From 7d99ddcb2bb09f1f54d91e6e20e42d217a5bccdf Mon Sep 17 00:00:00 2001 From: Eric Anholt Date: Fri, 26 Sep 2008 12:48:23 -0700 Subject: intel: Fix a number of memory leaks on context destroy. --- src/mesa/drivers/dri/i965/brw_draw.c | 10 +++++++++ src/mesa/drivers/dri/i965/brw_state_cache.c | 5 +++-- src/mesa/drivers/dri/i965/brw_vtbl.c | 34 +++++++++++++++++++++++++++++ 3 files changed, 47 insertions(+), 2 deletions(-) (limited to 'src/mesa/drivers/dri/i965') diff --git a/src/mesa/drivers/dri/i965/brw_draw.c b/src/mesa/drivers/dri/i965/brw_draw.c index 9a353fc7b6..39ce8eb4b6 100644 --- a/src/mesa/drivers/dri/i965/brw_draw.c +++ b/src/mesa/drivers/dri/i965/brw_draw.c @@ -409,8 +409,18 @@ void brw_draw_init( struct brw_context *brw ) void brw_draw_destroy( struct brw_context *brw ) { + int i; + if (brw->vb.upload.bo != NULL) { dri_bo_unreference(brw->vb.upload.bo); brw->vb.upload.bo = NULL; } + + for (i = 0; i < VERT_ATTRIB_MAX; i++) { + dri_bo_unreference(brw->vb.inputs[i].bo); + brw->vb.inputs[i].bo = NULL; + } + + dri_bo_unreference(brw->ib.bo); + brw->ib.bo = NULL; } diff --git a/src/mesa/drivers/dri/i965/brw_state_cache.c b/src/mesa/drivers/dri/i965/brw_state_cache.c index 1318dea594..d5b5166406 100644 --- a/src/mesa/drivers/dri/i965/brw_state_cache.c +++ b/src/mesa/drivers/dri/i965/brw_state_cache.c @@ -497,9 +497,10 @@ void brw_destroy_cache( struct brw_context *brw ) GLuint i; brw_clear_cache(brw); - for (i = 0; i < BRW_MAX_CACHE; i++) + for (i = 0; i < BRW_MAX_CACHE; i++) { + dri_bo_unreference(brw->cache.last_bo[i]); free(brw->cache.name[i]); - + } free(brw->cache.items); brw->cache.items = NULL; brw->cache.size = 0; diff --git a/src/mesa/drivers/dri/i965/brw_vtbl.c b/src/mesa/drivers/dri/i965/brw_vtbl.c index 2a03fc59f3..3780d3dad2 100644 --- a/src/mesa/drivers/dri/i965/brw_vtbl.c +++ b/src/mesa/drivers/dri/i965/brw_vtbl.c @@ -51,6 +51,12 @@ #include "brw_vs.h" #include +static void +dri_bo_release(dri_bo **bo) +{ + dri_bo_unreference(*bo); + *bo = NULL; +} /* called from intelDestroyContext() */ @@ -58,6 +64,7 @@ static void brw_destroy_context( struct intel_context *intel ) { GLcontext *ctx = &intel->ctx; struct brw_context *brw = brw_context(&intel->ctx); + int i; brw_destroy_metaops(brw); brw_destroy_state(brw); @@ -65,6 +72,33 @@ static void brw_destroy_context( struct intel_context *intel ) brw_ProgramCacheDestroy( ctx ); brw_FrameBufferTexDestroy( brw ); + + for (i = 0; i < brw->state.nr_draw_regions; i++) + intel_region_release(&brw->state.draw_regions[i]); + brw->state.nr_draw_regions = 0; + intel_region_release(&brw->state.depth_region); + + dri_bo_release(&brw->curbe.curbe_bo); + dri_bo_release(&brw->vs.prog_bo); + dri_bo_release(&brw->vs.state_bo); + dri_bo_release(&brw->gs.prog_bo); + dri_bo_release(&brw->gs.state_bo); + dri_bo_release(&brw->clip.prog_bo); + dri_bo_release(&brw->clip.state_bo); + dri_bo_release(&brw->clip.vp_bo); + dri_bo_release(&brw->sf.prog_bo); + dri_bo_release(&brw->sf.state_bo); + dri_bo_release(&brw->sf.vp_bo); + for (i = 0; i < BRW_MAX_TEX_UNIT; i++) + dri_bo_release(&brw->wm.sdc_bo[i]); + dri_bo_release(&brw->wm.bind_bo); + for (i = 0; i < BRW_WM_MAX_SURF; i++) + dri_bo_release(&brw->wm.surf_bo[i]); + dri_bo_release(&brw->wm.prog_bo); + dri_bo_release(&brw->wm.state_bo); + dri_bo_release(&brw->cc.prog_bo); + dri_bo_release(&brw->cc.state_bo); + dri_bo_release(&brw->cc.vp_bo); } /* called from intelDrawBuffer() -- cgit v1.2.3 From b5d59222ccbec9db23b6847737765a4dc0d8c47b Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Thu, 25 Sep 2008 17:19:47 -0700 Subject: Remove TNL-to-VP tracking from i965 The i965 driver previously had it's own set of code to convert fixed-function TNL state to a vertex program. Core Mesa has code to do this, so there is no reason to duplicate that effort in the driver. In fact, this duplication leads to bugs when other aspects of the Mesa infrastructure change. --- src/mesa/drivers/dri/i965/brw_context.c | 2 - src/mesa/drivers/dri/i965/brw_context.h | 5 - src/mesa/drivers/dri/i965/brw_state.h | 1 - src/mesa/drivers/dri/i965/brw_state_upload.c | 1 - src/mesa/drivers/dri/i965/brw_vs.h | 4 - src/mesa/drivers/dri/i965/brw_vs_tnl.c | 1653 +------------------------- src/mesa/drivers/dri/i965/brw_vtbl.c | 1 - 7 files changed, 2 insertions(+), 1665 deletions(-) (limited to 'src/mesa/drivers/dri/i965') diff --git a/src/mesa/drivers/dri/i965/brw_context.c b/src/mesa/drivers/dri/i965/brw_context.c index 92629016d9..5f60477176 100644 --- a/src/mesa/drivers/dri/i965/brw_context.c +++ b/src/mesa/drivers/dri/i965/brw_context.c @@ -152,8 +152,6 @@ GLboolean brwCreateContext( const __GLcontextModes *mesaVis, brw_draw_init( brw ); - brw_ProgramCacheInit( ctx ); - return GL_TRUE; } diff --git a/src/mesa/drivers/dri/i965/brw_context.h b/src/mesa/drivers/dri/i965/brw_context.h index b04487ecee..8bddc9da40 100644 --- a/src/mesa/drivers/dri/i965/brw_context.h +++ b/src/mesa/drivers/dri/i965/brw_context.h @@ -130,7 +130,6 @@ struct brw_context; #define BRW_NEW_CONTEXT 0x80 #define BRW_NEW_WM_INPUT_DIMENSIONS 0x100 #define BRW_NEW_INPUT_VARYING 0x200 -#define BRW_NEW_TNL_PROGRAM 0x400 #define BRW_NEW_PSP 0x800 #define BRW_NEW_METAOPS 0x1000 #define BRW_NEW_FENCE 0x2000 @@ -488,10 +487,6 @@ struct brw_context GLboolean active; } metaops; - /* Track fixed function t&l in a vertex program: - */ - struct gl_vertex_program *tnl_program; - struct brw_tnl_cache tnl_program_cache; /* Active vertex program: */ diff --git a/src/mesa/drivers/dri/i965/brw_state.h b/src/mesa/drivers/dri/i965/brw_state.h index 3ea6151ae9..4c04036ef0 100644 --- a/src/mesa/drivers/dri/i965/brw_state.h +++ b/src/mesa/drivers/dri/i965/brw_state.h @@ -74,7 +74,6 @@ const struct brw_tracked_state brw_wm_unit; const struct brw_tracked_state brw_psp_urb_cbs; const struct brw_tracked_state brw_active_vertprog; -const struct brw_tracked_state brw_tnl_vertprog; const struct brw_tracked_state brw_pipe_control; const struct brw_tracked_state brw_clear_surface_cache; diff --git a/src/mesa/drivers/dri/i965/brw_state_upload.c b/src/mesa/drivers/dri/i965/brw_state_upload.c index 7d4fd467b1..b6a52843a8 100644 --- a/src/mesa/drivers/dri/i965/brw_state_upload.c +++ b/src/mesa/drivers/dri/i965/brw_state_upload.c @@ -45,7 +45,6 @@ const struct brw_tracked_state *atoms[] = { &brw_check_fallback, - &brw_tnl_vertprog, &brw_active_vertprog, &brw_wm_input_sizes, &brw_vs_prog, diff --git a/src/mesa/drivers/dri/i965/brw_vs.h b/src/mesa/drivers/dri/i965/brw_vs.h index 41a33ffe38..22388ec99d 100644 --- a/src/mesa/drivers/dri/i965/brw_vs.h +++ b/src/mesa/drivers/dri/i965/brw_vs.h @@ -80,8 +80,4 @@ struct brw_vs_compile { void brw_vs_emit( struct brw_vs_compile *c ); - -void brw_ProgramCacheDestroy( GLcontext *ctx ); -void brw_ProgramCacheInit( GLcontext *ctx ); - #endif diff --git a/src/mesa/drivers/dri/i965/brw_vs_tnl.c b/src/mesa/drivers/dri/i965/brw_vs_tnl.c index 9b04f19112..eacc289f1f 100644 --- a/src/mesa/drivers/dri/i965/brw_vs_tnl.c +++ b/src/mesa/drivers/dri/i965/brw_vs_tnl.c @@ -33,1635 +33,15 @@ #include "main/glheader.h" #include "main/macros.h" #include "main/enums.h" -#include "shader/prog_parameter.h" -#include "shader/prog_print.h" -#include "shader/prog_statevars.h" #include "brw_vs.h" #include "brw_state.h" -struct state_key { - unsigned light_global_enabled:1; - unsigned light_local_viewer:1; - unsigned light_twoside:1; - unsigned light_color_material:1; - unsigned light_color_material_mask:12; - unsigned light_material_mask:12; - unsigned normalize:1; - unsigned rescale_normals:1; - unsigned fog_source_is_depth:1; - unsigned tnl_do_vertex_fog:1; - unsigned separate_specular:1; - unsigned fog_option:2; - unsigned point_attenuated:1; - unsigned texture_enabled_global:1; - unsigned fragprog_inputs_read:12; - - struct { - unsigned light_enabled:1; - unsigned light_eyepos3_is_zero:1; - unsigned light_spotcutoff_is_180:1; - unsigned light_attenuated:1; - unsigned texunit_really_enabled:1; - unsigned texmat_enabled:1; - unsigned texgen_enabled:4; - unsigned texgen_mode0:4; - unsigned texgen_mode1:4; - unsigned texgen_mode2:4; - unsigned texgen_mode3:4; - } unit[8]; -}; - - - -#define FOG_NONE 0 -#define FOG_LINEAR 1 -#define FOG_EXP 2 -#define FOG_EXP2 3 - -static GLuint translate_fog_mode( GLenum mode ) -{ - switch (mode) { - case GL_LINEAR: return FOG_LINEAR; - case GL_EXP: return FOG_EXP; - case GL_EXP2: return FOG_EXP2; - default: return FOG_NONE; - } -} - -#define TXG_NONE 0 -#define TXG_OBJ_LINEAR 1 -#define TXG_EYE_LINEAR 2 -#define TXG_SPHERE_MAP 3 -#define TXG_REFLECTION_MAP 4 -#define TXG_NORMAL_MAP 5 - -static GLuint translate_texgen( GLboolean enabled, GLenum mode ) -{ - if (!enabled) - return TXG_NONE; - - switch (mode) { - case GL_OBJECT_LINEAR: return TXG_OBJ_LINEAR; - case GL_EYE_LINEAR: return TXG_EYE_LINEAR; - case GL_SPHERE_MAP: return TXG_SPHERE_MAP; - case GL_REFLECTION_MAP_NV: return TXG_REFLECTION_MAP; - case GL_NORMAL_MAP_NV: return TXG_NORMAL_MAP; - default: return TXG_NONE; - } -} - -static void make_state_key( GLcontext *ctx, struct state_key *key ) -{ - struct brw_context *brw = brw_context(ctx); - const struct gl_fragment_program *fp = brw->fragment_program; - GLuint i; - - /* This now relies on texenvprogram.c being active: - */ - assert(fp); - - memset(key, 0, sizeof(*key)); - - /* BRW_NEW_FRAGMENT_PROGRAM */ - key->fragprog_inputs_read = fp->Base.InputsRead; - - /* _NEW_LIGHT */ - key->separate_specular = (brw->attribs.Light->Model.ColorControl == - GL_SEPARATE_SPECULAR_COLOR); - - /* _NEW_LIGHT */ - if (brw->attribs.Light->Enabled) { - key->light_global_enabled = 1; - - if (brw->attribs.Light->Model.LocalViewer) - key->light_local_viewer = 1; - - if (brw->attribs.Light->Model.TwoSide) - key->light_twoside = 1; - - if (brw->attribs.Light->ColorMaterialEnabled) { - key->light_color_material = 1; - key->light_color_material_mask = brw->attribs.Light->ColorMaterialBitmask; - } - - /* BRW_NEW_INPUT_VARYING */ - - /* For these programs, material values are stuffed into the - * generic slots: - */ - for (i = 0 ; i < MAT_ATTRIB_MAX ; i++) - if (brw->vb.info.varying & (1<<(VERT_ATTRIB_GENERIC0 + i))) - key->light_material_mask |= 1<attribs.Light->Light[i]; - - if (light->Enabled) { - key->unit[i].light_enabled = 1; - - if (light->EyePosition[3] == 0.0) - key->unit[i].light_eyepos3_is_zero = 1; - - if (light->SpotCutoff == 180.0) - key->unit[i].light_spotcutoff_is_180 = 1; - - if (light->ConstantAttenuation != 1.0 || - light->LinearAttenuation != 0.0 || - light->QuadraticAttenuation != 0.0) - key->unit[i].light_attenuated = 1; - } - } - } - - /* _NEW_TRANSFORM */ - if (brw->attribs.Transform->Normalize) - key->normalize = 1; - - if (brw->attribs.Transform->RescaleNormals) - key->rescale_normals = 1; - - /* BRW_NEW_FRAGMENT_PROGRAM */ - key->fog_option = translate_fog_mode(fp->FogOption); - if (key->fog_option) - key->fragprog_inputs_read |= FRAG_BIT_FOGC; - - /* _NEW_FOG */ - if (brw->attribs.Fog->FogCoordinateSource == GL_FRAGMENT_DEPTH_EXT) - key->fog_source_is_depth = 1; - - /* _NEW_HINT, ??? */ - if (1) - key->tnl_do_vertex_fog = 1; - - /* _NEW_POINT */ - if (brw->attribs.Point->_Attenuated) - key->point_attenuated = 1; - - /* _NEW_TEXTURE */ - if (brw->attribs.Texture->_TexGenEnabled || - brw->attribs.Texture->_TexMatEnabled || - brw->attribs.Texture->_EnabledUnits) - key->texture_enabled_global = 1; - - for (i = 0; i < MAX_TEXTURE_UNITS; i++) { - struct gl_texture_unit *texUnit = &brw->attribs.Texture->Unit[i]; - - if (texUnit->_ReallyEnabled) - key->unit[i].texunit_really_enabled = 1; - - if (brw->attribs.Texture->_TexMatEnabled & ENABLE_TEXMAT(i)) - key->unit[i].texmat_enabled = 1; - - if (texUnit->TexGenEnabled) { - key->unit[i].texgen_enabled = 1; - - key->unit[i].texgen_mode0 = - translate_texgen( texUnit->TexGenEnabled & (1<<0), - texUnit->GenModeS ); - key->unit[i].texgen_mode1 = - translate_texgen( texUnit->TexGenEnabled & (1<<1), - texUnit->GenModeT ); - key->unit[i].texgen_mode2 = - translate_texgen( texUnit->TexGenEnabled & (1<<2), - texUnit->GenModeR ); - key->unit[i].texgen_mode3 = - translate_texgen( texUnit->TexGenEnabled & (1<<3), - texUnit->GenModeQ ); - } - } -} - - - -/* Very useful debugging tool - produces annotated listing of - * generated program with line/function references for each - * instruction back into this file: - */ -#define DISASSEM 0 - -/* Should be tunable by the driver - do we want to do matrix - * multiplications with DP4's or with MUL/MAD's? SSE works better - * with the latter, drivers may differ. - */ -#define PREFER_DP4 1 - - -/* Use uregs to represent registers internally, translate to Mesa's - * expected formats on emit. - * - * NOTE: These are passed by value extensively in this file rather - * than as usual by pointer reference. If this disturbs you, try - * remembering they are just 32bits in size. - * - * GCC is smart enough to deal with these dword-sized structures in - * much the same way as if I had defined them as dwords and was using - * macros to access and set the fields. This is much nicer and easier - * to evolve. - */ -struct ureg { - GLuint file:4; - GLint idx:8; /* relative addressing may be negative */ - GLuint negate:1; - GLuint swz:12; - GLuint pad:7; -}; - - -struct tnl_program { - const struct state_key *state; - struct gl_vertex_program *program; - - GLuint nr_instructions; - GLuint temp_in_use; - GLuint temp_reserved; - - struct ureg eye_position; - struct ureg eye_position_normalized; - struct ureg eye_normal; - struct ureg identity; - - GLuint materials; - GLuint color_materials; -}; - - -const static struct ureg undef = { - PROGRAM_UNDEFINED, - ~0, - 0, - 0, - 0 -}; - -/* Local shorthand: - */ -#define X SWIZZLE_X -#define Y SWIZZLE_Y -#define Z SWIZZLE_Z -#define W SWIZZLE_W - - -/* Construct a ureg: - */ -static struct ureg make_ureg(GLuint file, GLint idx) -{ - struct ureg reg; - reg.file = file; - reg.idx = idx; - reg.negate = 0; - reg.swz = SWIZZLE_NOOP; - reg.pad = 0; - return reg; -} - - - -static struct ureg ureg_negate( struct ureg reg ) -{ - reg.negate ^= 1; - return reg; -} - - -static struct ureg swizzle( struct ureg reg, int x, int y, int z, int w ) -{ - reg.swz = MAKE_SWIZZLE4(GET_SWZ(reg.swz, x), - GET_SWZ(reg.swz, y), - GET_SWZ(reg.swz, z), - GET_SWZ(reg.swz, w)); - - return reg; -} - -static struct ureg swizzle1( struct ureg reg, int x ) -{ - return swizzle(reg, x, x, x, x); -} - -static struct ureg get_temp( struct tnl_program *p ) -{ - int bit = ffs( ~p->temp_in_use ); - if (!bit) { - fprintf(stderr, "%s: out of temporaries\n", __FILE__); - assert(0); - } - - if (bit > p->program->Base.NumTemporaries) - p->program->Base.NumTemporaries = bit; - - p->temp_in_use |= 1<<(bit-1); - return make_ureg(PROGRAM_TEMPORARY, bit-1); -} - -static struct ureg reserve_temp( struct tnl_program *p ) -{ - struct ureg temp = get_temp( p ); - p->temp_reserved |= 1<temp_in_use &= ~(1<temp_in_use |= p->temp_reserved; /* can't release reserved temps */ - } -} - -static void release_temps( struct tnl_program *p ) -{ - p->temp_in_use = p->temp_reserved; -} - - - -static struct ureg register_input( struct tnl_program *p, GLuint input ) -{ - assert(input < 32); - - p->program->Base.InputsRead |= (1<program->Base.OutputsWritten |= (1<program->Base.Parameters, values, 4, - &swizzle); - assert(swizzle == SWIZZLE_NOOP); /* Need to handle swizzle in reg setup */ - return make_ureg(PROGRAM_STATE_VAR, idx); -} - -#define register_const1f(p, s0) register_const4f(p, s0, 0, 0, 1) -#define register_scalar_const(p, s0) register_const4f(p, s0, s0, s0, s0) -#define register_const2f(p, s0, s1) register_const4f(p, s0, s1, 0, 1) -#define register_const3f(p, s0, s1, s2) register_const4f(p, s0, s1, s2, 1) - -static GLboolean is_undef( struct ureg reg ) -{ - return reg.file == PROGRAM_UNDEFINED; -} - -static struct ureg get_identity_param( struct tnl_program *p ) -{ - if (is_undef(p->identity)) - p->identity = register_const4f(p, 0,0,0,1); - - return p->identity; -} - -static struct ureg register_param5( struct tnl_program *p, - GLint s0, - GLint s1, - GLint s2, - GLint s3, - GLint s4) -{ - gl_state_index tokens[STATE_LENGTH]; - GLint idx; - tokens[0] = s0; - tokens[1] = s1; - tokens[2] = s2; - tokens[3] = s3; - tokens[4] = s4; - idx = _mesa_add_state_reference( p->program->Base.Parameters, tokens ); - return make_ureg(PROGRAM_STATE_VAR, idx); -} - - -#define register_param1(p,s0) register_param5(p,s0,0,0,0,0) -#define register_param2(p,s0,s1) register_param5(p,s0,s1,0,0,0) -#define register_param3(p,s0,s1,s2) register_param5(p,s0,s1,s2,0,0) -#define register_param4(p,s0,s1,s2,s3) register_param5(p,s0,s1,s2,s3,0) - - -static void register_matrix_param5( struct tnl_program *p, - GLint s0, /* matrix name */ - GLint s1, /* texture matrix number */ - GLint s2, /* first row */ - GLint s3, /* last row */ - GLint s4, /* modifier */ - struct ureg *matrix ) -{ - GLint i; - - /* This is a bit sad as the support is there to pull the whole - * matrix out in one go: - */ - for (i = 0; i <= s3 - s2; i++) - matrix[i] = register_param5( p, s0, s1, i, i, s4 ); -} - - -static void emit_arg( struct prog_src_register *src, - struct ureg reg ) -{ - src->File = reg.file; - src->Index = reg.idx; - src->Swizzle = reg.swz; - src->RelAddr = 0; - src->NegateBase = reg.negate; - src->Abs = 0; - src->NegateAbs = 0; -} - -static void emit_dst( struct prog_dst_register *dst, - struct ureg reg, GLuint mask ) -{ - dst->File = reg.file; - dst->Index = reg.idx; - /* allow zero as a shorthand for xyzw */ - dst->WriteMask = mask ? mask : WRITEMASK_XYZW; - dst->CondMask = 0; - dst->CondSwizzle = 0; - dst->CondSrc = 0; - dst->pad = 0; -} - -static void debug_insn( struct prog_instruction *inst, const char *fn, - GLuint line ) -{ - if (DISASSEM) { - static const char *last_fn; - - if (fn != last_fn) { - last_fn = fn; - _mesa_printf("%s:\n", fn); - } - - _mesa_printf("%d:\t", line); - _mesa_print_instruction(inst); - } -} - - -static void emit_op3fn(struct tnl_program *p, - GLuint op, - struct ureg dest, - GLuint mask, - struct ureg src0, - struct ureg src1, - struct ureg src2, - const char *fn, - GLuint line) -{ - GLuint nr = p->program->Base.NumInstructions++; - - if (nr >= p->nr_instructions) { - int new_nr_instructions = p->nr_instructions * 2; - - p->program->Base.Instructions = - _mesa_realloc(p->program->Base.Instructions, - sizeof(struct prog_instruction) * p->nr_instructions, - sizeof(struct prog_instruction) * new_nr_instructions); - p->nr_instructions = new_nr_instructions; - } - - { - struct prog_instruction *inst = &p->program->Base.Instructions[nr]; - memset(inst, 0, sizeof(*inst)); - inst->Opcode = op; - inst->StringPos = 0; - inst->Data = 0; - - emit_arg( &inst->SrcReg[0], src0 ); - emit_arg( &inst->SrcReg[1], src1 ); - emit_arg( &inst->SrcReg[2], src2 ); - - emit_dst( &inst->DstReg, dest, mask ); - - debug_insn(inst, fn, line); - } -} - - - -#define emit_op3(p, op, dst, mask, src0, src1, src2) \ - emit_op3fn(p, op, dst, mask, src0, src1, src2, __FUNCTION__, __LINE__) - -#define emit_op2(p, op, dst, mask, src0, src1) \ - emit_op3fn(p, op, dst, mask, src0, src1, undef, __FUNCTION__, __LINE__) - -#define emit_op1(p, op, dst, mask, src0) \ - emit_op3fn(p, op, dst, mask, src0, undef, undef, __FUNCTION__, __LINE__) - - -static struct ureg make_temp( struct tnl_program *p, struct ureg reg ) -{ - if (reg.file == PROGRAM_TEMPORARY && - !(p->temp_reserved & (1<eye_position)) { - struct ureg pos = register_input( p, VERT_ATTRIB_POS ); - struct ureg modelview[4]; - - p->eye_position = reserve_temp(p); - - if (PREFER_DP4) { - register_matrix_param5( p, STATE_MODELVIEW_MATRIX, 0, 0, 3, - 0, modelview ); - - emit_matrix_transform_vec4(p, p->eye_position, modelview, pos); - } - else { - register_matrix_param5( p, STATE_MODELVIEW_MATRIX, 0, 0, 3, - STATE_MATRIX_TRANSPOSE, modelview ); - - emit_transpose_matrix_transform_vec4(p, p->eye_position, modelview, pos); - } - } - - return p->eye_position; -} - - -#if 0 -static struct ureg get_eye_z( struct tnl_program *p ) -{ - if (!is_undef(p->eye_position)) { - return swizzle1(p->eye_position, Z); - } - else if (!is_undef(p->eye_z)) { - struct ureg pos = register_input( p, BRW_ATTRIB_POS ); - struct ureg modelview2; - - p->eye_z = reserve_temp(p); - - register_matrix_param6( p, STATE_MATRIX, STATE_MODELVIEW, 0, 2, 1, - STATE_MATRIX, &modelview2 ); - - emit_matrix_transform_vec4(p, p->eye_position, modelview, pos); - emit_op2(p, OPCODE_DP4, p->eye_z, WRITEMASK_Z, pos, modelview2); - } - - return swizzle1(p->eye_z, Z) -} -#endif - - - -static struct ureg get_eye_position_normalized( struct tnl_program *p ) -{ - if (is_undef(p->eye_position_normalized)) { - struct ureg eye = get_eye_position(p); - p->eye_position_normalized = reserve_temp(p); - emit_normalize_vec3(p, p->eye_position_normalized, eye); - } - - return p->eye_position_normalized; -} - - -static struct ureg get_eye_normal( struct tnl_program *p ) -{ - if (is_undef(p->eye_normal)) { - struct ureg normal = register_input(p, VERT_ATTRIB_NORMAL ); - struct ureg mvinv[3]; - - register_matrix_param5( p, STATE_MODELVIEW_MATRIX, 0, 0, 2, - STATE_MATRIX_INVTRANS, mvinv ); - - p->eye_normal = reserve_temp(p); - - /* Transform to eye space: - */ - emit_matrix_transform_vec3( p, p->eye_normal, mvinv, normal ); - - /* Normalize/Rescale: - */ - if (p->state->normalize) { - emit_normalize_vec3( p, p->eye_normal, p->eye_normal ); - } - else if (p->state->rescale_normals) { - struct ureg rescale = register_param2(p, STATE_INTERNAL, - STATE_NORMAL_SCALE); - - emit_op2( p, OPCODE_MUL, p->eye_normal, 0, p->eye_normal, - swizzle1(rescale, X)); - } - } - - return p->eye_normal; -} - - - -static void build_hpos( struct tnl_program *p ) -{ - struct ureg pos = register_input( p, VERT_ATTRIB_POS ); - struct ureg hpos = register_output( p, VERT_RESULT_HPOS ); - struct ureg mvp[4]; - - if (PREFER_DP4) { - register_matrix_param5( p, STATE_MVP_MATRIX, 0, 0, 3, - 0, mvp ); - emit_matrix_transform_vec4( p, hpos, mvp, pos ); - } - else { - register_matrix_param5( p, STATE_MVP_MATRIX, 0, 0, 3, - STATE_MATRIX_TRANSPOSE, mvp ); - emit_transpose_matrix_transform_vec4( p, hpos, mvp, pos ); - } -} - - -static GLuint material_attrib( GLuint side, GLuint property ) -{ - return (property - STATE_AMBIENT) * 2 + side; -} - -/* Get a bitmask of which material values vary on a per-vertex basis. - */ -static void set_material_flags( struct tnl_program *p ) -{ - p->color_materials = 0; - p->materials = 0; - - if (p->state->light_color_material) { - p->materials = - p->color_materials = p->state->light_color_material_mask; - } - - p->materials |= p->state->light_material_mask; -} - - -static struct ureg get_material( struct tnl_program *p, GLuint side, - GLuint property ) -{ - GLuint attrib = material_attrib(side, property); - - if (p->color_materials & (1<materials & (1<materials & SCENE_COLOR_BITS(side)) { - struct ureg lm_ambient = register_param1(p, STATE_LIGHTMODEL_AMBIENT); - struct ureg material_emission = get_material(p, side, STATE_EMISSION); - struct ureg material_ambient = get_material(p, side, STATE_AMBIENT); - struct ureg material_diffuse = get_material(p, side, STATE_DIFFUSE); - struct ureg tmp = make_temp(p, material_diffuse); - emit_op3(p, OPCODE_MAD, tmp, WRITEMASK_XYZ, lm_ambient, - material_ambient, material_emission); - return tmp; - } - else - return register_param2( p, STATE_LIGHTMODEL_SCENECOLOR, side ); -} - - -static struct ureg get_lightprod( struct tnl_program *p, GLuint light, - GLuint side, GLuint property ) -{ - GLuint attrib = material_attrib(side, property); - if (p->materials & (1<state->unit[i].light_spotcutoff_is_180) { - struct ureg spot_dir_norm = register_param3(p, STATE_INTERNAL, - STATE_LIGHT_SPOT_DIR_NORMALIZED, i); - struct ureg spot = get_temp(p); - struct ureg slt = get_temp(p); - - emit_op2(p, OPCODE_DP3, spot, 0, ureg_negate(VPpli), spot_dir_norm); - emit_op2(p, OPCODE_SLT, slt, 0, swizzle1(spot_dir_norm,W), spot); - emit_op2(p, OPCODE_POW, spot, 0, spot, swizzle1(attenuation, W)); - emit_op2(p, OPCODE_MUL, att, 0, slt, spot); - - release_temp(p, spot); - release_temp(p, slt); - } - - /* Calculate distance attenuation: - */ - if (p->state->unit[i].light_attenuated) { - - /* 1/d,d,d,1/d */ - emit_op1(p, OPCODE_RCP, dist, WRITEMASK_YZ, dist); - /* 1,d,d*d,1/d */ - emit_op2(p, OPCODE_MUL, dist, WRITEMASK_XZ, dist, swizzle1(dist,Y)); - /* 1/dist-atten */ - emit_op2(p, OPCODE_DP3, dist, 0, attenuation, dist); - - if (!p->state->unit[i].light_spotcutoff_is_180) { - /* dist-atten */ - emit_op1(p, OPCODE_RCP, dist, 0, dist); - /* spot-atten * dist-atten */ - emit_op2(p, OPCODE_MUL, att, 0, dist, att); - } else { - /* dist-atten */ - emit_op1(p, OPCODE_RCP, att, 0, dist); - } - } - - return att; -} - - - - - -/* Need to add some addtional parameters to allow lighting in object - * space - STATE_SPOT_DIRECTION and STATE_HALF_VECTOR implicitly assume eye - * space lighting. - */ -static void build_lighting( struct tnl_program *p ) -{ - const GLboolean twoside = p->state->light_twoside; - const GLboolean separate = p->state->separate_specular; - GLuint nr_lights = 0, count = 0; - struct ureg normal = get_eye_normal(p); - struct ureg lit = get_temp(p); - struct ureg dots = get_temp(p); - struct ureg _col0 = undef, _col1 = undef; - struct ureg _bfc0 = undef, _bfc1 = undef; - GLuint i; - - for (i = 0; i < MAX_LIGHTS; i++) - if (p->state->unit[i].light_enabled) - nr_lights++; - - set_material_flags(p); - - { - struct ureg shininess = get_material(p, 0, STATE_SHININESS); - emit_op1(p, OPCODE_MOV, dots, WRITEMASK_W, swizzle1(shininess,X)); - release_temp(p, shininess); - - _col0 = make_temp(p, get_scenecolor(p, 0)); - if (separate) - _col1 = make_temp(p, get_identity_param(p)); - else - _col1 = _col0; - - } - - if (twoside) { - struct ureg shininess = get_material(p, 1, STATE_SHININESS); - emit_op1(p, OPCODE_MOV, dots, WRITEMASK_Z, - ureg_negate(swizzle1(shininess,X))); - release_temp(p, shininess); - - _bfc0 = make_temp(p, get_scenecolor(p, 1)); - if (separate) - _bfc1 = make_temp(p, get_identity_param(p)); - else - _bfc1 = _bfc0; - } - - - /* If no lights, still need to emit the scenecolor. - */ - /* KW: changed to do this always - v1.17 "Fix lighting alpha result"? - */ - if (p->state->fragprog_inputs_read & FRAG_BIT_COL0) - { - struct ureg res0 = register_output( p, VERT_RESULT_COL0 ); - emit_op1(p, OPCODE_MOV, res0, 0, _col0); - - if (twoside) { - struct ureg res0 = register_output( p, VERT_RESULT_BFC0 ); - emit_op1(p, OPCODE_MOV, res0, 0, _bfc0); - } - } - - if (separate && (p->state->fragprog_inputs_read & FRAG_BIT_COL1)) { - - struct ureg res1 = register_output( p, VERT_RESULT_COL1 ); - emit_op1(p, OPCODE_MOV, res1, 0, _col1); - - if (twoside) { - struct ureg res1 = register_output( p, VERT_RESULT_BFC1 ); - emit_op1(p, OPCODE_MOV, res1, 0, _bfc1); - } - } - - if (nr_lights == 0) { - release_temps(p); - return; - } - - - for (i = 0; i < MAX_LIGHTS; i++) { - if (p->state->unit[i].light_enabled) { - struct ureg half = undef; - struct ureg att = undef, VPpli = undef; - - count++; - - if (p->state->unit[i].light_eyepos3_is_zero) { - /* Can used precomputed constants in this case. - * Attenuation never applies to infinite lights. - */ - VPpli = register_param3(p, STATE_LIGHT, i, - STATE_LIGHT_POSITION_NORMALIZED); - if (p->state->light_local_viewer) { - struct ureg eye_hat = get_eye_position_normalized(p); - half = get_temp(p); - emit_op2(p, OPCODE_SUB, half, 0, VPpli, eye_hat); - emit_normalize_vec3(p, half, half); - } else { - half = register_param3(p, STATE_LIGHT, i, STATE_HALF_VECTOR); - } - } - else { - struct ureg Ppli = register_param3(p, STATE_LIGHT, i, - STATE_POSITION); - struct ureg V = get_eye_position(p); - struct ureg dist = get_temp(p); - struct ureg tmpPpli = get_temp(p); - - VPpli = get_temp(p); - half = get_temp(p); - - /* In homogeneous object coordinates - */ - emit_op1(p, OPCODE_RCP, dist, 0, swizzle1(Ppli, W)); - emit_op2(p, OPCODE_MUL, tmpPpli, 0, Ppli, dist); - - /* Calulate VPpli vector - */ - emit_op2(p, OPCODE_SUB, VPpli, 0, tmpPpli, V); - - /* Normalize VPpli. The dist value also used in - * attenuation below. - */ - emit_op2(p, OPCODE_DP3, dist, 0, VPpli, VPpli); - emit_op1(p, OPCODE_RSQ, dist, 0, dist); - emit_op2(p, OPCODE_MUL, VPpli, 0, VPpli, dist); - - - /* Calculate attenuation: - */ - if (!p->state->unit[i].light_spotcutoff_is_180 || - p->state->unit[i].light_attenuated) { - att = calculate_light_attenuation(p, i, VPpli, dist); - } - - - /* Calculate viewer direction, or use infinite viewer: - */ - if (p->state->light_local_viewer) { - struct ureg eye_hat = get_eye_position_normalized(p); - emit_op2(p, OPCODE_SUB, half, 0, VPpli, eye_hat); - } - else { - struct ureg z_dir = swizzle(get_identity_param(p),X,Y,W,Z); - emit_op2(p, OPCODE_ADD, half, 0, VPpli, z_dir); - } - - emit_normalize_vec3(p, half, half); - - release_temp(p, dist); - release_temp(p, tmpPpli); - } - - /* Calculate dot products: - */ - emit_op2(p, OPCODE_DP3, dots, WRITEMASK_X, normal, VPpli); - emit_op2(p, OPCODE_DP3, dots, WRITEMASK_Y, normal, half); - - - /* Front face lighting: - */ - { - struct ureg ambient = get_lightprod(p, i, 0, STATE_AMBIENT); - struct ureg diffuse = get_lightprod(p, i, 0, STATE_DIFFUSE); - struct ureg specular = get_lightprod(p, i, 0, STATE_SPECULAR); - struct ureg res0, res1; - GLuint mask0, mask1; - - emit_op1(p, OPCODE_LIT, lit, 0, dots); - - if (!is_undef(att)) - emit_op2(p, OPCODE_MUL, lit, 0, lit, att); - - - mask0 = 0; - mask1 = 0; - res0 = _col0; - res1 = _col1; - - if (count == nr_lights) { - if (separate) { - mask0 = WRITEMASK_XYZ; - mask1 = WRITEMASK_XYZ; - - if (p->state->fragprog_inputs_read & FRAG_BIT_COL0) - res0 = register_output( p, VERT_RESULT_COL0 ); - - if (p->state->fragprog_inputs_read & FRAG_BIT_COL1) - res1 = register_output( p, VERT_RESULT_COL1 ); - } - else { - mask1 = WRITEMASK_XYZ; - - if (p->state->fragprog_inputs_read & FRAG_BIT_COL0) - res1 = register_output( p, VERT_RESULT_COL0 ); - } - } - - emit_op3(p, OPCODE_MAD, _col0, 0, swizzle1(lit,X), ambient, _col0); - emit_op3(p, OPCODE_MAD, res0, mask0, swizzle1(lit,Y), diffuse, _col0); - emit_op3(p, OPCODE_MAD, res1, mask1, swizzle1(lit,Z), specular, _col1); - - release_temp(p, ambient); - release_temp(p, diffuse); - release_temp(p, specular); - } - - /* Back face lighting: - */ - if (twoside) { - struct ureg ambient = get_lightprod(p, i, 1, STATE_AMBIENT); - struct ureg diffuse = get_lightprod(p, i, 1, STATE_DIFFUSE); - struct ureg specular = get_lightprod(p, i, 1, STATE_SPECULAR); - struct ureg res0, res1; - GLuint mask0, mask1; - - emit_op1(p, OPCODE_LIT, lit, 0, ureg_negate(swizzle(dots,X,Y,W,Z))); - - if (!is_undef(att)) - emit_op2(p, OPCODE_MUL, lit, 0, lit, att); - - mask0 = 0; - mask1 = 0; - res0 = _bfc0; - res1 = _bfc1; - - if (count == nr_lights) { - if (separate) { - mask0 = WRITEMASK_XYZ; - mask1 = WRITEMASK_XYZ; - if (p->state->fragprog_inputs_read & FRAG_BIT_COL0) - res0 = register_output( p, VERT_RESULT_BFC0 ); - - if (p->state->fragprog_inputs_read & FRAG_BIT_COL1) - res1 = register_output( p, VERT_RESULT_BFC1 ); - } - else { - mask1 = WRITEMASK_XYZ; - - if (p->state->fragprog_inputs_read & FRAG_BIT_COL0) - res1 = register_output( p, VERT_RESULT_BFC0 ); - } - } - - emit_op3(p, OPCODE_MAD, _bfc0, 0, swizzle1(lit,X), ambient, _bfc0); - emit_op3(p, OPCODE_MAD, res0, mask0, swizzle1(lit,Y), diffuse, _bfc0); - emit_op3(p, OPCODE_MAD, res1, mask1, swizzle1(lit,Z), specular, _bfc1); - - release_temp(p, ambient); - release_temp(p, diffuse); - release_temp(p, specular); - } - - release_temp(p, half); - release_temp(p, VPpli); - release_temp(p, att); - } - } - - release_temps( p ); -} - - -static void build_fog( struct tnl_program *p ) -{ - struct ureg fog = register_output(p, VERT_RESULT_FOGC); - struct ureg input; - GLuint useabs = p->state->fog_source_is_depth && p->state->fog_option && - (p->state->fog_option != FOG_EXP2); - - if (p->state->fog_source_is_depth) { - input = swizzle1(get_eye_position(p), Z); - } - else { - input = swizzle1(register_input(p, VERT_ATTRIB_FOG), X); - if (p->state->fog_option && - p->state->tnl_do_vertex_fog) - input = swizzle1(register_input(p, VERT_ATTRIB_FOG), X); - else - input = register_input(p, VERT_ATTRIB_FOG); - } - - if (p->state->fog_option && - p->state->tnl_do_vertex_fog) { - struct ureg params = register_param2(p, STATE_INTERNAL, - STATE_FOG_PARAMS_OPTIMIZED); - struct ureg tmp = get_temp(p); - struct ureg id = get_identity_param(p); - - emit_op1(p, OPCODE_MOV, fog, 0, id); - - if (useabs) { - emit_op1(p, OPCODE_ABS, tmp, 0, input); - } - - switch (p->state->fog_option) { - case FOG_LINEAR: { - emit_op3(p, OPCODE_MAD, tmp, 0, useabs ? tmp : input, - swizzle1(params,X), swizzle1(params,Y)); - emit_op2(p, OPCODE_MAX, tmp, 0, tmp, swizzle1(id,X)); /* saturate */ - emit_op2(p, OPCODE_MIN, fog, WRITEMASK_X, tmp, swizzle1(id,W)); - break; - } - case FOG_EXP: - emit_op2(p, OPCODE_MUL, tmp, 0, useabs ? tmp : input, - swizzle1(params,Z)); - emit_op1(p, OPCODE_EX2, fog, WRITEMASK_X, ureg_negate(tmp)); - break; - case FOG_EXP2: - emit_op2(p, OPCODE_MUL, tmp, 0, input, swizzle1(params,W)); - emit_op2(p, OPCODE_MUL, tmp, 0, tmp, tmp); - emit_op1(p, OPCODE_EX2, fog, WRITEMASK_X, ureg_negate(tmp)); - break; - } - - release_temp(p, tmp); - } - else { - /* results = incoming fog coords (compute fog per-fragment later) - * - * KW: Is it really necessary to do anything in this case? - */ - emit_op1(p, useabs ? OPCODE_ABS : OPCODE_MOV, fog, 0, input); - } -} - -static void build_reflect_texgen( struct tnl_program *p, - struct ureg dest, - GLuint writemask ) -{ - struct ureg normal = get_eye_normal(p); - struct ureg eye_hat = get_eye_position_normalized(p); - struct ureg tmp = get_temp(p); - - /* n.u */ - emit_op2(p, OPCODE_DP3, tmp, 0, normal, eye_hat); - /* 2n.u */ - emit_op2(p, OPCODE_ADD, tmp, 0, tmp, tmp); - /* (-2n.u)n + u */ - emit_op3(p, OPCODE_MAD, dest, writemask, ureg_negate(tmp), normal, eye_hat); - - release_temp(p, tmp); -} - -static void build_sphere_texgen( struct tnl_program *p, - struct ureg dest, - GLuint writemask ) -{ - struct ureg normal = get_eye_normal(p); - struct ureg eye_hat = get_eye_position_normalized(p); - struct ureg tmp = get_temp(p); - struct ureg half = register_scalar_const(p, .5); - struct ureg r = get_temp(p); - struct ureg inv_m = get_temp(p); - struct ureg id = get_identity_param(p); - - /* Could share the above calculations, but it would be - * a fairly odd state for someone to set (both sphere and - * reflection active for different texture coordinate - * components. Of course - if two texture units enable - * reflect and/or sphere, things start to tilt in favour - * of seperating this out: - */ - - /* n.u */ - emit_op2(p, OPCODE_DP3, tmp, 0, normal, eye_hat); - /* 2n.u */ - emit_op2(p, OPCODE_ADD, tmp, 0, tmp, tmp); - /* (-2n.u)n + u */ - emit_op3(p, OPCODE_MAD, r, 0, ureg_negate(tmp), normal, eye_hat); - /* r + 0,0,1 */ - emit_op2(p, OPCODE_ADD, tmp, 0, r, swizzle(id,X,Y,W,Z)); - /* rx^2 + ry^2 + (rz+1)^2 */ - emit_op2(p, OPCODE_DP3, tmp, 0, tmp, tmp); - /* 2/m */ - emit_op1(p, OPCODE_RSQ, tmp, 0, tmp); - /* 1/m */ - emit_op2(p, OPCODE_MUL, inv_m, 0, tmp, half); - /* r/m + 1/2 */ - emit_op3(p, OPCODE_MAD, dest, writemask, r, inv_m, half); - - release_temp(p, tmp); - release_temp(p, r); - release_temp(p, inv_m); -} - - -static void build_texture_transform( struct tnl_program *p ) -{ - GLuint i, j; - - for (i = 0; i < MAX_TEXTURE_UNITS; i++) { - - if (!(p->state->fragprog_inputs_read & (FRAG_BIT_TEX0<state->unit[i].texgen_enabled || - p->state->unit[i].texmat_enabled) { - - GLuint texmat_enabled = p->state->unit[i].texmat_enabled; - struct ureg out = register_output(p, VERT_RESULT_TEX0 + i); - struct ureg out_texgen = undef; - - if (p->state->unit[i].texgen_enabled) { - GLuint copy_mask = 0; - GLuint sphere_mask = 0; - GLuint reflect_mask = 0; - GLuint normal_mask = 0; - GLuint modes[4]; - - if (texmat_enabled) - out_texgen = get_temp(p); - else - out_texgen = out; - - modes[0] = p->state->unit[i].texgen_mode0; - modes[1] = p->state->unit[i].texgen_mode1; - modes[2] = p->state->unit[i].texgen_mode2; - modes[3] = p->state->unit[i].texgen_mode3; - - for (j = 0; j < 4; j++) { - switch (modes[j]) { - case TXG_OBJ_LINEAR: { - struct ureg obj = register_input(p, VERT_ATTRIB_POS); - struct ureg plane = - register_param3(p, STATE_TEXGEN, i, - STATE_TEXGEN_OBJECT_S + j); - - emit_op2(p, OPCODE_DP4, out_texgen, WRITEMASK_X << j, - obj, plane ); - break; - } - case TXG_EYE_LINEAR: { - struct ureg eye = get_eye_position(p); - struct ureg plane = - register_param3(p, STATE_TEXGEN, i, - STATE_TEXGEN_EYE_S + j); - - emit_op2(p, OPCODE_DP4, out_texgen, WRITEMASK_X << j, - eye, plane ); - break; - } - case TXG_SPHERE_MAP: - sphere_mask |= WRITEMASK_X << j; - break; - case TXG_REFLECTION_MAP: - reflect_mask |= WRITEMASK_X << j; - break; - case TXG_NORMAL_MAP: - normal_mask |= WRITEMASK_X << j; - break; - case TXG_NONE: - copy_mask |= WRITEMASK_X << j; - } - - } - - - if (sphere_mask) { - build_sphere_texgen(p, out_texgen, sphere_mask); - } - - if (reflect_mask) { - build_reflect_texgen(p, out_texgen, reflect_mask); - } - - if (normal_mask) { - struct ureg normal = get_eye_normal(p); - emit_op1(p, OPCODE_MOV, out_texgen, normal_mask, normal ); - } - - if (copy_mask) { - struct ureg in = register_input(p, VERT_ATTRIB_TEX0+i); - emit_op1(p, OPCODE_MOV, out_texgen, copy_mask, in ); - } - } - - if (texmat_enabled) { - struct ureg texmat[4]; - struct ureg in = (!is_undef(out_texgen) ? - out_texgen : - register_input(p, VERT_ATTRIB_TEX0+i)); - if (PREFER_DP4) { - register_matrix_param5( p, STATE_TEXTURE_MATRIX, i, 0, 3, - 0, texmat ); - emit_matrix_transform_vec4( p, out, texmat, in ); - } - else { - register_matrix_param5( p, STATE_TEXTURE_MATRIX, i, 0, 3, - STATE_MATRIX_TRANSPOSE, texmat ); - emit_transpose_matrix_transform_vec4( p, out, texmat, in ); - } - } - - release_temps(p); - } - else { - emit_passthrough(p, VERT_ATTRIB_TEX0+i, VERT_RESULT_TEX0+i); - } - } -} - - -/* Seems like it could be tighter: - */ -static void build_pointsize( struct tnl_program *p ) -{ - struct ureg eye = get_eye_position(p); - struct ureg state_size = register_param1(p, STATE_POINT_SIZE); - struct ureg state_attenuation = register_param1(p, STATE_POINT_ATTENUATION); - struct ureg out = register_output(p, VERT_RESULT_PSIZ); - struct ureg ut = get_temp(p); - - /* 1, Z, Z * Z, 1 */ - emit_op1(p, OPCODE_MOV, ut, WRITEMASK_XW, swizzle1(get_identity_param(p), W)); - emit_op1(p, OPCODE_ABS, ut, WRITEMASK_YZ, swizzle1(eye, Z)); - emit_op2(p, OPCODE_MUL, ut, WRITEMASK_Z, ut, ut); - - - /* p1 + p2 * dist + p3 * dist * dist, 0 */ - emit_op2(p, OPCODE_DP3, ut, WRITEMASK_X, ut, state_attenuation); - - /* 1 / sqrt(factor) */ - emit_op1(p, OPCODE_RSQ, ut, WRITEMASK_X, ut ); - - /* ut = pointSize / factor */ - emit_op2(p, OPCODE_MUL, ut, WRITEMASK_X, ut, state_size); - - /* Clamp to min/max - state_size.[yz] - */ - emit_op2(p, OPCODE_MAX, ut, WRITEMASK_X, ut, swizzle1(state_size, Y)); - emit_op2(p, OPCODE_MIN, out, 0, swizzle1(ut, X), swizzle1(state_size, Z)); - - release_temp(p, ut); -} - -static void build_tnl_program( struct tnl_program *p ) -{ - /* Emit the program, starting with modelviewproject: - */ - build_hpos(p); - - /* Lighting calculations: - */ - if (p->state->fragprog_inputs_read & (FRAG_BIT_COL0|FRAG_BIT_COL1)) { - if (p->state->light_global_enabled) - build_lighting(p); - else { - if (p->state->fragprog_inputs_read & FRAG_BIT_COL0) - emit_passthrough(p, VERT_ATTRIB_COLOR0, VERT_RESULT_COL0); - - if (p->state->fragprog_inputs_read & FRAG_BIT_COL1) - emit_passthrough(p, VERT_ATTRIB_COLOR1, VERT_RESULT_COL1); - } - } - - if ((p->state->fragprog_inputs_read & FRAG_BIT_FOGC) || - p->state->fog_option != FOG_NONE) - build_fog(p); - - if (p->state->fragprog_inputs_read & FRAG_BITS_TEX_ANY) - build_texture_transform(p); - - if (p->state->point_attenuated) - build_pointsize(p); - - /* Finish up: - */ - emit_op1(p, OPCODE_END, undef, 0, undef); - - /* Disassemble: - */ - if (DISASSEM) { - _mesa_printf ("\n"); - } -} - - -static void build_new_tnl_program( const struct state_key *key, - struct gl_vertex_program *program, - GLuint max_temps) -{ - struct tnl_program p; - - _mesa_memset(&p, 0, sizeof(p)); - p.state = key; - p.program = program; - p.eye_position = undef; - p.eye_position_normalized = undef; - p.eye_normal = undef; - p.identity = undef; - p.temp_in_use = 0; - p.nr_instructions = 16; - - if (max_temps >= sizeof(int) * 8) - p.temp_reserved = 0; - else - p.temp_reserved = ~((1<Base.Instructions = - _mesa_malloc(sizeof(struct prog_instruction) * p.nr_instructions); - p.program->Base.String = 0; - p.program->Base.NumInstructions = - p.program->Base.NumTemporaries = - p.program->Base.NumParameters = - p.program->Base.NumAttributes = p.program->Base.NumAddressRegs = 0; - p.program->Base.Parameters = _mesa_new_parameter_list(); - p.program->Base.InputsRead = 0; - p.program->Base.OutputsWritten = 0; - - build_tnl_program( &p ); -} - -static void *search_cache( struct brw_tnl_cache *cache, - GLuint hash, - const void *key, - GLuint keysize) -{ - struct brw_tnl_cache_item *c; - - for (c = cache->items[hash % cache->size]; c; c = c->next) { - if (c->hash == hash && memcmp(c->key, key, keysize) == 0) - return c->data; - } - - return NULL; -} - -static void rehash( struct brw_tnl_cache *cache ) -{ - struct brw_tnl_cache_item **items; - struct brw_tnl_cache_item *c, *next; - GLuint size, i; - - size = cache->size * 3; - items = (struct brw_tnl_cache_item**) _mesa_malloc(size * sizeof(*items)); - _mesa_memset(items, 0, size * sizeof(*items)); - - for (i = 0; i < cache->size; i++) - for (c = cache->items[i]; c; c = next) { - next = c->next; - c->next = items[c->hash % size]; - items[c->hash % size] = c; - } - - FREE(cache->items); - cache->items = items; - cache->size = size; -} - -static void cache_item( struct brw_tnl_cache *cache, - GLuint hash, - const struct state_key *key, - void *data ) -{ - struct brw_tnl_cache_item *c = MALLOC(sizeof(*c)); - c->hash = hash; - - c->key = malloc(sizeof(*key)); - memcpy(c->key, key, sizeof(*key)); - - c->data = data; - - if (++cache->n_items > cache->size * 1.5) - rehash(cache); - - c->next = cache->items[hash % cache->size]; - cache->items[hash % cache->size] = c; -} - - -static GLuint hash_key( struct state_key *key ) -{ - GLuint *ikey = (GLuint *)key; - GLuint hash = 0, i; - - /* I'm sure this can be improved on, but speed is important: - */ - for (i = 0; i < sizeof(*key)/sizeof(GLuint); i++) - hash += ikey[i]; - - return hash; -} - -static void prepare_tnl_program( struct brw_context *brw ) -{ - GLcontext *ctx = &brw->intel.ctx; - struct state_key key; - GLuint hash; - struct gl_vertex_program *old = brw->tnl_program; - - /* _NEW_PROGRAM */ - if (brw->attribs.VertexProgram->_Current) - return; - - /* Grab all the relevent state and put it in a single structure: - */ - make_state_key(ctx, &key); - hash = hash_key(&key); - - /* Look for an already-prepared program for this state: - */ - brw->tnl_program = (struct gl_vertex_program *) - search_cache( &brw->tnl_program_cache, hash, &key, sizeof(key) ); - - /* OK, we'll have to build a new one: - */ - if (!brw->tnl_program) { - brw->tnl_program = (struct gl_vertex_program *) - ctx->Driver.NewProgram(ctx, GL_VERTEX_PROGRAM_ARB, 0); - - build_new_tnl_program( &key, brw->tnl_program, -/* ctx->Const.MaxVertexProgramTemps */ - 32 - ); - - if (ctx->Driver.ProgramStringNotify) - ctx->Driver.ProgramStringNotify( ctx, GL_VERTEX_PROGRAM_ARB, - &brw->tnl_program->Base ); - - cache_item( &brw->tnl_program_cache, - hash, &key, brw->tnl_program ); - } - - if (old != brw->tnl_program) - brw->state.dirty.brw |= BRW_NEW_TNL_PROGRAM; - return; -} - -/* Note: See brw_draw.c - the vertex program must not rely on - * brw->primitive or brw->reduced_prim. - */ -const struct brw_tracked_state brw_tnl_vertprog = { - .dirty = { - .mesa = (_NEW_PROGRAM | - _NEW_LIGHT | - _NEW_TRANSFORM | - _NEW_FOG | - _NEW_HINT | - _NEW_POINT | - _NEW_TEXTURE | - _NEW_TEXTURE_MATRIX), - .brw = (BRW_NEW_FRAGMENT_PROGRAM | - BRW_NEW_INPUT_VARYING), - .cache = 0 - }, - .prepare = prepare_tnl_program -}; - - - - static void prepare_active_vertprog( struct brw_context *brw ) { const struct gl_vertex_program *prev = brw->vertex_program; - /* NEW_PROGRAM */ - if (brw->attribs.VertexProgram->_Current) { - brw->vertex_program = brw->attribs.VertexProgram->_Current; - } - else { - /* BRW_NEW_TNL_PROGRAM */ - brw->vertex_program = brw->tnl_program; - } + brw->vertex_program = brw->attribs.VertexProgram->_Current; if (brw->vertex_program != prev) brw->state.dirty.brw |= BRW_NEW_VERTEX_PROGRAM; @@ -1672,37 +52,8 @@ static void prepare_active_vertprog( struct brw_context *brw ) const struct brw_tracked_state brw_active_vertprog = { .dirty = { .mesa = _NEW_PROGRAM, - .brw = BRW_NEW_TNL_PROGRAM, + .brw = 0, .cache = 0 }, .prepare = prepare_active_vertprog }; - - -void brw_ProgramCacheInit( GLcontext *ctx ) -{ - struct brw_context *brw = brw_context(ctx); - - brw->tnl_program_cache.size = 17; - brw->tnl_program_cache.n_items = 0; - brw->tnl_program_cache.items = (struct brw_tnl_cache_item **) - _mesa_calloc(brw->tnl_program_cache.size * - sizeof(struct brw_tnl_cache_item)); -} - -void brw_ProgramCacheDestroy( GLcontext *ctx ) -{ - struct brw_context *brw = brw_context(ctx); - struct brw_tnl_cache_item *c, *next; - GLuint i; - - for (i = 0; i < brw->tnl_program_cache.size; i++) - for (c = brw->tnl_program_cache.items[i]; c; c = next) { - next = c->next; - FREE(c->key); - FREE(c->data); - FREE(c); - } - - FREE(brw->tnl_program_cache.items); -} diff --git a/src/mesa/drivers/dri/i965/brw_vtbl.c b/src/mesa/drivers/dri/i965/brw_vtbl.c index 3780d3dad2..cd074dfed6 100644 --- a/src/mesa/drivers/dri/i965/brw_vtbl.c +++ b/src/mesa/drivers/dri/i965/brw_vtbl.c @@ -70,7 +70,6 @@ static void brw_destroy_context( struct intel_context *intel ) brw_destroy_state(brw); brw_draw_destroy( brw ); - brw_ProgramCacheDestroy( ctx ); brw_FrameBufferTexDestroy( brw ); for (i = 0; i < brw->state.nr_draw_regions; i++) -- cgit v1.2.3 From df6ae3f0a39f95cb1199ac16c98be2bf9a3bc96f Mon Sep 17 00:00:00 2001 From: Eric Anholt Date: Wed, 1 Oct 2008 14:14:06 -0700 Subject: i965: Fix overwriting of depth override for SetTexOffset. Fixes black borders around windows in compiz. Bug #17233. --- src/mesa/drivers/dri/i965/brw_wm_surface_state.c | 1 - 1 file changed, 1 deletion(-) (limited to 'src/mesa/drivers/dri/i965') diff --git a/src/mesa/drivers/dri/i965/brw_wm_surface_state.c b/src/mesa/drivers/dri/i965/brw_wm_surface_state.c index 47127c0421..3790b50c97 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_surface_state.c +++ b/src/mesa/drivers/dri/i965/brw_wm_surface_state.c @@ -274,7 +274,6 @@ brw_update_texture_surface( GLcontext *ctx, GLuint unit ) key.width = firstImage->Width; key.height = firstImage->Height; key.cpp = intelObj->mt->cpp; - key.depth = firstImage->Depth; key.tiling = intelObj->mt->region->tiling; dri_bo_unreference(brw->wm.surf_bo[unit + MAX_DRAW_BUFFERS]); -- cgit v1.2.3 From 17fdd1ab3b48da8dff742e626c3c59bc89cafeed Mon Sep 17 00:00:00 2001 From: Eric Anholt Date: Wed, 1 Oct 2008 16:58:38 -0700 Subject: i965: sampler default color ends up in texture cache, not instructions. See volume 4, SAMPLER_BORDER_COLOR_STATE programming notes. --- src/mesa/drivers/dri/i965/brw_wm_sampler_state.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src/mesa/drivers/dri/i965') diff --git a/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c b/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c index e1db31ec08..977b90ad43 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c +++ b/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c @@ -304,7 +304,7 @@ static void upload_wm_samplers( struct brw_context *brw ) continue; dri_bo_emit_reloc(brw->wm.sampler_bo, - I915_GEM_DOMAIN_INSTRUCTION, 0, + I915_GEM_DOMAIN_SAMPLER, 0, 0, i * sizeof(struct brw_sampler_state) + offsetof(struct brw_sampler_state, ss2), -- cgit v1.2.3 From 94d3a30df759bb7c2724fdcee9e89a350d3a4d8b Mon Sep 17 00:00:00 2001 From: "Xiang, Haihao" Date: Wed, 8 Oct 2008 09:30:12 +0800 Subject: i965: Fix a potential assertion failure. --- src/mesa/drivers/dri/i965/brw_draw_upload.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'src/mesa/drivers/dri/i965') diff --git a/src/mesa/drivers/dri/i965/brw_draw_upload.c b/src/mesa/drivers/dri/i965/brw_draw_upload.c index 303eaac5cf..cc3d9396ac 100644 --- a/src/mesa/drivers/dri/i965/brw_draw_upload.c +++ b/src/mesa/drivers/dri/i965/brw_draw_upload.c @@ -365,8 +365,10 @@ static void brw_prepare_vertices(struct brw_context *brw) if (i == 0) { /* Position array not properly enabled: */ - if (input->glarray->StrideB == 0) - return; + if (input->glarray->StrideB == 0) { + intel->Fallback = 1; + return; + } interleave = input->glarray->StrideB; ptr = input->glarray->Ptr; -- cgit v1.2.3 From c157cfc6376f7469ab272b18868183e5ff9ac754 Mon Sep 17 00:00:00 2001 From: Eric Anholt Date: Mon, 6 Oct 2008 17:34:51 -0700 Subject: i965: Add ARB_occlusion_query support. --- src/mesa/drivers/dri/i965/Makefile | 1 + src/mesa/drivers/dri/i965/brw_context.c | 4 + src/mesa/drivers/dri/i965/brw_context.h | 30 +++- src/mesa/drivers/dri/i965/brw_draw.c | 2 +- src/mesa/drivers/dri/i965/brw_draw_upload.c | 3 + src/mesa/drivers/dri/i965/brw_queryobj.c | 263 ++++++++++++++++++++++++++++ src/mesa/drivers/dri/i965/brw_vtbl.c | 10 ++ src/mesa/drivers/dri/intel/intel_context.c | 51 +----- src/mesa/drivers/dri/intel/intel_reg.h | 19 ++ 9 files changed, 331 insertions(+), 52 deletions(-) create mode 100644 src/mesa/drivers/dri/i965/brw_queryobj.c (limited to 'src/mesa/drivers/dri/i965') diff --git a/src/mesa/drivers/dri/i965/Makefile b/src/mesa/drivers/dri/i965/Makefile index c2d555cd0c..7bc16a02c8 100644 --- a/src/mesa/drivers/dri/i965/Makefile +++ b/src/mesa/drivers/dri/i965/Makefile @@ -51,6 +51,7 @@ DRIVER_SOURCES = \ brw_metaops.c \ brw_misc_state.c \ brw_program.c \ + brw_queryobj.c \ brw_sf.c \ brw_sf_emit.c \ brw_sf_state.c \ diff --git a/src/mesa/drivers/dri/i965/brw_context.c b/src/mesa/drivers/dri/i965/brw_context.c index 5f60477176..474158b484 100644 --- a/src/mesa/drivers/dri/i965/brw_context.c +++ b/src/mesa/drivers/dri/i965/brw_context.c @@ -33,6 +33,7 @@ #include "main/imports.h" #include "main/api_noop.h" #include "main/vtxfmt.h" +#include "main/simple_list.h" #include "shader/shader_api.h" #include "brw_context.h" @@ -68,6 +69,7 @@ static void brwInitDriverFunctions( struct dd_function_table *functions ) brwInitFragProgFuncs( functions ); brwInitProgFuncs( functions ); + brw_init_queryobj_functions(functions); } @@ -150,6 +152,8 @@ GLboolean brwCreateContext( const __GLcontextModes *mesaVis, ctx->VertexProgram._MaintainTnlProgram = GL_TRUE; ctx->FragmentProgram._MaintainTexEnvProgram = GL_TRUE; + make_empty_list(&brw->query.active_head); + brw_draw_init( brw ); return GL_TRUE; diff --git a/src/mesa/drivers/dri/i965/brw_context.h b/src/mesa/drivers/dri/i965/brw_context.h index 8bddc9da40..1c6a0dede0 100644 --- a/src/mesa/drivers/dri/i965/brw_context.h +++ b/src/mesa/drivers/dri/i965/brw_context.h @@ -410,7 +410,22 @@ struct brw_tnl_cache { GLuint size, n_items; }; +struct brw_query_object { + struct gl_query_object Base; + /** Doubly linked list of active query objects in the context. */ + struct brw_query_object *prev, *next; + + /** Last query BO associated with this query. */ + dri_bo *bo; + /** First index in bo with query data for this object. */ + int first_index; + /** Last index in bo with query data for this object. */ + int last_index; + + /* Total count of pixels from previous BOs */ + unsigned int count; +}; struct brw_context { @@ -626,7 +641,12 @@ struct brw_context dri_bo *vp_bo; } cc; - + struct { + struct brw_query_object active_head; + dri_bo *bo; + int index; + GLboolean active; + } query; /* Used to give every program string a unique id */ GLuint program_id; @@ -651,7 +671,13 @@ GLboolean brwCreateContext( const __GLcontextModes *mesaVis, __DRIcontextPrivate *driContextPriv, void *sharedContextPrivate); - +/*====================================================================== + * brw_queryobj.c + */ +void brw_init_queryobj_functions(struct dd_function_table *functions); +void brw_prepare_query_begin(struct brw_context *brw); +void brw_emit_query_begin(struct brw_context *brw); +void brw_emit_query_end(struct brw_context *brw); /*====================================================================== * brw_state.c diff --git a/src/mesa/drivers/dri/i965/brw_draw.c b/src/mesa/drivers/dri/i965/brw_draw.c index 39ce8eb4b6..6c71b4abcf 100644 --- a/src/mesa/drivers/dri/i965/brw_draw.c +++ b/src/mesa/drivers/dri/i965/brw_draw.c @@ -382,7 +382,6 @@ void brw_draw_prims( GLcontext *ctx, return; } - /* Make a first attempt at drawing: */ retval = brw_try_draw_prims(ctx, arrays, prim, nr_prims, ib, min_index, max_index); @@ -395,6 +394,7 @@ void brw_draw_prims( GLcontext *ctx, _swsetup_Wakeup(ctx); _tnl_draw_prims(ctx, arrays, prim, nr_prims, ib, min_index, max_index); } + } void brw_draw_init( struct brw_context *brw ) diff --git a/src/mesa/drivers/dri/i965/brw_draw_upload.c b/src/mesa/drivers/dri/i965/brw_draw_upload.c index cc3d9396ac..7b88b5eaa1 100644 --- a/src/mesa/drivers/dri/i965/brw_draw_upload.c +++ b/src/mesa/drivers/dri/i965/brw_draw_upload.c @@ -415,6 +415,8 @@ static void brw_prepare_vertices(struct brw_context *brw) copy_array_to_vbo_array(brw, upload[i], upload[i]->element_size); } } + + brw_prepare_query_begin(brw); } static void brw_emit_vertices(struct brw_context *brw) @@ -435,6 +437,7 @@ static void brw_emit_vertices(struct brw_context *brw) enabled[nr_enabled++] = input; } + brw_emit_query_begin(brw); /* Now emit VB and VEP state packets. * diff --git a/src/mesa/drivers/dri/i965/brw_queryobj.c b/src/mesa/drivers/dri/i965/brw_queryobj.c new file mode 100644 index 0000000000..a1a1353dee --- /dev/null +++ b/src/mesa/drivers/dri/i965/brw_queryobj.c @@ -0,0 +1,263 @@ +/* + * Copyright © 2008 Intel Corporation + * + * 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. + * + * Authors: + * Eric Anholt + * + */ + +/** @file support for ARB_query_object + * + * ARB_query_object is implemented by using the PIPE_CONTROL command to stall + * execution on the completion of previous depth tests, and write the + * current PS_DEPTH_COUNT to a buffer object. + * + * We use before and after counts when drawing during a query so that + * we don't pick up other clients' query data in ours. To reduce overhead, + * a single BO is used to record the query data for all active queries at + * once. This also gives us a simple bound on how much batchbuffer space is + * required for handling queries, so that we can be sure that we won't + * have to emit a batchbuffer without getting the ending PS_DEPTH_COUNT. + */ +#include "main/simple_list.h" +#include "main/imports.h" + +#include "brw_context.h" +#include "intel_batchbuffer.h" +#include "intel_reg.h" + +/** Waits on the query object's BO and totals the results for this query */ +static void +brw_queryobj_get_results(struct brw_query_object *query) +{ + int i; + uint64_t *results; + + if (query->bo == NULL) + return; + + /* Map and count the pixels from the current query BO */ + dri_bo_map(query->bo, GL_FALSE); + results = query->bo->virtual; + for (i = query->first_index; i <= query->last_index; i++) { + query->Base.Result += results[i * 2 + 1] - results[i * 2]; + } + dri_bo_unmap(query->bo); + + dri_bo_unreference(query->bo); + query->bo = NULL; +} + +static struct gl_query_object * +brw_new_query_object(GLcontext *ctx, GLuint id) +{ + struct brw_query_object *query; + + query = _mesa_calloc(sizeof(struct brw_query_object)); + + query->Base.Id = id; + query->Base.Result = 0; + query->Base.Active = GL_FALSE; + query->Base.Ready = GL_TRUE; + + return &query->Base; +} + +static void +brw_delete_query(GLcontext *ctx, struct gl_query_object *q) +{ + struct brw_query_object *query = (struct brw_query_object *)q; + + dri_bo_unreference(query->bo); + _mesa_free(query); +} + +static void +brw_begin_query(GLcontext *ctx, struct gl_query_object *q) +{ + struct brw_context *brw = brw_context(ctx); + struct intel_context *intel = intel_context(ctx); + struct brw_query_object *query = (struct brw_query_object *)q; + + /* Reset our driver's tracking of query state. */ + dri_bo_unreference(query->bo); + query->bo = NULL; + query->first_index = -1; + query->last_index = -1; + + insert_at_head(&brw->query.active_head, query); + intel->stats_wm++; +} + +/** + * Begin the ARB_occlusion_query query on a query object. + */ +static void +brw_end_query(GLcontext *ctx, struct gl_query_object *q) +{ + struct brw_context *brw = brw_context(ctx); + struct intel_context *intel = intel_context(ctx); + struct brw_query_object *query = (struct brw_query_object *)q; + + /* Flush the batchbuffer in case it has writes to our query BO. + * Have later queries write to a new query BO so that further rendering + * doesn't delay the collection of our results. + */ + if (query->bo) { + brw_emit_query_end(brw); + intel_batchbuffer_flush(intel->batch); + + dri_bo_unreference(brw->query.bo); + brw->query.bo = NULL; + } + + remove_from_list(query); + + intel->stats_wm--; +} + +static void brw_wait_query(GLcontext *ctx, struct gl_query_object *q) +{ + struct brw_query_object *query = (struct brw_query_object *)q; + + brw_queryobj_get_results(query); + query->Base.Ready = GL_TRUE; +} + +static void brw_check_query(GLcontext *ctx, struct gl_query_object *q) +{ + /* XXX: Need to expose dri_bo_is_idle from bufmgr. */ +#if 0 + struct brw_query_object *query = (struct brw_query_object *)q; + + if (dri_bo_is_idle(query->bo)) { + brw_queryobj_get_results(query); + query->Base.Ready = GL_TRUE; + } +#else + brw_wait_query(ctx, q); +#endif +} + +/** Called to set up the query BO and account for its aperture space */ +void +brw_prepare_query_begin(struct brw_context *brw) +{ + struct intel_context *intel = &brw->intel; + dri_bo *aper_array[] = { + intel->batch->buf, + brw->query.bo, + }; + + /* Skip if we're not doing any queries. */ + if (is_empty_list(&brw->query.active_head)) + return; + + /* Get a new query BO if we're going to need it. */ + if (brw->query.bo == NULL || + brw->query.index * 2 + 1 >= 4096 / sizeof(uint64_t)) { + dri_bo_unreference(brw->query.bo); + brw->query.bo = NULL; + + brw->query.bo = dri_bo_alloc(intel->bufmgr, "query", 4096, 1); + brw->query.index = 0; + } + + if (dri_bufmgr_check_aperture_space(aper_array, ARRAY_SIZE(aper_array))) + intel_batchbuffer_flush(intel->batch); +} + +/** Called just before primitive drawing to get a beginning PS_DEPTH_COUNT. */ +void +brw_emit_query_begin(struct brw_context *brw) +{ + struct intel_context *intel = &brw->intel; + struct brw_query_object *query; + + /* Skip if we're not doing any queries, or we've emitted the start. */ + if (brw->query.active || is_empty_list(&brw->query.active_head)) + return; + + BEGIN_BATCH(4, IGNORE_CLIPRECTS); + OUT_BATCH(_3DSTATE_PIPE_CONTROL | + PIPE_CONTROL_DEPTH_STALL | + PIPE_CONTROL_WRITE_DEPTH_COUNT); + /* This object could be mapped cacheable, but we don't have an exposed + * mechanism to support that. Since it's going uncached, tell GEM that + * we're writing to it. The usual clflush should be all that's required + * to pick up the results. + */ + OUT_RELOC(brw->query.bo, + I915_GEM_DOMAIN_INSTRUCTION, I915_GEM_DOMAIN_INSTRUCTION, + PIPE_CONTROL_GLOBAL_GTT_WRITE | + ((brw->query.index * 2) * sizeof(uint64_t))); + OUT_BATCH(0); + OUT_BATCH(0); + ADVANCE_BATCH(); + + foreach(query, &brw->query.active_head) { + if (query->bo != brw->query.bo) { + if (query->bo != NULL) + brw_queryobj_get_results(query); + dri_bo_reference(brw->query.bo); + query->bo = brw->query.bo; + query->first_index = brw->query.index; + } + query->last_index = brw->query.index; + } + brw->query.active = GL_TRUE; +} + +/** Called at batchbuffer flush to get an ending PS_DEPTH_COUNT */ +void +brw_emit_query_end(struct brw_context *brw) +{ + struct intel_context *intel = &brw->intel; + + if (!brw->query.active) + return; + + BEGIN_BATCH(4, IGNORE_CLIPRECTS); + OUT_BATCH(_3DSTATE_PIPE_CONTROL | + PIPE_CONTROL_DEPTH_STALL | + PIPE_CONTROL_WRITE_DEPTH_COUNT); + OUT_RELOC(brw->query.bo, + I915_GEM_DOMAIN_INSTRUCTION, I915_GEM_DOMAIN_INSTRUCTION, + PIPE_CONTROL_GLOBAL_GTT_WRITE | + ((brw->query.index * 2 + 1) * sizeof(uint64_t))); + OUT_BATCH(0); + OUT_BATCH(0); + ADVANCE_BATCH(); + + brw->query.active = GL_FALSE; + brw->query.index++; +} + +void brw_init_queryobj_functions(struct dd_function_table *functions) +{ + functions->NewQueryObject = brw_new_query_object; + functions->DeleteQuery = brw_delete_query; + functions->BeginQuery = brw_begin_query; + functions->EndQuery = brw_end_query; + functions->CheckQuery = brw_check_query; + functions->WaitQuery = brw_wait_query; +} diff --git a/src/mesa/drivers/dri/i965/brw_vtbl.c b/src/mesa/drivers/dri/i965/brw_vtbl.c index cd074dfed6..a64e437860 100644 --- a/src/mesa/drivers/dri/i965/brw_vtbl.c +++ b/src/mesa/drivers/dri/i965/brw_vtbl.c @@ -120,6 +120,15 @@ static void brw_set_draw_region( struct intel_context *intel, brw->state.nr_draw_regions = num_regions; } +/* called from intel_batchbuffer_flush and children before sending a + * batchbuffer off. + */ +static void brw_finish_batch(struct intel_context *intel) +{ + struct brw_context *brw = brw_context(&intel->ctx); + + brw_emit_query_end(brw); +} /* called from intelFlushBatchLocked */ @@ -218,6 +227,7 @@ void brwInitVtbl( struct brw_context *brw ) brw->intel.vtbl.note_fence = brw_note_fence; brw->intel.vtbl.note_unlock = brw_note_unlock; brw->intel.vtbl.new_batch = brw_new_batch; + brw->intel.vtbl.finish_batch = brw_finish_batch; brw->intel.vtbl.destroy = brw_destroy_context; brw->intel.vtbl.set_draw_region = brw_set_draw_region; brw->intel.vtbl.flush_cmd = brw_flush_cmd; diff --git a/src/mesa/drivers/dri/intel/intel_context.c b/src/mesa/drivers/dri/intel/intel_context.c index 9dc32e487f..e6c0d3175e 100644 --- a/src/mesa/drivers/dri/intel/intel_context.c +++ b/src/mesa/drivers/dri/intel/intel_context.c @@ -394,6 +394,7 @@ static const struct dri_extension brw_extensions[] = { { "GL_ARB_fragment_program", NULL }, { "GL_ARB_fragment_program_shadow", NULL }, { "GL_ARB_fragment_shader", NULL }, + { "GL_ARB_occlusion_query", GL_ARB_occlusion_query_functions }, { "GL_ARB_point_sprite", NULL }, { "GL_ARB_shader_objects", GL_ARB_shader_objects_functions }, { "GL_ARB_shading_language_100", GL_VERSION_2_0_functions }, @@ -407,12 +408,9 @@ static const struct dri_extension brw_extensions[] = { { NULL, NULL } }; -#ifdef I915_MMIO_READ -static const struct dri_extension arb_oc_extensions[] = { - { "GL_ARB_occlusion_query", GL_ARB_occlusion_query_functions }, +static const struct dri_extension arb_oq_extensions[] = { { NULL, NULL } }; -#endif static const struct dri_extension ttm_extensions[] = { { "GL_ARB_pixel_buffer_object", NULL }, @@ -437,13 +435,6 @@ void intelInitExtensions(GLcontext *ctx, GLboolean enable_imaging) if (intel == NULL || intel->ttm) driInitExtensions(ctx, ttm_extensions, GL_FALSE); -#ifdef I915_MMIO_READ - if (intel == NULL || - (IS_965(intel->intelScreen->deviceID) && - intel->intelScreen->drmMinor >= 8)) - driInitExtensions(ctx, arb_oc_extensions, GL_FALSE); -#endif - if (intel == NULL || IS_965(intel->intelScreen->deviceID)) driInitExtensions(ctx, brw_extensions, GL_FALSE); } @@ -540,39 +531,6 @@ intelFinish(GLcontext * ctx) } } -#ifdef I915_MMIO_READ -static void -intelBeginQuery(GLcontext *ctx, struct gl_query_object *q) -{ - struct intel_context *intel = intel_context( ctx ); - struct drm_i915_mmio io = { - .read_write = I915_MMIO_READ, - .reg = MMIO_REGS_PS_DEPTH_COUNT, - .data = &q->Result - }; - intel->stats_wm++; - intelFinish(&intel->ctx); - drmCommandWrite(intel->driFd, DRM_I915_MMIO, &io, sizeof(io)); -} - -static void -intelEndQuery(GLcontext *ctx, struct gl_query_object *q) -{ - struct intel_context *intel = intel_context( ctx ); - GLuint64EXT tmp; - struct drm_i915_mmio io = { - .read_write = I915_MMIO_READ, - .reg = MMIO_REGS_PS_DEPTH_COUNT, - .data = &tmp - }; - intelFinish(&intel->ctx); - drmCommandWrite(intel->driFd, DRM_I915_MMIO, &io, sizeof(io)); - q->Result = tmp - q->Result; - q->Ready = GL_TRUE; - intel->stats_wm--; -} -#endif - void intelInitDriverFunctions(struct dd_function_table *functions) { @@ -589,11 +547,6 @@ intelInitDriverFunctions(struct dd_function_table *functions) functions->CopyConvolutionFilter1D = _swrast_CopyConvolutionFilter1D; functions->CopyConvolutionFilter2D = _swrast_CopyConvolutionFilter2D; -#ifdef I915_MMIO_READ - functions->BeginQuery = intelBeginQuery; - functions->EndQuery = intelEndQuery; -#endif - intelInitTextureFuncs(functions); intelInitStateFuncs(functions); intelInitBufferFuncs(functions); diff --git a/src/mesa/drivers/dri/intel/intel_reg.h b/src/mesa/drivers/dri/intel/intel_reg.h index 96af7e1a03..c21f408093 100644 --- a/src/mesa/drivers/dri/intel/intel_reg.h +++ b/src/mesa/drivers/dri/intel/intel_reg.h @@ -44,6 +44,25 @@ #define _3DSTATE_LOAD_STATE_IMMEDIATE_1 (CMD_3D | (0x1d<<24) | (0x04<<16)) #define I1_LOAD_S(n) (1<<(4+n)) +/** @{ + * + * PIPE_CONTROL operation, a combination MI_FLUSH and register write with + * additional flushing control. + */ +#define _3DSTATE_PIPE_CONTROL (CMD_3D | (3 << 27) | (2 << 24) | 2) +#define PIPE_CONTROL_NO_WRITE (0 << 14) +#define PIPE_CONTROL_WRITE_IMMEDIATE (1 << 14) +#define PIPE_CONTROL_WRITE_DEPTH_COUNT (2 << 14) +#define PIPE_CONTROL_WRITE_TIMESTAMP (3 << 14) +#define PIPE_CONTROL_DEPTH_STALL (1 << 13) +#define PIPE_CONTROL_WRITE_FLUSH (1 << 12) +#define PIPE_CONTROL_INSTRUCTION_FLUSH (1 << 11) +#define PIPE_CONTROL_INTERRUPT_ENABLE (1 << 8) +#define PIPE_CONTROL_PPGTT_WRITE (0 << 2) +#define PIPE_CONTROL_GLOBAL_GTT_WRITE (1 << 2) + +/** @} */ + /** @{ * 915 definitions */ -- cgit v1.2.3 From 91221483a633d6230a4f8d2500ed180428754215 Mon Sep 17 00:00:00 2001 From: Eric Anholt Date: Thu, 9 Oct 2008 10:23:47 -0700 Subject: i965: Actually hook up the accelerated DrawPixels support. --- src/mesa/drivers/dri/i965/Makefile | 3 ++- src/mesa/drivers/dri/intel/intel_pixel.c | 2 +- src/mesa/drivers/dri/intel/intel_pixel_draw.c | 1 - 3 files changed, 3 insertions(+), 3 deletions(-) (limited to 'src/mesa/drivers/dri/i965') diff --git a/src/mesa/drivers/dri/i965/Makefile b/src/mesa/drivers/dri/i965/Makefile index 7bc16a02c8..005460f354 100644 --- a/src/mesa/drivers/dri/i965/Makefile +++ b/src/mesa/drivers/dri/i965/Makefile @@ -18,8 +18,9 @@ DRIVER_SOURCES = \ intel_screen.c \ intel_span.c \ intel_pixel.c \ - intel_pixel_copy.c \ intel_pixel_bitmap.c \ + intel_pixel_copy.c \ + intel_pixel_draw.c \ intel_state.c \ intel_tex.c \ intel_tex_copy.c \ diff --git a/src/mesa/drivers/dri/intel/intel_pixel.c b/src/mesa/drivers/dri/intel/intel_pixel.c index f39fac13cf..b267ffd890 100644 --- a/src/mesa/drivers/dri/intel/intel_pixel.c +++ b/src/mesa/drivers/dri/intel/intel_pixel.c @@ -181,9 +181,9 @@ intelInitPixelFuncs(struct dd_function_table *functions) if (!getenv("INTEL_NO_BLIT")) { functions->Bitmap = intelBitmap; functions->CopyPixels = intelCopyPixels; + functions->DrawPixels = intelDrawPixels; #ifdef I915 functions->ReadPixels = intelReadPixels; - functions->DrawPixels = intelDrawPixels; #endif } } diff --git a/src/mesa/drivers/dri/intel/intel_pixel_draw.c b/src/mesa/drivers/dri/intel/intel_pixel_draw.c index be213e7b96..b60dad7460 100644 --- a/src/mesa/drivers/dri/intel/intel_pixel_draw.c +++ b/src/mesa/drivers/dri/intel/intel_pixel_draw.c @@ -51,7 +51,6 @@ #include "intel_regions.h" #include "intel_pixel.h" #include "intel_buffer_objects.h" -#include "intel_tris.h" static GLboolean -- cgit v1.2.3 From 7216679c1998b49ff5b08e6b43f8d5779415bf54 Mon Sep 17 00:00:00 2001 From: Eric Anholt Date: Thu, 9 Oct 2008 11:45:58 -0700 Subject: i965: Accelerate depth textures with border color. The fallback was introduced to fix bug #16697, but made the test it was fixing run excessively long. --- src/mesa/drivers/dri/i965/brw_fallback.c | 5 +---- src/mesa/drivers/dri/i965/brw_wm_sampler_state.c | 21 +++++++++++++++++++-- 2 files changed, 20 insertions(+), 6 deletions(-) (limited to 'src/mesa/drivers/dri/i965') diff --git a/src/mesa/drivers/dri/i965/brw_fallback.c b/src/mesa/drivers/dri/i965/brw_fallback.c index 2f6b7febbd..4ea660a51a 100644 --- a/src/mesa/drivers/dri/i965/brw_fallback.c +++ b/src/mesa/drivers/dri/i965/brw_fallback.c @@ -74,10 +74,7 @@ static GLboolean do_check_fallback(struct brw_context *brw) if (texUnit->_ReallyEnabled) { struct intel_texture_object *intelObj = intel_texture_object(texUnit->_Current); struct gl_texture_image *texImage = intelObj->base.Image[0][intelObj->firstLevel]; - if (texImage->Border || - ((texImage->_BaseFormat == GL_DEPTH_COMPONENT) && - ((texImage->TexObject->WrapS == GL_CLAMP_TO_BORDER) || - (texImage->TexObject->WrapT == GL_CLAMP_TO_BORDER)))) { + if (texImage->Border) { DBG("FALLBACK: texture border\n"); return GL_TRUE; } diff --git a/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c b/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c index 977b90ad43..f12ef47a7d 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c +++ b/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c @@ -229,6 +229,9 @@ brw_wm_sampler_populate_key(struct brw_context *brw, struct wm_sampler_entry *entry = &key->sampler[unit]; struct gl_texture_unit *texUnit = &brw->attribs.Texture->Unit[unit]; struct gl_texture_object *texObj = texUnit->_Current; + struct intel_texture_object *intelObj = intel_texture_object(texObj); + struct gl_texture_image *firstImage = + texObj->Image[0][intelObj->firstLevel]; entry->wrap_r = texObj->WrapR; entry->wrap_s = texObj->WrapS; @@ -244,8 +247,22 @@ brw_wm_sampler_populate_key(struct brw_context *brw, entry->comparefunc = texObj->CompareFunc; dri_bo_unreference(brw->wm.sdc_bo[unit]); - brw->wm.sdc_bo[unit] = upload_default_color(brw, texObj->BorderColor); - + if (firstImage->_BaseFormat == GL_DEPTH_COMPONENT) { + float bordercolor[4] = { + texObj->BorderColor[0], + texObj->BorderColor[0], + texObj->BorderColor[0], + texObj->BorderColor[0] + }; + /* GL specs that border color for depth textures is taken from the + * R channel, while the hardware uses A. Spam R into all the + * channels for safety. + */ + brw->wm.sdc_bo[unit] = upload_default_color(brw, bordercolor); + } else { + brw->wm.sdc_bo[unit] = upload_default_color(brw, + texObj->BorderColor); + } key->sampler_count = unit + 1; } } -- cgit v1.2.3