - ps_codec->as_process[i].s_entropy.pi4_mb_skip_run = (WORD32 *) (pu1_buf + size); - size += sizeof(WORD32); - size = ALIGN8(size); - - /* entropy map */ - ps_codec->as_process[i].s_entropy.pu1_entropy_map = - (UWORD8 *) (pu1_buf + size + max_mb_cols); - /* size in bytes to store entropy status of an entire frame */ - size += (max_mb_cols * max_mb_rows); - /* add an additional 1 row of bytes to evade the special case of row 0 - */ - size += max_mb_cols; - size = ALIGN128(size); - - /* bit stream ptr */ - ps_codec->as_process[i].s_entropy.ps_bitstrm = (bitstrm_t *) (pu1_buf + size); - size += sizeof(ps_codec->as_process[i].s_entropy.ps_bitstrm[0]); - size = ALIGN128(size); - -#if ENABLE_RE_ENC_AS_SKIP - ps_codec->as_process[i].s_entropy.ps_bitstrm_after_slice_hdr = - (bitstrm_t *) (pu1_buf + size); - size += sizeof(ps_codec->as_process[i].s_entropy.ps_bitstrm_after_slice_hdr[0]); - size = ALIGN128(size); -#endif - - /* nnz luma */ - ps_codec->as_process[i].s_entropy.pu1_top_nnz_luma = (UWORD8(*)[4])(pu1_buf + size); - size += (max_mb_cols * 4 * sizeof(UWORD8)); - size = ALIGN128(size); - - /* nnz chroma */ - ps_codec->as_process[i].s_entropy.pu1_top_nnz_cbcr = (UWORD8(*)[4])(pu1_buf + size); - size += (max_mb_cols * 4 * sizeof(UWORD8)); - size = ALIGN128(size); - - /* ps_mb_qp_ctxt */ - ps_codec->as_process[i].s_entropy.ps_mb_qp_ctxt = (mb_qp_ctxt_t *) (pu1_buf + size); - size += ALIGN128(sizeof(ps_codec->as_process[i].s_entropy.ps_mb_qp_ctxt[0])); - - offset = size; - - /* cabac Context */ - ps_codec->as_process[i].s_entropy.ps_cabac = ps_cabac; - } - else - { - /* base ptr */ - UWORD8 *pu1_buf = ps_mem_rec->pv_base; - - /* reset size */ - size = offset; - - /* skip mb run */ - ps_codec->as_process[i].s_entropy.pi4_mb_skip_run = (WORD32 *) (pu1_buf + size); - size += sizeof(WORD32); - size = ALIGN8(size); - - /* entropy map */ - ps_codec->as_process[i].s_entropy.pu1_entropy_map = - (UWORD8 *) (pu1_buf + size + max_mb_cols); - /* size in bytes to store entropy status of an entire frame */ - size += (max_mb_cols * max_mb_rows); - /* add an additional 1 row of bytes to evade the special case of row 0 - */ - size += max_mb_cols; - size = ALIGN128(size); - - /* bit stream ptr */ - ps_codec->as_process[i].s_entropy.ps_bitstrm = (bitstrm_t *) (pu1_buf + size); - size += sizeof(ps_codec->as_process[i].s_entropy.ps_bitstrm[0]); - size = ALIGN128(size); + /* base ptr */ + UWORD8 *pu1_buf = ps_mem_rec->pv_base; + + /* reset size */ + size = 0; + + /* skip mb run */ + ps_codec->as_process[i].s_entropy.pi4_mb_skip_run = (WORD32 *) (pu1_buf + size); + size += sizeof(WORD32); + size = ALIGN8(size); + + /* entropy map */ + ps_codec->as_process[i].s_entropy.pu1_entropy_map = + (UWORD8 *) (pu1_buf + size + max_mb_cols); + /* size in bytes to store entropy status of an entire frame */ + size += (max_mb_cols * max_mb_rows); + /* add an additional 1 row of bytes to evade the special case of row 0 + */ + size += max_mb_cols; + size = ALIGN128(size); + + /* bit stream ptr */ + ps_codec->as_process[i].s_entropy.ps_bitstrm = (bitstrm_t *) (pu1_buf + size); + size += sizeof(ps_codec->as_process[i].s_entropy.ps_bitstrm[0]); + size = ALIGN128(size); #if ENABLE_RE_ENC_AS_SKIP - ps_codec->as_process[i].s_entropy.ps_bitstrm_after_slice_hdr = - (bitstrm_t *) (pu1_buf + size); - size += sizeof(ps_codec->as_process[i].s_entropy.ps_bitstrm_after_slice_hdr[0]); - size = ALIGN128(size); + ps_codec->as_process[i].s_entropy.ps_bitstrm_after_slice_hdr = + (bitstrm_t *) (pu1_buf + size); + size += sizeof(ps_codec->as_process[i].s_entropy.ps_bitstrm_after_slice_hdr[0]); + size = ALIGN128(size); #endif - /* nnz luma */ - ps_codec->as_process[i].s_entropy.pu1_top_nnz_luma = - (UWORD8(*)[4])(UWORD8(*)[4])(pu1_buf + size); - size += (max_mb_cols * 4 * sizeof(UWORD8)); - size = ALIGN128(size); + /* nnz luma */ + ps_codec->as_process[i].s_entropy.pu1_top_nnz_luma = (UWORD8(*)[4])(pu1_buf + size); + size += (max_mb_cols * 4 * sizeof(UWORD8)); + size = ALIGN128(size); - /* nnz chroma */ - ps_codec->as_process[i].s_entropy.pu1_top_nnz_cbcr = (UWORD8(*)[4])(pu1_buf + size); - size += (max_mb_cols * 4 * sizeof(UWORD8)); - size = ALIGN128(size); + /* nnz chroma */ + ps_codec->as_process[i].s_entropy.pu1_top_nnz_cbcr = (UWORD8(*)[4])(pu1_buf + size); + size += (max_mb_cols * 4 * sizeof(UWORD8)); + size = ALIGN128(size); - /* ps_mb_qp_ctxt */ - ps_codec->as_process[i].s_entropy.ps_mb_qp_ctxt = (mb_qp_ctxt_t *) (pu1_buf + size); - size = ALIGN128(sizeof(ps_codec->as_process[i].s_entropy.ps_mb_qp_ctxt[0])); + /* ps_mb_qp_ctxt */ + ps_codec->as_process[i].s_entropy.ps_mb_qp_ctxt = (mb_qp_ctxt_t *) (pu1_buf + size); + size += ALIGN128(sizeof(ps_codec->as_process[i].s_entropy.ps_mb_qp_ctxt[0])); - /* cabac Context */ - ps_codec->as_process[i].s_entropy.ps_cabac = ps_cabac; - } + /* cabac Context */ + ps_codec->as_process[i].s_entropy.ps_cabac = ps_cabac; } ps_codec->as_process[0].s_entropy.ps_cabac->ps_mb_map_ctxt_inc_base = ps_mb_map_ctxt_inc; } @@ -3631,16 +3577,8 @@ static WORD32 isvce_init_mem_rec(iv_obj_t *ps_codec_obj, void *pv_api_ip, void * for(i = 0; i < MAX_PROCESS_CTXT; i++) { - if(i < MAX_PROCESS_CTXT / MAX_CTXT_SETS) - { - ps_codec->as_process[i].pv_pic_mb_coeff_data = pu1_buf; - ps_codec->as_process[i].s_entropy.pv_pic_mb_coeff_data = pu1_buf; - } - else - { - ps_codec->as_process[i].pv_pic_mb_coeff_data = pu1_buf + size; - ps_codec->as_process[i].s_entropy.pv_pic_mb_coeff_data = pu1_buf + size; - } + ps_codec->as_process[i].pv_pic_mb_coeff_data = pu1_buf; + ps_codec->as_process[i].s_entropy.pv_pic_mb_coeff_data = pu1_buf; } } @@ -3667,16 +3605,8 @@ static WORD32 isvce_init_mem_rec(iv_obj_t *ps_codec_obj, void *pv_api_ip, void * for(i = 0; i < MAX_PROCESS_CTXT; i++) { - if(i < MAX_PROCESS_CTXT / MAX_CTXT_SETS) - { - ps_codec->as_process[i].pv_pic_mb_header_data = pu1_buf; - ps_codec->as_process[i].s_entropy.pv_pic_mb_header_data = pu1_buf; - } - else - { - ps_codec->as_process[i].pv_pic_mb_header_data = pu1_buf + size; - ps_codec->as_process[i].s_entropy.pv_pic_mb_header_data = pu1_buf + size; - } + ps_codec->as_process[i].pv_pic_mb_header_data = pu1_buf; + ps_codec->as_process[i].s_entropy.pv_pic_mb_header_data = pu1_buf; } } @@ -3729,17 +3659,7 @@ static WORD32 isvce_init_mem_rec(iv_obj_t *ps_codec_obj, void *pv_api_ip, void * for(i = 0; i < MAX_PROCESS_CTXT; i++) { - if(i < MAX_PROCESS_CTXT / MAX_CTXT_SETS) - { - ps_codec->as_process[i].ps_svc_nalu_ext_base = ps_mem_rec->pv_base; - } - else - { - WORD32 size = SVC_MAX_SLICE_HDR_CNT * sizeof(slice_header_t); - void *pv_buf = (UWORD8 *) ps_mem_rec->pv_base + size; - - ps_codec->as_process[i].ps_svc_nalu_ext_base = pv_buf; - } + ps_codec->as_process[i].ps_svc_nalu_ext_base = ps_mem_rec->pv_base; } } @@ -3757,18 +3677,7 @@ static WORD32 isvce_init_mem_rec(iv_obj_t *ps_codec_obj, void *pv_api_ip, void * for(i = 0; i < MAX_PROCESS_CTXT; i++) { - if(i < MAX_PROCESS_CTXT / MAX_CTXT_SETS) - { - ps_codec->as_process[i].ps_slice_hdr_base = ps_mem_rec->pv_base; - } - else - { - /* temp var */ - WORD32 size = SVC_MAX_SLICE_HDR_CNT * sizeof(slice_header_t); - void *pv_buf = (UWORD8 *) ps_mem_rec->pv_base + size; - - ps_codec->as_process[i].ps_slice_hdr_base = pv_buf; - } + ps_codec->as_process[i].ps_slice_hdr_base = ps_mem_rec->pv_base; } } @@ -3789,14 +3698,7 @@ static WORD32 isvce_init_mem_rec(iv_obj_t *ps_codec_obj, void *pv_api_ip, void * for(i = 0; i < MAX_PROCESS_CTXT; i++) { - if(i < MAX_PROCESS_CTXT / MAX_CTXT_SETS) - { - ps_codec->as_process[i].pu1_is_intra_coded = pu1_buf; - } - else - { - ps_codec->as_process[i].pu1_is_intra_coded = pu1_buf + max_mb_cnt; - } + ps_codec->as_process[i].pu1_is_intra_coded = pu1_buf; } ps_codec->pu2_intr_rfrsh_map = (UWORD16 *) (pu1_buf + max_mb_cnt * MAX_CTXT_SETS); @@ -3805,22 +3707,14 @@ static WORD32 isvce_init_mem_rec(iv_obj_t *ps_codec_obj, void *pv_api_ip, void * ps_mem_rec = &ps_mem_rec_base[ISVCE_MEM_REC_SLICE_MAP]; { /* pointer to storage space */ - UWORD8 *pu1_buf_ping, *pu1_buf_pong; + UWORD8 *pu1_buf_ping; /* init pointer */ pu1_buf_ping = ps_mem_rec->pv_base; - pu1_buf_pong = pu1_buf_ping + ALIGN64(max_mb_cnt); for(i = 0; i < MAX_PROCESS_CTXT; i++) { - if(i < MAX_PROCESS_CTXT / MAX_CTXT_SETS) - { - ps_codec->as_process[i].pu1_slice_idx = pu1_buf_ping; - } - else - { - ps_codec->as_process[i].pu1_slice_idx = pu1_buf_pong; - } + ps_codec->as_process[i].pu1_slice_idx = pu1_buf_ping; } } @@ -3862,25 +3756,9 @@ static WORD32 isvce_init_mem_rec(iv_obj_t *ps_codec_obj, void *pv_api_ip, void * /* pointer to storage space */ UWORD8 *pu1_buf = ps_mem_rec->pv_base; - /* total size of the mem record */ - WORD32 total_size = 0; - - /* size in bytes to mb core coding status of an entire frame */ - total_size = max_mb_cnt; - - /* add an additional 1 row of bytes to evade the special case of row 0 */ - total_size += max_mb_cols; - for(i = 0; i < MAX_PROCESS_CTXT; i++) { - if(i < MAX_PROCESS_CTXT / MAX_CTXT_SETS) - { - ps_codec->as_process[i].pu1_proc_map = pu1_buf + max_mb_cols; - } - else - { - ps_codec->as_process[i].pu1_proc_map = pu1_buf + total_size + max_mb_cols; - } + ps_codec->as_process[i].pu1_proc_map = pu1_buf + max_mb_cols; } } @@ -3903,14 +3781,7 @@ static WORD32 isvce_init_mem_rec(iv_obj_t *ps_codec_obj, void *pv_api_ip, void * for(i = 0; i < MAX_PROCESS_CTXT; i++) { - if(i < MAX_PROCESS_CTXT / MAX_CTXT_SETS) - { - ps_codec->as_process[i].pu1_deblk_map = pu1_buf + max_mb_cols; - } - else - { - ps_codec->as_process[i].pu1_deblk_map = pu1_buf + total_size + max_mb_cols; - } + ps_codec->as_process[i].pu1_deblk_map = pu1_buf + max_mb_cols; } } @@ -3919,25 +3790,9 @@ static WORD32 isvce_init_mem_rec(iv_obj_t *ps_codec_obj, void *pv_api_ip, void * /* pointer to storage space */ UWORD8 *pu1_buf = (UWORD8 *) ps_mem_rec->pv_base; - /* total size of the mem record */ - WORD32 total_size = 0; - - /* size in bytes to mb core coding status of an entire frame */ - total_size = max_mb_cnt; - - /* add an additional 1 row of bytes to evade the special case of row 0 */ - total_size += max_mb_cols; - for(i = 0; i < MAX_PROCESS_CTXT; i++) { - if(i < MAX_PROCESS_CTXT / MAX_CTXT_SETS) - { - ps_codec->as_process[i].pu1_me_map = pu1_buf + max_mb_cols; - } - else - { - ps_codec->as_process[i].pu1_me_map = pu1_buf + total_size + max_mb_cols; - } + ps_codec->as_process[i].pu1_me_map = pu1_buf + max_mb_cols; } } diff --git a/encoder/svc/isvce_encode.c b/encoder/svc/isvce_encode.c index 8c6aa11..7d57ced 100644 --- a/encoder/svc/isvce_encode.c +++ b/encoder/svc/isvce_encode.c @@ -111,39 +111,6 @@ /** ****************************************************************************** * -* @brief This function puts the current thread to sleep for a duration -* of sleep_us -* -* @par Description -* ithread_yield() method causes the calling thread to yield execution to -*another thread that is ready to run on the current processor. The operating -*system selects the thread to yield to. ithread_usleep blocks the current thread -*for the specified number of milliseconds. In other words, yield just says, end -*my timeslice prematurely, look around for other threads to run. If there is -*nothing better than me, continue. Sleep says I don't want to run for x -* milliseconds. Even if no other thread wants to run, don't make me run. -* -* @param[in] sleep_us -* thread sleep duration -* -* @returns error_status -* -****************************************************************************** -*/ -IH264E_ERROR_T isvce_wait_for_thread(UWORD32 sleep_us) -{ - /* yield thread */ - ithread_yield(); - - /* put thread to sleep */ - ithread_sleep(sleep_us); - - return IH264E_SUCCESS; -} - -/** -****************************************************************************** -* * @brief * Encodes in synchronous api mode * diff --git a/encoder/svc/isvce_function_selector_generic.c b/encoder/svc/isvce_function_selector_generic.c index 044bbeb..c3a48bf 100644 --- a/encoder/svc/isvce_function_selector_generic.c +++ b/encoder/svc/isvce_function_selector_generic.c @@ -265,14 +265,10 @@ void isvce_init_function_ptr_generic(isvce_codec_t *ps_codec) ps_codec->pf_compute_sad_16x8 = ime_compute_sad_16x8; /* memory handling operations */ - ps_mem_fxns->pf_mem_cpy = ih264_memcpy; ps_mem_fxns->pf_mem_cpy_mul8 = ih264_memcpy_mul_8; - ps_mem_fxns->pf_mem_set = ih264_memset; ps_mem_fxns->pf_mem_set_mul8 = ih264_memset_mul_8; ps_mem_fxns->pf_copy_2d = isvc_copy_2d; ps_mem_fxns->pf_memset_2d = isvc_memset_2d; - ps_mem_fxns->pf_16bit_interleaved_copy = isvc_16bit_interleaved_copy; - ps_mem_fxns->pf_16bit_interleaved_memset = isvc_16bit_interleaved_memset; ps_mem_fxns->pf_nonzero_checker = isvc_is_nonzero_blk; /* sad me level functions */ diff --git a/encoder/svc/isvce_ilp_mv.c b/encoder/svc/isvce_ilp_mv.c index 9aa45a3..e9a7c17 100644 --- a/encoder/svc/isvce_ilp_mv.c +++ b/encoder/svc/isvce_ilp_mv.c @@ -204,7 +204,7 @@ void isvce_ilp_mv_ctxt_init(isvce_codec_t *ps_codec, iv_mem_rec_t *ps_mem_rec) if(u1_num_spatial_layers > 1) { - ilp_mv_layer_state_t *ps_layer_states; + ilp_mv_layer_state_t *ps_layer_states = NULL; ilp_mv_mb_state_t *aps_luma_mb_states[MAX_NUM_SPATIAL_LAYERS]; DOUBLE d_spatial_res_ratio = ps_codec->s_cfg.s_svc_params.d_spatial_res_ratio; diff --git a/encoder/svc/isvce_structs.h b/encoder/svc/isvce_structs.h index 9d8f6c5..68619bc 100644 --- a/encoder/svc/isvce_structs.h +++ b/encoder/svc/isvce_structs.h @@ -1958,10 +1958,6 @@ typedef struct inter_pred_fxns_t typedef struct mem_fxns_t { - FT_MEMCPY *pf_mem_cpy; - - FT_MEMSET *pf_mem_set; - FT_MEMCPY *pf_mem_cpy_mul8; FT_MEMSET *pf_mem_set_mul8; @@ -1970,10 +1966,6 @@ typedef struct mem_fxns_t FT_MEMSET_2D *pf_memset_2d; - FT_16BIT_INTERLEAVED_COPY *pf_16bit_interleaved_copy; - - FT_16BIT_INTERLEAVED_MEMSET *pf_16bit_interleaved_memset; - FT_NONZERO_CHECKER *pf_nonzero_checker; } mem_fxns_t; diff --git a/encoder/svc/isvce_utils.c b/encoder/svc/isvce_utils.c index 485a52e..81ec2c0 100644 --- a/encoder/svc/isvce_utils.c +++ b/encoder/svc/isvce_utils.c @@ -528,7 +528,7 @@ WORD32 isvce_svc_inp_params_validate(isvce_init_ip_t *ps_ip, isvce_cfg_params_t * ******************************************************************************* */ -WORD32 isvce_svc_rc_params_validate(isvce_cfg_params_t *ps_cfg) +static WORD32 isvce_svc_rc_params_validate(isvce_cfg_params_t *ps_cfg) { WORD32 i; @@ -3449,6 +3449,82 @@ void isvce_derive_nghbr_avbl_of_mbs(isvce_process_ctxt_t *ps_proc) } } +static IH264E_ERROR_T isvce_rc_init_wrapper(isvce_codec_t *ps_codec) +{ + WORD32 i; + + isvce_cfg_params_t *ps_cfg = &ps_codec->s_cfg; + + IH264E_ERROR_T err = isvce_svc_rc_params_validate(ps_cfg); + + if(IH264E_SUCCESS != err) + { + return err; + } + + for(i = 0; i < ps_cfg->s_svc_params.u1_num_spatial_layers; i++) + { + UWORD8 au1_init_qp[MAX_PIC_TYPE]; + UWORD8 au1_min_max_qp[2 * MAX_PIC_TYPE]; + + au1_init_qp[0] = gau1_h264_to_mpeg2_qmap[ps_cfg->au4_i_qp[i]]; + au1_init_qp[1] = gau1_h264_to_mpeg2_qmap[ps_cfg->au4_p_qp[i]]; + au1_init_qp[2] = gau1_h264_to_mpeg2_qmap[ps_cfg->au4_b_qp[i]]; + + au1_min_max_qp[2 * I_PIC] = gau1_h264_to_mpeg2_qmap[ps_cfg->au4_i_qp_min[i]]; + au1_min_max_qp[2 * I_PIC + 1] = gau1_h264_to_mpeg2_qmap[ps_cfg->au4_i_qp_max[i]]; + + au1_min_max_qp[2 * P_PIC] = gau1_h264_to_mpeg2_qmap[ps_cfg->au4_p_qp_min[i]]; + au1_min_max_qp[2 * P_PIC + 1] = gau1_h264_to_mpeg2_qmap[ps_cfg->au4_p_qp_max[i]]; + + au1_min_max_qp[2 * B_PIC] = gau1_h264_to_mpeg2_qmap[ps_cfg->au4_b_qp_min[i]]; + au1_min_max_qp[2 * B_PIC + 1] = gau1_h264_to_mpeg2_qmap[ps_cfg->au4_b_qp_max[i]]; + + switch(ps_cfg->e_rc_mode) + { + case IVE_RC_STORAGE: + { + ps_codec->s_rate_control.e_rc_type = VBR_STORAGE; + break; + } + case IVE_RC_CBR_NON_LOW_DELAY: + { + ps_codec->s_rate_control.e_rc_type = CBR_NLDRC; + break; + } + case IVE_RC_CBR_LOW_DELAY: + { + ps_codec->s_rate_control.e_rc_type = CBR_LDRC; + break; + } + case IVE_RC_NONE: + { + ps_codec->s_rate_control.e_rc_type = CONST_QP; + break; + } + default: + { + break; + } + } + + for(i = 0; i < MAX_NUM_SPATIAL_LAYERS; i++) + { + isvce_rc_init(ps_codec->s_rate_control.apps_rate_control_api[i], + ps_codec->s_rate_control.pps_frame_time, + ps_codec->s_rate_control.pps_time_stamp, + ps_codec->s_rate_control.pps_pd_frm_rate, ps_cfg->u4_max_framerate, + ps_cfg->u4_src_frame_rate, ps_cfg->u4_tgt_frame_rate, + ps_codec->s_rate_control.e_rc_type, ps_cfg->au4_target_bitrate[i], + ps_cfg->au4_max_bitrate[i], ps_cfg->au4_vbv_buffer_delay[i], + ps_cfg->u4_i_frm_interval, ps_cfg->u4_num_bframes + 1, au1_init_qp, + ps_cfg->u4_num_bframes + 2, au1_min_max_qp, ps_cfg->u4_max_level); + } + } + + return IH264E_SUCCESS; +} + /** ******************************************************************************* * @@ -3576,7 +3652,7 @@ IH264E_ERROR_T isvce_codec_init(isvce_codec_t *ps_codec) } { - WORD32 i4_err_code = isvce_svc_rc_params_validate(&ps_codec->s_cfg); + WORD32 i4_err_code = isvce_rc_init_wrapper(ps_codec); if(IH264E_SUCCESS != i4_err_code) { @@ -3584,82 +3660,6 @@ IH264E_ERROR_T isvce_codec_init(isvce_codec_t *ps_codec) } } - /****************************************************/ - /* INITIALIZE RATE CONTROL */ - /****************************************************/ - { - for(i = 0; i < MAX_NUM_SPATIAL_LAYERS; i++) - { - UWORD8 au1_init_qp[MAX_PIC_TYPE]; - UWORD8 au1_min_max_qp[2 * MAX_PIC_TYPE]; - UWORD8 au1_min_max_avc_qp[2 * MAX_PIC_TYPE]; - - /* update rc lib with modified qp */ - au1_init_qp[0] = gau1_h264_to_mpeg2_qmap[ps_codec->s_cfg.au4_i_qp[i]]; - au1_init_qp[1] = gau1_h264_to_mpeg2_qmap[ps_codec->s_cfg.au4_p_qp[i]]; - au1_init_qp[2] = gau1_h264_to_mpeg2_qmap[ps_codec->s_cfg.au4_b_qp[i]]; - - au1_min_max_qp[2 * I_PIC] = gau1_h264_to_mpeg2_qmap[ps_codec->s_cfg.au4_i_qp_min[i]]; - au1_min_max_qp[2 * I_PIC + 1] = - gau1_h264_to_mpeg2_qmap[ps_codec->s_cfg.au4_i_qp_max[i]]; - - au1_min_max_qp[2 * P_PIC] = gau1_h264_to_mpeg2_qmap[ps_codec->s_cfg.au4_p_qp_min[i]]; - au1_min_max_qp[2 * P_PIC + 1] = - gau1_h264_to_mpeg2_qmap[ps_codec->s_cfg.au4_p_qp_max[i]]; - - au1_min_max_qp[2 * B_PIC] = gau1_h264_to_mpeg2_qmap[ps_codec->s_cfg.au4_b_qp_min[i]]; - au1_min_max_qp[2 * B_PIC + 1] = - gau1_h264_to_mpeg2_qmap[ps_codec->s_cfg.au4_b_qp_max[i]]; - - /* get rc mode */ - switch(ps_codec->s_cfg.e_rc_mode) - { - case IVE_RC_STORAGE: - ps_codec->s_rate_control.e_rc_type = VBR_STORAGE; - break; - case IVE_RC_CBR_NON_LOW_DELAY: - ps_codec->s_rate_control.e_rc_type = CBR_NLDRC; - break; - case IVE_RC_CBR_LOW_DELAY: - ps_codec->s_rate_control.e_rc_type = CBR_LDRC; - break; - case IVE_RC_NONE: - ps_codec->s_rate_control.e_rc_type = CONST_QP; - break; - default: - break; - } - - ps_codec->u1_enable_init_qp = DEFAULT_INIT_QP; - - /* init rate control */ - isvce_rc_init( - ps_codec->s_rate_control.apps_rate_control_api[i], - ps_codec->s_rate_control.pps_frame_time, ps_codec->s_rate_control.pps_time_stamp, - ps_codec->s_rate_control.pps_pd_frm_rate, ps_codec->s_cfg.u4_max_framerate, - ps_codec->s_cfg.u4_src_frame_rate, ps_codec->s_cfg.u4_tgt_frame_rate, - ps_codec->s_rate_control.e_rc_type, ps_codec->s_cfg.au4_target_bitrate[i], - ps_codec->s_cfg.au4_max_bitrate[i], ps_codec->s_cfg.au4_vbv_buffer_delay[i], - ps_codec->s_cfg.u4_i_frm_interval, ps_codec->s_cfg.u4_num_bframes + 1, au1_init_qp, - ps_codec->s_cfg.u4_num_bframes + 2, au1_min_max_qp, - MAX(ps_codec->s_cfg.u4_max_level, - (UWORD32) ih264e_get_min_level(ps_codec->s_cfg.u4_max_wd, - ps_codec->s_cfg.u4_max_ht))); - - au1_min_max_avc_qp[2 * I_PIC] = ps_codec->s_cfg.au4_i_qp_min[i]; - au1_min_max_avc_qp[2 * I_PIC + 1] = ps_codec->s_cfg.au4_i_qp_max[i]; - - au1_min_max_avc_qp[2 * P_PIC] = ps_codec->s_cfg.au4_p_qp_min[i]; - au1_min_max_avc_qp[2 * P_PIC + 1] = ps_codec->s_cfg.au4_p_qp_max[i]; - - au1_min_max_avc_qp[2 * B_PIC] = ps_codec->s_cfg.au4_b_qp_min[i]; - au1_min_max_avc_qp[2 * B_PIC + 1] = ps_codec->s_cfg.au4_b_qp_max[i]; - - irc_change_qp_constraints(ps_codec->s_rate_control.apps_rate_control_api[i], - au1_min_max_qp, au1_min_max_avc_qp); - } - } - /* recon stride */ ps_codec->i4_rec_strd = ALIGN16(ps_codec->s_cfg.u4_max_wd) + PAD_WD; @@ -3774,7 +3774,6 @@ IH264E_ERROR_T isvce_codec_update_config(isvce_codec_t *ps_codec, isvce_cfg_para } else if(ps_cfg->e_cmd == ISVCE_CMD_CTL_SET_FRAMERATE) { - /* temp var */ UWORD32 u4_src_ticks, u4_tgt_ticks; u4_src_ticks = ih264e_frame_time_get_src_ticks(ps_codec->s_rate_control.pps_frame_time); @@ -4122,71 +4121,7 @@ IH264E_ERROR_T isvce_codec_update_config(isvce_codec_t *ps_codec, isvce_cfg_para /* reset RC model */ if(u4_init_rc) { - for(i = 0; i < ps_cfg->s_svc_params.u1_num_spatial_layers; i++) - { - /* init qp */ - UWORD8 au1_init_qp[MAX_PIC_TYPE]; - - /* min max qp */ - UWORD8 au1_min_max_qp[2 * MAX_PIC_TYPE]; - - /* init i,p,b qp */ - au1_init_qp[0] = gau1_h264_to_mpeg2_qmap[ps_codec->s_cfg.au4_i_qp[i]]; - au1_init_qp[1] = gau1_h264_to_mpeg2_qmap[ps_codec->s_cfg.au4_p_qp[i]]; - au1_init_qp[2] = gau1_h264_to_mpeg2_qmap[ps_codec->s_cfg.au4_b_qp[i]]; - - /* init min max qp */ - au1_min_max_qp[2 * I_PIC] = gau1_h264_to_mpeg2_qmap[ps_codec->s_cfg.au4_i_qp_min[i]]; - au1_min_max_qp[2 * I_PIC + 1] = - gau1_h264_to_mpeg2_qmap[ps_codec->s_cfg.au4_i_qp_max[i]]; - - au1_min_max_qp[2 * P_PIC] = gau1_h264_to_mpeg2_qmap[ps_codec->s_cfg.au4_p_qp_min[i]]; - au1_min_max_qp[2 * P_PIC + 1] = - gau1_h264_to_mpeg2_qmap[ps_codec->s_cfg.au4_p_qp_max[i]]; - - au1_min_max_qp[2 * B_PIC] = gau1_h264_to_mpeg2_qmap[ps_codec->s_cfg.au4_b_qp_min[i]]; - au1_min_max_qp[2 * B_PIC + 1] = - gau1_h264_to_mpeg2_qmap[ps_codec->s_cfg.au4_b_qp_max[i]]; - - /* get rc mode */ - switch(ps_codec->s_cfg.e_rc_mode) - { - case IVE_RC_STORAGE: - ps_codec->s_rate_control.e_rc_type = VBR_STORAGE; - break; - - case IVE_RC_CBR_NON_LOW_DELAY: - ps_codec->s_rate_control.e_rc_type = CBR_NLDRC; - break; - - case IVE_RC_CBR_LOW_DELAY: - ps_codec->s_rate_control.e_rc_type = CBR_LDRC; - break; - - case IVE_RC_NONE: - ps_codec->s_rate_control.e_rc_type = CONST_QP; - break; - - default: - break; - } - - /* init rate control */ - for(i = 0; i < MAX_NUM_SPATIAL_LAYERS; i++) - { - isvce_rc_init( - ps_codec->s_rate_control.apps_rate_control_api[i], - ps_codec->s_rate_control.pps_frame_time, - ps_codec->s_rate_control.pps_time_stamp, - ps_codec->s_rate_control.pps_pd_frm_rate, ps_codec->s_cfg.u4_max_framerate, - ps_codec->s_cfg.u4_src_frame_rate, ps_codec->s_cfg.u4_tgt_frame_rate, - ps_codec->s_rate_control.e_rc_type, ps_codec->s_cfg.au4_target_bitrate[i], - ps_codec->s_cfg.au4_max_bitrate[i], ps_codec->s_cfg.au4_vbv_buffer_delay[i], - ps_codec->s_cfg.u4_i_frm_interval, ps_codec->s_cfg.u4_num_bframes + 1, - au1_init_qp, ps_codec->s_cfg.u4_num_bframes + 2, au1_min_max_qp, - ps_codec->s_cfg.u4_max_level); - } - } + err = isvce_rc_init_wrapper(ps_codec); } return err; diff --git a/encoder/svc/isvce_utils.h b/encoder/svc/isvce_utils.h index 497809e..ad14446 100644 --- a/encoder/svc/isvce_utils.h +++ b/encoder/svc/isvce_utils.h @@ -180,8 +180,6 @@ extern WORD32 isvce_svc_au_props_validate(svc_inp_params_t *ps_svc_inp_params, U extern WORD32 isvce_svc_inp_params_validate(isvce_init_ip_t *ps_ip, isvce_cfg_params_t *ps_cfg); -extern WORD32 isvce_svc_rc_params_validate(isvce_cfg_params_t *ps_cfg); - extern WORD32 isvce_svc_frame_params_validate( rate_control_api_t *aps_rate_control_api[MAX_NUM_SPATIAL_LAYERS], UWORD8 u1_num_spatial_layers); diff --git a/encoder/x86/svc/isvce_function_selector_sse42.c b/encoder/x86/svc/isvce_function_selector_sse42.c index 709155f..03f1cf0 100644 --- a/encoder/x86/svc/isvce_function_selector_sse42.c +++ b/encoder/x86/svc/isvce_function_selector_sse42.c @@ -128,20 +128,13 @@ void isvce_init_function_ptr_sse42(isvce_codec_t *ps_codec) isvc_iquant_itrans_recon_res_chroma_4x4_sse42; ps_enc_loop_fxns->apf_iquant_itrans_recon_chroma_4x4[1] = isvc_iquant_itrans_recon_res_chroma_4x4_with_res_acc_sse42; - ps_enc_loop_fxns->apf_iquant_itrans_recon_chroma_4x4[2] = - isvc_iquant_itrans_recon_chroma_4x4_sse42; - ps_enc_loop_fxns->apf_iquant_itrans_recon_4x4_dc[0] = isvc_iquant_itrans_recon_res_dc_4x4_sse42; - ps_enc_loop_fxns->apf_iquant_itrans_recon_4x4_dc[1] = - isvc_iquant_itrans_recon_res_dc_with_res_acc_4x4_sse42; ps_enc_loop_fxns->apf_iquant_itrans_recon_4x4_dc[2] = isvc_iquant_itrans_recon_dc_4x4_sse42; ps_enc_loop_fxns->apf_iquant_itrans_recon_chroma_4x4_dc[0] = isvc_iquant_itrans_recon_res_chroma_4x4_dc_sse42; ps_enc_loop_fxns->apf_iquant_itrans_recon_chroma_4x4_dc[1] = isvc_iquant_itrans_recon_res_chroma_4x4_dc_with_res_acc_sse42; - ps_enc_loop_fxns->apf_iquant_itrans_recon_chroma_4x4_dc[2] = - isvc_iquant_itrans_recon_chroma_4x4_dc_sse42; ps_enc_loop_fxns->pf_ihadamard_scaling_4x4 = ih264_ihadamard_scaling_4x4_sse42; |