diff options
author | Zhao Yakui <yakui.zhao@intel.com> | 2012-12-24 15:08:39 +0800 |
---|---|---|
committer | Xiang, Haihao <haihao.xiang@intel.com> | 2013-01-17 13:08:39 +0800 |
commit | 788f649bc66b773e3af1e23870431c19c2c1136d (patch) | |
tree | b48c3c76a2fad4ab543a894678d46135a206cefd /src/shaders | |
parent | cfd70d7afcc22e28bfce8805d7de1e5f759fed8b (diff) |
Add the MV prediction to optimize VME parameter on Haswell
Signed-off-by: Zhao Yakui <yakui.zhao@intel.com>
Diffstat (limited to 'src/shaders')
-rw-r--r-- | src/shaders/vme/inter_frame_haswell.asm | 338 | ||||
-rw-r--r-- | src/shaders/vme/inter_frame_haswell.g75b | 150 | ||||
-rw-r--r-- | src/shaders/vme/vme75.inc | 44 |
3 files changed, 532 insertions, 0 deletions
diff --git a/src/shaders/vme/inter_frame_haswell.asm b/src/shaders/vme/inter_frame_haswell.asm index 5bb8ba7..36b394a 100644 --- a/src/shaders/vme/inter_frame_haswell.asm +++ b/src/shaders/vme/inter_frame_haswell.asm @@ -15,6 +15,9 @@ // Now, begin source code.... // +#define SAVE_RET add (1) RETURN_REG<1>:ud ip:ud 32:ud +#define RETURN mov (1) ip:ud RETURN_REG<0,1,0>:ud + /* * __START */ @@ -73,6 +76,294 @@ mov (1) read1_header.8<1>:UD BLOCK_8X4 {align1}; mov (8) msg_reg0.0<1>:UD read1_header.0<8,8,1>:UD {align1}; send (8) msg_ind CHROMA_COL<1>:UB null read(BIND_IDX_CBCR, 0, 0, 4) mlen 1 rlen 1 {align1}; +mov (8) mb_mvp_ref.0<1>:ud 0:ud {align1}; +and.z.f0.0 (1) null:uw mb_hwdep<0,1,0>:uw 0x04:uw {align1}; +(f0.0) jmpi (1) __mb_hwdep_end; +/* read back the data for MB A */ +/* the layout of MB result is: rx.0(Available). rx.4(MVa), rX.8(MVb), rX.16(Pred_L0 flag), +* rX.18 (Pred_L1 flag), rX.20(Forward reference ID), rX.22(Backwared reference ID) +*/ +mov (8) mba_result.0<1>:ud 0x0:ud {align1}; +mov (8) mbb_result.0<1>:ud 0x0:ud {align1}; +mov (8) mbc_result.0<1>:ud 0x0:ud {align1}; +mba_start: +mov (8) mb_msg0.0<1>:ud 0:ud {align1}; +and.z.f0.0 (1) null:uw input_mb_intra_ub<0,1,0>:ub INTRA_PRED_AVAIL_FLAG_AE:uw {align1}; +/* MB A doesn't exist. Zero MV. mba_flag is zero and ref ID = -1 */ +(f0.0) mov (2) mba_result.20<1>:w -1:w {align1}; +(f0.0) jmpi (1) mbb_start; +mov (1) mba_result.0<1>:d MB_AVAIL {align1}; +mov (2) tmp_reg0.0<1>:UW orig_xy_ub<2,2,1>:UB {align1}; +add (1) tmp_reg0.0<1>:w tmp_reg0.0<0,1,0>:w -1:w {align1}; +mul (1) mb_msg0.8<1>:UD w_in_mb_uw<0,1,0>:UW tmp_reg0.2<0,1,0>:UW {align1}; +add (1) mb_msg0.8<1>:UD mb_msg0.8<0,1,0>:UD tmp_reg0.0<0,1,0>:uw {align1}; +mul (1) mb_msg0.8<1>:UD mb_msg0.8<0,1,0>:UD 24:UD {align1}; +mov (1) mb_msg0.20<1>:UB thread_id_ub {align1}; /* dispatch id */ + +/* bind index 3, read 4 oword (64bytes), msg type: 0(OWord Block Read) */ +send (16) + mb_ind + mb_wb.0<1>:ud + NULL + data_port( + OBR_CACHE_TYPE, + OBR_MESSAGE_TYPE, + OBR_CONTROL_4, + OBR_BIND_IDX, + OBR_WRITE_COMMIT_CATEGORY, + OBR_HEADER_PRESENT + ) + mlen 1 + rlen 2 + {align1}; + +/* TODO: RefID is required after multi-references are added */ +cmp.l.f0.0 (1) null:w mb_intra_wb.16<0,1,0>:uw mb_inter_wb.8<0,1,0>:uw {align1}; +(f0.0) mov (2) mba_result.20<1>:w -1:w {align1}; +(f0.0) jmpi (1) mbb_start; + +add (1) mb_msg0.8<1>:UD mb_msg0.8<0,1,0>:ud 3:ud {align1}; +/* Read MV for MB A */ +/* bind index 3, read 8 oword (128bytes), msg type: 0(OWord Block Read) */ +send (16) + mb_ind + mb_mv0.0<1>:ud + NULL + data_port( + OBR_CACHE_TYPE, + OBR_MESSAGE_TYPE, + OBR_CONTROL_8, + OBR_BIND_IDX, + OBR_WRITE_COMMIT_CATEGORY, + OBR_HEADER_PRESENT + ) + mlen 1 + rlen 4 + {align1}; +/* TODO: RefID is required after multi-references are added */ +/* MV */ +mov (2) mba_result.4<1>:ud mb_mv1.8<2,2,1>:ud {align1}; +mov (1) mba_result.16<1>:w MB_PRED_FLAG {align1}; + +mbb_start: +mov (8) mb_msg0.0<1>:ud 0:ud {align1}; +and.z.f0.0 (1) null:uw input_mb_intra_ub<0,1,0>:ub INTRA_PRED_AVAIL_FLAG_B:uw {align1}; +/* MB B doesn't exist. Zero MV. mba_flag is zero */ +/* If MB B doesn't exist, neight of MB C nor D exists */ +(f0.0) mov (2) mbb_result.20<1>:w -1:w {align1}; +(f0.0) mov (2) mbc_result.20<1>:w -1:w {align1}; +(f0.0) jmpi (1) mb_mvp_start; +mov (1) mbb_result.0<1>:d MB_AVAIL {align1}; +mov (2) tmp_reg0.0<1>:UW orig_xy_ub<2,2,1>:UB {align1}; +add (1) tmp_reg0.2<1>:w tmp_reg0.2<0,1,0>:w -1:w {align1}; +mul (1) mb_msg0.8<1>:UD w_in_mb_uw<0,1,0>:UW tmp_reg0.2<0,1,0>:UW {align1}; +add (1) mb_msg0.8<1>:UD mb_msg0.8<0,1,0>:UD tmp_reg0.0<0,1,0>:uw {align1}; +mul (1) mb_msg0.8<1>:UD mb_msg0.8<0,1,0>:UD 24:UD {align1}; +mov (1) mb_msg0.20<1>:UB thread_id_ub {align1}; /* dispatch id */ + +/* bind index 3, read 4 oword (64bytes), msg type: 0(OWord Block Read) */ +send (16) + mb_ind + mb_wb.0<1>:ud + NULL + data_port( + OBR_CACHE_TYPE, + OBR_MESSAGE_TYPE, + OBR_CONTROL_4, + OBR_BIND_IDX, + OBR_WRITE_COMMIT_CATEGORY, + OBR_HEADER_PRESENT + ) + mlen 1 + rlen 2 + {align1}; + +/* TODO: RefID is required after multi-references are added */ +cmp.l.f0.0 (1) null:w mb_intra_wb.16<0,1,0>:uw mb_inter_wb.8<0,1,0>:uw {align1}; +(f0.0) mov (2) mbb_result.20<1>:w -1:w {align1}; +(f0.0) jmpi (1) mbc_start; +add (1) mb_msg0.8<1>:UD mb_msg0.8<0,1,0>:ud 3:ud {align1}; +/* Read MV for MB B */ +/* bind index 3, read 8 oword (128bytes), msg type: 0(OWord Block Read) */ +send (16) + mb_ind + mb_mv0.0<1>:ud + NULL + data_port( + OBR_CACHE_TYPE, + OBR_MESSAGE_TYPE, + OBR_CONTROL_8, + OBR_BIND_IDX, + OBR_WRITE_COMMIT_CATEGORY, + OBR_HEADER_PRESENT + ) + mlen 1 + rlen 4 + {align1}; +/* TODO: RefID is required after multi-references are added */ +mov (2) mbb_result.4<1>:ud mb_mv2.16<2,2,1>:ud {align1}; +mov (1) mbb_result.16<1>:w MB_PRED_FLAG {align1}; + +mbc_start: +mov (8) mb_msg0.0<1>:ud 0:ud {align1}; +and.z.f0.0 (1) null:uw input_mb_intra_ub<0,1,0>:ub INTRA_PRED_AVAIL_FLAG_C:uw {align1}; +/* MB C doesn't exist. Zero MV. mba_flag is zero */ +/* Based on h264 spec the MB D will be replaced if MB C doesn't exist */ +(f0.0) jmpi (1) mbd_start; +mov (1) mbc_result.0<1>:d MB_AVAIL {align1}; +mov (2) tmp_reg0.0<1>:UW orig_xy_ub<2,2,1>:UB {align1}; +add (1) tmp_reg0.2<1>:w tmp_reg0.2<0,1,0>:w -1:w {align1}; +add (1) tmp_reg0.0<1>:w tmp_reg0.0<0,1,0>:w 1:w {align1}; +mul (1) mb_msg0.8<1>:UD w_in_mb_uw<0,1,0>:UW tmp_reg0.2<0,1,0>:UW {align1}; +add (1) mb_msg0.8<1>:UD mb_msg0.8<0,1,0>:UD tmp_reg0.0<0,1,0>:uw {align1}; +mul (1) mb_msg0.8<1>:UD mb_msg0.8<0,1,0>:UD 24:UD {align1}; +mov (1) mb_msg0.20<1>:UB thread_id_ub {align1}; /* dispatch id */ + +/* bind index 3, read 4 oword (64bytes), msg type: 0(OWord Block Read) */ +send (16) + mb_ind + mb_wb.0<1>:ud + NULL + data_port( + OBR_CACHE_TYPE, + OBR_MESSAGE_TYPE, + OBR_CONTROL_4, + OBR_BIND_IDX, + OBR_WRITE_COMMIT_CATEGORY, + OBR_HEADER_PRESENT + ) + mlen 1 + rlen 2 + {align1}; + +/* TODO: RefID is required after multi-references are added */ +cmp.l.f0.0 (1) null:w mb_intra_wb.16<0,1,0>:uw mb_inter_wb.8<0,1,0>:uw {align1}; +(f0.0) mov (2) mbc_result.20<1>:w -1:w {align1}; +(f0.0) jmpi (1) mb_mvp_start; +add (1) mb_msg0.8<1>:UD mb_msg0.8<0,1,0>:ud 3:ud {align1}; +/* Read MV for MB C */ +/* bind index 3, read 8 oword (128bytes), msg type: 0(OWord Block Read) */ +send (16) + mb_ind + mb_mv0.0<1>:ud + NULL + data_port( + OBR_CACHE_TYPE, + OBR_MESSAGE_TYPE, + OBR_CONTROL_8, + OBR_BIND_IDX, + OBR_WRITE_COMMIT_CATEGORY, + OBR_HEADER_PRESENT + ) + mlen 1 + rlen 4 + {align1}; +/* TODO: RefID is required after multi-references are added */ +/* Forward MV */ +mov (2) mbc_result.4<1>:ud mb_mv2.16<2,2,1>:ud {align1}; +mov (1) mbc_result.16<1>:w MB_PRED_FLAG {align1}; + +jmpi (1) mb_mvp_start; +mbd_start: +mov (1) mbc_result.0<1>:d MB_AVAIL {align1}; +mov (2) tmp_reg0.0<1>:UW orig_xy_ub<2,2,1>:UB {align1}; +add (2) tmp_reg0.0<1>:w tmp_reg0.0<2,2,1>:w -1:w {align1}; +mul (1) mb_msg0.8<1>:UD w_in_mb_uw<0,1,0>:UW tmp_reg0.2<0,1,0>:UW {align1}; +add (1) mb_msg0.8<1>:UD mb_msg0.8<0,1,0>:UD tmp_reg0.0<0,1,0>:uw {align1}; +mul (1) mb_msg0.8<1>:UD mb_msg0.8<0,1,0>:UD 24:UD {align1}; +mov (1) mb_msg0.20<1>:UB thread_id_ub {align1}; /* dispatch id */ + +/* bind index 3, read 4 oword (64bytes), msg type: 0(OWord Block Read) */ +send (16) + mb_ind + mb_wb.0<1>:ud + NULL + data_port( + OBR_CACHE_TYPE, + OBR_MESSAGE_TYPE, + OBR_CONTROL_4, + OBR_BIND_IDX, + OBR_WRITE_COMMIT_CATEGORY, + OBR_HEADER_PRESENT + ) + mlen 1 + rlen 2 + {align1}; + +cmp.l.f0.0 (1) null:w mb_intra_wb.16<0,1,0>:uw mb_inter_wb.8<0,1,0>:uw {align1}; +(f0.0) mov (2) mbc_result.20<1>:w -1:w {align1}; +(f0.0) jmpi (1) mb_mvp_start; + +add (1) mb_msg0.8<1>:UD mb_msg0.8<0,1,0>:ud 3:ud {align1}; +/* Read MV for MB D */ +/* bind index 3, read 8 oword (128bytes), msg type: 0(OWord Block Read) */ +send (16) + mb_ind + mb_mv0.0<1>:ub + NULL + data_port( + OBR_CACHE_TYPE, + OBR_MESSAGE_TYPE, + OBR_CONTROL_8, + OBR_BIND_IDX, + OBR_WRITE_COMMIT_CATEGORY, + OBR_HEADER_PRESENT + ) + mlen 1 + rlen 4 + {align1}; + +/* TODO: RefID is required after multi-references are added */ + +/* Forward MV */ +mov (2) mbc_result.4<1>:ud mb_mv3.24<2,2,1>:ud {align1}; +mov (1) mbc_result.18<1>:w MB_PRED_FLAG {align1}; + +mb_mvp_start: +/*TODO: Add the skip prediction */ +/* Check whether both MB and C are invailable */ +add (1) tmp_reg0.0<1>:d mbb_result.0<0,1,0>:d mbc_result.0<0,1,0>:d {align1}; +cmp.z.f0.0 (1) null:d tmp_reg0.0<0,1,0>:d 0:d {align1}; +(-f0.0) jmpi (1) mb_median_start; +cmp.nz.f0.0 (1) null:d mba_result.0<0,1,0>:d 1:d {align1}; +(f0.0) mov (1) mbb_result.4<1>:ud mba_result.4<0,1,0>:ud {align1}; +(f0.0) mov (1) mbc_result.4<1>:ud mba_result.4<0,1,0>:ud {align1}; +(f0.0) mov (1) mbb_result.20<1>:uw mba_result.20<0,1,0>:uw {align1}; +(f0.0) mov (1) mbc_result.20<1>:uw mba_result.20<0,1,0>:uw {align1}; +(f0.0) mov (1) mb_mvp_ref.0<1>:ud mba_result.4<0,1,0>:ud {align1}; +(-f0.0) mov (1) mb_mvp_ref.0<1>:ud 0:ud {align1}; +jmpi (1) __mb_hwdep_end; + +mb_median_start: +/* check whether only one neighbour MB has the same ref ID with the current MB */ +mov (8) tmp_reg0.0<1>:ud 0:ud {align1}; +cmp.z.f0.0 (1) null:d mba_result.20<1>:w 0:w {align1}; +(f0.0) add (1) tmp_reg0.0<1>:w tmp_reg0.0<1>:w 1:w {align1}; +(f0.0) mov (1) tmp_reg0.4<1>:ud mba_result.4<0,1,0>:ud {align1}; +cmp.z.f0.0 (1) null:d mbb_result.20<1>:w 0:w {align1}; +(f0.0) add (1) tmp_reg0.0<1>:w tmp_reg0.0<1>:w 1:w {align1}; +(f0.0) mov (1) tmp_reg0.4<1>:ud mbb_result.4<0,1,0>:ud {align1}; +cmp.z.f0.0 (1) null:d mbc_result.20<1>:w 0:w {align1}; +(f0.0) add (1) tmp_reg0.0<1>:w tmp_reg0.0<1>:w 1:w {align1}; +(f0.0) mov (1) tmp_reg0.4<1>:ud mbc_result.4<0,1,0>:ud {align1}; +cmp.e.f0.0 (1) null:d tmp_reg0.0<1>:w 1:w {align1}; +(f0.0) mov (1) mb_mvp_ref.0<1>:ud tmp_reg0.4<0,1,0>:ud {align1}; +(f0.0) jmpi (1) __mb_hwdep_end; + +mov (1) INPUT_ARG0.0<1>:w mba_result.4<0,1,0>:w {align1}; +mov (1) INPUT_ARG0.4<1>:w mbb_result.4<0,1,0>:w {align1}; +mov (1) INPUT_ARG0.8<1>:w mbc_result.4<0,1,0>:w {align1}; +SAVE_RET {align1}; + jmpi (1) word_imedian; +mov (1) mb_mvp_ref.0<1>:w RET_ARG<0,1,0>:w {align1}; +mov (1) INPUT_ARG0.0<1>:w mba_result.6<0,1,0>:w {align1}; +mov (1) INPUT_ARG0.4<1>:w mbb_result.6<0,1,0>:w {align1}; +mov (1) INPUT_ARG0.8<1>:w mbc_result.6<0,1,0>:w {align1}; +SAVE_RET {align1}; +jmpi (1) word_imedian; +mov (1) mb_mvp_ref.2<1>:w RET_ARG<0,1,0>:w {align1}; + +__mb_hwdep_end: /* m2, get the MV/Mb cost passed from constant buffer when spawning thread by MEDIA_OBJECT */ mov (8) vme_m2<1>:UD r1.0<8,8,1>:UD {align1}; @@ -196,6 +487,9 @@ mov (1) vme_m1.0<1>:UD ADAPTIVE_SEARCH_ENABLE:ud {align1} ; /* the Max MV number is passed by constant buffer */ mov (1) vme_m1.4<1>:UB r4.28<0,1,0>:UB {align1}; mov (1) vme_m1.8<1>:UD START_CENTER + SEARCH_PATH_LEN:UD {align1}; +/* Set the MV cost center */ +mov (1) vme_m1.16<1>:ud mb_mvp_ref.0<0,1,0>:ud {align1}; +mov (1) vme_m1.20<1>:ud mb_mvp_ref.0<0,1,0>:ud {align1}; mov (8) vme_msg_1.0<1>:UD vme_m1.0<8,8,1>:UD {align1}; mov (8) vme_msg_2<1>:UD vme_m2.0<8,8,1>:UD {align1}; @@ -348,3 +642,47 @@ __EXIT: */ mov (8) ts_msg_reg0<1>:UD r0<8,8,1>:UD {align1}; send (16) ts_msg_ind acc0<1>UW null thread_spawner(0, 0, 1) mlen 1 rlen 0 {align1 EOT}; + + + nop ; + nop ; +/* Compare three word data to get the min value */ +word_imin: + cmp.le.f0.0 (1) null:w INPUT_ARG0.0<0,1,0>:w INPUT_ARG0.4<0,1,0>:w {align1}; + (f0.0) mov (1) TEMP_VAR0.0<1>:w INPUT_ARG0.0<0,1,0>:w {align1}; + (-f0.0) mov (1) TEMP_VAR0.0<1>:w INPUT_ARG0.4<0,1,0>:w {align1}; + cmp.le.f0.0 (1) null:w TEMP_VAR0.0<0,1,0>:w INPUT_ARG0.8<0,1,0>:w {align1}; + (f0.0) mov (1) RET_ARG<1>:w TEMP_VAR0.0<0,1,0>:w {align1}; + (-f0.0) mov (1) RET_ARG<1>:w INPUT_ARG0.8<0,1,0>:w {align1}; + RETURN {align1}; + +/* Compare three word data to get the max value */ +word_imax: + cmp.ge.f0.0 (1) null:w INPUT_ARG0.0<0,1,0>:w INPUT_ARG0.4<0,1,0>:w {align1}; + (f0.0) mov (1) TEMP_VAR0.0<1>:w INPUT_ARG0.0<0,1,0>:w {align1}; + (-f0.0) mov (1) TEMP_VAR0.0<1>:w INPUT_ARG0.4<0,1,0>:w {align1}; + cmp.ge.f0.0 (1) null:w TEMP_VAR0.0<0,1,0>:w INPUT_ARG0.8<0,1,0>:w {align1}; + (f0.0) mov (1) RET_ARG<1>:w TEMP_VAR0.0<0,1,0>:w {align1}; + (-f0.0) mov (1) RET_ARG<1>:w INPUT_ARG0.8<0,1,0>:w {align1}; + RETURN {align1}; + +word_imedian: + cmp.ge.f0.0 (1) null:w INPUT_ARG0.0<0,1,0>:w INPUT_ARG0.4<0,1,0>:w {align1}; + (f0.0) jmpi (1) cmp_a_ge_b; + cmp.ge.f0.0 (1) null:w INPUT_ARG0.0<0,1,0>:w INPUT_ARG0.8<0,1,0>:w {align1}; + (f0.0) mov (1) RET_ARG<1>:w INPUT_ARG0.0<0,1,0>:w {align1}; + (f0.0) jmpi (1) cmp_end; + cmp.ge.f0.0 (1) null:w INPUT_ARG0.4<0,1,0>:w INPUT_ARG0.8<0,1,0>:w {align1}; + (f0.0) mov (1) RET_ARG<1>:w INPUT_ARG0.8<0,1,0>:w {align1}; + (-f0.0) mov (1) RET_ARG<1>:w INPUT_ARG0.4<0,1,0>:w {align1}; + jmpi (1) cmp_end; +cmp_a_ge_b: + cmp.ge.f0.0 (1) null:w INPUT_ARG0.4<0,1,0>:w INPUT_ARG0.8<0,1,0>:w {align1}; + (f0.0) mov (1) RET_ARG<1>:w INPUT_ARG0.4<0,1,0>:w {align1}; + (f0.0) jmpi (1) cmp_end; + cmp.ge.f0.0 (1) null:w INPUT_ARG0.0<0,1,0>:w INPUT_ARG0.8<0,1,0>:w {align1}; + (f0.0) mov (1) RET_ARG<1>:w INPUT_ARG0.8<0,1,0>:w {align1}; + (-f0.0) mov (1) RET_ARG<1>:w INPUT_ARG0.0<0,1,0>:w {align1}; +cmp_end: + RETURN {align1}; + diff --git a/src/shaders/vme/inter_frame_haswell.g75b b/src/shaders/vme/inter_frame_haswell.g75b index 5ae0aad..1ef526c 100644 --- a/src/shaders/vme/inter_frame_haswell.g75b +++ b/src/shaders/vme/inter_frame_haswell.g75b @@ -33,6 +33,122 @@ { 0x00000001, 0x242800e1, 0x00000000, 0x00070003 }, { 0x00600001, 0x28000021, 0x008d0420, 0x00000000 }, { 0x04600031, 0x26201cb1, 0x00000800, 0x02190006 }, + { 0x00600001, 0x2ac00061, 0x00000000, 0x00000000 }, + { 0x01000005, 0x20002d28, 0x000000a6, 0x00040004 }, + { 0x00010020, 0x34001c00, 0x00001400, 0x00000710 }, + { 0x00600001, 0x2ae00061, 0x00000000, 0x00000000 }, + { 0x00600001, 0x2b000061, 0x00000000, 0x00000000 }, + { 0x00600001, 0x2b200061, 0x00000000, 0x00000000 }, + { 0x00600001, 0x2b400061, 0x00000000, 0x00000000 }, + { 0x01000005, 0x20002e28, 0x000000a5, 0x00600060 }, + { 0x00210001, 0x2af401ed, 0x00000000, 0xffffffff }, + { 0x00010020, 0x34001c00, 0x00001400, 0x000000f0 }, + { 0x00000001, 0x2ae000e5, 0x00000000, 0x00000001 }, + { 0x00200001, 0x24000229, 0x004500a0, 0x00000000 }, + { 0x00000040, 0x24003dad, 0x00000400, 0xffffffff }, + { 0x00000041, 0x2b482521, 0x000000a2, 0x00000402 }, + { 0x00000040, 0x2b482421, 0x00000b48, 0x00000400 }, + { 0x00000041, 0x2b480c21, 0x00000b48, 0x00000018 }, + { 0x00000001, 0x2b540231, 0x00000014, 0x00000000 }, + { 0x0a800031, 0x2b601ca1, 0x00000b40, 0x02280303 }, + { 0x05000010, 0x2000252c, 0x00000b70, 0x00000b88 }, + { 0x00210001, 0x2af401ed, 0x00000000, 0xffffffff }, + { 0x00010020, 0x34001c00, 0x00001400, 0x00000040 }, + { 0x00000040, 0x2b480c21, 0x00000b48, 0x00000003 }, + { 0x0a800031, 0x2ba01ca1, 0x00000b40, 0x02480403 }, + { 0x00200001, 0x2ae40021, 0x00450bc8, 0x00000000 }, + { 0x00000001, 0x2af001ed, 0x00000000, 0x00010001 }, + { 0x00600001, 0x2b400061, 0x00000000, 0x00000000 }, + { 0x01000005, 0x20002e28, 0x000000a5, 0x00100010 }, + { 0x00210001, 0x2b1401ed, 0x00000000, 0xffffffff }, + { 0x00210001, 0x2b3401ed, 0x00000000, 0xffffffff }, + { 0x00010020, 0x34001c00, 0x00001400, 0x00000320 }, + { 0x00000001, 0x2b0000e5, 0x00000000, 0x00000001 }, + { 0x00200001, 0x24000229, 0x004500a0, 0x00000000 }, + { 0x00000040, 0x24023dad, 0x00000402, 0xffffffff }, + { 0x00000041, 0x2b482521, 0x000000a2, 0x00000402 }, + { 0x00000040, 0x2b482421, 0x00000b48, 0x00000400 }, + { 0x00000041, 0x2b480c21, 0x00000b48, 0x00000018 }, + { 0x00000001, 0x2b540231, 0x00000014, 0x00000000 }, + { 0x0a800031, 0x2b601ca1, 0x00000b40, 0x02280303 }, + { 0x05000010, 0x2000252c, 0x00000b70, 0x00000b88 }, + { 0x00210001, 0x2b1401ed, 0x00000000, 0xffffffff }, + { 0x00010020, 0x34001c00, 0x00001400, 0x00000040 }, + { 0x00000040, 0x2b480c21, 0x00000b48, 0x00000003 }, + { 0x0a800031, 0x2ba01ca1, 0x00000b40, 0x02480403 }, + { 0x00200001, 0x2b040021, 0x00450bf0, 0x00000000 }, + { 0x00000001, 0x2b1001ed, 0x00000000, 0x00010001 }, + { 0x00600001, 0x2b400061, 0x00000000, 0x00000000 }, + { 0x01000005, 0x20002e28, 0x000000a5, 0x00080008 }, + { 0x00010020, 0x34001c00, 0x00001400, 0x00000110 }, + { 0x00000001, 0x2b2000e5, 0x00000000, 0x00000001 }, + { 0x00200001, 0x24000229, 0x004500a0, 0x00000000 }, + { 0x00000040, 0x24023dad, 0x00000402, 0xffffffff }, + { 0x00000040, 0x24003dad, 0x00000400, 0x00010001 }, + { 0x00000041, 0x2b482521, 0x000000a2, 0x00000402 }, + { 0x00000040, 0x2b482421, 0x00000b48, 0x00000400 }, + { 0x00000041, 0x2b480c21, 0x00000b48, 0x00000018 }, + { 0x00000001, 0x2b540231, 0x00000014, 0x00000000 }, + { 0x0a800031, 0x2b601ca1, 0x00000b40, 0x02280303 }, + { 0x05000010, 0x2000252c, 0x00000b70, 0x00000b88 }, + { 0x00210001, 0x2b3401ed, 0x00000000, 0xffffffff }, + { 0x00010020, 0x34001c00, 0x00001400, 0x00000140 }, + { 0x00000040, 0x2b480c21, 0x00000b48, 0x00000003 }, + { 0x0a800031, 0x2ba01ca1, 0x00000b40, 0x02480403 }, + { 0x00200001, 0x2b240021, 0x00450bf0, 0x00000000 }, + { 0x00000001, 0x2b3001ed, 0x00000000, 0x00010001 }, + { 0x00000020, 0x34001c00, 0x00001400, 0x000000f0 }, + { 0x00000001, 0x2b2000e5, 0x00000000, 0x00000001 }, + { 0x00200001, 0x24000229, 0x004500a0, 0x00000000 }, + { 0x00200040, 0x24003dad, 0x00450400, 0xffffffff }, + { 0x00000041, 0x2b482521, 0x000000a2, 0x00000402 }, + { 0x00000040, 0x2b482421, 0x00000b48, 0x00000400 }, + { 0x00000041, 0x2b480c21, 0x00000b48, 0x00000018 }, + { 0x00000001, 0x2b540231, 0x00000014, 0x00000000 }, + { 0x0a800031, 0x2b601ca1, 0x00000b40, 0x02280303 }, + { 0x05000010, 0x2000252c, 0x00000b70, 0x00000b88 }, + { 0x00210001, 0x2b3401ed, 0x00000000, 0xffffffff }, + { 0x00010020, 0x34001c00, 0x00001400, 0x00000040 }, + { 0x00000040, 0x2b480c21, 0x00000b48, 0x00000003 }, + { 0x0a800031, 0x2ba01cb1, 0x00000b40, 0x02480403 }, + { 0x00200001, 0x2b240021, 0x00450c18, 0x00000000 }, + { 0x00000001, 0x2b3201ed, 0x00000000, 0x00010001 }, + { 0x00000040, 0x240014a5, 0x00000b00, 0x00000b20 }, + { 0x01000010, 0x20001ca4, 0x00000400, 0x00000000 }, + { 0x00110020, 0x34001c00, 0x00001400, 0x00000080 }, + { 0x02000010, 0x20001ca4, 0x00000ae0, 0x00000001 }, + { 0x00010001, 0x2b040021, 0x00000ae4, 0x00000000 }, + { 0x00010001, 0x2b240021, 0x00000ae4, 0x00000000 }, + { 0x00010001, 0x2b140129, 0x00000af4, 0x00000000 }, + { 0x00010001, 0x2b340129, 0x00000af4, 0x00000000 }, + { 0x00010001, 0x2ac00021, 0x00000ae4, 0x00000000 }, + { 0x00110001, 0x2ac00061, 0x00000000, 0x00000000 }, + { 0x00000020, 0x34001c00, 0x00001400, 0x00000190 }, + { 0x00600001, 0x24000061, 0x00000000, 0x00000000 }, + { 0x01000010, 0x20003da4, 0x00200af4, 0x00000000 }, + { 0x00010040, 0x24003dad, 0x00200400, 0x00010001 }, + { 0x00010001, 0x24040021, 0x00000ae4, 0x00000000 }, + { 0x01000010, 0x20003da4, 0x00200b14, 0x00000000 }, + { 0x00010040, 0x24003dad, 0x00200400, 0x00010001 }, + { 0x00010001, 0x24040021, 0x00000b04, 0x00000000 }, + { 0x01000010, 0x20003da4, 0x00200b34, 0x00000000 }, + { 0x00010040, 0x24003dad, 0x00200400, 0x00010001 }, + { 0x00010001, 0x24040021, 0x00000b24, 0x00000000 }, + { 0x01000010, 0x20003da4, 0x00200400, 0x00010001 }, + { 0x00010001, 0x2ac00021, 0x00000404, 0x00000000 }, + { 0x00010020, 0x34001c00, 0x00001400, 0x000000c0 }, + { 0x00000001, 0x2fa001ad, 0x00000ae4, 0x00000000 }, + { 0x00000001, 0x2fa401ad, 0x00000b04, 0x00000000 }, + { 0x00000001, 0x2fa801ad, 0x00000b24, 0x00000000 }, + { 0x00000040, 0x2fe00c01, 0x00001400, 0x00000020 }, + { 0x00000020, 0x34001c00, 0x00001400, 0x000007b0 }, + { 0x00000001, 0x2ac001ad, 0x00000fe4, 0x00000000 }, + { 0x00000001, 0x2fa001ad, 0x00000ae6, 0x00000000 }, + { 0x00000001, 0x2fa401ad, 0x00000b06, 0x00000000 }, + { 0x00000001, 0x2fa801ad, 0x00000b26, 0x00000000 }, + { 0x00000040, 0x2fe00c01, 0x00001400, 0x00000020 }, + { 0x00000020, 0x34001c00, 0x00001400, 0x00000750 }, + { 0x00000001, 0x2ac201ad, 0x00000fe4, 0x00000000 }, { 0x00600001, 0x25600021, 0x008d0020, 0x00000000 }, { 0x00600001, 0x28400021, 0x008d0560, 0x00000000 }, { 0x00600001, 0x28600061, 0x00000000, 0x00000000 }, @@ -81,6 +197,8 @@ { 0x00000001, 0x24600061, 0x00000000, 0x00000002 }, { 0x00000001, 0x24640231, 0x0000009c, 0x00000000 }, { 0x00000001, 0x24680061, 0x00000000, 0x30003030 }, + { 0x00000001, 0x24700021, 0x00000ac0, 0x00000000 }, + { 0x00000001, 0x24740021, 0x00000ac0, 0x00000000 }, { 0x00600001, 0x28200021, 0x008d0460, 0x00000000 }, { 0x00600001, 0x28400021, 0x008d0560, 0x00000000 }, { 0x00000001, 0x28600061, 0x00000000, 0x01010101 }, @@ -131,3 +249,35 @@ { 0x0a800031, 0x20001cac, 0x00000800, 0x040a0203 }, { 0x00600001, 0x2e000021, 0x008d0000, 0x00000000 }, { 0x07800031, 0x24001ca8, 0x00000e00, 0x82000010 }, + { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, + { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, + { 0x06000010, 0x200035ac, 0x00000fa0, 0x00000fa4 }, + { 0x00010001, 0x2f6001ad, 0x00000fa0, 0x00000000 }, + { 0x00110001, 0x2f6001ad, 0x00000fa4, 0x00000000 }, + { 0x06000010, 0x200035ac, 0x00000f60, 0x00000fa8 }, + { 0x00010001, 0x2fe401ad, 0x00000f60, 0x00000000 }, + { 0x00110001, 0x2fe401ad, 0x00000fa8, 0x00000000 }, + { 0x00000001, 0x34000020, 0x00000fe0, 0x00000000 }, + { 0x04000010, 0x200035ac, 0x00000fa0, 0x00000fa4 }, + { 0x00010001, 0x2f6001ad, 0x00000fa0, 0x00000000 }, + { 0x00110001, 0x2f6001ad, 0x00000fa4, 0x00000000 }, + { 0x04000010, 0x200035ac, 0x00000f60, 0x00000fa8 }, + { 0x00010001, 0x2fe401ad, 0x00000f60, 0x00000000 }, + { 0x00110001, 0x2fe401ad, 0x00000fa8, 0x00000000 }, + { 0x00000001, 0x34000020, 0x00000fe0, 0x00000000 }, + { 0x04000010, 0x200035ac, 0x00000fa0, 0x00000fa4 }, + { 0x00010020, 0x34001c00, 0x00001400, 0x00000070 }, + { 0x04000010, 0x200035ac, 0x00000fa0, 0x00000fa8 }, + { 0x00010001, 0x2fe401ad, 0x00000fa0, 0x00000000 }, + { 0x00010020, 0x34001c00, 0x00001400, 0x000000a0 }, + { 0x04000010, 0x200035ac, 0x00000fa4, 0x00000fa8 }, + { 0x00010001, 0x2fe401ad, 0x00000fa8, 0x00000000 }, + { 0x00110001, 0x2fe401ad, 0x00000fa4, 0x00000000 }, + { 0x00000020, 0x34001c00, 0x00001400, 0x00000060 }, + { 0x04000010, 0x200035ac, 0x00000fa4, 0x00000fa8 }, + { 0x00010001, 0x2fe401ad, 0x00000fa4, 0x00000000 }, + { 0x00010020, 0x34001c00, 0x00001400, 0x00000030 }, + { 0x04000010, 0x200035ac, 0x00000fa0, 0x00000fa8 }, + { 0x00010001, 0x2fe401ad, 0x00000fa8, 0x00000000 }, + { 0x00110001, 0x2fe401ad, 0x00000fa0, 0x00000000 }, + { 0x00000001, 0x34000020, 0x00000fe0, 0x00000000 }, diff --git a/src/shaders/vme/vme75.inc b/src/shaders/vme/vme75.inc index 35033c0..a65a4ee 100644 --- a/src/shaders/vme/vme75.inc +++ b/src/shaders/vme/vme75.inc @@ -265,3 +265,47 @@ define(`BIND_IDX_CBCR', `6') define(`LUMA_CHROMA_MODE', `0x0') define(`LUMA_INTRA_MODE', `0x1') define(`LUMA_INTRA_DISABLE', `0x2') + +define(`RETURN_REG', `r127.0') +define(`RET_ARG', `r127.4') + +/* Now at most two registers are used for input parameter */ +define(`INPUT_ARG0', `r125') +define(`INPUT_ARG1', `r126') + +/* Two temporal registers are used in the function */ +define(`TEMP_VAR0', `r123') +define(`TEMP_VAR1', `r124') + + +define(`OBR_MESSAGE_TYPE', `0') +define(`OBR_CACHE_TYPE', `10') +define(`OBR_BIND_IDX', `BIND_IDX_OUTPUT') + +define(`OBR_CONTROL_0', `0') /* 1 OWord, low 128 bits */ +define(`OBR_CONTROL_1', `1') /* 1 OWord, high 128 bits */ +define(`OBR_CONTROL_2', `2') /* 2 OWords */ +define(`OBR_CONTROL_4', `3') /* 4 OWords */ +define(`OBR_CONTROL_8', `4') /* 8 OWords */ +define(`OBR_WRITE_COMMIT_CATEGORY', `0') /* category on SNB+ for Data port */ +define(`OBR_HEADER_PRESENT', `1') + +define(`mb_hwdep', `r5.6') +define(`MB_AVAIL', `1:d') +define(`MB_PRED_FLAG', `1:w') + +define(`mb_mvp_ref', `r86') +define(`mba_result', `r87') +define(`mbb_result', `r88') +define(`mbc_result', `r89') +define(`mb_ind', `90') +define(`mb_msg0', `r90') +define(`mb_wb', `r91') +define(`mb_intra_wb', `r91') +define(`mb_inter_wb', `r92') +define(`mb_mv0', `r93') +define(`mb_mv1', `r94') +define(`mb_mv2', `r95') +define(`mb_mv3', `r96') +define(`mb_ref', `r97') + |