diff options
author | Zhao Yakui <yakui.zhao@intel.com> | 2013-08-12 15:13:23 +0800 |
---|---|---|
committer | Xiang, Haihao <haihao.xiang@intel.com> | 2013-09-06 15:33:20 +0800 |
commit | 7264aeb6c682b0bc352303befdcec536cca02fec (patch) | |
tree | 04f47f1494a13b64ad76ca3b9626f25fa354d64e | |
parent | 6d6c227ca19ffaa7454cc748b5864f2024974b33 (diff) |
Optimize the VME shader for MPEG2 on Haswell
Signed-off-by: Zhao Yakui <yakui.zhao@intel.com>
(cherry picked from commit 19e93152f0e10f94ecaf3ddecf95c1dc7b97dfed)
-rw-r--r-- | src/shaders/vme/mpeg2_inter_haswell.asm | 145 | ||||
-rw-r--r-- | src/shaders/vme/mpeg2_inter_haswell.g75b | 82 | ||||
-rw-r--r-- | src/shaders/vme/vme75_mpeg2.inc | 10 |
3 files changed, 234 insertions, 3 deletions
diff --git a/src/shaders/vme/mpeg2_inter_haswell.asm b/src/shaders/vme/mpeg2_inter_haswell.asm index cc9efa4..0e91a04 100644 --- a/src/shaders/vme/mpeg2_inter_haswell.asm +++ b/src/shaders/vme/mpeg2_inter_haswell.asm @@ -414,6 +414,7 @@ send (8) mlen sic_vme_msg_length rlen vme_wb_length {align1}; + /* * Oword Block Write message */ @@ -540,6 +541,11 @@ send (8) rlen vme_wb_length {align1}; +and.z.f0.0 (1) null:uw mb_hwdep<0,1,0>:uw 0x04:uw {align1}; +(-f0.0) jmpi (1) vme_run_again; +nop; +vme_mv_output: + add (1) obw_m0.8<1>:UD obw_m0.8<0,1,0>:UD 0x02:UD {align1}; mov (8) msg_reg0.0<1>:UD obw_m0<8,8,1>:UD {align1}; /* write FME info */ @@ -568,6 +574,7 @@ send (16) rlen obw_wb_length {align1}; + /* Write FME/BME MV */ add (1) obw_m0.8<1>:UD obw_m0.8<0,1,0>:UD 0x01:UD {align1}; mov (8) msg_reg0.0<1>:UD obw_m0.0<8,8,1>:UD {align1}; @@ -617,6 +624,7 @@ send (16) rlen obw_wb_length {align1}; + /* Issue message fence so that the previous write message is committed */ send (16) mb_ind @@ -713,3 +721,140 @@ add (2) RET_ARG<1>:w TEMP_VAR0.0<2,2,1>:w TEMP_VAR1.0<2,2,1>:w {align1}; RETURN {align1}; nop; nop; + +vme_run_again: + +asr (2) mb_ref_win.0<1>:w mb_mvp_ref.0<2,2,1>:w 2:w {align1}; +mov (2) tmp_reg0.0<1>:w mb_ref_win.0<2,2,1>:w {align1}; +add (2) mb_ref_win.8<1>:w mb_ref_win.0<2,2,1>:w 3:w {align1}; +and (2) mb_ref_win.16<1>:uw mb_ref_win.8<2,2,1>:uw 0xFFFC:uw {align1}; + +cmp.l.f0.0 (1) null:w tmp_reg0.0<0,1,0>:w 0:w {align1}; +(f0.0) mul (1) tmp_reg0.0<1>:w tmp_reg0.0<0,1,0>:w -1:w {align1}; +cmp.l.f0.0 (1) null:w tmp_reg0.2<0,1,0>:w 0:w {align1}; +(f0.0) mul (1) tmp_reg0.2<1>:w tmp_reg0.2<0,1,0>:w -1:w {align1}; + +cmp.ge.f0.0 (1) null:w tmp_reg0.0<0,1,0>:w 4:w {align1}; +(f0.0) jmpi (1) vme_start; +cmp.ge.f0.0 (1) null:w tmp_reg0.2<0,1,0>:w 4:w {align1}; +(f0.0) jmpi (1) vme_start; + +jmpi (1) vme_done; + +vme_start: + mov (8) tmp_vme_wb0.0<1>:ud vme_wb0.0<8,8,1>:ud {align1}; + mov (8) tmp_vme_wb1.0<1>:ud vme_wb1.0<8,8,1>:ud {align1}; + +/* Calibrate the ref window for MPEG2 */ +mov (1) vme_m0.0<1>:W -16:W {align1}; +mov (1) vme_m0.2<1>:W -12:W {align1}; +mov (4) INPUT_ARG0.0<1>:ud vme_m0.0<4,4,1>:ud {align1}; +add (2) INPUT_ARG0.0<1>:w INPUT_ARG0.0<2,2,1>:w mb_ref_win.16<2,2,1>:w {align1}; +mov (8) INPUT_ARG1.0<1>:ud pic_ref.0<8,8,1>:ud {align1}; + +SAVE_RET {align1}; +jmpi (1) ref_boundary_check; +mov (2) vme_m0.0<1>:w RET_ARG<2,2,1>:w {align1}; + +/* IME search */ +mov (1) vme_m0.12<1>:UD SEARCH_CTRL_SINGLE + INTER_PART_MASK + INTER_SAD_HAAR:UD {align1}; /* 16x16 Source, harr */ +mov (1) vme_m0.22<1>:UW REF_REGION_SIZE {align1}; /* Reference Width&Height, 48x40 */ + +mov (1) vme_m0.4<1>:UD vme_m0.0<0,1,0>:UD {align1}; + +mov (8) vme_msg_0.0<1>:UD vme_m0.0<8,8,1>:UD {align1}; + +mov (8) vme_m1.0<1>:ud 0x0:UD {align1}; + +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 mv_cc_ref.0<0,1,0>:ud {align1}; +mov (1) vme_m1.20<1>:ud mv_cc_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}; +/* M3/M4 search path */ + +mov (1) vme_msg_3.0<1>:UD 0x01010101:UD {align1}; +mov (1) vme_msg_3.4<1>:UD 0x10010101:UD {align1}; +mov (1) vme_msg_3.8<1>:UD 0x0F0F0F0F:UD {align1}; +mov (1) vme_msg_3.12<1>:UD 0x100F0F0F:UD {align1}; +mov (1) vme_msg_3.16<1>:UD 0x01010101:UD {align1}; +mov (1) vme_msg_3.20<1>:UD 0x10010101:UD {align1}; +mov (1) vme_msg_3.24<1>:UD 0x0F0F0F0F:UD {align1}; +mov (1) vme_msg_3.28<1>:UD 0x100F0F0F:UD {align1}; + +mov (1) vme_msg_4.0<1>:UD 0x01010101:UD {align1}; +mov (1) vme_msg_4.4<1>:UD 0x10010101:UD {align1}; +mov (1) vme_msg_4.8<1>:UD 0x0F0F0F0F:UD {align1}; +mov (1) vme_msg_4.12<1>:UD 0x000F0F0F:UD {align1}; + +mov (4) vme_msg_4.16<1>:UD 0x0:UD {align1}; + +send (8) + vme_msg_ind + vme_wb<1>:UD + null + vme( + BIND_IDX_VME, + 0, + 0, + VME_IME_MESSAGE_TYPE + ) + mlen ime_vme_msg_length + rlen vme_wb_length {align1}; + +/* Set Macroblock-shape/mode for FBR */ + +mov (1) vme_m2.20<1>:UD 0x0:UD {align1}; +mov (1) vme_m2.21<1>:UB vme_wb.25<0,1,0>:UB {align1}; +mov (1) vme_m2.22<1>:UB vme_wb.26<0,1,0>:UB {align1}; + +and (1) tmp_reg0.0<1>:UW vme_wb.0<0,1,0>:UW 0x03:UW {align1}; +mov (1) vme_m2.20<1>:UB tmp_reg0.0<0,1,0>:UB {align1}; + +/* Send FBR message into CRE */ + +mov (8) vme_msg_3.0<1>:UD vme_wb1.0<8,8,1>:UD {align1}; +mov (8) vme_msg_4.0<1>:ud vme_wb2.0<8,8,1>:ud {align1}; +mov (8) vme_msg_5.0<1>:ud vme_wb3.0<8,8,1>:ud {align1}; +mov (8) vme_msg_6.0<1>:ud vme_wb4.0<8,8,1>:ud {align1}; + +mov (1) vme_m0.12<1>:UD INTER_SAD_HAAR + SUB_PEL_MODE_HALF + FBR_BME_DISABLE:UD {align1}; /* 16x16 Source, 1/2 pixel, harr, BME disable */ + +/* Bilinear filter */ +mov (1) tmp_reg0.0<1>:uw 0x04:uw {align1}; +add (1) vme_m1.30<1>:ub vme_m1.30<0,1,0>:ub tmp_reg0.0<0,1,0>:ub {align1}; + +mov (8) vme_msg_0.0<1>:UD vme_m0.0<8,8,1>:UD {align1}; +mov (8) vme_msg_1.0<1>:UD vme_m1.0<8,8,1>:UD {align1}; + +mov (8) vme_msg_2.0<1>:UD vme_m2.0<8,8,1>:UD {align1}; + +/* after verification it will be passed by using payload */ +send (8) + vme_msg_ind + vme_wb<1>:UD + null + cre( + BIND_IDX_VME, + VME_FBR_MESSAGE_TYPE + ) + mlen fbr_vme_msg_length + rlen vme_wb_length + {align1}; + +cmp.l.f0.0 (1) null:uw vme_wb0.8<0,1,0>:uw tmp_vme_wb0.8<0,1,0>:uw {align1}; +(f0.0) jmpi (1) vme_done; +mov (8) vme_wb0.0<1>:ud tmp_vme_wb0.0<8,8,1>:ud {align1}; +mov (8) vme_wb1.0<1>:ud tmp_vme_wb1.0<8,8,1>:ud {align1}; + +vme_done: + jmpi (1) vme_mv_output; +nop; +nop; +nop; + diff --git a/src/shaders/vme/mpeg2_inter_haswell.g75b b/src/shaders/vme/mpeg2_inter_haswell.g75b index 277f116..d6625e3 100644 --- a/src/shaders/vme/mpeg2_inter_haswell.g75b +++ b/src/shaders/vme/mpeg2_inter_haswell.g75b @@ -122,13 +122,13 @@ { 0x00000001, 0x2fa401ad, 0x00000b04, 0x00000000 }, { 0x00000001, 0x2fa801ad, 0x00000b24, 0x00000000 }, { 0x00000040, 0x2fe00c01, 0x00001400, 0x00000020 }, - { 0x00000020, 0x34001c00, 0x00001400, 0x00000730 }, + { 0x00000020, 0x34001c00, 0x00001400, 0x00000760 }, { 0x00000001, 0x2ac001ad, 0x00000fe4, 0x00000000 }, { 0x00000001, 0x2fa001ad, 0x00000ae6, 0x00000000 }, { 0x00000001, 0x2fa401ad, 0x00000b06, 0x00000000 }, { 0x00000001, 0x2fa801ad, 0x00000b26, 0x00000000 }, { 0x00000040, 0x2fe00c01, 0x00001400, 0x00000020 }, - { 0x00000020, 0x34001c00, 0x00001400, 0x000006d0 }, + { 0x00000020, 0x34001c00, 0x00001400, 0x00000700 }, { 0x00000001, 0x2ac201ad, 0x00000fe4, 0x00000000 }, { 0x00200001, 0x2a2001ad, 0x00450ae4, 0x00000000 }, { 0x00000001, 0x244001ed, 0x00000000, 0xfff0fff0 }, @@ -137,7 +137,7 @@ { 0x00000001, 0x2fa80021, 0x00000448, 0x00000000 }, { 0x00600001, 0x2fc00021, 0x008d0a40, 0x00000000 }, { 0x00000040, 0x2fe00c01, 0x00001400, 0x00000020 }, - { 0x00000020, 0x34001c00, 0x00001400, 0x00000760 }, + { 0x00000020, 0x34001c00, 0x00001400, 0x00000790 }, { 0x00200001, 0x244001ad, 0x00450fe4, 0x00000000 }, { 0x00600001, 0x25600021, 0x008d0020, 0x00000000 }, { 0x00600001, 0x28400021, 0x008d0560, 0x00000000 }, @@ -215,6 +215,9 @@ { 0x00600001, 0x28200021, 0x008d0460, 0x00000000 }, { 0x00600001, 0x28400021, 0x008d0560, 0x00000000 }, { 0x0d600031, 0x21801ca1, 0x00000800, 0x0e786000 }, + { 0x01000005, 0x20002d28, 0x000000a6, 0x00040004 }, + { 0x00110020, 0x34001c00, 0x00001400, 0x000004a0 }, + { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, { 0x00000040, 0x24880c21, 0x00000488, 0x00000002 }, { 0x00600001, 0x28000021, 0x008d0480, 0x00000000 }, { 0x00000001, 0x28200021, 0x00000180, 0x00000000 }, @@ -288,3 +291,76 @@ { 0x00000001, 0x34000020, 0x00000fe0, 0x00000000 }, { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, + { 0x0020000c, 0x2a803dad, 0x00450ac0, 0x00020002 }, + { 0x00200001, 0x240001ad, 0x00450a80, 0x00000000 }, + { 0x00200040, 0x2a883dad, 0x00450a80, 0x00030003 }, + { 0x00200005, 0x2a902d29, 0x00450a88, 0xfffcfffc }, + { 0x05000010, 0x20003dac, 0x00000400, 0x00000000 }, + { 0x00010041, 0x24003dad, 0x00000400, 0xffffffff }, + { 0x05000010, 0x20003dac, 0x00000402, 0x00000000 }, + { 0x00010041, 0x24023dad, 0x00000402, 0xffffffff }, + { 0x04000010, 0x20003dac, 0x00000400, 0x00040004 }, + { 0x00010020, 0x34001c00, 0x00001400, 0x00000030 }, + { 0x04000010, 0x20003dac, 0x00000402, 0x00040004 }, + { 0x00010020, 0x34001c00, 0x00001400, 0x00000010 }, + { 0x00000020, 0x34001c00, 0x00001400, 0x00000380 }, + { 0x00600001, 0x2c800021, 0x008d0180, 0x00000000 }, + { 0x00600001, 0x2ca00021, 0x008d01a0, 0x00000000 }, + { 0x00000001, 0x244001ed, 0x00000000, 0xfff0fff0 }, + { 0x00000001, 0x244201ed, 0x00000000, 0xfff4fff4 }, + { 0x00400001, 0x2fa00021, 0x00690440, 0x00000000 }, + { 0x00200040, 0x2fa035ad, 0x00450fa0, 0x00450a90 }, + { 0x00600001, 0x2fc00021, 0x008d0a40, 0x00000000 }, + { 0x00000040, 0x2fe00c01, 0x00001400, 0x00000020 }, + { 0x00000020, 0x34001c00, 0x00001400, 0xfffffca0 }, + { 0x00200001, 0x244001ad, 0x00450fe4, 0x00000000 }, + { 0x00000001, 0x244c0061, 0x00000000, 0x7e200000 }, + { 0x00000001, 0x24560169, 0x00000000, 0x28302830 }, + { 0x00000001, 0x24440021, 0x00000440, 0x00000000 }, + { 0x00600001, 0x28000021, 0x008d0440, 0x00000000 }, + { 0x00600001, 0x24600061, 0x00000000, 0x00000000 }, + { 0x00000001, 0x24600061, 0x00000000, 0x00000002 }, + { 0x00000001, 0x24640231, 0x0000009c, 0x00000000 }, + { 0x00000001, 0x24680061, 0x00000000, 0x30003030 }, + { 0x00000001, 0x24700021, 0x00000a20, 0x00000000 }, + { 0x00000001, 0x24740021, 0x00000a20, 0x00000000 }, + { 0x00600001, 0x28200021, 0x008d0460, 0x00000000 }, + { 0x00600001, 0x28400021, 0x008d0560, 0x00000000 }, + { 0x00000001, 0x28600061, 0x00000000, 0x01010101 }, + { 0x00000001, 0x28640061, 0x00000000, 0x10010101 }, + { 0x00000001, 0x28680061, 0x00000000, 0x0f0f0f0f }, + { 0x00000001, 0x286c0061, 0x00000000, 0x100f0f0f }, + { 0x00000001, 0x28700061, 0x00000000, 0x01010101 }, + { 0x00000001, 0x28740061, 0x00000000, 0x10010101 }, + { 0x00000001, 0x28780061, 0x00000000, 0x0f0f0f0f }, + { 0x00000001, 0x287c0061, 0x00000000, 0x100f0f0f }, + { 0x00000001, 0x28800061, 0x00000000, 0x01010101 }, + { 0x00000001, 0x28840061, 0x00000000, 0x10010101 }, + { 0x00000001, 0x28880061, 0x00000000, 0x0f0f0f0f }, + { 0x00000001, 0x288c0061, 0x00000000, 0x000f0f0f }, + { 0x00400001, 0x28900061, 0x00000000, 0x00000000 }, + { 0x08600031, 0x21801ca1, 0x00000800, 0x0a784000 }, + { 0x00000001, 0x25740061, 0x00000000, 0x00000000 }, + { 0x00000001, 0x25750231, 0x00000199, 0x00000000 }, + { 0x00000001, 0x25760231, 0x0000019a, 0x00000000 }, + { 0x00000005, 0x24002d29, 0x00000180, 0x00030003 }, + { 0x00000001, 0x25740231, 0x00000400, 0x00000000 }, + { 0x00600001, 0x28600021, 0x008d01a0, 0x00000000 }, + { 0x00600001, 0x28800021, 0x008d01c0, 0x00000000 }, + { 0x00600001, 0x28a00021, 0x008d01e0, 0x00000000 }, + { 0x00600001, 0x28c00021, 0x008d0200, 0x00000000 }, + { 0x00000001, 0x244c0061, 0x00000000, 0x00241000 }, + { 0x00000001, 0x24000169, 0x00000000, 0x00040004 }, + { 0x00000040, 0x247e4631, 0x0000047e, 0x00000400 }, + { 0x00600001, 0x28000021, 0x008d0440, 0x00000000 }, + { 0x00600001, 0x28200021, 0x008d0460, 0x00000000 }, + { 0x00600001, 0x28400021, 0x008d0560, 0x00000000 }, + { 0x0d600031, 0x21801ca1, 0x00000800, 0x0e786000 }, + { 0x05000010, 0x20002528, 0x00000188, 0x00000c88 }, + { 0x00010020, 0x34001c00, 0x00001400, 0x00000020 }, + { 0x00600001, 0x21800021, 0x008d0c80, 0x00000000 }, + { 0x00600001, 0x21a00021, 0x008d0ca0, 0x00000000 }, + { 0x00000020, 0x34001c00, 0x00001400, 0xfffff710 }, + { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, + { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, + { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, diff --git a/src/shaders/vme/vme75_mpeg2.inc b/src/shaders/vme/vme75_mpeg2.inc index b297a49..b638056 100644 --- a/src/shaders/vme/vme75_mpeg2.inc +++ b/src/shaders/vme/vme75_mpeg2.inc @@ -21,3 +21,13 @@ define(`pic_ref', `r82') define(`INTRA16_DC_PRED', `0xBB') /* Cost center ref */ define(`mv_cc_ref', `r81') +define(`tmp_vme_wb0', `r100') +define(`tmp_vme_wb1', `r101') +define(`tmp_vme_wb2', `r102') +define(`tmp_vme_wb3', `r103') +define(`tmp_vme_wb4', `r104') +define(`tmp_vme_wb5', `r105') +define(`tmp_vme_wb6', `r106') +define(`tmp_vme_wb7', `r107') +define(`tmp_vme_wb8', `r108') +define(`tmp_vme_wb9', `r109') |