R6xx/R7xx Xv: switch packed over to Yang's new shader code

2 files changed, 99 insertions, 107 deletions

diff --git a/src/r600_exa.c b/src/r600_exa.c
index 34e67d83..542d42d7 100644
--- a/src/r600_exa.c
+++ b/src/r600_exa.c
@@ -2841,7 +2841,7 @@ R600LoadShaders(ScrnInfoPtr pScrn, ScreenPtr pScreen)
     // 2
     ps[i++] = CF_ALLOC_IMP_EXP_DWORD0(ARRAY_BASE(CF_PIXEL_MRT0),
 				      TYPE(SQ_EXPORT_PIXEL),
-				      RW_GPR(3),
+				      RW_GPR(2),
 				      RW_REL(ABSOLUTE),
 				      INDEX_GPR(0),
 				      ELEM_SIZE(3));
@@ -2856,96 +2856,88 @@ R600LoadShaders(ScrnInfoPtr pScrn, ScreenPtr pScreen)
 					   CF_INST(SQ_CF_INST_EXPORT_DONE),
 					   WHOLE_QUAD_MODE(0),
 					   BARRIER(1));
+    /* Undo scaling of Y'CbCr values
+     *  Y' is scaled from 16:235
+     *  Cb/Cr are scaled from 16:240
+     */
     // 3 - alu 0
-    // DP4 gpr[2].x gpr[1].x c[0].x
+    // MULADD gpr[1].x gpr[1].x c[3].x c[3].y
     ps[i++] = ALU_DWORD0(SRC0_SEL(1),
 			 SRC0_REL(ABSOLUTE),
 			 SRC0_ELEM(ELEM_X),
 			 SRC0_NEG(0),
-			 SRC1_SEL(256),
+			 SRC1_SEL(259),
 			 SRC1_REL(ABSOLUTE),
 			 SRC1_ELEM(ELEM_X),
 			 SRC1_NEG(0),
 			 INDEX_MODE(SQ_INDEX_LOOP),
 			 PRED_SEL(SQ_PRED_SEL_OFF),
 			 LAST(0));
-    ps[i++] = ALU_DWORD1_OP2(info->ChipFamily,
-			     SRC0_ABS(0),
-			     SRC1_ABS(0),
-			     UPDATE_EXECUTE_MASK(0),
-			     UPDATE_PRED(0),
-			     WRITE_MASK(1),
-			     FOG_MERGE(0),
-			     OMOD(SQ_ALU_OMOD_OFF),
-			     ALU_INST(SQ_OP2_INST_DOT4),
-			     BANK_SWIZZLE(SQ_ALU_VEC_102),
-			     DST_GPR(2),
+    ps[i++] = ALU_DWORD1_OP3(SRC2_SEL(259),
+			     SRC2_REL(ABSOLUTE),
+			     SRC2_ELEM(ELEM_Y),
+			     SRC2_NEG(0),
+			     ALU_INST(SQ_OP3_INST_MULADD),
+			     BANK_SWIZZLE(SQ_ALU_VEC_012),
+			     DST_GPR(1),
 			     DST_REL(ABSOLUTE),
 			     DST_ELEM(ELEM_X),
 			     CLAMP(1));
     // 4 - alu 1
-    // DP4 gpr[2].y gpr[1].y c[0].y
+    // MULADD gpr[1].y gpr[1].y c[3].z c[3].w
     ps[i++] = ALU_DWORD0(SRC0_SEL(1),
 			 SRC0_REL(ABSOLUTE),
 			 SRC0_ELEM(ELEM_Y),
 			 SRC0_NEG(0),
-			 SRC1_SEL(256),
+			 SRC1_SEL(259),
 			 SRC1_REL(ABSOLUTE),
-			 SRC1_ELEM(ELEM_Y),
+			 SRC1_ELEM(ELEM_Z),
 			 SRC1_NEG(0),
 			 INDEX_MODE(SQ_INDEX_LOOP),
 			 PRED_SEL(SQ_PRED_SEL_OFF),
 			 LAST(0));
-    ps[i++] = ALU_DWORD1_OP2(info->ChipFamily,
-			     SRC0_ABS(0),
-			     SRC1_ABS(0),
-			     UPDATE_EXECUTE_MASK(0),
-			     UPDATE_PRED(0),
-			     WRITE_MASK(0),
-			     FOG_MERGE(0),
-			     OMOD(SQ_ALU_OMOD_OFF),
-			     ALU_INST(SQ_OP2_INST_DOT4),
-			     BANK_SWIZZLE(SQ_ALU_VEC_102),
-			     DST_GPR(2),
+    ps[i++] = ALU_DWORD1_OP3(SRC2_SEL(259),
+			     SRC2_REL(ABSOLUTE),
+			     SRC2_ELEM(ELEM_W),
+			     SRC2_NEG(0),
+			     ALU_INST(SQ_OP3_INST_MULADD),
+			     BANK_SWIZZLE(SQ_ALU_VEC_012),
+			     DST_GPR(1),
 			     DST_REL(ABSOLUTE),
 			     DST_ELEM(ELEM_Y),
-			     CLAMP(1));
+			     CLAMP(0));
     // 5 - alu 2
-    // DP4 gpr[2].z gpr[1].z c[0].z
+    // MULADD gpr[1].z gpr[1].z c[3].z c[3].w
     ps[i++] = ALU_DWORD0(SRC0_SEL(1),
 			 SRC0_REL(ABSOLUTE),
 			 SRC0_ELEM(ELEM_Z),
 			 SRC0_NEG(0),
-			 SRC1_SEL(256),
+			 SRC1_SEL(259),
 			 SRC1_REL(ABSOLUTE),
 			 SRC1_ELEM(ELEM_Z),
 			 SRC1_NEG(0),
 			 INDEX_MODE(SQ_INDEX_LOOP),
 			 PRED_SEL(SQ_PRED_SEL_OFF),
 			 LAST(0));
-    ps[i++] = ALU_DWORD1_OP2(info->ChipFamily,
-			     SRC0_ABS(0),
-			     SRC1_ABS(0),
-			     UPDATE_EXECUTE_MASK(0),
-			     UPDATE_PRED(0),
-			     WRITE_MASK(0),
-			     FOG_MERGE(0),
-			     OMOD(SQ_ALU_OMOD_OFF),
-			     ALU_INST(SQ_OP2_INST_DOT4),
-			     BANK_SWIZZLE(SQ_ALU_VEC_102),
-			     DST_GPR(2),
+    ps[i++] = ALU_DWORD1_OP3(SRC2_SEL(259),
+			     SRC2_REL(ABSOLUTE),
+			     SRC2_ELEM(ELEM_W),
+			     SRC2_NEG(0),
+			     ALU_INST(SQ_OP3_INST_MULADD),
+			     BANK_SWIZZLE(SQ_ALU_VEC_012),
+			     DST_GPR(1),
 			     DST_REL(ABSOLUTE),
 			     DST_ELEM(ELEM_Z),
-			     CLAMP(1));
+			     CLAMP(0));
     // 6 - alu 3
-    // DP4 gpr[2].w gpr[1].w c[0].w
-    ps[i++] = ALU_DWORD0(SRC0_SEL(1),
+    // MOV gpr[1].w 0.0
+    ps[i++] = ALU_DWORD0(SRC0_SEL(SQ_ALU_SRC_0),
 			 SRC0_REL(ABSOLUTE),
-			 SRC0_ELEM(ELEM_W),
+			 SRC0_ELEM(ELEM_X),
 			 SRC0_NEG(0),
-			 SRC1_SEL(256),
+			 SRC1_SEL(SQ_ALU_SRC_0),
 			 SRC1_REL(ABSOLUTE),
-			 SRC1_ELEM(ELEM_W),
+			 SRC1_ELEM(ELEM_X),
 			 SRC1_NEG(0),
 			 INDEX_MODE(SQ_INDEX_LOOP),
 			 PRED_SEL(SQ_PRED_SEL_OFF),
@@ -2955,22 +2947,22 @@ R600LoadShaders(ScrnInfoPtr pScrn, ScreenPtr pScreen)
 			     SRC1_ABS(0),
 			     UPDATE_EXECUTE_MASK(0),
 			     UPDATE_PRED(0),
-			     WRITE_MASK(0),
+			     WRITE_MASK(1),
 			     FOG_MERGE(0),
 			     OMOD(SQ_ALU_OMOD_OFF),
 			     ALU_INST(SQ_OP2_INST_DOT4),
-			     BANK_SWIZZLE(SQ_ALU_VEC_021),
-			     DST_GPR(2),
+			     BANK_SWIZZLE(SQ_ALU_VEC_102),
+			     DST_GPR(1),
 			     DST_REL(ABSOLUTE),
 			     DST_ELEM(ELEM_W),
-			     CLAMP(1));
+			     CLAMP(0));
     // 7 - alu 4
-    // DP4 gpr[2].x gpr[1].x c[1].x
+    // DP4 gpr[2].x gpr[1].x c[0].x
     ps[i++] = ALU_DWORD0(SRC0_SEL(1),
 			 SRC0_REL(ABSOLUTE),
 			 SRC0_ELEM(ELEM_X),
 			 SRC0_NEG(0),
-			 SRC1_SEL(257),
+			 SRC1_SEL(256),
 			 SRC1_REL(ABSOLUTE),
 			 SRC1_ELEM(ELEM_X),
 			 SRC1_NEG(0),
@@ -2982,7 +2974,7 @@ R600LoadShaders(ScrnInfoPtr pScrn, ScreenPtr pScreen)
 			     SRC1_ABS(0),
 			     UPDATE_EXECUTE_MASK(0),
 			     UPDATE_PRED(0),
-			     WRITE_MASK(0),
+			     WRITE_MASK(1),
 			     FOG_MERGE(0),
 			     OMOD(SQ_ALU_OMOD_OFF),
 			     ALU_INST(SQ_OP2_INST_DOT4),
@@ -2992,12 +2984,12 @@ R600LoadShaders(ScrnInfoPtr pScrn, ScreenPtr pScreen)
 			     DST_ELEM(ELEM_X),
 			     CLAMP(1));
     // 8 - alu 5
-    // DP4 gpr[2].y gpr[1].y c[1].y
+    // DP4 gpr[2].y gpr[1].y c[0].y
     ps[i++] = ALU_DWORD0(SRC0_SEL(1),
 			 SRC0_REL(ABSOLUTE),
 			 SRC0_ELEM(ELEM_Y),
 			 SRC0_NEG(0),
-			 SRC1_SEL(257),
+			 SRC1_SEL(256),
 			 SRC1_REL(ABSOLUTE),
 			 SRC1_ELEM(ELEM_Y),
 			 SRC1_NEG(0),
@@ -3009,7 +3001,7 @@ R600LoadShaders(ScrnInfoPtr pScrn, ScreenPtr pScreen)
 			     SRC1_ABS(0),
 			     UPDATE_EXECUTE_MASK(0),
 			     UPDATE_PRED(0),
-			     WRITE_MASK(1),
+			     WRITE_MASK(0),
 			     FOG_MERGE(0),
 			     OMOD(SQ_ALU_OMOD_OFF),
 			     ALU_INST(SQ_OP2_INST_DOT4),
@@ -3019,12 +3011,12 @@ R600LoadShaders(ScrnInfoPtr pScrn, ScreenPtr pScreen)
 			     DST_ELEM(ELEM_Y),
 			     CLAMP(1));
     // 9 - alu 6
-    // DP4 gpr[2].z gpr[1].z c[1].z
+    // DP4 gpr[2].z gpr[1].z c[0].z
     ps[i++] = ALU_DWORD0(SRC0_SEL(1),
 			 SRC0_REL(ABSOLUTE),
 			 SRC0_ELEM(ELEM_Z),
 			 SRC0_NEG(0),
-			 SRC1_SEL(257),
+			 SRC1_SEL(256),
 			 SRC1_REL(ABSOLUTE),
 			 SRC1_ELEM(ELEM_Z),
 			 SRC1_NEG(0),
@@ -3046,12 +3038,12 @@ R600LoadShaders(ScrnInfoPtr pScrn, ScreenPtr pScreen)
 			     DST_ELEM(ELEM_Z),
 			     CLAMP(1));
     // 10 - alu 7
-    // DP4 gpr[2].w gpr[1].w c[1].w
+    // DP4 gpr[2].w gpr[1].w c[0].w
     ps[i++] = ALU_DWORD0(SRC0_SEL(1),
 			 SRC0_REL(ABSOLUTE),
 			 SRC0_ELEM(ELEM_W),
 			 SRC0_NEG(0),
-			 SRC1_SEL(257),
+			 SRC1_SEL(256),
 			 SRC1_REL(ABSOLUTE),
 			 SRC1_ELEM(ELEM_W),
 			 SRC1_NEG(0),
@@ -3073,12 +3065,12 @@ R600LoadShaders(ScrnInfoPtr pScrn, ScreenPtr pScreen)
 			     DST_ELEM(ELEM_W),
 			     CLAMP(1));
     // 11 - alu 8
-    // DP4 gpr[2].x gpr[1].x c[2].x
+    // DP4 gpr[2].x gpr[1].x c[1].x
     ps[i++] = ALU_DWORD0(SRC0_SEL(1),
 			 SRC0_REL(ABSOLUTE),
 			 SRC0_ELEM(ELEM_X),
 			 SRC0_NEG(0),
-			 SRC1_SEL(258),
+			 SRC1_SEL(257),
 			 SRC1_REL(ABSOLUTE),
 			 SRC1_ELEM(ELEM_X),
 			 SRC1_NEG(0),
@@ -3100,12 +3092,12 @@ R600LoadShaders(ScrnInfoPtr pScrn, ScreenPtr pScreen)
 			     DST_ELEM(ELEM_X),
 			     CLAMP(1));
     // 12 - alu 9
-    // DP4 gpr[2].y gpr[1].y c[2].y
+    // DP4 gpr[2].y gpr[1].y c[1].y
     ps[i++] = ALU_DWORD0(SRC0_SEL(1),
 			 SRC0_REL(ABSOLUTE),
 			 SRC0_ELEM(ELEM_Y),
 			 SRC0_NEG(0),
-			 SRC1_SEL(258),
+			 SRC1_SEL(257),
 			 SRC1_REL(ABSOLUTE),
 			 SRC1_ELEM(ELEM_Y),
 			 SRC1_NEG(0),
@@ -3117,7 +3109,7 @@ R600LoadShaders(ScrnInfoPtr pScrn, ScreenPtr pScreen)
 			     SRC1_ABS(0),
 			     UPDATE_EXECUTE_MASK(0),
 			     UPDATE_PRED(0),
-			     WRITE_MASK(0),
+			     WRITE_MASK(1),
 			     FOG_MERGE(0),
 			     OMOD(SQ_ALU_OMOD_OFF),
 			     ALU_INST(SQ_OP2_INST_DOT4),
@@ -3127,12 +3119,12 @@ R600LoadShaders(ScrnInfoPtr pScrn, ScreenPtr pScreen)
 			     DST_ELEM(ELEM_Y),
 			     CLAMP(1));
     // 13 - alu 10
-    // DP4 gpr[2].z gpr[1].z c[2].z
+    // DP4 gpr[2].z gpr[1].z c[1].z
     ps[i++] = ALU_DWORD0(SRC0_SEL(1),
 			 SRC0_REL(ABSOLUTE),
 			 SRC0_ELEM(ELEM_Z),
 			 SRC0_NEG(0),
-			 SRC1_SEL(258),
+			 SRC1_SEL(257),
 			 SRC1_REL(ABSOLUTE),
 			 SRC1_ELEM(ELEM_Z),
 			 SRC1_NEG(0),
@@ -3144,7 +3136,7 @@ R600LoadShaders(ScrnInfoPtr pScrn, ScreenPtr pScreen)
 			     SRC1_ABS(0),
 			     UPDATE_EXECUTE_MASK(0),
 			     UPDATE_PRED(0),
-			     WRITE_MASK(1),
+			     WRITE_MASK(0),
 			     FOG_MERGE(0),
 			     OMOD(SQ_ALU_OMOD_OFF),
 			     ALU_INST(SQ_OP2_INST_DOT4),
@@ -3154,12 +3146,12 @@ R600LoadShaders(ScrnInfoPtr pScrn, ScreenPtr pScreen)
 			     DST_ELEM(ELEM_Z),
 			     CLAMP(1));
     // 14 - alu 11
-    // DP4 gpr[2].w gpr[1].w c[2].w
+    // DP4 gpr[2].w gpr[1].w c[1].w
     ps[i++] = ALU_DWORD0(SRC0_SEL(1),
 			 SRC0_REL(ABSOLUTE),
 			 SRC0_ELEM(ELEM_W),
 			 SRC0_NEG(0),
-			 SRC1_SEL(258),
+			 SRC1_SEL(257),
 			 SRC1_REL(ABSOLUTE),
 			 SRC1_ELEM(ELEM_W),
 			 SRC1_NEG(0),
@@ -3181,12 +3173,12 @@ R600LoadShaders(ScrnInfoPtr pScrn, ScreenPtr pScreen)
 			     DST_ELEM(ELEM_W),
 			     CLAMP(1));
     // 15 - alu 12
-    // MOV gpr[3].x gpr[2].x
-    ps[i++] = ALU_DWORD0(SRC0_SEL(2),
+    // DP4 gpr[2].x gpr[1].x c[2].x
+    ps[i++] = ALU_DWORD0(SRC0_SEL(1),
 			 SRC0_REL(ABSOLUTE),
 			 SRC0_ELEM(ELEM_X),
 			 SRC0_NEG(0),
-			 SRC1_SEL(0),
+			 SRC1_SEL(258),
 			 SRC1_REL(ABSOLUTE),
 			 SRC1_ELEM(ELEM_X),
 			 SRC1_NEG(0),
@@ -3198,24 +3190,24 @@ R600LoadShaders(ScrnInfoPtr pScrn, ScreenPtr pScreen)
 			     SRC1_ABS(0),
 			     UPDATE_EXECUTE_MASK(0),
 			     UPDATE_PRED(0),
-			     WRITE_MASK(1),
+			     WRITE_MASK(0),
 			     FOG_MERGE(0),
 			     OMOD(SQ_ALU_OMOD_OFF),
-			     ALU_INST(SQ_OP2_INST_MOV),
-			     BANK_SWIZZLE(SQ_ALU_VEC_210),
-			     DST_GPR(3),
+			     ALU_INST(SQ_OP2_INST_DOT4),
+			     BANK_SWIZZLE(SQ_ALU_VEC_102),
+			     DST_GPR(2),
 			     DST_REL(ABSOLUTE),
 			     DST_ELEM(ELEM_X),
-			     CLAMP(0));
+			     CLAMP(1));
     // 16 - alu 13
-    // MOV gpr[3].y gpr[2].y
-    ps[i++] = ALU_DWORD0(SRC0_SEL(2),
+    // DP4 gpr[2].y gpr[1].y c[2].y
+    ps[i++] = ALU_DWORD0(SRC0_SEL(1),
 			 SRC0_REL(ABSOLUTE),
 			 SRC0_ELEM(ELEM_Y),
 			 SRC0_NEG(0),
-			 SRC1_SEL(0),
+			 SRC1_SEL(258),
 			 SRC1_REL(ABSOLUTE),
-			 SRC1_ELEM(ELEM_X),
+			 SRC1_ELEM(ELEM_Y),
 			 SRC1_NEG(0),
 			 INDEX_MODE(SQ_INDEX_LOOP),
 			 PRED_SEL(SQ_PRED_SEL_OFF),
@@ -3225,24 +3217,24 @@ R600LoadShaders(ScrnInfoPtr pScrn, ScreenPtr pScreen)
 			     SRC1_ABS(0),
 			     UPDATE_EXECUTE_MASK(0),
 			     UPDATE_PRED(0),
-			     WRITE_MASK(1),
+			     WRITE_MASK(0),
 			     FOG_MERGE(0),
 			     OMOD(SQ_ALU_OMOD_OFF),
-			     ALU_INST(SQ_OP2_INST_MOV),
-			     BANK_SWIZZLE(SQ_ALU_VEC_210),
-			     DST_GPR(3),
+			     ALU_INST(SQ_OP2_INST_DOT4),
+			     BANK_SWIZZLE(SQ_ALU_VEC_102),
+			     DST_GPR(2),
 			     DST_REL(ABSOLUTE),
 			     DST_ELEM(ELEM_Y),
-			     CLAMP(0));
+			     CLAMP(1));
     // 17 - alu 14
-    // MOV gpr[3].z gpr[2].z
-    ps[i++] = ALU_DWORD0(SRC0_SEL(2),
+    // DP4 gpr[2].z gpr[1].z c[2].z
+    ps[i++] = ALU_DWORD0(SRC0_SEL(1),
 			 SRC0_REL(ABSOLUTE),
 			 SRC0_ELEM(ELEM_Z),
 			 SRC0_NEG(0),
-			 SRC1_SEL(0),
+			 SRC1_SEL(258),
 			 SRC1_REL(ABSOLUTE),
-			 SRC1_ELEM(ELEM_X),
+			 SRC1_ELEM(ELEM_Z),
 			 SRC1_NEG(0),
 			 INDEX_MODE(SQ_INDEX_LOOP),
 			 PRED_SEL(SQ_PRED_SEL_OFF),
@@ -3255,21 +3247,21 @@ R600LoadShaders(ScrnInfoPtr pScrn, ScreenPtr pScreen)
 			     WRITE_MASK(1),
 			     FOG_MERGE(0),
 			     OMOD(SQ_ALU_OMOD_OFF),
-			     ALU_INST(SQ_OP2_INST_MOV),
-			     BANK_SWIZZLE(SQ_ALU_VEC_210),
-			     DST_GPR(3),
+			     ALU_INST(SQ_OP2_INST_DOT4),
+			     BANK_SWIZZLE(SQ_ALU_VEC_102),
+			     DST_GPR(2),
 			     DST_REL(ABSOLUTE),
 			     DST_ELEM(ELEM_Z),
-			     CLAMP(0));
+			     CLAMP(1));
     // 18 - alu 15
-    // MOV gpr[3].w gpr[2].w
-    ps[i++] = ALU_DWORD0(SRC0_SEL(2),
+    // DP4 gpr[2].w gpr[1].w c[2].w
+    ps[i++] = ALU_DWORD0(SRC0_SEL(1),
 			 SRC0_REL(ABSOLUTE),
 			 SRC0_ELEM(ELEM_W),
 			 SRC0_NEG(0),
-			 SRC1_SEL(0),
+			 SRC1_SEL(258),
 			 SRC1_REL(ABSOLUTE),
-			 SRC1_ELEM(ELEM_X),
+			 SRC1_ELEM(ELEM_W),
 			 SRC1_NEG(0),
 			 INDEX_MODE(SQ_INDEX_LOOP),
 			 PRED_SEL(SQ_PRED_SEL_OFF),
@@ -3279,15 +3271,15 @@ R600LoadShaders(ScrnInfoPtr pScrn, ScreenPtr pScreen)
 			     SRC1_ABS(0),
 			     UPDATE_EXECUTE_MASK(0),
 			     UPDATE_PRED(0),
-			     WRITE_MASK(1),
+			     WRITE_MASK(0),
 			     FOG_MERGE(0),
 			     OMOD(SQ_ALU_OMOD_OFF),
-			     ALU_INST(SQ_OP2_INST_MOV),
-			     BANK_SWIZZLE(SQ_ALU_VEC_012),
-			     DST_GPR(3),
+			     ALU_INST(SQ_OP2_INST_DOT4),
+			     BANK_SWIZZLE(SQ_ALU_VEC_021),
+			     DST_GPR(2),
 			     DST_REL(ABSOLUTE),
 			     DST_ELEM(ELEM_W),
-			     CLAMP(0));
+			     CLAMP(1));
     // 19 - alignment
     ps[i++] = 0x00000000;
     ps[i++] = 0x00000000;
diff --git a/src/r600_textured_videofuncs.c b/src/r600_textured_videofuncs.c
index 4a7391c8..c06512a5 100644
--- a/src/r600_textured_videofuncs.c
+++ b/src/r600_textured_videofuncs.c
@@ -202,7 +202,7 @@ R600DisplayTexturedVideo(ScrnInfoPtr pScrn, RADEONPortPrivPtr pPriv)
 			accel_state->ps_size, accel_state->ps_mc_addr);
 
     ps_conf.shader_addr         = accel_state->ps_mc_addr;
-    ps_conf.num_gprs            = 4;
+    ps_conf.num_gprs            = 3;
     ps_conf.stack_size          = 0;
     ps_conf.uncached_first_inst = 1;
     ps_conf.clamp_consts        = 0;