diff options
| author | Nicolai Hähnle <[email protected]> | 2009-09-06 13:15:04 +0200 |
|---|---|---|
| committer | Nicolai Hähnle <[email protected]> | 2009-09-06 13:15:04 +0200 |
| commit | f02f63997ce65530788a6dfcb28f11790a14d938 (patch) | |
| tree | fc6aedb5256bfb84eb170cb82addd2b6605510f8 /src/gallium | |
| parent | e95e76e1255a3ad0ce604271301d090337b2e82b (diff) | |
| parent | 9778731732b4753e79a1b786c65325a52392411d (diff) | |
Merge branch 'master' into r300-compiler
Conflicts:
src/gallium/drivers/r300/r300_tgsi_to_rc.c
Diffstat (limited to 'src/gallium')
97 files changed, 3913 insertions, 2753 deletions
diff --git a/src/gallium/auxiliary/tgsi/tgsi-instruction-set.txt b/src/gallium/auxiliary/tgsi/tgsi-instruction-set.txt index 802ec371189..eb492076b7d 100644 --- a/src/gallium/auxiliary/tgsi/tgsi-instruction-set.txt +++ b/src/gallium/auxiliary/tgsi/tgsi-instruction-set.txt @@ -187,11 +187,7 @@ TGSI Instruction Specification 1.2.6 CND0 - Condition Zero - dst.x = (src2.x >= 0.0) ? src0.x : src1.x - dst.y = (src2.y >= 0.0) ? src0.y : src1.y - dst.z = (src2.z >= 0.0) ? src0.z : src1.z - dst.w = (src2.w >= 0.0) ? src0.w : src1.w - + Removed. Use (CMP src2, src1, src0) instead. 1.2.7 DOT2ADD - 2-component Dot Product And Add @@ -1031,12 +1027,12 @@ TGSI Instruction Specification 1.18.1 EXPP - Approximate Exponential Base 2 - Alias for EXP. + Use EXP. See also 1.19.3. 1.18.2 LOGP - Logarithm Base 2 - Alias for LG2. + Use LOG. See also 1.19.4. 1.19 vs_2_0 @@ -1053,6 +1049,16 @@ TGSI Instruction Specification Alias for ARR. +1.19.3 EXPP - Approximate Exponential Base 2 + + Use EX2. + + +1.19.4 LOGP - Logarithm Base 2 + + Use LG2. + + 2 Explanation of symbols used ============================== diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump.c b/src/gallium/auxiliary/tgsi/tgsi_dump.c index 05b07a3a73e..111d95b6665 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/tgsi_dump.c @@ -43,6 +43,7 @@ struct dump_ctx struct tgsi_iterate_context iter; uint instno; + int indent; uint indentation; @@ -335,14 +336,6 @@ tgsi_dump_immediate( iter_immediate( &ctx.iter, (struct tgsi_full_immediate *)imm ); } -static void -indent(struct dump_ctx *ctx) -{ - uint i; - for (i = 0; i < ctx->indentation; i++) - TXT(" "); -} - static boolean iter_instruction( struct tgsi_iterate_context *iter, @@ -350,22 +343,19 @@ iter_instruction( { struct dump_ctx *ctx = (struct dump_ctx *) iter; uint instno = ctx->instno++; - + const struct tgsi_opcode_info *info = tgsi_get_opcode_info( inst->Instruction.Opcode ); uint i; boolean first_reg = TRUE; INSTID( instno ); TXT( ": " ); - - /* update indentation */ - if (inst->Instruction.Opcode == TGSI_OPCODE_ENDIF || - inst->Instruction.Opcode == TGSI_OPCODE_ENDFOR || - inst->Instruction.Opcode == TGSI_OPCODE_ENDLOOP) { - ctx->indentation -= indent_spaces; - } - indent(ctx); - - TXT( tgsi_get_opcode_info( inst->Instruction.Opcode )->mnemonic ); + + ctx->indent -= info->pre_dedent; + for(i = 0; (int)i < ctx->indent; ++i) + TXT( " " ); + ctx->indent += info->post_indent; + + TXT( info->mnemonic ); switch (inst->Instruction.Saturate) { case TGSI_SAT_NONE: @@ -526,6 +516,7 @@ tgsi_dump_instruction( struct dump_ctx ctx; ctx.instno = instno; + ctx.indent = 0; ctx.printf = dump_ctx_printf; ctx.indentation = 0; @@ -559,6 +550,7 @@ tgsi_dump( ctx.iter.epilog = NULL; ctx.instno = 0; + ctx.indent = 0; ctx.printf = dump_ctx_printf; ctx.indentation = 0; @@ -612,6 +604,7 @@ tgsi_dump_str( ctx.base.iter.epilog = NULL; ctx.base.instno = 0; + ctx.base.indent = 0; ctx.base.printf = &str_dump_ctx_printf; ctx.base.indentation = 0; diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c index 711e86d6edf..60ffa188df6 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c @@ -2329,16 +2329,6 @@ exec_instruction( } break; - case TGSI_OPCODE_CND0: - FOR_EACH_ENABLED_CHANNEL(*inst, chan_index) { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); - FETCH(&r[2], 2, chan_index); - micro_le(&r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[2], &r[0], &r[1]); - STORE(&r[0], 0, chan_index); - } - break; - case TGSI_OPCODE_DP2A: FETCH( &r[0], 0, CHAN_X ); FETCH( &r[1], 1, CHAN_X ); @@ -3104,6 +3094,12 @@ exec_instruction( break; case TGSI_OPCODE_BGNFOR: + assert(mach->LoopCounterStackTop < TGSI_EXEC_MAX_LOOP_NESTING); + for (chan_index = 0; chan_index < 3; chan_index++) { + FETCH( &mach->LoopCounterStack[mach->LoopCounterStackTop].xyzw[chan_index], 0, chan_index ); + } + STORE( &mach->LoopCounterStack[mach->LoopCounterStackTop].xyzw[CHAN_Y], 0, CHAN_X ); + ++mach->LoopCounterStackTop; /* fall-through (for now) */ case TGSI_OPCODE_BGNLOOP: /* push LoopMask and ContMasks */ @@ -3111,10 +3107,58 @@ exec_instruction( mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask; assert(mach->ContStackTop < TGSI_EXEC_MAX_LOOP_NESTING); mach->ContStack[mach->ContStackTop++] = mach->ContMask; + assert(mach->LoopLabelStackTop < TGSI_EXEC_MAX_LOOP_NESTING); + mach->LoopLabelStack[mach->LoopLabelStackTop++] = *pc - 1; break; case TGSI_OPCODE_ENDFOR: - /* fall-through (for now at least) */ + assert(mach->LoopCounterStackTop > 0); + micro_sub( &mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_X], + &mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_X], + &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); + /* update LoopMask */ + if( mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_X].f[0] <= 0) { + mach->LoopMask &= ~0x1; + } + if( mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_X].f[1] <= 0 ) { + mach->LoopMask &= ~0x2; + } + if( mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_X].f[2] <= 0 ) { + mach->LoopMask &= ~0x4; + } + if( mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_X].f[3] <= 0 ) { + mach->LoopMask &= ~0x8; + } + micro_add( &mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_Y], + &mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_Y], + &mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[CHAN_Z]); + assert(mach->LoopLabelStackTop > 0); + inst = mach->Instructions + mach->LoopLabelStack[mach->LoopLabelStackTop - 1]; + STORE( &mach->LoopCounterStack[mach->LoopCounterStackTop].xyzw[CHAN_Y], 0, CHAN_X ); + /* Restore ContMask, but don't pop */ + assert(mach->ContStackTop > 0); + mach->ContMask = mach->ContStack[mach->ContStackTop - 1]; + UPDATE_EXEC_MASK(mach); + if (mach->ExecMask) { + /* repeat loop: jump to instruction just past BGNLOOP */ + assert(mach->LoopLabelStackTop > 0); + *pc = mach->LoopLabelStack[mach->LoopLabelStackTop - 1] + 1; + } + else { + /* exit loop: pop LoopMask */ + assert(mach->LoopStackTop > 0); + mach->LoopMask = mach->LoopStack[--mach->LoopStackTop]; + /* pop ContMask */ + assert(mach->ContStackTop > 0); + mach->ContMask = mach->ContStack[--mach->ContStackTop]; + assert(mach->LoopLabelStackTop > 0); + --mach->LoopLabelStackTop; + assert(mach->LoopCounterStackTop > 0); + --mach->LoopCounterStackTop; + } + UPDATE_EXEC_MASK(mach); + break; + case TGSI_OPCODE_ENDLOOP: /* Restore ContMask, but don't pop */ assert(mach->ContStackTop > 0); @@ -3122,7 +3166,8 @@ exec_instruction( UPDATE_EXEC_MASK(mach); if (mach->ExecMask) { /* repeat loop: jump to instruction just past BGNLOOP */ - *pc = inst->InstructionExtLabel.Label + 1; + assert(mach->LoopLabelStackTop > 0); + *pc = mach->LoopLabelStack[mach->LoopLabelStackTop - 1] + 1; } else { /* exit loop: pop LoopMask */ @@ -3131,6 +3176,8 @@ exec_instruction( /* pop ContMask */ assert(mach->ContStackTop > 0); mach->ContMask = mach->ContStack[--mach->ContStackTop]; + assert(mach->LoopLabelStackTop > 0); + --mach->LoopLabelStackTop; } UPDATE_EXEC_MASK(mach); break; diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.h b/src/gallium/auxiliary/tgsi/tgsi_exec.h index fd9ef6f35df..3baa94dbdde 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_exec.h +++ b/src/gallium/auxiliary/tgsi/tgsi_exec.h @@ -232,6 +232,14 @@ struct tgsi_exec_machine uint LoopStack[TGSI_EXEC_MAX_LOOP_NESTING]; int LoopStackTop; + /** Loop label stack */ + uint LoopLabelStack[TGSI_EXEC_MAX_LOOP_NESTING]; + int LoopLabelStackTop; + + /** Loop counter stack (x = count, y = current, z = step) */ + struct tgsi_exec_vector LoopCounterStack[TGSI_EXEC_MAX_LOOP_NESTING]; + int LoopCounterStackTop; + /** Loop continue mask stack (see comments in tgsi_exec.c) */ uint ContStack[TGSI_EXEC_MAX_LOOP_NESTING]; int ContStackTop; diff --git a/src/gallium/auxiliary/tgsi/tgsi_info.c b/src/gallium/auxiliary/tgsi/tgsi_info.c index ccf4b205ffb..17af4cb7ad2 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_info.c +++ b/src/gallium/auxiliary/tgsi/tgsi_info.c @@ -31,125 +31,125 @@ static const struct tgsi_opcode_info opcode_info[TGSI_OPCODE_LAST] = { - { 1, 1, 0, 0, "ARL", TGSI_OPCODE_ARL }, - { 1, 1, 0, 0, "MOV", TGSI_OPCODE_MOV }, - { 1, 1, 0, 0, "LIT", TGSI_OPCODE_LIT }, - { 1, 1, 0, 0, "RCP", TGSI_OPCODE_RCP }, - { 1, 1, 0, 0, "RSQ", TGSI_OPCODE_RSQ }, - { 1, 1, 0, 0, "EXP", TGSI_OPCODE_EXP }, - { 1, 1, 0, 0, "LOG", TGSI_OPCODE_LOG }, - { 1, 2, 0, 0, "MUL", TGSI_OPCODE_MUL }, - { 1, 2, 0, 0, "ADD", TGSI_OPCODE_ADD }, - { 1, 2, 0, 0, "DP3", TGSI_OPCODE_DP3 }, - { 1, 2, 0, 0, "DP4", TGSI_OPCODE_DP4 }, - { 1, 2, 0, 0, "DST", TGSI_OPCODE_DST }, - { 1, 2, 0, 0, "MIN", TGSI_OPCODE_MIN }, - { 1, 2, 0, 0, "MAX", TGSI_OPCODE_MAX }, - { 1, 2, 0, 0, "SLT", TGSI_OPCODE_SLT }, - { 1, 2, 0, 0, "SGE", TGSI_OPCODE_SGE }, - { 1, 3, 0, 0, "MAD", TGSI_OPCODE_MAD }, - { 1, 2, 0, 0, "SUB", TGSI_OPCODE_SUB }, - { 1, 3, 0, 0, "LRP", TGSI_OPCODE_LRP }, - { 1, 3, 0, 0, "CND", TGSI_OPCODE_CND }, - { 1, 3, 0, 0, "CND0", TGSI_OPCODE_CND0 }, - { 1, 3, 0, 0, "DP2A", TGSI_OPCODE_DP2A }, - { 0, 0, 0, 0, "", 22 }, /* removed */ - { 0, 0, 0, 0, "", 23 }, /* removed */ - { 1, 1, 0, 0, "FRC", TGSI_OPCODE_FRC }, - { 1, 3, 0, 0, "CLAMP", TGSI_OPCODE_CLAMP }, - { 1, 1, 0, 0, "FLR", TGSI_OPCODE_FLR }, - { 1, 1, 0, 0, "ROUND", TGSI_OPCODE_ROUND }, - { 1, 1, 0, 0, "EX2", TGSI_OPCODE_EX2 }, - { 1, 1, 0, 0, "LG2", TGSI_OPCODE_LG2 }, - { 1, 2, 0, 0, "POW", TGSI_OPCODE_POW }, - { 1, 2, 0, 0, "XPD", TGSI_OPCODE_XPD }, - { 0, 0, 0, 0, "", 32 }, /* removed */ - { 1, 1, 0, 0, "ABS", TGSI_OPCODE_ABS }, - { 1, 1, 0, 0, "RCC", TGSI_OPCODE_RCC }, - { 1, 2, 0, 0, "DPH", TGSI_OPCODE_DPH }, - { 1, 1, 0, 0, "COS", TGSI_OPCODE_COS }, - { 1, 1, 0, 0, "DDX", TGSI_OPCODE_DDX }, - { 1, 1, 0, 0, "DDY", TGSI_OPCODE_DDY }, - { 0, 0, 0, 0, "KILP", TGSI_OPCODE_KILP }, - { 1, 1, 0, 0, "PK2H", TGSI_OPCODE_PK2H }, - { 1, 1, 0, 0, "PK2US", TGSI_OPCODE_PK2US }, - { 1, 1, 0, 0, "PK4B", TGSI_OPCODE_PK4B }, - { 1, 1, 0, 0, "PK4UB", TGSI_OPCODE_PK4UB }, - { 1, 2, 0, 0, "RFL", TGSI_OPCODE_RFL }, - { 1, 2, 0, 0, "SEQ", TGSI_OPCODE_SEQ }, - { 1, 2, 0, 0, "SFL", TGSI_OPCODE_SFL }, - { 1, 2, 0, 0, "SGT", TGSI_OPCODE_SGT }, - { 1, 1, 0, 0, "SIN", TGSI_OPCODE_SIN }, - { 1, 2, 0, 0, "SLE", TGSI_OPCODE_SLE }, - { 1, 2, 0, 0, "SNE", TGSI_OPCODE_SNE }, - { 1, 2, 0, 0, "STR", TGSI_OPCODE_STR }, - { 1, 2, 1, 0, "TEX", TGSI_OPCODE_TEX }, - { 1, 4, 1, 0, "TXD", TGSI_OPCODE_TXD }, - { 1, 2, 1, 0, "TXP", TGSI_OPCODE_TXP }, - { 1, 1, 0, 0, "UP2H", TGSI_OPCODE_UP2H }, - { 1, 1, 0, 0, "UP2US", TGSI_OPCODE_UP2US }, - { 1, 1, 0, 0, "UP4B", TGSI_OPCODE_UP4B }, - { 1, 1, 0, 0, "UP4UB", TGSI_OPCODE_UP4UB }, - { 1, 3, 0, 0, "X2D", TGSI_OPCODE_X2D }, - { 1, 1, 0, 0, "ARA", TGSI_OPCODE_ARA }, - { 1, 1, 0, 0, "ARR", TGSI_OPCODE_ARR }, - { 0, 1, 0, 0, "BRA", TGSI_OPCODE_BRA }, - { 0, 0, 0, 1, "CAL", TGSI_OPCODE_CAL }, - { 0, 0, 0, 0, "RET", TGSI_OPCODE_RET }, - { 1, 1, 0, 0, "SSG", TGSI_OPCODE_SSG }, - { 1, 3, 0, 0, "CMP", TGSI_OPCODE_CMP }, - { 1, 1, 0, 0, "SCS", TGSI_OPCODE_SCS }, - { 1, 2, 1, 0, "TXB", TGSI_OPCODE_TXB }, - { 1, 1, 0, 0, "NRM", TGSI_OPCODE_NRM }, - { 1, 2, 0, 0, "DIV", TGSI_OPCODE_DIV }, - { 1, 2, 0, 0, "DP2", TGSI_OPCODE_DP2 }, - { 1, 2, 1, 0, "TXL", TGSI_OPCODE_TXL }, - { 0, 0, 0, 0, "BRK", TGSI_OPCODE_BRK }, - { 0, 1, 0, 1, "IF", TGSI_OPCODE_IF }, - { 1, 1, 0, 0, "BGNFOR", TGSI_OPCODE_BGNFOR }, - { 0, 1, 0, 0, "REP", TGSI_OPCODE_REP }, - { 0, 0, 0, 1, "ELSE", TGSI_OPCODE_ELSE }, - { 0, 0, 0, 0, "ENDIF", TGSI_OPCODE_ENDIF }, - { 1, 0, 0, 0, "ENDFOR", TGSI_OPCODE_ENDFOR }, - { 0, 0, 0, 0, "ENDREP", TGSI_OPCODE_ENDREP }, - { 0, 1, 0, 0, "PUSHA", TGSI_OPCODE_PUSHA }, - { 1, 0, 0, 0, "POPA", TGSI_OPCODE_POPA }, - { 1, 1, 0, 0, "CEIL", TGSI_OPCODE_CEIL }, - { 1, 1, 0, 0, "I2F", TGSI_OPCODE_I2F }, - { 1, 1, 0, 0, "NOT", TGSI_OPCODE_NOT }, - { 1, 1, 0, 0, "TRUNC", TGSI_OPCODE_TRUNC }, - { 1, 2, 0, 0, "SHL", TGSI_OPCODE_SHL }, - { 1, 2, 0, 0, "SHR", TGSI_OPCODE_SHR }, - { 1, 2, 0, 0, "AND", TGSI_OPCODE_AND }, - { 1, 2, 0, 0, "OR", TGSI_OPCODE_OR }, - { 1, 2, 0, 0, "MOD", TGSI_OPCODE_MOD }, - { 1, 2, 0, 0, "XOR", TGSI_OPCODE_XOR }, - { 1, 3, 0, 0, "SAD", TGSI_OPCODE_SAD }, - { 1, 2, 1, 0, "TXF", TGSI_OPCODE_TXF }, - { 1, 2, 1, 0, "TXQ", TGSI_OPCODE_TXQ }, - { 0, 0, 0, 0, "CONT", TGSI_OPCODE_CONT }, - { 0, 0, 0, 0, "EMIT", TGSI_OPCODE_EMIT }, - { 0, 0, 0, 0, "ENDPRIM", TGSI_OPCODE_ENDPRIM }, - { 0, 0, 0, 1, "BGNLOOP", TGSI_OPCODE_BGNLOOP }, - { 0, 0, 0, 0, "BGNSUB", TGSI_OPCODE_BGNSUB }, - { 0, 0, 0, 1, "ENDLOOP", TGSI_OPCODE_ENDLOOP }, - { 0, 0, 0, 0, "ENDSUB", TGSI_OPCODE_ENDSUB }, - { 1, 1, 0, 0, "NOISE1", TGSI_OPCODE_NOISE1 }, - { 1, 1, 0, 0, "NOISE2", TGSI_OPCODE_NOISE2 }, - { 1, 1, 0, 0, "NOISE3", TGSI_OPCODE_NOISE3 }, - { 1, 1, 0, 0, "NOISE4", TGSI_OPCODE_NOISE4 }, - { 0, 0, 0, 0, "NOP", TGSI_OPCODE_NOP }, - { 0, 0, 0, 0, "", 108 }, /* removed */ - { 0, 0, 0, 0, "", 109 }, /* removed */ - { 0, 0, 0, 0, "", 110 }, /* removed */ - { 0, 0, 0, 0, "", 111 }, /* removed */ - { 1, 1, 0, 0, "NRM4", TGSI_OPCODE_NRM4 }, - { 0, 1, 0, 0, "CALLNZ", TGSI_OPCODE_CALLNZ }, - { 0, 1, 0, 0, "IFC", TGSI_OPCODE_IFC }, - { 0, 1, 0, 0, "BREAKC", TGSI_OPCODE_BREAKC }, - { 0, 1, 0, 0, "KIL", TGSI_OPCODE_KIL }, - { 0, 0, 0, 0, "END", TGSI_OPCODE_END }, - { 1, 1, 0, 0, "SWZ", TGSI_OPCODE_SWZ } + { 1, 1, 0, 0, 0, 0, "ARL", TGSI_OPCODE_ARL }, + { 1, 1, 0, 0, 0, 0, "MOV", TGSI_OPCODE_MOV }, + { 1, 1, 0, 0, 0, 0, "LIT", TGSI_OPCODE_LIT }, + { 1, 1, 0, 0, 0, 0, "RCP", TGSI_OPCODE_RCP }, + { 1, 1, 0, 0, 0, 0, "RSQ", TGSI_OPCODE_RSQ }, + { 1, 1, 0, 0, 0, 0, "EXP", TGSI_OPCODE_EXP }, + { 1, 1, 0, 0, 0, 0, "LOG", TGSI_OPCODE_LOG }, + { 1, 2, 0, 0, 0, 0, "MUL", TGSI_OPCODE_MUL }, + { 1, 2, 0, 0, 0, 0, "ADD", TGSI_OPCODE_ADD }, + { 1, 2, 0, 0, 0, 0, "DP3", TGSI_OPCODE_DP3 }, + { 1, 2, 0, 0, 0, 0, "DP4", TGSI_OPCODE_DP4 }, + { 1, 2, 0, 0, 0, 0, "DST", TGSI_OPCODE_DST }, + { 1, 2, 0, 0, 0, 0, "MIN", TGSI_OPCODE_MIN }, + { 1, 2, 0, 0, 0, 0, "MAX", TGSI_OPCODE_MAX }, + { 1, 2, 0, 0, 0, 0, "SLT", TGSI_OPCODE_SLT }, + { 1, 2, 0, 0, 0, 0, "SGE", TGSI_OPCODE_SGE }, + { 1, 3, 0, 0, 0, 0, "MAD", TGSI_OPCODE_MAD }, + { 1, 2, 0, 0, 0, 0, "SUB", TGSI_OPCODE_SUB }, + { 1, 3, 0, 0, 0, 0, "LRP", TGSI_OPCODE_LRP }, + { 1, 3, 0, 0, 0, 0, "CND", TGSI_OPCODE_CND }, + { 0, 0, 0, 0, 0, 0, "", 20 }, /* removed */ + { 1, 3, 0, 0, 0, 0, "DP2A", TGSI_OPCODE_DP2A }, + { 0, 0, 0, 0, 0, 0, "", 22 }, /* removed */ + { 0, 0, 0, 0, 0, 0, "", 23 }, /* removed */ + { 1, 1, 0, 0, 0, 0, "FRC", TGSI_OPCODE_FRC }, + { 1, 3, 0, 0, 0, 0, "CLAMP", TGSI_OPCODE_CLAMP }, + { 1, 1, 0, 0, 0, 0, "FLR", TGSI_OPCODE_FLR }, + { 1, 1, 0, 0, 0, 0, "ROUND", TGSI_OPCODE_ROUND }, + { 1, 1, 0, 0, 0, 0, "EX2", TGSI_OPCODE_EX2 }, + { 1, 1, 0, 0, 0, 0, "LG2", TGSI_OPCODE_LG2 }, + { 1, 2, 0, 0, 0, 0, "POW", TGSI_OPCODE_POW }, + { 1, 2, 0, 0, 0, 0, "XPD", TGSI_OPCODE_XPD }, + { 0, 0, 0, 0, 0, 0, "", 32 }, /* removed */ + { 1, 1, 0, 0, 0, 0, "ABS", TGSI_OPCODE_ABS }, + { 1, 1, 0, 0, 0, 0, "RCC", TGSI_OPCODE_RCC }, + { 1, 2, 0, 0, 0, 0, "DPH", TGSI_OPCODE_DPH }, + { 1, 1, 0, 0, 0, 0, "COS", TGSI_OPCODE_COS }, + { 1, 1, 0, 0, 0, 0, "DDX", TGSI_OPCODE_DDX }, + { 1, 1, 0, 0, 0, 0, "DDY", TGSI_OPCODE_DDY }, + { 0, 0, 0, 0, 0, 0, "KILP", TGSI_OPCODE_KILP }, + { 1, 1, 0, 0, 0, 0, "PK2H", TGSI_OPCODE_PK2H }, + { 1, 1, 0, 0, 0, 0, "PK2US", TGSI_OPCODE_PK2US }, + { 1, 1, 0, 0, 0, 0, "PK4B", TGSI_OPCODE_PK4B }, + { 1, 1, 0, 0, 0, 0, "PK4UB", TGSI_OPCODE_PK4UB }, + { 1, 2, 0, 0, 0, 0, "RFL", TGSI_OPCODE_RFL }, + { 1, 2, 0, 0, 0, 0, "SEQ", TGSI_OPCODE_SEQ }, + { 1, 2, 0, 0, 0, 0, "SFL", TGSI_OPCODE_SFL }, + { 1, 2, 0, 0, 0, 0, "SGT", TGSI_OPCODE_SGT }, + { 1, 1, 0, 0, 0, 0, "SIN", TGSI_OPCODE_SIN }, + { 1, 2, 0, 0, 0, 0, "SLE", TGSI_OPCODE_SLE }, + { 1, 2, 0, 0, 0, 0, "SNE", TGSI_OPCODE_SNE }, + { 1, 2, 0, 0, 0, 0, "STR", TGSI_OPCODE_STR }, + { 1, 2, 1, 0, 0, 0, "TEX", TGSI_OPCODE_TEX }, + { 1, 4, 1, 0, 0, 0, "TXD", TGSI_OPCODE_TXD }, + { 1, 2, 1, 0, 0, 0, "TXP", TGSI_OPCODE_TXP }, + { 1, 1, 0, 0, 0, 0, "UP2H", TGSI_OPCODE_UP2H }, + { 1, 1, 0, 0, 0, 0, "UP2US", TGSI_OPCODE_UP2US }, + { 1, 1, 0, 0, 0, 0, "UP4B", TGSI_OPCODE_UP4B }, + { 1, 1, 0, 0, 0, 0, "UP4UB", TGSI_OPCODE_UP4UB }, + { 1, 3, 0, 0, 0, 0, "X2D", TGSI_OPCODE_X2D }, + { 1, 1, 0, 0, 0, 0, "ARA", TGSI_OPCODE_ARA }, + { 1, 1, 0, 0, 0, 0, "ARR", TGSI_OPCODE_ARR }, + { 0, 1, 0, 0, 0, 0, "BRA", TGSI_OPCODE_BRA }, + { 0, 0, 0, 1, 0, 0, "CAL", TGSI_OPCODE_CAL }, + { 0, 0, 0, 0, 0, 0, "RET", TGSI_OPCODE_RET }, + { 1, 1, 0, 0, 0, 0, "SSG", TGSI_OPCODE_SSG }, + { 1, 3, 0, 0, 0, 0, "CMP", TGSI_OPCODE_CMP }, + { 1, 1, 0, 0, 0, 0, "SCS", TGSI_OPCODE_SCS }, + { 1, 2, 1, 0, 0, 0, "TXB", TGSI_OPCODE_TXB }, + { 1, 1, 0, 0, 0, 0, "NRM", TGSI_OPCODE_NRM }, + { 1, 2, 0, 0, 0, 0, "DIV", TGSI_OPCODE_DIV }, + { 1, 2, 0, 0, 0, 0, "DP2", TGSI_OPCODE_DP2 }, + { 1, 2, 1, 0, 0, 0, "TXL", TGSI_OPCODE_TXL }, + { 0, 0, 0, 0, 0, 0, "BRK", TGSI_OPCODE_BRK }, + { 0, 1, 0, 1, 0, 1, "IF", TGSI_OPCODE_IF }, + { 1, 1, 0, 0, 0, 1, "BGNFOR", TGSI_OPCODE_BGNFOR }, + { 0, 1, 0, 0, 0, 1, "REP", TGSI_OPCODE_REP }, + { 0, 0, 0, 1, 1, 1, "ELSE", TGSI_OPCODE_ELSE }, + { 0, 0, 0, 0, 1, 0, "ENDIF", TGSI_OPCODE_ENDIF }, + { 1, 0, 0, 0, 1, 0, "ENDFOR", TGSI_OPCODE_ENDFOR }, + { 0, 0, 0, 0, 1, 0, "ENDREP", TGSI_OPCODE_ENDREP }, + { 0, 1, 0, 0, 0, 0, "PUSHA", TGSI_OPCODE_PUSHA }, + { 1, 0, 0, 0, 0, 0, "POPA", TGSI_OPCODE_POPA }, + { 1, 1, 0, 0, 0, 0, "CEIL", TGSI_OPCODE_CEIL }, + { 1, 1, 0, 0, 0, 0, "I2F", TGSI_OPCODE_I2F }, + { 1, 1, 0, 0, 0, 0, "NOT", TGSI_OPCODE_NOT }, + { 1, 1, 0, 0, 0, 0, "TRUNC", TGSI_OPCODE_TRUNC }, + { 1, 2, 0, 0, 0, 0, "SHL", TGSI_OPCODE_SHL }, + { 1, 2, 0, 0, 0, 0, "SHR", TGSI_OPCODE_SHR }, + { 1, 2, 0, 0, 0, 0, "AND", TGSI_OPCODE_AND }, + { 1, 2, 0, 0, 0, 0, "OR", TGSI_OPCODE_OR }, + { 1, 2, 0, 0, 0, 0, "MOD", TGSI_OPCODE_MOD }, + { 1, 2, 0, 0, 0, 0, "XOR", TGSI_OPCODE_XOR }, + { 1, 3, 0, 0, 0, 0, "SAD", TGSI_OPCODE_SAD }, + { 1, 2, 1, 0, 0, 0, "TXF", TGSI_OPCODE_TXF }, + { 1, 2, 1, 0, 0, 0, "TXQ", TGSI_OPCODE_TXQ }, + { 0, 0, 0, 0, 0, 0, "CONT", TGSI_OPCODE_CONT }, + { 0, 0, 0, 0, 0, 0, "EMIT", TGSI_OPCODE_EMIT }, + { 0, 0, 0, 0, 0, 0, "ENDPRIM", TGSI_OPCODE_ENDPRIM }, + { 0, 0, 0, 1, 0, 1, "BGNLOOP", TGSI_OPCODE_BGNLOOP }, + { 0, 0, 0, 0, 0, 1, "BGNSUB", TGSI_OPCODE_BGNSUB }, + { 0, 0, 0, 1, 1, 0, "ENDLOOP", TGSI_OPCODE_ENDLOOP }, + { 0, 0, 0, 0, 1, 0, "ENDSUB", TGSI_OPCODE_ENDSUB }, + { 1, 1, 0, 0, 0, 0, "NOISE1", TGSI_OPCODE_NOISE1 }, + { 1, 1, 0, 0, 0, 0, "NOISE2", TGSI_OPCODE_NOISE2 }, + { 1, 1, 0, 0, 0, 0, "NOISE3", TGSI_OPCODE_NOISE3 }, + { 1, 1, 0, 0, 0, 0, "NOISE4", TGSI_OPCODE_NOISE4 }, + { 0, 0, 0, 0, 0, 0, "NOP", TGSI_OPCODE_NOP }, + { 0, 0, 0, 0, 0, 0, "", 108 }, /* removed */ + { 0, 0, 0, 0, 0, 0, "", 109 }, /* removed */ + { 0, 0, 0, 0, 0, 0, "", 110 }, /* removed */ + { 0, 0, 0, 0, 0, 0, "", 111 }, /* removed */ + { 1, 1, 0, 0, 0, 0, "NRM4", TGSI_OPCODE_NRM4 }, + { 0, 1, 0, 0, 0, 0, "CALLNZ", TGSI_OPCODE_CALLNZ }, + { 0, 1, 0, 0, 0, 0, "IFC", TGSI_OPCODE_IFC }, + { 0, 1, 0, 0, 0, 0, "BREAKC", TGSI_OPCODE_BREAKC }, + { 0, 1, 0, 0, 0, 0, "KIL", TGSI_OPCODE_KIL }, + { 0, 0, 0, 0, 0, 0, "END", TGSI_OPCODE_END }, + { 1, 1, 0, 0, 0, 0, "SWZ", TGSI_OPCODE_SWZ } }; const struct tgsi_opcode_info * diff --git a/src/gallium/auxiliary/tgsi/tgsi_info.h b/src/gallium/auxiliary/tgsi/tgsi_info.h index b2375c69710..74713c3b98a 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_info.h +++ b/src/gallium/auxiliary/tgsi/tgsi_info.h @@ -36,10 +36,12 @@ extern "C" { struct tgsi_opcode_info { - uint num_dst; - uint num_src; - boolean is_tex; - boolean is_branch; + unsigned num_dst:3; + unsigned num_src:3; + unsigned is_tex:1; + unsigned is_branch:1; + int pre_dedent:2; + int post_indent:2; const char *mnemonic; uint opcode; }; diff --git a/src/gallium/auxiliary/tgsi/tgsi_opcode_tmp.h b/src/gallium/auxiliary/tgsi/tgsi_opcode_tmp.h index ed594a3e2c7..e7bcf4bf754 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_opcode_tmp.h +++ b/src/gallium/auxiliary/tgsi/tgsi_opcode_tmp.h @@ -60,7 +60,6 @@ OP13(MAD) OP12(SUB) OP13(LRP) OP13(CND) -OP13(CND0) OP13(DP2A) OP11(FRC) OP13(CLAMP) diff --git a/src/gallium/auxiliary/tgsi/tgsi_sanity.c b/src/gallium/auxiliary/tgsi/tgsi_sanity.c index 4fe8553c423..8a13885da9b 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_sanity.c +++ b/src/gallium/auxiliary/tgsi/tgsi_sanity.c @@ -199,10 +199,10 @@ iter_instruction( } if (info->num_dst != inst->Instruction.NumDstRegs) { - report_error( ctx, "Invalid number of destination operands, should be %u", info->num_dst ); + report_error( ctx, "%s: Invalid number of destination operands, should be %u", info->mnemonic, info->num_dst ); } if (info->num_src != inst->Instruction.NumSrcRegs) { - report_error( ctx, "Invalid number of source operands, should be %u", info->num_src ); + report_error( ctx, "%s: Invalid number of source operands, should be %u", info->mnemonic, info->num_src ); } /* Check destination and source registers' validity. diff --git a/src/gallium/auxiliary/tgsi/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/tgsi_sse2.c index 46f2387c158..3cdf8b9f359 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/tgsi_sse2.c @@ -2089,10 +2089,6 @@ emit_instruction( return 0; break; - case TGSI_OPCODE_CND0: - return 0; - break; - case TGSI_OPCODE_DP2A: FETCH( func, *inst, 0, 0, CHAN_X ); /* xmm0 = src[0].x */ FETCH( func, *inst, 1, 1, CHAN_X ); /* xmm1 = src[1].x */ diff --git a/src/gallium/auxiliary/tgsi/tgsi_ureg.c b/src/gallium/auxiliary/tgsi/tgsi_ureg.c index c0a0627e0b2..f7096bd8e2c 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_ureg.c +++ b/src/gallium/auxiliary/tgsi/tgsi_ureg.c @@ -29,6 +29,7 @@ #include "pipe/p_context.h" #include "pipe/p_state.h" #include "tgsi/tgsi_ureg.h" +#include "tgsi/tgsi_info.h" #include "tgsi/tgsi_dump.h" #include "util/u_memory.h" #include "util/u_math.h" @@ -71,6 +72,7 @@ struct ureg_tokens { #define UREG_MAX_OUTPUT PIPE_MAX_ATTRIBS #define UREG_MAX_IMMEDIATE 32 #define UREG_MAX_TEMP 256 +#define UREG_MAX_ADDR 2 #define DOMAIN_DECL 0 #define DOMAIN_INSN 1 @@ -99,11 +101,15 @@ struct ureg_program } immediate[UREG_MAX_IMMEDIATE]; unsigned nr_immediates; + struct ureg_src sampler[PIPE_MAX_SAMPLERS]; + unsigned nr_samplers; + unsigned temps_active[UREG_MAX_TEMP / 32]; unsigned nr_temps; + unsigned nr_addrs; + unsigned nr_constants; - unsigned nr_samplers; unsigned nr_instructions; struct ureg_tokens domain[2]; @@ -187,6 +193,8 @@ ureg_dst_register( unsigned file, dst.File = file; dst.WriteMask = TGSI_WRITEMASK_XYZW; dst.Indirect = 0; + dst.IndirectIndex = 0; + dst.IndirectSwizzle = 0; dst.Saturate = 0; dst.Index = index; dst.Pad1 = 0; @@ -208,6 +216,8 @@ ureg_src_register( unsigned file, src.SwizzleW = TGSI_SWIZZLE_W; src.Pad = 0; src.Indirect = 0; + src.IndirectIndex = 0; + src.IndirectSwizzle = 0; src.Absolute = 0; src.Index = index; src.Negate = 0; @@ -254,6 +264,7 @@ ureg_DECL_fs_input( struct ureg_program *ureg, unsigned index, unsigned interp ) { + assert(ureg->processor == TGSI_PROCESSOR_FRAGMENT); return ureg_DECL_input( ureg, name, index, interp ); } @@ -263,6 +274,7 @@ ureg_DECL_vs_input( struct ureg_program *ureg, unsigned name, unsigned index ) { + assert(ureg->processor == TGSI_PROCESSOR_VERTEX); return ureg_DECL_input( ureg, name, index, TGSI_INTERPOLATE_CONSTANT ); } @@ -346,11 +358,36 @@ void ureg_release_temporary( struct ureg_program *ureg, } +/* Allocate a new address register. + */ +struct ureg_dst ureg_DECL_address( struct ureg_program *ureg ) +{ + if (ureg->nr_addrs < UREG_MAX_ADDR) + return ureg_dst_register( TGSI_FILE_ADDRESS, ureg->nr_addrs++ ); + + assert( 0 ); + return ureg_dst_register( TGSI_FILE_ADDRESS, 0 ); +} + /* Allocate a new sampler. */ -struct ureg_src ureg_DECL_sampler( struct ureg_program *ureg ) +struct ureg_src ureg_DECL_sampler( struct ureg_program *ureg, + unsigned nr ) { - return ureg_src_register( TGSI_FILE_SAMPLER, ureg->nr_samplers++ ); + unsigned i; + + for (i = 0; i < ureg->nr_samplers; i++) + if (ureg->sampler[i].Index == nr) + return ureg->sampler[i]; + + if (i < PIPE_MAX_SAMPLERS) { + ureg->sampler[i] = ureg_src_register( TGSI_FILE_SAMPLER, nr ); + ureg->nr_samplers++; + return ureg->sampler[i]; + } + + assert( 0 ); + return ureg->sampler[0]; } @@ -363,6 +400,8 @@ static int match_or_expand_immediate( const float *v, unsigned *swizzle ) { unsigned i, j; + + *swizzle = 0; for (i = 0; i < nr; i++) { boolean found = FALSE; @@ -394,8 +433,8 @@ struct ureg_src ureg_DECL_immediate( struct ureg_program *ureg, const float *v, unsigned nr ) { - unsigned i; - unsigned swizzle = 0; + unsigned i, j; + unsigned swizzle; /* Could do a first pass where we examine all existing immediates * without expanding. @@ -423,6 +462,12 @@ struct ureg_src ureg_DECL_immediate( struct ureg_program *ureg, set_bad( ureg ); out: + /* Make sure that all referenced elements are from this immediate. + * Has the effect of making size-one immediates into scalars. + */ + for (j = nr; j < 4; j++) + swizzle |= (swizzle & 0x3) << (j * 2); + return ureg_swizzle( ureg_src_register( TGSI_FILE_IMMEDIATE, i ), (swizzle >> 0) & 0x3, (swizzle >> 2) & 0x3, @@ -442,31 +487,39 @@ ureg_emit_src( struct ureg_program *ureg, union tgsi_any_token *out = get_tokens( ureg, DOMAIN_INSN, size ); unsigned n = 0; + assert(src.File != TGSI_FILE_NULL); + assert(src.File != TGSI_FILE_OUTPUT); + assert(src.File < TGSI_FILE_COUNT); + out[n].value = 0; out[n].src.File = src.File; out[n].src.SwizzleX = src.SwizzleX; out[n].src.SwizzleY = src.SwizzleY; out[n].src.SwizzleZ = src.SwizzleZ; out[n].src.SwizzleW = src.SwizzleW; - out[n].src.Indirect = src.Indirect; out[n].src.Index = src.Index; + out[n].src.Negate = src.Negate; n++; if (src.Absolute) { + out[0].src.Extended = 1; + out[0].src.Negate = 0; out[n].value = 0; + out[n].src_ext_mod.Type = TGSI_SRC_REGISTER_EXT_TYPE_MOD; out[n].src_ext_mod.Absolute = 1; + out[n].src_ext_mod.Negate = src.Negate; n++; } if (src.Indirect) { + out[0].src.Indirect = 1; out[n].value = 0; out[n].src.File = TGSI_FILE_ADDRESS; - out[n].src.SwizzleX = TGSI_SWIZZLE_X; - out[n].src.SwizzleY = TGSI_SWIZZLE_X; - out[n].src.SwizzleZ = TGSI_SWIZZLE_X; - out[n].src.SwizzleW = TGSI_SWIZZLE_X; - out[n].src.Indirect = 0; - out[n].src.Index = 0; + out[n].src.SwizzleX = src.IndirectSwizzle; + out[n].src.SwizzleY = src.IndirectSwizzle; + out[n].src.SwizzleZ = src.IndirectSwizzle; + out[n].src.SwizzleW = src.IndirectSwizzle; + out[n].src.Index = src.IndirectIndex; n++; } @@ -484,6 +537,13 @@ ureg_emit_dst( struct ureg_program *ureg, union tgsi_any_token *out = get_tokens( ureg, DOMAIN_INSN, size ); unsigned n = 0; + assert(dst.File != TGSI_FILE_NULL); + assert(dst.File != TGSI_FILE_CONSTANT); + assert(dst.File != TGSI_FILE_INPUT); + assert(dst.File != TGSI_FILE_SAMPLER); + assert(dst.File != TGSI_FILE_IMMEDIATE); + assert(dst.File < TGSI_FILE_COUNT); + out[n].value = 0; out[n].dst.File = dst.File; out[n].dst.WriteMask = dst.WriteMask; @@ -494,12 +554,11 @@ ureg_emit_dst( struct ureg_program *ureg, if (dst.Indirect) { out[n].value = 0; out[n].src.File = TGSI_FILE_ADDRESS; - out[n].src.SwizzleX = TGSI_SWIZZLE_X; - out[n].src.SwizzleY = TGSI_SWIZZLE_X; - out[n].src.SwizzleZ = TGSI_SWIZZLE_X; - out[n].src.SwizzleW = TGSI_SWIZZLE_X; - out[n].src.Indirect = 0; - out[n].src.Index = 0; + out[n].src.SwizzleX = dst.IndirectSwizzle; + out[n].src.SwizzleY = dst.IndirectSwizzle; + out[n].src.SwizzleZ = dst.IndirectSwizzle; + out[n].src.SwizzleW = dst.IndirectSwizzle; + out[n].src.Index = dst.IndirectIndex; n++; } @@ -523,7 +582,6 @@ ureg_emit_insn(struct ureg_program *ureg, out[0].insn.NrTokens = 0; out[0].insn.Opcode = opcode; out[0].insn.Saturate = saturate; - out[0].insn.NrTokens = 0; out[0].insn.NumDstRegs = num_dst; out[0].insn.NumSrcRegs = num_src; out[0].insn.Padding = 0; @@ -542,6 +600,9 @@ ureg_emit_label(struct ureg_program *ureg, { union tgsi_any_token *out, *insn; + if(!label_token) + return; + out = get_tokens( ureg, DOMAIN_INSN, 1 ); insn = retrieve_token( ureg, DOMAIN_INSN, insn_token ); @@ -617,6 +678,17 @@ ureg_insn(struct ureg_program *ureg, unsigned insn, i; boolean saturate; +#ifdef DEBUG + { + const struct tgsi_opcode_info *info = tgsi_get_opcode_info( opcode ); + assert(info); + if(info) { + assert(nr_dst == info->num_dst); + assert(nr_src == info->num_src); + } + } +#endif + saturate = nr_dst ? dst[0].Saturate : FALSE; insn = ureg_emit_insn( ureg, opcode, saturate, nr_dst, nr_src ); @@ -723,10 +795,10 @@ static void emit_decls( struct ureg_program *ureg ) TGSI_INTERPOLATE_CONSTANT ); } - if (ureg->nr_samplers) { + for (i = 0; i < ureg->nr_samplers; i++) { emit_decl_range( ureg, TGSI_FILE_SAMPLER, - 0, ureg->nr_samplers ); + ureg->sampler[i].Index, 1 ); } if (ureg->nr_constants) { @@ -741,6 +813,12 @@ static void emit_decls( struct ureg_program *ureg ) 0, ureg->nr_temps ); } + if (ureg->nr_addrs) { + emit_decl_range( ureg, + TGSI_FILE_ADDRESS, + 0, ureg->nr_addrs ); + } + for (i = 0; i < ureg->nr_immediates; i++) { emit_immediate( ureg, ureg->immediate[i].v ); @@ -764,7 +842,7 @@ static void copy_instructions( struct ureg_program *ureg ) static void -fixup_header_size(struct ureg_program *ureg ) +fixup_header_size(struct ureg_program *ureg) { union tgsi_any_token *out = retrieve_token( ureg, DOMAIN_DECL, 1 ); diff --git a/src/gallium/auxiliary/tgsi/tgsi_ureg.h b/src/gallium/auxiliary/tgsi/tgsi_ureg.h index 8836a1ea0eb..acbca59040c 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_ureg.h +++ b/src/gallium/auxiliary/tgsi/tgsi_ureg.h @@ -31,6 +31,10 @@ #include "pipe/p_compiler.h" #include "pipe/p_shader_tokens.h" +#ifdef __cplusplus +extern "C" { +#endif + struct ureg_program; /* Almost a tgsi_src_register, but we need to pull in the Absolute @@ -48,6 +52,8 @@ struct ureg_src unsigned Absolute : 1; /* BOOL */ int Index : 16; /* SINT */ unsigned Negate : 1; /* BOOL */ + int IndirectIndex : 16; /* SINT */ + int IndirectSwizzle : 2; /* TGSI_SWIZZLE_ */ }; /* Very similar to a tgsi_dst_register, removing unsupported fields @@ -64,6 +70,8 @@ struct ureg_dst int Index : 16; /* SINT */ unsigned Pad1 : 5; unsigned Pad2 : 1; /* BOOL */ + int IndirectIndex : 16; /* SINT */ + int IndirectSwizzle : 2; /* TGSI_SWIZZLE_ */ }; struct pipe_context; @@ -131,12 +139,21 @@ void ureg_release_temporary( struct ureg_program *ureg, struct ureg_dst tmp ); +struct ureg_dst +ureg_DECL_address( struct ureg_program * ); + +/* Supply an index to the sampler declaration as this is the hook to + * the external pipe_sampler state. Users of this function probably + * don't want just any sampler, but a specific one which they've set + * up state for in the context. + */ struct ureg_src -ureg_DECL_sampler( struct ureg_program * ); +ureg_DECL_sampler( struct ureg_program *, + unsigned index ); static INLINE struct ureg_src -ureg_DECL_immediate4f( struct ureg_program *ureg, +ureg_imm4f( struct ureg_program *ureg, float a, float b, float c, float d) { @@ -149,7 +166,7 @@ ureg_DECL_immediate4f( struct ureg_program *ureg, } static INLINE struct ureg_src -ureg_DECL_immediate3f( struct ureg_program *ureg, +ureg_imm3f( struct ureg_program *ureg, float a, float b, float c) { @@ -161,7 +178,7 @@ ureg_DECL_immediate3f( struct ureg_program *ureg, } static INLINE struct ureg_src -ureg_DECL_immediate2f( struct ureg_program *ureg, +ureg_imm2f( struct ureg_program *ureg, float a, float b) { float v[2]; @@ -171,7 +188,7 @@ ureg_DECL_immediate2f( struct ureg_program *ureg, } static INLINE struct ureg_src -ureg_DECL_immediate1f( struct ureg_program *ureg, +ureg_imm1f( struct ureg_program *ureg, float a) { float v[1]; @@ -392,6 +409,7 @@ static INLINE void ureg_##op( struct ureg_program *ureg, \ static INLINE struct ureg_src ureg_negate( struct ureg_src reg ) { + assert(reg.File != TGSI_FILE_NULL); reg.Negate ^= 1; return reg; } @@ -399,6 +417,7 @@ ureg_negate( struct ureg_src reg ) static INLINE struct ureg_src ureg_abs( struct ureg_src reg ) { + assert(reg.File != TGSI_FILE_NULL); reg.Absolute = 1; reg.Negate = 0; return reg; @@ -413,6 +432,12 @@ ureg_swizzle( struct ureg_src reg, (reg.SwizzleZ << 4) | (reg.SwizzleW << 6)); + assert(reg.File != TGSI_FILE_NULL); + assert(x < 4); + assert(y < 4); + assert(z < 4); + assert(w < 4); + reg.SwizzleX = (swz >> (x*2)) & 0x3; reg.SwizzleY = (swz >> (y*2)) & 0x3; reg.SwizzleZ = (swz >> (z*2)) & 0x3; @@ -430,6 +455,7 @@ static INLINE struct ureg_dst ureg_writemask( struct ureg_dst reg, unsigned writemask ) { + assert(reg.File != TGSI_FILE_NULL); reg.WriteMask &= writemask; return reg; } @@ -437,10 +463,33 @@ ureg_writemask( struct ureg_dst reg, static INLINE struct ureg_dst ureg_saturate( struct ureg_dst reg ) { + assert(reg.File != TGSI_FILE_NULL); reg.Saturate = 1; return reg; } +static INLINE struct ureg_dst +ureg_dst_indirect( struct ureg_dst reg, struct ureg_src addr ) +{ + assert(reg.File != TGSI_FILE_NULL); + assert(addr.File == TGSI_FILE_ADDRESS); + reg.Indirect = 1; + reg.IndirectIndex = addr.Index; + reg.IndirectSwizzle = addr.SwizzleX; + return reg; +} + +static INLINE struct ureg_src +ureg_src_indirect( struct ureg_src reg, struct ureg_src addr ) +{ + assert(reg.File != TGSI_FILE_NULL); + assert(addr.File == TGSI_FILE_ADDRESS); + reg.Indirect = 1; + reg.IndirectIndex = addr.Index; + reg.IndirectSwizzle = addr.SwizzleX; + return reg; +} + static INLINE struct ureg_dst ureg_dst( struct ureg_src src ) { @@ -449,6 +498,8 @@ ureg_dst( struct ureg_src src ) dst.File = src.File; dst.WriteMask = TGSI_WRITEMASK_XYZW; dst.Indirect = src.Indirect; + dst.IndirectIndex = src.IndirectIndex; + dst.IndirectSwizzle = src.IndirectSwizzle; dst.Saturate = 0; dst.Index = src.Index; dst.Pad1 = 0; @@ -469,6 +520,8 @@ ureg_src( struct ureg_dst dst ) src.SwizzleW = TGSI_SWIZZLE_W; src.Pad = 0; src.Indirect = dst.Indirect; + src.IndirectIndex = dst.IndirectIndex; + src.IndirectSwizzle = dst.IndirectSwizzle; src.Absolute = 0; src.Index = dst.Index; src.Negate = 0; @@ -478,4 +531,60 @@ ureg_src( struct ureg_dst dst ) +static INLINE struct ureg_dst +ureg_dst_undef( void ) +{ + struct ureg_dst dst; + + dst.File = TGSI_FILE_NULL; + dst.WriteMask = 0; + dst.Indirect = 0; + dst.IndirectIndex = 0; + dst.IndirectSwizzle = 0; + dst.Saturate = 0; + dst.Index = 0; + dst.Pad1 = 0; + dst.Pad2 = 0; + + return dst; +} + +static INLINE struct ureg_src +ureg_src_undef( void ) +{ + struct ureg_src src; + + src.File = TGSI_FILE_NULL; + src.SwizzleX = 0; + src.SwizzleY = 0; + src.SwizzleZ = 0; + src.SwizzleW = 0; + src.Pad = 0; + src.Indirect = 0; + src.IndirectIndex = 0; + src.IndirectSwizzle = 0; + src.Absolute = 0; + src.Index = 0; + src.Negate = 0; + + return src; +} + +static INLINE boolean +ureg_src_is_undef( struct ureg_src src ) +{ + return src.File == TGSI_FILE_NULL; +} + +static INLINE boolean +ureg_dst_is_undef( struct ureg_dst dst ) +{ + return dst.File == TGSI_FILE_NULL; +} + + +#ifdef __cplusplus +} +#endif + #endif diff --git a/src/gallium/auxiliary/tgsi/tgsi_util.h b/src/gallium/auxiliary/tgsi/tgsi_util.h index 7877f345587..21eb656327e 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_util.h +++ b/src/gallium/auxiliary/tgsi/tgsi_util.h @@ -32,6 +32,10 @@ extern "C" { #endif +struct tgsi_src_register; +struct tgsi_src_register_ext_swz; +struct tgsi_full_src_register; + void * tgsi_align_128bit( void *unaligned ); diff --git a/src/gallium/auxiliary/util/u_blit.c b/src/gallium/auxiliary/util/u_blit.c index cda6dbd46d7..c516317d701 100644 --- a/src/gallium/auxiliary/util/u_blit.c +++ b/src/gallium/auxiliary/util/u_blit.c @@ -62,7 +62,7 @@ struct blit_state struct pipe_viewport_state viewport; void *vs; - void *fs; + void *fs[TGSI_WRITEMASK_XYZW + 1]; struct pipe_buffer *vbuf; /**< quad vertices */ unsigned vbuf_slot; @@ -125,7 +125,7 @@ util_create_blit(struct pipe_context *pipe, struct cso_context *cso) } /* fragment shader */ - ctx->fs = util_make_fragment_tex_shader(pipe); + ctx->fs[TGSI_WRITEMASK_XYZW] = util_make_fragment_tex_shader(pipe); ctx->vbuf = NULL; /* init vertex data that doesn't change */ @@ -146,9 +146,13 @@ void util_destroy_blit(struct blit_state *ctx) { struct pipe_context *pipe = ctx->pipe; + unsigned i; pipe->delete_vs_state(pipe, ctx->vs); - pipe->delete_fs_state(pipe, ctx->fs); + + for (i = 0; i < Elements(ctx->fs); i++) + if (ctx->fs[i]) + pipe->delete_fs_state(pipe, ctx->fs[i]); pipe_buffer_reference(&ctx->vbuf, NULL); @@ -299,14 +303,15 @@ regions_overlap(int srcX0, int srcY0, * XXX need some control over blitting Z and/or stencil. */ void -util_blit_pixels(struct blit_state *ctx, - struct pipe_surface *src, - int srcX0, int srcY0, - int srcX1, int srcY1, - struct pipe_surface *dst, - int dstX0, int dstY0, - int dstX1, int dstY1, - float z, uint filter) +util_blit_pixels_writemask(struct blit_state *ctx, + struct pipe_surface *src, + int srcX0, int srcY0, + int srcX1, int srcY1, + struct pipe_surface *dst, + int dstX0, int dstY0, + int dstX1, int dstY1, + float z, uint filter, + uint writemask) { struct pipe_context *pipe = ctx->pipe; struct pipe_screen *screen = pipe->screen; @@ -426,8 +431,11 @@ util_blit_pixels(struct blit_state *ctx, /* texture */ cso_set_sampler_textures(ctx->cso, 1, &tex); + if (ctx->fs[writemask] == NULL) + ctx->fs[writemask] = util_make_fragment_tex_shader_writemask(pipe, writemask); + /* shaders */ - cso_set_fragment_shader_handle(ctx->cso, ctx->fs); + cso_set_fragment_shader_handle(ctx->cso, ctx->fs[writemask]); cso_set_vertex_shader_handle(ctx->cso, ctx->vs); /* drawing dest */ @@ -462,6 +470,27 @@ util_blit_pixels(struct blit_state *ctx, } +void +util_blit_pixels(struct blit_state *ctx, + struct pipe_surface *src, + int srcX0, int srcY0, + int srcX1, int srcY1, + struct pipe_surface *dst, + int dstX0, int dstY0, + int dstX1, int dstY1, + float z, uint filter ) +{ + util_blit_pixels_writemask( ctx, src, + srcX0, srcY0, + srcX1, srcY1, + dst, + dstX0, dstY0, + dstX1, dstY1, + z, filter, + TGSI_WRITEMASK_XYZW ); +} + + /* Release vertex buffer at end of frame to avoid synchronous * rendering. */ @@ -535,7 +564,7 @@ util_blit_pixels_tex(struct blit_state *ctx, cso_set_sampler_textures(ctx->cso, 1, &tex); /* shaders */ - cso_set_fragment_shader_handle(ctx->cso, ctx->fs); + cso_set_fragment_shader_handle(ctx->cso, ctx->fs[TGSI_WRITEMASK_XYZW]); cso_set_vertex_shader_handle(ctx->cso, ctx->vs); /* drawing dest */ diff --git a/src/gallium/auxiliary/util/u_blit.h b/src/gallium/auxiliary/util/u_blit.h index c35beceda8d..a102021529e 100644 --- a/src/gallium/auxiliary/util/u_blit.h +++ b/src/gallium/auxiliary/util/u_blit.h @@ -60,6 +60,17 @@ util_blit_pixels(struct blit_state *ctx, int dstX1, int dstY1, float z, uint filter); +void +util_blit_pixels_writemask(struct blit_state *ctx, + struct pipe_surface *src, + int srcX0, int srcY0, + int srcX1, int srcY1, + struct pipe_surface *dst, + int dstX0, int dstY0, + int dstX1, int dstY1, + float z, uint filter, + uint writemask); + extern void util_blit_pixels_tex(struct blit_state *ctx, struct pipe_texture *tex, diff --git a/src/gallium/auxiliary/util/u_math.h b/src/gallium/auxiliary/util/u_math.h index b0807c13392..4c6c2bc00e1 100644 --- a/src/gallium/auxiliary/util/u_math.h +++ b/src/gallium/auxiliary/util/u_math.h @@ -374,6 +374,10 @@ unsigned ffs( unsigned u ) #define ffs __builtin_ffs #endif +#ifdef __MINGW32__ +#define ffs __builtin_ffs +#endif + /* Could also binary search for the highest bit. */ diff --git a/src/gallium/auxiliary/util/u_simple_shaders.c b/src/gallium/auxiliary/util/u_simple_shaders.c index d54a1d8c746..1b8da9b6853 100644 --- a/src/gallium/auxiliary/util/u_simple_shaders.c +++ b/src/gallium/auxiliary/util/u_simple_shaders.c @@ -88,11 +88,14 @@ util_make_vertex_passthrough_shader(struct pipe_context *pipe, /** * Make simple fragment texture shader: - * TEX OUT[0], IN[0], SAMP[0], 2D; + * IMM {0,0,0,1} // (if writemask != 0xf) + * MOV OUT[0], IMM[0] // (if writemask != 0xf) + * TEX OUT[0].writemask, IN[0], SAMP[0], 2D; * END; */ void * -util_make_fragment_tex_shader(struct pipe_context *pipe) +util_make_fragment_tex_shader_writemask(struct pipe_context *pipe, + unsigned writemask ) { struct ureg_program *ureg; struct ureg_src sampler; @@ -103,7 +106,7 @@ util_make_fragment_tex_shader(struct pipe_context *pipe) if (ureg == NULL) return NULL; - sampler = ureg_DECL_sampler( ureg ); + sampler = ureg_DECL_sampler( ureg, 0 ); tex = ureg_DECL_fs_input( ureg, TGSI_SEMANTIC_GENERIC, 0, @@ -119,6 +122,13 @@ util_make_fragment_tex_shader(struct pipe_context *pipe) return ureg_create_shader_and_destroy( ureg, pipe ); } +void * +util_make_fragment_tex_shader(struct pipe_context *pipe ) +{ + return util_make_fragment_tex_shader_writemask( pipe, + TGSI_WRITEMASK_XYZW ); +} + diff --git a/src/gallium/auxiliary/util/u_simple_shaders.h b/src/gallium/auxiliary/util/u_simple_shaders.h index 6f8d96af9bc..d2e80d6eb4d 100644 --- a/src/gallium/auxiliary/util/u_simple_shaders.h +++ b/src/gallium/auxiliary/util/u_simple_shaders.h @@ -50,6 +50,10 @@ util_make_vertex_passthrough_shader(struct pipe_context *pipe, extern void * +util_make_fragment_tex_shader_writemask(struct pipe_context *pipe, + unsigned writemask ); + +extern void * util_make_fragment_tex_shader(struct pipe_context *pipe); diff --git a/src/gallium/drivers/cell/spu/spu_exec.c b/src/gallium/drivers/cell/spu/spu_exec.c index 6db9501128c..0eaae2e451b 100644 --- a/src/gallium/drivers/cell/spu/spu_exec.c +++ b/src/gallium/drivers/cell/spu/spu_exec.c @@ -1150,10 +1150,6 @@ exec_instruction( ASSERT (0); break; - case TGSI_OPCODE_CND0: - ASSERT (0); - break; - case TGSI_OPCODE_DP2A: ASSERT (0); break; diff --git a/src/gallium/drivers/i915simple/Makefile b/src/gallium/drivers/i915simple/Makefile index 8870b398661..fb533c17961 100644 --- a/src/gallium/drivers/i915simple/Makefile +++ b/src/gallium/drivers/i915simple/Makefile @@ -5,6 +5,7 @@ LIBNAME = i915simple C_SOURCES = \ i915_blit.c \ + i915_buffer.c \ i915_clear.c \ i915_flush.c \ i915_context.c \ diff --git a/src/gallium/drivers/i915simple/SConscript b/src/gallium/drivers/i915simple/SConscript index 2366e1247f2..778c4ed0fde 100644 --- a/src/gallium/drivers/i915simple/SConscript +++ b/src/gallium/drivers/i915simple/SConscript @@ -6,6 +6,7 @@ i915simple = env.ConvenienceLibrary( target = 'i915simple', source = [ 'i915_blit.c', + 'i915_buffer.c', 'i915_clear.c', 'i915_context.c', 'i915_debug.c', diff --git a/src/gallium/drivers/i915simple/i915_batch.h b/src/gallium/drivers/i915simple/i915_batch.h index c6e68ea38a2..b813784723f 100644 --- a/src/gallium/drivers/i915simple/i915_batch.h +++ b/src/gallium/drivers/i915simple/i915_batch.h @@ -28,89 +28,20 @@ #ifndef I915_BATCH_H #define I915_BATCH_H -#include "i915_winsys.h" +#include "intel_batchbuffer.h" -struct i915_batchbuffer -{ - struct pipe_buffer *buffer; - struct i915_winsys *winsys; +#define BEGIN_BATCH(dwords, relocs) \ + (intel_batchbuffer_check(i915->batch, dwords, relocs)) - unsigned char *map; - unsigned char *ptr; +#define OUT_BATCH(dword) \ + intel_batchbuffer_dword(i915->batch, dword) - size_t size; - size_t actual_size; +#define OUT_RELOC(buf, usage, offset) \ + intel_batchbuffer_reloc(i915->batch, buf, usage, offset) - size_t relocs; - size_t max_relocs; -}; - -static INLINE boolean -i915_batchbuffer_check( struct i915_batchbuffer *batch, - size_t dwords, - size_t relocs ) -{ - return dwords * 4 <= batch->size - (batch->ptr - batch->map) && - relocs <= (batch->max_relocs - batch->relocs); -} - -static INLINE size_t -i915_batchbuffer_space( struct i915_batchbuffer *batch ) -{ - return batch->size - (batch->ptr - batch->map); -} - -static INLINE void -i915_batchbuffer_dword( struct i915_batchbuffer *batch, - unsigned dword ) -{ - if (i915_batchbuffer_space(batch) < 4) - return; - - *(unsigned *)batch->ptr = dword; - batch->ptr += 4; -} - -static INLINE void -i915_batchbuffer_write( struct i915_batchbuffer *batch, - void *data, - size_t size ) -{ - if (i915_batchbuffer_space(batch) < size) - return; - - memcpy(data, batch->ptr, size); - batch->ptr += size; -} - -static INLINE void -i915_batchbuffer_reloc( struct i915_batchbuffer *batch, - struct pipe_buffer *buffer, - size_t flags, - size_t offset ) -{ - batch->winsys->batch_reloc( batch->winsys, buffer, flags, offset ); -} - -static INLINE void -i915_batchbuffer_flush( struct i915_batchbuffer *batch, - struct pipe_fence_handle **fence ) -{ - batch->winsys->batch_flush( batch->winsys, fence ); -} - -#define BEGIN_BATCH( dwords, relocs ) \ - (i915_batchbuffer_check( i915->batch, dwords, relocs )) - -#define OUT_BATCH( dword ) \ - i915_batchbuffer_dword( i915->batch, dword ) - -#define OUT_RELOC( buf, flags, delta ) \ - i915_batchbuffer_reloc( i915->batch, buf, flags, delta ) - -#define FLUSH_BATCH(fence) do { \ - i915->winsys->batch_flush( i915->winsys, fence ); \ - i915->hardware_dirty = ~0; \ +#define FLUSH_BATCH(fence) do { \ + intel_batchbuffer_flush(i915->batch, fence); \ + i915->hardware_dirty = ~0; \ } while (0) #endif diff --git a/src/gallium/drivers/i915simple/i915_blit.c b/src/gallium/drivers/i915simple/i915_blit.c index 448a4708ce8..83dfc335288 100644 --- a/src/gallium/drivers/i915simple/i915_blit.c +++ b/src/gallium/drivers/i915simple/i915_blit.c @@ -26,8 +26,6 @@ **************************************************************************/ -#include "i915_context.h" -#include "i915_winsys.h" #include "i915_blit.h" #include "i915_reg.h" #include "i915_batch.h" @@ -37,33 +35,33 @@ void i915_fill_blit(struct i915_context *i915, - unsigned cpp, - unsigned short dst_pitch, - struct pipe_buffer *dst_buffer, - unsigned dst_offset, - short x, short y, - short w, short h, - unsigned color) + unsigned cpp, + unsigned short dst_pitch, + struct intel_buffer *dst_buffer, + unsigned dst_offset, + short x, short y, + short w, short h, + unsigned color) { unsigned BR13, CMD; I915_DBG(i915, - "%s dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n", - __FUNCTION__, - dst_buffer, dst_pitch, dst_offset, x, y, w, h); + "%s dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n", + __FUNCTION__, + dst_buffer, dst_pitch, dst_offset, x, y, w, h); switch (cpp) { case 1: case 2: case 3: BR13 = (((int) dst_pitch) & 0xffff) | - (0xF0 << 16) | (1 << 24); + (0xF0 << 16) | (1 << 24); CMD = XY_COLOR_BLT_CMD; break; case 4: BR13 = (((int) dst_pitch) & 0xffff) | - (0xF0 << 16) | (1 << 24) | (1 << 25); + (0xF0 << 16) | (1 << 24) | (1 << 25); CMD = (XY_COLOR_BLT_CMD | XY_COLOR_BLT_WRITE_ALPHA | XY_COLOR_BLT_WRITE_RGB); break; @@ -79,25 +77,24 @@ i915_fill_blit(struct i915_context *i915, OUT_BATCH(BR13); OUT_BATCH((y << 16) | x); OUT_BATCH(((y + h) << 16) | (x + w)); - OUT_RELOC( dst_buffer, I915_BUFFER_ACCESS_WRITE, dst_offset); + OUT_RELOC(dst_buffer, INTEL_USAGE_2D_TARGET, dst_offset); OUT_BATCH(color); FLUSH_BATCH(NULL); } - void -i915_copy_blit( struct i915_context *i915, - unsigned do_flip, - unsigned cpp, - unsigned short src_pitch, - struct pipe_buffer *src_buffer, - unsigned src_offset, - unsigned short dst_pitch, - struct pipe_buffer *dst_buffer, - unsigned dst_offset, - short src_x, short src_y, - short dst_x, short dst_y, - short w, short h ) +i915_copy_blit(struct i915_context *i915, + unsigned do_flip, + unsigned cpp, + unsigned short src_pitch, + struct intel_buffer *src_buffer, + unsigned src_offset, + unsigned short dst_pitch, + struct intel_buffer *dst_buffer, + unsigned dst_offset, + short src_x, short src_y, + short dst_x, short dst_y, + short w, short h) { unsigned CMD, BR13; int dst_y2 = dst_y + h; @@ -105,32 +102,30 @@ i915_copy_blit( struct i915_context *i915, I915_DBG(i915, - "%s src:buf(%p)/%d+%d %d,%d dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n", - __FUNCTION__, - src_buffer, src_pitch, src_offset, src_x, src_y, - dst_buffer, dst_pitch, dst_offset, dst_x, dst_y, w, h); + "%s src:buf(%p)/%d+%d %d,%d dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n", + __FUNCTION__, + src_buffer, src_pitch, src_offset, src_x, src_y, + dst_buffer, dst_pitch, dst_offset, dst_x, dst_y, w, h); switch (cpp) { case 1: case 2: case 3: BR13 = (((int) dst_pitch) & 0xffff) | - (0xCC << 16) | (1 << 24); + (0xCC << 16) | (1 << 24); CMD = XY_SRC_COPY_BLT_CMD; break; case 4: BR13 = (((int) dst_pitch) & 0xffff) | - (0xCC << 16) | (1 << 24) | (1 << 25); - CMD = - (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | - XY_SRC_COPY_BLT_WRITE_RGB); + (0xCC << 16) | (1 << 24) | (1 << 25); + CMD = (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | + XY_SRC_COPY_BLT_WRITE_RGB); break; default: return; } - if (dst_y2 < dst_y || - dst_x2 < dst_x) { + if (dst_y2 < dst_y || dst_x2 < dst_x) { return; } @@ -140,7 +135,6 @@ i915_copy_blit( struct i915_context *i915, */ assert (dst_pitch > 0 && src_pitch > 0); - if (!BEGIN_BATCH(8, 2)) { FLUSH_BATCH(NULL); assert(BEGIN_BATCH(8, 2)); @@ -149,11 +143,9 @@ i915_copy_blit( struct i915_context *i915, OUT_BATCH(BR13); OUT_BATCH((dst_y << 16) | dst_x); OUT_BATCH((dst_y2 << 16) | dst_x2); - OUT_RELOC(dst_buffer, I915_BUFFER_ACCESS_WRITE, dst_offset); + OUT_RELOC(dst_buffer, INTEL_USAGE_2D_TARGET, dst_offset); OUT_BATCH((src_y << 16) | src_x); OUT_BATCH(((int) src_pitch & 0xffff)); - OUT_RELOC(src_buffer, I915_BUFFER_ACCESS_READ, src_offset); + OUT_RELOC(src_buffer, INTEL_USAGE_2D_SOURCE, src_offset); FLUSH_BATCH(NULL); } - - diff --git a/src/gallium/drivers/i915simple/i915_blit.h b/src/gallium/drivers/i915simple/i915_blit.h index 0bb34538611..8ce3220cfd9 100644 --- a/src/gallium/drivers/i915simple/i915_blit.h +++ b/src/gallium/drivers/i915simple/i915_blit.h @@ -32,24 +32,24 @@ extern void i915_copy_blit(struct i915_context *i915, unsigned do_flip, - unsigned cpp, - unsigned short src_pitch, - struct pipe_buffer *src_buffer, - unsigned src_offset, - unsigned short dst_pitch, - struct pipe_buffer *dst_buffer, - unsigned dst_offset, - short srcx, short srcy, - short dstx, short dsty, - short w, short h ); + unsigned cpp, + unsigned short src_pitch, + struct intel_buffer *src_buffer, + unsigned src_offset, + unsigned short dst_pitch, + struct intel_buffer *dst_buffer, + unsigned dst_offset, + short srcx, short srcy, + short dstx, short dsty, + short w, short h); extern void i915_fill_blit(struct i915_context *i915, - unsigned cpp, - unsigned short dst_pitch, - struct pipe_buffer *dst_buffer, - unsigned dst_offset, - short x, short y, - short w, short h, unsigned color); + unsigned cpp, + unsigned short dst_pitch, + struct intel_buffer *dst_buffer, + unsigned dst_offset, + short x, short y, + short w, short h, unsigned color); #endif diff --git a/src/gallium/drivers/i915simple/i915_buffer.c b/src/gallium/drivers/i915simple/i915_buffer.c new file mode 100644 index 00000000000..effeba12972 --- /dev/null +++ b/src/gallium/drivers/i915simple/i915_buffer.c @@ -0,0 +1,136 @@ +/************************************************************************** + * + * Copyright © 2009 Jakob Bornecrantz + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "util/u_memory.h" +#include "i915_screen.h" +#include "i915_buffer.h" + +struct intel_buffer; + +struct i915_buffer +{ + struct pipe_buffer base; + + struct intel_buffer *ibuf; /** hw buffer */ + + void *data; /**< user and malloc data */ + boolean own; /**< we own the data incase of malloc */ +}; + +static INLINE struct i915_buffer * +i915_buffer(struct pipe_buffer *buffer) +{ + return (struct i915_buffer *)buffer; +} + +static struct pipe_buffer * +i915_buffer_create(struct pipe_screen *screen, + unsigned alignment, + unsigned usage, + unsigned size) +{ + struct i915_buffer *buf = CALLOC_STRUCT(i915_buffer); + + if (!buf) + return NULL; + + pipe_reference_init(&buf->base.reference, 1); + buf->base.alignment = alignment; + buf->base.screen = screen; + buf->base.usage = usage; + buf->base.size = size; + buf->data = MALLOC(size); + buf->own = TRUE; + + if (!buf->data) + goto err; + + return &buf->base; + +err: + FREE(buf); + return NULL; +} + +static struct pipe_buffer * +i915_user_buffer_create(struct pipe_screen *screen, + void *ptr, + unsigned bytes) +{ + struct i915_buffer *buf = CALLOC_STRUCT(i915_buffer); + + if (!buf) + return NULL; + + pipe_reference_init(&buf->base.reference, 1); + buf->base.alignment = 0; + buf->base.screen = screen; + buf->base.usage = 0; + buf->base.size = bytes; + buf->data = ptr; + buf->own = FALSE; + + return &buf->base; +} + +static void * +i915_buffer_map(struct pipe_screen *screen, + struct pipe_buffer *buffer, + unsigned usage) +{ + struct i915_buffer *buf = i915_buffer(buffer); + assert(!buf->ibuf); + return buf->data; +} + +static void +i915_buffer_unmap(struct pipe_screen *screen, + struct pipe_buffer *buffer) +{ + struct i915_buffer *buf = i915_buffer(buffer); + assert(!buf->ibuf); +} + +static void +i915_buffer_destroy(struct pipe_buffer *buffer) +{ + struct i915_buffer *buf = i915_buffer(buffer); + assert(!buf->ibuf); + + if (buf->own) + FREE(buf->data); + FREE(buf); +} + +void i915_init_screen_buffer_functions(struct i915_screen *screen) +{ + screen->base.buffer_create = i915_buffer_create; + screen->base.user_buffer_create = i915_user_buffer_create; + screen->base.buffer_map = i915_buffer_map; + screen->base.buffer_map_range = NULL; + screen->base.buffer_flush_mapped_range = NULL; + screen->base.buffer_unmap = i915_buffer_unmap; + screen->base.buffer_destroy = i915_buffer_destroy; +} diff --git a/src/gallium/drivers/i915simple/i915_buffer.h b/src/gallium/drivers/i915simple/i915_buffer.h new file mode 100644 index 00000000000..80fda7c62fd --- /dev/null +++ b/src/gallium/drivers/i915simple/i915_buffer.h @@ -0,0 +1,31 @@ +/************************************************************************** + * + * Copyright © 2009 Jakob Bornecrantz + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef I915_BUFFER_H +#define I915_BUFFER_H + +void i915_init_screen_buffer_functions(struct i915_screen *screen); + +#endif diff --git a/src/gallium/drivers/i915simple/i915_context.c b/src/gallium/drivers/i915simple/i915_context.c index bf69c8e9f53..b43f7352456 100644 --- a/src/gallium/drivers/i915simple/i915_context.c +++ b/src/gallium/drivers/i915simple/i915_context.c @@ -26,8 +26,8 @@ **************************************************************************/ #include "i915_context.h" -#include "i915_winsys.h" #include "i915_state.h" +#include "i915_screen.h" #include "i915_batch.h" #include "i915_texture.h" #include "i915_reg.h" @@ -40,66 +40,58 @@ #include "pipe/p_screen.h" -static void i915_destroy( struct pipe_context *pipe ) -{ - struct i915_context *i915 = i915_context( pipe ); - - draw_destroy( i915->draw ); - - if(i915->winsys->destroy) - i915->winsys->destroy(i915->winsys); - - FREE( i915 ); -} +/* + * Draw functions + */ static boolean i915_draw_range_elements(struct pipe_context *pipe, - struct pipe_buffer *indexBuffer, - unsigned indexSize, - unsigned min_index, - unsigned max_index, - unsigned prim, unsigned start, unsigned count) + struct pipe_buffer *indexBuffer, + unsigned indexSize, + unsigned min_index, + unsigned max_index, + unsigned prim, unsigned start, unsigned count) { - struct i915_context *i915 = i915_context( pipe ); + struct i915_context *i915 = i915_context(pipe); struct draw_context *draw = i915->draw; unsigned i; if (i915->dirty) - i915_update_derived( i915 ); + i915_update_derived(i915); /* * Map vertex buffers */ for (i = 0; i < i915->num_vertex_buffers; i++) { - void *buf - = pipe_buffer_map(pipe->screen, - i915->vertex_buffer[i].buffer, - PIPE_BUFFER_USAGE_CPU_READ); + void *buf = pipe_buffer_map(pipe->screen, i915->vertex_buffer[i].buffer, + PIPE_BUFFER_USAGE_CPU_READ); draw_set_mapped_vertex_buffer(draw, i, buf); } - /* Map index buffer, if present */ + + /* + * Map index buffer, if present + */ if (indexBuffer) { - void *mapped_indexes - = pipe_buffer_map(pipe->screen, indexBuffer, - PIPE_BUFFER_USAGE_CPU_READ); + void *mapped_indexes = pipe_buffer_map(pipe->screen, indexBuffer, + PIPE_BUFFER_USAGE_CPU_READ); draw_set_mapped_element_buffer_range(draw, indexSize, - min_index, - max_index, - mapped_indexes); - } - else { - /* no index/element buffer */ + min_index, + max_index, + mapped_indexes); + } else { draw_set_mapped_element_buffer(draw, 0, NULL); } draw_set_mapped_constant_buffer(draw, i915->current.constants[PIPE_SHADER_VERTEX], - ( i915->current.num_user_constants[PIPE_SHADER_VERTEX] * - 4 * sizeof(float) )); + (i915->current.num_user_constants[PIPE_SHADER_VERTEX] * + 4 * sizeof(float))); - /* draw! */ + /* + * Do the drawing + */ draw_arrays(i915->draw, prim, start, count); /* @@ -109,6 +101,7 @@ i915_draw_range_elements(struct pipe_context *pipe, pipe_buffer_unmap(pipe->screen, i915->vertex_buffer[i].buffer); draw_set_mapped_vertex_buffer(draw, i, NULL); } + if (indexBuffer) { pipe_buffer_unmap(pipe->screen, indexBuffer); draw_set_mapped_element_buffer_range(draw, 0, start, start + count - 1, NULL); @@ -118,28 +111,34 @@ i915_draw_range_elements(struct pipe_context *pipe, } static boolean -i915_draw_elements( struct pipe_context *pipe, - struct pipe_buffer *indexBuffer, - unsigned indexSize, - unsigned prim, unsigned start, unsigned count) +i915_draw_elements(struct pipe_context *pipe, + struct pipe_buffer *indexBuffer, + unsigned indexSize, + unsigned prim, unsigned start, unsigned count) { - return i915_draw_range_elements( pipe, indexBuffer, - indexSize, - 0, 0xffffffff, - prim, start, count ); + return i915_draw_range_elements(pipe, indexBuffer, + indexSize, + 0, 0xffffffff, + prim, start, count); } -static boolean i915_draw_arrays( struct pipe_context *pipe, - unsigned prim, unsigned start, unsigned count) +static boolean +i915_draw_arrays(struct pipe_context *pipe, + unsigned prim, unsigned start, unsigned count) { return i915_draw_elements(pipe, NULL, 0, prim, start, count); } +/* + * Is referenced functions + */ + + static unsigned int -i915_is_texture_referenced( struct pipe_context *pipe, - struct pipe_texture *texture, - unsigned face, unsigned level) +i915_is_texture_referenced(struct pipe_context *pipe, + struct pipe_texture *texture, + unsigned face, unsigned level) { /** * FIXME: Return the corrent result. We can't alays return referenced @@ -153,8 +152,8 @@ i915_is_texture_referenced( struct pipe_context *pipe, } static unsigned int -i915_is_buffer_referenced( struct pipe_context *pipe, - struct pipe_buffer *buf) +i915_is_buffer_referenced(struct pipe_context *pipe, + struct pipe_buffer *buf) { /** * FIXME: Return the corrent result. We can't alays return referenced @@ -168,9 +167,25 @@ i915_is_buffer_referenced( struct pipe_context *pipe, } -struct pipe_context *i915_create_context( struct pipe_screen *screen, - struct pipe_winsys *pipe_winsys, - struct i915_winsys *i915_winsys ) +/* + * Generic context functions + */ + + +static void i915_destroy(struct pipe_context *pipe) +{ + struct i915_context *i915 = i915_context(pipe); + + draw_destroy(i915->draw); + + if(i915->batch) + i915->iws->batchbuffer_destroy(i915->batch); + + FREE(i915); +} + +struct pipe_context * +i915_create_context(struct pipe_screen *screen) { struct i915_context *i915; @@ -178,21 +193,20 @@ struct pipe_context *i915_create_context( struct pipe_screen *screen, if (i915 == NULL) return NULL; - i915->winsys = i915_winsys; - i915->pipe.winsys = pipe_winsys; - i915->pipe.screen = screen; + i915->iws = i915_screen(screen)->iws; + i915->base.winsys = NULL; + i915->base.screen = screen; - i915->pipe.destroy = i915_destroy; + i915->base.destroy = i915_destroy; - i915->pipe.clear = i915_clear; + i915->base.clear = i915_clear; + i915->base.draw_arrays = i915_draw_arrays; + i915->base.draw_elements = i915_draw_elements; + i915->base.draw_range_elements = i915_draw_range_elements; - i915->pipe.draw_arrays = i915_draw_arrays; - i915->pipe.draw_elements = i915_draw_elements; - i915->pipe.draw_range_elements = i915_draw_range_elements; - - i915->pipe.is_texture_referenced = i915_is_texture_referenced; - i915->pipe.is_buffer_referenced = i915_is_buffer_referenced; + i915->base.is_texture_referenced = i915_is_texture_referenced; + i915->base.is_buffer_referenced = i915_is_buffer_referenced; /* * Create drawing context and plug our rendering stage into it. @@ -201,27 +215,23 @@ struct pipe_context *i915_create_context( struct pipe_screen *screen, assert(i915->draw); if (!debug_get_bool_option("I915_NO_VBUF", FALSE)) { draw_set_rasterize_stage(i915->draw, i915_draw_vbuf_stage(i915)); - } - else { + } else { draw_set_rasterize_stage(i915->draw, i915_draw_render_stage(i915)); } i915_init_surface_functions(i915); i915_init_state_functions(i915); i915_init_flush_functions(i915); - i915_init_texture_functions(i915); - draw_install_aaline_stage(i915->draw, &i915->pipe); - draw_install_aapoint_stage(i915->draw, &i915->pipe); + draw_install_aaline_stage(i915->draw, &i915->base); + draw_install_aapoint_stage(i915->draw, &i915->base); i915->dirty = ~0; i915->hardware_dirty = ~0; /* Batch stream debugging is a bit hacked up at the moment: */ - i915->batch = i915_winsys->batch_get(i915_winsys); - i915->batch->winsys = i915_winsys; + i915->batch = i915->iws->batchbuffer_create(i915->iws); - return &i915->pipe; + return &i915->base; } - diff --git a/src/gallium/drivers/i915simple/i915_context.h b/src/gallium/drivers/i915simple/i915_context.h index b6983ba86e7..234b441ce6e 100644 --- a/src/gallium/drivers/i915simple/i915_context.h +++ b/src/gallium/drivers/i915simple/i915_context.h @@ -38,6 +38,11 @@ #include "tgsi/tgsi_scan.h" +struct intel_winsys; +struct intel_buffer; +struct intel_batchbuffer; + + #define I915_TEX_UNITS 8 #define I915_DYNAMIC_MODES4 0 @@ -182,7 +187,6 @@ struct i915_sampler_state { unsigned maxlod; }; - struct i915_texture { struct pipe_texture base; @@ -192,7 +196,8 @@ struct i915_texture { unsigned depth_stride; /* per-image on i945? */ unsigned total_nblocksy; - unsigned tiled; + unsigned sw_tiled; /**< tiled with software flags */ + unsigned hw_tiled; /**< tiled with hardware fences */ unsigned nr_images[PIPE_MAX_TEXTURE_LEVELS]; @@ -206,15 +211,15 @@ struct i915_texture { /* The data is held here: */ - struct pipe_buffer *buffer; + struct intel_buffer *buffer; }; -struct i915_batchbuffer; - struct i915_context { - struct pipe_context pipe; - struct i915_winsys *winsys; + struct pipe_context base; + + struct intel_winsys *iws; + struct draw_context *draw; /* The most recent drawing state as set by the driver: @@ -243,10 +248,10 @@ struct i915_context unsigned num_vertex_elements; unsigned num_vertex_buffers; - struct i915_batchbuffer *batch; + struct intel_batchbuffer *batch; /** Vertex buffer */ - struct pipe_buffer *vbo; + struct intel_buffer *vbo; size_t vbo_offset; unsigned vbo_flushed; diff --git a/src/gallium/drivers/i915simple/i915_debug.c b/src/gallium/drivers/i915simple/i915_debug.c index e08582efaba..ce92d1af9a7 100644 --- a/src/gallium/drivers/i915simple/i915_debug.c +++ b/src/gallium/drivers/i915simple/i915_debug.c @@ -27,7 +27,6 @@ #include "i915_reg.h" #include "i915_context.h" -#include "i915_winsys.h" #include "i915_debug.h" #include "i915_batch.h" #include "pipe/internal/p_winsys_screen.h" @@ -864,7 +863,7 @@ static boolean i915_debug_packet( struct debug_stream *stream ) void -i915_dump_batchbuffer( struct i915_batchbuffer *batch ) +i915_dump_batchbuffer( struct intel_batchbuffer *batch ) { struct debug_stream stream; unsigned *start = (unsigned*)batch->map; diff --git a/src/gallium/drivers/i915simple/i915_debug.h b/src/gallium/drivers/i915simple/i915_debug.h index 16ca7277c7a..dd9b86e17b5 100644 --- a/src/gallium/drivers/i915simple/i915_debug.h +++ b/src/gallium/drivers/i915simple/i915_debug.h @@ -104,9 +104,9 @@ I915_DBG( #endif -struct i915_batchbuffer; +struct intel_batchbuffer; -void i915_dump_batchbuffer( struct i915_batchbuffer *i915 ); +void i915_dump_batchbuffer( struct intel_batchbuffer *i915 ); void i915_debug_init( struct i915_context *i915 ); diff --git a/src/gallium/drivers/i915simple/i915_flush.c b/src/gallium/drivers/i915simple/i915_flush.c index 472e0ab7740..1582168eba5 100644 --- a/src/gallium/drivers/i915simple/i915_flush.c +++ b/src/gallium/drivers/i915simple/i915_flush.c @@ -45,6 +45,7 @@ static void i915_flush( struct pipe_context *pipe, draw_flush(i915->draw); +#if 0 /* Do we need to emit an MI_FLUSH command to flush the hardware * caches? */ @@ -63,6 +64,13 @@ static void i915_flush( struct pipe_context *pipe, } OUT_BATCH( flush ); } +#endif + +#if 0 + if (i915->batch->map == i915->batch->ptr) { + return; + } +#endif /* If there are no flags, just flush pending commands to hardware: */ @@ -74,5 +82,5 @@ static void i915_flush( struct pipe_context *pipe, void i915_init_flush_functions( struct i915_context *i915 ) { - i915->pipe.flush = i915_flush; + i915->base.flush = i915_flush; } diff --git a/src/gallium/drivers/i915simple/i915_prim_emit.c b/src/gallium/drivers/i915simple/i915_prim_emit.c index 8f1f58b2dd1..d9a5c40ab97 100644 --- a/src/gallium/drivers/i915simple/i915_prim_emit.c +++ b/src/gallium/drivers/i915simple/i915_prim_emit.c @@ -32,7 +32,6 @@ #include "util/u_pack_color.h" #include "i915_context.h" -#include "i915_winsys.h" #include "i915_reg.h" #include "i915_state.h" #include "i915_batch.h" diff --git a/src/gallium/drivers/i915simple/i915_prim_vbuf.c b/src/gallium/drivers/i915simple/i915_prim_vbuf.c index 9bdd91f2881..508f4560e48 100644 --- a/src/gallium/drivers/i915simple/i915_prim_vbuf.c +++ b/src/gallium/drivers/i915simple/i915_prim_vbuf.c @@ -42,13 +42,11 @@ #include "draw/draw_vbuf.h" #include "util/u_debug.h" #include "pipe/p_inlines.h" -#include "pipe/internal/p_winsys_screen.h" #include "util/u_math.h" #include "util/u_memory.h" #include "i915_context.h" #include "i915_reg.h" -#include "i915_winsys.h" #include "i915_batch.h" #include "i915_state.h" @@ -59,7 +57,7 @@ struct i915_vbuf_render { struct vbuf_render base; - struct i915_context *i915; + struct i915_context *i915; /** Vertex size in bytes */ size_t vertex_size; @@ -74,7 +72,7 @@ struct i915_vbuf_render { unsigned fallback; /* Stuff for the vbo */ - struct pipe_buffer *vbo; + struct intel_buffer *vbo; size_t vbo_size; size_t vbo_offset; void *vbo_ptr; @@ -87,36 +85,34 @@ struct i915_vbuf_render { * Basically a cast wrapper. */ static INLINE struct i915_vbuf_render * -i915_vbuf_render( struct vbuf_render *render ) +i915_vbuf_render(struct vbuf_render *render) { assert(render); return (struct i915_vbuf_render *)render; } - static const struct vertex_info * -i915_vbuf_render_get_vertex_info( struct vbuf_render *render ) +i915_vbuf_render_get_vertex_info(struct vbuf_render *render) { struct i915_vbuf_render *i915_render = i915_vbuf_render(render); struct i915_context *i915 = i915_render->i915; if (i915->dirty) { /* make sure we have up to date vertex layout */ - i915_update_derived( i915 ); + i915_update_derived(i915); } return &i915->current.vertex_info; } - static boolean -i915_vbuf_render_allocate_vertices( struct vbuf_render *render, - ushort vertex_size, - ushort nr_vertices ) +i915_vbuf_render_allocate_vertices(struct vbuf_render *render, + ushort vertex_size, + ushort nr_vertices) { struct i915_vbuf_render *i915_render = i915_vbuf_render(render); struct i915_context *i915 = i915_render->i915; - struct pipe_screen *screen = i915->pipe.screen; + struct intel_winsys *iws = i915->iws; size_t size = (size_t)vertex_size * (size_t)nr_vertices; /* FIXME: handle failure */ @@ -125,17 +121,17 @@ i915_vbuf_render_allocate_vertices( struct vbuf_render *render, if (i915_render->vbo_size > size + i915_render->vbo_offset && !i915->vbo_flushed) { } else { i915->vbo_flushed = 0; - if (i915_render->vbo) - pipe_buffer_reference(&i915_render->vbo, NULL); + if (i915_render->vbo) { + iws->buffer_destroy(iws, i915_render->vbo); + i915_render->vbo = NULL; + } } if (!i915_render->vbo) { i915_render->vbo_size = MAX2(size, i915_render->vbo_alloc_size); i915_render->vbo_offset = 0; - i915_render->vbo = pipe_buffer_create(screen, - 64, - I915_BUFFER_USAGE_LIT_VERTEX, - i915_render->vbo_size); + i915_render->vbo = iws->buffer_create(iws, i915_render->vbo_size, 64, + INTEL_NEW_VERTEX); } @@ -149,40 +145,37 @@ i915_vbuf_render_allocate_vertices( struct vbuf_render *render, return TRUE; } - static void * -i915_vbuf_render_map_vertices( struct vbuf_render *render ) +i915_vbuf_render_map_vertices(struct vbuf_render *render) { struct i915_vbuf_render *i915_render = i915_vbuf_render(render); struct i915_context *i915 = i915_render->i915; - struct pipe_screen *screen = i915->pipe.screen; + struct intel_winsys *iws = i915->iws; if (i915->vbo_flushed) debug_printf("%s bad vbo flush occured stalling on hw\n"); - i915_render->vbo_ptr = pipe_buffer_map(screen, - i915_render->vbo, - PIPE_BUFFER_USAGE_CPU_WRITE); + i915_render->vbo_ptr = iws->buffer_map(iws, i915_render->vbo, TRUE); return (unsigned char *)i915_render->vbo_ptr + i915->vbo_offset; } static void -i915_vbuf_render_unmap_vertices( struct vbuf_render *render, - ushort min_index, - ushort max_index ) +i915_vbuf_render_unmap_vertices(struct vbuf_render *render, + ushort min_index, + ushort max_index) { struct i915_vbuf_render *i915_render = i915_vbuf_render(render); struct i915_context *i915 = i915_render->i915; - struct pipe_screen *screen = i915->pipe.screen; + struct intel_winsys *iws = i915->iws; i915_render->vbo_max_used = MAX2(i915_render->vbo_max_used, i915_render->vertex_size * (max_index + 1)); - pipe_buffer_unmap(screen, i915_render->vbo); + iws->buffer_unmap(iws, i915_render->vbo); } static boolean -i915_vbuf_render_set_primitive( struct vbuf_render *render, - unsigned prim ) +i915_vbuf_render_set_primitive(struct vbuf_render *render, + unsigned prim) { struct i915_vbuf_render *i915_render = i915_vbuf_render(render); i915_render->prim = prim; @@ -234,15 +227,13 @@ i915_vbuf_render_set_primitive( struct vbuf_render *render, } } - - /** * Used for fallbacks in draw_arrays */ static void -draw_arrays_generate_indices( struct vbuf_render *render, - unsigned start, uint nr, - unsigned type ) +draw_arrays_generate_indices(struct vbuf_render *render, + unsigned start, uint nr, + unsigned type) { struct i915_vbuf_render *i915_render = i915_vbuf_render(render); struct i915_context *i915 = i915_render->i915; @@ -251,29 +242,29 @@ draw_arrays_generate_indices( struct vbuf_render *render, switch(type) { case 0: for (i = start; i+1 < end; i += 2) - OUT_BATCH( (i+0) | (i+1) << 16 ); + OUT_BATCH((i+0) | (i+1) << 16); if (i < end) - OUT_BATCH( i ); + OUT_BATCH(i); break; case PIPE_PRIM_LINE_LOOP: if (nr >= 2) { - for (i = start + 1; i < end; i++) - OUT_BATCH( (i-0) | (i+0) << 16 ); - OUT_BATCH( (i-0) | ( start) << 16 ); + for (i = start + 1; i < end; i++) + OUT_BATCH((i-0) | (i+0) << 16); + OUT_BATCH((i-0) | ( start) << 16); } break; case PIPE_PRIM_QUADS: for (i = start; i + 3 < end; i += 4) { - OUT_BATCH( (i+0) | (i+1) << 16 ); - OUT_BATCH( (i+3) | (i+1) << 16 ); - OUT_BATCH( (i+2) | (i+3) << 16 ); + OUT_BATCH((i+0) | (i+1) << 16); + OUT_BATCH((i+3) | (i+1) << 16); + OUT_BATCH((i+2) | (i+3) << 16); } break; case PIPE_PRIM_QUAD_STRIP: for (i = start; i + 3 < end; i += 2) { - OUT_BATCH( (i+0) | (i+1) << 16 ); - OUT_BATCH( (i+3) | (i+2) << 16 ); - OUT_BATCH( (i+0) | (i+3) << 16 ); + OUT_BATCH((i+0) | (i+1) << 16); + OUT_BATCH((i+3) | (i+2) << 16); + OUT_BATCH((i+0) | (i+3) << 16); } break; default: @@ -282,16 +273,16 @@ draw_arrays_generate_indices( struct vbuf_render *render, } static unsigned -draw_arrays_calc_nr_indices( uint nr, unsigned type ) +draw_arrays_calc_nr_indices(uint nr, unsigned type) { switch (type) { case 0: return nr; case PIPE_PRIM_LINE_LOOP: if (nr >= 2) - return nr * 2; + return nr * 2; else - return 0; + return 0; case PIPE_PRIM_QUADS: return (nr / 4) * 6; case PIPE_PRIM_QUAD_STRIP: @@ -303,64 +294,64 @@ draw_arrays_calc_nr_indices( uint nr, unsigned type ) } static void -draw_arrays_fallback( struct vbuf_render *render, - unsigned start, - uint nr ) +draw_arrays_fallback(struct vbuf_render *render, + unsigned start, + uint nr) { struct i915_vbuf_render *i915_render = i915_vbuf_render(render); struct i915_context *i915 = i915_render->i915; unsigned nr_indices; if (i915->dirty) - i915_update_derived( i915 ); + i915_update_derived(i915); if (i915->hardware_dirty) - i915_emit_hardware_state( i915 ); + i915_emit_hardware_state(i915); - nr_indices = draw_arrays_calc_nr_indices( nr, i915_render->fallback ); + nr_indices = draw_arrays_calc_nr_indices(nr, i915_render->fallback); if (!nr_indices) return; - if (!BEGIN_BATCH( 1 + (nr_indices + 1)/2, 1 )) { + if (!BEGIN_BATCH(1 + (nr_indices + 1)/2, 1)) { FLUSH_BATCH(NULL); /* Make sure state is re-emitted after a flush: */ - i915_update_derived( i915 ); - i915_emit_hardware_state( i915 ); + i915_update_derived(i915); + i915_emit_hardware_state(i915); i915->vbo_flushed = 1; - if (!BEGIN_BATCH( 1 + (nr_indices + 1)/2, 1 )) { - assert(0); - goto out; + if (!BEGIN_BATCH(1 + (nr_indices + 1)/2, 1)) { + assert(0); + goto out; } } - OUT_BATCH( _3DPRIMITIVE | - PRIM_INDIRECT | - i915_render->hwprim | - PRIM_INDIRECT_ELTS | - nr_indices ); + OUT_BATCH(_3DPRIMITIVE | + PRIM_INDIRECT | + i915_render->hwprim | + PRIM_INDIRECT_ELTS | + nr_indices); - draw_arrays_generate_indices( render, start, nr, i915_render->fallback ); + draw_arrays_generate_indices(render, start, nr, i915_render->fallback); out: return; } static void -i915_vbuf_render_draw_arrays( struct vbuf_render *render, - unsigned start, - uint nr ) +i915_vbuf_render_draw_arrays(struct vbuf_render *render, + unsigned start, + uint nr) { struct i915_vbuf_render *i915_render = i915_vbuf_render(render); if (i915_render->fallback) { - draw_arrays_fallback( render, start, nr ); + draw_arrays_fallback(render, start, nr); return; } /* JB: TODO submit direct cmds */ - draw_arrays_fallback( render, start, nr ); + draw_arrays_fallback(render, start, nr); } /** @@ -368,10 +359,10 @@ i915_vbuf_render_draw_arrays( struct vbuf_render *render, * If type is zero normal operation assumed. */ static void -draw_generate_indices( struct vbuf_render *render, - const ushort *indices, - uint nr_indices, - unsigned type ) +draw_generate_indices(struct vbuf_render *render, + const ushort *indices, + uint nr_indices, + unsigned type) { struct i915_vbuf_render *i915_render = i915_vbuf_render(render); struct i915_context *i915 = i915_render->i915; @@ -380,31 +371,31 @@ draw_generate_indices( struct vbuf_render *render, switch(type) { case 0: for (i = 0; i + 1 < nr_indices; i += 2) { - OUT_BATCH( indices[i] | indices[i+1] << 16 ); + OUT_BATCH(indices[i] | indices[i+1] << 16); } if (i < nr_indices) { - OUT_BATCH( indices[i] ); + OUT_BATCH(indices[i]); } break; case PIPE_PRIM_LINE_LOOP: if (nr_indices >= 2) { - for (i = 1; i < nr_indices; i++) - OUT_BATCH( indices[i-1] | indices[i] << 16 ); - OUT_BATCH( indices[i-1] | indices[0] << 16 ); + for (i = 1; i < nr_indices; i++) + OUT_BATCH(indices[i-1] | indices[i] << 16); + OUT_BATCH(indices[i-1] | indices[0] << 16); } break; case PIPE_PRIM_QUADS: for (i = 0; i + 3 < nr_indices; i += 4) { - OUT_BATCH( indices[i+0] | indices[i+1] << 16 ); - OUT_BATCH( indices[i+3] | indices[i+1] << 16 ); - OUT_BATCH( indices[i+2] | indices[i+3] << 16 ); + OUT_BATCH(indices[i+0] | indices[i+1] << 16); + OUT_BATCH(indices[i+3] | indices[i+1] << 16); + OUT_BATCH(indices[i+2] | indices[i+3] << 16); } break; case PIPE_PRIM_QUAD_STRIP: for (i = 0; i + 3 < nr_indices; i += 2) { - OUT_BATCH( indices[i+0] | indices[i+1] << 16 ); - OUT_BATCH( indices[i+3] | indices[i+2] << 16 ); - OUT_BATCH( indices[i+0] | indices[i+3] << 16 ); + OUT_BATCH(indices[i+0] | indices[i+1] << 16); + OUT_BATCH(indices[i+3] | indices[i+2] << 16); + OUT_BATCH(indices[i+0] | indices[i+3] << 16); } break; default: @@ -414,16 +405,16 @@ draw_generate_indices( struct vbuf_render *render, } static unsigned -draw_calc_nr_indices( uint nr_indices, unsigned type ) +draw_calc_nr_indices(uint nr_indices, unsigned type) { switch (type) { case 0: return nr_indices; case PIPE_PRIM_LINE_LOOP: if (nr_indices >= 2) - return nr_indices * 2; + return nr_indices * 2; else - return 0; + return 0; case PIPE_PRIM_QUADS: return (nr_indices / 4) * 6; case PIPE_PRIM_QUAD_STRIP: @@ -435,9 +426,9 @@ draw_calc_nr_indices( uint nr_indices, unsigned type ) } static void -i915_vbuf_render_draw( struct vbuf_render *render, - const ushort *indices, - uint nr_indices) +i915_vbuf_render_draw(struct vbuf_render *render, + const ushort *indices, + uint nr_indices) { struct i915_vbuf_render *i915_render = i915_vbuf_render(render); struct i915_context *i915 = i915_render->i915; @@ -445,48 +436,47 @@ i915_vbuf_render_draw( struct vbuf_render *render, save_nr_indices = nr_indices; - nr_indices = draw_calc_nr_indices( nr_indices, i915_render->fallback ); + nr_indices = draw_calc_nr_indices(nr_indices, i915_render->fallback); if (!nr_indices) return; if (i915->dirty) - i915_update_derived( i915 ); + i915_update_derived(i915); if (i915->hardware_dirty) - i915_emit_hardware_state( i915 ); + i915_emit_hardware_state(i915); - if (!BEGIN_BATCH( 1 + (nr_indices + 1)/2, 1 )) { + if (!BEGIN_BATCH(1 + (nr_indices + 1)/2, 1)) { FLUSH_BATCH(NULL); /* Make sure state is re-emitted after a flush: */ - i915_update_derived( i915 ); - i915_emit_hardware_state( i915 ); + i915_update_derived(i915); + i915_emit_hardware_state(i915); i915->vbo_flushed = 1; - if (!BEGIN_BATCH( 1 + (nr_indices + 1)/2, 1 )) { - assert(0); - goto out; + if (!BEGIN_BATCH(1 + (nr_indices + 1)/2, 1)) { + assert(0); + goto out; } } - OUT_BATCH( _3DPRIMITIVE | - PRIM_INDIRECT | - i915_render->hwprim | - PRIM_INDIRECT_ELTS | - nr_indices ); - draw_generate_indices( render, - indices, - save_nr_indices, - i915_render->fallback ); + OUT_BATCH(_3DPRIMITIVE | + PRIM_INDIRECT | + i915_render->hwprim | + PRIM_INDIRECT_ELTS | + nr_indices); + draw_generate_indices(render, + indices, + save_nr_indices, + i915_render->fallback); out: return; } - static void -i915_vbuf_render_release_vertices( struct vbuf_render *render ) +i915_vbuf_render_release_vertices(struct vbuf_render *render) { struct i915_vbuf_render *i915_render = i915_vbuf_render(render); struct i915_context *i915 = i915_render->i915; @@ -499,23 +489,21 @@ i915_vbuf_render_release_vertices( struct vbuf_render *render ) i915->dirty |= I915_NEW_VBO; } - static void -i915_vbuf_render_destroy( struct vbuf_render *render ) +i915_vbuf_render_destroy(struct vbuf_render *render) { struct i915_vbuf_render *i915_render = i915_vbuf_render(render); FREE(i915_render); } - /** * Create a new primitive render. */ static struct vbuf_render * -i915_vbuf_render_create( struct i915_context *i915 ) +i915_vbuf_render_create(struct i915_context *i915) { struct i915_vbuf_render *i915_render = CALLOC_STRUCT(i915_vbuf_render); - struct pipe_screen *screen = i915->pipe.screen; + struct intel_winsys *iws = i915->iws; i915_render->i915 = i915; @@ -539,23 +527,19 @@ i915_vbuf_render_create( struct i915_context *i915 ) i915_render->vbo_alloc_size = 128 * 4096; i915_render->vbo_size = i915_render->vbo_alloc_size; i915_render->vbo_offset = 0; - i915_render->vbo = pipe_buffer_create(screen, - 64, - I915_BUFFER_USAGE_LIT_VERTEX, - i915_render->vbo_size); - i915_render->vbo_ptr = pipe_buffer_map(screen, - i915_render->vbo, - PIPE_BUFFER_USAGE_CPU_WRITE); - pipe_buffer_unmap(screen, i915_render->vbo); + i915_render->vbo = iws->buffer_create(iws, i915_render->vbo_size, 64, + INTEL_NEW_VERTEX); + /* TODO JB: is this realy needed? */ + i915_render->vbo_ptr = iws->buffer_map(iws, i915_render->vbo, TRUE); + iws->buffer_unmap(iws, i915_render->vbo); return &i915_render->base; } - /** * Create a new primitive vbuf/render stage. */ -struct draw_stage *i915_draw_vbuf_stage( struct i915_context *i915 ) +struct draw_stage *i915_draw_vbuf_stage(struct i915_context *i915) { struct vbuf_render *render; struct draw_stage *stage; @@ -564,7 +548,7 @@ struct draw_stage *i915_draw_vbuf_stage( struct i915_context *i915 ) if(!render) return NULL; - stage = draw_vbuf_stage( i915->draw, render ); + stage = draw_vbuf_stage(i915->draw, render); if(!stage) { render->destroy(render); return NULL; diff --git a/src/gallium/drivers/i915simple/i915_screen.c b/src/gallium/drivers/i915simple/i915_screen.c index a3de38d5860..9f017a14cca 100644 --- a/src/gallium/drivers/i915simple/i915_screen.c +++ b/src/gallium/drivers/i915simple/i915_screen.c @@ -26,33 +26,36 @@ **************************************************************************/ -#include "util/u_memory.h" -#include "util/u_simple_screen.h" -#include "pipe/internal/p_winsys_screen.h" #include "pipe/p_inlines.h" +#include "util/u_memory.h" #include "util/u_string.h" #include "i915_reg.h" #include "i915_context.h" #include "i915_screen.h" +#include "i915_buffer.h" #include "i915_texture.h" -#include "i915_winsys.h" +#include "intel_winsys.h" + + +/* + * Probe functions + */ static const char * -i915_get_vendor( struct pipe_screen *pscreen ) +i915_get_vendor(struct pipe_screen *screen) { return "Tungsten Graphics, Inc."; } - static const char * -i915_get_name( struct pipe_screen *pscreen ) +i915_get_name(struct pipe_screen *screen) { static char buffer[128]; const char *chipset; - switch (i915_screen(pscreen)->pci_id) { + switch (i915_screen(screen)->pci_id) { case PCI_CHIP_I915_G: chipset = "915G"; break; @@ -86,7 +89,6 @@ i915_get_name( struct pipe_screen *pscreen ) return buffer; } - static int i915_get_param(struct pipe_screen *screen, int param) { @@ -122,7 +124,6 @@ i915_get_param(struct pipe_screen *screen, int param) } } - static float i915_get_paramf(struct pipe_screen *screen, int param) { @@ -148,13 +149,12 @@ i915_get_paramf(struct pipe_screen *screen, int param) } } - static boolean -i915_is_format_supported( struct pipe_screen *screen, - enum pipe_format format, - enum pipe_texture_target target, - unsigned tex_usage, - unsigned geom_flags ) +i915_is_format_supported(struct pipe_screen *screen, + enum pipe_format format, + enum pipe_texture_target target, + unsigned tex_usage, + unsigned geom_flags) { static const enum pipe_format tex_supported[] = { PIPE_FORMAT_R8G8B8A8_UNORM, @@ -173,7 +173,6 @@ i915_is_format_supported( struct pipe_screen *screen, PIPE_FORMAT_A8R8G8B8_UNORM, PIPE_FORMAT_R5G6B5_UNORM, PIPE_FORMAT_S8Z24_UNORM, - /*PIPE_FORMAT_R16G16B16A16_SNORM,*/ PIPE_FORMAT_NONE /* list terminator */ }; const enum pipe_format *list; @@ -193,122 +192,73 @@ i915_is_format_supported( struct pipe_screen *screen, } -static void -i915_destroy_screen( struct pipe_screen *screen ) -{ - struct pipe_winsys *winsys = screen->winsys; - - if(winsys->destroy) - winsys->destroy(winsys); - - FREE(screen); -} +/* + * Fence functions + */ -static struct pipe_transfer* -i915_get_tex_transfer(struct pipe_screen *screen, - struct pipe_texture *texture, - unsigned face, unsigned level, unsigned zslice, - enum pipe_transfer_usage usage, unsigned x, unsigned y, - unsigned w, unsigned h) +static void +i915_fence_reference(struct pipe_screen *screen, + struct pipe_fence_handle **ptr, + struct pipe_fence_handle *fence) { - struct i915_texture *tex = (struct i915_texture *)texture; - struct i915_transfer *trans; - unsigned offset; /* in bytes */ + struct i915_screen *is = i915_screen(screen); - if (texture->target == PIPE_TEXTURE_CUBE) { - offset = tex->image_offset[level][face]; - } - else if (texture->target == PIPE_TEXTURE_3D) { - offset = tex->image_offset[level][zslice]; - } - else { - offset = tex->image_offset[level][0]; - assert(face == 0); - assert(zslice == 0); - } - - trans = CALLOC_STRUCT(i915_transfer); - if (trans) { - pipe_texture_reference(&trans->base.texture, texture); - trans->base.format = trans->base.format; - trans->base.x = x; - trans->base.y = y; - trans->base.width = w; - trans->base.height = h; - trans->base.block = texture->block; - trans->base.nblocksx = texture->nblocksx[level]; - trans->base.nblocksy = texture->nblocksy[level]; - trans->base.stride = tex->stride; - trans->offset = offset; - trans->base.usage = usage; - } - return &trans->base; + is->iws->fence_reference(is->iws, ptr, fence); } -static void -i915_tex_transfer_destroy(struct pipe_transfer *trans) +static int +i915_fence_signalled(struct pipe_screen *screen, + struct pipe_fence_handle *fence, + unsigned flags) { - pipe_texture_reference(&trans->texture, NULL); - FREE(trans); + struct i915_screen *is = i915_screen(screen); + + return is->iws->fence_signalled(is->iws, fence); } -static void * -i915_transfer_map( struct pipe_screen *screen, - struct pipe_transfer *transfer ) +static int +i915_fence_finish(struct pipe_screen *screen, + struct pipe_fence_handle *fence, + unsigned flags) { - struct i915_texture *tex = (struct i915_texture *)transfer->texture; - char *map; - unsigned flags = 0; + struct i915_screen *is = i915_screen(screen); - if (transfer->usage != PIPE_TRANSFER_WRITE) - flags |= PIPE_BUFFER_USAGE_CPU_READ; + return is->iws->fence_finish(is->iws, fence); +} - if (transfer->usage != PIPE_TRANSFER_READ) - flags |= PIPE_BUFFER_USAGE_CPU_WRITE; - map = pipe_buffer_map( screen, tex->buffer, flags ); - if (map == NULL) - return NULL; +/* + * Generic functions + */ - if (transfer->texture && - (flags & PIPE_BUFFER_USAGE_CPU_WRITE)) - { - /* Do something to notify contexts of a texture change. - */ - /* i915_screen(screen)->timestamp++; */ - } - - return map + i915_transfer(transfer)->offset + - transfer->y / transfer->block.height * transfer->stride + - transfer->x / transfer->block.width * transfer->block.size; -} static void -i915_transfer_unmap(struct pipe_screen *screen, - struct pipe_transfer *transfer) +i915_destroy_screen(struct pipe_screen *screen) { - struct i915_texture *tex = (struct i915_texture *)transfer->texture; - pipe_buffer_unmap( screen, tex->buffer ); -} + struct i915_screen *is = i915_screen(screen); + if (is->iws) + is->iws->destroy(is->iws); + FREE(is); +} /** * Create a new i915_screen object */ struct pipe_screen * -i915_create_screen(struct pipe_winsys *winsys, uint pci_id) +i915_create_screen(struct intel_winsys *iws, uint pci_id) { - struct i915_screen *i915screen = CALLOC_STRUCT(i915_screen); + struct i915_screen *is = CALLOC_STRUCT(i915_screen); - if (!i915screen) + if (!is) return NULL; switch (pci_id) { case PCI_CHIP_I915_G: case PCI_CHIP_I915_GM: - i915screen->is_i945 = FALSE; + is->is_i945 = FALSE; break; case PCI_CHIP_I945_G: @@ -317,7 +267,7 @@ i915_create_screen(struct pipe_winsys *winsys, uint pci_id) case PCI_CHIP_G33_G: case PCI_CHIP_Q33_G: case PCI_CHIP_Q35_G: - i915screen->is_i945 = TRUE; + is->is_i945 = TRUE; break; default: @@ -326,24 +276,25 @@ i915_create_screen(struct pipe_winsys *winsys, uint pci_id) return NULL; } - i915screen->pci_id = pci_id; + is->pci_id = pci_id; + is->iws = iws; + + is->base.winsys = NULL; - i915screen->screen.winsys = winsys; + is->base.destroy = i915_destroy_screen; - i915screen->screen.destroy = i915_destroy_screen; + is->base.get_name = i915_get_name; + is->base.get_vendor = i915_get_vendor; + is->base.get_param = i915_get_param; + is->base.get_paramf = i915_get_paramf; + is->base.is_format_supported = i915_is_format_supported; - i915screen->screen.get_name = i915_get_name; - i915screen->screen.get_vendor = i915_get_vendor; - i915screen->screen.get_param = i915_get_param; - i915screen->screen.get_paramf = i915_get_paramf; - i915screen->screen.is_format_supported = i915_is_format_supported; - i915screen->screen.get_tex_transfer = i915_get_tex_transfer; - i915screen->screen.tex_transfer_destroy = i915_tex_transfer_destroy; - i915screen->screen.transfer_map = i915_transfer_map; - i915screen->screen.transfer_unmap = i915_transfer_unmap; + is->base.fence_reference = i915_fence_reference; + is->base.fence_signalled = i915_fence_signalled; + is->base.fence_finish = i915_fence_finish; - i915_init_screen_texture_functions(&i915screen->screen); - u_simple_screen_init(&i915screen->screen); + i915_init_screen_texture_functions(is); + i915_init_screen_buffer_functions(is); - return &i915screen->screen; + return &is->base; } diff --git a/src/gallium/drivers/i915simple/i915_screen.h b/src/gallium/drivers/i915simple/i915_screen.h index 5284c325951..5126485caa7 100644 --- a/src/gallium/drivers/i915simple/i915_screen.h +++ b/src/gallium/drivers/i915simple/i915_screen.h @@ -25,17 +25,14 @@ * **************************************************************************/ - #ifndef I915_SCREEN_H #define I915_SCREEN_H - +#include "pipe/p_state.h" #include "pipe/p_screen.h" -#ifdef __cplusplus -extern "C" { -#endif +struct intel_winsys; /** @@ -43,13 +40,14 @@ extern "C" { */ struct i915_screen { - struct pipe_screen screen; + struct pipe_screen base; + + struct intel_winsys *iws; boolean is_i945; uint pci_id; }; - /** * Subclass of pipe_transfer */ @@ -61,7 +59,11 @@ struct i915_transfer }; -/** cast wrappers */ +/* + * Cast wrappers + */ + + static INLINE struct i915_screen * i915_screen(struct pipe_screen *pscreen) { @@ -69,14 +71,10 @@ i915_screen(struct pipe_screen *pscreen) } static INLINE struct i915_transfer * -i915_transfer( struct pipe_transfer *transfer ) +i915_transfer(struct pipe_transfer *transfer) { return (struct i915_transfer *)transfer; } -#ifdef __cplusplus -} -#endif - #endif /* I915_SCREEN_H */ diff --git a/src/gallium/drivers/i915simple/i915_state.c b/src/gallium/drivers/i915simple/i915_state.c index 273e74002aa..0087dfa410f 100644 --- a/src/gallium/drivers/i915simple/i915_state.c +++ b/src/gallium/drivers/i915simple/i915_state.c @@ -518,7 +518,7 @@ static void i915_set_constant_buffer(struct pipe_context *pipe, const struct pipe_constant_buffer *buf) { struct i915_context *i915 = i915_context(pipe); - struct pipe_winsys *ws = pipe->winsys; + struct pipe_screen *screen = pipe->screen; draw_flush(i915->draw); assert(shader < PIPE_SHADER_TYPES); @@ -536,10 +536,10 @@ static void i915_set_constant_buffer(struct pipe_context *pipe, if (buf) { void *mapped; if (buf->buffer && buf->buffer->size && - (mapped = ws->buffer_map(ws, buf->buffer, + (mapped = pipe_buffer_map(screen, buf->buffer, PIPE_BUFFER_USAGE_CPU_READ))) { memcpy(i915->current.constants[shader], mapped, buf->buffer->size); - ws->buffer_unmap(ws, buf->buffer); + pipe_buffer_unmap(screen, buf->buffer); i915->current.num_user_constants[shader] = buf->buffer->size / (4 * sizeof(float)); } @@ -751,38 +751,38 @@ static void i915_set_edgeflags(struct pipe_context *pipe, void i915_init_state_functions( struct i915_context *i915 ) { - i915->pipe.set_edgeflags = i915_set_edgeflags; - i915->pipe.create_blend_state = i915_create_blend_state; - i915->pipe.bind_blend_state = i915_bind_blend_state; - i915->pipe.delete_blend_state = i915_delete_blend_state; - - i915->pipe.create_sampler_state = i915_create_sampler_state; - i915->pipe.bind_sampler_states = i915_bind_sampler_states; - i915->pipe.delete_sampler_state = i915_delete_sampler_state; - - i915->pipe.create_depth_stencil_alpha_state = i915_create_depth_stencil_state; - i915->pipe.bind_depth_stencil_alpha_state = i915_bind_depth_stencil_state; - i915->pipe.delete_depth_stencil_alpha_state = i915_delete_depth_stencil_state; - - i915->pipe.create_rasterizer_state = i915_create_rasterizer_state; - i915->pipe.bind_rasterizer_state = i915_bind_rasterizer_state; - i915->pipe.delete_rasterizer_state = i915_delete_rasterizer_state; - i915->pipe.create_fs_state = i915_create_fs_state; - i915->pipe.bind_fs_state = i915_bind_fs_state; - i915->pipe.delete_fs_state = i915_delete_fs_state; - i915->pipe.create_vs_state = i915_create_vs_state; - i915->pipe.bind_vs_state = i915_bind_vs_state; - i915->pipe.delete_vs_state = i915_delete_vs_state; - - i915->pipe.set_blend_color = i915_set_blend_color; - i915->pipe.set_clip_state = i915_set_clip_state; - i915->pipe.set_constant_buffer = i915_set_constant_buffer; - i915->pipe.set_framebuffer_state = i915_set_framebuffer_state; - - i915->pipe.set_polygon_stipple = i915_set_polygon_stipple; - i915->pipe.set_scissor_state = i915_set_scissor_state; - i915->pipe.set_sampler_textures = i915_set_sampler_textures; - i915->pipe.set_viewport_state = i915_set_viewport_state; - i915->pipe.set_vertex_buffers = i915_set_vertex_buffers; - i915->pipe.set_vertex_elements = i915_set_vertex_elements; + i915->base.set_edgeflags = i915_set_edgeflags; + i915->base.create_blend_state = i915_create_blend_state; + i915->base.bind_blend_state = i915_bind_blend_state; + i915->base.delete_blend_state = i915_delete_blend_state; + + i915->base.create_sampler_state = i915_create_sampler_state; + i915->base.bind_sampler_states = i915_bind_sampler_states; + i915->base.delete_sampler_state = i915_delete_sampler_state; + + i915->base.create_depth_stencil_alpha_state = i915_create_depth_stencil_state; + i915->base.bind_depth_stencil_alpha_state = i915_bind_depth_stencil_state; + i915->base.delete_depth_stencil_alpha_state = i915_delete_depth_stencil_state; + + i915->base.create_rasterizer_state = i915_create_rasterizer_state; + i915->base.bind_rasterizer_state = i915_bind_rasterizer_state; + i915->base.delete_rasterizer_state = i915_delete_rasterizer_state; + i915->base.create_fs_state = i915_create_fs_state; + i915->base.bind_fs_state = i915_bind_fs_state; + i915->base.delete_fs_state = i915_delete_fs_state; + i915->base.create_vs_state = i915_create_vs_state; + i915->base.bind_vs_state = i915_bind_vs_state; + i915->base.delete_vs_state = i915_delete_vs_state; + + i915->base.set_blend_color = i915_set_blend_color; + i915->base.set_clip_state = i915_set_clip_state; + i915->base.set_constant_buffer = i915_set_constant_buffer; + i915->base.set_framebuffer_state = i915_set_framebuffer_state; + + i915->base.set_polygon_stipple = i915_set_polygon_stipple; + i915->base.set_scissor_state = i915_set_scissor_state; + i915->base.set_sampler_textures = i915_set_sampler_textures; + i915->base.set_viewport_state = i915_set_viewport_state; + i915->base.set_vertex_buffers = i915_set_vertex_buffers; + i915->base.set_vertex_elements = i915_set_vertex_elements; } diff --git a/src/gallium/drivers/i915simple/i915_state_emit.c b/src/gallium/drivers/i915simple/i915_state_emit.c index 1e1fb968b47..a3d4e3b04e5 100644 --- a/src/gallium/drivers/i915simple/i915_state_emit.c +++ b/src/gallium/drivers/i915simple/i915_state_emit.c @@ -28,7 +28,6 @@ #include "i915_reg.h" #include "i915_context.h" -#include "i915_winsys.h" #include "i915_batch.h" #include "i915_reg.h" @@ -107,7 +106,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) 6 ) * 3/2; /* plus 50% margin */ const unsigned relocs = ( I915_TEX_UNITS + - 3 + 3 ) * 3/2; /* plus 50% margin */ #if 0 @@ -123,9 +122,9 @@ i915_emit_hardware_state(struct i915_context *i915 ) if (i915->hardware_dirty & I915_HW_INVARIENT) { OUT_BATCH(_3DSTATE_AA_CMD | - AA_LINE_ECAAR_WIDTH_ENABLE | - AA_LINE_ECAAR_WIDTH_1_0 | - AA_LINE_REGION_WIDTH_ENABLE | AA_LINE_REGION_WIDTH_1_0); + AA_LINE_ECAAR_WIDTH_ENABLE | + AA_LINE_ECAAR_WIDTH_1_0 | + AA_LINE_REGION_WIDTH_ENABLE | AA_LINE_REGION_WIDTH_1_0); OUT_BATCH(_3DSTATE_DFLT_DIFFUSE_CMD); OUT_BATCH(0); @@ -137,24 +136,24 @@ i915_emit_hardware_state(struct i915_context *i915 ) OUT_BATCH(0); OUT_BATCH(_3DSTATE_COORD_SET_BINDINGS | - CSB_TCB(0, 0) | - CSB_TCB(1, 1) | - CSB_TCB(2, 2) | - CSB_TCB(3, 3) | - CSB_TCB(4, 4) | - CSB_TCB(5, 5) | - CSB_TCB(6, 6) | - CSB_TCB(7, 7)); + CSB_TCB(0, 0) | + CSB_TCB(1, 1) | + CSB_TCB(2, 2) | + CSB_TCB(3, 3) | + CSB_TCB(4, 4) | + CSB_TCB(5, 5) | + CSB_TCB(6, 6) | + CSB_TCB(7, 7)); OUT_BATCH(_3DSTATE_RASTER_RULES_CMD | - ENABLE_POINT_RASTER_RULE | - OGL_POINT_RASTER_RULE | - ENABLE_LINE_STRIP_PROVOKE_VRTX | - ENABLE_TRI_FAN_PROVOKE_VRTX | - LINE_STRIP_PROVOKE_VRTX(1) | - TRI_FAN_PROVOKE_VRTX(2) | - ENABLE_TEXKILL_3D_4D | - TEXKILL_4D); + ENABLE_POINT_RASTER_RULE | + OGL_POINT_RASTER_RULE | + ENABLE_LINE_STRIP_PROVOKE_VRTX | + ENABLE_TRI_FAN_PROVOKE_VRTX | + LINE_STRIP_PROVOKE_VRTX(1) | + TRI_FAN_PROVOKE_VRTX(2) | + ENABLE_TEXKILL_3D_4D | + TEXKILL_4D); /* Need to initialize this to zero. */ @@ -173,21 +172,21 @@ i915_emit_hardware_state(struct i915_context *i915 ) if (i915->hardware_dirty & I915_HW_IMMEDIATE) { OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 | - I1_LOAD_S(0) | - I1_LOAD_S(1) | - I1_LOAD_S(2) | - I1_LOAD_S(4) | - I1_LOAD_S(5) | - I1_LOAD_S(6) | - (5)); + I1_LOAD_S(0) | + I1_LOAD_S(1) | + I1_LOAD_S(2) | + I1_LOAD_S(4) | + I1_LOAD_S(5) | + I1_LOAD_S(6) | + (5)); if(i915->vbo) OUT_RELOC(i915->vbo, - I915_BUFFER_ACCESS_READ, + INTEL_USAGE_VERTEX, i915->current.immediate[I915_IMMEDIATE_S0]); else - /* FIXME: we should not do this */ - OUT_BATCH(0); + /* FIXME: we should not do this */ + OUT_BATCH(0); OUT_BATCH(i915->current.immediate[I915_IMMEDIATE_S1]); OUT_BATCH(i915->current.immediate[I915_IMMEDIATE_S2]); OUT_BATCH(i915->current.immediate[I915_IMMEDIATE_S4]); @@ -200,7 +199,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) { int i; for (i = 0; i < I915_MAX_DYNAMIC; i++) { - OUT_BATCH(i915->current.dynamic[i]); + OUT_BATCH(i915->current.dynamic[i]); } } @@ -211,68 +210,68 @@ i915_emit_hardware_state(struct i915_context *i915 ) struct pipe_surface *depth_surface = i915->framebuffer.zsbuf; if (cbuf_surface) { - unsigned ctile = BUF_3D_USE_FENCE; + unsigned ctile = BUF_3D_USE_FENCE; struct i915_texture *tex = (struct i915_texture *) cbuf_surface->texture; assert(tex); - if (tex && tex->tiled) { - ctile = BUF_3D_TILED_SURFACE; - } + if (tex && tex->sw_tiled) { + ctile = BUF_3D_TILED_SURFACE; + } - OUT_BATCH(_3DSTATE_BUF_INFO_CMD); + OUT_BATCH(_3DSTATE_BUF_INFO_CMD); - OUT_BATCH(BUF_3D_ID_COLOR_BACK | - BUF_3D_PITCH(tex->stride) | /* pitch in bytes */ - ctile); + OUT_BATCH(BUF_3D_ID_COLOR_BACK | + BUF_3D_PITCH(tex->stride) | /* pitch in bytes */ + ctile); - OUT_RELOC(tex->buffer, - I915_BUFFER_ACCESS_WRITE, - cbuf_surface->offset); + OUT_RELOC(tex->buffer, + INTEL_USAGE_RENDER, + cbuf_surface->offset); } /* What happens if no zbuf?? */ if (depth_surface) { - unsigned ztile = BUF_3D_USE_FENCE; + unsigned ztile = BUF_3D_USE_FENCE; struct i915_texture *tex = (struct i915_texture *) depth_surface->texture; assert(tex); - if (tex && tex->tiled) { - ztile = BUF_3D_TILED_SURFACE; - } + if (tex && tex->sw_tiled) { + ztile = BUF_3D_TILED_SURFACE; + } - OUT_BATCH(_3DSTATE_BUF_INFO_CMD); + OUT_BATCH(_3DSTATE_BUF_INFO_CMD); - OUT_BATCH(BUF_3D_ID_DEPTH | - BUF_3D_PITCH(tex->stride) | /* pitch in bytes */ - ztile); + OUT_BATCH(BUF_3D_ID_DEPTH | + BUF_3D_PITCH(tex->stride) | /* pitch in bytes */ + ztile); - OUT_RELOC(tex->buffer, - I915_BUFFER_ACCESS_WRITE, - depth_surface->offset); + OUT_RELOC(tex->buffer, + INTEL_USAGE_RENDER, + depth_surface->offset); } { - unsigned cformat, zformat = 0; + unsigned cformat, zformat = 0; - if (cbuf_surface) + if (cbuf_surface) cformat = cbuf_surface->format; else cformat = PIPE_FORMAT_A8R8G8B8_UNORM; /* arbitrary */ cformat = translate_format(cformat); - if (depth_surface) - zformat = translate_depth_format( i915->framebuffer.zsbuf->format ); + if (depth_surface) + zformat = translate_depth_format( i915->framebuffer.zsbuf->format ); - OUT_BATCH(_3DSTATE_DST_BUF_VARS_CMD); - OUT_BATCH(DSTORG_HORT_BIAS(0x8) | /* .5 */ - DSTORG_VERT_BIAS(0x8) | /* .5 */ - LOD_PRECLAMP_OGL | - TEX_DEFAULT_COLOR_OGL | - cformat | - zformat ); + OUT_BATCH(_3DSTATE_DST_BUF_VARS_CMD); + OUT_BATCH(DSTORG_HORT_BIAS(0x8) | /* .5 */ + DSTORG_VERT_BIAS(0x8) | /* .5 */ + LOD_PRECLAMP_OGL | + TEX_DEFAULT_COLOR_OGL | + cformat | + zformat ); } } @@ -290,16 +289,13 @@ i915_emit_hardware_state(struct i915_context *i915 ) OUT_BATCH(enabled); for (unit = 0; unit < I915_TEX_UNITS; unit++) { if (enabled & (1 << unit)) { - struct pipe_buffer *buf = - i915->texture[unit]->buffer; + struct intel_buffer *buf = i915->texture[unit]->buffer; uint offset = 0; assert(buf); count++; - OUT_RELOC(buf, - I915_BUFFER_ACCESS_READ, - offset); + OUT_RELOC(buf, INTEL_USAGE_SAMPLER, offset); OUT_BATCH(i915->current.texbuffer[unit][0]); /* MS3 */ OUT_BATCH(i915->current.texbuffer[unit][1]); /* MS4 */ } @@ -315,20 +311,20 @@ i915_emit_hardware_state(struct i915_context *i915 ) if (i915->hardware_dirty & I915_HW_SAMPLER) { if (i915->current.sampler_enable_nr) { - int i; - - OUT_BATCH( _3DSTATE_SAMPLER_STATE | - (3 * i915->current.sampler_enable_nr) ); - - OUT_BATCH( i915->current.sampler_enable_flags ); - - for (i = 0; i < I915_TEX_UNITS; i++) { - if (i915->current.sampler_enable_flags & (1<<i)) { - OUT_BATCH( i915->current.sampler[i][0] ); - OUT_BATCH( i915->current.sampler[i][1] ); - OUT_BATCH( i915->current.sampler[i][2] ); - } - } + int i; + + OUT_BATCH( _3DSTATE_SAMPLER_STATE | + (3 * i915->current.sampler_enable_nr) ); + + OUT_BATCH( i915->current.sampler_enable_flags ); + + for (i = 0; i < I915_TEX_UNITS; i++) { + if (i915->current.sampler_enable_flags & (1<<i)) { + OUT_BATCH( i915->current.sampler[i][0] ); + OUT_BATCH( i915->current.sampler[i][1] ); + OUT_BATCH( i915->current.sampler[i][2] ); + } + } } } #endif diff --git a/src/gallium/drivers/i915simple/i915_state_sampler.c b/src/gallium/drivers/i915simple/i915_state_sampler.c index 3667ed1afa7..c5e9084d12e 100644 --- a/src/gallium/drivers/i915simple/i915_state_sampler.c +++ b/src/gallium/drivers/i915simple/i915_state_sampler.c @@ -247,7 +247,7 @@ i915_update_texture(struct i915_context *i915, assert(format); assert(pitch); - if (tex->tiled) { + if (tex->sw_tiled) { assert(!((pitch - 1) & pitch)); tiled = MS3_TILED_SURFACE; } diff --git a/src/gallium/drivers/i915simple/i915_surface.c b/src/gallium/drivers/i915simple/i915_surface.c index 09b2c499b8f..ab8331f3e64 100644 --- a/src/gallium/drivers/i915simple/i915_surface.c +++ b/src/gallium/drivers/i915simple/i915_surface.c @@ -89,6 +89,6 @@ i915_surface_fill(struct pipe_context *pipe, void i915_init_surface_functions(struct i915_context *i915) { - i915->pipe.surface_copy = i915_surface_copy; - i915->pipe.surface_fill = i915_surface_fill; + i915->base.surface_copy = i915_surface_copy; + i915->base.surface_fill = i915_surface_fill; } diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index 0dc6a547463..6a6c6542717 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -42,12 +42,14 @@ #include "i915_texture.h" #include "i915_debug.h" #include "i915_screen.h" -#include "i915_winsys.h" +#include "intel_winsys.h" + /* * Helper function and arrays */ + /** * Initial offset for Cube map. */ @@ -133,7 +135,7 @@ i915_miptree_set_level_info(struct i915_texture *tex, static void i915_miptree_set_image_offset(struct i915_texture *tex, - unsigned level, unsigned img, unsigned x, unsigned y) + unsigned level, unsigned img, unsigned x, unsigned y) { if (img == 0 && level == 0) assert(x == 0 && y == 0); @@ -150,7 +152,7 @@ i915_miptree_set_image_offset(struct i915_texture *tex, /* - * Layout functions + * i915 layout functions, some used by i945 */ @@ -165,37 +167,174 @@ i915_scanout_layout(struct i915_texture *tex) if (pt->last_level > 0 || pt->block.size != 4) return 0; - i915_miptree_set_level_info( tex, 0, 1, - tex->base.width[0], - tex->base.height[0], - 1 ); - i915_miptree_set_image_offset( tex, 0, 0, 0, 0 ); + i915_miptree_set_level_info(tex, 0, 1, + tex->base.width[0], + tex->base.height[0], + 1); + i915_miptree_set_image_offset(tex, 0, 0, 0, 0); -#if 0 /* TODO use this code when backend is smarter */ if (tex->base.width[0] >= 240) { tex->stride = power_of_two(tex->base.nblocksx[0] * pt->block.size); tex->total_nblocksy = round_up(tex->base.nblocksy[0], 8); -#else - if (tex->base.width[0] >= 240) { - tex->stride = 2048 * 4; - tex->total_nblocksy = round_up(tex->base.nblocksy[0], 8); -#endif + tex->hw_tiled = INTEL_TILE_X; } else if (tex->base.width[0] == 64 && tex->base.height[0] == 64) { tex->stride = power_of_two(tex->base.nblocksx[0] * pt->block.size); tex->total_nblocksy = round_up(tex->base.nblocksy[0], 8); } else { - return 0; + return FALSE; } debug_printf("%s size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__, tex->base.width[0], tex->base.height[0], pt->block.size, tex->stride, tex->total_nblocksy, tex->stride * tex->total_nblocksy); - return 1; + return TRUE; +} + +static void +i915_miptree_layout_2d(struct i915_texture *tex) +{ + struct pipe_texture *pt = &tex->base; + unsigned level; + unsigned width = pt->width[0]; + unsigned height = pt->height[0]; + unsigned nblocksx = pt->nblocksx[0]; + unsigned nblocksy = pt->nblocksy[0]; + + tex->stride = round_up(pt->nblocksx[0] * pt->block.size, 4); + tex->total_nblocksy = 0; + + for (level = 0; level <= pt->last_level; level++) { + i915_miptree_set_level_info(tex, level, 1, width, height, 1); + i915_miptree_set_image_offset(tex, level, 0, 0, tex->total_nblocksy); + + nblocksy = round_up(MAX2(2, nblocksy), 2); + + tex->total_nblocksy += nblocksy; + + width = minify(width); + height = minify(height); + nblocksx = pf_get_nblocksx(&pt->block, width); + nblocksy = pf_get_nblocksy(&pt->block, height); + } +} + +static void +i915_miptree_layout_3d(struct i915_texture *tex) +{ + struct pipe_texture *pt = &tex->base; + unsigned level; + + unsigned width = pt->width[0]; + unsigned height = pt->height[0]; + unsigned depth = pt->depth[0]; + unsigned nblocksx = pt->nblocksx[0]; + unsigned nblocksy = pt->nblocksy[0]; + unsigned stack_nblocksy = 0; + + /* Calculate the size of a single slice. + */ + tex->stride = round_up(pt->nblocksx[0] * pt->block.size, 4); + + /* XXX: hardware expects/requires 9 levels at minimum. + */ + for (level = 0; level <= MAX2(8, pt->last_level); level++) { + i915_miptree_set_level_info(tex, level, depth, width, height, depth); + + stack_nblocksy += MAX2(2, nblocksy); + + width = minify(width); + height = minify(height); + depth = minify(depth); + nblocksx = pf_get_nblocksx(&pt->block, width); + nblocksy = pf_get_nblocksy(&pt->block, height); + } + + /* Fixup depth image_offsets: + */ + depth = pt->depth[0]; + for (level = 0; level <= pt->last_level; level++) { + unsigned i; + for (i = 0; i < depth; i++) + i915_miptree_set_image_offset(tex, level, i, 0, i * stack_nblocksy); + + depth = minify(depth); + } + + /* Multiply slice size by texture depth for total size. It's + * remarkable how wasteful of memory the i915 texture layouts + * are. They are largely fixed in the i945. + */ + tex->total_nblocksy = stack_nblocksy * pt->depth[0]; } static void -i945_miptree_layout_2d( struct i915_texture *tex ) +i915_miptree_layout_cube(struct i915_texture *tex) +{ + struct pipe_texture *pt = &tex->base; + unsigned width = pt->width[0], height = pt->height[0]; + const unsigned nblocks = pt->nblocksx[0]; + unsigned level; + unsigned face; + + assert(width == height); /* cubemap images are square */ + + /* double pitch for cube layouts */ + tex->stride = round_up(nblocks * pt->block.size * 2, 4); + tex->total_nblocksy = nblocks * 4; + + for (level = 0; level <= pt->last_level; level++) { + i915_miptree_set_level_info(tex, level, 6, width, height, 1); + width /= 2; + height /= 2; + } + + for (face = 0; face < 6; face++) { + unsigned x = initial_offsets[face][0] * nblocks; + unsigned y = initial_offsets[face][1] * nblocks; + unsigned d = nblocks; + + for (level = 0; level <= pt->last_level; level++) { + i915_miptree_set_image_offset(tex, level, face, x, y); + d >>= 1; + x += step_offsets[face][0] * d; + y += step_offsets[face][1] * d; + } + } +} + +static boolean +i915_miptree_layout(struct i915_texture * tex) +{ + struct pipe_texture *pt = &tex->base; + + switch (pt->target) { + case PIPE_TEXTURE_1D: + case PIPE_TEXTURE_2D: + i915_miptree_layout_2d(tex); + break; + case PIPE_TEXTURE_3D: + i915_miptree_layout_3d(tex); + break; + case PIPE_TEXTURE_CUBE: + i915_miptree_layout_cube(tex); + break; + default: + assert(0); + return FALSE; + } + + return TRUE; +} + + +/* + * i945 layout functions + */ + + +static void +i945_miptree_layout_2d(struct i915_texture *tex) { struct pipe_texture *pt = &tex->base; const int align_x = 2, align_y = 4; @@ -210,7 +349,7 @@ i945_miptree_layout_2d( struct i915_texture *tex ) /* used for scanouts that need special layouts */ if (tex->base.tex_usage & PIPE_TEXTURE_USAGE_PRIMARY) if (i915_scanout_layout(tex)) - return; + return; tex->stride = round_up(pt->nblocksx[0] * pt->block.size, 4); @@ -221,11 +360,11 @@ i945_miptree_layout_2d( struct i915_texture *tex ) */ if (pt->last_level > 0) { unsigned mip1_nblocksx - = align(pf_get_nblocksx(&pt->block, minify(width)), align_x) + = align(pf_get_nblocksx(&pt->block, minify(width)), align_x) + pf_get_nblocksx(&pt->block, minify(minify(width))); if (mip1_nblocksx > nblocksx) - tex->stride = mip1_nblocksx * pt->block.size; + tex->stride = mip1_nblocksx * pt->block.size; } /* Pitch must be a whole number of dwords @@ -247,10 +386,10 @@ i945_miptree_layout_2d( struct i915_texture *tex ) /* Layout_below: step right after second mipmap level. */ if (level == 1) { - x += align(nblocksx, align_x); + x += align(nblocksx, align_x); } else { - y += nblocksy; + y += nblocksy; } width = minify(width); @@ -261,6 +400,63 @@ i945_miptree_layout_2d( struct i915_texture *tex ) } static void +i945_miptree_layout_3d(struct i915_texture *tex) +{ + struct pipe_texture *pt = &tex->base; + unsigned width = pt->width[0]; + unsigned height = pt->height[0]; + unsigned depth = pt->depth[0]; + unsigned nblocksx = pt->nblocksx[0]; + unsigned nblocksy = pt->nblocksy[0]; + unsigned pack_x_pitch, pack_x_nr; + unsigned pack_y_pitch; + unsigned level; + + tex->stride = round_up(pt->nblocksx[0] * pt->block.size, 4); + tex->total_nblocksy = 0; + + pack_y_pitch = MAX2(pt->nblocksy[0], 2); + pack_x_pitch = tex->stride / pt->block.size; + pack_x_nr = 1; + + for (level = 0; level <= pt->last_level; level++) { + int x = 0; + int y = 0; + unsigned q, j; + + i915_miptree_set_level_info(tex, level, depth, width, height, depth); + + for (q = 0; q < depth;) { + for (j = 0; j < pack_x_nr && q < depth; j++, q++) { + i915_miptree_set_image_offset(tex, level, q, x, y + tex->total_nblocksy); + x += pack_x_pitch; + } + + x = 0; + y += pack_y_pitch; + } + + tex->total_nblocksy += y; + + if (pack_x_pitch > 4) { + pack_x_pitch >>= 1; + pack_x_nr <<= 1; + assert(pack_x_pitch * pack_x_nr * pt->block.size <= tex->stride); + } + + if (pack_y_pitch > 2) { + pack_y_pitch >>= 1; + } + + width = minify(width); + height = minify(height); + depth = minify(depth); + nblocksx = pf_get_nblocksx(&pt->block, width); + nblocksy = pf_get_nblocksy(&pt->block, height); + } +} + +static void i945_miptree_layout_cube(struct i915_texture *tex) { struct pipe_texture *pt = &tex->base; @@ -362,224 +558,41 @@ i945_miptree_layout_cube(struct i915_texture *tex) } static boolean -i915_miptree_layout(struct i915_texture * tex) -{ - struct pipe_texture *pt = &tex->base; - unsigned level; - - switch (pt->target) { - case PIPE_TEXTURE_CUBE: { - const unsigned nblocks = pt->nblocksx[0]; - unsigned face; - unsigned width = pt->width[0], height = pt->height[0]; - - assert(width == height); /* cubemap images are square */ - - /* double pitch for cube layouts */ - tex->stride = round_up(nblocks * pt->block.size * 2, 4); - tex->total_nblocksy = nblocks * 4; - - for (level = 0; level <= pt->last_level; level++) { - i915_miptree_set_level_info(tex, level, 6, - width, height, - 1); - width /= 2; - height /= 2; - } - - for (face = 0; face < 6; face++) { - unsigned x = initial_offsets[face][0] * nblocks; - unsigned y = initial_offsets[face][1] * nblocks; - unsigned d = nblocks; - - for (level = 0; level <= pt->last_level; level++) { - i915_miptree_set_image_offset(tex, level, face, x, y); - d >>= 1; - x += step_offsets[face][0] * d; - y += step_offsets[face][1] * d; - } - } - break; - } - case PIPE_TEXTURE_3D:{ - unsigned width = pt->width[0]; - unsigned height = pt->height[0]; - unsigned depth = pt->depth[0]; - unsigned nblocksx = pt->nblocksx[0]; - unsigned nblocksy = pt->nblocksy[0]; - unsigned stack_nblocksy = 0; - - /* Calculate the size of a single slice. - */ - tex->stride = round_up(pt->nblocksx[0] * pt->block.size, 4); - - /* XXX: hardware expects/requires 9 levels at minimum. - */ - for (level = 0; level <= MAX2(8, pt->last_level); - level++) { - i915_miptree_set_level_info(tex, level, depth, - width, height, depth); - - - stack_nblocksy += MAX2(2, nblocksy); - - width = minify(width); - height = minify(height); - depth = minify(depth); - nblocksx = pf_get_nblocksx(&pt->block, width); - nblocksy = pf_get_nblocksy(&pt->block, height); - } - - /* Fixup depth image_offsets: - */ - depth = pt->depth[0]; - for (level = 0; level <= pt->last_level; level++) { - unsigned i; - for (i = 0; i < depth; i++) - i915_miptree_set_image_offset(tex, level, i, - 0, i * stack_nblocksy); - - depth = minify(depth); - } - - - /* Multiply slice size by texture depth for total size. It's - * remarkable how wasteful of memory the i915 texture layouts - * are. They are largely fixed in the i945. - */ - tex->total_nblocksy = stack_nblocksy * pt->depth[0]; - break; - } - - default:{ - unsigned width = pt->width[0]; - unsigned height = pt->height[0]; - unsigned nblocksx = pt->nblocksx[0]; - unsigned nblocksy = pt->nblocksy[0]; - - tex->stride = round_up(pt->nblocksx[0] * pt->block.size, 4); - tex->total_nblocksy = 0; - - for (level = 0; level <= pt->last_level; level++) { - i915_miptree_set_level_info(tex, level, 1, - width, height, 1); - i915_miptree_set_image_offset(tex, level, 0, - 0, tex->total_nblocksy); - - nblocksy = round_up(MAX2(2, nblocksy), 2); - - tex->total_nblocksy += nblocksy; - - width = minify(width); - height = minify(height); - nblocksx = pf_get_nblocksx(&pt->block, width); - nblocksy = pf_get_nblocksy(&pt->block, height); - } - break; - } - } - /* - DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, - tex->pitch, - tex->total_nblocksy, pt->block.size, tex->stride * tex->total_nblocksy); - */ - - return TRUE; -} - - -static boolean i945_miptree_layout(struct i915_texture * tex) { struct pipe_texture *pt = &tex->base; - unsigned level; switch (pt->target) { + case PIPE_TEXTURE_1D: + case PIPE_TEXTURE_2D: + i945_miptree_layout_2d(tex); + break; + case PIPE_TEXTURE_3D: + i945_miptree_layout_3d(tex); + break; case PIPE_TEXTURE_CUBE: i945_miptree_layout_cube(tex); break; - case PIPE_TEXTURE_3D:{ - unsigned width = pt->width[0]; - unsigned height = pt->height[0]; - unsigned depth = pt->depth[0]; - unsigned nblocksx = pt->nblocksx[0]; - unsigned nblocksy = pt->nblocksy[0]; - unsigned pack_x_pitch, pack_x_nr; - unsigned pack_y_pitch; - - tex->stride = round_up(pt->nblocksx[0] * pt->block.size, 4); - tex->total_nblocksy = 0; - - pack_y_pitch = MAX2(pt->nblocksy[0], 2); - pack_x_pitch = tex->stride / pt->block.size; - pack_x_nr = 1; - - for (level = 0; level <= pt->last_level; level++) { - unsigned nr_images = pt->target == PIPE_TEXTURE_3D ? depth : 6; - int x = 0; - int y = 0; - unsigned q, j; - - i915_miptree_set_level_info(tex, level, nr_images, - width, height, depth); - - for (q = 0; q < nr_images;) { - for (j = 0; j < pack_x_nr && q < nr_images; j++, q++) { - i915_miptree_set_image_offset(tex, level, q, x, y + tex->total_nblocksy); - x += pack_x_pitch; - } - - x = 0; - y += pack_y_pitch; - } - - - tex->total_nblocksy += y; - - if (pack_x_pitch > 4) { - pack_x_pitch >>= 1; - pack_x_nr <<= 1; - assert(pack_x_pitch * pack_x_nr * pt->block.size <= tex->stride); - } - - if (pack_y_pitch > 2) { - pack_y_pitch >>= 1; - } - - width = minify(width); - height = minify(height); - depth = minify(depth); - nblocksx = pf_get_nblocksx(&pt->block, width); - nblocksy = pf_get_nblocksy(&pt->block, height); - } - break; - } - - case PIPE_TEXTURE_1D: - case PIPE_TEXTURE_2D: -// case PIPE_TEXTURE_RECTANGLE: - i945_miptree_layout_2d(tex); - break; default: assert(0); return FALSE; } - /* - DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, - tex->pitch, - tex->total_nblocksy, pt->block.size, tex->stride * tex->total_nblocksy); - */ - return TRUE; } +/* + * Screen texture functions + */ + + static struct pipe_texture * i915_texture_create(struct pipe_screen *screen, const struct pipe_texture *templat) { - struct i915_screen *i915screen = i915_screen(screen); + struct i915_screen *is = i915_screen(screen); + struct intel_winsys *iws = is->iws; struct i915_texture *tex = CALLOC_STRUCT(i915_texture); size_t tex_size; unsigned buf_usage = 0; @@ -594,27 +607,35 @@ i915_texture_create(struct pipe_screen *screen, tex->base.nblocksx[0] = pf_get_nblocksx(&tex->base.block, tex->base.width[0]); tex->base.nblocksy[0] = pf_get_nblocksy(&tex->base.block, tex->base.height[0]); - if (i915screen->is_i945) { + if (is->is_i945) { if (!i945_miptree_layout(tex)) - goto fail; + goto fail; } else { if (!i915_miptree_layout(tex)) - goto fail; + goto fail; } tex_size = tex->stride * tex->total_nblocksy; - buf_usage = PIPE_BUFFER_USAGE_PIXEL; - /* for scanouts and cursors, cursors don't have the scanout tag */ - if (templat->tex_usage & PIPE_TEXTURE_USAGE_PRIMARY && templat->width[0] != 64) - buf_usage |= I915_BUFFER_USAGE_SCANOUT; - tex->buffer = screen->buffer_create(screen, 64, buf_usage, tex_size); + /* for scanouts and cursors, cursors arn't scanouts */ + if (templat->tex_usage & PIPE_TEXTURE_USAGE_PRIMARY && templat->width[0] != 64) + buf_usage = INTEL_NEW_SCANOUT; + else + buf_usage = INTEL_NEW_TEXTURE; + tex->buffer = iws->buffer_create(iws, tex_size, 64, buf_usage); if (!tex->buffer) goto fail; + /* setup any hw fences */ + if (tex->hw_tiled) { + assert(tex->sw_tiled == INTEL_TILE_NONE); + iws->buffer_set_fence_reg(iws, tex->buffer, tex->stride, tex->hw_tiled); + } + + #if 0 void *ptr = ws->buffer_map(ws, tex->buffer, PIPE_BUFFER_USAGE_CPU_WRITE); @@ -629,18 +650,56 @@ fail: return NULL; } +static struct pipe_texture * +i915_texture_blanket(struct pipe_screen * screen, + const struct pipe_texture *base, + const unsigned *stride, + struct pipe_buffer *buffer) +{ +#if 0 + struct i915_texture *tex; + assert(screen); + + /* Only supports one type */ + if (base->target != PIPE_TEXTURE_2D || + base->last_level != 0 || + base->depth[0] != 1) { + return NULL; + } + + tex = CALLOC_STRUCT(i915_texture); + if (!tex) + return NULL; + + tex->base = *base; + pipe_reference_init(&tex->base.reference, 1); + tex->base.screen = screen; + + tex->stride = stride[0]; + + i915_miptree_set_level_info(tex, 0, 1, base->width[0], base->height[0], 1); + i915_miptree_set_image_offset(tex, 0, 0, 0, 0); + + pipe_buffer_reference(&tex->buffer, buffer); + + return &tex->base; +#else + return NULL; +#endif +} static void i915_texture_destroy(struct pipe_texture *pt) { struct i915_texture *tex = (struct i915_texture *)pt; + struct intel_winsys *iws = i915_screen(pt->screen)->iws; uint i; /* DBG("%s deleting %p\n", __FUNCTION__, (void *) tex); */ - pipe_buffer_reference(&tex->buffer, NULL); + iws->buffer_destroy(iws, tex->buffer); for (i = 0; i < PIPE_MAX_TEXTURE_LEVELS; i++) if (tex->image_offset[i]) @@ -649,6 +708,12 @@ i915_texture_destroy(struct pipe_texture *pt) FREE(tex); } + +/* + * Screen surface functions + */ + + static struct pipe_surface * i915_get_tex_surface(struct pipe_screen *screen, struct pipe_texture *pt, @@ -684,11 +749,122 @@ i915_get_tex_surface(struct pipe_screen *screen, return ps; } -static struct pipe_texture * -i915_texture_blanket(struct pipe_screen * screen, - const struct pipe_texture *base, - const unsigned *stride, - struct pipe_buffer *buffer) +static void +i915_tex_surface_destroy(struct pipe_surface *surf) +{ + pipe_texture_reference(&surf->texture, NULL); + FREE(surf); +} + + +/* + * Screen transfer functions + */ + + +static struct pipe_transfer* +i915_get_tex_transfer(struct pipe_screen *screen, + struct pipe_texture *texture, + unsigned face, unsigned level, unsigned zslice, + enum pipe_transfer_usage usage, unsigned x, unsigned y, + unsigned w, unsigned h) +{ + struct i915_texture *tex = (struct i915_texture *)texture; + struct i915_transfer *trans; + unsigned offset; /* in bytes */ + + if (texture->target == PIPE_TEXTURE_CUBE) { + offset = tex->image_offset[level][face]; + } + else if (texture->target == PIPE_TEXTURE_3D) { + offset = tex->image_offset[level][zslice]; + } + else { + offset = tex->image_offset[level][0]; + assert(face == 0); + assert(zslice == 0); + } + + trans = CALLOC_STRUCT(i915_transfer); + if (trans) { + pipe_texture_reference(&trans->base.texture, texture); + trans->base.format = trans->base.format; + trans->base.x = x; + trans->base.y = y; + trans->base.width = w; + trans->base.height = h; + trans->base.block = texture->block; + trans->base.nblocksx = texture->nblocksx[level]; + trans->base.nblocksy = texture->nblocksy[level]; + trans->base.stride = tex->stride; + trans->offset = offset; + trans->base.usage = usage; + } + return &trans->base; +} + +static void * +i915_transfer_map(struct pipe_screen *screen, + struct pipe_transfer *transfer) +{ + struct i915_texture *tex = (struct i915_texture *)transfer->texture; + struct intel_winsys *iws = i915_screen(tex->base.screen)->iws; + char *map; + boolean write = FALSE; + + if (transfer->usage != PIPE_TRANSFER_READ) + write = TRUE; + + map = iws->buffer_map(iws, tex->buffer, write); + if (map == NULL) + return NULL; + + return map + i915_transfer(transfer)->offset + + transfer->y / transfer->block.height * transfer->stride + + transfer->x / transfer->block.width * transfer->block.size; +} + +static void +i915_transfer_unmap(struct pipe_screen *screen, + struct pipe_transfer *transfer) +{ + struct i915_texture *tex = (struct i915_texture *)transfer->texture; + struct intel_winsys *iws = i915_screen(tex->base.screen)->iws; + iws->buffer_unmap(iws, tex->buffer); +} + +static void +i915_tex_transfer_destroy(struct pipe_transfer *trans) +{ + pipe_texture_reference(&trans->texture, NULL); + FREE(trans); +} + + +/* + * Other texture functions + */ + + +void +i915_init_screen_texture_functions(struct i915_screen *is) +{ + is->base.texture_create = i915_texture_create; + is->base.texture_blanket = i915_texture_blanket; + is->base.texture_destroy = i915_texture_destroy; + is->base.get_tex_surface = i915_get_tex_surface; + is->base.tex_surface_destroy = i915_tex_surface_destroy; + is->base.get_tex_transfer = i915_get_tex_transfer; + is->base.transfer_map = i915_transfer_map; + is->base.transfer_unmap = i915_transfer_unmap; + is->base.tex_transfer_destroy = i915_tex_transfer_destroy; +} + +struct pipe_texture * +i915_texture_blanket_intel(struct pipe_screen *screen, + struct pipe_texture *base, + unsigned stride, + struct intel_buffer *buffer) { struct i915_texture *tex; assert(screen); @@ -708,52 +884,28 @@ i915_texture_blanket(struct pipe_screen * screen, pipe_reference_init(&tex->base.reference, 1); tex->base.screen = screen; - tex->stride = stride[0]; + tex->stride = stride; i915_miptree_set_level_info(tex, 0, 1, base->width[0], base->height[0], 1); i915_miptree_set_image_offset(tex, 0, 0, 0, 0); - pipe_buffer_reference(&tex->buffer, buffer); + tex->buffer = buffer; return &tex->base; } -void -i915_init_texture_functions(struct i915_context *i915) -{ -// i915->pipe.texture_update = i915_texture_update; -} - -static void -i915_tex_surface_destroy(struct pipe_surface *surf) -{ - pipe_texture_reference(&surf->texture, NULL); - FREE(surf); -} - -void -i915_init_screen_texture_functions(struct pipe_screen *screen) -{ - screen->texture_create = i915_texture_create; - screen->texture_destroy = i915_texture_destroy; - screen->get_tex_surface = i915_get_tex_surface; - screen->texture_blanket = i915_texture_blanket; - screen->tex_surface_destroy = i915_tex_surface_destroy; -} - -boolean i915_get_texture_buffer( struct pipe_texture *texture, - struct pipe_buffer **buf, - unsigned *stride ) +boolean +i915_get_texture_buffer_intel(struct pipe_texture *texture, + struct intel_buffer **buffer, + unsigned *stride) { struct i915_texture *tex = (struct i915_texture *)texture; - if (!tex) + if (!texture) return FALSE; - pipe_buffer_reference(buf, tex->buffer); - - if (stride) - *stride = tex->stride; + *stride = tex->stride; + *buffer = tex->buffer; return TRUE; } diff --git a/src/gallium/drivers/i915simple/i915_texture.h b/src/gallium/drivers/i915simple/i915_texture.h index 7225016a9f4..51a1dd984c8 100644 --- a/src/gallium/drivers/i915simple/i915_texture.h +++ b/src/gallium/drivers/i915simple/i915_texture.h @@ -28,16 +28,9 @@ #ifndef I915_TEXTURE_H #define I915_TEXTURE_H -struct i915_context; -struct pipe_screen; - +struct i915_screen; extern void -i915_init_texture_functions(struct i915_context *i915); - - -extern void -i915_init_screen_texture_functions(struct pipe_screen *screen); - +i915_init_screen_texture_functions(struct i915_screen *is); #endif /* I915_TEXTURE_H */ diff --git a/src/gallium/drivers/i915simple/i915_winsys.h b/src/gallium/drivers/i915simple/i915_winsys.h deleted file mode 100644 index 711db91c367..00000000000 --- a/src/gallium/drivers/i915simple/i915_winsys.h +++ /dev/null @@ -1,144 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \file - * This is the interface that i915simple requires any window system - * hosting it to implement. This is the only include file in i915simple - * which is public. - * - * This isn't currently true as the winsys needs i915_batchbuffer.h - */ - -#ifndef I915_WINSYS_H -#define I915_WINSYS_H - - -#include "pipe/p_defines.h" - - -#ifdef __cplusplus -extern "C" { -#endif - - -/* Pipe drivers are independent of both GL and the window system. - * The window system provides a buffer manager and a set of additional - * hooks for things like command buffer submission, etc. - * - * There clearly has to be some agreement between the window system - * driver and the hardware driver about the format of command buffers, - * etc. - */ - -struct i915_batchbuffer; -struct pipe_texture; -struct pipe_buffer; -struct pipe_fence_handle; -struct pipe_winsys; -struct pipe_screen; - - -/** - * Additional winsys interface for i915simple. - * - * It is an over-simple batchbuffer mechanism. Will want to improve the - * performance of this, perhaps based on the cmdstream stuff. It - * would be pretty impossible to implement swz on top of this - * interface. - * - * Will also need additions/changes to implement static/dynamic - * indirect state. - */ -struct i915_winsys { - - void (*destroy)( struct i915_winsys *sws ); - - /** - * Get the current batch buffer from the winsys. - */ - struct i915_batchbuffer *(*batch_get)( struct i915_winsys *sws ); - - /** - * Emit a relocation to a buffer. - * - * Used not only when the buffer addresses are not pinned, but also to - * ensure refered buffers will not be destroyed until the current batch - * buffer execution is finished. - * - * The access flags is a combination of I915_BUFFER_ACCESS_WRITE and - * I915_BUFFER_ACCESS_READ macros. - */ - void (*batch_reloc)( struct i915_winsys *sws, - struct pipe_buffer *buf, - unsigned access_flags, - unsigned delta ); - - /** - * Flush the batch. - */ - void (*batch_flush)( struct i915_winsys *sws, - struct pipe_fence_handle **fence ); -}; - -#define I915_BUFFER_ACCESS_WRITE 0x1 -#define I915_BUFFER_ACCESS_READ 0x2 - -#define I915_BUFFER_USAGE_LIT_VERTEX (PIPE_BUFFER_USAGE_CUSTOM << 0) -#define I915_BUFFER_USAGE_SCANOUT (PIPE_BUFFER_USAGE_CUSTOM << 1) - - -/** - * Create i915 pipe_screen. - */ -struct pipe_screen *i915_create_screen( struct pipe_winsys *winsys, - uint pci_id ); - -/** - * Create a i915 pipe_context. - */ -struct pipe_context *i915_create_context( struct pipe_screen *screen, - struct pipe_winsys *winsys, - struct i915_winsys *i915 ); - -/** - * Used for the winsys to get the buffer used for a texture - * and also the stride used for the texture. - * - * Buffer is referenced for you so you need to unref after use. - * - * This is needed for example kms. - */ -boolean i915_get_texture_buffer( struct pipe_texture *texture, - struct pipe_buffer **buf, - unsigned *stride ); - -#ifdef __cplusplus -} -#endif - -#endif diff --git a/src/gallium/drivers/i915simple/intel_batchbuffer.h b/src/gallium/drivers/i915simple/intel_batchbuffer.h new file mode 100644 index 00000000000..db12dfd2ac2 --- /dev/null +++ b/src/gallium/drivers/i915simple/intel_batchbuffer.h @@ -0,0 +1,87 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_BATCH_H +#define INTEL_BATCH_H + +#include "intel_winsys.h" + +static INLINE boolean +intel_batchbuffer_check(struct intel_batchbuffer *batch, + size_t dwords, + size_t relocs) +{ + return dwords * 4 <= batch->size - (batch->ptr - batch->map) && + relocs <= (batch->max_relocs - batch->relocs); +} + +static INLINE size_t +intel_batchbuffer_space(struct intel_batchbuffer *batch) +{ + return batch->size - (batch->ptr - batch->map); +} + +static INLINE void +intel_batchbuffer_dword(struct intel_batchbuffer *batch, + unsigned dword) +{ + if (intel_batchbuffer_space(batch) < 4) + return; + + *(unsigned *)batch->ptr = dword; + batch->ptr += 4; +} + +static INLINE void +intel_batchbuffer_write(struct intel_batchbuffer *batch, + void *data, + size_t size) +{ + if (intel_batchbuffer_space(batch) < size) + return; + + memcpy(data, batch->ptr, size); + batch->ptr += size; +} + +static INLINE int +intel_batchbuffer_reloc(struct intel_batchbuffer *batch, + struct intel_buffer *buffer, + enum intel_buffer_usage usage, + size_t offset) +{ + return batch->iws->batchbuffer_reloc(batch, buffer, usage, offset); +} + +static INLINE void +intel_batchbuffer_flush(struct intel_batchbuffer *batch, + struct pipe_fence_handle **fence) +{ + batch->iws->batchbuffer_flush(batch, fence); +} + +#endif diff --git a/src/gallium/drivers/i915simple/intel_winsys.h b/src/gallium/drivers/i915simple/intel_winsys.h new file mode 100644 index 00000000000..f949f52a9ce --- /dev/null +++ b/src/gallium/drivers/i915simple/intel_winsys.h @@ -0,0 +1,219 @@ +/************************************************************************** + * + * Copyright © 2009 Jakob Bornecrantz + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_WINSYS_H +#define INTEL_WINSYS_H + +#include "pipe/p_compiler.h" + +struct intel_winsys; +struct intel_buffer; +struct intel_batchbuffer; +struct pipe_texture; +struct pipe_fence_handle; + +enum intel_buffer_usage +{ + /* use on textures */ + INTEL_USAGE_RENDER = 0x01, + INTEL_USAGE_SAMPLER = 0x02, + INTEL_USAGE_2D_TARGET = 0x04, + INTEL_USAGE_2D_SOURCE = 0x08, + /* use on vertex */ + INTEL_USAGE_VERTEX = 0x10, +}; + +enum intel_buffer_type +{ + INTEL_NEW_TEXTURE, + INTEL_NEW_SCANOUT, /**< a texture used for scanning out from */ + INTEL_NEW_VERTEX, +}; + +enum intel_buffer_tile +{ + INTEL_TILE_NONE, + INTEL_TILE_X, + INTEL_TILE_Y, +}; + +struct intel_batchbuffer { + + struct intel_winsys *iws; + + /** + * Values exported to speed up the writing the batchbuffer, + * instead of having to go trough a accesor function for + * each dword written. + */ + /*{@*/ + uint8_t *map; + uint8_t *ptr; + size_t size; + + size_t relocs; + size_t max_relocs; + /*@}*/ +}; + +struct intel_winsys { + + /** + * Batchbuffer functions. + */ + /*@{*/ + /** + * Create a new batchbuffer. + */ + struct intel_batchbuffer *(*batchbuffer_create)(struct intel_winsys *iws); + + /** + * Emit a relocation to a buffer. + * Target position in batchbuffer is the same as ptr. + * + * @batch + * @reloc buffer address to be inserted into target. + * @usage how is the hardware going to use the buffer. + * @offset add this to the reloc buffers address + * @target buffer where to write the address, null for batchbuffer. + */ + int (*batchbuffer_reloc)(struct intel_batchbuffer *batch, + struct intel_buffer *reloc, + enum intel_buffer_usage usage, + unsigned offset); + + /** + * Flush a bufferbatch. + */ + void (*batchbuffer_flush)(struct intel_batchbuffer *batch, + struct pipe_fence_handle **fence); + + /** + * Destroy a batchbuffer. + */ + void (*batchbuffer_destroy)(struct intel_batchbuffer *batch); + /*@}*/ + + + /** + * Buffer functions. + */ + /*@{*/ + /** + * Create a buffer. + */ + struct intel_buffer *(*buffer_create)(struct intel_winsys *iws, + unsigned size, unsigned alignment, + enum intel_buffer_type type); + + /** + * Fence a buffer with a fence reg. + * Not to be confused with pipe_fence_handle. + */ + int (*buffer_set_fence_reg)(struct intel_winsys *iws, + struct intel_buffer *buffer, + unsigned stride, + enum intel_buffer_tile tile); + + /** + * Map a buffer. + */ + void *(*buffer_map)(struct intel_winsys *iws, + struct intel_buffer *buffer, + boolean write); + + /** + * Unmap a buffer. + */ + void (*buffer_unmap)(struct intel_winsys *iws, + struct intel_buffer *buffer); + + void (*buffer_destroy)(struct intel_winsys *iws, + struct intel_buffer *buffer); + /*@}*/ + + + /** + * Fence functions. + */ + /*@{*/ + /** + * Reference fence and set ptr to fence. + */ + void (*fence_reference)(struct intel_winsys *iws, + struct pipe_fence_handle **ptr, + struct pipe_fence_handle *fence); + + /** + * Check if a fence has finished. + */ + int (*fence_signalled)(struct intel_winsys *iws, + struct pipe_fence_handle *fence); + + /** + * Wait on a fence to finish. + */ + int (*fence_finish)(struct intel_winsys *iws, + struct pipe_fence_handle *fence); + /*@}*/ + + + /** + * Destroy the winsys. + */ + void (*destroy)(struct intel_winsys *iws); +}; + + +/** + * Create i915 pipe_screen. + */ +struct pipe_screen *i915_create_screen(struct intel_winsys *iws, unsigned pci_id); + +/** + * Create a i915 pipe_context. + */ +struct pipe_context *i915_create_context(struct pipe_screen *screen); + +/** + * Get the intel_winsys buffer backing the texture. + * + * TODO UGLY + */ +boolean i915_get_texture_buffer_intel(struct pipe_texture *texture, + struct intel_buffer **buffer, + unsigned *stride); + +/** + * Wrap a intel_winsys buffer with a texture blanket. + * + * TODO UGLY + */ +struct pipe_texture * i915_texture_blanket_intel(struct pipe_screen *screen, + struct pipe_texture *tmplt, + unsigned pitch, + struct intel_buffer *buffer); + +#endif diff --git a/src/gallium/drivers/llvmpipe/SConscript b/src/gallium/drivers/llvmpipe/SConscript index 6bceb84da47..5c29bdac56e 100644 --- a/src/gallium/drivers/llvmpipe/SConscript +++ b/src/gallium/drivers/llvmpipe/SConscript @@ -2,8 +2,12 @@ Import('*') env = env.Clone() +env.Tool('llvm') +if 'LLVM_VERSION' not in env: + print 'warning: LLVM not found: not building llvmpipe' + Return() + env.Tool('udis86') -env.ParseConfig('llvm-config --cppflags') llvmpipe = env.ConvenienceLibrary( target = 'llvmpipe', @@ -57,8 +61,6 @@ llvmpipe = env.ConvenienceLibrary( env = env.Clone() -env['LINK'] = env['CXX'] -env.ParseConfig('llvm-config --libs jit interpreter nativecodegen bitwriter') env.Prepend(LIBS = [llvmpipe] + auxiliaries) env.Program( diff --git a/src/gallium/drivers/llvmpipe/lp_bld_arit.c b/src/gallium/drivers/llvmpipe/lp_bld_arit.c index 710e6246d74..09a57ff33d5 100644 --- a/src/gallium/drivers/llvmpipe/lp_bld_arit.c +++ b/src/gallium/drivers/llvmpipe/lp_bld_arit.c @@ -52,6 +52,7 @@ #include "lp_bld_type.h" #include "lp_bld_const.h" #include "lp_bld_intr.h" +#include "lp_bld_logic.h" #include "lp_bld_arit.h" @@ -98,11 +99,8 @@ lp_build_min_simple(struct lp_build_context *bld, if(intrinsic) return lp_build_intrinsic_binary(bld->builder, intrinsic, lp_build_vec_type(bld->type), a, b); - if(type.floating) - cond = LLVMBuildFCmp(bld->builder, LLVMRealULT, a, b, ""); - else - cond = LLVMBuildICmp(bld->builder, type.sign ? LLVMIntSLT : LLVMIntULT, a, b, ""); - return LLVMBuildSelect(bld->builder, cond, a, b, ""); + cond = lp_build_cmp(bld, PIPE_FUNC_LESS, a, b); + return lp_build_select(bld, cond, a, b); } @@ -149,11 +147,8 @@ lp_build_max_simple(struct lp_build_context *bld, if(intrinsic) return lp_build_intrinsic_binary(bld->builder, intrinsic, lp_build_vec_type(bld->type), a, b); - if(type.floating) - cond = LLVMBuildFCmp(bld->builder, LLVMRealULT, a, b, ""); - else - cond = LLVMBuildICmp(bld->builder, type.sign ? LLVMIntSLT : LLVMIntULT, a, b, ""); - return LLVMBuildSelect(bld->builder, cond, b, a, ""); + cond = lp_build_cmp(bld, PIPE_FUNC_GREATER, a, b); + return lp_build_select(bld, cond, a, b); } diff --git a/src/gallium/drivers/llvmpipe/lp_bld_logic.c b/src/gallium/drivers/llvmpipe/lp_bld_logic.c index 1e1ecf805bc..8631efd6c3e 100644 --- a/src/gallium/drivers/llvmpipe/lp_bld_logic.c +++ b/src/gallium/drivers/llvmpipe/lp_bld_logic.c @@ -33,7 +33,6 @@ */ -#include "pipe/p_defines.h" #include "lp_bld_type.h" #include "lp_bld_const.h" #include "lp_bld_intr.h" @@ -52,6 +51,8 @@ lp_build_cmp(struct lp_build_context *bld, LLVMValueRef zeros = LLVMConstNull(int_vec_type); LLVMValueRef ones = LLVMConstAllOnes(int_vec_type); LLVMValueRef cond; + LLVMValueRef res; + unsigned i; if(func == PIPE_FUNC_NEVER) return zeros; @@ -68,7 +69,6 @@ lp_build_cmp(struct lp_build_context *bld, LLVMValueRef args[3]; unsigned cc; boolean swap; - LLVMValueRef res; swap = FALSE; switch(func) { @@ -219,7 +219,28 @@ lp_build_cmp(struct lp_build_context *bld, assert(0); return bld->undef; } + +#if 0 + /* XXX: Although valid IR, no LLVM target currently support this */ cond = LLVMBuildFCmp(bld->builder, op, a, b, ""); + res = LLVMBuildSelect(bld->builder, cond, ones, zeros, ""); +#else + debug_printf("%s: warning: using slow element-wise vector comparison\n", + __FUNCTION__); + res = LLVMGetUndef(int_vec_type); + for(i = 0; i < type.length; ++i) { + LLVMValueRef index = LLVMConstInt(LLVMInt32Type(), i, 0); + cond = LLVMBuildFCmp(bld->builder, op, + LLVMBuildExtractElement(bld->builder, a, index, ""), + LLVMBuildExtractElement(bld->builder, b, index, ""), + ""); + cond = LLVMBuildSelect(bld->builder, cond, + LLVMConstExtractElement(ones, index), + LLVMConstExtractElement(zeros, index), + ""); + res = LLVMBuildInsertElement(bld->builder, res, cond, index, ""); + } +#endif } else { LLVMIntPredicate op; @@ -246,10 +267,31 @@ lp_build_cmp(struct lp_build_context *bld, assert(0); return bld->undef; } + +#if 0 + /* XXX: Although valid IR, no LLVM target currently support this */ cond = LLVMBuildICmp(bld->builder, op, a, b, ""); + res = LLVMBuildSelect(bld->builder, cond, ones, zeros, ""); +#else + debug_printf("%s: warning: using slow element-wise vector comparison\n", + __FUNCTION__); + res = LLVMGetUndef(int_vec_type); + for(i = 0; i < type.length; ++i) { + LLVMValueRef index = LLVMConstInt(LLVMInt32Type(), i, 0); + cond = LLVMBuildICmp(bld->builder, op, + LLVMBuildExtractElement(bld->builder, a, index, ""), + LLVMBuildExtractElement(bld->builder, b, index, ""), + ""); + cond = LLVMBuildSelect(bld->builder, cond, + LLVMConstExtractElement(ones, index), + LLVMConstExtractElement(zeros, index), + ""); + res = LLVMBuildInsertElement(bld->builder, res, cond, index, ""); + } +#endif } - return LLVMBuildSelect(bld->builder, cond, ones, zeros, ""); + return res; } diff --git a/src/gallium/drivers/llvmpipe/lp_bld_logic.h b/src/gallium/drivers/llvmpipe/lp_bld_logic.h index 0989f9f2078..29b9e1c45b8 100644 --- a/src/gallium/drivers/llvmpipe/lp_bld_logic.h +++ b/src/gallium/drivers/llvmpipe/lp_bld_logic.h @@ -39,6 +39,8 @@ #include <llvm-c/Core.h> +#include "pipe/p_defines.h" /* For PIPE_FUNC_xxx */ + union lp_type type; struct lp_build_context; diff --git a/src/gallium/drivers/llvmpipe/lp_bld_tgsi_soa.c b/src/gallium/drivers/llvmpipe/lp_bld_tgsi_soa.c index bce26607f90..d4d18febec7 100644 --- a/src/gallium/drivers/llvmpipe/lp_bld_tgsi_soa.c +++ b/src/gallium/drivers/llvmpipe/lp_bld_tgsi_soa.c @@ -687,10 +687,6 @@ emit_instruction( return 0; break; - case TGSI_OPCODE_CND0: - return 0; - break; - case TGSI_OPCODE_DP2A: tmp0 = emit_fetch( bld, inst, 0, CHAN_X ); /* xmm0 = src[0].x */ tmp1 = emit_fetch( bld, inst, 1, CHAN_X ); /* xmm1 = src[1].x */ diff --git a/src/gallium/drivers/llvmpipe/lp_buffer.c b/src/gallium/drivers/llvmpipe/lp_buffer.c new file mode 100644 index 00000000000..66f1f8e1383 --- /dev/null +++ b/src/gallium/drivers/llvmpipe/lp_buffer.c @@ -0,0 +1,150 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "util/u_memory.h" +#include "util/u_math.h" + +#include "lp_winsys.h" +#include "lp_screen.h" +#include "lp_texture.h" +#include "lp_buffer.h" + + +static void * +llvmpipe_buffer_map(struct pipe_screen *screen, + struct pipe_buffer *buf, + unsigned flags) +{ + struct llvmpipe_buffer *llvmpipe_buf = llvmpipe_buffer(buf); + return llvmpipe_buf->data; +} + + +static void +llvmpipe_buffer_unmap(struct pipe_screen *screen, + struct pipe_buffer *buf) +{ +} + + +static void +llvmpipe_buffer_destroy(struct pipe_buffer *buf) +{ + struct llvmpipe_buffer *sbuf = llvmpipe_buffer(buf); + + if (!sbuf->userBuffer) + align_free(sbuf->data); + + FREE(sbuf); +} + + +static struct pipe_buffer * +llvmpipe_buffer_create(struct pipe_screen *screen, + unsigned alignment, + unsigned usage, + unsigned size) +{ + struct llvmpipe_buffer *buffer = CALLOC_STRUCT(llvmpipe_buffer); + + pipe_reference_init(&buffer->base.reference, 1); + buffer->base.screen = screen; + buffer->base.alignment = MAX2(alignment, 16); + buffer->base.usage = usage; + buffer->base.size = size; + + buffer->data = align_malloc(size, alignment); + + return &buffer->base; +} + + +/** + * Create buffer which wraps user-space data. + */ +static struct pipe_buffer * +llvmpipe_user_buffer_create(struct pipe_screen *screen, + void *ptr, + unsigned bytes) +{ + struct llvmpipe_buffer *buffer; + + buffer = CALLOC_STRUCT(llvmpipe_buffer); + if(!buffer) + return NULL; + + pipe_reference_init(&buffer->base.reference, 1); + buffer->base.screen = screen; + buffer->base.size = bytes; + buffer->userBuffer = TRUE; + buffer->data = ptr; + + return &buffer->base; +} + + +static void +llvmpipe_fence_reference(struct pipe_screen *screen, + struct pipe_fence_handle **ptr, + struct pipe_fence_handle *fence) +{ +} + + +static int +llvmpipe_fence_signalled(struct pipe_screen *screen, + struct pipe_fence_handle *fence, + unsigned flag) +{ + return 0; +} + + +static int +llvmpipe_fence_finish(struct pipe_screen *screen, + struct pipe_fence_handle *fence, + unsigned flag) +{ + return 0; +} + + +void +llvmpipe_init_screen_buffer_funcs(struct pipe_screen *screen) +{ + screen->buffer_create = llvmpipe_buffer_create; + screen->user_buffer_create = llvmpipe_user_buffer_create; + screen->buffer_map = llvmpipe_buffer_map; + screen->buffer_unmap = llvmpipe_buffer_unmap; + screen->buffer_destroy = llvmpipe_buffer_destroy; + + screen->fence_reference = llvmpipe_fence_reference; + screen->fence_signalled = llvmpipe_fence_signalled; + screen->fence_finish = llvmpipe_fence_finish; + +} diff --git a/src/gallium/drivers/llvmpipe/lp_buffer.h b/src/gallium/drivers/llvmpipe/lp_buffer.h new file mode 100644 index 00000000000..d6b8184a0b9 --- /dev/null +++ b/src/gallium/drivers/llvmpipe/lp_buffer.h @@ -0,0 +1,55 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef LP_BUFFER_H +#define LP_BUFFER_H + +#include "pipe/p_compiler.h" +#include "pipe/p_state.h" + + +struct llvmpipe_buffer +{ + struct pipe_buffer base; + boolean userBuffer; /** Is this a user-space buffer? */ + void *data; +}; + + +/** Cast wrapper */ +static INLINE struct llvmpipe_buffer * +llvmpipe_buffer( struct pipe_buffer *buf ) +{ + return (struct llvmpipe_buffer *)buf; +} + + +void +llvmpipe_init_screen_buffer_funcs(struct pipe_screen *screen); + + +#endif /* LP_BUFFER_H */ diff --git a/src/gallium/drivers/nouveau/nouveau_screen.h b/src/gallium/drivers/nouveau/nouveau_screen.h index 9968b078966..ebfc67ad1c1 100644 --- a/src/gallium/drivers/nouveau/nouveau_screen.h +++ b/src/gallium/drivers/nouveau/nouveau_screen.h @@ -22,4 +22,15 @@ nouveau_bo(struct pipe_buffer *pb) int nouveau_screen_init(struct nouveau_screen *, struct nouveau_device *); void nouveau_screen_fini(struct nouveau_screen *); +struct nouveau_miptree { + struct pipe_texture base; + struct nouveau_bo *bo; +}; + +static inline struct nouveau_miptree * +nouveau_miptree(struct pipe_texture *pt) +{ + return (struct nouveau_miptree *)pt; +} + #endif diff --git a/src/gallium/drivers/nv50/nv50_context.h b/src/gallium/drivers/nv50/nv50_context.h index 4de6e8cfa29..1e9e8e49bfb 100644 --- a/src/gallium/drivers/nv50/nv50_context.h +++ b/src/gallium/drivers/nv50/nv50_context.h @@ -75,9 +75,7 @@ struct nv50_miptree_level { }; struct nv50_miptree { - struct pipe_texture base; - - struct nouveau_bo *bo; + struct nouveau_miptree base; struct nv50_miptree_level level[PIPE_MAX_TEXTURE_LEVELS]; int image_nr; diff --git a/src/gallium/drivers/nv50/nv50_miptree.c b/src/gallium/drivers/nv50/nv50_miptree.c index dd1b0303bd5..03b9243b828 100644 --- a/src/gallium/drivers/nv50/nv50_miptree.c +++ b/src/gallium/drivers/nv50/nv50_miptree.c @@ -31,15 +31,15 @@ nv50_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *tmp) { struct nouveau_device *dev = nouveau_screen(pscreen)->device; struct nv50_miptree *mt = CALLOC_STRUCT(nv50_miptree); - struct pipe_texture *pt = &mt->base; + struct pipe_texture *pt = &mt->base.base; unsigned width = tmp->width[0], height = tmp->height[0]; unsigned depth = tmp->depth[0]; uint32_t tile_mode, tile_flags, tile_h; int ret, i, l; - mt->base = *tmp; - pipe_reference_init(&mt->base.reference, 1); - mt->base.screen = pscreen; + *pt = *tmp; + pipe_reference_init(&pt->reference, 1); + pt->screen = pscreen; switch (pt->format) { case PIPE_FORMAT_Z32_FLOAT: @@ -116,13 +116,14 @@ nv50_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *tmp) } ret = nouveau_bo_new_tile(dev, NOUVEAU_BO_VRAM, 256, mt->total_size, - mt->level[0].tile_mode, tile_flags, &mt->bo); + mt->level[0].tile_mode, tile_flags, + &mt->base.bo); if (ret) { FREE(mt); return NULL; } - return &mt->base; + return pt; } static struct pipe_texture * @@ -141,15 +142,15 @@ nv50_miptree_blanket(struct pipe_screen *pscreen, const struct pipe_texture *pt, if (!mt) return NULL; - mt->base = *pt; - pipe_reference_init(&mt->base.reference, 1); - mt->base.screen = pscreen; + mt->base.base = *pt; + pipe_reference_init(&mt->base.base.reference, 1); + mt->base.base.screen = pscreen; mt->image_nr = 1; mt->level[0].pitch = *stride; mt->level[0].image_offset = CALLOC(1, sizeof(unsigned)); - nouveau_bo_ref(bo, &mt->bo); - return &mt->base; + nouveau_bo_ref(bo, &mt->base.bo); + return &mt->base.base; } static void @@ -157,7 +158,7 @@ nv50_miptree_destroy(struct pipe_texture *pt) { struct nv50_miptree *mt = nv50_miptree(pt); - nouveau_bo_ref(NULL, &mt->bo); + nouveau_bo_ref(NULL, &mt->base.bo); FREE(mt); } diff --git a/src/gallium/drivers/nv50/nv50_program.c b/src/gallium/drivers/nv50/nv50_program.c index 289c3485e08..4a838529de7 100644 --- a/src/gallium/drivers/nv50/nv50_program.c +++ b/src/gallium/drivers/nv50/nv50_program.c @@ -1106,10 +1106,10 @@ convert_to_long(struct nv50_pc *pc, struct nv50_program_exec *e) m = 0xffff7fff; break; case 0x8: - /* INTERP */ - m = ~0x02000000; - if (e->inst[0] & 0x02000000) - q = 0x00020000; + /* INTERP (move centroid, perspective and flat bits) */ + m = ~0x03000100; + q = (e->inst[0] & (3 << 24)) >> (24 - 16); + q |= (e->inst[0] & (1 << 8)) << (18 - 8); break; case 0x9: /* RCP */ @@ -1495,6 +1495,7 @@ nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok) } break; case TGSI_OPCODE_MOV: + case TGSI_OPCODE_SWZ: for (c = 0; c < 4; c++) { if (!(mask & (1 << c))) continue; diff --git a/src/gallium/drivers/nv50/nv50_state_validate.c b/src/gallium/drivers/nv50/nv50_state_validate.c index 99d5b96e456..344c2cf6dde 100644 --- a/src/gallium/drivers/nv50/nv50_state_validate.c +++ b/src/gallium/drivers/nv50/nv50_state_validate.c @@ -33,7 +33,7 @@ nv50_state_validate_fb(struct nv50_context *nv50) for (i = 0; i < fb->nr_cbufs; i++) { struct pipe_texture *pt = fb->cbufs[i]->texture; - struct nouveau_bo *bo = nv50_miptree(pt)->bo; + struct nouveau_bo *bo = nv50_miptree(pt)->base.bo; if (!gw) { w = fb->cbufs[i]->width; @@ -75,7 +75,7 @@ nv50_state_validate_fb(struct nv50_context *nv50) if (fb->zsbuf) { struct pipe_texture *pt = fb->zsbuf->texture; - struct nouveau_bo *bo = nv50_miptree(pt)->bo; + struct nouveau_bo *bo = nv50_miptree(pt)->base.bo; if (!gw) { w = fb->zsbuf->width; diff --git a/src/gallium/drivers/nv50/nv50_surface.c b/src/gallium/drivers/nv50/nv50_surface.c index edaf4b055a1..b266324f58d 100644 --- a/src/gallium/drivers/nv50/nv50_surface.c +++ b/src/gallium/drivers/nv50/nv50_surface.c @@ -53,7 +53,7 @@ nv50_surface_set(struct nv50_screen *screen, struct pipe_surface *ps, int dst) struct nv50_miptree *mt = nv50_miptree(ps->texture); struct nouveau_channel *chan = screen->eng2d->channel; struct nouveau_grobj *eng2d = screen->eng2d; - struct nouveau_bo *bo = nv50_miptree(ps->texture)->bo; + struct nouveau_bo *bo = nv50_miptree(ps->texture)->base.bo; int format, mthd = dst ? NV50_2D_DST_FORMAT : NV50_2D_SRC_FORMAT; int flags = NOUVEAU_BO_VRAM | (dst ? NOUVEAU_BO_WR : NOUVEAU_BO_RD); diff --git a/src/gallium/drivers/nv50/nv50_tex.c b/src/gallium/drivers/nv50/nv50_tex.c index 14c68b96e14..033cb50c115 100644 --- a/src/gallium/drivers/nv50/nv50_tex.c +++ b/src/gallium/drivers/nv50/nv50_tex.c @@ -29,7 +29,7 @@ static int nv50_tex_construct(struct nv50_context *nv50, struct nouveau_stateobj *so, struct nv50_miptree *mt, int unit) { - switch (mt->base.format) { + switch (mt->base.base.format) { case PIPE_FORMAT_A8R8G8B8_UNORM: so_data(so, NV50TIC_0_0_MAPA_C3 | NV50TIC_0_0_TYPEA_UNORM | NV50TIC_0_0_MAPR_C2 | NV50TIC_0_0_TYPER_UNORM | @@ -118,18 +118,18 @@ nv50_tex_construct(struct nv50_context *nv50, struct nouveau_stateobj *so, return 1; } - so_reloc(so, mt->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_LOW | + so_reloc(so, mt->base.bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_LOW | NOUVEAU_BO_RD, 0, 0); if (nv50->sampler[unit]->normalized) - so_data (so, 0xd0005000 | mt->bo->tile_mode << 22); + so_data (so, 0xd0005000 | mt->base.bo->tile_mode << 22); else - so_data (so, 0x5001d000 | mt->bo->tile_mode << 22); + so_data (so, 0x5001d000 | mt->base.bo->tile_mode << 22); so_data (so, 0x00300000); - so_data (so, mt->base.width[0]); - so_data (so, (mt->base.last_level << 28) | - (mt->base.depth[0] << 16) | mt->base.height[0]); + so_data (so, mt->base.base.width[0]); + so_data (so, (mt->base.base.last_level << 28) | + (mt->base.base.depth[0] << 16) | mt->base.base.height[0]); so_data (so, 0x03000000); - so_data (so, mt->base.last_level << 4); + so_data (so, mt->base.base.last_level << 4); return 0; } diff --git a/src/gallium/drivers/nv50/nv50_transfer.c b/src/gallium/drivers/nv50/nv50_transfer.c index d2b5e4d75d4..e9c3562194b 100644 --- a/src/gallium/drivers/nv50/nv50_transfer.c +++ b/src/gallium/drivers/nv50/nv50_transfer.c @@ -148,8 +148,8 @@ nv50_transfer_new(struct pipe_screen *pscreen, struct pipe_texture *pt, tx->base.usage = usage; tx->level_pitch = lvl->pitch; - tx->level_width = mt->base.width[level]; - tx->level_height = mt->base.height[level]; + tx->level_width = mt->base.base.width[level]; + tx->level_height = mt->base.base.height[level]; tx->level_offset = lvl->image_offset[image]; tx->level_tiling = lvl->tile_mode; tx->level_x = x; @@ -162,7 +162,7 @@ nv50_transfer_new(struct pipe_screen *pscreen, struct pipe_texture *pt, } if (usage != PIPE_TRANSFER_WRITE) { - nv50_transfer_rect_m2mf(pscreen, mt->bo, tx->level_offset, + nv50_transfer_rect_m2mf(pscreen, mt->base.bo, tx->level_offset, tx->level_pitch, tx->level_tiling, x, y, tx->level_width, tx->level_height, @@ -188,7 +188,7 @@ nv50_transfer_del(struct pipe_transfer *ptx) nv50_transfer_rect_m2mf(pscreen, tx->bo, 0, tx->base.stride, tx->bo->tile_mode, 0, 0, tx->base.width, tx->base.height, - mt->bo, tx->level_offset, + mt->base.bo, tx->level_offset, tx->level_pitch, tx->level_tiling, tx->level_x, tx->level_y, tx->level_width, tx->level_height, diff --git a/src/gallium/drivers/nv50/nv50_vbo.c b/src/gallium/drivers/nv50/nv50_vbo.c index 36f1b24b2f0..eeed148c7b9 100644 --- a/src/gallium/drivers/nv50/nv50_vbo.c +++ b/src/gallium/drivers/nv50/nv50_vbo.c @@ -178,7 +178,7 @@ nv50_draw_elements_inline_u08(struct nv50_context *nv50, uint8_t *map, BEGIN_RING(chan, tesla, 0x400015f0, nr >> 1); for (i = 0; i < nr; i += 2) - OUT_RING (chan, (map[1] << 16) | map[0]); + OUT_RING (chan, (map[i + 1] << 16) | map[i]); count -= nr; map += nr; @@ -207,7 +207,7 @@ nv50_draw_elements_inline_u16(struct nv50_context *nv50, uint16_t *map, BEGIN_RING(chan, tesla, 0x400015f0, nr >> 1); for (i = 0; i < nr; i += 2) - OUT_RING (chan, (map[1] << 16) | map[0]); + OUT_RING (chan, (map[i + 1] << 16) | map[i]); count -= nr; map += nr; @@ -313,18 +313,18 @@ nv50_vbo_static_attrib(struct nv50_context *nv50, unsigned attrib, so_data (so, fui(v[3])); break; case 3: - so_method(so, tesla, NV50TCL_VTX_ATTR_3F_X(attrib), 4); + so_method(so, tesla, NV50TCL_VTX_ATTR_3F_X(attrib), 3); so_data (so, fui(v[0])); so_data (so, fui(v[1])); so_data (so, fui(v[2])); break; case 2: - so_method(so, tesla, NV50TCL_VTX_ATTR_2F_X(attrib), 4); + so_method(so, tesla, NV50TCL_VTX_ATTR_2F_X(attrib), 2); so_data (so, fui(v[0])); so_data (so, fui(v[1])); break; case 1: - so_method(so, tesla, NV50TCL_VTX_ATTR_1F(attrib), 4); + so_method(so, tesla, NV50TCL_VTX_ATTR_1F(attrib), 1); so_data (so, fui(v[0])); break; default: diff --git a/src/gallium/drivers/r300/r300_render.c b/src/gallium/drivers/r300/r300_render.c index cd458d019ae..aced4ab8877 100644 --- a/src/gallium/drivers/r300/r300_render.c +++ b/src/gallium/drivers/r300/r300_render.c @@ -233,7 +233,8 @@ static void r300_render_draw(struct vbuf_render* render, OUT_CS_INDEX_RELOC(index_buffer, 0, count, RADEON_GEM_DOMAIN_GTT, 0, 0); END_CS; */ - BEGIN_CS(2 + (count+1)/2); + BEGIN_CS(4 + (count+1)/2); + OUT_CS_REG(R300_VAP_VF_MAX_VTX_INDX, count); OUT_CS_PKT3(R300_PACKET3_3D_DRAW_INDX_2, (count+1)/2); OUT_CS(R300_VAP_VF_CNTL__PRIM_WALK_INDICES | (count << 16) | r300render->hwprim); diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c index c16cadd0407..81808017f86 100644 --- a/src/gallium/drivers/r300/r300_state.c +++ b/src/gallium/drivers/r300/r300_state.c @@ -429,6 +429,9 @@ static void r300_bind_rs_state(struct pipe_context* pipe, void* state) r300->rs_state = rs; r300->dirty_state |= R300_NEW_RASTERIZER; + r300->dirty_state |= R300_NEW_RS_BLOCK; + r300->dirty_state |= R300_NEW_SCISSOR; + r300->dirty_state |= R300_NEW_VIEWPORT; } /* Free rasterizer state. */ diff --git a/src/gallium/drivers/r300/r300_state_invariant.c b/src/gallium/drivers/r300/r300_state_invariant.c index 1e92374a4e9..7d822fec483 100644 --- a/src/gallium/drivers/r300/r300_state_invariant.c +++ b/src/gallium/drivers/r300/r300_state_invariant.c @@ -38,7 +38,9 @@ void r300_emit_invariant_state(struct r300_context* r300) /*** Graphics Backend (GB) ***/ /* Various GB enables */ - OUT_CS_REG(R300_GB_ENABLE, 0x0); + OUT_CS_REG(R300_GB_ENABLE, R300_GB_POINT_STUFF_ENABLE | + R300_GB_LINE_STUFF_ENABLE | + R300_GB_TRIANGLE_STUFF_ENABLE); /* Subpixel multisampling for AA * These are commented out because glisse's CS checker doesn't like them. * I presume these will be re-enabled later. diff --git a/src/gallium/drivers/r300/r300_surface.h b/src/gallium/drivers/r300/r300_surface.h index d01f0b143f5..f9e98b2ec9c 100644 --- a/src/gallium/drivers/r300/r300_surface.h +++ b/src/gallium/drivers/r300/r300_surface.h @@ -73,9 +73,9 @@ static struct r300_rs_state rs_clear_state = { }; static struct r300_rs_block r3xx_rs_block_clear_state = { - .ip[0] = R500_RS_SEL_S(R300_RS_SEL_K0) | - R500_RS_SEL_T(R300_RS_SEL_K0) | - R500_RS_SEL_R(R300_RS_SEL_K0) | + .ip[0] = R500_RS_SEL_S(R300_RS_SEL_C0) | + R500_RS_SEL_T(R300_RS_SEL_C0) | + R500_RS_SEL_R(R300_RS_SEL_C0) | R500_RS_SEL_Q(R300_RS_SEL_K1), .inst[0] = R300_RS_INST_COL_CN_WRITE, .count = R300_IT_COUNT(0) | R300_IC_COUNT(1) | R300_HIRES_EN, diff --git a/src/gallium/drivers/trace/tr_context.c b/src/gallium/drivers/trace/tr_context.c index 4ab718f233d..ae0af4d0557 100644 --- a/src/gallium/drivers/trace/tr_context.c +++ b/src/gallium/drivers/trace/tr_context.c @@ -1277,8 +1277,10 @@ trace_context_create(struct pipe_screen *_screen, tr_ctx->base.set_sampler_textures = trace_context_set_sampler_textures; tr_ctx->base.set_vertex_buffers = trace_context_set_vertex_buffers; tr_ctx->base.set_vertex_elements = trace_context_set_vertex_elements; - tr_ctx->base.surface_copy = trace_context_surface_copy; - tr_ctx->base.surface_fill = trace_context_surface_fill; + if (pipe->surface_copy) + tr_ctx->base.surface_copy = trace_context_surface_copy; + if (pipe->surface_fill) + tr_ctx->base.surface_fill = trace_context_surface_fill; tr_ctx->base.clear = trace_context_clear; tr_ctx->base.flush = trace_context_flush; tr_ctx->base.is_texture_referenced = trace_is_texture_referenced; diff --git a/src/gallium/include/pipe/p_shader_tokens.h b/src/gallium/include/pipe/p_shader_tokens.h index f0ba4fb308c..5fa6c9af30b 100644 --- a/src/gallium/include/pipe/p_shader_tokens.h +++ b/src/gallium/include/pipe/p_shader_tokens.h @@ -184,7 +184,7 @@ union tgsi_immediate_data #define TGSI_OPCODE_SUB 17 #define TGSI_OPCODE_LRP 18 #define TGSI_OPCODE_CND 19 -#define TGSI_OPCODE_CND0 20 + /* gap */ #define TGSI_OPCODE_DP2A 21 /* gap */ #define TGSI_OPCODE_FRC 24 diff --git a/src/gallium/state_trackers/dri/dri_drawable.c b/src/gallium/state_trackers/dri/dri_drawable.c index bcfd1c06fec..5cec9e329d9 100644 --- a/src/gallium/state_trackers/dri/dri_drawable.c +++ b/src/gallium/state_trackers/dri/dri_drawable.c @@ -189,12 +189,10 @@ dri_get_buffers(__DRIdrawablePrivate * dPriv) format = drawable->color_format; break; case __DRI_BUFFER_DEPTH: - index = ST_SURFACE_DEPTH; - format = drawable->depth_format; - break; + case __DRI_BUFFER_DEPTH_STENCIL: case __DRI_BUFFER_STENCIL: index = ST_SURFACE_DEPTH; - format = drawable->stencil_format; + format = drawable->depth_stencil_format; break; case __DRI_BUFFER_ACCUM: default: @@ -215,6 +213,18 @@ dri_get_buffers(__DRIdrawablePrivate * dPriv) dri_drawable->w, dri_drawable->h, buffers[i].pitch); + switch (buffers[i].attachment) { + case __DRI_BUFFER_FAKE_FRONT_LEFT: + case __DRI_BUFFER_BACK_LEFT: + drawable->color_format = surface->format; + break; + case __DRI_BUFFER_DEPTH: + case __DRI_BUFFER_DEPTH_STENCIL: + case __DRI_BUFFER_STENCIL: + drawable->depth_stencil_format = surface->format; + break; + } + st_set_framebuffer_surface(drawable->stfb, index, surface); pipe_surface_reference(&surface, NULL); } @@ -241,9 +251,7 @@ void dri2_set_tex_buffer2(__DRIcontext *pDRICtx, GLint target, st_get_framebuffer_surface(drawable->stfb, ST_SURFACE_FRONT_LEFT, &ps); st_bind_texture_surface(ps, target == GL_TEXTURE_2D ? ST_TEXTURE_2D : - ST_TEXTURE_RECT, 0, - format == GLX_TEXTURE_FORMAT_RGBA_EXT ? - PIPE_FORMAT_R8G8B8A8_UNORM : PIPE_FORMAT_R8G8B8X8_UNORM); + ST_TEXTURE_RECT, 0, drawable->color_format); } void dri2_set_tex_buffer(__DRIcontext *pDRICtx, GLint target, @@ -311,43 +319,31 @@ dri_create_buffer(__DRIscreenPrivate * sPriv, switch(visual->depthBits) { default: case 0: - drawable->depth_format = PIPE_FORMAT_NONE; + drawable->depth_stencil_format = PIPE_FORMAT_NONE; break; case 16: - drawable->depth_format = PIPE_FORMAT_Z16_UNORM; + drawable->depth_stencil_format = PIPE_FORMAT_Z16_UNORM; break; case 24: if (visual->stencilBits == 0) { - drawable->depth_format = (screen->d_depth_bits_last) ? + drawable->depth_stencil_format = (screen->d_depth_bits_last) ? PIPE_FORMAT_X8Z24_UNORM: PIPE_FORMAT_Z24X8_UNORM; } else { - drawable->depth_format = (screen->sd_depth_bits_last) ? + drawable->depth_stencil_format = (screen->sd_depth_bits_last) ? PIPE_FORMAT_S8Z24_UNORM: PIPE_FORMAT_Z24S8_UNORM; } break; case 32: - drawable->depth_format = PIPE_FORMAT_Z32_UNORM; - break; - } - - switch(visual->stencilBits) { - default: - case 0: - drawable->stencil_format = PIPE_FORMAT_NONE; - break; - case 8: - drawable->stencil_format = (screen->sd_depth_bits_last) ? - PIPE_FORMAT_S8Z24_UNORM: - PIPE_FORMAT_Z24S8_UNORM; + drawable->depth_stencil_format = PIPE_FORMAT_Z32_UNORM; break; } drawable->stfb = st_create_framebuffer(visual, drawable->color_format, - drawable->depth_format, - drawable->stencil_format, + drawable->depth_stencil_format, + drawable->depth_stencil_format, dPriv->w, dPriv->h, (void *)drawable); if (drawable->stfb == NULL) @@ -366,9 +362,11 @@ dri_create_buffer(__DRIscreenPrivate * sPriv, drawable->attachments[i++] = __DRI_BUFFER_BACK_LEFT; else drawable->attachments[i++] = __DRI_BUFFER_FAKE_FRONT_LEFT; - if (visual->depthBits) + if (visual->depthBits && visual->stencilBits) + drawable->attachments[i++] = __DRI_BUFFER_DEPTH_STENCIL; + else if (visual->depthBits) drawable->attachments[i++] = __DRI_BUFFER_DEPTH; - if (visual->stencilBits) + else if (visual->stencilBits) drawable->attachments[i++] = __DRI_BUFFER_STENCIL; drawable->num_attachments = i; diff --git a/src/gallium/state_trackers/dri/dri_drawable.h b/src/gallium/state_trackers/dri/dri_drawable.h index 9f9cb29b974..b910930db42 100644 --- a/src/gallium/state_trackers/dri/dri_drawable.h +++ b/src/gallium/state_trackers/dri/dri_drawable.h @@ -63,8 +63,7 @@ struct dri_drawable unsigned int cur_fences; enum pipe_format color_format; - enum pipe_format depth_format; - enum pipe_format stencil_format; + enum pipe_format depth_stencil_format; }; static INLINE struct dri_drawable * diff --git a/src/gallium/state_trackers/glx/xlib/xm_api.c b/src/gallium/state_trackers/glx/xlib/xm_api.c index f4c5893427d..957002ddd55 100644 --- a/src/gallium/state_trackers/glx/xlib/xm_api.c +++ b/src/gallium/state_trackers/glx/xlib/xm_api.c @@ -286,6 +286,18 @@ choose_pixel_format(XMesaVisual v) return PIPE_FORMAT_B8G8R8A8_UNORM; } } + else if ( GET_REDMASK(v) == 0x0000ff00 + && GET_GREENMASK(v) == 0x00ff0000 + && GET_BLUEMASK(v) == 0xff000000 + && v->BitsPerPixel == 32) { + if (native_byte_order) { + /* no byteswapping needed */ + return PIPE_FORMAT_B8G8R8A8_UNORM; + } + else { + return PIPE_FORMAT_A8R8G8B8_UNORM; + } + } else if ( GET_REDMASK(v) == 0xf800 && GET_GREENMASK(v) == 0x07e0 && GET_BLUEMASK(v) == 0x001f @@ -656,7 +668,7 @@ XMesaVisual XMesaCreateVisual( Display *display, * at a later time. */ v->visinfo = (XVisualInfo *) MALLOC(sizeof(*visinfo)); - if(!v->visinfo) { + if (!v->visinfo) { _mesa_free(v); return NULL; } @@ -743,7 +755,7 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) { static GLboolean firstTime = GL_TRUE; static struct pipe_screen *screen = NULL; - struct pipe_context *pipe; + struct pipe_context *pipe = NULL; XMesaContext c; GLcontext *mesaCtx; uint pf; @@ -769,8 +781,7 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) if (screen == NULL) goto fail; - pipe = driver.create_pipe_context( screen, - (void *)c ); + pipe = driver.create_pipe_context(screen, (void *) c); if (pipe == NULL) goto fail; @@ -783,23 +794,15 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) mesaCtx = c->st->ctx; c->st->ctx->DriverCtx = c; -#if 00 - _mesa_enable_sw_extensions(mesaCtx); - _mesa_enable_1_3_extensions(mesaCtx); - _mesa_enable_1_4_extensions(mesaCtx); - _mesa_enable_1_5_extensions(mesaCtx); - _mesa_enable_2_0_extensions(mesaCtx); -#endif - return c; - fail: +fail: if (c->st) st_destroy_context(c->st); else if (pipe) pipe->destroy(pipe); - FREE(c); + _mesa_free(c); return NULL; } @@ -1153,7 +1156,7 @@ void XMesaFlush( XMesaContext c ) XMesaBuffer XMesaFindBuffer( Display *dpy, Drawable d ) { XMesaBuffer b; - for (b=XMesaBufferList; b; b=b->Next) { + for (b = XMesaBufferList; b; b = b->Next) { if (b->drawable == d && b->xm_visual->display == dpy) { return b; } diff --git a/src/gallium/state_trackers/xorg/xorg_composite.c b/src/gallium/state_trackers/xorg/xorg_composite.c index af4ae05fb4a..c708ac31702 100644 --- a/src/gallium/state_trackers/xorg/xorg_composite.c +++ b/src/gallium/state_trackers/xorg/xorg_composite.c @@ -2,9 +2,10 @@ #include "xorg_exa_tgsi.h" -#include <cso_cache/cso_context.h> +#include "cso_cache/cso_context.h" +#include "util/u_draw_quad.h" -#include <pipe/p_inlines.h> +#include "pipe/p_inlines.h" struct xorg_composite_blend { int op:8; @@ -39,6 +40,24 @@ static const struct xorg_composite_blend xorg_blends[] = { PIPE_BLENDFACTOR_INV_SRC_ALPHA, PIPE_BLENDFACTOR_INV_SRC_ALPHA }, }; +static INLINE void +pixel_to_float4(PictFormatPtr format, + CARD32 pixel, float *color) +{ + CARD32 r, g, b, a; + + debug_assert(format->type == PictTypeDirect); + + r = (pixel >> format->direct.red) & format->direct.redMask; + g = (pixel >> format->direct.green) & format->direct.greenMask; + b = (pixel >> format->direct.blue) & format->direct.blueMask; + a = (pixel >> format->direct.alpha) & format->direct.alphaMask; + color[0] = ((float)r) / ((float)format->direct.redMask); + color[1] = ((float)g) / ((float)format->direct.greenMask); + color[2] = ((float)b) / ((float)format->direct.blueMask); + color[3] = ((float)a) / ((float)format->direct.alphaMask); +} + struct acceleration_info { int op : 16; int with_mask : 1; @@ -75,19 +94,168 @@ blend_for_op(int op) return xorg_blends[BLEND_OP_OVER]; } -static void -draw_texture(struct exa_context *exa) +static INLINE int +render_repeat_to_gallium(int mode) { -#if 0 - if (buf) { - util_draw_vertex_buffer(pipe, buf, 0, - PIPE_PRIM_TRIANGLE_FAN, - 4, /* verts */ - 2); /* attribs/vert */ - - pipe_buffer_reference(&buf, NULL); + switch(mode) { + case RepeatNone: + return PIPE_TEX_WRAP_CLAMP; + case RepeatNormal: + return PIPE_TEX_WRAP_REPEAT; + case RepeatReflect: + return PIPE_TEX_WRAP_MIRROR_REPEAT; + case RepeatPad: + return PIPE_TEX_WRAP_CLAMP_TO_EDGE; + default: + debug_printf("Unsupported repeat mode\n"); } -#endif + return PIPE_TEX_WRAP_REPEAT; +} + + +static INLINE void +setup_vertex0(float vertex[2][4], float x, float y, + float color[4]) +{ + vertex[0][0] = x; + vertex[0][1] = y; + vertex[0][2] = 0.f; /*z*/ + vertex[0][3] = 1.f; /*w*/ + + vertex[1][0] = color[0]; /*r*/ + vertex[1][1] = color[1]; /*g*/ + vertex[1][2] = color[2]; /*b*/ + vertex[1][3] = color[3]; /*a*/ +} + +static struct pipe_buffer * +setup_vertex_data0(struct exa_context *ctx, + int srcX, int srcY, int maskX, int maskY, + int dstX, int dstY, int width, int height) +{ + float vertices[4][2][4]; + + /* 1st vertex */ + setup_vertex0(vertices[0], dstX, dstY, + ctx->solid_color); + /* 2nd vertex */ + setup_vertex0(vertices[1], dstX + width, dstY, + ctx->solid_color); + /* 3rd vertex */ + setup_vertex0(vertices[2], dstX + width, dstY + height, + ctx->solid_color); + /* 4th vertex */ + setup_vertex0(vertices[3], dstX, dstY + height, + ctx->solid_color); + + return pipe_user_buffer_create(ctx->ctx->screen, + vertices, + sizeof(vertices)); +} + +static INLINE void +setup_vertex1(float vertex[2][4], float x, float y, float s, float t) +{ + vertex[0][0] = x; + vertex[0][1] = y; + vertex[0][2] = 0.f; /*z*/ + vertex[0][3] = 1.f; /*w*/ + + vertex[1][0] = s; /*s*/ + vertex[1][1] = t; /*t*/ + vertex[1][2] = 0.f; /*r*/ + vertex[1][3] = 1.f; /*q*/ +} + +static struct pipe_buffer * +setup_vertex_data1(struct exa_context *ctx, + int srcX, int srcY, int maskX, int maskY, + int dstX, int dstY, int width, int height) +{ + float vertices[4][2][4]; + float s0, t0, s1, t1; + struct pipe_texture *src = ctx->bound_textures[0]; + + s0 = srcX / src->width[0]; + s1 = srcX + width / src->width[0]; + t0 = srcY / src->height[0]; + t1 = srcY + height / src->height[0]; + + /* 1st vertex */ + setup_vertex1(vertices[0], dstX, dstY, + s0, t0); + /* 2nd vertex */ + setup_vertex1(vertices[1], dstX + width, dstY, + s1, t0); + /* 3rd vertex */ + setup_vertex1(vertices[2], dstX + width, dstY + height, + s1, t1); + /* 4th vertex */ + setup_vertex1(vertices[3], dstX, dstY + height, + s0, t1); + + return pipe_user_buffer_create(ctx->ctx->screen, + vertices, + sizeof(vertices)); +} + + +static INLINE void +setup_vertex2(float vertex[3][4], float x, float y, + float s0, float t0, float s1, float t1) +{ + vertex[0][0] = x; + vertex[0][1] = y; + vertex[0][2] = 0.f; /*z*/ + vertex[0][3] = 1.f; /*w*/ + + vertex[1][0] = s0; /*s*/ + vertex[1][1] = t0; /*t*/ + vertex[1][2] = 0.f; /*r*/ + vertex[1][3] = 1.f; /*q*/ + + vertex[2][0] = s1; /*s*/ + vertex[2][1] = t1; /*t*/ + vertex[2][2] = 0.f; /*r*/ + vertex[2][3] = 1.f; /*q*/ +} + +static struct pipe_buffer * +setup_vertex_data2(struct exa_context *ctx, + int srcX, int srcY, int maskX, int maskY, + int dstX, int dstY, int width, int height) +{ + float vertices[4][3][4]; + float st0[4], st1[4]; + struct pipe_texture *src = ctx->bound_textures[0]; + struct pipe_texture *mask = ctx->bound_textures[0]; + + st0[0] = srcX / src->width[0]; + st0[1] = srcY / src->height[0]; + st0[2] = srcX + width / src->width[0]; + st0[3] = srcY + height / src->height[0]; + + st1[0] = maskX / mask->width[0]; + st1[1] = maskY / mask->height[0]; + st1[2] = maskX + width / mask->width[0]; + st1[3] = maskY + height / mask->height[0]; + + /* 1st vertex */ + setup_vertex2(vertices[0], dstX, dstY, + st0[0], st0[1], st1[0], st1[1]); + /* 2nd vertex */ + setup_vertex2(vertices[1], dstX + width, dstY, + st0[2], st0[1], st1[2], st1[1]); + /* 3rd vertex */ + setup_vertex2(vertices[2], dstX + width, dstY + height, + st0[2], st0[3], st1[2], st1[3]); + /* 4th vertex */ + setup_vertex2(vertices[3], dstX, dstY + height, + st0[0], st0[3], st1[0], st1[3]); + + return pipe_user_buffer_create(ctx->ctx->screen, + vertices, + sizeof(vertices)); } boolean xorg_composite_accelerated(int op, @@ -172,27 +340,10 @@ set_viewport(struct exa_context *exa, int width, int height, static void bind_viewport_state(struct exa_context *exa, PicturePtr pDstPicture) { - const int param_bytes = 8 * sizeof(float); int width = pDstPicture->pDrawable->width; int height = pDstPicture->pDrawable->height; - float vs_consts[8] = { - 2.f/width, 2.f/height, 1, 1, - -1, -1, 0, 0 - }; - struct pipe_constant_buffer *cbuf = &exa->vs_const_buffer; - - set_viewport(exa, width, height, Y0_BOTTOM); - - pipe_buffer_reference(&cbuf->buffer, NULL); - cbuf->buffer = pipe_buffer_create(exa->ctx->screen, 16, - PIPE_BUFFER_USAGE_CONSTANT, - param_bytes); - if (cbuf->buffer) { - pipe_buffer_write(exa->ctx->screen, cbuf->buffer, - 0, param_bytes, vs_consts); - } - exa->ctx->set_constant_buffer(exa->ctx, PIPE_SHADER_VERTEX, 0, cbuf); + set_viewport(exa, width, height, Y0_TOP); } static void @@ -240,8 +391,20 @@ bind_shaders(struct exa_context *exa, int op, struct xorg_shader shader; if (pSrcPicture) { - vs_traits |= VS_COMPOSITE; - fs_traits |= FS_COMPOSITE; + if (pSrcPicture->pSourcePict) { + if (pSrcPicture->pSourcePict->type == SourcePictTypeSolidFill) { + fs_traits |= FS_SOLID_FILL; + vs_traits |= VS_SOLID_FILL; + pixel_to_float4(pSrcPicture->pFormat, + pSrcPicture->pSourcePict->solidFill.color, + exa->solid_color); + } else { + debug_assert("!gradients not supported"); + } + } else { + fs_traits |= FS_COMPOSITE; + vs_traits |= VS_COMPOSITE; + } } if (pMaskPicture) { @@ -264,35 +427,98 @@ bind_samplers(struct exa_context *exa, int op, struct exa_pixmap_priv *pDst) { struct pipe_sampler_state *samplers[PIPE_MAX_SAMPLERS]; - struct pipe_texture *textures[PIPE_MAX_SAMPLERS]; struct pipe_sampler_state src_sampler, mask_sampler; + exa->num_bound_samplers = 0; + memset(&src_sampler, 0, sizeof(struct pipe_sampler_state)); memset(&mask_sampler, 0, sizeof(struct pipe_sampler_state)); if (pSrcPicture && pSrc) { - src_sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; - src_sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + unsigned src_wrap = render_repeat_to_gallium( + pSrcPicture->repeatType); + src_sampler.wrap_s = src_wrap; + src_sampler.wrap_t = src_wrap; src_sampler.min_img_filter = PIPE_TEX_MIPFILTER_NEAREST; src_sampler.mag_img_filter = PIPE_TEX_MIPFILTER_NEAREST; src_sampler.normalized_coords = 1; samplers[0] = &src_sampler; - textures[0] = pSrc->tex; + exa->bound_textures[0] = pSrc->tex; + ++exa->num_bound_samplers; } if (pMaskPicture && pMask) { - mask_sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; - mask_sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + unsigned mask_wrap = render_repeat_to_gallium( + pMaskPicture->repeatType); + mask_sampler.wrap_s = mask_wrap; + mask_sampler.wrap_t = mask_wrap; mask_sampler.min_img_filter = PIPE_TEX_MIPFILTER_NEAREST; mask_sampler.mag_img_filter = PIPE_TEX_MIPFILTER_NEAREST; mask_sampler.normalized_coords = 1; samplers[1] = &mask_sampler; - textures[1] = pMask->tex; + exa->bound_textures[1] = pMask->tex; + ++exa->num_bound_samplers; } - cso_set_samplers(exa->cso, 3, + cso_set_samplers(exa->cso, exa->num_bound_samplers, (const struct pipe_sampler_state **)samplers); - cso_set_sampler_textures(exa->cso, 3, textures); + cso_set_sampler_textures(exa->cso, exa->num_bound_samplers, + exa->bound_textures); +} + +static void +setup_vs_constant_buffer(struct exa_context *exa, + int width, int height) +{ + const int param_bytes = 8 * sizeof(float); + float vs_consts[8] = { + 2.f/width, 2.f/height, 1, 1, + -1, -1, 0, 0 + }; + struct pipe_constant_buffer *cbuf = &exa->vs_const_buffer; + + pipe_buffer_reference(&cbuf->buffer, NULL); + cbuf->buffer = pipe_buffer_create(exa->ctx->screen, 16, + PIPE_BUFFER_USAGE_CONSTANT, + param_bytes); + + if (cbuf->buffer) { + pipe_buffer_write(exa->ctx->screen, cbuf->buffer, + 0, param_bytes, vs_consts); + } + exa->ctx->set_constant_buffer(exa->ctx, PIPE_SHADER_VERTEX, 0, cbuf); +} + + +static void +setup_fs_constant_buffer(struct exa_context *exa) +{ + const int param_bytes = 4 * sizeof(float); + float fs_consts[8] = { + 0, 0, 0, 1, + }; + struct pipe_constant_buffer *cbuf = &exa->fs_const_buffer; + + pipe_buffer_reference(&cbuf->buffer, NULL); + cbuf->buffer = pipe_buffer_create(exa->ctx->screen, 16, + PIPE_BUFFER_USAGE_CONSTANT, + param_bytes); + + if (cbuf->buffer) { + pipe_buffer_write(exa->ctx->screen, cbuf->buffer, + 0, param_bytes, fs_consts); + } + exa->ctx->set_constant_buffer(exa->ctx, PIPE_SHADER_FRAGMENT, 0, cbuf); +} + +static void +setup_constant_buffers(struct exa_context *exa, PicturePtr pDstPicture) +{ + int width = pDstPicture->pDrawable->width; + int height = pDstPicture->pDrawable->height; + + setup_vs_constant_buffer(exa, width, height); + setup_fs_constant_buffer(exa); } boolean xorg_composite_bind_state(struct exa_context *exa, @@ -309,8 +535,10 @@ boolean xorg_composite_bind_state(struct exa_context *exa, bind_blend_state(exa, op, pSrcPicture, pMaskPicture); bind_rasterizer_state(exa); bind_shaders(exa, op, pSrcPicture, pMaskPicture); - bind_samplers(exa, op, pSrcPicture, pMaskPicture, pDstPicture, - pSrc, pMask, pDst); + bind_samplers(exa, op, pSrcPicture, pMaskPicture, + pDstPicture, pSrc, pMask, pDst); + + setup_constant_buffers(exa, pDstPicture); return FALSE; } @@ -320,5 +548,37 @@ void xorg_composite(struct exa_context *exa, int srcX, int srcY, int maskX, int maskY, int dstX, int dstY, int width, int height) { + struct pipe_context *pipe = exa->ctx; + struct pipe_buffer *buf = 0; + + if (exa->num_bound_samplers == 0 ) { /* solid fill */ + buf = setup_vertex_data0(exa, + srcX, srcY, maskX, maskY, + dstX, dstY, width, height); + } else if (exa->num_bound_samplers == 1 ) /* src */ + buf = setup_vertex_data1(exa, + srcX, srcY, maskX, maskY, + dstX, dstY, width, height); + else if (exa->num_bound_samplers == 2) /* src + mask */ + buf = setup_vertex_data2(exa, + srcX, srcY, maskX, maskY, + dstX, dstY, width, height); + else if (exa->num_bound_samplers == 3) { /* src + mask + dst */ + debug_assert(!"src/mask/dst not handled right now"); +#if 0 + buf = setup_vertex_data2(exa, + srcX, srcY, maskX, maskY, + dstX, dstY, width, height); +#endif + } + + if (buf) { + util_draw_vertex_buffer(pipe, buf, 0, + PIPE_PRIM_TRIANGLE_FAN, + 4, /* verts */ + 1 + exa->num_bound_samplers); /* attribs/vert */ + + pipe_buffer_reference(&buf, NULL); + } } diff --git a/src/gallium/state_trackers/xorg/xorg_dri2.c b/src/gallium/state_trackers/xorg/xorg_dri2.c index 3b90421de9f..6431a0fe254 100644 --- a/src/gallium/state_trackers/xorg/xorg_dri2.c +++ b/src/gallium/state_trackers/xorg/xorg_dri2.c @@ -33,6 +33,7 @@ #include "xf86_OSproc.h" #include "xorg_tracker.h" +#include "xorg_exa.h" #include "dri2.h" @@ -47,62 +48,57 @@ typedef struct { struct pipe_fence_handle *fence; } *BufferPrivatePtr; -static DRI2BufferPtr -driCreateBuffers(DrawablePtr pDraw, unsigned int *attachments, int count) +static Bool +driDoCreateBuffer(DrawablePtr pDraw, DRI2BufferPtr buffer, unsigned int format) { - struct pipe_texture *depth, *tex; - struct pipe_buffer *buf; + struct pipe_texture *tex = NULL; ScreenPtr pScreen = pDraw->pScreen; ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum]; modesettingPtr ms = modesettingPTR(pScrn); - BufferPrivatePtr privates; - DRI2BufferPtr buffers; + struct exa_pixmap_priv *exa_priv; + BufferPrivatePtr private = buffer->driverPrivate; PixmapPtr pPixmap; unsigned stride, handle; - boolean have_depth = FALSE, have_stencil = FALSE; - int i; - - buffers = xcalloc(count, sizeof *buffers); - if (!buffers) - goto fail_buffers; - - privates = xcalloc(count, sizeof *privates); - if (!privates) - goto fail_privates; - - for (i = 0; i < count; i++) { - if (attachments[i] == DRI2BufferDepth) - have_depth = TRUE; - else if (attachments[i] == DRI2BufferStencil) - have_stencil = TRUE; - } - - if (have_stencil && !have_depth) - FatalError("Doesn't support only stencil yet\n"); - depth = NULL; - for (i = 0; i < count; i++) { - pPixmap = NULL; - tex = NULL; - buf = NULL; - if (attachments[i] == DRI2BufferFrontLeft) { - if (pDraw->type == DRAWABLE_PIXMAP) - pPixmap = (PixmapPtr) pDraw; - else - pPixmap = (*pScreen->GetWindowPixmap)((WindowPtr) pDraw); - pPixmap->refcnt++; - } else if (attachments[i] == DRI2BufferStencil) { - pipe_texture_reference(&tex, depth); - } else if (attachments[i] == DRI2BufferDepth) { + if (pDraw->type == DRAWABLE_PIXMAP) + pPixmap = (PixmapPtr) pDraw; + else + pPixmap = (*pScreen->GetWindowPixmap)((WindowPtr) pDraw); + exa_priv = exaGetPixmapDriverPrivate(pPixmap); + + switch (buffer->attachment) { + default: + if (buffer->attachment != DRI2BufferFakeFrontLeft || + pDraw->type != DRAWABLE_PIXMAP) { + private->pPixmap = (*pScreen->CreatePixmap)(pScreen, pDraw->width, + pDraw->height, + pDraw->depth, + 0); + } + break; + case DRI2BufferFrontLeft: + break; + case DRI2BufferStencil: +#if defined(DRI2INFOREC_VERSION) && DRI2INFOREC_VERSION > 2 + case DRI2BufferDepthStencil: + if (exa_priv->depth_stencil_tex && + !pf_is_depth_stencil(exa_priv->depth_stencil_tex->format)) + exa_priv->depth_stencil_tex = NULL; + /* Fall through */ +#endif + case DRI2BufferDepth: + if (exa_priv->depth_stencil_tex) + pipe_texture_reference(&tex, exa_priv->depth_stencil_tex); + else { struct pipe_texture template; memset(&template, 0, sizeof(template)); template.target = PIPE_TEXTURE_2D; - if (have_stencil) + if (buffer->attachment == DRI2BufferDepth) template.format = ms->ds_depth_bits_last ? - PIPE_FORMAT_S8Z24_UNORM : PIPE_FORMAT_Z24S8_UNORM; - else - template.format = ms->d_depth_bits_last ? PIPE_FORMAT_X8Z24_UNORM : PIPE_FORMAT_Z24X8_UNORM; + else + template.format = ms->ds_depth_bits_last ? + PIPE_FORMAT_S8Z24_UNORM : PIPE_FORMAT_Z24S8_UNORM; pf_get_block(template.format, &template.block); template.width[0] = pDraw->width; template.height[0] = pDraw->height; @@ -111,41 +107,119 @@ driCreateBuffers(DrawablePtr pDraw, unsigned int *attachments, int count) template.tex_usage = PIPE_TEXTURE_USAGE_DEPTH_STENCIL | PIPE_TEXTURE_USAGE_DISPLAY_TARGET; tex = ms->screen->texture_create(ms->screen, &template); - depth = tex; - } else if (attachments[i] == DRI2BufferFakeFrontLeft && - pDraw->type == DRAWABLE_PIXMAP) { - pPixmap = (PixmapPtr) pDraw; - pPixmap->refcnt++; - } else { - pPixmap = (*pScreen->CreatePixmap)(pScreen, pDraw->width, - pDraw->height, - pDraw->depth, - 0); + pipe_texture_reference(&exa_priv->depth_stencil_tex, tex); } + break; + } - if (pPixmap) { - xorg_exa_set_shared_usage(pPixmap); - pScreen->ModifyPixmapHeader(pPixmap, 0, 0, 0, 0, 0, NULL); - tex = xorg_exa_get_texture(pPixmap); - } + if (!private->pPixmap) { + private->pPixmap = pPixmap; + pPixmap->refcnt++; + } + + if (!tex) { + xorg_exa_set_shared_usage(private->pPixmap); + pScreen->ModifyPixmapHeader(private->pPixmap, 0, 0, 0, 0, 0, NULL); + tex = xorg_exa_get_texture(private->pPixmap); + } + + if (!tex) + FatalError("NO TEXTURE IN DRI2\n"); + + ms->api->shared_handle_from_texture(ms->api, ms->screen, tex, &stride, &handle); - if (!tex) - FatalError("NO TEXTURE IN DRI2\n"); + buffer->name = handle; + buffer->pitch = stride; + buffer->cpp = 4; + buffer->driverPrivate = private; + buffer->flags = 0; /* not tiled */ + private->tex = tex; + + return TRUE; +} + +static void +driDoDestroyBuffer(DrawablePtr pDraw, DRI2BufferPtr buffer) +{ + ScreenPtr pScreen = pDraw->pScreen; + ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum]; + modesettingPtr ms = modesettingPTR(pScrn); + BufferPrivatePtr private = buffer->driverPrivate; + struct exa_pixmap_priv *exa_priv = exaGetPixmapDriverPrivate(private->pPixmap); + + pipe_texture_reference(&private->tex, NULL); + ms->screen->fence_reference(ms->screen, &private->fence, NULL); + pipe_texture_reference(&exa_priv->depth_stencil_tex, NULL); + (*pScreen->DestroyPixmap)(private->pPixmap); +} - ms->api->shared_handle_from_texture(ms->api, ms->screen, tex, &stride, &handle); +#if defined(DRI2INFOREC_VERSION) && DRI2INFOREC_VERSION > 2 - buffers[i].name = handle; +static DRI2BufferPtr +driCreateBuffer(DrawablePtr pDraw, unsigned int attachment, unsigned int format) +{ + DRI2BufferPtr buffer; + BufferPrivatePtr private; + + buffer = xcalloc(1, sizeof *buffer); + if (!buffer) + return NULL; + + private = xcalloc(1, sizeof *private); + if (!private) { + goto fail; + } + + buffer->attachment = attachment; + buffer->driverPrivate = private; + + if (driDoCreateBuffer(pDraw, buffer, format)) + return buffer; + + xfree(private); +fail: + xfree(buffer); + return NULL; +} + +static void +driDestroyBuffer(DrawablePtr pDraw, DRI2BufferPtr buffer) +{ + driDoDestroyBuffer(pDraw, buffer); + + xfree(buffer->driverPrivate); + xfree(buffer); +} + +#else /* DRI2INFOREC_VERSION <= 2 */ + +static DRI2BufferPtr +driCreateBuffers(DrawablePtr pDraw, unsigned int *attachments, int count) +{ + BufferPrivatePtr privates; + DRI2BufferPtr buffers; + int i; + + buffers = xcalloc(count, sizeof *buffers); + if (!buffers) + goto fail_buffers; + + privates = xcalloc(count, sizeof *privates); + if (!privates) + goto fail_privates; + + for (i = 0; i < count; i++) { buffers[i].attachment = attachments[i]; - buffers[i].pitch = stride; - buffers[i].cpp = 4; buffers[i].driverPrivate = &privates[i]; - buffers[i].flags = 0; /* not tiled */ - privates[i].pPixmap = pPixmap; - privates[i].tex = tex; + + if (!driDoCreateBuffer(pDraw, &buffers[i], 0)) + goto fail; } return buffers; +fail: + xfree(privates); fail_privates: xfree(buffers); fail_buffers: @@ -155,21 +229,10 @@ fail_buffers: static void driDestroyBuffers(DrawablePtr pDraw, DRI2BufferPtr buffers, int count) { - ScreenPtr pScreen = pDraw->pScreen; - ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum]; - modesettingPtr ms = modesettingPTR(pScrn); - BufferPrivatePtr private; int i; - (void)ms; for (i = 0; i < count; i++) { - private = buffers[i].driverPrivate; - - pipe_texture_reference(&private->tex, NULL); - ms->screen->fence_reference(ms->screen, &private->fence, NULL); - - if (private->pPixmap) - (*pScreen->DestroyPixmap)(private->pPixmap); + driDoDestroyBuffer(pDraw, &buffers[i]); } if (buffers) { @@ -178,6 +241,8 @@ driDestroyBuffers(DrawablePtr pDraw, DRI2BufferPtr buffers, int count) } } +#endif /* DRI2INFOREC_VERSION */ + static void driCopyRegion(DrawablePtr pDraw, RegionPtr pRegion, DRI2BufferPtr pDestBuffer, DRI2BufferPtr pSrcBuffer) @@ -273,15 +338,25 @@ driScreenInit(ScreenPtr pScreen) modesettingPtr ms = modesettingPTR(pScrn); DRI2InfoRec dri2info; +#if defined(DRI2INFOREC_VERSION) + dri2info.version = DRI2INFOREC_VERSION; +#else dri2info.version = 1; +#endif dri2info.fd = ms->fd; dri2info.driverName = pScrn->driverName; dri2info.deviceName = "/dev/dri/card0"; /* FIXME */ +#if defined(DRI2INFOREC_VERSION) && DRI2INFOREC_VERSION > 2 + dri2info.CreateBuffer = driCreateBuffer; + dri2info.DestroyBuffer = driDestroyBuffer; +#else dri2info.CreateBuffers = driCreateBuffers; dri2info.DestroyBuffers = driDestroyBuffers; +#endif dri2info.CopyRegion = driCopyRegion; + dri2info.Wait = NULL; ms->d_depth_bits_last = ms->screen->is_format_supported(ms->screen, PIPE_FORMAT_X8Z24_UNORM, diff --git a/src/gallium/state_trackers/xorg/xorg_driver.c b/src/gallium/state_trackers/xorg/xorg_driver.c index bc3fd54bdb4..643b6b3b9e4 100644 --- a/src/gallium/state_trackers/xorg/xorg_driver.c +++ b/src/gallium/state_trackers/xorg/xorg_driver.c @@ -33,10 +33,8 @@ #include "xf86.h" #include "xf86_OSproc.h" #include "compiler.h" -#include "xf86RAC.h" #include "xf86PciInfo.h" #include "xf86Pci.h" -#include "xf86Resources.h" #include "mipointer.h" #include "micmap.h" #include <X11/extensions/randr.h> @@ -85,41 +83,9 @@ static const OptionInfoRec Options[] = { }; /* - * Functions that might be needed - */ - -static const char *exaSymbols[] = { - "exaGetVersion", - "exaDriverInit", - "exaDriverFini", - "exaOffscreenAlloc", - "exaOffscreenFree", - "exaWaitSync", - NULL -}; - -static const char *fbSymbols[] = { - "fbPictureInit", - "fbScreenInit", - NULL -}; - -static const char *ddcSymbols[] = { - "xf86PrintEDID", - "xf86SetDDCproperties", - NULL -}; - -/* * Exported Xorg driver functions to winsys */ -void -xorg_tracker_loader_ref_sym_lists() -{ - LoaderRefSymLists(exaSymbols, fbSymbols, ddcSymbols, NULL); -} - const OptionInfoRec * xorg_tracker_available_options(int chipid, int busid) { @@ -288,10 +254,6 @@ PreInit(ScrnInfoPtr pScrn, int flags) } else ms->entityPrivate = NULL; - if (xf86RegisterResources(ms->pEnt->index, NULL, ResNone)) { - return FALSE; - } - if (xf86IsEntityShared(pScrn->entityList[0])) { if (xf86IsPrimInitDone(pScrn->entityList[0])) { /* do something */ @@ -312,7 +274,6 @@ PreInit(ScrnInfoPtr pScrn, int flags) if (ms->fd < 0) return FALSE; - pScrn->racMemFlags = RAC_FB | RAC_COLORMAP; pScrn->monitor = pScrn->confScreen->monitor; pScrn->progClock = TRUE; pScrn->rgbBits = 8; @@ -398,8 +359,6 @@ PreInit(ScrnInfoPtr pScrn, int flags) return FALSE; } - xf86LoaderReqSymLists(fbSymbols, NULL); - xf86LoadSubModule(pScrn, "exa"); #ifdef DRI2 diff --git a/src/gallium/state_trackers/xorg/xorg_exa.c b/src/gallium/state_trackers/xorg/xorg_exa.c index 18afa01dbd5..a17a71f23a8 100644 --- a/src/gallium/state_trackers/xorg/xorg_exa.c +++ b/src/gallium/state_trackers/xorg/xorg_exa.c @@ -98,6 +98,68 @@ ExaMarkSync(ScreenPtr pScreen) } static Bool +ExaDownloadFromScreen(PixmapPtr pPix, int x, int y, int w, int h, char *dst, + int dst_pitch) +{ + ScreenPtr pScreen = pPix->drawable.pScreen; + ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum]; + modesettingPtr ms = modesettingPTR(pScrn); + struct exa_context *exa = ms->exa; + struct exa_pixmap_priv *priv = exaGetPixmapDriverPrivate(pPix); + struct pipe_transfer *transfer; + + if (!priv || !priv->tex) + return FALSE; + + if (exa->ctx->is_texture_referenced(exa->ctx, priv->tex, 0, 0) & + PIPE_REFERENCED_FOR_WRITE) + exa->ctx->flush(exa->ctx, 0, NULL); + + transfer = exa->scrn->get_tex_transfer(exa->scrn, priv->tex, 0, 0, 0, + PIPE_TRANSFER_READ, x, y, w, h); + if (!transfer) + return FALSE; + + util_copy_rect((unsigned char*)dst, &priv->tex->block, dst_pitch, 0, 0, + w, h, exa->scrn->transfer_map(exa->scrn, transfer), + transfer->stride, 0, 0); + + exa->scrn->transfer_unmap(exa->scrn, transfer); + exa->scrn->tex_transfer_destroy(transfer); + + return TRUE; +} + +static Bool +ExaUploadToScreen(PixmapPtr pPix, int x, int y, int w, int h, char *src, + int src_pitch) +{ + ScreenPtr pScreen = pPix->drawable.pScreen; + ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum]; + modesettingPtr ms = modesettingPTR(pScrn); + struct exa_context *exa = ms->exa; + struct exa_pixmap_priv *priv = exaGetPixmapDriverPrivate(pPix); + struct pipe_transfer *transfer; + + if (!priv || !priv->tex) + return FALSE; + + transfer = exa->scrn->get_tex_transfer(exa->scrn, priv->tex, 0, 0, 0, + PIPE_TRANSFER_WRITE, x, y, w, h); + if (!transfer) + return FALSE; + + util_copy_rect(exa->scrn->transfer_map(exa->scrn, transfer), + &priv->tex->block, transfer->stride, 0, 0, w, h, + (unsigned char*)src, src_pitch, 0, 0); + + exa->scrn->transfer_unmap(exa->scrn, transfer); + exa->scrn->tex_transfer_destroy(transfer); + + return TRUE; +} + +static Bool ExaPrepareAccess(PixmapPtr pPix, int index) { ScreenPtr pScreen = pPix->drawable.pScreen; @@ -188,9 +250,6 @@ ExaPrepareSolid(PixmapPtr pPixmap, int alu, Pixel planeMask, Pixel fg) struct exa_pixmap_priv *priv = exaGetPixmapDriverPrivate(pPixmap); struct exa_context *exa = ms->exa; - if (1) - return FALSE; - if (pPixmap->drawable.depth < 15) return FALSE; @@ -430,7 +489,7 @@ ExaModifyPixmapHeader(PixmapPtr pPixmap, int width, int height, modesettingPtr ms = modesettingPTR(pScrn); struct exa_context *exa = ms->exa; - if (!priv) + if (!priv || pPixData) return FALSE; if (depth <= 0) @@ -452,65 +511,69 @@ ExaModifyPixmapHeader(PixmapPtr pPixmap, int width, int height, bitsPerPixel, devKind, NULL); /* Deal with screen resize */ - if (priv->tex && (priv->tex->width[0] != width || - priv->tex->height[0] != height || - priv->tex_flags != priv->flags)) { - pipe_texture_reference(&priv->tex, NULL); - } + if (!priv->tex || + (priv->tex->width[0] != width || + priv->tex->height[0] != height || + priv->tex_flags != priv->flags)) { + struct pipe_texture *texture = NULL; - if (!priv->tex #ifdef DRM_MODE_FEATURE_DIRTYFB - && priv->flags + if (priv->flags) #endif - ) { - struct pipe_texture template; - - memset(&template, 0, sizeof(template)); - template.target = PIPE_TEXTURE_2D; - exa_get_pipe_format(depth, &template.format, &bitsPerPixel); - pf_get_block(template.format, &template.block); - template.width[0] = width; - template.height[0] = height; - template.depth[0] = 1; - template.last_level = 0; - template.tex_usage = PIPE_TEXTURE_USAGE_RENDER_TARGET | priv->flags; - priv->tex_flags = priv->flags; - priv->tex = exa->scrn->texture_create(exa->scrn, &template); - } - + { + struct pipe_texture template; + + memset(&template, 0, sizeof(template)); + template.target = PIPE_TEXTURE_2D; + exa_get_pipe_format(depth, &template.format, &bitsPerPixel); + pf_get_block(template.format, &template.block); + template.width[0] = width; + template.height[0] = height; + template.depth[0] = 1; + template.last_level = 0; + template.tex_usage = PIPE_TEXTURE_USAGE_RENDER_TARGET | priv->flags; + priv->tex_flags = priv->flags; + texture = exa->scrn->texture_create(exa->scrn, &template); + + if (priv->tex) { + struct pipe_surface *dst_surf; + + dst_surf = exa->scrn->get_tex_surface(exa->scrn, texture, 0, 0, 0, + PIPE_BUFFER_USAGE_GPU_WRITE); + priv->src_surf = exa_gpu_surface(exa, priv); + exa->ctx->surface_copy(exa->ctx, dst_surf, 0, 0, priv->src_surf, + 0, 0, min(width, texture->width[0]), + min(height, texture->height[0])); + exa->scrn->tex_surface_destroy(dst_surf); + exa->scrn->tex_surface_destroy(priv->src_surf); + priv->src_surf = NULL; + } else if (pPixmap->devPrivate.ptr) { + struct pipe_transfer *transfer; + + if (priv->map_count != 0) + FatalError("doing ExaModifyPixmapHeader on mapped buffer\n"); + + transfer = + exa->scrn->get_tex_transfer(exa->scrn, texture, 0, 0, 0, + PIPE_TRANSFER_WRITE, + 0, 0, width, height); + util_copy_rect(exa->scrn->transfer_map(exa->scrn, transfer), + &texture->block, transfer->stride, 0, 0, + width, height, pPixmap->devPrivate.ptr, + pPixmap->devKind, 0, 0); + exa->scrn->transfer_unmap(exa->scrn, transfer); + exa->scrn->tex_transfer_destroy(transfer); + } + } #ifdef DRM_MODE_FEATURE_DIRTYFB - if (!priv->tex) { - if (pPixData) - pPixmap->devPrivate.ptr = pPixData; - else - pPixmap->devPrivate.ptr = xalloc(pPixmap->drawable.height * pPixmap->devKind); - } else + else { + xfree(pPixmap->devPrivate.ptr); + pPixmap->devPrivate.ptr = xalloc(pPixmap->drawable.height * + pPixmap->devKind); + } #endif - if (pPixData) { - struct pipe_transfer *transfer = - exa->scrn->get_tex_transfer(exa->scrn, priv->tex, 0, 0, 0, - PIPE_TRANSFER_WRITE, - 0, 0, width, height); - util_copy_rect(exa->scrn->transfer_map(exa->scrn, transfer), - &priv->tex->block, transfer->stride, 0, 0, - width, height, pPixData, pPixmap->devKind, 0, 0); - exa->scrn->transfer_unmap(exa->scrn, transfer); - exa->scrn->tex_transfer_destroy(transfer); - } else if (priv->tex && pPixmap->devPrivate.ptr) { - struct pipe_transfer *transfer; - - if (priv->map_count != 0) - FatalError("doing ExaModifyPixmapHeader on mapped buffer\n"); - - transfer = - exa->scrn->get_tex_transfer(exa->scrn, priv->tex, 0, 0, 0, - PIPE_TRANSFER_WRITE, - 0, 0, width, height); - util_copy_rect(exa->scrn->transfer_map(exa->scrn, transfer), - &priv->tex->block, transfer->stride, 0, 0, - width, height, pPixmap->devPrivate.ptr, pPixmap->devKind, 0, 0); - exa->scrn->transfer_unmap(exa->scrn, transfer); - exa->scrn->tex_transfer_destroy(transfer); + + pipe_texture_reference(&priv->tex, texture); } return TRUE; @@ -598,6 +661,8 @@ xorg_exa_init(ScrnInfoPtr pScrn) pExa->Composite = ExaComposite; pExa->DoneComposite = ExaDoneComposite; pExa->PixmapIsOffscreen = ExaPixmapIsOffscreen; + pExa->DownloadFromScreen = ExaDownloadFromScreen; + pExa->UploadToScreen = ExaUploadToScreen; pExa->PrepareAccess = ExaPrepareAccess; pExa->FinishAccess = ExaFinishAccess; pExa->CreatePixmap = ExaCreatePixmap; diff --git a/src/gallium/state_trackers/xorg/xorg_exa.h b/src/gallium/state_trackers/xorg/xorg_exa.h index 0a93fa0bd7d..5b515be1397 100644 --- a/src/gallium/state_trackers/xorg/xorg_exa.h +++ b/src/gallium/state_trackers/xorg/xorg_exa.h @@ -8,6 +8,9 @@ struct cso_context; struct xorg_shaders; +/* src + mask + dst */ +#define MAX_EXA_SAMPLERS 3 + struct exa_context { ExaDriverPtr pExa; @@ -18,6 +21,11 @@ struct exa_context struct pipe_constant_buffer vs_const_buffer; struct pipe_constant_buffer fs_const_buffer; + + struct pipe_texture *bound_textures[MAX_EXA_SAMPLERS]; + int num_bound_samplers; + + float solid_color[4]; }; @@ -27,6 +35,7 @@ struct exa_pixmap_priv int tex_flags; struct pipe_texture *tex; + struct pipe_texture *depth_stencil_tex; unsigned int color; struct pipe_surface *src_surf; /* for copies */ diff --git a/src/gallium/state_trackers/xorg/xorg_exa_tgsi.c b/src/gallium/state_trackers/xorg/xorg_exa_tgsi.c index b5288bde4fb..cfee10c3b30 100644 --- a/src/gallium/state_trackers/xorg/xorg_exa_tgsi.c +++ b/src/gallium/state_trackers/xorg/xorg_exa_tgsi.c @@ -18,22 +18,24 @@ #include "cso_cache/cso_hash.h" /* Vertex shader: - * IN[0] = src_pos - * IN[1] = mask_pos - * IN[2] = dst_pos + * IN[0] = vertex pos + * IN[1] = src tex coord | solid fill color + * IN[2] = mask tex coord + * IN[3] = dst tex coord * CONST[0] = (2/dst_width, 2/dst_height, 1, 1) * CONST[1] = (-1, -1, 0, 0) * - * OUT[0] = src_pos - * OUT[1] = mask_pos - * OUT[2] = dst_pos + * OUT[0] = vertex pos + * OUT[1] = src tex coord | solid fill color + * OUT[2] = mask tex coord + * OUT[3] = dst tex coord */ /* Fragment shader: * SAMP[0] = src * SAMP[1] = mask * SAMP[2] = dst - * IN[0] = pos src + * IN[0] = pos src | solid fill color * IN[1] = pos mask * IN[2] = pos dst * CONST[0] = (0, 0, 0, 1) @@ -84,6 +86,150 @@ vs_normalize_coords(struct ureg_program *ureg, struct ureg_src coords, return ret; } +static void +linear_gradient(struct ureg_program *ureg, + struct ureg_dst out, + struct ureg_src pos, + struct ureg_src sampler, + struct ureg_src coords, + struct ureg_src const0124, + struct ureg_src matrow0, + struct ureg_src matrow1, + struct ureg_src matrow2) +{ + struct ureg_dst temp0 = ureg_DECL_temporary(ureg); + struct ureg_dst temp1 = ureg_DECL_temporary(ureg); + struct ureg_dst temp2 = ureg_DECL_temporary(ureg); + struct ureg_dst temp3 = ureg_DECL_temporary(ureg); + struct ureg_dst temp4 = ureg_DECL_temporary(ureg); + struct ureg_dst temp5 = ureg_DECL_temporary(ureg); + + ureg_MOV(ureg, + ureg_writemask(temp0, TGSI_WRITEMASK_XY), pos); + ureg_MOV(ureg, + ureg_writemask(temp0, TGSI_WRITEMASK_Z), + ureg_scalar(const0124, TGSI_SWIZZLE_Y)); + + ureg_DP3(ureg, temp1, matrow0, ureg_src(temp0)); + ureg_DP3(ureg, temp2, matrow1, ureg_src(temp0)); + ureg_DP3(ureg, temp3, matrow2, ureg_src(temp0)); + ureg_RCP(ureg, temp3, ureg_src(temp3)); + ureg_MUL(ureg, temp1, ureg_src(temp1), ureg_src(temp3)); + ureg_MUL(ureg, temp2, ureg_src(temp2), ureg_src(temp3)); + + ureg_MOV(ureg, ureg_writemask(temp4, TGSI_WRITEMASK_X), + ureg_src(temp1)); + ureg_MOV(ureg, ureg_writemask(temp4, TGSI_WRITEMASK_Y), + ureg_src(temp2)); + + ureg_MUL(ureg, temp0, + ureg_scalar(coords, TGSI_SWIZZLE_Y), + ureg_scalar(ureg_src(temp4), TGSI_SWIZZLE_Y)); + ureg_MAD(ureg, temp1, + ureg_scalar(coords, TGSI_SWIZZLE_X), + ureg_scalar(ureg_src(temp4), TGSI_SWIZZLE_X), + ureg_src(temp0)); + + ureg_MUL(ureg, temp2, + ureg_src(temp1), + ureg_scalar(coords, TGSI_SWIZZLE_Z)); + + ureg_TEX(ureg, out, + TGSI_TEXTURE_1D, ureg_src(temp2), sampler); + + ureg_release_temporary(ureg, temp0); + ureg_release_temporary(ureg, temp1); + ureg_release_temporary(ureg, temp2); + ureg_release_temporary(ureg, temp3); + ureg_release_temporary(ureg, temp4); + ureg_release_temporary(ureg, temp5); +} + + +static void +radial_gradient(struct ureg_program *ureg, + struct ureg_dst out, + struct ureg_src pos, + struct ureg_src sampler, + struct ureg_src coords, + struct ureg_src const0124, + struct ureg_src matrow0, + struct ureg_src matrow1, + struct ureg_src matrow2) +{ + struct ureg_dst temp0 = ureg_DECL_temporary(ureg); + struct ureg_dst temp1 = ureg_DECL_temporary(ureg); + struct ureg_dst temp2 = ureg_DECL_temporary(ureg); + struct ureg_dst temp3 = ureg_DECL_temporary(ureg); + struct ureg_dst temp4 = ureg_DECL_temporary(ureg); + struct ureg_dst temp5 = ureg_DECL_temporary(ureg); + + ureg_MOV(ureg, + ureg_writemask(temp0, TGSI_WRITEMASK_XY), + pos); + ureg_MOV(ureg, + ureg_writemask(temp0, TGSI_WRITEMASK_Z), + ureg_scalar(const0124, TGSI_SWIZZLE_Y)); + + ureg_DP3(ureg, temp1, matrow0, ureg_src(temp0)); + ureg_DP3(ureg, temp2, matrow1, ureg_src(temp0)); + ureg_DP3(ureg, temp3, matrow2, ureg_src(temp0)); + ureg_RCP(ureg, temp3, ureg_src(temp3)); + ureg_MUL(ureg, temp1, ureg_src(temp1), ureg_src(temp3)); + ureg_MUL(ureg, temp2, ureg_src(temp2), ureg_src(temp3)); + + ureg_MOV(ureg, ureg_writemask(temp5, TGSI_WRITEMASK_X), + ureg_src(temp1)); + ureg_MOV(ureg, ureg_writemask(temp5, TGSI_WRITEMASK_Y), + ureg_src(temp2)); + + ureg_MUL(ureg, temp0, ureg_scalar(coords, TGSI_SWIZZLE_Y), + ureg_scalar(ureg_src(temp5), TGSI_SWIZZLE_Y)); + ureg_MAD(ureg, temp1, + ureg_scalar(coords, TGSI_SWIZZLE_X), + ureg_scalar(ureg_src(temp5), TGSI_SWIZZLE_X), + ureg_src(temp0)); + ureg_ADD(ureg, temp1, + ureg_src(temp1), ureg_src(temp1)); + ureg_MUL(ureg, temp3, + ureg_scalar(ureg_src(temp5), TGSI_SWIZZLE_Y), + ureg_scalar(ureg_src(temp5), TGSI_SWIZZLE_Y)); + ureg_MAD(ureg, temp4, + ureg_scalar(ureg_src(temp5), TGSI_SWIZZLE_X), + ureg_scalar(ureg_src(temp5), TGSI_SWIZZLE_X), + ureg_src(temp3)); + ureg_MOV(ureg, temp4, ureg_negate(ureg_src(temp4))); + ureg_MUL(ureg, temp2, + ureg_scalar(coords, TGSI_SWIZZLE_Z), + ureg_src(temp4)); + ureg_MUL(ureg, temp0, + ureg_scalar(const0124, TGSI_SWIZZLE_W), + ureg_src(temp2)); + ureg_MUL(ureg, temp3, + ureg_src(temp1), ureg_src(temp1)); + ureg_SUB(ureg, temp2, + ureg_src(temp3), ureg_src(temp0)); + ureg_RSQ(ureg, temp2, ureg_abs(ureg_src(temp2))); + ureg_RCP(ureg, temp2, ureg_src(temp2)); + ureg_SUB(ureg, temp1, + ureg_src(temp2), ureg_src(temp1)); + ureg_ADD(ureg, temp0, + ureg_scalar(coords, TGSI_SWIZZLE_Z), + ureg_scalar(coords, TGSI_SWIZZLE_Z)); + ureg_RCP(ureg, temp0, ureg_src(temp0)); + ureg_MUL(ureg, temp2, + ureg_src(temp1), ureg_src(temp0)); + ureg_TEX(ureg, out, TGSI_TEXTURE_1D, + ureg_src(temp2), sampler); + + ureg_release_temporary(ureg, temp0); + ureg_release_temporary(ureg, temp1); + ureg_release_temporary(ureg, temp2); + ureg_release_temporary(ureg, temp3); + ureg_release_temporary(ureg, temp4); + ureg_release_temporary(ureg, temp5); +} + static void * create_vs(struct pipe_context *pipe, unsigned vs_traits) @@ -92,6 +238,9 @@ create_vs(struct pipe_context *pipe, struct ureg_src src; struct ureg_dst dst; struct ureg_src const0, const1; + boolean is_fill = vs_traits & VS_FILL; + boolean is_composite = vs_traits & VS_COMPOSITE; + boolean has_mask = vs_traits & VS_MASK; ureg = ureg_create(TGSI_PROCESSOR_VERTEX); if (ureg == NULL) @@ -100,18 +249,34 @@ create_vs(struct pipe_context *pipe, const0 = ureg_DECL_constant(ureg); const1 = ureg_DECL_constant(ureg); - if ((vs_traits & VS_COMPOSITE)) { + /* it has to be either a fill or a composite op */ + debug_assert(is_fill ^ is_composite); + + src = ureg_DECL_vs_input(ureg, + TGSI_SEMANTIC_POSITION, 0); + dst = ureg_DECL_output(ureg, TGSI_SEMANTIC_POSITION, 0); + src = vs_normalize_coords(ureg, src, + const0, const1); + ureg_MOV(ureg, dst, src); + + + if (is_composite) { src = ureg_DECL_vs_input(ureg, - TGSI_SEMANTIC_POSITION, 0); - dst = ureg_DECL_output(ureg, TGSI_SEMANTIC_POSITION, 0); - src = vs_normalize_coords(ureg, src, - const0, const1); + TGSI_SEMANTIC_GENERIC, 1); + dst = ureg_DECL_output(ureg, TGSI_SEMANTIC_GENERIC, 1); ureg_MOV(ureg, dst, src); } - if ((vs_traits & VS_MASK)) { + if (is_fill) { src = ureg_DECL_vs_input(ureg, - TGSI_SEMANTIC_POSITION, 1); - dst = ureg_DECL_output(ureg, TGSI_SEMANTIC_POSITION, 1); + TGSI_SEMANTIC_COLOR, 1); + dst = ureg_DECL_output(ureg, TGSI_SEMANTIC_COLOR, 1); + ureg_MOV(ureg, dst, src); + } + + if (has_mask) { + src = ureg_DECL_vs_input(ureg, + TGSI_SEMANTIC_GENERIC, 2); + dst = ureg_DECL_output(ureg, TGSI_SEMANTIC_POSITION, 2); ureg_MOV(ureg, dst, src); } @@ -125,27 +290,50 @@ create_fs(struct pipe_context *pipe, unsigned fs_traits) { struct ureg_program *ureg; - struct ureg_src dst_sampler, src_sampler, mask_sampler; - struct ureg_src dst_pos, src_pos, mask_pos; - struct ureg_src src, mask; + struct ureg_src /*dst_sampler,*/ src_sampler, mask_sampler; + struct ureg_src /*dst_pos,*/ src_input, mask_pos; + struct ureg_dst src, mask; struct ureg_dst out; + boolean has_mask = fs_traits & FS_MASK; + boolean is_fill = fs_traits & FS_FILL; + boolean is_composite = fs_traits & FS_COMPOSITE; + boolean is_solid = fs_traits & FS_SOLID_FILL; + boolean is_lingrad = fs_traits & FS_LINGRAD_FILL; + boolean is_radgrad = fs_traits & FS_RADGRAD_FILL; ureg = ureg_create(TGSI_PROCESSOR_FRAGMENT); if (ureg == NULL) return 0; + /* it has to be either a fill or a composite op */ + debug_assert(is_fill ^ is_composite); + out = ureg_DECL_output(ureg, TGSI_SEMANTIC_COLOR, 0); - src_sampler = ureg_DECL_sampler(ureg); - src_pos = ureg_DECL_fs_input(ureg, - TGSI_SEMANTIC_POSITION, - 0, - TGSI_INTERPOLATE_PERSPECTIVE); + if (is_composite) { + src_sampler = ureg_DECL_sampler(ureg, 0); + src_input = ureg_DECL_fs_input(ureg, + TGSI_SEMANTIC_POSITION, + 0, + TGSI_INTERPOLATE_PERSPECTIVE); + } + if (is_fill) { + if (is_solid) + src_input = ureg_DECL_fs_input(ureg, + TGSI_SEMANTIC_COLOR, + 0, + TGSI_INTERPOLATE_PERSPECTIVE); + else + src_input = ureg_DECL_fs_input(ureg, + TGSI_SEMANTIC_POSITION, + 0, + TGSI_INTERPOLATE_PERSPECTIVE); + } - if ((fs_traits & FS_MASK)) { - mask_sampler = ureg_DECL_sampler(ureg); + if (has_mask) { + mask_sampler = ureg_DECL_sampler(ureg, 1); mask_pos = ureg_DECL_fs_input(ureg, TGSI_SEMANTIC_POSITION, 1, @@ -153,21 +341,63 @@ create_fs(struct pipe_context *pipe, } #if 0 /* unused right now */ - dst_sampler = ureg_DECL_sampler(ureg); + dst_sampler = ureg_DECL_sampler(ureg, 2); dst_pos = ureg_DECL_fs_input(ureg, TGSI_SEMANTIC_POSITION, 2, TGSI_INTERPOLATE_PERSPECTIVE); #endif - if ((fs_traits & FS_MASK)) { - ureg_TEX(ureg, ureg_dst(mask), + if (is_composite) { + if (has_mask) + src = ureg_DECL_temporary(ureg); + else + src = out; + ureg_TEX(ureg, src, + TGSI_TEXTURE_2D, src_input, src_sampler); + } else if (is_fill) { + if (is_solid) { + if (has_mask) + src = ureg_dst(src_input); + else + ureg_MOV(ureg, out, src_input); + } else if (is_lingrad || is_radgrad) { + struct ureg_src coords, const0124, + matrow0, matrow1, matrow2; + + if (has_mask) + src = ureg_DECL_temporary(ureg); + else + src = out; + + coords = ureg_DECL_constant(ureg); + const0124 = ureg_DECL_constant(ureg); + matrow0 = ureg_DECL_constant(ureg); + matrow1 = ureg_DECL_constant(ureg); + matrow2 = ureg_DECL_constant(ureg); + + if (is_lingrad) { + linear_gradient(ureg, src, + src_input, src_sampler, + coords, const0124, + matrow0, matrow1, matrow2); + } else if (is_radgrad) { + radial_gradient(ureg, src, + src_input, src_sampler, + coords, const0124, + matrow0, matrow1, matrow2); + } + } else + debug_assert(!"Unknown fill type!"); + } + + if (has_mask) { + mask = ureg_DECL_temporary(ureg); + ureg_TEX(ureg, mask, TGSI_TEXTURE_2D, mask_pos, mask_sampler); /* src IN mask */ - src_in_mask(ureg, out, src, mask); - } else { - ureg_TEX(ureg, out, - TGSI_TEXTURE_2D, src_pos, src_sampler); + src_in_mask(ureg, out, ureg_src(src), ureg_src(mask)); + ureg_release_temporary(ureg, mask); } ureg_END(ureg); diff --git a/src/gallium/state_trackers/xorg/xorg_exa_tgsi.h b/src/gallium/state_trackers/xorg/xorg_exa_tgsi.h index 003e5d8caf4..1535a0c8c30 100644 --- a/src/gallium/state_trackers/xorg/xorg_exa_tgsi.h +++ b/src/gallium/state_trackers/xorg/xorg_exa_tgsi.h @@ -6,16 +6,24 @@ enum xorg_vs_traits { VS_COMPOSITE = 1 << 0, VS_MASK = 1 << 1, - VS_FILL = 1 << 2 - /*VS_TRANSFORM = 1 << 3*/ + VS_SOLID_FILL = 1 << 2, + VS_LINGRAD_FILL = 1 << 3, + VS_RADGRAD_FILL = 1 << 4, + VS_FILL = (VS_SOLID_FILL | + VS_LINGRAD_FILL | + VS_RADGRAD_FILL) + /*VS_TRANSFORM = 1 << 5*/ }; enum xorg_fs_traits { FS_COMPOSITE = 1 << 0, FS_MASK = 1 << 1, - FS_FILL = 1 << 2, - FS_LINEAR_GRADIENT = 1 << 3, - FS_RADIAL_GRADIENT = 1 << 4 + FS_SOLID_FILL = 1 << 2, + FS_LINGRAD_FILL = 1 << 3, + FS_RADGRAD_FILL = 1 << 4, + FS_FILL = (FS_SOLID_FILL | + FS_LINGRAD_FILL | + FS_RADGRAD_FILL) }; struct xorg_shader { diff --git a/src/gallium/state_trackers/xorg/xorg_winsys.h b/src/gallium/state_trackers/xorg/xorg_winsys.h index d523080e90f..47ee4b9ffd8 100644 --- a/src/gallium/state_trackers/xorg/xorg_winsys.h +++ b/src/gallium/state_trackers/xorg/xorg_winsys.h @@ -37,7 +37,6 @@ #include "xorg-server.h" #include "xf86.h" -#include "xf86Resources.h" #include "pciaccess.h" #ifndef XSERVER_LIBPCIACCESS @@ -46,6 +45,5 @@ void xorg_tracker_set_functions(ScrnInfoPtr scrn); const OptionInfoRec * xorg_tracker_available_options(int chipid, int busid); -void xorg_tracker_loader_ref_sym_lists(void); #endif diff --git a/src/gallium/winsys/drm/intel/gem/Makefile b/src/gallium/winsys/drm/intel/gem/Makefile index 7ab1a2a771b..0d6d4e37dbd 100644 --- a/src/gallium/winsys/drm/intel/gem/Makefile +++ b/src/gallium/winsys/drm/intel/gem/Makefile @@ -4,10 +4,10 @@ include $(TOP)/configs/current LIBNAME = inteldrm C_SOURCES = \ - intel_be_batchbuffer.c \ - intel_be_context.c \ - intel_be_device.c \ - intel_be_api.c + intel_drm_batchbuffer.c \ + intel_drm_buffer.c \ + intel_drm_fence.c \ + intel_drm_api.c LIBRARY_INCLUDES = $(shell pkg-config libdrm --cflags-only-I) diff --git a/src/gallium/winsys/drm/intel/gem/SConscript b/src/gallium/winsys/drm/intel/gem/SConscript index ea8a2e55f62..26717f391fa 100644 --- a/src/gallium/winsys/drm/intel/gem/SConscript +++ b/src/gallium/winsys/drm/intel/gem/SConscript @@ -3,10 +3,10 @@ Import('*') env = drienv.Clone() inteldrm_sources = [ - 'intel_be_api.c', - 'intel_be_batchbuffer.c', - 'intel_be_context.c', - 'intel_be_device.c', + 'intel_drm_api.c', + 'intel_drm_batchbuffer.c', + 'intel_drm_buffer.c', + 'intel_drm_fence.c', ] inteldrm = env.ConvenienceLibrary( diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_api.c b/src/gallium/winsys/drm/intel/gem/intel_be_api.c deleted file mode 100644 index 3e90088d5e0..00000000000 --- a/src/gallium/winsys/drm/intel/gem/intel_be_api.c +++ /dev/null @@ -1,36 +0,0 @@ - -#include "intel_be_api.h" -#include "i915simple/i915_winsys.h" -#include "identity/id_drm.h" -#include "trace/tr_drm.h" - -static void destroy(struct drm_api *api) -{ - -} - -struct drm_api intel_be_drm_api = -{ - /* intel_be_context.c */ - .create_context = intel_be_create_context, - /* intel_be_device.c */ - .create_screen = intel_be_create_screen, - .texture_from_shared_handle = intel_be_texture_from_shared_handle, - .shared_handle_from_texture = intel_be_shared_handle_from_texture, - .local_handle_from_texture = intel_be_local_handle_from_texture, - .destroy = destroy, -}; - -struct drm_api * -drm_api_create() -{ -#ifdef DEBUG -#if 0 - return identity_drm_create(&intel_be_drm_api); -#else - return trace_drm_create(&intel_be_drm_api); -#endif -#else - return &intel_be_drm_api; -#endif -} diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_api.h b/src/gallium/winsys/drm/intel/gem/intel_be_api.h deleted file mode 100644 index f286b62eb8c..00000000000 --- a/src/gallium/winsys/drm/intel/gem/intel_be_api.h +++ /dev/null @@ -1,18 +0,0 @@ - -#ifndef _INTEL_BE_API_H_ -#define _INTEL_BE_API_H_ - -#include "pipe/p_compiler.h" - -#include "state_tracker/drm_api.h" - -#include "intel_be_device.h" - -extern struct drm_api intel_be_drm_api; - -struct pipe_screen *intel_be_create_screen(struct drm_api *api, int drmFD, - struct drm_create_screen_arg *arg); -struct pipe_context *intel_be_create_context(struct drm_api *api, - struct pipe_screen *screen); - -#endif diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_batchbuffer.c b/src/gallium/winsys/drm/intel/gem/intel_be_batchbuffer.c deleted file mode 100644 index c4a79586e68..00000000000 --- a/src/gallium/winsys/drm/intel/gem/intel_be_batchbuffer.c +++ /dev/null @@ -1,156 +0,0 @@ - -#include "i915simple/i915_debug.h" -#include "intel_be_batchbuffer.h" -#include "intel_be_context.h" -#include "intel_be_device.h" -#include "intel_be_fence.h" -#include <errno.h> - -#include "util/u_memory.h" - -struct intel_be_batchbuffer * -intel_be_batchbuffer_alloc(struct intel_be_context *intel) -{ - struct intel_be_batchbuffer *batch = CALLOC_STRUCT(intel_be_batchbuffer); - - - batch->base.buffer = NULL; - batch->base.winsys = &intel->base; - batch->base.map = NULL; - batch->base.ptr = NULL; - batch->base.size = 0; - batch->base.actual_size = intel->device->max_batch_size; - batch->base.relocs = 0; - batch->base.max_relocs = 100;/*INTEL_DEFAULT_RELOCS;*/ - - batch->intel = intel; - batch->device = intel->device; - - intel_be_batchbuffer_reset(batch); - - return batch; -} - -void -intel_be_batchbuffer_reset(struct intel_be_batchbuffer *batch) -{ - struct intel_be_context *intel = intel_be_context(batch->base.winsys); - struct intel_be_device *dev = intel->device; - - if (batch->bo) - drm_intel_bo_unreference(batch->bo); - batch->bo = drm_intel_bo_alloc(dev->pools.gem, - "gallium3d_batch_buffer", - batch->base.actual_size, - 4096); - drm_intel_bo_map(batch->bo, TRUE); - batch->base.map = batch->bo->virtual; - - memset(batch->base.map, 0, batch->base.actual_size); - batch->base.ptr = batch->base.map; - batch->base.size = batch->base.actual_size - BATCH_RESERVED; - batch->base.relocs = 0; -} - -int -intel_be_offset_relocation(struct intel_be_batchbuffer *batch, - unsigned pre_add, - drm_intel_bo *bo, - uint32_t read_domains, - uint32_t write_domain) -{ - unsigned offset; - int ret = 0; - - assert(batch->base.relocs < batch->base.max_relocs); - - offset = (unsigned)(batch->base.ptr - batch->base.map); - - ret = drm_intel_bo_emit_reloc(batch->bo, offset, - bo, pre_add, - read_domains, - write_domain); - - ((uint32_t*)batch->base.ptr)[0] = bo->offset + pre_add; - batch->base.ptr += 4; - - if (!ret) - batch->base.relocs++; - - return ret; -} - -void -intel_be_batchbuffer_flush(struct intel_be_batchbuffer *batch, - struct intel_be_fence **fence) -{ - struct i915_batchbuffer *i915 = &batch->base; - unsigned used = 0; - int ret = 0; - int i; - - assert(i915_batchbuffer_space(i915) >= 0); - - used = batch->base.ptr - batch->base.map; - assert((used & 3) == 0); - - if (used & 4) { - i915_batchbuffer_dword(i915, (0x0<<29)|(0x4<<23)|(1<<0)); // MI_FLUSH | FLUSH_MAP_CACHE; - i915_batchbuffer_dword(i915, (0x0<<29)|(0x0<<23)); // MI_NOOP - i915_batchbuffer_dword(i915, (0x0<<29)|(0xA<<23)); // MI_BATCH_BUFFER_END; - } else { - i915_batchbuffer_dword(i915, (0x0<<29)|(0x4<<23)|(1<<0)); //MI_FLUSH | FLUSH_MAP_CACHE; - i915_batchbuffer_dword(i915, (0x0<<29)|(0xA<<23)); // MI_BATCH_BUFFER_END; - } - - used = batch->base.ptr - batch->base.map; - - drm_intel_bo_unmap(batch->bo); - - /* Do the sending to HW */ - ret = drm_intel_bo_exec(batch->bo, used, NULL, 0, 0); - assert(ret == 0); - - if (batch->device->dump_cmd) { - unsigned *ptr; - drm_intel_bo_map(batch->bo, FALSE); - ptr = (unsigned*)batch->bo->virtual; - - debug_printf("%s:\n", __func__); - for (i = 0; i < used / 4; i++, ptr++) { - debug_printf("\t%08x: %08x\n", i*4, *ptr); - } - - drm_intel_bo_unmap(batch->bo); - } else { - /* TODO figgure out why the gpu hangs if we don't run sync */ - drm_intel_bo_map(batch->bo, FALSE); - drm_intel_bo_unmap(batch->bo); - } - - intel_be_batchbuffer_reset(batch); - - if (fence) { - if (*fence) - intel_be_fence_reference(fence, NULL); - - (*fence) = CALLOC_STRUCT(intel_be_fence); - pipe_reference_init(&(*fence)->reference, 1); - (*fence)->bo = NULL; - } -} - -void -intel_be_batchbuffer_finish(struct intel_be_batchbuffer *batch) -{ - -} - -void -intel_be_batchbuffer_free(struct intel_be_batchbuffer *batch) -{ - if (batch->bo) - drm_intel_bo_unreference(batch->bo); - - free(batch); -} diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_batchbuffer.h b/src/gallium/winsys/drm/intel/gem/intel_be_batchbuffer.h deleted file mode 100644 index 195bf8dee77..00000000000 --- a/src/gallium/winsys/drm/intel/gem/intel_be_batchbuffer.h +++ /dev/null @@ -1,55 +0,0 @@ - -#ifndef INTEL_BE_BATCHBUFFER_H -#define INTEL_BE_BATCHBUFFER_H - -#include "i915simple/i915_batch.h" - -#include "drm.h" -#include "intel_bufmgr.h" - -#define BATCH_RESERVED 16 - -#define INTEL_DEFAULT_RELOCS 100 -#define INTEL_MAX_RELOCS 400 - -#define INTEL_BATCH_NO_CLIPRECTS 0x1 -#define INTEL_BATCH_CLIPRECTS 0x2 - -struct intel_be_context; -struct intel_be_device; -struct intel_be_fence; - -struct intel_be_batchbuffer -{ - struct i915_batchbuffer base; - - struct intel_be_context *intel; - struct intel_be_device *device; - - drm_intel_bo *bo; -}; - -struct intel_be_batchbuffer * -intel_be_batchbuffer_alloc(struct intel_be_context *intel); - -void -intel_be_batchbuffer_free(struct intel_be_batchbuffer *batch); - -void -intel_be_batchbuffer_finish(struct intel_be_batchbuffer *batch); - -void -intel_be_batchbuffer_flush(struct intel_be_batchbuffer *batch, - struct intel_be_fence **fence); - -void -intel_be_batchbuffer_reset(struct intel_be_batchbuffer *batch); - -int -intel_be_offset_relocation(struct intel_be_batchbuffer *batch, - unsigned pre_add, - drm_intel_bo *bo, - uint32_t read_domains, - uint32_t write_doman); - -#endif diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_context.c b/src/gallium/winsys/drm/intel/gem/intel_be_context.c deleted file mode 100644 index ff4518f8680..00000000000 --- a/src/gallium/winsys/drm/intel/gem/intel_be_context.c +++ /dev/null @@ -1,117 +0,0 @@ - -#include "pipe/p_screen.h" - -#include "softpipe/sp_winsys.h" - -#include "intel_be_device.h" -#include "intel_be_context.h" -#include "intel_be_batchbuffer.h" - -#include "i915_drm.h" - -#include "intel_be_api.h" - -static struct i915_batchbuffer * -intel_be_batch_get(struct i915_winsys *sws) -{ - struct intel_be_context *intel = intel_be_context(sws); - return &intel->batch->base; -} - -static void -intel_be_batch_reloc(struct i915_winsys *sws, - struct pipe_buffer *buf, - unsigned access_flags, - unsigned delta) -{ - struct intel_be_context *intel = intel_be_context(sws); - drm_intel_bo *bo = intel_bo(buf); - int ret; - uint32_t read = 0; - uint32_t write = 0; - - if (access_flags & I915_BUFFER_ACCESS_WRITE) { - write = I915_GEM_DOMAIN_RENDER; - read = I915_GEM_DOMAIN_RENDER; - } - - if (access_flags & I915_BUFFER_ACCESS_READ) { - read |= I915_GEM_DOMAIN_SAMPLER; - } - - ret = intel_be_offset_relocation(intel->batch, - delta, - bo, - read, - write); - assert(ret == 0); - - /* TODO change return type */ - /* return ret; */ -} - -static void -intel_be_batch_flush(struct i915_winsys *sws, - struct pipe_fence_handle **fence) -{ - struct intel_be_context *intel = intel_be_context(sws); - struct intel_be_fence **f = (struct intel_be_fence **)fence; - - intel_be_batchbuffer_flush(intel->batch, f); -} - - -/* - * Misc functions. - */ - -static void -intel_be_destroy_context(struct i915_winsys *winsys) -{ - struct intel_be_context *intel = intel_be_context(winsys); - - intel_be_batchbuffer_free(intel->batch); - - free(intel); -} - -boolean -intel_be_init_context(struct intel_be_context *intel, struct intel_be_device *device) -{ - assert(intel); - assert(device); - intel->device = device; - - intel->base.batch_get = intel_be_batch_get; - intel->base.batch_reloc = intel_be_batch_reloc; - intel->base.batch_flush = intel_be_batch_flush; - - intel->base.destroy = intel_be_destroy_context; - - intel->batch = intel_be_batchbuffer_alloc(intel); - - return true; -} - -struct pipe_context * -intel_be_create_context(struct drm_api *api, struct pipe_screen *screen) -{ - struct intel_be_context *intel; - struct pipe_context *pipe; - struct intel_be_device *device = intel_be_device(screen->winsys); - - intel = (struct intel_be_context *)malloc(sizeof(*intel)); - memset(intel, 0, sizeof(*intel)); - - intel_be_init_context(intel, device); - - if (device->softpipe) - pipe = softpipe_create(screen); - else - pipe = i915_create_context(screen, &device->base, &intel->base); - - if (pipe) - pipe->priv = intel; - - return pipe; -} diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_context.h b/src/gallium/winsys/drm/intel/gem/intel_be_context.h deleted file mode 100644 index 5a369669c0d..00000000000 --- a/src/gallium/winsys/drm/intel/gem/intel_be_context.h +++ /dev/null @@ -1,31 +0,0 @@ - -#ifndef INTEL_BE_CONTEXT_H -#define INTEL_BE_CONTEXT_H - -#include "i915simple/i915_winsys.h" - -struct intel_be_context -{ - /** Interface to i915simple driver */ - struct i915_winsys base; - - struct intel_be_device *device; - struct intel_be_batchbuffer *batch; -}; - -static INLINE struct intel_be_context * -intel_be_context(struct i915_winsys *sws) -{ - return (struct intel_be_context *)sws; -} - -/** - * Intialize a allocated intel_be_context struct. - * - * Remember to set the hardware_* functions. - */ -boolean -intel_be_init_context(struct intel_be_context *intel, - struct intel_be_device *device); - -#endif diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_device.c b/src/gallium/winsys/drm/intel/gem/intel_be_device.c deleted file mode 100644 index 2d8e9f1a24e..00000000000 --- a/src/gallium/winsys/drm/intel/gem/intel_be_device.c +++ /dev/null @@ -1,474 +0,0 @@ - -#include "intel_be_device.h" - -#include "pipe/internal/p_winsys_screen.h" -#include "pipe/p_defines.h" -#include "pipe/p_state.h" -#include "pipe/p_inlines.h" -#include "util/u_memory.h" -#include "util/u_debug.h" -#include "util/u_math.h" - -#include "intel_be_fence.h" - -#include "i915simple/i915_winsys.h" -#include "softpipe/sp_winsys.h" - -#include "intel_be_api.h" -#include <stdio.h> - -#define I915_TILING_X 1 - -/* - * Buffer - */ - -static void * -intel_be_buffer_map(struct pipe_winsys *winsys, - struct pipe_buffer *buf, - unsigned flags) -{ - struct intel_be_buffer *buffer = intel_be_buffer(buf); - drm_intel_bo *bo = intel_bo(buf); - int write = 0; - int ret = 0; - - if (flags & PIPE_BUFFER_USAGE_DONTBLOCK) { - /* Remove this when drm_intel_bo_map supports DONTBLOCK - */ - return NULL; - } - - if (flags & PIPE_BUFFER_USAGE_CPU_WRITE) - write = 1; - - if (buffer->map_count) - goto out; - - if (buffer->map_gtt) - ret = drm_intel_gem_bo_map_gtt(bo); - else - ret = drm_intel_bo_map(bo, write); - - buffer->ptr = bo->virtual; - -out: - if (ret) - return NULL; - - buffer->map_count++; - return buffer->ptr; -} - -static void -intel_be_buffer_unmap(struct pipe_winsys *winsys, - struct pipe_buffer *buf) -{ - struct intel_be_buffer *buffer = intel_be_buffer(buf); - - if (--buffer->map_count) - return; - - if (buffer->map_gtt) - drm_intel_gem_bo_unmap_gtt(intel_bo(buf)); - else - drm_intel_bo_unmap(intel_bo(buf)); -} - -static void -intel_be_buffer_destroy(struct pipe_buffer *buf) -{ - drm_intel_bo_unreference(intel_bo(buf)); - free(buf); -} - -static struct pipe_buffer * -intel_be_buffer_create(struct pipe_winsys *winsys, - unsigned alignment, - unsigned usage, - unsigned size) -{ - struct intel_be_buffer *buffer = CALLOC_STRUCT(intel_be_buffer); - struct intel_be_device *dev = intel_be_device(winsys); - drm_intel_bufmgr *pool; - char *name; - - if (!buffer) - return NULL; - - pipe_reference_init(&buffer->base.reference, 1); - buffer->base.alignment = alignment; - buffer->base.usage = usage; - buffer->base.size = size; - buffer->flinked = FALSE; - buffer->flink = 0; - buffer->map_gtt = FALSE; - - if (usage & I915_BUFFER_USAGE_SCANOUT) { - /* Scanout buffer */ - name = "gallium3d_scanout"; - pool = dev->pools.gem; - } else if (usage & (PIPE_BUFFER_USAGE_VERTEX | PIPE_BUFFER_USAGE_CONSTANT)) { - /* Local buffer */ - name = "gallium3d_local"; - pool = dev->pools.gem; - } else if (usage & PIPE_BUFFER_USAGE_CUSTOM) { - /* For vertex buffers */ - name = "gallium3d_internal_vertex"; - pool = dev->pools.gem; - } else { - /* Regular buffers */ - name = "gallium3d_regular"; - pool = dev->pools.gem; - } - - buffer->bo = drm_intel_bo_alloc(pool, name, size, alignment); - if (usage & I915_BUFFER_USAGE_SCANOUT) { - unsigned tiling = I915_TILING_X; - unsigned stride = 2048 * 4; /* TODO do something smarter here */ - drm_intel_bo_set_tiling(buffer->bo, &tiling, stride); - buffer->map_gtt = TRUE; - } - - if (!buffer->bo) - goto err; - - return &buffer->base; - -err: - free(buffer); - return NULL; -} - -static struct pipe_buffer * -intel_be_user_buffer_create(struct pipe_winsys *winsys, void *ptr, unsigned bytes) -{ - struct intel_be_buffer *buffer = CALLOC_STRUCT(intel_be_buffer); - struct intel_be_device *dev = intel_be_device(winsys); - int ret; - - if (!buffer) - return NULL; - - pipe_reference_init(&buffer->base.reference, 1); - buffer->base.alignment = 0; - buffer->base.usage = 0; - buffer->base.size = bytes; - - buffer->bo = drm_intel_bo_alloc(dev->pools.gem, - "gallium3d_user_buffer", - bytes, 0); - - if (!buffer->bo) - goto err; - - ret = drm_intel_bo_subdata(buffer->bo, - 0, bytes, ptr); - - if (ret) - goto err; - - return &buffer->base; - -err: - free(buffer); - return NULL; -} - -static struct pipe_buffer * -intel_be_surface_buffer_create(struct pipe_winsys *winsys, - unsigned width, unsigned height, - enum pipe_format format, - unsigned usage, - unsigned tex_usage, - unsigned *stride) -{ - struct pipe_format_block block; - unsigned buf_usage = 0; - unsigned buf_stride = 0; - unsigned buf_size = 0; - - pf_get_block(format, &block); - buf_stride = pf_get_stride(&block, width); - buf_stride = align(buf_stride, 64); - - if (tex_usage & PIPE_TEXTURE_USAGE_PRIMARY) { - /* TODO more checks */ - assert(buf_stride <= 2048*4); - assert(height % 8 == 0); - buf_stride = 2048 * 4; - buf_usage |= I915_BUFFER_USAGE_SCANOUT; - } - - buf_size = buf_stride * height; - *stride = buf_stride; - - return intel_be_buffer_create(winsys, - 0, - buf_usage, - buf_size); -} - -static struct pipe_buffer * -intel_be_buffer_from_handle(struct drm_api *api, - struct pipe_screen *screen, - const char* name, unsigned handle) -{ - struct intel_be_device *dev = intel_be_device(screen->winsys); - struct intel_be_buffer *buffer = CALLOC_STRUCT(intel_be_buffer); - - if (!buffer) - return NULL; - - buffer->bo = drm_intel_bo_gem_create_from_name(dev->pools.gem, name, handle); - - if (!buffer->bo) - goto err; - - pipe_reference_init(&buffer->base.reference, 1); - buffer->base.screen = screen; - buffer->base.alignment = buffer->bo->align; - buffer->base.usage = PIPE_BUFFER_USAGE_GPU_READ | - PIPE_BUFFER_USAGE_GPU_WRITE | - PIPE_BUFFER_USAGE_CPU_READ | - PIPE_BUFFER_USAGE_CPU_WRITE; - buffer->base.size = buffer->bo->size; - - return &buffer->base; - -err: - free(buffer); - return NULL; -} - -struct pipe_texture * -intel_be_texture_from_shared_handle(struct drm_api *api, - struct pipe_screen *screen, - struct pipe_texture *templ, - const char* name, - unsigned pitch, - unsigned handle) -{ - struct pipe_buffer *buffer; - - buffer = intel_be_buffer_from_handle(api, - screen, - name, - handle); - if (!buffer) - return NULL; - - return screen->texture_blanket(screen, templ, &pitch, buffer); -} - -static boolean -intel_be_get_texture_buffer(struct drm_api *api, - struct pipe_texture *texture, - struct pipe_buffer **buffer, - unsigned *stride) -{ - struct intel_be_device *dev; - - if (!texture) - return FALSE; - - dev = intel_be_device(texture->screen->winsys); - if (dev->softpipe) - return softpipe_get_texture_buffer(texture, buffer, stride); - else - return i915_get_texture_buffer(texture, buffer, stride); -} - -boolean -intel_be_shared_handle_from_texture(struct drm_api *api, - struct pipe_screen *screen, - struct pipe_texture *texture, - unsigned *pitch, - unsigned *handle) -{ - struct pipe_buffer *buffer = NULL; - struct intel_be_buffer *buf; - if (!intel_be_get_texture_buffer(api, - texture, - &buffer, - pitch)) - return FALSE; - - buf = intel_be_buffer(buffer); - if (!buf->flinked) { - if (drm_intel_bo_flink(intel_bo(buffer), &buf->flink)) - return FALSE; - buf->flinked = TRUE; - } - - *handle = buf->flink; - - pipe_buffer_reference(&buffer, NULL); - - return TRUE; -} - -boolean -intel_be_local_handle_from_texture(struct drm_api *api, - struct pipe_screen *screen, - struct pipe_texture *texture, - unsigned *pitch, - unsigned *handle) -{ - struct pipe_buffer *buffer = NULL; - if (!intel_be_get_texture_buffer(api, - texture, - &buffer, - pitch)) - return FALSE; - - *handle = intel_bo(buffer)->handle; - - pipe_buffer_reference(&buffer, NULL); - - return TRUE; -} - -/* - * Fence - */ - -static void -intel_be_fence_refunref(struct pipe_winsys *sws, - struct pipe_fence_handle **ptr, - struct pipe_fence_handle *fence) -{ - struct intel_be_fence **p = (struct intel_be_fence **)ptr; - struct intel_be_fence *f = (struct intel_be_fence *)fence; - - intel_be_fence_reference(p, f); -} - -static int -intel_be_fence_signalled(struct pipe_winsys *sws, - struct pipe_fence_handle *fence, - unsigned flag) -{ - assert(0); - - return 0; -} - -static int -intel_be_fence_finish(struct pipe_winsys *sws, - struct pipe_fence_handle *fence, - unsigned flag) -{ - struct intel_be_fence *f = (struct intel_be_fence *)fence; - - /* fence already expired */ - if (!f->bo) - return 0; - - drm_intel_bo_wait_rendering(f->bo); - drm_intel_bo_unreference(f->bo); - f->bo = NULL; - - return 0; -} - -/* - * Misc functions - */ - -static void -intel_be_destroy_winsys(struct pipe_winsys *winsys) -{ - struct intel_be_device *dev = intel_be_device(winsys); - - drm_intel_bufmgr_destroy(dev->pools.gem); - - free(dev); -} - -boolean -intel_be_init_device(struct intel_be_device *dev, int fd, unsigned id) -{ - dev->fd = fd; - dev->id = id; - dev->max_batch_size = 16 * 4096; - dev->max_vertex_size = 128 * 4096; - - dev->base.buffer_create = intel_be_buffer_create; - dev->base.user_buffer_create = intel_be_user_buffer_create; - dev->base.buffer_map = intel_be_buffer_map; - dev->base.buffer_unmap = intel_be_buffer_unmap; - dev->base.buffer_destroy = intel_be_buffer_destroy; - - /* Used by softpipe */ - dev->base.surface_buffer_create = intel_be_surface_buffer_create; - - dev->base.fence_reference = intel_be_fence_refunref; - dev->base.fence_signalled = intel_be_fence_signalled; - dev->base.fence_finish = intel_be_fence_finish; - - dev->base.destroy = intel_be_destroy_winsys; - - dev->pools.gem = drm_intel_bufmgr_gem_init(dev->fd, dev->max_batch_size); - - dev->softpipe = debug_get_bool_option("INTEL_SOFTPIPE", FALSE); - dev->dump_cmd = debug_get_bool_option("INTEL_DUMP_CMD", FALSE); - - return true; -} - -static void -intel_be_get_device_id(unsigned int *device_id) -{ - char path[512]; - FILE *file; - void *shutup_gcc; - - /* - * FIXME: Fix this up to use a drm ioctl or whatever. - */ - - snprintf(path, sizeof(path), "/sys/class/drm/card0/device/device"); - file = fopen(path, "r"); - if (!file) { - return; - } - - shutup_gcc = fgets(path, sizeof(path), file); - sscanf(path, "%x", device_id); - fclose(file); -} - -struct pipe_screen * -intel_be_create_screen(struct drm_api *api, int drmFD, - struct drm_create_screen_arg *arg) -{ - struct intel_be_device *dev; - struct pipe_screen *screen; - unsigned int deviceID; - - if (arg != NULL) { - switch(arg->mode) { - case DRM_CREATE_NORMAL: - break; - default: - return NULL; - } - } - - /* Allocate the private area */ - dev = malloc(sizeof(*dev)); - if (!dev) - return NULL; - memset(dev, 0, sizeof(*dev)); - - intel_be_get_device_id(&deviceID); - intel_be_init_device(dev, drmFD, deviceID); - - if (dev->softpipe) { - screen = softpipe_create_screen(&dev->base); - } else - screen = i915_create_screen(&dev->base, deviceID); - - return screen; -} diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_device.h b/src/gallium/winsys/drm/intel/gem/intel_be_device.h deleted file mode 100644 index 15916b0b102..00000000000 --- a/src/gallium/winsys/drm/intel/gem/intel_be_device.h +++ /dev/null @@ -1,107 +0,0 @@ - -#ifndef INTEL_DRM_DEVICE_H -#define INTEL_DRM_DEVICE_H - -#include "pipe/internal/p_winsys_screen.h" -#include "pipe/p_context.h" - -#include "drm.h" -#include "intel_bufmgr.h" - -struct drm_api; - -/* - * Device - */ - -struct intel_be_device -{ - struct pipe_winsys base; - - boolean softpipe; - boolean dump_cmd; - - int fd; /**< Drm file discriptor */ - - unsigned id; - - size_t max_batch_size; - size_t max_vertex_size; - - struct { - drm_intel_bufmgr *gem; - } pools; -}; - -static INLINE struct intel_be_device * -intel_be_device(struct pipe_winsys *winsys) -{ - return (struct intel_be_device *)winsys; -} - -boolean -intel_be_init_device(struct intel_be_device *device, int fd, unsigned id); - -/* - * Buffer - */ - -struct intel_be_buffer { - struct pipe_buffer base; - void *ptr; - unsigned map_count; - boolean map_gtt; - - drm_intel_bo *bo; - boolean flinked; - unsigned flink; -}; - -/** - * Create a texture from a shared drm handle. - */ -struct pipe_texture * -intel_be_texture_from_shared_handle(struct drm_api *api, - struct pipe_screen *screen, - struct pipe_texture *templ, - const char* name, - unsigned pitch, - unsigned handle); - -/** - * Gets a shared handle from a texture. - * - * If texture is destroyed handle may become invalid. - */ -boolean -intel_be_shared_handle_from_texture(struct drm_api *api, - struct pipe_screen *screen, - struct pipe_texture *texture, - unsigned *pitch, - unsigned *handle); - -/** - * Gets the local handle from a texture. As used by KMS. - * - * If texture is destroyed handle may become invalid. - */ -boolean -intel_be_local_handle_from_texture(struct drm_api *api, - struct pipe_screen *screen, - struct pipe_texture *texture, - unsigned *pitch, - unsigned *handle); - -static INLINE struct intel_be_buffer * -intel_be_buffer(struct pipe_buffer *buf) -{ - return (struct intel_be_buffer *)buf; -} - -static INLINE drm_intel_bo * -intel_bo(struct pipe_buffer *buf) -{ - return intel_be_buffer(buf)->bo; -} - -#endif diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_fence.h b/src/gallium/winsys/drm/intel/gem/intel_be_fence.h deleted file mode 100644 index a8abb01a9ed..00000000000 --- a/src/gallium/winsys/drm/intel/gem/intel_be_fence.h +++ /dev/null @@ -1,34 +0,0 @@ - -#ifndef INTEL_BE_FENCE_H -#define INTEL_BE_FENCE_H - -#include "pipe/p_defines.h" - -#include "drm.h" -#include "intel_bufmgr.h" - -/** - * Because gem does not have fence's we have to create our own fences. - * - * They work by keeping the batchbuffer around and checking if that has - * been idled. If bo is NULL fence has expired. - */ -struct intel_be_fence -{ - struct pipe_reference reference; - drm_intel_bo *bo; -}; - -static INLINE void -intel_be_fence_reference(struct intel_be_fence **ptr, struct intel_be_fence *f) -{ - struct intel_be_fence *old_fence = *ptr; - - if (pipe_reference((struct pipe_reference**)ptr, &f->reference)) { - if (old_fence->bo) - drm_intel_bo_unreference(old_fence->bo); - free(old_fence); - } -} - -#endif diff --git a/src/gallium/winsys/drm/intel/gem/intel_drm_api.c b/src/gallium/winsys/drm/intel/gem/intel_drm_api.c new file mode 100644 index 00000000000..4c5a1d2ea89 --- /dev/null +++ b/src/gallium/winsys/drm/intel/gem/intel_drm_api.c @@ -0,0 +1,202 @@ + +#include "state_tracker/drm_api.h" + +#include "intel_drm_winsys.h" +#include "util/u_memory.h" + +#include "i915simple/i915_context.h" +#include "i915simple/i915_screen.h" + + +/* + * Helper functions + */ + + +static void +intel_drm_get_device_id(unsigned int *device_id) +{ + char path[512]; + FILE *file; + void *shutup_gcc; + + /* + * FIXME: Fix this up to use a drm ioctl or whatever. + */ + + snprintf(path, sizeof(path), "/sys/class/drm/card0/device/device"); + file = fopen(path, "r"); + if (!file) { + return; + } + + shutup_gcc = fgets(path, sizeof(path), file); + sscanf(path, "%x", device_id); + fclose(file); +} + +static struct intel_buffer * +intel_drm_buffer_from_handle(struct intel_drm_winsys *idws, + const char* name, unsigned handle) +{ + struct intel_drm_buffer *buf = CALLOC_STRUCT(intel_drm_buffer); + + if (!buf) + return NULL; + + buf->magic = 0xDEAD1337; + buf->bo = drm_intel_bo_gem_create_from_name(idws->pools.gem, name, handle); + buf->flinked = TRUE; + buf->flink = handle; + + if (!buf->bo) + goto err; + + return (struct intel_buffer *)buf; + +err: + FREE(buf); + return NULL; +} + + +/* + * Exported functions + */ + + +static struct pipe_texture * +intel_drm_texture_from_shared_handle(struct drm_api *api, + struct pipe_screen *screen, + struct pipe_texture *templ, + const char* name, + unsigned pitch, + unsigned handle) +{ + struct intel_drm_winsys *idws = intel_drm_winsys(i915_screen(screen)->iws); + struct intel_buffer *buffer; + + buffer = intel_drm_buffer_from_handle(idws, name, handle); + if (!buffer) + return NULL; + + return i915_texture_blanket_intel(screen, templ, pitch, buffer); +} + +static boolean +intel_drm_shared_handle_from_texture(struct drm_api *api, + struct pipe_screen *screen, + struct pipe_texture *texture, + unsigned *pitch, + unsigned *handle) +{ + struct intel_drm_buffer *buf = NULL; + struct intel_buffer *buffer = NULL; + if (!i915_get_texture_buffer_intel(texture, &buffer, pitch)) + return FALSE; + + buf = intel_drm_buffer(buffer); + if (!buf->flinked) { + if (drm_intel_bo_flink(buf->bo, &buf->flink)) + return FALSE; + buf->flinked = TRUE; + } + + *handle = buf->flink; + + return TRUE; +} + +static boolean +intel_drm_local_handle_from_texture(struct drm_api *api, + struct pipe_screen *screen, + struct pipe_texture *texture, + unsigned *pitch, + unsigned *handle) +{ + struct intel_buffer *buffer = NULL; + if (!i915_get_texture_buffer_intel(texture, &buffer, pitch)) + return FALSE; + + *handle = intel_drm_buffer(buffer)->bo->handle; + + return TRUE; +} + +static void +intel_drm_winsys_destroy(struct intel_winsys *iws) +{ + struct intel_drm_winsys *idws = intel_drm_winsys(iws); + + drm_intel_bufmgr_destroy(idws->pools.gem); + + FREE(idws); +} + +static struct pipe_screen * +intel_drm_create_screen(struct drm_api *api, int drmFD, + struct drm_create_screen_arg *arg) +{ + struct intel_drm_winsys *idws; + unsigned int deviceID; + + if (arg != NULL) { + switch(arg->mode) { + case DRM_CREATE_NORMAL: + break; + default: + return NULL; + } + } + + idws = CALLOC_STRUCT(intel_drm_winsys); + if (!idws) + return NULL; + + intel_drm_get_device_id(&deviceID); + + intel_drm_winsys_init_batchbuffer_functions(idws); + intel_drm_winsys_init_buffer_functions(idws); + intel_drm_winsys_init_fence_functions(idws); + + idws->fd = drmFD; + idws->id = deviceID; + idws->max_batch_size = 16 * 4096; + + idws->base.destroy = intel_drm_winsys_destroy; + + idws->pools.gem = drm_intel_bufmgr_gem_init(idws->fd, idws->max_batch_size); + + idws->softpipe = FALSE; + idws->dump_cmd = debug_get_bool_option("INTEL_DUMP_CMD", FALSE); + + return i915_create_screen(&idws->base, deviceID); +} + +static struct pipe_context * +intel_drm_create_context(struct drm_api *api, struct pipe_screen *screen) +{ + return i915_create_context(screen); +} + +static void +destroy(struct drm_api *api) +{ + +} + +struct drm_api intel_drm_api = +{ + .create_context = intel_drm_create_context, + .create_screen = intel_drm_create_screen, + .texture_from_shared_handle = intel_drm_texture_from_shared_handle, + .shared_handle_from_texture = intel_drm_shared_handle_from_texture, + .local_handle_from_texture = intel_drm_local_handle_from_texture, + .destroy = destroy, +}; + +struct drm_api * +drm_api_create() +{ + return &intel_drm_api; +} diff --git a/src/gallium/winsys/drm/intel/gem/intel_drm_batchbuffer.c b/src/gallium/winsys/drm/intel/gem/intel_drm_batchbuffer.c new file mode 100644 index 00000000000..5ca3ad9762d --- /dev/null +++ b/src/gallium/winsys/drm/intel/gem/intel_drm_batchbuffer.c @@ -0,0 +1,214 @@ + +#include "intel_drm_winsys.h" +#include "util/u_memory.h" + +#include "i915_drm.h" + +#define BATCH_RESERVED 16 + +#define INTEL_DEFAULT_RELOCS 100 +#define INTEL_MAX_RELOCS 400 + +#define INTEL_BATCH_NO_CLIPRECTS 0x1 +#define INTEL_BATCH_CLIPRECTS 0x2 + +#undef INTEL_RUN_SYNC + +struct intel_drm_batchbuffer +{ + struct intel_batchbuffer base; + + size_t actual_size; + + drm_intel_bo *bo; +}; + +static INLINE struct intel_drm_batchbuffer * +intel_drm_batchbuffer(struct intel_batchbuffer *batch) +{ + return (struct intel_drm_batchbuffer *)batch; +} + +static void +intel_drm_batchbuffer_reset(struct intel_drm_batchbuffer *batch) +{ + struct intel_drm_winsys *idws = intel_drm_winsys(batch->base.iws); + + if (batch->bo) + drm_intel_bo_unreference(batch->bo); + batch->bo = drm_intel_bo_alloc(idws->pools.gem, + "gallium3d_batchbuffer", + batch->actual_size, + 4096); + drm_intel_bo_map(batch->bo, TRUE); + batch->base.map = batch->bo->virtual; + + memset(batch->base.map, 0, batch->actual_size); + batch->base.ptr = batch->base.map; + batch->base.size = batch->actual_size - BATCH_RESERVED; + batch->base.relocs = 0; +} + +static struct intel_batchbuffer * +intel_drm_batchbuffer_create(struct intel_winsys *iws) +{ + struct intel_drm_winsys *idws = intel_drm_winsys(iws); + struct intel_drm_batchbuffer *batch = CALLOC_STRUCT(intel_drm_batchbuffer); + + batch->base.map = NULL; + batch->base.ptr = NULL; + batch->base.size = 0; + + batch->base.relocs = 0; + batch->base.max_relocs = 300;/*INTEL_DEFAULT_RELOCS;*/ + + batch->base.iws = iws; + + batch->actual_size = idws->max_batch_size; + + intel_drm_batchbuffer_reset(batch); + + return &batch->base; +} + +static int +intel_drm_batchbuffer_reloc(struct intel_batchbuffer *ibatch, + struct intel_buffer *buffer, + enum intel_buffer_usage usage, + unsigned pre_add) +{ + struct intel_drm_batchbuffer *batch = intel_drm_batchbuffer(ibatch); + unsigned write_domain = 0; + unsigned read_domain = 0; + unsigned offset; + int ret = 0; + + assert(batch->base.relocs < batch->base.max_relocs); + + if (usage == INTEL_USAGE_SAMPLER) { + write_domain = 0; + read_domain = I915_GEM_DOMAIN_SAMPLER; + + } else if (usage == INTEL_USAGE_RENDER) { + write_domain = I915_GEM_DOMAIN_RENDER; + read_domain = I915_GEM_DOMAIN_RENDER; + + } else if (usage == INTEL_USAGE_2D_TARGET) { + write_domain = I915_GEM_DOMAIN_RENDER; + read_domain = I915_GEM_DOMAIN_RENDER; + + } else if (usage == INTEL_USAGE_2D_SOURCE) { + write_domain = 0; + read_domain = I915_GEM_DOMAIN_RENDER; + + } else if (usage == INTEL_USAGE_VERTEX) { + write_domain = 0; + read_domain = I915_GEM_DOMAIN_VERTEX; + + } else { + assert(0); + return -1; + } + + offset = (unsigned)(batch->base.ptr - batch->base.map); + + ret = drm_intel_bo_emit_reloc(batch->bo, offset, + intel_bo(buffer), pre_add, + read_domain, + write_domain); + + ((uint32_t*)batch->base.ptr)[0] = intel_bo(buffer)->offset + pre_add; + batch->base.ptr += 4; + + if (!ret) + batch->base.relocs++; + + return ret; +} + +static void +intel_drm_batchbuffer_flush(struct intel_batchbuffer *ibatch, + struct pipe_fence_handle **fence) +{ + struct intel_drm_batchbuffer *batch = intel_drm_batchbuffer(ibatch); + unsigned used = 0; + int ret = 0; + int i; + + assert(intel_batchbuffer_space(ibatch) >= 0); + + used = batch->base.ptr - batch->base.map; + assert((used & 3) == 0); + + if (used & 4) { + // MI_FLUSH | FLUSH_MAP_CACHE; + intel_batchbuffer_dword(ibatch, (0x0<<29)|(0x4<<23)|(1<<0)); + // MI_NOOP + intel_batchbuffer_dword(ibatch, (0x0<<29)|(0x0<<23)); + // MI_BATCH_BUFFER_END; + intel_batchbuffer_dword(ibatch, (0x0<<29)|(0xA<<23)); + } else { + //MI_FLUSH | FLUSH_MAP_CACHE; + intel_batchbuffer_dword(ibatch, (0x0<<29)|(0x4<<23)|(1<<0)); + // MI_BATCH_BUFFER_END; + intel_batchbuffer_dword(ibatch, (0x0<<29)|(0xA<<23)); + } + + used = batch->base.ptr - batch->base.map; + + drm_intel_bo_unmap(batch->bo); + + /* Do the sending to HW */ + ret = drm_intel_bo_exec(batch->bo, used, NULL, 0, 0); + assert(ret == 0); + + if (intel_drm_winsys(ibatch->iws)->dump_cmd) { + unsigned *ptr; + drm_intel_bo_map(batch->bo, FALSE); + ptr = (unsigned*)batch->bo->virtual; + + debug_printf("%s:\n", __func__); + for (i = 0; i < used / 4; i++, ptr++) { + debug_printf("\t%08x: %08x\n", i*4, *ptr); + } + + drm_intel_bo_unmap(batch->bo); + } else { +#ifdef INTEL_RUN_SYNC + drm_intel_bo_map(batch->bo, FALSE); + drm_intel_bo_unmap(batch->bo); +#endif + } + + if (fence) { + ibatch->iws->fence_reference(ibatch->iws, fence, NULL); + +#ifdef INTEL_RUN_SYNC + /* we run synced to GPU so just pass null */ + (*fence) = intel_drm_fence_create(NULL); +#else + (*fence) = intel_drm_fence_create(batch->bo); +#endif + } + + intel_drm_batchbuffer_reset(batch); +} + +static void +intel_drm_batchbuffer_destroy(struct intel_batchbuffer *ibatch) +{ + struct intel_drm_batchbuffer *batch = intel_drm_batchbuffer(ibatch); + + if (batch->bo) + drm_intel_bo_unreference(batch->bo); + + free(batch); +} + +void intel_drm_winsys_init_batchbuffer_functions(struct intel_drm_winsys *idws) +{ + idws->base.batchbuffer_create = intel_drm_batchbuffer_create; + idws->base.batchbuffer_reloc = intel_drm_batchbuffer_reloc; + idws->base.batchbuffer_flush = intel_drm_batchbuffer_flush; + idws->base.batchbuffer_destroy = intel_drm_batchbuffer_destroy; +} diff --git a/src/gallium/winsys/drm/intel/gem/intel_drm_buffer.c b/src/gallium/winsys/drm/intel/gem/intel_drm_buffer.c new file mode 100644 index 00000000000..e017cd2e982 --- /dev/null +++ b/src/gallium/winsys/drm/intel/gem/intel_drm_buffer.c @@ -0,0 +1,134 @@ + +#include "intel_drm_winsys.h" +#include "util/u_memory.h" + +#include "i915_drm.h" + +static struct intel_buffer * +intel_drm_buffer_create(struct intel_winsys *iws, + unsigned size, unsigned alignment, + enum intel_buffer_type type) +{ + struct intel_drm_buffer *buf = CALLOC_STRUCT(intel_drm_buffer); + struct intel_drm_winsys *idws = intel_drm_winsys(iws); + drm_intel_bufmgr *pool; + char *name; + + if (!buf) + return NULL; + + buf->magic = 0xDEAD1337; + buf->flinked = FALSE; + buf->flink = 0; + buf->map_gtt = FALSE; + + if (type == INTEL_NEW_TEXTURE) { + name = "gallium3d_texture"; + pool = idws->pools.gem; + } else if (type == INTEL_NEW_VERTEX) { + name = "gallium3d_vertex"; + pool = idws->pools.gem; + } else if (type == INTEL_NEW_SCANOUT) { + name = "gallium3d_scanout"; + pool = idws->pools.gem; + buf->map_gtt = TRUE; + } else { + assert(0); + name = "gallium3d_unknown"; + pool = idws->pools.gem; + } + + buf->bo = drm_intel_bo_alloc(pool, name, size, alignment); + + if (!buf->bo) + goto err; + + return (struct intel_buffer *)buf; + +err: + assert(0); + FREE(buf); + return NULL; +} + +static int +intel_drm_buffer_set_fence_reg(struct intel_winsys *iws, + struct intel_buffer *buffer, + unsigned stride, + enum intel_buffer_tile tile) +{ + assert(I915_TILING_NONE == INTEL_TILE_NONE); + assert(I915_TILING_X == INTEL_TILE_X); + assert(I915_TILING_Y == INTEL_TILE_Y); + + return drm_intel_bo_set_tiling(intel_bo(buffer), &tile, stride); +} + +static void * +intel_drm_buffer_map(struct intel_winsys *iws, + struct intel_buffer *buffer, + boolean write) +{ + struct intel_drm_buffer *buf = intel_drm_buffer(buffer); + drm_intel_bo *bo = intel_bo(buffer); + int ret = 0; + + assert(bo); + + if (buf->map_count) + goto out; + + if (buf->map_gtt) + ret = drm_intel_gem_bo_map_gtt(bo); + else + ret = drm_intel_bo_map(bo, write); + + buf->ptr = bo->virtual; + + assert(ret == 0); +out: + if (ret) + return NULL; + + buf->map_count++; + return buf->ptr; +} + +static void +intel_drm_buffer_unmap(struct intel_winsys *iws, + struct intel_buffer *buffer) +{ + struct intel_drm_buffer *buf = intel_drm_buffer(buffer); + + if (--buf->map_count) + return; + + if (buf->map_gtt) + drm_intel_gem_bo_unmap_gtt(intel_bo(buffer)); + else + drm_intel_bo_unmap(intel_bo(buffer)); +} + +static void +intel_drm_buffer_destroy(struct intel_winsys *iws, + struct intel_buffer *buffer) +{ + drm_intel_bo_unreference(intel_bo(buffer)); + +#ifdef DEBUG + intel_drm_buffer(buffer)->magic = 0; + intel_drm_buffer(buffer)->bo = NULL; +#endif + + FREE(buffer); +} + +void +intel_drm_winsys_init_buffer_functions(struct intel_drm_winsys *idws) +{ + idws->base.buffer_create = intel_drm_buffer_create; + idws->base.buffer_set_fence_reg = intel_drm_buffer_set_fence_reg; + idws->base.buffer_map = intel_drm_buffer_map; + idws->base.buffer_unmap = intel_drm_buffer_unmap; + idws->base.buffer_destroy = intel_drm_buffer_destroy; +} diff --git a/src/gallium/winsys/drm/intel/gem/intel_drm_fence.c b/src/gallium/winsys/drm/intel/gem/intel_drm_fence.c new file mode 100644 index 00000000000..e70bfe7b44e --- /dev/null +++ b/src/gallium/winsys/drm/intel/gem/intel_drm_fence.c @@ -0,0 +1,81 @@ + +#include "intel_drm_winsys.h" +#include "util/u_memory.h" +#include "pipe/p_refcnt.h" + +/** + * Because gem does not have fence's we have to create our own fences. + * + * They work by keeping the batchbuffer around and checking if that has + * been idled. If bo is NULL fence has expired. + */ +struct intel_drm_fence +{ + struct pipe_reference reference; + drm_intel_bo *bo; +}; + + +struct pipe_fence_handle * +intel_drm_fence_create(drm_intel_bo *bo) +{ + struct intel_drm_fence *fence = CALLOC_STRUCT(intel_drm_fence); + + pipe_reference_init(&fence->reference, 1); + /* bo is null if fence already expired */ + if (bo) { + drm_intel_bo_reference(bo); + fence->bo = bo; + } + + return (struct pipe_fence_handle *)fence; +} + +static void +intel_drm_fence_reference(struct intel_winsys *iws, + struct pipe_fence_handle **ptr, + struct pipe_fence_handle *fence) +{ + struct intel_drm_fence *old = (struct intel_drm_fence *)*ptr; + struct intel_drm_fence *f = (struct intel_drm_fence *)fence; + + if (pipe_reference((struct pipe_reference**)ptr, &f->reference)) { + if (old->bo) + drm_intel_bo_unreference(old->bo); + FREE(old); + } +} + +static int +intel_drm_fence_signalled(struct intel_winsys *iws, + struct pipe_fence_handle *fence) +{ + assert(0); + + return 0; +} + +static int +intel_drm_fence_finish(struct intel_winsys *iws, + struct pipe_fence_handle *fence) +{ + struct intel_drm_fence *f = (struct intel_drm_fence *)fence; + + /* fence already expired */ + if (!f->bo) + return 0; + + drm_intel_bo_wait_rendering(f->bo); + drm_intel_bo_unreference(f->bo); + f->bo = NULL; + + return 0; +} + +void +intel_drm_winsys_init_fence_functions(struct intel_drm_winsys *idws) +{ + idws->base.fence_reference = intel_drm_fence_reference; + idws->base.fence_signalled = intel_drm_fence_signalled; + idws->base.fence_finish = intel_drm_fence_finish; +} diff --git a/src/gallium/winsys/drm/intel/gem/intel_drm_winsys.h b/src/gallium/winsys/drm/intel/gem/intel_drm_winsys.h new file mode 100644 index 00000000000..415c45feea0 --- /dev/null +++ b/src/gallium/winsys/drm/intel/gem/intel_drm_winsys.h @@ -0,0 +1,78 @@ + +#ifndef INTEL_DRM_WINSYS_H +#define INTEL_DRM_WINSYS_H + +#include "i915simple/intel_batchbuffer.h" + +#include "drm.h" +#include "intel_bufmgr.h" + + +/* + * Winsys + */ + + +struct intel_drm_winsys +{ + struct intel_winsys base; + + boolean softpipe; + boolean dump_cmd; + + int fd; /**< Drm file discriptor */ + + unsigned id; + + size_t max_batch_size; + + struct { + drm_intel_bufmgr *gem; + } pools; +}; + +static INLINE struct intel_drm_winsys * +intel_drm_winsys(struct intel_winsys *iws) +{ + return (struct intel_drm_winsys *)iws; +} + +struct intel_drm_winsys * intel_drm_winsys_create(int fd, unsigned pci_id); +struct pipe_fence_handle * intel_drm_fence_create(drm_intel_bo *bo); + +void intel_drm_winsys_init_batchbuffer_functions(struct intel_drm_winsys *idws); +void intel_drm_winsys_init_buffer_functions(struct intel_drm_winsys *idws); +void intel_drm_winsys_init_fence_functions(struct intel_drm_winsys *idws); + + +/* + * Buffer + */ + + +struct intel_drm_buffer { + unsigned magic; + + drm_intel_bo *bo; + + void *ptr; + unsigned map_count; + boolean map_gtt; + + boolean flinked; + unsigned flink; +}; + +static INLINE struct intel_drm_buffer * +intel_drm_buffer(struct intel_buffer *buffer) +{ + return (struct intel_drm_buffer *)buffer; +} + +static INLINE drm_intel_bo * +intel_bo(struct intel_buffer *buffer) +{ + return intel_drm_buffer(buffer)->bo; +} + +#endif diff --git a/src/gallium/winsys/drm/intel/xorg/intel_xorg.c b/src/gallium/winsys/drm/intel/xorg/intel_xorg.c index 28107f4b806..369dc356cf8 100644 --- a/src/gallium/winsys/drm/intel/xorg/intel_xorg.c +++ b/src/gallium/winsys/drm/intel/xorg/intel_xorg.c @@ -47,8 +47,8 @@ static SymTabRec intel_xorg_chipsets[] = { }; static PciChipsets intel_xorg_pci_devices[] = { - {PCI_MATCH_ANY, PCI_MATCH_ANY, RES_SHARED_VGA}, - {-1, -1, RES_UNDEFINED} + {PCI_MATCH_ANY, PCI_MATCH_ANY, NULL}, + {-1, -1, NULL} }; static XF86ModuleVersionInfo intel_xorg_version = { @@ -105,12 +105,6 @@ intel_xorg_setup(pointer module, pointer opts, int *errmaj, int *errmin) xf86AddDriver(&modesetting, module, HaveDriverFuncs); /* - * Tell the loader about symbols from other modules that this module - * might refer to. - */ - xorg_tracker_loader_ref_sym_lists(); - - /* * The return value must be non-NULL on success even though there * is no TearDownProc. */ diff --git a/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c index 3167ccfdc16..091cbbcfed1 100644 --- a/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c +++ b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c @@ -12,51 +12,37 @@ #include "nouveau/nouveau_screen.h" static struct pipe_surface * -dri_surface_from_handle(struct drm_api *api, struct pipe_screen *screen, - unsigned handle, - enum pipe_format format, - unsigned width, - unsigned height, - unsigned pitch) +dri_surface_from_handle(struct drm_api *api, struct pipe_screen *pscreen, + unsigned handle, enum pipe_format format, + unsigned width, unsigned height, unsigned pitch) { - struct pipe_surface *surface = NULL; - struct pipe_texture *texture = NULL; - struct pipe_texture templat; - struct pipe_buffer *buf = NULL; - - buf = api->buffer_from_handle(api, screen, "front buffer", handle); - if (!buf) - return NULL; - - memset(&templat, 0, sizeof(templat)); - templat.tex_usage = PIPE_TEXTURE_USAGE_PRIMARY | - NOUVEAU_TEXTURE_USAGE_LINEAR; - templat.target = PIPE_TEXTURE_2D; - templat.last_level = 0; - templat.depth[0] = 1; - templat.format = format; - templat.width[0] = width; - templat.height[0] = height; - pf_get_block(templat.format, &templat.block); - - texture = screen->texture_blanket(screen, - &templat, - &pitch, - buf); - - /* we don't need the buffer from this point on */ - pipe_buffer_reference(&buf, NULL); - - if (!texture) - return NULL; - - surface = screen->get_tex_surface(screen, texture, 0, 0, 0, - PIPE_BUFFER_USAGE_GPU_READ | - PIPE_BUFFER_USAGE_GPU_WRITE); - - /* we don't need the texture from this point on */ - pipe_texture_reference(&texture, NULL); - return surface; + struct pipe_surface *ps = NULL; + struct pipe_texture *pt = NULL; + struct pipe_texture tmpl; + + memset(&tmpl, 0, sizeof(tmpl)); + tmpl.tex_usage = PIPE_TEXTURE_USAGE_PRIMARY | + NOUVEAU_TEXTURE_USAGE_LINEAR; + tmpl.target = PIPE_TEXTURE_2D; + tmpl.last_level = 0; + tmpl.depth[0] = 1; + tmpl.format = format; + tmpl.width[0] = width; + tmpl.height[0] = height; + pf_get_block(tmpl.format, &tmpl.block); + + pt = api->texture_from_shared_handle(api, pscreen, &tmpl, + "front buffer", pitch, handle); + if (!pt) + return NULL; + + ps = pscreen->get_tex_surface(pscreen, pt, 0, 0, 0, + PIPE_BUFFER_USAGE_GPU_READ | + PIPE_BUFFER_USAGE_GPU_WRITE); + + /* we don't need the texture from this point on */ + pipe_texture_reference(&pt, NULL); + return ps; } static struct pipe_surface * @@ -205,16 +191,10 @@ nouveau_drm_create_context(struct drm_api *api, struct pipe_screen *pscreen) return nvws->pctx[i]; } -static boolean -nouveau_drm_pb_from_pt(struct drm_api *api, struct pipe_texture *pt, - struct pipe_buffer **ppb, unsigned *stride) -{ - return false; -} - -static struct pipe_buffer * -nouveau_drm_pb_from_handle(struct drm_api *api, struct pipe_screen *pscreen, - const char *name, unsigned handle) +static struct pipe_texture * +nouveau_drm_pt_from_name(struct drm_api *api, struct pipe_screen *pscreen, + struct pipe_texture *templ, const char *name, + unsigned stride, unsigned handle) { struct nouveau_device *dev = nouveau_screen(pscreen)->device; struct pipe_buffer *pb; @@ -238,41 +218,42 @@ nouveau_drm_pb_from_handle(struct drm_api *api, struct pipe_screen *pscreen, pb->usage = PIPE_BUFFER_USAGE_GPU_READ_WRITE | PIPE_BUFFER_USAGE_CPU_READ_WRITE; pb->size = nouveau_bo(pb)->size; - return pb; + return pscreen->texture_blanket(pscreen, templ, &stride, pb); } static boolean -nouveau_drm_handle_from_pb(struct drm_api *api, struct pipe_screen *pscreen, - struct pipe_buffer *pb, unsigned *handle) +nouveau_drm_name_from_pt(struct drm_api *api, struct pipe_screen *pscreen, + struct pipe_texture *pt, unsigned *stride, + unsigned *handle) { - struct nouveau_bo *bo = nouveau_bo(pb); + struct nouveau_miptree *mt = nouveau_miptree(pt); - if (!bo) - return FALSE; + if (!mt || !mt->bo) + return false; - *handle = bo->handle; - return TRUE; + return nouveau_bo_handle_get(mt->bo, handle) == 0; } static boolean -nouveau_drm_name_from_pb(struct drm_api *api, struct pipe_screen *pscreen, - struct pipe_buffer *pb, unsigned *handle) +nouveau_drm_handle_from_pt(struct drm_api *api, struct pipe_screen *pscreen, + struct pipe_texture *pt, unsigned *stride, + unsigned *handle) { - struct nouveau_bo *bo = nouveau_bo(pb); + struct nouveau_miptree *mt = nouveau_miptree(pt); - if (!bo) - return FALSE; + if (!mt || !mt->bo) + return false; - return nouveau_bo_handle_get(bo, handle) == 0; + *handle = mt->bo->handle; + return true; } struct drm_api drm_api_hooks = { .create_screen = nouveau_drm_create_screen, .create_context = nouveau_drm_create_context, - .buffer_from_texture = nouveau_drm_pb_from_pt, - .buffer_from_handle = nouveau_drm_pb_from_handle, - .handle_from_buffer = nouveau_drm_handle_from_pb, - .global_handle_from_buffer = nouveau_drm_name_from_pb, + .texture_from_shared_handle = nouveau_drm_pt_from_name, + .shared_handle_from_texture = nouveau_drm_name_from_pt, + .local_handle_from_texture = nouveau_drm_handle_from_pt, }; struct drm_api * diff --git a/src/gallium/winsys/drm/radeon/xorg/radeon_xorg.c b/src/gallium/winsys/drm/radeon/xorg/radeon_xorg.c index f2c6ee5f9a9..837f2aa8fec 100644 --- a/src/gallium/winsys/drm/radeon/xorg/radeon_xorg.c +++ b/src/gallium/winsys/drm/radeon/xorg/radeon_xorg.c @@ -48,8 +48,8 @@ static SymTabRec radeon_xorg_chipsets[] = { }; static PciChipsets radeon_xorg_pci_devices[] = { - {PCI_MATCH_ANY, PCI_MATCH_ANY, RES_SHARED_VGA}, - {-1, -1, RES_UNDEFINED} + {PCI_MATCH_ANY, PCI_MATCH_ANY, NULL}, + {-1, -1, NULL} }; static XF86ModuleVersionInfo radeon_xorg_version = { @@ -106,12 +106,6 @@ radeon_xorg_setup(pointer module, pointer opts, int *errmaj, int *errmin) xf86AddDriver(&modesetting, module, HaveDriverFuncs); /* - * Tell the loader about symbols from other modules that this module - * might refer to. - */ - xorg_tracker_loader_ref_sym_lists(); - - /* * The return value must be non-NULL on success even though there * is no TearDownProc. */ diff --git a/src/gallium/winsys/xlib/SConscript b/src/gallium/winsys/xlib/SConscript index 518fd2b5a84..467d595d33b 100644 --- a/src/gallium/winsys/xlib/SConscript +++ b/src/gallium/winsys/xlib/SConscript @@ -30,12 +30,12 @@ if env['platform'] == 'linux' \ drivers += [softpipe] if 'llvmpipe' in env['drivers']: - env.Append(CPPDEFINES = 'GALLIUM_LLVMPIPE') - env.Tool('udis86') - env.ParseConfig('llvm-config --libs jit interpreter nativecodegen') - env['LINK'] = env['CXX'] - sources += ['xlib_llvmpipe.c'] - drivers += [llvmpipe] + env.Tool('llvm') + if 'LLVM_VERSION' in env: + env.Append(CPPDEFINES = 'GALLIUM_LLVMPIPE') + env.Tool('udis86') + sources += ['xlib_llvmpipe.c'] + drivers += [llvmpipe] if 'i965simple' in env['drivers']: env.Append(CPPDEFINES = 'GALLIUM_I965SIMPLE') diff --git a/src/gallium/winsys/xlib/xlib_brw_screen.c b/src/gallium/winsys/xlib/xlib_brw_screen.c index 6f3861e2cd6..ef545796f3c 100644 --- a/src/gallium/winsys/xlib/xlib_brw_screen.c +++ b/src/gallium/winsys/xlib/xlib_brw_screen.c @@ -365,7 +365,7 @@ void xlib_brw_buffer_subdata_typed( struct pipe_winsys *pws, unsigned data_type ) { unsigned aub_type = DW_GENERAL_STATE; - unsigned aub_sub_type; + unsigned aub_sub_type = 0; switch (data_type) { case BRW_CC_VP: |