diff options
Diffstat (limited to 'src/mesa/drivers/dri/r300')
45 files changed, 757 insertions, 553 deletions
diff --git a/src/mesa/drivers/dri/r300/compiler/r300_fragprog_emit.c b/src/mesa/drivers/dri/r300/compiler/r300_fragprog_emit.c index 3b2b06fc2b..8be32ea91f 100644 --- a/src/mesa/drivers/dri/r300/compiler/r300_fragprog_emit.c +++ b/src/mesa/drivers/dri/r300/compiler/r300_fragprog_emit.c @@ -74,7 +74,7 @@ static void use_temporary(struct r300_fragment_program_code *code, unsigned int code->pixsize = index; } -static unsigned int use_source(struct r300_fragment_program_code* code, struct radeon_pair_instruction_source src) +static unsigned int use_source(struct r300_fragment_program_code* code, struct rc_pair_instruction_source src) { if (src.File == RC_FILE_CONSTANT) { return src.Index | (1 << 5); @@ -133,6 +133,8 @@ static unsigned int translate_alpha_opcode(struct r300_fragment_program_compiler */ static int emit_alu(struct r300_emit_state * emit, struct rc_pair_instruction* inst) { + int ip; + int j; PROG_CODE; if (code->alu.length >= c->Base.max_alu_insts) { @@ -140,20 +142,20 @@ static int emit_alu(struct r300_emit_state * emit, struct rc_pair_instruction* i return 0; } - int ip = code->alu.length++; - int j; + ip = code->alu.length++; code->alu.inst[ip].rgb_inst = translate_rgb_opcode(c, inst->RGB.Opcode); code->alu.inst[ip].alpha_inst = translate_alpha_opcode(c, inst->Alpha.Opcode); for(j = 0; j < 3; ++j) { unsigned int src = use_source(code, inst->RGB.Src[j]); + unsigned int arg; code->alu.inst[ip].rgb_addr |= src << (6*j); src = use_source(code, inst->Alpha.Src[j]); code->alu.inst[ip].alpha_addr |= src << (6*j); - unsigned int arg = r300FPTranslateRGBSwizzle(inst->RGB.Arg[j].Source, inst->RGB.Arg[j].Swizzle); + arg = r300FPTranslateRGBSwizzle(inst->RGB.Arg[j].Source, inst->RGB.Arg[j].Swizzle); arg |= inst->RGB.Arg[j].Abs << 6; arg |= inst->RGB.Arg[j].Negate << 5; code->alu.inst[ip].rgb_inst |= arg << (7*j); @@ -259,6 +261,10 @@ static int finish_node(struct r300_emit_state * emit) { struct r300_fragment_program_compiler * c = emit->compiler; struct r300_fragment_program_code *code = &emit->compiler->code->code.r300; + unsigned alu_offset; + unsigned alu_end; + unsigned tex_offset; + unsigned tex_end; if (code->alu.length == emit->node_first_alu) { /* Generate a single NOP for this node */ @@ -268,10 +274,10 @@ static int finish_node(struct r300_emit_state * emit) return 0; } - unsigned alu_offset = emit->node_first_alu; - unsigned alu_end = code->alu.length - alu_offset - 1; - unsigned tex_offset = emit->node_first_tex; - unsigned tex_end = code->tex.length - tex_offset - 1; + alu_offset = emit->node_first_alu; + alu_end = code->alu.length - alu_offset - 1; + tex_offset = emit->node_first_tex; + tex_end = code->tex.length - tex_offset - 1; if (code->tex.length == emit->node_first_tex) { if (emit->current_node > 0) { @@ -334,6 +340,9 @@ static int begin_tex(struct r300_emit_state * emit) static int emit_tex(struct r300_emit_state * emit, struct rc_instruction * inst) { + unsigned int unit; + unsigned int dest; + unsigned int opcode; PROG_CODE; if (code->tex.length >= R300_PFS_MAX_TEX_INST) { @@ -341,9 +350,8 @@ static int emit_tex(struct r300_emit_state * emit, struct rc_instruction * inst) return 0; } - unsigned int unit = inst->U.I.TexSrcUnit; - unsigned int dest = inst->U.I.DstReg.Index; - unsigned int opcode; + unit = inst->U.I.TexSrcUnit; + dest = inst->U.I.DstReg.Index; switch(inst->U.I.Opcode) { case RC_OPCODE_KIL: opcode = R300_TEX_OP_KIL; break; diff --git a/src/mesa/drivers/dri/r300/compiler/r300_fragprog_swizzle.c b/src/mesa/drivers/dri/r300/compiler/r300_fragprog_swizzle.c index caa48fe478..2d28b06539 100644 --- a/src/mesa/drivers/dri/r300/compiler/r300_fragprog_swizzle.c +++ b/src/mesa/drivers/dri/r300/compiler/r300_fragprog_swizzle.c @@ -94,6 +94,9 @@ static const struct swizzle_data* lookup_native_swizzle(unsigned int swizzle) */ static int r300_swizzle_is_native(rc_opcode opcode, struct rc_src_register reg) { + unsigned int relevant; + int j; + if (reg.Abs) reg.Negate = RC_MASK_NONE; @@ -101,8 +104,6 @@ static int r300_swizzle_is_native(rc_opcode opcode, struct rc_src_register reg) opcode == RC_OPCODE_TEX || opcode == RC_OPCODE_TXB || opcode == RC_OPCODE_TXP) { - int j; - if (reg.Abs || reg.Negate) return 0; @@ -117,8 +118,7 @@ static int r300_swizzle_is_native(rc_opcode opcode, struct rc_src_register reg) return 1; } - unsigned int relevant = 0; - int j; + relevant = 0; for(j = 0; j < 3; ++j) if (GET_SWZ(reg.Swizzle, j) != RC_SWIZZLE_UNUSED) @@ -154,9 +154,10 @@ static void r300_swizzle_split( unsigned int matchcount = 0; unsigned int matchmask = 0; for(comp = 0; comp < 3; ++comp) { + unsigned int swz; if (!GET_BIT(mask, comp)) continue; - unsigned int swz = GET_SWZ(src.Swizzle, comp); + swz = GET_SWZ(src.Swizzle, comp); if (swz == RC_SWIZZLE_UNUSED) continue; if (swz == GET_SWZ(sd->hash, comp)) { diff --git a/src/mesa/drivers/dri/r300/compiler/r3xx_vertprog.c b/src/mesa/drivers/dri/r300/compiler/r3xx_vertprog.c index 08785716db..d7d49e514b 100644 --- a/src/mesa/drivers/dri/r300/compiler/r3xx_vertprog.c +++ b/src/mesa/drivers/dri/r300/compiler/r3xx_vertprog.c @@ -392,9 +392,9 @@ static void ei_if(struct r300_vertex_program_compiler * compiler, * don't already have one. */ if (!compiler->PredicateMask) { unsigned int writemasks[RC_REGISTER_MAX_INDEX]; - memset(writemasks, 0, sizeof(writemasks)); struct rc_instruction * inst; unsigned int i; + memset(writemasks, 0, sizeof(writemasks)); for(inst = compiler->Base.Program.Instructions.Next; inst != &compiler->Base.Program.Instructions; inst = inst->Next) { diff --git a/src/mesa/drivers/dri/r300/compiler/r500_fragprog.c b/src/mesa/drivers/dri/r300/compiler/r500_fragprog.c index 627ce374ef..289bb87ae5 100644 --- a/src/mesa/drivers/dri/r300/compiler/r500_fragprog.c +++ b/src/mesa/drivers/dri/r300/compiler/r500_fragprog.c @@ -39,10 +39,12 @@ int r500_transform_IF( struct rc_instruction * inst, void* data) { + struct rc_instruction * inst_mov; + if (inst->U.I.Opcode != RC_OPCODE_IF) return 0; - struct rc_instruction * inst_mov = rc_insert_new_instruction(c, inst->Prev); + inst_mov = rc_insert_new_instruction(c, inst->Prev); inst_mov->U.I.Opcode = RC_OPCODE_MOV; inst_mov->U.I.DstReg.WriteMask = 0; inst_mov->U.I.WriteALUResult = RC_ALURESULT_W; @@ -251,12 +253,11 @@ void r500FragmentProgramDump(struct radeon_compiler *c, void *user) { struct r300_fragment_program_compiler *compiler = (struct r300_fragment_program_compiler*)c; struct r500_fragment_program_code *code = &compiler->code->code.r500; - fprintf(stderr, "R500 Fragment Program:\n--------\n"); - int n, i; uint32_t inst; uint32_t inst0; char *str = NULL; + fprintf(stderr, "R500 Fragment Program:\n--------\n"); for (n = 0; n < code->inst_end+1; n++) { inst0 = inst = code->inst[n].inst0; diff --git a/src/mesa/drivers/dri/r300/compiler/r500_fragprog_emit.c b/src/mesa/drivers/dri/r300/compiler/r500_fragprog_emit.c index 54cff9169a..6f101c68eb 100644 --- a/src/mesa/drivers/dri/r300/compiler/r500_fragprog_emit.c +++ b/src/mesa/drivers/dri/r300/compiler/r500_fragprog_emit.c @@ -198,7 +198,7 @@ static void use_temporary(struct r500_fragment_program_code* code, unsigned int code->max_temp_idx = index; } -static unsigned int use_source(struct r500_fragment_program_code* code, struct radeon_pair_instruction_source src) +static unsigned int use_source(struct r500_fragment_program_code* code, struct rc_pair_instruction_source src) { if (src.File == RC_FILE_CONSTANT) { return src.Index | 0x100; @@ -227,6 +227,7 @@ static void alu_nop(struct r300_fragment_program_compiler *c, int ip) */ static void emit_paired(struct r300_fragment_program_compiler *c, struct rc_pair_instruction *inst) { + int ip; PROG_CODE; if (code->inst_end >= c->Base.max_alu_insts-1) { @@ -234,7 +235,7 @@ static void emit_paired(struct r300_fragment_program_compiler *c, struct rc_pair return; } - int ip = ++code->inst_end; + ip = ++code->inst_end; /* Quirk: MDH/MDV (DDX/DDY) need a NOP on previous non-TEX instructions. */ if (inst->RGB.Opcode == RC_OPCODE_DDX || inst->Alpha.Opcode == RC_OPCODE_DDX || @@ -250,7 +251,7 @@ static void emit_paired(struct r300_fragment_program_compiler *c, struct rc_pair if (inst->RGB.OutputWriteMask || inst->Alpha.OutputWriteMask || inst->Alpha.DepthWriteMask) { code->inst[ip].inst0 = R500_INST_TYPE_OUT; if (inst->WriteALUResult) { - error("%s: cannot write output and ALU result at the same time"); + error("Cannot write output and ALU result at the same time"); return; } } else { @@ -357,6 +358,7 @@ static unsigned int translate_strq_swizzle(unsigned int swizzle) */ static int emit_tex(struct r300_fragment_program_compiler *c, struct rc_sub_instruction *inst) { + int ip; PROG_CODE; if (code->inst_end >= c->Base.max_alu_insts-1) { @@ -364,7 +366,7 @@ static int emit_tex(struct r300_fragment_program_compiler *c, struct rc_sub_inst return 0; } - int ip = ++code->inst_end; + ip = ++code->inst_end; code->inst[ip].inst0 = R500_INST_TYPE_TEX | (inst->DstReg.WriteMask << 11) @@ -407,12 +409,14 @@ static int emit_tex(struct r300_fragment_program_compiler *c, struct rc_sub_inst static void emit_flowcontrol(struct emit_state * s, struct rc_instruction * inst) { + unsigned int newip; + if (s->Code->inst_end >= s->C->max_alu_insts-1) { rc_error(s->C, "emit_tex: Too many instructions"); return; } - unsigned int newip = ++s->Code->inst_end; + newip = ++s->Code->inst_end; /* Currently all loops use the same integer constant to intialize * the loop variables. */ @@ -623,6 +627,8 @@ void r500BuildFragmentProgramHwCode(struct radeon_compiler *c, void *user) if (code->inst_end == -1 || (code->inst[code->inst_end].inst0 & R500_INST_TYPE_MASK) != R500_INST_TYPE_OUT) { + int ip; + /* This may happen when dead-code elimination is disabled or * when most of the fragment program logic is leading to a KIL */ if (code->inst_end >= compiler->Base.max_alu_insts-1) { @@ -630,7 +636,7 @@ void r500BuildFragmentProgramHwCode(struct radeon_compiler *c, void *user) return; } - int ip = ++code->inst_end; + ip = ++code->inst_end; code->inst[ip].inst0 = R500_INST_TYPE_OUT | R500_INST_TEX_SEM_WAIT; } diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_code.c b/src/mesa/drivers/dri/r300/compiler/radeon_code.c index 0eab18c344..6842fb873b 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_code.c +++ b/src/mesa/drivers/dri/r300/compiler/radeon_code.c @@ -28,6 +28,7 @@ #include "radeon_code.h" #include <stdlib.h> +#include <stdio.h> #include <string.h> #include "radeon_program.h" @@ -171,3 +172,16 @@ unsigned rc_constants_add_immediate_scalar(struct rc_constant_list * c, float da return rc_constants_add(c, &constant); } + +void rc_constants_print(struct rc_constant_list * c) +{ + unsigned int i; + for(i = 0; i < c->Count; i++) { + if (c->Constants[i].Type == RC_CONSTANT_IMMEDIATE) { + float * values = c->Constants[i].u.Immediate; + fprintf(stderr, "CONST[%u] = " + "{ %10.4f %10.4f %10.4f %10.4f }\n", + i, values[0],values[1], values[2], values[3]); + } + } +} diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_code.h b/src/mesa/drivers/dri/r300/compiler/radeon_code.h index f76676fae8..cfb6df2cd7 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_code.h +++ b/src/mesa/drivers/dri/r300/compiler/radeon_code.h @@ -60,6 +60,7 @@ enum { RC_STATE_R300_WINDOW_DIMENSION, RC_STATE_R300_TEXRECT_FACTOR, + RC_STATE_R300_TEXSCALE_FACTOR, RC_STATE_R300_VIEWPORT_SCALE, RC_STATE_R300_VIEWPORT_OFFSET }; @@ -89,6 +90,7 @@ unsigned rc_constants_add(struct rc_constant_list * c, struct rc_constant * cons unsigned rc_constants_add_state(struct rc_constant_list * c, unsigned state1, unsigned state2); unsigned rc_constants_add_immediate_vec4(struct rc_constant_list * c, const float * data); unsigned rc_constants_add_immediate_scalar(struct rc_constant_list * c, float data, unsigned * swizzle); +void rc_constants_print(struct rc_constant_list * c); /** * Compare functions. @@ -126,11 +128,7 @@ struct r300_fragment_program_external_state { struct { /** * If the sampler is used as a shadow sampler, - * this field is: - * 0 - GL_LUMINANCE - * 1 - GL_INTENSITY - * 2 - GL_ALPHA - * depending on the depth texture mode. + * this field contains swizzle depending on the depth texture mode. */ unsigned depth_texture_swizzle:12; @@ -150,13 +148,9 @@ struct r300_fragment_program_external_state { unsigned compare_mode_enabled : 1; /** - * If the sampler needs to fake NPOT, this field is set. - */ - unsigned fake_npot : 1; - - /** - * If the sampler will recieve non-normalized coords, - * this field is set. + * If the sampler will receive non-normalized coords, + * this field is set. The scaling factor is given by + * RC_STATE_R300_TEXRECT_FACTOR. */ unsigned non_normalized_coords : 1; @@ -166,7 +160,13 @@ struct r300_fragment_program_external_state { * If this field is \ref RC_WRAP_NONE (aka 0), no wrapping maths * will be performed on the coordinates. */ - unsigned wrap_mode : 2; + unsigned wrap_mode : 3; + + /** + * The coords are scaled after applying the wrap mode emulation + * and right before texture fetch. The scaling factor is given by + * RC_STATE_R300_TEXSCALE_FACTOR. */ + unsigned clamp_and_scale_before_fetch : 1; } unit[16]; }; diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_dataflow.c b/src/mesa/drivers/dri/r300/compiler/radeon_dataflow.c index e73700f84a..a27d395587 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_dataflow.c +++ b/src/mesa/drivers/dri/r300/compiler/radeon_dataflow.c @@ -29,49 +29,32 @@ #include "radeon_program.h" +struct read_write_mask_data { + void * UserData; + rc_read_write_mask_fn Cb; +}; + static void reads_normal_callback( - rc_read_write_chan_fn cb, + void * userdata, struct rc_instruction * fullinst, - struct rc_src_register src, - void * userdata) + struct rc_src_register * src) { + struct read_write_mask_data * cb_data = userdata; unsigned int refmask = 0; unsigned int chan; for(chan = 0; chan < 4; chan++) { - refmask |= 1 << GET_SWZ(src.Swizzle, chan); + refmask |= 1 << GET_SWZ(src->Swizzle, chan); } refmask &= RC_MASK_XYZW; - if (refmask) - cb(userdata, fullinst, src.File, src.Index, refmask); - - if (refmask && src.RelAddr) - cb(userdata, fullinst, RC_FILE_ADDRESS, 0, RC_MASK_X); -} - -static void reads_normal(struct rc_instruction * fullinst, rc_read_write_chan_fn cb, void * userdata) -{ - struct rc_sub_instruction * inst = &fullinst->U.I; - const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->Opcode); - - for(unsigned int src = 0; src < opcode->NumSrcRegs; ++src) { - - if (inst->SrcReg[src].File == RC_FILE_NONE) - return; + if (refmask) { + cb_data->Cb(cb_data->UserData, fullinst, src->File, + src->Index, refmask); + } - if (inst->SrcReg[src].File == RC_FILE_PRESUB) { - unsigned int i; - unsigned int srcp_regs = rc_presubtract_src_reg_count( - inst->PreSub.Opcode); - for( i = 0; i < srcp_regs; i++) { - reads_normal_callback(cb, fullinst, - inst->PreSub.SrcReg[i], - userdata); - } - } else { - reads_normal_callback(cb, fullinst, - inst->SrcReg[src], userdata); - } + if (refmask && src->RelAddr) { + cb_data->Cb(cb_data->UserData, fullinst, RC_FILE_ADDRESS, 0, + RC_MASK_X); } } @@ -112,7 +95,7 @@ static void pair_get_src_refmasks(unsigned int * refmasks, } } -static void reads_pair(struct rc_instruction * fullinst, rc_read_write_mask_fn cb, void * userdata) +static void reads_pair(struct rc_instruction * fullinst, rc_read_write_mask_fn cb, void * userdata) { struct rc_pair_instruction * inst = &fullinst->U.P; unsigned int refmasks[3] = { 0, 0, 0 }; @@ -143,6 +126,74 @@ static void reads_pair(struct rc_instruction * fullinst, rc_read_write_mask_fn } } +static void pair_sub_for_all_args( + struct rc_instruction * fullinst, + struct rc_pair_sub_instruction * sub, + rc_pair_read_arg_fn cb, + void * userdata) +{ + int i; + const struct rc_opcode_info * info = rc_get_opcode_info(sub->Opcode); + + for(i = 0; i < info->NumSrcRegs; i++) { + cb(userdata, fullinst, &sub->Arg[i]); + } +} + +/* This function calls the callback function (cb) for each source used by + * the instruction. + * */ +void rc_for_all_reads_src( + struct rc_instruction * inst, + rc_read_src_fn cb, + void * userdata) +{ + const struct rc_opcode_info * opcode = + rc_get_opcode_info(inst->U.I.Opcode); + + /* This function only works with normal instructions. */ + if (inst->Type != RC_INSTRUCTION_NORMAL) { + assert(0); + return; + } + + for(unsigned int src = 0; src < opcode->NumSrcRegs; ++src) { + + if (inst->U.I.SrcReg[src].File == RC_FILE_NONE) + continue; + + if (inst->U.I.SrcReg[src].File == RC_FILE_PRESUB) { + unsigned int i; + unsigned int srcp_regs = rc_presubtract_src_reg_count( + inst->U.I.PreSub.Opcode); + for( i = 0; i < srcp_regs; i++) { + cb(userdata, inst, &inst->U.I.PreSub.SrcReg[i]); + } + } else { + cb(userdata, inst, &inst->U.I.SrcReg[src]); + } + } +} + +/** + * This function calls the callback function (cb) for each arg of the RGB and + * alpha components. + */ +void rc_pair_for_all_reads_arg( + struct rc_instruction * inst, + rc_pair_read_arg_fn cb, + void * userdata) +{ + /* This function only works with pair instructions. */ + if (inst->Type != RC_INSTRUCTION_PAIR) { + assert(0); + return; + } + + pair_sub_for_all_args(inst, &inst->U.P.RGB, cb, userdata); + pair_sub_for_all_args(inst, &inst->U.P.Alpha, cb, userdata); +} + /** * Calls a callback function for all register reads. * @@ -153,7 +204,11 @@ static void reads_pair(struct rc_instruction * fullinst, rc_read_write_mask_fn void rc_for_all_reads_mask(struct rc_instruction * inst, rc_read_write_mask_fn cb, void * userdata) { if (inst->Type == RC_INSTRUCTION_NORMAL) { - reads_normal(inst, cb, userdata); + struct read_write_mask_data cb_data; + cb_data.UserData = userdata; + cb_data.Cb = cb; + + rc_for_all_reads_src(inst, reads_normal_callback, &cb_data); } else { reads_pair(inst, cb, userdata); } diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_dataflow.h b/src/mesa/drivers/dri/r300/compiler/radeon_dataflow.h index 795d9cc2b6..d10ae3c7b7 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_dataflow.h +++ b/src/mesa/drivers/dri/r300/compiler/radeon_dataflow.h @@ -33,6 +33,8 @@ struct radeon_compiler; struct rc_instruction; struct rc_swizzle_caps; +struct rc_src_register; +struct rc_pair_instruction_arg; /** @@ -49,6 +51,16 @@ typedef void (*rc_read_write_mask_fn)(void * userdata, struct rc_instruction * i void rc_for_all_reads_mask(struct rc_instruction * inst, rc_read_write_mask_fn cb, void * userdata); void rc_for_all_writes_mask(struct rc_instruction * inst, rc_read_write_mask_fn cb, void * userdata); +typedef void (*rc_read_src_fn)(void * userdata, struct rc_instruction * inst, + struct rc_src_register * src); +void rc_for_all_reads_src(struct rc_instruction * inst, rc_read_src_fn cb, + void * userdata); + +typedef void (*rc_pair_read_arg_fn)(void * userdata, + struct rc_instruction * inst, struct rc_pair_instruction_arg * arg); +void rc_pair_for_all_reads_arg(struct rc_instruction * inst, + rc_pair_read_arg_fn cb, void * userdata); + typedef void (*rc_remap_register_fn)(void * userdata, struct rc_instruction * inst, rc_register_file * pfile, unsigned int * pindex); void rc_remap_registers(struct rc_instruction * inst, rc_remap_register_fn cb, void * userdata); diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_dataflow_deadcode.c b/src/mesa/drivers/dri/r300/compiler/radeon_dataflow_deadcode.c index 9d17b4772a..87906f37b1 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_dataflow_deadcode.c +++ b/src/mesa/drivers/dri/r300/compiler/radeon_dataflow_deadcode.c @@ -106,10 +106,12 @@ static void push_loop(struct deadcode_state * s) static void push_branch(struct deadcode_state * s) { + struct branchinfo * branch; + memory_pool_array_reserve(&s->C->Pool, struct branchinfo, s->BranchStack, s->BranchStackSize, s->BranchStackReserved, 1); - struct branchinfo * branch = &s->BranchStack[s->BranchStackSize++]; + branch = &s->BranchStack[s->BranchStackSize++]; branch->HaveElse = 0; memcpy(&branch->StoreEndif, &s->R, sizeof(s->R)); } @@ -152,6 +154,7 @@ static void update_instruction(struct deadcode_state * s, struct rc_instruction const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->U.I.Opcode); struct instruction_state * insts = &s->Instructions[inst->IP]; unsigned int usedmask = 0; + unsigned int srcmasks[3]; if (opcode->HasDstReg) { unsigned char * pused = get_used_ptr(s, inst->U.I.DstReg.File, inst->U.I.DstReg.Index); @@ -180,7 +183,6 @@ static void update_instruction(struct deadcode_state * s, struct rc_instruction } } - unsigned int srcmasks[3]; rc_compute_sources_for_writemask(inst, usedmask, srcmasks); for(unsigned int src = 0; src < opcode->NumSrcRegs; ++src) { @@ -219,6 +221,7 @@ void rc_dataflow_deadcode(struct radeon_compiler * c, void *user) unsigned int nr_instructions; unsigned has_temp_reladdr_src = 0; rc_dataflow_mark_outputs_fn dce = (rc_dataflow_mark_outputs_fn)user; + unsigned int ip; /* Give up if there is relative addressing of destination operands. */ for(struct rc_instruction * inst = c->Program.Instructions.Next; @@ -349,12 +352,14 @@ void rc_dataflow_deadcode(struct radeon_compiler * c, void *user) update_instruction(&s, inst); } - unsigned int ip = 0; + ip = 0; for(struct rc_instruction * inst = c->Program.Instructions.Next; inst != &c->Program.Instructions; inst = inst->Next, ++ip) { const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->U.I.Opcode); int dead = 1; + unsigned int srcmasks[3]; + unsigned int usemask; if (!opcode->HasDstReg) { dead = 0; @@ -376,8 +381,7 @@ void rc_dataflow_deadcode(struct radeon_compiler * c, void *user) continue; } - unsigned int srcmasks[3]; - unsigned int usemask = s.Instructions[ip].WriteMask; + usemask = s.Instructions[ip].WriteMask; if (inst->U.I.WriteALUResult == RC_ALURESULT_X) usemask |= RC_MASK_X; diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_emulate_branches.c b/src/mesa/drivers/dri/r300/compiler/radeon_emulate_branches.c index 222e5b7e02..7bede344f3 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_emulate_branches.c +++ b/src/mesa/drivers/dri/r300/compiler/radeon_emulate_branches.c @@ -57,18 +57,21 @@ struct emulate_branch_state { static void handle_if(struct emulate_branch_state * s, struct rc_instruction * inst) { + struct branch_info * branch; + struct rc_instruction * inst_mov; + memory_pool_array_reserve(&s->C->Pool, struct branch_info, s->Branches, s->BranchCount, s->BranchReserved, 1); DBG("%s\n", __FUNCTION__); - struct branch_info * branch = &s->Branches[s->BranchCount++]; + branch = &s->Branches[s->BranchCount++]; memset(branch, 0, sizeof(struct branch_info)); branch->If = inst; /* Make a safety copy of the decision register, because we will need * it at ENDIF time and it might be overwritten in both branches. */ - struct rc_instruction * inst_mov = rc_insert_new_instruction(s->C, inst->Prev); + inst_mov = rc_insert_new_instruction(s->C, inst->Prev); inst_mov->U.I.Opcode = RC_OPCODE_MOV; inst_mov->U.I.DstReg.File = RC_FILE_TEMPORARY; inst_mov->U.I.DstReg.Index = rc_find_free_temporary(s->C); @@ -84,6 +87,8 @@ static void handle_if(struct emulate_branch_state * s, struct rc_instruction * i static void handle_else(struct emulate_branch_state * s, struct rc_instruction * inst) { + struct branch_info * branch; + if (!s->BranchCount) { rc_error(s->C, "Encountered ELSE outside of branches"); return; @@ -91,7 +96,7 @@ static void handle_else(struct emulate_branch_state * s, struct rc_instruction * DBG("%s\n", __FUNCTION__); - struct branch_info * branch = &s->Branches[s->BranchCount - 1]; + branch = &s->Branches[s->BranchCount - 1]; branch->Else = inst; } @@ -191,6 +196,10 @@ static void inject_cmp(struct emulate_branch_state * s, static void handle_endif(struct emulate_branch_state * s, struct rc_instruction * inst) { + struct branch_info * branch; + struct register_proxies IfProxies; + struct register_proxies ElseProxies; + if (!s->BranchCount) { rc_error(s->C, "Encountered ENDIF outside of branches"); return; @@ -198,9 +207,7 @@ static void handle_endif(struct emulate_branch_state * s, struct rc_instruction DBG("%s\n", __FUNCTION__); - struct branch_info * branch = &s->Branches[s->BranchCount - 1]; - struct register_proxies IfProxies; - struct register_proxies ElseProxies; + branch = &s->Branches[s->BranchCount - 1]; memset(&IfProxies, 0, sizeof(IfProxies)); memset(&ElseProxies, 0, sizeof(ElseProxies)); @@ -261,16 +268,19 @@ static void remap_output_function(void * userdata, struct rc_instruction * inst, */ static void fix_output_writes(struct emulate_branch_state * s, struct rc_instruction * inst) { + const struct rc_opcode_info * opcode; + if (!s->BranchCount) return; - const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->U.I.Opcode); + opcode = rc_get_opcode_info(inst->U.I.Opcode); if (!opcode->HasDstReg) return; if (inst->U.I.DstReg.File == RC_FILE_OUTPUT) { struct remap_output_data remap; + struct rc_instruction * inst_mov; remap.Output = inst->U.I.DstReg.Index; remap.Temporary = rc_find_free_temporary(s->C); @@ -281,7 +291,7 @@ static void fix_output_writes(struct emulate_branch_state * s, struct rc_instruc rc_remap_registers(inst, &remap_output_function, &remap); } - struct rc_instruction * inst_mov = rc_insert_new_instruction(s->C, s->C->Program.Instructions.Prev); + inst_mov = rc_insert_new_instruction(s->C, s->C->Program.Instructions.Prev); inst_mov->U.I.Opcode = RC_OPCODE_MOV; inst_mov->U.I.DstReg.File = RC_FILE_OUTPUT; inst_mov->U.I.DstReg.Index = remap.Output; @@ -299,12 +309,13 @@ static void fix_output_writes(struct emulate_branch_state * s, struct rc_instruc void rc_emulate_branches(struct radeon_compiler *c, void *user) { struct emulate_branch_state s; + struct rc_instruction * ptr; memset(&s, 0, sizeof(s)); s.C = c; /* Untypical loop because we may remove the current instruction */ - struct rc_instruction * ptr = c->Program.Instructions.Next; + ptr = c->Program.Instructions.Next; while(ptr != &c->Program.Instructions) { struct rc_instruction * inst = ptr; ptr = ptr->Next; diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_emulate_loops.c b/src/mesa/drivers/dri/r300/compiler/radeon_emulate_loops.c index cd4fcbabb9..205eecd112 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_emulate_loops.c +++ b/src/mesa/drivers/dri/r300/compiler/radeon_emulate_loops.c @@ -386,8 +386,6 @@ static int build_loop_info(struct radeon_compiler * c, struct loop_info * loop, case RC_OPCODE_SNE: break; default: - rc_error(c, "%s: expected conditional", - __FUNCTION__); return 0; } loop->Cond = loop->If->Prev; @@ -431,8 +429,10 @@ static int transform_loop(struct emulate_loop_state * s, loop = &s->Loops[s->LoopCount++]; - if (!build_loop_info(s->C, loop, inst)) + if (!build_loop_info(s->C, loop, inst)) { + rc_error(s->C, "Failed to build loop info\n"); return 0; + } if(try_unroll_loop(s->C, loop)){ return 1; @@ -511,11 +511,12 @@ void rc_emulate_loops(struct radeon_compiler *c, void *user) * loops are unrolled first. */ for( i = s->LoopCount - 1; i >= 0; i-- ){ + unsigned int iterations; + if(!s->Loops[i].EndLoop){ continue; } - unsigned int iterations = loop_max_possible_iterations( - s->C, &s->Loops[i]); + iterations = loop_max_possible_iterations(s->C, &s->Loops[i]); unroll_loop(s->C, &s->Loops[i], iterations); } } diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_optimize.c b/src/mesa/drivers/dri/r300/compiler/radeon_optimize.c index c15a9b1c45..41769e347e 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_optimize.c +++ b/src/mesa/drivers/dri/r300/compiler/radeon_optimize.c @@ -163,7 +163,8 @@ static void copy_propagate(struct radeon_compiler * c, struct rc_instruction * i if (inst_mov->U.I.DstReg.File != RC_FILE_TEMPORARY || inst_mov->U.I.DstReg.RelAddr || - inst_mov->U.I.WriteALUResult) + inst_mov->U.I.WriteALUResult || + inst_mov->U.I.SaturateMode) return; memset(&s, 0, sizeof(s)); @@ -410,27 +411,34 @@ static void constant_folding(struct radeon_compiler * c, struct rc_instruction * /* Replace 0.0, 1.0 and 0.5 immediates by their explicit swizzles */ for(unsigned int src = 0; src < opcode->NumSrcRegs; ++src) { + struct rc_constant * constant; + struct rc_src_register newsrc; + int have_real_reference; + if (inst->U.I.SrcReg[src].File != RC_FILE_CONSTANT || inst->U.I.SrcReg[src].RelAddr || inst->U.I.SrcReg[src].Index >= c->Program.Constants.Count) continue; - struct rc_constant * constant = + constant = &c->Program.Constants.Constants[inst->U.I.SrcReg[src].Index]; if (constant->Type != RC_CONSTANT_IMMEDIATE) continue; - struct rc_src_register newsrc = inst->U.I.SrcReg[src]; - int have_real_reference = 0; + newsrc = inst->U.I.SrcReg[src]; + have_real_reference = 0; for(unsigned int chan = 0; chan < 4; ++chan) { unsigned int swz = GET_SWZ(newsrc.Swizzle, chan); + unsigned int newswz; + float imm; + float baseimm; + if (swz >= 4) continue; - unsigned int newswz; - float imm = constant->u.Immediate[swz]; - float baseimm = imm; + imm = constant->u.Immediate[swz]; + baseimm = imm; if (imm < 0.0) baseimm = -baseimm; @@ -475,8 +483,8 @@ static void constant_folding(struct radeon_compiler * c, struct rc_instruction * } /** - * This function returns a writemask that indicates wich components are - * read by src and also written by dst. + * If src and dst use the same register, this function returns a writemask that + * indicates wich components are read by src. Otherwise zero is returned. */ static unsigned int src_reads_dst_mask(struct rc_src_register src, struct rc_dst_register dst) @@ -563,10 +571,18 @@ static int presub_helper( * s->Inst->U.I.DstReg, because if it does we must not * remove s->Inst. */ for(i = 0; i < info->NumSrcRegs; i++) { - if(s->Inst->U.I.DstReg.WriteMask != - src_reads_dst_mask(inst->U.I.SrcReg[i], - s->Inst->U.I.DstReg)) { - continue; + unsigned int mask = src_reads_dst_mask( + inst->U.I.SrcReg[i], s->Inst->U.I.DstReg); + /* XXX We could be more aggressive here using + * presubtract. It is okay if SrcReg[i] only reads + * from some of the mask components. */ + if(s->Inst->U.I.DstReg.WriteMask != mask) { + if (s->Inst->U.I.DstReg.WriteMask & mask) { + can_remove = 0; + break; + } else { + continue; + } } if (cant_sub || !can_use_presub) { can_remove = 0; @@ -626,6 +642,21 @@ static void presub_replace_add(struct peephole_state *s, inst->U.I.SrcReg[src_index].Index = presub_opcode; } +static int is_presub_candidate(struct rc_instruction * inst) +{ + const struct rc_opcode_info * info = rc_get_opcode_info(inst->U.I.Opcode); + unsigned int i; + + if (inst->U.I.PreSub.Opcode != RC_PRESUB_NONE || inst->U.I.SaturateMode) + return 0; + + for(i = 0; i < info->NumSrcRegs; i++) { + if (src_reads_dst_mask(inst->U.I.SrcReg[i], inst->U.I.DstReg)) + return 0; + } + return 1; +} + static int peephole_add_presub_add( struct radeon_compiler * c, struct rc_instruction * inst_add) @@ -635,10 +666,7 @@ static int peephole_add_presub_add( unsigned int i; struct peephole_state s; - if (inst_add->U.I.PreSub.Opcode != RC_PRESUB_NONE) - return 0; - - if (inst_add->U.I.SaturateMode) + if (!is_presub_candidate(inst_add)) return 0; if (inst_add->U.I.SrcReg[0].Swizzle != inst_add->U.I.SrcReg[1].Swizzle) @@ -705,10 +733,7 @@ static int peephole_add_presub_inv( unsigned int i, swz, mask; struct peephole_state s; - if (inst_add->U.I.PreSub.Opcode != RC_PRESUB_NONE) - return 0; - - if (inst_add->U.I.SaturateMode) + if (!is_presub_candidate(inst_add)) return 0; mask = inst_add->U.I.DstReg.WriteMask; diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_pair_regalloc.c b/src/mesa/drivers/dri/r300/compiler/radeon_pair_regalloc.c index c73845512f..91524f5ec6 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_pair_regalloc.c +++ b/src/mesa/drivers/dri/r300/compiler/radeon_pair_regalloc.c @@ -322,12 +322,13 @@ void rc_pair_regalloc_inputs_only(struct radeon_compiler *cc, void *user) { struct r300_fragment_program_compiler *c = (struct r300_fragment_program_compiler*)cc; struct regalloc_state s; + int temp_reg_offset; compute_live_intervals(cc, &s); c->AllocateHwInputs(c, &alloc_input, &s); - int temp_reg_offset = 0; + temp_reg_offset = 0; for (unsigned i = 0; i < RC_REGISTER_MAX_INDEX; i++) { if (s.Input[i].Allocated && temp_reg_offset <= s.Input[i].Index) temp_reg_offset = s.Input[i].Index + 1; diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_pair_schedule.c b/src/mesa/drivers/dri/r300/compiler/radeon_pair_schedule.c index 5269d65985..bfe3553460 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_pair_schedule.c +++ b/src/mesa/drivers/dri/r300/compiler/radeon_pair_schedule.c @@ -237,11 +237,12 @@ static void commit_alu_instruction(struct schedule_state * s, struct schedule_in static void emit_all_tex(struct schedule_state * s, struct rc_instruction * before) { struct schedule_instruction *readytex; + struct rc_instruction * inst_begin; assert(s->ReadyTEX); /* Node marker for R300 */ - struct rc_instruction * inst_begin = rc_insert_new_instruction(s->C, before->Prev); + inst_begin = rc_insert_new_instruction(s->C, before->Prev); inst_begin->U.I.Opcode = RC_OPCODE_BEGIN_TEX; /* Link texture instructions back in */ @@ -280,6 +281,8 @@ static int destructive_merge_instructions( struct rc_pair_instruction * alpha) { const struct rc_opcode_info * opcode; + const struct rc_opcode_info * rgb_info; + assert(rgb->Alpha.Opcode == RC_OPCODE_NOP); assert(alpha->RGB.Opcode == RC_OPCODE_NOP); @@ -288,8 +291,7 @@ static int destructive_merge_instructions( * src1. */ /* Merge the rgb presubtract registers. */ - const struct rc_opcode_info * rgb_info = - rc_get_opcode_info(rgb->RGB.Opcode); + rgb_info = rc_get_opcode_info(rgb->RGB.Opcode); if (alpha->RGB.Src[RC_PAIR_PRESUB_SRC].Used) { unsigned int srcp_src; unsigned int srcp_regs; @@ -301,9 +303,9 @@ static int destructive_merge_instructions( unsigned int arg; int free_source; unsigned int one_way = 0; - struct radeon_pair_instruction_source srcp = + struct rc_pair_instruction_source srcp = alpha->RGB.Src[srcp_src]; - struct radeon_pair_instruction_source temp; + struct rc_pair_instruction_source temp; /* 2nd arg of 1 means this is an rgb source. * 3rd arg of 0 means this is not an alpha source. */ free_source = rc_pair_alloc_source(rgb, 1, 0, @@ -366,9 +368,9 @@ static int destructive_merge_instructions( unsigned int arg; int free_source; unsigned int one_way = 0; - struct radeon_pair_instruction_source srcp = + struct rc_pair_instruction_source srcp = alpha->Alpha.Src[srcp_src]; - struct radeon_pair_instruction_source temp; + struct rc_pair_instruction_source temp; /* 2nd arg of 0 means this is not an rgb source. * 3rd arg of 1 means this is an alpha source. */ free_source = rc_pair_alloc_source(rgb, 0, 1, @@ -424,6 +426,7 @@ static int destructive_merge_instructions( unsigned int oldsrc = alpha->Alpha.Arg[arg].Source; rc_register_file file = 0; unsigned int index = 0; + int source; if (alpha->Alpha.Arg[arg].Swizzle < 3) { srcrgb = 1; @@ -435,7 +438,7 @@ static int destructive_merge_instructions( index = alpha->Alpha.Src[oldsrc].Index; } - int source = rc_pair_alloc_source(rgb, srcrgb, srcalpha, file, index); + source = rc_pair_alloc_source(rgb, srcrgb, srcalpha, file, index); if (source < 0) return 0; @@ -475,6 +478,12 @@ static int merge_instructions(struct rc_pair_instruction * rgb, struct rc_pair_i { struct rc_pair_instruction backup; + /*Instructions can't write output registers and ALU result at the + * same time. */ + if ((rgb->WriteALUResult && alpha->Alpha.OutputWriteMask) + || (rgb->RGB.OutputWriteMask && alpha->WriteALUResult)) { + return 0; + } memcpy(&backup, rgb, sizeof(struct rc_pair_instruction)); if (destructive_merge_instructions(rgb, alpha)) @@ -597,6 +606,7 @@ static void scan_read(void * data, struct rc_instruction * inst, { struct schedule_state * s = data; struct reg_value * v = get_reg_value(s, file, index, chan); + struct reg_value_reader * reader; if (!v) return; @@ -610,7 +620,7 @@ static void scan_read(void * data, struct rc_instruction * inst, DBG("%i: read %i[%i] chan %i\n", s->Current->Instruction->IP, file, index, chan); - struct reg_value_reader * reader = memory_pool_malloc(&s->C->Pool, sizeof(*reader)); + reader = memory_pool_malloc(&s->C->Pool, sizeof(*reader)); reader->Reader = s->Current; reader->Next = v->Readers; v->Readers = reader; @@ -630,13 +640,14 @@ static void scan_write(void * data, struct rc_instruction * inst, { struct schedule_state * s = data; struct reg_value ** pv = get_reg_valuep(s, file, index, chan); + struct reg_value * newv; if (!pv) return; DBG("%i: write %i[%i] chan %i\n", s->Current->Instruction->IP, file, index, chan); - struct reg_value * newv = memory_pool_malloc(&s->C->Pool, sizeof(*newv)); + newv = memory_pool_malloc(&s->C->Pool, sizeof(*newv)); memset(newv, 0, sizeof(*newv)); newv->Writer = s->Current; @@ -659,12 +670,13 @@ static void schedule_block(struct r300_fragment_program_compiler * c, struct rc_instruction * begin, struct rc_instruction * end) { struct schedule_state s; + unsigned int ip; memset(&s, 0, sizeof(s)); s.C = &c->Base; /* Scan instructions for data dependencies */ - unsigned int ip = 0; + ip = 0; for(struct rc_instruction * inst = begin; inst != end; inst = inst->Next) { s.Current = memory_pool_malloc(&c->Base.Pool, sizeof(*s.Current)); memset(s.Current, 0, sizeof(struct schedule_instruction)); @@ -716,12 +728,14 @@ void rc_pair_schedule(struct radeon_compiler *cc, void *user) struct r300_fragment_program_compiler *c = (struct r300_fragment_program_compiler*)cc; struct rc_instruction * inst = c->Base.Program.Instructions.Next; while(inst != &c->Base.Program.Instructions) { + struct rc_instruction * first; + if (is_controlflow(inst)) { inst = inst->Next; continue; } - struct rc_instruction * first = inst; + first = inst; while(inst != &c->Base.Program.Instructions && !is_controlflow(inst)) inst = inst->Next; diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_pair_translate.c b/src/mesa/drivers/dri/r300/compiler/radeon_pair_translate.c index ff82584466..c549be5218 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_pair_translate.c +++ b/src/mesa/drivers/dri/r300/compiler/radeon_pair_translate.c @@ -148,9 +148,12 @@ static void set_pair_instruction(struct r300_fragment_program_compiler *c, struct rc_pair_instruction * pair, struct rc_sub_instruction * inst) { + int needrgb, needalpha, istranscendent; + const struct rc_opcode_info * opcode; + int i; + memset(pair, 0, sizeof(struct rc_pair_instruction)); - int needrgb, needalpha, istranscendent; classify_instruction(inst, &needrgb, &needalpha, &istranscendent); if (needrgb) { @@ -167,8 +170,7 @@ static void set_pair_instruction(struct r300_fragment_program_compiler *c, pair->Alpha.Saturate = 1; } - const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->Opcode); - int i; + opcode = rc_get_opcode_info(inst->Opcode); /* Presubtract handling: * We need to make sure that the values used by the presubtract @@ -330,15 +332,18 @@ void rc_pair_translate(struct radeon_compiler *cc, void *user) for(struct rc_instruction * inst = c->Base.Program.Instructions.Next; inst != &c->Base.Program.Instructions; inst = inst->Next) { + const struct rc_opcode_info * opcode; + struct rc_sub_instruction copy; + if (inst->Type != RC_INSTRUCTION_NORMAL) continue; - const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->U.I.Opcode); + opcode = rc_get_opcode_info(inst->U.I.Opcode); if (opcode->HasTexture || opcode->IsFlowControl || opcode->Opcode == RC_OPCODE_KIL) continue; - struct rc_sub_instruction copy = inst->U.I; + copy = inst->U.I; check_opcode_support(c, ©); diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_program_alu.c b/src/mesa/drivers/dri/r300/compiler/radeon_program_alu.c index 4d3e26f28c..39408845d5 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_program_alu.c +++ b/src/mesa/drivers/dri/r300/compiler/radeon_program_alu.c @@ -874,13 +874,15 @@ int r300_transform_trig_simple(struct radeon_compiler* c, struct rc_instruction* inst, void* unused) { + unsigned int constants[2]; + unsigned int tempreg; + if (inst->U.I.Opcode != RC_OPCODE_COS && inst->U.I.Opcode != RC_OPCODE_SIN && inst->U.I.Opcode != RC_OPCODE_SCS) return 0; - unsigned int constants[2]; - unsigned int tempreg = rc_find_free_temporary(c); + tempreg = rc_find_free_temporary(c); sincos_constants(c, constants); @@ -918,6 +920,8 @@ int r300_transform_trig_simple(struct radeon_compiler* c, swizzle_wwww(srcreg(RC_FILE_TEMPORARY, tempreg)), constants); } else { + struct rc_dst_register dst; + emit3(c, inst->Prev, RC_OPCODE_MAD, 0, dstregtmpmask(tempreg, RC_MASK_XY), swizzle_xxxx(inst->U.I.SrcReg[0]), swizzle_zzzz(srcreg(RC_FILE_CONSTANT, constants[1])), @@ -929,7 +933,7 @@ int r300_transform_trig_simple(struct radeon_compiler* c, swizzle_wwww(srcreg(RC_FILE_CONSTANT, constants[1])), negate(swizzle_zzzz(srcreg(RC_FILE_CONSTANT, constants[0])))); - struct rc_dst_register dst = inst->U.I.DstReg; + dst = inst->U.I.DstReg; dst.WriteMask = inst->U.I.DstReg.WriteMask & RC_MASK_X; sin_approx(c, inst, dst, @@ -988,16 +992,16 @@ int radeonTransformTrigScale(struct radeon_compiler* c, struct rc_instruction* inst, void* unused) { - if (inst->U.I.Opcode != RC_OPCODE_COS && - inst->U.I.Opcode != RC_OPCODE_SIN && - inst->U.I.Opcode != RC_OPCODE_SCS) - return 0; - static const float RCP_2PI = 0.15915494309189535; unsigned int temp; unsigned int constant; unsigned int constant_swizzle; + if (inst->U.I.Opcode != RC_OPCODE_COS && + inst->U.I.Opcode != RC_OPCODE_SIN && + inst->U.I.Opcode != RC_OPCODE_SCS) + return 0; + temp = rc_find_free_temporary(c); constant = rc_constants_add_immediate_scalar(&c->Program.Constants, RCP_2PI, &constant_swizzle); @@ -1020,6 +1024,10 @@ int r300_transform_trig_scale_vertex(struct radeon_compiler *c, struct rc_instruction *inst, void *unused) { + static const float cons[4] = {0.15915494309189535, 0.5, 6.28318530717959, -3.14159265358979}; + unsigned int temp; + unsigned int constant; + if (inst->U.I.Opcode != RC_OPCODE_COS && inst->U.I.Opcode != RC_OPCODE_SIN && inst->U.I.Opcode != RC_OPCODE_SCS) @@ -1030,10 +1038,6 @@ int r300_transform_trig_scale_vertex(struct radeon_compiler *c, * repeat(x) = frac(x / 2PI + 0.5) * 2PI - PI */ - static const float cons[4] = {0.15915494309189535, 0.5, 6.28318530717959, -3.14159265358979}; - unsigned int temp; - unsigned int constant; - temp = rc_find_free_temporary(c); constant = rc_constants_add_immediate_vec4(&c->Program.Constants, cons); diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_program_pair.h b/src/mesa/drivers/dri/r300/compiler/radeon_program_pair.h index e0061e454b..01cdb15424 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_program_pair.h +++ b/src/mesa/drivers/dri/r300/compiler/radeon_program_pair.h @@ -55,52 +55,35 @@ struct radeon_compiler; */ #define RC_PAIR_PRESUB_SRC 3 -struct radeon_pair_instruction_source { +struct rc_pair_instruction_source { unsigned int Used:1; unsigned int File:3; unsigned int Index:RC_REGISTER_INDEX_BITS; }; -struct radeon_pair_instruction_rgb { - unsigned int Opcode:8; - unsigned int DestIndex:RC_REGISTER_INDEX_BITS; - unsigned int WriteMask:3; - unsigned int Target:2; - unsigned int OutputWriteMask:3; - unsigned int Saturate:1; - - struct radeon_pair_instruction_source Src[4]; - - struct { - unsigned int Source:2; - unsigned int Swizzle:9; - unsigned int Abs:1; - unsigned int Negate:1; - } Arg[3]; +struct rc_pair_instruction_arg { + unsigned int Source:2; + unsigned int Swizzle:9; + unsigned int Abs:1; + unsigned int Negate:1; }; -struct radeon_pair_instruction_alpha { +struct rc_pair_sub_instruction { unsigned int Opcode:8; unsigned int DestIndex:RC_REGISTER_INDEX_BITS; - unsigned int WriteMask:1; - unsigned int Target:2; - unsigned int OutputWriteMask:1; + unsigned int WriteMask:3; + unsigned int Target:2; + unsigned int OutputWriteMask:3; unsigned int DepthWriteMask:1; unsigned int Saturate:1; - struct radeon_pair_instruction_source Src[4]; - - struct { - unsigned int Source:2; - unsigned int Swizzle:3; - unsigned int Abs:1; - unsigned int Negate:1; - } Arg[3]; + struct rc_pair_instruction_source Src[4]; + struct rc_pair_instruction_arg Arg[3]; }; struct rc_pair_instruction { - struct radeon_pair_instruction_rgb RGB; - struct radeon_pair_instruction_alpha Alpha; + struct rc_pair_sub_instruction RGB; + struct rc_pair_sub_instruction Alpha; unsigned int WriteALUResult:2; unsigned int ALUResultCompare:3; @@ -108,7 +91,7 @@ struct rc_pair_instruction { }; typedef void (*rc_pair_foreach_src_fn) - (void *, struct radeon_pair_instruction_source *); + (void *, struct rc_pair_instruction_source *); typedef enum { RC_PAIR_SOURCE_NONE = 0, diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_program_tex.c b/src/mesa/drivers/dri/r300/compiler/radeon_program_tex.c index ddce590ee6..530afa5e08 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_program_tex.c +++ b/src/mesa/drivers/dri/r300/compiler/radeon_program_tex.c @@ -1,5 +1,6 @@ /* * Copyright (C) 2010 Corbin Simpson + * Copyright (C) 2010 Marek Olšák <maraeo@gmail.com> * * All Rights Reserved. * @@ -46,31 +47,61 @@ static struct rc_src_register shadow_ambient(struct r300_fragment_program_compil return reg; } -static void lower_texture_rect(struct r300_fragment_program_compiler *compiler, - struct rc_instruction *inst) +static void scale_texcoords(struct r300_fragment_program_compiler *compiler, + struct rc_instruction *inst, + unsigned state_constant) { - struct rc_instruction *inst_rect; + struct rc_instruction *inst_mov; + unsigned temp = rc_find_free_temporary(&compiler->Base); - if (inst->U.I.TexSrcTarget == RC_TEXTURE_RECT || - compiler->state.unit[inst->U.I.TexSrcUnit].non_normalized_coords) { - inst_rect = rc_insert_new_instruction(&compiler->Base, inst->Prev); + inst_mov = rc_insert_new_instruction(&compiler->Base, inst->Prev); - inst_rect->U.I.Opcode = RC_OPCODE_MUL; - inst_rect->U.I.DstReg.File = RC_FILE_TEMPORARY; - inst_rect->U.I.DstReg.Index = temp; - inst_rect->U.I.SrcReg[0] = inst->U.I.SrcReg[0]; - inst_rect->U.I.SrcReg[1].File = RC_FILE_CONSTANT; - inst_rect->U.I.SrcReg[1].Index = - rc_constants_add_state(&compiler->Base.Program.Constants, - RC_STATE_R300_TEXRECT_FACTOR, inst->U.I.TexSrcUnit); + inst_mov->U.I.Opcode = RC_OPCODE_MUL; + inst_mov->U.I.DstReg.File = RC_FILE_TEMPORARY; + inst_mov->U.I.DstReg.Index = temp; + inst_mov->U.I.SrcReg[0] = inst->U.I.SrcReg[0]; + inst_mov->U.I.SrcReg[1].File = RC_FILE_CONSTANT; + inst_mov->U.I.SrcReg[1].Index = + rc_constants_add_state(&compiler->Base.Program.Constants, + state_constant, inst->U.I.TexSrcUnit); - reset_srcreg(&inst->U.I.SrcReg[0]); - inst->U.I.SrcReg[0].File = RC_FILE_TEMPORARY; - inst->U.I.SrcReg[0].Index = temp; + reset_srcreg(&inst->U.I.SrcReg[0]); + inst->U.I.SrcReg[0].File = RC_FILE_TEMPORARY; + inst->U.I.SrcReg[0].Index = temp; +} - inst->U.I.TexSrcTarget = RC_TEXTURE_2D; - } +static void projective_divide(struct r300_fragment_program_compiler *compiler, + struct rc_instruction *inst) +{ + struct rc_instruction *inst_mul, *inst_rcp; + + unsigned temp = rc_find_free_temporary(&compiler->Base); + + inst_rcp = rc_insert_new_instruction(&compiler->Base, inst->Prev); + inst_rcp->U.I.Opcode = RC_OPCODE_RCP; + inst_rcp->U.I.DstReg.File = RC_FILE_TEMPORARY; + inst_rcp->U.I.DstReg.Index = temp; + inst_rcp->U.I.DstReg.WriteMask = RC_MASK_W; + inst_rcp->U.I.SrcReg[0] = inst->U.I.SrcReg[0]; + /* Because the input can be arbitrarily swizzled, + * read the component mapped to W. */ + inst_rcp->U.I.SrcReg[0].Swizzle = + RC_MAKE_SWIZZLE_SMEAR(GET_SWZ(inst->U.I.SrcReg[0].Swizzle, 3)); + + inst_mul = rc_insert_new_instruction(&compiler->Base, inst->Prev); + inst_mul->U.I.Opcode = RC_OPCODE_MUL; + inst_mul->U.I.DstReg.File = RC_FILE_TEMPORARY; + inst_mul->U.I.DstReg.Index = temp; + inst_mul->U.I.SrcReg[0] = inst->U.I.SrcReg[0]; + inst_mul->U.I.SrcReg[1].File = RC_FILE_TEMPORARY; + inst_mul->U.I.SrcReg[1].Index = temp; + inst_mul->U.I.SrcReg[1].Swizzle = RC_SWIZZLE_WWWW; + + reset_srcreg(&inst->U.I.SrcReg[0]); + inst->U.I.Opcode = RC_OPCODE_TEX; + inst->U.I.SrcReg[0].File = RC_FILE_TEMPORARY; + inst->U.I.SrcReg[0].Index = temp; } /** @@ -88,6 +119,9 @@ int radeonTransformTEX( { struct r300_fragment_program_compiler *compiler = (struct r300_fragment_program_compiler*)data; + rc_wrap_mode wrapmode = compiler->state.unit[inst->U.I.TexSrcUnit].wrap_mode; + int is_rect = inst->U.I.TexSrcTarget == RC_TEXTURE_RECT || + compiler->state.unit[inst->U.I.TexSrcUnit].non_normalized_coords; if (inst->U.I.Opcode != RC_OPCODE_TEX && inst->U.I.Opcode != RC_OPCODE_TXB && @@ -141,16 +175,18 @@ int radeonTransformTEX( inst_rcp->U.I.DstReg.Index = tmp_recip_w; inst_rcp->U.I.DstReg.WriteMask = RC_MASK_W; inst_rcp->U.I.SrcReg[0] = inst->U.I.SrcReg[0]; - inst_rcp->U.I.SrcReg[0].Swizzle = RC_SWIZZLE_WWWW; + inst_rcp->U.I.SrcReg[0].Swizzle = + RC_MAKE_SWIZZLE_SMEAR(GET_SWZ(inst->U.I.SrcReg[0].Swizzle, 3)); } - /* Perspective-divide r by W (if it's TXP) and add the texture sample (see below). */ + /* Perspective-divide Z by W (if it's TXP) and add the texture sample (see below). */ tmp_sum = rc_find_free_temporary(c); inst_mad = rc_insert_new_instruction(c, inst_rcp ? inst_rcp : inst); inst_mad->U.I.DstReg.File = RC_FILE_TEMPORARY; inst_mad->U.I.DstReg.Index = tmp_sum; inst_mad->U.I.SrcReg[0] = inst->U.I.SrcReg[0]; - inst_mad->U.I.SrcReg[0].Swizzle = RC_SWIZZLE_ZZZZ; + inst_mad->U.I.SrcReg[0].Swizzle = + RC_MAKE_SWIZZLE_SMEAR(GET_SWZ(inst->U.I.SrcReg[0].Swizzle, 2)); if (inst->U.I.Opcode == RC_OPCODE_TXP) { inst_mad->U.I.Opcode = RC_OPCODE_MAD; inst_mad->U.I.SrcReg[1].File = RC_FILE_TEMPORARY; @@ -206,11 +242,22 @@ int radeonTransformTEX( } } - /* Texture wrap modes don't work on NPOT textures or texrects. - * - * The game plan is simple. We have two flags, fake_npot and - * non_normalized_coords, as well as a tex target. The RECT tex target - * will make the emitted code use non-scaled texcoords. + /* R300 cannot sample from rectangles and the wrap mode fallback needs + * normalized coordinates anyway. */ + if (inst->U.I.Opcode != RC_OPCODE_KIL && + is_rect && (!c->is_r500 || wrapmode != RC_WRAP_NONE)) { + scale_texcoords(compiler, inst, RC_STATE_R300_TEXRECT_FACTOR); + inst->U.I.TexSrcTarget = RC_TEXTURE_2D; + } + + /* Divide by W if needed. */ + if (inst->U.I.Opcode == RC_OPCODE_TXP && + (wrapmode == RC_WRAP_REPEAT || wrapmode == RC_WRAP_MIRRORED_REPEAT || + compiler->state.unit[inst->U.I.TexSrcUnit].clamp_and_scale_before_fetch)) { + projective_divide(compiler, inst); + } + + /* Texture wrap modes don't work on NPOT textures. * * Non-wrapped/clamped texcoords with NPOT are free in HW. Repeat and * mirroring are not. If we need to repeat, we do: @@ -235,128 +282,140 @@ int radeonTransformTEX( * ~ C & M. ;) */ if (inst->U.I.Opcode != RC_OPCODE_KIL && - (inst->U.I.TexSrcTarget == RC_TEXTURE_RECT || - compiler->state.unit[inst->U.I.TexSrcUnit].fake_npot || - compiler->state.unit[inst->U.I.TexSrcUnit].non_normalized_coords)) { - rc_wrap_mode wrapmode = compiler->state.unit[inst->U.I.TexSrcUnit].wrap_mode; - - /* R300 cannot sample from rectangles. */ - if (!c->is_r500) { - lower_texture_rect(compiler, inst); - } - - if (compiler->state.unit[inst->U.I.TexSrcUnit].fake_npot && - wrapmode != RC_WRAP_NONE) { + wrapmode != RC_WRAP_NONE) { + struct rc_instruction *inst_mov; + unsigned temp = rc_find_free_temporary(c); + + if (wrapmode == RC_WRAP_REPEAT) { + /* Both instructions will be paired up. */ + struct rc_instruction *inst_frc = rc_insert_new_instruction(c, inst->Prev); + + inst_frc->U.I.Opcode = RC_OPCODE_FRC; + inst_frc->U.I.DstReg.File = RC_FILE_TEMPORARY; + inst_frc->U.I.DstReg.Index = temp; + inst_frc->U.I.DstReg.WriteMask = RC_MASK_XYZ; + inst_frc->U.I.SrcReg[0] = inst->U.I.SrcReg[0]; + } else if (wrapmode == RC_WRAP_MIRRORED_REPEAT) { + /* + * Function: + * f(v) = 1 - abs(frac(v * 0.5) * 2 - 1) + * + * Code: + * MUL temp, src0, 0.5 + * FRC temp, temp + * MAD temp, temp, 2, -1 + * ADD temp, 1, -abs(temp) + */ + + struct rc_instruction *inst_mul, *inst_frc, *inst_mad, *inst_add; + unsigned two, two_swizzle; + + inst_mul = rc_insert_new_instruction(c, inst->Prev); + + inst_mul->U.I.Opcode = RC_OPCODE_MUL; + inst_mul->U.I.DstReg.File = RC_FILE_TEMPORARY; + inst_mul->U.I.DstReg.Index = temp; + inst_mul->U.I.DstReg.WriteMask = RC_MASK_XYZ; + inst_mul->U.I.SrcReg[0] = inst->U.I.SrcReg[0]; + inst_mul->U.I.SrcReg[1].Swizzle = RC_SWIZZLE_HHHH; + + inst_frc = rc_insert_new_instruction(c, inst->Prev); + + inst_frc->U.I.Opcode = RC_OPCODE_FRC; + inst_frc->U.I.DstReg.File = RC_FILE_TEMPORARY; + inst_frc->U.I.DstReg.Index = temp; + inst_frc->U.I.DstReg.WriteMask = RC_MASK_XYZ; + inst_frc->U.I.SrcReg[0].File = RC_FILE_TEMPORARY; + inst_frc->U.I.SrcReg[0].Index = temp; + inst_frc->U.I.SrcReg[0].Swizzle = RC_SWIZZLE_XYZ0; + + two = rc_constants_add_immediate_scalar(&c->Program.Constants, 2, &two_swizzle); + inst_mad = rc_insert_new_instruction(c, inst->Prev); + + inst_mad->U.I.Opcode = RC_OPCODE_MAD; + inst_mad->U.I.DstReg.File = RC_FILE_TEMPORARY; + inst_mad->U.I.DstReg.Index = temp; + inst_mad->U.I.DstReg.WriteMask = RC_MASK_XYZ; + inst_mad->U.I.SrcReg[0].File = RC_FILE_TEMPORARY; + inst_mad->U.I.SrcReg[0].Index = temp; + inst_mad->U.I.SrcReg[0].Swizzle = RC_SWIZZLE_XYZ0; + inst_mad->U.I.SrcReg[1].File = RC_FILE_CONSTANT; + inst_mad->U.I.SrcReg[1].Index = two; + inst_mad->U.I.SrcReg[1].Swizzle = two_swizzle; + inst_mad->U.I.SrcReg[2].Swizzle = RC_SWIZZLE_1111; + inst_mad->U.I.SrcReg[2].Negate = RC_MASK_XYZ; + + inst_add = rc_insert_new_instruction(c, inst->Prev); + + inst_add->U.I.Opcode = RC_OPCODE_ADD; + inst_add->U.I.DstReg.File = RC_FILE_TEMPORARY; + inst_add->U.I.DstReg.Index = temp; + inst_add->U.I.DstReg.WriteMask = RC_MASK_XYZ; + inst_add->U.I.SrcReg[0].Swizzle = RC_SWIZZLE_1111; + inst_add->U.I.SrcReg[1].File = RC_FILE_TEMPORARY; + inst_add->U.I.SrcReg[1].Index = temp; + inst_add->U.I.SrcReg[1].Swizzle = RC_SWIZZLE_XYZ0; + inst_add->U.I.SrcReg[1].Abs = 1; + inst_add->U.I.SrcReg[1].Negate = RC_MASK_XYZ; + } else if (wrapmode == RC_WRAP_MIRRORED_CLAMP) { + /* + * Mirrored clamp modes are bloody simple, we just use abs + * to mirror [0, 1] into [-1, 0]. This works for + * all modes i.e. CLAMP, CLAMP_TO_EDGE, and CLAMP_TO_BORDER. + */ struct rc_instruction *inst_mov; - unsigned temp = rc_find_free_temporary(c); - - /* For NPOT fallback, we need normalized coordinates anyway. */ - if (c->is_r500) { - lower_texture_rect(compiler, inst); - } - - if (wrapmode == RC_WRAP_REPEAT) { - /* Both instructions will be paired up. */ - struct rc_instruction *inst_frc = rc_insert_new_instruction(c, inst->Prev); - - inst_frc->U.I.Opcode = RC_OPCODE_FRC; - inst_frc->U.I.DstReg.File = RC_FILE_TEMPORARY; - inst_frc->U.I.DstReg.Index = temp; - inst_frc->U.I.DstReg.WriteMask = RC_MASK_XYZ; - inst_frc->U.I.SrcReg[0] = inst->U.I.SrcReg[0]; - } else if (wrapmode == RC_WRAP_MIRRORED_REPEAT) { - /* - * Function: - * f(v) = 1 - abs(frac(v * 0.5) * 2 - 1) - * - * Code: - * MUL temp, src0, 0.5 - * FRC temp, temp - * MAD temp, temp, 2, -1 - * ADD temp, 1, -abs(temp) - */ - - struct rc_instruction *inst_mul, *inst_frc, *inst_mad, *inst_add; - unsigned two, two_swizzle; - - inst_mul = rc_insert_new_instruction(c, inst->Prev); - - inst_mul->U.I.Opcode = RC_OPCODE_MUL; - inst_mul->U.I.DstReg.File = RC_FILE_TEMPORARY; - inst_mul->U.I.DstReg.Index = temp; - inst_mul->U.I.DstReg.WriteMask = RC_MASK_XYZ; - inst_mul->U.I.SrcReg[0] = inst->U.I.SrcReg[0]; - inst_mul->U.I.SrcReg[1].Swizzle = RC_SWIZZLE_HHHH; - - inst_frc = rc_insert_new_instruction(c, inst->Prev); - - inst_frc->U.I.Opcode = RC_OPCODE_FRC; - inst_frc->U.I.DstReg.File = RC_FILE_TEMPORARY; - inst_frc->U.I.DstReg.Index = temp; - inst_frc->U.I.DstReg.WriteMask = RC_MASK_XYZ; - inst_frc->U.I.SrcReg[0].File = RC_FILE_TEMPORARY; - inst_frc->U.I.SrcReg[0].Index = temp; - inst_frc->U.I.SrcReg[0].Swizzle = RC_SWIZZLE_XYZ0; - - two = rc_constants_add_immediate_scalar(&c->Program.Constants, 2, &two_swizzle); - inst_mad = rc_insert_new_instruction(c, inst->Prev); - - inst_mad->U.I.Opcode = RC_OPCODE_MAD; - inst_mad->U.I.DstReg.File = RC_FILE_TEMPORARY; - inst_mad->U.I.DstReg.Index = temp; - inst_mad->U.I.DstReg.WriteMask = RC_MASK_XYZ; - inst_mad->U.I.SrcReg[0].File = RC_FILE_TEMPORARY; - inst_mad->U.I.SrcReg[0].Index = temp; - inst_mad->U.I.SrcReg[0].Swizzle = RC_SWIZZLE_XYZ0; - inst_mad->U.I.SrcReg[1].File = RC_FILE_CONSTANT; - inst_mad->U.I.SrcReg[1].Index = two; - inst_mad->U.I.SrcReg[1].Swizzle = two_swizzle; - inst_mad->U.I.SrcReg[2].Swizzle = RC_SWIZZLE_1111; - inst_mad->U.I.SrcReg[2].Negate = RC_MASK_XYZ; - - inst_add = rc_insert_new_instruction(c, inst->Prev); - - inst_add->U.I.Opcode = RC_OPCODE_ADD; - inst_add->U.I.DstReg.File = RC_FILE_TEMPORARY; - inst_add->U.I.DstReg.Index = temp; - inst_add->U.I.DstReg.WriteMask = RC_MASK_XYZ; - inst_add->U.I.SrcReg[0].Swizzle = RC_SWIZZLE_1111; - inst_add->U.I.SrcReg[1].File = RC_FILE_TEMPORARY; - inst_add->U.I.SrcReg[1].Index = temp; - inst_add->U.I.SrcReg[1].Swizzle = RC_SWIZZLE_XYZ0; - inst_add->U.I.SrcReg[1].Abs = 1; - inst_add->U.I.SrcReg[1].Negate = RC_MASK_XYZ; - } else if (wrapmode == RC_WRAP_MIRRORED_CLAMP) { - /* - * Mirrored clamp modes are bloody simple, we just use abs - * to mirror [0, 1] into [-1, 0]. This works for - * all modes i.e. CLAMP, CLAMP_TO_EDGE, and CLAMP_TO_BORDER. - */ - struct rc_instruction *inst_mov; - - inst_mov = rc_insert_new_instruction(c, inst->Prev); - - inst_mov->U.I.Opcode = RC_OPCODE_MOV; - inst_mov->U.I.DstReg.File = RC_FILE_TEMPORARY; - inst_mov->U.I.DstReg.Index = temp; - inst_mov->U.I.DstReg.WriteMask = RC_MASK_XYZ; - inst_mov->U.I.SrcReg[0] = inst->U.I.SrcReg[0]; - inst_mov->U.I.SrcReg[0].Abs = 1; - } - /* Preserve W for TXP/TXB. */ inst_mov = rc_insert_new_instruction(c, inst->Prev); inst_mov->U.I.Opcode = RC_OPCODE_MOV; inst_mov->U.I.DstReg.File = RC_FILE_TEMPORARY; inst_mov->U.I.DstReg.Index = temp; - inst_mov->U.I.DstReg.WriteMask = RC_MASK_W; + inst_mov->U.I.DstReg.WriteMask = RC_MASK_XYZ; inst_mov->U.I.SrcReg[0] = inst->U.I.SrcReg[0]; - - reset_srcreg(&inst->U.I.SrcReg[0]); - inst->U.I.SrcReg[0].File = RC_FILE_TEMPORARY; - inst->U.I.SrcReg[0].Index = temp; + inst_mov->U.I.SrcReg[0].Abs = 1; } + + /* Preserve W for TXP/TXB. */ + inst_mov = rc_insert_new_instruction(c, inst->Prev); + + inst_mov->U.I.Opcode = RC_OPCODE_MOV; + inst_mov->U.I.DstReg.File = RC_FILE_TEMPORARY; + inst_mov->U.I.DstReg.Index = temp; + inst_mov->U.I.DstReg.WriteMask = RC_MASK_W; + inst_mov->U.I.SrcReg[0] = inst->U.I.SrcReg[0]; + + reset_srcreg(&inst->U.I.SrcReg[0]); + inst->U.I.SrcReg[0].File = RC_FILE_TEMPORARY; + inst->U.I.SrcReg[0].Index = temp; + } + + if (inst->U.I.Opcode != RC_OPCODE_KIL && + compiler->state.unit[inst->U.I.TexSrcUnit].clamp_and_scale_before_fetch) { + struct rc_instruction *inst_mov; + unsigned temp = rc_find_free_temporary(c); + + /* Saturate XYZ. */ + inst_mov = rc_insert_new_instruction(c, inst->Prev); + inst_mov->U.I.Opcode = RC_OPCODE_MOV; + inst_mov->U.I.SaturateMode = RC_SATURATE_ZERO_ONE; + inst_mov->U.I.DstReg.File = RC_FILE_TEMPORARY; + inst_mov->U.I.DstReg.Index = temp; + inst_mov->U.I.DstReg.WriteMask = RC_MASK_XYZ; + inst_mov->U.I.SrcReg[0] = inst->U.I.SrcReg[0]; + + /* Copy W. */ + inst_mov = rc_insert_new_instruction(c, inst->Prev); + inst_mov->U.I.Opcode = RC_OPCODE_MOV; + inst_mov->U.I.DstReg.File = RC_FILE_TEMPORARY; + inst_mov->U.I.DstReg.Index = temp; + inst_mov->U.I.DstReg.WriteMask = RC_MASK_W; + inst_mov->U.I.SrcReg[0] = inst->U.I.SrcReg[0]; + + reset_srcreg(&inst->U.I.SrcReg[0]); + inst->U.I.SrcReg[0].File = RC_FILE_TEMPORARY; + inst->U.I.SrcReg[0].Index = temp; + + scale_texcoords(compiler, inst, RC_STATE_R300_TEXSCALE_FACTOR); } /* Cannot write texture to output registers (all chips) or with masks (non-r500) */ diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_remove_constants.c b/src/mesa/drivers/dri/r300/compiler/radeon_remove_constants.c index d111319d3d..d6c808ad81 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_remove_constants.c +++ b/src/mesa/drivers/dri/r300/compiler/radeon_remove_constants.c @@ -28,107 +28,105 @@ #include "radeon_remove_constants.h" #include "radeon_dataflow.h" +struct mark_used_data { + unsigned char * const_used; + unsigned * has_rel_addr; +}; + static void remap_regs(void * userdata, struct rc_instruction * inst, rc_register_file * pfile, unsigned int * pindex) { - unsigned *inv_remap_table = userdata; + unsigned *inv_remap_table = userdata; + + if (*pfile == RC_FILE_CONSTANT) { + *pindex = inv_remap_table[*pindex]; + } +} + +static void mark_used(void * userdata, struct rc_instruction * inst, + struct rc_src_register * src) +{ + struct mark_used_data * d = userdata; - if (*pfile == RC_FILE_CONSTANT) { - *pindex = inv_remap_table[*pindex]; - } + if (src->File == RC_FILE_CONSTANT) { + if (src->RelAddr) { + *d->has_rel_addr = 1; + } else { + d->const_used[src->Index] = 1; + } + } } void rc_remove_unused_constants(struct radeon_compiler *c, void *user) { unsigned **out_remap_table = (unsigned**)user; - unsigned char *const_used; - unsigned *remap_table; - unsigned *inv_remap_table; - unsigned has_rel_addr = 0; - unsigned is_identity = 1; + unsigned char *const_used; + unsigned *remap_table; + unsigned *inv_remap_table; + unsigned has_rel_addr = 0; + unsigned is_identity = 1; unsigned are_externals_remapped = 0; - struct rc_constant *constants = c->Program.Constants.Constants; + struct rc_constant *constants = c->Program.Constants.Constants; + struct mark_used_data d; + unsigned new_count; + + if (!c->Program.Constants.Count) { + *out_remap_table = NULL; + return; + } - if (!c->Program.Constants.Count) { - *out_remap_table = NULL; - return; - } + const_used = malloc(c->Program.Constants.Count); + memset(const_used, 0, c->Program.Constants.Count); - const_used = malloc(c->Program.Constants.Count); - memset(const_used, 0, c->Program.Constants.Count); + d.const_used = const_used; + d.has_rel_addr = &has_rel_addr; /* Pass 1: Mark used constants. */ - for (struct rc_instruction *inst = c->Program.Instructions.Next; - inst != &c->Program.Instructions; inst = inst->Next) { - const struct rc_opcode_info *opcode = rc_get_opcode_info(inst->U.I.Opcode); - - /* XXX: This loop and the if statement after it should be - * replaced by a call to one of the rc_for_all_reads_* functions. - * The reason it does not use one of those functions now is - * because none of them have RelAddr as an argument. */ - for (unsigned i = 0; i < opcode->NumSrcRegs; i++) { - if (inst->U.I.SrcReg[i].File == RC_FILE_CONSTANT) { - if (inst->U.I.SrcReg[i].RelAddr) { - has_rel_addr = 1; - } else { - const_used[inst->U.I.SrcReg[i].Index] = 1; - } - } - } - if (inst->U.I.PreSub.Opcode != RC_PRESUB_NONE) { - unsigned int i; - unsigned int srcp_regs = rc_presubtract_src_reg_count( - inst->U.I.PreSub.Opcode); - for( i = 0; i < srcp_regs; i++) { - if (inst->U.I.PreSub.SrcReg[i].File == - RC_FILE_CONSTANT) { - const_used[ - inst->U.I.PreSub.SrcReg[i].Index] = 1; - } - } - } - } + for (struct rc_instruction *inst = c->Program.Instructions.Next; + inst != &c->Program.Instructions; inst = inst->Next) { + rc_for_all_reads_src(inst, mark_used, &d); + } - /* Pass 2: If there is relative addressing, mark all externals as used. */ - if (has_rel_addr) { - for (unsigned i = 0; i < c->Program.Constants.Count; i++) - if (constants[i].Type == RC_CONSTANT_EXTERNAL) - const_used[i] = 1; - } + /* Pass 2: If there is relative addressing, mark all externals as used. */ + if (has_rel_addr) { + for (unsigned i = 0; i < c->Program.Constants.Count; i++) + if (constants[i].Type == RC_CONSTANT_EXTERNAL) + const_used[i] = 1; + } - /* Pass 3: Make the remapping table and remap constants. + /* Pass 3: Make the remapping table and remap constants. * This pass removes unused constants simply by overwriting them by other constants. */ - remap_table = malloc(c->Program.Constants.Count * sizeof(unsigned)); - inv_remap_table = malloc(c->Program.Constants.Count * sizeof(unsigned)); - unsigned new_count = 0; + remap_table = malloc(c->Program.Constants.Count * sizeof(unsigned)); + inv_remap_table = malloc(c->Program.Constants.Count * sizeof(unsigned)); + new_count = 0; - for (unsigned i = 0; i < c->Program.Constants.Count; i++) { - if (const_used[i]) { - remap_table[new_count] = i; - inv_remap_table[i] = new_count; + for (unsigned i = 0; i < c->Program.Constants.Count; i++) { + if (const_used[i]) { + remap_table[new_count] = i; + inv_remap_table[i] = new_count; - if (i != new_count) { + if (i != new_count) { if (constants[i].Type == RC_CONSTANT_EXTERNAL) are_externals_remapped = 1; constants[new_count] = constants[i]; - is_identity = 0; - } - new_count++; - } - } + is_identity = 0; + } + new_count++; + } + } /* is_identity ==> new_count == old_count * !is_identity ==> new_count < old_count */ assert( is_identity || new_count < c->Program.Constants.Count); assert(!(has_rel_addr && are_externals_remapped)); - /* Pass 4: Redirect reads of all constants to their new locations. */ - if (!is_identity) { - for (struct rc_instruction *inst = c->Program.Instructions.Next; - inst != &c->Program.Instructions; inst = inst->Next) { - rc_remap_registers(inst, remap_regs, inv_remap_table); - } + /* Pass 4: Redirect reads of all constants to their new locations. */ + if (!is_identity) { + for (struct rc_instruction *inst = c->Program.Instructions.Next; + inst != &c->Program.Instructions; inst = inst->Next) { + rc_remap_registers(inst, remap_regs, inv_remap_table); + } } @@ -138,12 +136,15 @@ void rc_remove_unused_constants(struct radeon_compiler *c, void *user) c->Program.Constants.Count = new_count; if (are_externals_remapped) { - *out_remap_table = remap_table; - } else { - *out_remap_table = NULL; - free(remap_table); - } - - free(const_used); - free(inv_remap_table); + *out_remap_table = remap_table; + } else { + *out_remap_table = NULL; + free(remap_table); + } + + free(const_used); + free(inv_remap_table); + + if (c->Debug) + rc_constants_print(&c->Program.Constants); } diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_rename_regs.c b/src/mesa/drivers/dri/r300/compiler/radeon_rename_regs.c index 31d25f9ab8..60e228be5b 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_rename_regs.c +++ b/src/mesa/drivers/dri/r300/compiler/radeon_rename_regs.c @@ -104,13 +104,14 @@ void rc_rename_regs(struct radeon_compiler *c, void *user) inst != &c->Program.Instructions; inst = inst->Next) { const struct rc_opcode_info * info; + unsigned int old_index, temp_index; + struct rc_dst_register * dst; if(inst->Type != RC_INSTRUCTION_NORMAL) { rc_error(c, "%s only works with normal instructions.", __FUNCTION__); return; } - unsigned int old_index, temp_index; - struct rc_dst_register * dst = &inst->U.I.DstReg; + dst = &inst->U.I.DstReg; info = rc_get_opcode_info(inst->U.I.Opcode); if(!info->HasDstReg || dst->File != RC_FILE_TEMPORARY) { continue; diff --git a/src/mesa/drivers/dri/r300/r300_blit.c b/src/mesa/drivers/dri/r300/r300_blit.c index 74aef765e3..9fd8e8fde5 100644 --- a/src/mesa/drivers/dri/r300/r300_blit.c +++ b/src/mesa/drivers/dri/r300/r300_blit.c @@ -569,7 +569,7 @@ unsigned r300_check_blit(gl_format dst_format) * @param[in] height region height * @param[in] flip_y set if y coords of the source image need to be flipped */ -unsigned r300_blit(GLcontext *ctx, +unsigned r300_blit(struct gl_context *ctx, struct radeon_bo *src_bo, intptr_t src_offset, gl_format src_mesaformat, diff --git a/src/mesa/drivers/dri/r300/r300_blit.h b/src/mesa/drivers/dri/r300/r300_blit.h index 39b157a57b..286dbe1856 100644 --- a/src/mesa/drivers/dri/r300/r300_blit.h +++ b/src/mesa/drivers/dri/r300/r300_blit.h @@ -32,7 +32,7 @@ void r300_blit_init(struct r300_context *r300); unsigned r300_check_blit(gl_format mesa_format); -unsigned r300_blit(GLcontext *ctx, +unsigned r300_blit(struct gl_context *ctx, struct radeon_bo *src_bo, intptr_t src_offset, gl_format src_mesaformat, diff --git a/src/mesa/drivers/dri/r300/r300_cmdbuf.c b/src/mesa/drivers/dri/r300/r300_cmdbuf.c index c40802aec6..8a2f5ce021 100644 --- a/src/mesa/drivers/dri/r300/r300_cmdbuf.c +++ b/src/mesa/drivers/dri/r300/r300_cmdbuf.c @@ -69,7 +69,7 @@ static unsigned packet0_count(r300ContextPtr r300, uint32_t *pkt) #define vpu_count(ptr) (((drm_r300_cmd_header_t*)(ptr))->vpu.count) #define r500fp_count(ptr) (((drm_r300_cmd_header_t*)(ptr))->r500fp.count) -static int check_vpu(GLcontext *ctx, struct radeon_state_atom *atom) +static int check_vpu(struct gl_context *ctx, struct radeon_state_atom *atom) { r300ContextPtr r300 = R300_CONTEXT(ctx); int cnt; @@ -83,7 +83,7 @@ static int check_vpu(GLcontext *ctx, struct radeon_state_atom *atom) return cnt ? (cnt * 4) + extra : 0; } -static int check_vpp(GLcontext *ctx, struct radeon_state_atom *atom) +static int check_vpp(struct gl_context *ctx, struct radeon_state_atom *atom) { r300ContextPtr r300 = R300_CONTEXT(ctx); int cnt; @@ -114,7 +114,7 @@ void r300_emit_vpu(struct r300_context *r300, END_BATCH(); } -static void emit_vpu_state(GLcontext *ctx, struct radeon_state_atom * atom) +static void emit_vpu_state(struct gl_context *ctx, struct radeon_state_atom * atom) { r300ContextPtr r300 = R300_CONTEXT(ctx); drm_r300_cmd_header_t cmd; @@ -126,7 +126,7 @@ static void emit_vpu_state(GLcontext *ctx, struct radeon_state_atom * atom) r300_emit_vpu(r300, &atom->cmd[1], vpu_count(atom->cmd) * 4, addr); } -static void emit_vpp_state(GLcontext *ctx, struct radeon_state_atom * atom) +static void emit_vpp_state(struct gl_context *ctx, struct radeon_state_atom * atom) { r300ContextPtr r300 = R300_CONTEXT(ctx); drm_r300_cmd_header_t cmd; @@ -158,7 +158,7 @@ void r500_emit_fp(struct r300_context *r300, END_BATCH(); } -static void emit_r500fp_atom(GLcontext *ctx, struct radeon_state_atom * atom) +static void emit_r500fp_atom(struct gl_context *ctx, struct radeon_state_atom * atom) { r300ContextPtr r300 = R300_CONTEXT(ctx); drm_r300_cmd_header_t cmd; @@ -179,7 +179,7 @@ static void emit_r500fp_atom(GLcontext *ctx, struct radeon_state_atom * atom) r500_emit_fp(r300, &atom->cmd[1], count, addr, type, clamp); } -static int check_tex_offsets(GLcontext *ctx, struct radeon_state_atom * atom) +static int check_tex_offsets(struct gl_context *ctx, struct radeon_state_atom * atom) { r300ContextPtr r300 = R300_CONTEXT(ctx); int numtmus = packet0_count(r300, r300->hw.tex.offset.cmd); @@ -200,7 +200,7 @@ static int check_tex_offsets(GLcontext *ctx, struct radeon_state_atom * atom) return dw; } -static void emit_tex_offsets(GLcontext *ctx, struct radeon_state_atom * atom) +static void emit_tex_offsets(struct gl_context *ctx, struct radeon_state_atom * atom) { r300ContextPtr r300 = R300_CONTEXT(ctx); BATCH_LOCALS(&r300->radeon); @@ -249,7 +249,7 @@ static void emit_tex_offsets(GLcontext *ctx, struct radeon_state_atom * atom) } } -void r300_emit_scissor(GLcontext *ctx) +void r300_emit_scissor(struct gl_context *ctx) { r300ContextPtr r300 = R300_CONTEXT(ctx); BATCH_LOCALS(&r300->radeon); @@ -287,7 +287,7 @@ void r300_emit_scissor(GLcontext *ctx) OUT_BATCH((x2 << R300_SCISSORS_X_SHIFT)|(y2 << R300_SCISSORS_Y_SHIFT)); END_BATCH(); } -static int check_cb_offset(GLcontext *ctx, struct radeon_state_atom * atom) +static int check_cb_offset(struct gl_context *ctx, struct radeon_state_atom * atom) { r300ContextPtr r300 = R300_CONTEXT(ctx); uint32_t dw = 6 + 3 + 16; @@ -411,7 +411,7 @@ void r300_emit_cb_setup(struct r300_context *r300, END_BATCH(); } -static void emit_cb_offset_atom(GLcontext *ctx, struct radeon_state_atom * atom) +static void emit_cb_offset_atom(struct gl_context *ctx, struct radeon_state_atom * atom) { r300ContextPtr r300 = R300_CONTEXT(ctx); struct radeon_renderbuffer *rrb; @@ -433,7 +433,7 @@ static void emit_cb_offset_atom(GLcontext *ctx, struct radeon_state_atom * atom) } } -static int check_zb_offset(GLcontext *ctx, struct radeon_state_atom * atom) +static int check_zb_offset(struct gl_context *ctx, struct radeon_state_atom * atom) { r300ContextPtr r300 = R300_CONTEXT(ctx); uint32_t dw; @@ -443,7 +443,7 @@ static int check_zb_offset(GLcontext *ctx, struct radeon_state_atom * atom) return dw; } -static void emit_zb_offset(GLcontext *ctx, struct radeon_state_atom * atom) +static void emit_zb_offset(struct gl_context *ctx, struct radeon_state_atom * atom) { r300ContextPtr r300 = R300_CONTEXT(ctx); BATCH_LOCALS(&r300->radeon); @@ -476,7 +476,7 @@ static void emit_zb_offset(GLcontext *ctx, struct radeon_state_atom * atom) END_BATCH(); } -static void emit_zstencil_format(GLcontext *ctx, struct radeon_state_atom * atom) +static void emit_zstencil_format(struct gl_context *ctx, struct radeon_state_atom * atom) { r300ContextPtr r300 = R300_CONTEXT(ctx); BATCH_LOCALS(&r300->radeon); @@ -504,17 +504,17 @@ static void emit_zstencil_format(GLcontext *ctx, struct radeon_state_atom * atom END_BATCH(); } -static int check_never(GLcontext *ctx, struct radeon_state_atom *atom) +static int check_never(struct gl_context *ctx, struct radeon_state_atom *atom) { return 0; } -static int check_always(GLcontext *ctx, struct radeon_state_atom *atom) +static int check_always(struct gl_context *ctx, struct radeon_state_atom *atom) { return atom->cmd_size; } -static int check_variable(GLcontext *ctx, struct radeon_state_atom *atom) +static int check_variable(struct gl_context *ctx, struct radeon_state_atom *atom) { r300ContextPtr r300 = R300_CONTEXT(ctx); int cnt; @@ -525,7 +525,7 @@ static int check_variable(GLcontext *ctx, struct radeon_state_atom *atom) return cnt ? cnt + 1 : 0; } -static int check_r500fp(GLcontext *ctx, struct radeon_state_atom *atom) +static int check_r500fp(struct gl_context *ctx, struct radeon_state_atom *atom) { int cnt; r300ContextPtr r300 = R300_CONTEXT(ctx); @@ -537,7 +537,7 @@ static int check_r500fp(GLcontext *ctx, struct radeon_state_atom *atom) return cnt ? (cnt * 6) + extra : 0; } -static int check_r500fp_const(GLcontext *ctx, struct radeon_state_atom *atom) +static int check_r500fp_const(struct gl_context *ctx, struct radeon_state_atom *atom) { int cnt; r300ContextPtr r300 = R300_CONTEXT(ctx); diff --git a/src/mesa/drivers/dri/r300/r300_cmdbuf.h b/src/mesa/drivers/dri/r300/r300_cmdbuf.h index 0e68da928e..7e6b8c5de6 100644 --- a/src/mesa/drivers/dri/r300/r300_cmdbuf.h +++ b/src/mesa/drivers/dri/r300/r300_cmdbuf.h @@ -45,7 +45,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #define SCISSORS_BUFSZ (3) void r300InitCmdBuf(r300ContextPtr r300); -void r300_emit_scissor(GLcontext *ctx); +void r300_emit_scissor(struct gl_context *ctx); void r300_emit_vpu(struct r300_context *ctx, uint32_t *data, diff --git a/src/mesa/drivers/dri/r300/r300_context.c b/src/mesa/drivers/dri/r300/r300_context.c index 3d2f8928fa..9fbd36bfe6 100644 --- a/src/mesa/drivers/dri/r300/r300_context.c +++ b/src/mesa/drivers/dri/r300/r300_context.c @@ -134,7 +134,6 @@ static const struct dri_extension card_extensions[] = { {"GL_MESAX_texture_float", NULL}, {"GL_NV_blend_square", NULL}, {"GL_NV_vertex_program", GL_NV_vertex_program_functions}, - {"GL_SGIS_generate_mipmap", NULL}, {NULL, NULL} /* *INDENT-ON* */ }; @@ -220,7 +219,7 @@ static void r300_vtbl_pre_emit_atoms(radeonContextPtr radeon) end_3d(radeon); } -static void r300_fallback(GLcontext *ctx, GLuint bit, GLboolean mode) +static void r300_fallback(struct gl_context *ctx, GLuint bit, GLboolean mode) { r300ContextPtr r300 = R300_CONTEXT(ctx); if (mode) @@ -332,7 +331,7 @@ static void r300_init_vtbl(radeonContextPtr radeon) } } -static void r300InitConstValues(GLcontext *ctx, radeonScreenPtr screen) +static void r300InitConstValues(struct gl_context *ctx, radeonScreenPtr screen) { r300ContextPtr r300 = R300_CONTEXT(ctx); @@ -440,7 +439,7 @@ static void r300ParseOptions(r300ContextPtr r300, radeonScreenPtr screen) r300->options = options; } -static void r300InitGLExtensions(GLcontext *ctx) +static void r300InitGLExtensions(struct gl_context *ctx) { r300ContextPtr r300 = R300_CONTEXT(ctx); @@ -478,7 +477,7 @@ static void r300InitIoctlFuncs(struct dd_function_table *functions) /* Create the device specific rendering context. */ GLboolean r300CreateContext(gl_api api, - const __GLcontextModes * glVisual, + const struct gl_config * glVisual, __DRIcontext * driContextPriv, void *sharedContextPrivate) { @@ -486,7 +485,7 @@ GLboolean r300CreateContext(gl_api api, radeonScreenPtr screen = (radeonScreenPtr) (sPriv->private); struct dd_function_table functions; r300ContextPtr r300; - GLcontext *ctx; + struct gl_context *ctx; assert(glVisual); assert(driContextPriv); diff --git a/src/mesa/drivers/dri/r300/r300_context.h b/src/mesa/drivers/dri/r300/r300_context.h index 99540e3354..349a3d412f 100644 --- a/src/mesa/drivers/dri/r300/r300_context.h +++ b/src/mesa/drivers/dri/r300/r300_context.h @@ -463,9 +463,9 @@ struct r300_swtcl_info { }; struct r300_vtable { - void (* SetupRSUnit)(GLcontext *ctx); - void (* SetupFragmentShaderTextures)(GLcontext *ctx, int *tmu_mappings); - void (* SetupPixelShader)(GLcontext *ctx); + void (* SetupRSUnit)(struct gl_context *ctx); + void (* SetupFragmentShaderTextures)(struct gl_context *ctx, int *tmu_mappings); + void (* SetupPixelShader)(struct gl_context *ctx); }; struct r300_vertex_buffer { @@ -544,7 +544,7 @@ struct r300_context { extern void r300DestroyContext(__DRIcontext * driContextPriv); extern GLboolean r300CreateContext(gl_api api, - const __GLcontextModes * glVisual, + const struct gl_config * glVisual, __DRIcontext * driContextPriv, void *sharedContextPrivate); @@ -552,7 +552,7 @@ extern void r300InitShaderFuncs(struct dd_function_table *functions); extern void r300InitShaderFunctions(r300ContextPtr r300); -extern void r300InitDraw(GLcontext *ctx); +extern void r300InitDraw(struct gl_context *ctx); #define r300PackFloat32 radeonPackFloat32 #define r300PackFloat24 radeonPackFloat24 diff --git a/src/mesa/drivers/dri/r300/r300_draw.c b/src/mesa/drivers/dri/r300/r300_draw.c index 5ae9f49840..81769e1ee5 100644 --- a/src/mesa/drivers/dri/r300/r300_draw.c +++ b/src/mesa/drivers/dri/r300/r300_draw.c @@ -75,7 +75,7 @@ static int getTypeSize(GLenum type) } } -static void r300FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer *mesa_ind_buf) +static void r300FixupIndexBuffer(struct gl_context *ctx, const struct _mesa_index_buffer *mesa_ind_buf) { r300ContextPtr r300 = R300_CONTEXT(ctx); GLvoid *src_ptr; @@ -143,7 +143,7 @@ static void r300FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer } -static void r300SetupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer *mesa_ind_buf) +static void r300SetupIndexBuffer(struct gl_context *ctx, const struct _mesa_index_buffer *mesa_ind_buf) { r300ContextPtr r300 = R300_CONTEXT(ctx); @@ -219,7 +219,7 @@ static void r300SetupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer * Convert attribute data type to float * If the attribute uses named buffer object replace the bo with newly allocated bo */ -static void r300ConvertAttrib(GLcontext *ctx, int count, const struct gl_client_array *input, struct vertex_attribute *attr) +static void r300ConvertAttrib(struct gl_context *ctx, int count, const struct gl_client_array *input, struct vertex_attribute *attr) { r300ContextPtr r300 = R300_CONTEXT(ctx); const GLvoid *src_ptr; @@ -290,7 +290,7 @@ static void r300ConvertAttrib(GLcontext *ctx, int count, const struct gl_client_ } } -static void r300AlignDataToDword(GLcontext *ctx, const struct gl_client_array *input, int count, struct vertex_attribute *attr) +static void r300AlignDataToDword(struct gl_context *ctx, const struct gl_client_array *input, int count, struct vertex_attribute *attr) { r300ContextPtr r300 = R300_CONTEXT(ctx); const int dst_stride = (input->StrideB + 3) & ~3; @@ -328,7 +328,7 @@ static void r300AlignDataToDword(GLcontext *ctx, const struct gl_client_array *i attr->stride = dst_stride; } -static void r300TranslateAttrib(GLcontext *ctx, GLuint attr, int count, const struct gl_client_array *input) +static void r300TranslateAttrib(struct gl_context *ctx, GLuint attr, int count, const struct gl_client_array *input) { r300ContextPtr r300 = R300_CONTEXT(ctx); struct r300_vertex_buffer *vbuf = &r300->vbuf; @@ -467,7 +467,7 @@ static void r300TranslateAttrib(GLcontext *ctx, GLuint attr, int count, const st ++vbuf->num_attribs; } -static void r300SetVertexFormat(GLcontext *ctx, const struct gl_client_array *arrays[], int count) +static void r300SetVertexFormat(struct gl_context *ctx, const struct gl_client_array *arrays[], int count) { r300ContextPtr r300 = R300_CONTEXT(ctx); struct r300_vertex_buffer *vbuf = &r300->vbuf; @@ -497,7 +497,7 @@ static void r300SetVertexFormat(GLcontext *ctx, const struct gl_client_array *ar return; } -static void r300AllocDmaRegions(GLcontext *ctx, const struct gl_client_array *input[], int count) +static void r300AllocDmaRegions(struct gl_context *ctx, const struct gl_client_array *input[], int count) { r300ContextPtr r300 = R300_CONTEXT(ctx); struct r300_vertex_buffer *vbuf = &r300->vbuf; @@ -578,7 +578,7 @@ static void r300AllocDmaRegions(GLcontext *ctx, const struct gl_client_array *in } -static void r300FreeData(GLcontext *ctx) +static void r300FreeData(struct gl_context *ctx) { /* Need to zero tcl.aos[n].bo and tcl.elt_dma_bo * to prevent double unref in radeonReleaseArrays @@ -604,7 +604,7 @@ static void r300FreeData(GLcontext *ctx) } } -static GLuint r300PredictTryDrawPrimsSize(GLcontext *ctx, +static GLuint r300PredictTryDrawPrimsSize(struct gl_context *ctx, GLuint nr_prims, const struct _mesa_prim *prim) { struct r300_context *r300 = R300_CONTEXT(ctx); @@ -641,7 +641,7 @@ static GLuint r300PredictTryDrawPrimsSize(GLcontext *ctx, return dwords; } -static GLboolean r300TryDrawPrims(GLcontext *ctx, +static GLboolean r300TryDrawPrims(struct gl_context *ctx, const struct gl_client_array *arrays[], const struct _mesa_prim *prim, GLuint nr_prims, @@ -707,7 +707,7 @@ static GLboolean r300TryDrawPrims(GLcontext *ctx, return GL_TRUE; } -static void r300DrawPrims(GLcontext *ctx, +static void r300DrawPrims(struct gl_context *ctx, const struct gl_client_array *arrays[], const struct _mesa_prim *prim, GLuint nr_prims, @@ -741,7 +741,7 @@ static void r300DrawPrims(GLcontext *ctx, _tnl_draw_prims(ctx, arrays, prim, nr_prims, ib, min_index, max_index); } -void r300InitDraw(GLcontext *ctx) +void r300InitDraw(struct gl_context *ctx) { struct vbo_context *vbo = vbo_context(ctx); diff --git a/src/mesa/drivers/dri/r300/r300_emit.c b/src/mesa/drivers/dri/r300/r300_emit.c index a24d431611..f392006ced 100644 --- a/src/mesa/drivers/dri/r300/r300_emit.c +++ b/src/mesa/drivers/dri/r300/r300_emit.c @@ -48,14 +48,14 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "r300_emit.h" -GLuint r300VAPInputCntl0(GLcontext * ctx, GLuint InputsRead) +GLuint r300VAPInputCntl0(struct gl_context * ctx, GLuint InputsRead) { /* No idea what this value means. I have seen other values written to * this register... */ return 0x5555; } -GLuint r300VAPInputCntl1(GLcontext * ctx, GLuint InputsRead) +GLuint r300VAPInputCntl1(struct gl_context * ctx, GLuint InputsRead) { GLuint i, vic_1 = 0; @@ -76,7 +76,7 @@ GLuint r300VAPInputCntl1(GLcontext * ctx, GLuint InputsRead) return vic_1; } -GLuint r300VAPOutputCntl0(GLcontext * ctx, GLuint vp_writes) +GLuint r300VAPOutputCntl0(struct gl_context * ctx, GLuint vp_writes) { GLuint ret = 0; @@ -100,7 +100,7 @@ GLuint r300VAPOutputCntl0(GLcontext * ctx, GLuint vp_writes) return ret; } -GLuint r300VAPOutputCntl1(GLcontext * ctx, GLuint vp_writes) +GLuint r300VAPOutputCntl1(struct gl_context * ctx, GLuint vp_writes) { GLuint i, ret = 0, first_free_texcoord = 0; diff --git a/src/mesa/drivers/dri/r300/r300_emit.h b/src/mesa/drivers/dri/r300/r300_emit.h index a456d8867c..8911ab7728 100644 --- a/src/mesa/drivers/dri/r300/r300_emit.h +++ b/src/mesa/drivers/dri/r300/r300_emit.h @@ -220,9 +220,9 @@ extern int r300NumVerts(r300ContextPtr rmesa, int num_verts, int prim); extern void r300EmitCacheFlush(r300ContextPtr rmesa); -extern GLuint r300VAPInputCntl0(GLcontext * ctx, GLuint InputsRead); -extern GLuint r300VAPInputCntl1(GLcontext * ctx, GLuint InputsRead); -extern GLuint r300VAPOutputCntl0(GLcontext * ctx, GLuint vp_writes); -extern GLuint r300VAPOutputCntl1(GLcontext * ctx, GLuint vp_writes); +extern GLuint r300VAPInputCntl0(struct gl_context * ctx, GLuint InputsRead); +extern GLuint r300VAPInputCntl1(struct gl_context * ctx, GLuint InputsRead); +extern GLuint r300VAPOutputCntl0(struct gl_context * ctx, GLuint vp_writes); +extern GLuint r300VAPOutputCntl1(struct gl_context * ctx, GLuint vp_writes); #endif diff --git a/src/mesa/drivers/dri/r300/r300_fragprog_common.c b/src/mesa/drivers/dri/r300/r300_fragprog_common.c index 4af91f114d..4e457b51eb 100644 --- a/src/mesa/drivers/dri/r300/r300_fragprog_common.c +++ b/src/mesa/drivers/dri/r300/r300_fragprog_common.c @@ -208,7 +208,7 @@ static void allocate_hw_inputs( } -static void translate_fragment_program(GLcontext *ctx, struct r300_fragment_program_cont *cont, struct r300_fragment_program *fp) +static void translate_fragment_program(struct gl_context *ctx, struct r300_fragment_program_cont *cont, struct r300_fragment_program *fp) { r300ContextPtr r300 = R300_CONTEXT(ctx); struct r300_fragment_program_compiler compiler; @@ -278,7 +278,7 @@ static void translate_fragment_program(GLcontext *ctx, struct r300_fragment_prog rc_destroy(&compiler.Base); } -struct r300_fragment_program *r300SelectAndTranslateFragmentShader(GLcontext *ctx) +struct r300_fragment_program *r300SelectAndTranslateFragmentShader(struct gl_context *ctx) { r300ContextPtr r300 = R300_CONTEXT(ctx); struct r300_fragment_program_cont *fp_list; diff --git a/src/mesa/drivers/dri/r300/r300_fragprog_common.h b/src/mesa/drivers/dri/r300/r300_fragprog_common.h index 3d64c08cee..cfa5acf433 100644 --- a/src/mesa/drivers/dri/r300/r300_fragprog_common.h +++ b/src/mesa/drivers/dri/r300/r300_fragprog_common.h @@ -32,6 +32,6 @@ #include "r300_context.h" -struct r300_fragment_program *r300SelectAndTranslateFragmentShader(GLcontext *ctx); +struct r300_fragment_program *r300SelectAndTranslateFragmentShader(struct gl_context *ctx); #endif diff --git a/src/mesa/drivers/dri/r300/r300_render.c b/src/mesa/drivers/dri/r300/r300_render.c index cf89ab7ec3..821318e7a5 100644 --- a/src/mesa/drivers/dri/r300/r300_render.c +++ b/src/mesa/drivers/dri/r300/r300_render.c @@ -321,7 +321,7 @@ static void r300FireAOS(r300ContextPtr rmesa, int vertex_count, int type) END_BATCH(); } -void r300RunRenderPrimitive(GLcontext * ctx, int start, int end, int prim) +void r300RunRenderPrimitive(struct gl_context * ctx, int start, int end, int prim) { r300ContextPtr rmesa = R300_CONTEXT(ctx); BATCH_LOCALS(&rmesa->radeon); @@ -444,7 +444,7 @@ static const char *getFallbackString(r300ContextPtr rmesa, uint32_t bit) } } -void r300SwitchFallback(GLcontext *ctx, uint32_t bit, GLboolean mode) +void r300SwitchFallback(struct gl_context *ctx, uint32_t bit, GLboolean mode) { TNLcontext *tnl = TNL_CONTEXT(ctx); r300ContextPtr rmesa = R300_CONTEXT(ctx); diff --git a/src/mesa/drivers/dri/r300/r300_render.h b/src/mesa/drivers/dri/r300/r300_render.h index 581e9fa0cc..5a78592c75 100644 --- a/src/mesa/drivers/dri/r300/r300_render.h +++ b/src/mesa/drivers/dri/r300/r300_render.h @@ -63,8 +63,8 @@ extern const struct tnl_pipeline_stage _r300_render_stage; -extern void r300SwitchFallback(GLcontext *ctx, uint32_t bit, GLboolean mode); +extern void r300SwitchFallback(struct gl_context *ctx, uint32_t bit, GLboolean mode); -extern void r300RunRenderPrimitive(GLcontext * ctx, int start, int end, int prim); +extern void r300RunRenderPrimitive(struct gl_context * ctx, int start, int end, int prim); #endif diff --git a/src/mesa/drivers/dri/r300/r300_shader.c b/src/mesa/drivers/dri/r300/r300_shader.c index a9bddf0577..f2bbac5b85 100644 --- a/src/mesa/drivers/dri/r300/r300_shader.c +++ b/src/mesa/drivers/dri/r300/r300_shader.c @@ -32,7 +32,7 @@ #include "r300_context.h" #include "r300_fragprog_common.h" -static void freeFragProgCache(GLcontext *ctx, struct r300_fragment_program_cont *cache) +static void freeFragProgCache(struct gl_context *ctx, struct r300_fragment_program_cont *cache) { struct r300_fragment_program *tmp, *fp = cache->progs; @@ -44,7 +44,7 @@ static void freeFragProgCache(GLcontext *ctx, struct r300_fragment_program_cont } } -static void freeVertProgCache(GLcontext *ctx, struct r300_vertex_program_cont *cache) +static void freeVertProgCache(struct gl_context *ctx, struct r300_vertex_program_cont *cache) { struct r300_vertex_program *tmp, *vp = cache->progs; @@ -57,7 +57,7 @@ static void freeVertProgCache(GLcontext *ctx, struct r300_vertex_program_cont *c } } -static struct gl_program *r300NewProgram(GLcontext * ctx, GLenum target, +static struct gl_program *r300NewProgram(struct gl_context * ctx, GLenum target, GLuint id) { struct r300_vertex_program_cont *vp; @@ -81,7 +81,7 @@ static struct gl_program *r300NewProgram(GLcontext * ctx, GLenum target, return NULL; } -static void r300DeleteProgram(GLcontext * ctx, struct gl_program *prog) +static void r300DeleteProgram(struct gl_context * ctx, struct gl_program *prog) { struct r300_vertex_program_cont *vp = (struct r300_vertex_program_cont *)prog; struct r300_fragment_program_cont *fp = (struct r300_fragment_program_cont *)prog; @@ -99,7 +99,7 @@ static void r300DeleteProgram(GLcontext * ctx, struct gl_program *prog) } static GLboolean -r300ProgramStringNotify(GLcontext * ctx, GLenum target, struct gl_program *prog) +r300ProgramStringNotify(struct gl_context * ctx, GLenum target, struct gl_program *prog) { struct r300_vertex_program_cont *vp = (struct r300_vertex_program_cont *)prog; struct r300_fragment_program_cont *fp = (struct r300_fragment_program_cont *)prog; @@ -123,7 +123,7 @@ r300ProgramStringNotify(GLcontext * ctx, GLenum target, struct gl_program *prog) } static GLboolean -r300IsProgramNative(GLcontext * ctx, GLenum target, struct gl_program *prog) +r300IsProgramNative(struct gl_context * ctx, GLenum target, struct gl_program *prog) { if (target == GL_FRAGMENT_PROGRAM_ARB) { struct r300_fragment_program *fp = r300SelectAndTranslateFragmentShader(ctx); diff --git a/src/mesa/drivers/dri/r300/r300_state.c b/src/mesa/drivers/dri/r300/r300_state.c index 0113eecaa3..ab8c1df5f7 100644 --- a/src/mesa/drivers/dri/r300/r300_state.c +++ b/src/mesa/drivers/dri/r300/r300_state.c @@ -62,7 +62,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "r300_render.h" #include "r300_vertprog.h" -static void r300BlendColor(GLcontext * ctx, const GLfloat cf[4]) +static void r300BlendColor(struct gl_context * ctx, const GLfloat cf[4]) { r300ContextPtr rmesa = R300_CONTEXT(ctx); @@ -204,7 +204,7 @@ static void r300SetBlendCntl(r300ContextPtr r300, int func, int eqn, } } -static void r300SetBlendState(GLcontext * ctx) +static void r300SetBlendState(struct gl_context * ctx) { r300ContextPtr r300 = R300_CONTEXT(ctx); int func = (R300_BLEND_GL_ONE << R300_SRC_BLEND_SHIFT) | @@ -302,13 +302,13 @@ static void r300SetBlendState(GLcontext * ctx) R300_ALPHA_BLEND_ENABLE), funcA, eqnA); } -static void r300BlendEquationSeparate(GLcontext * ctx, +static void r300BlendEquationSeparate(struct gl_context * ctx, GLenum modeRGB, GLenum modeA) { r300SetBlendState(ctx); } -static void r300BlendFuncSeparate(GLcontext * ctx, +static void r300BlendFuncSeparate(struct gl_context * ctx, GLenum sfactorRGB, GLenum dfactorRGB, GLenum sfactorA, GLenum dfactorA) { @@ -331,7 +331,7 @@ static GLuint translate_logicop(GLenum logicop) * Used internally to update the r300->hw hardware state to match the * current OpenGL state. */ -static void r300SetLogicOpState(GLcontext *ctx) +static void r300SetLogicOpState(struct gl_context *ctx) { r300ContextPtr r300 = R300_CONTEXT(ctx); R300_STATECHANGE(r300, rop); @@ -347,13 +347,13 @@ static void r300SetLogicOpState(GLcontext *ctx) * Called by Mesa when an application program changes the LogicOp state * via glLogicOp. */ -static void r300LogicOpcode(GLcontext *ctx, GLenum logicop) +static void r300LogicOpcode(struct gl_context *ctx, GLenum logicop) { if (RGBA_LOGICOP_ENABLED(ctx)) r300SetLogicOpState(ctx); } -static void r300ClipPlane( GLcontext *ctx, GLenum plane, const GLfloat *eq ) +static void r300ClipPlane( struct gl_context *ctx, GLenum plane, const GLfloat *eq ) { r300ContextPtr rmesa = R300_CONTEXT(ctx); GLint p; @@ -373,7 +373,7 @@ static void r300ClipPlane( GLcontext *ctx, GLenum plane, const GLfloat *eq ) rmesa->hw.vpucp[p].cmd[R300_VPUCP_W] = ip[3]; } -static void r300SetClipPlaneState(GLcontext * ctx, GLenum cap, GLboolean state) +static void r300SetClipPlaneState(struct gl_context * ctx, GLenum cap, GLboolean state) { r300ContextPtr r300 = R300_CONTEXT(ctx); GLuint p; @@ -395,7 +395,7 @@ static void r300SetClipPlaneState(GLcontext * ctx, GLenum cap, GLboolean state) /** * Update our tracked culling state based on Mesa's state. */ -static void r300UpdateCulling(GLcontext * ctx) +static void r300UpdateCulling(struct gl_context * ctx) { r300ContextPtr r300 = R300_CONTEXT(ctx); uint32_t val = 0; @@ -435,7 +435,7 @@ static void r300UpdateCulling(GLcontext * ctx) r300->hw.cul.cmd[R300_CUL_CULL] = val; } -static void r300SetPolygonOffsetState(GLcontext * ctx, GLboolean state) +static void r300SetPolygonOffsetState(struct gl_context * ctx, GLboolean state) { r300ContextPtr r300 = R300_CONTEXT(ctx); @@ -447,14 +447,14 @@ static void r300SetPolygonOffsetState(GLcontext * ctx, GLboolean state) } } -static GLboolean current_fragment_program_writes_depth(GLcontext* ctx) +static GLboolean current_fragment_program_writes_depth(struct gl_context* ctx) { r300ContextPtr r300 = R300_CONTEXT(ctx); return ctx->FragmentProgram._Current && r300->selected_fp->code.writes_depth; } -static void r300SetEarlyZState(GLcontext * ctx) +static void r300SetEarlyZState(struct gl_context * ctx) { r300ContextPtr r300 = R300_CONTEXT(ctx); GLuint topZ = R300_ZTOP_ENABLE; @@ -499,7 +499,7 @@ static void r300SetEarlyZState(GLcontext * ctx) } } -static void r300SetAlphaState(GLcontext * ctx) +static void r300SetAlphaState(struct gl_context * ctx) { r300ContextPtr r300 = R300_CONTEXT(ctx); GLubyte refByte; @@ -549,7 +549,7 @@ static void r300SetAlphaState(GLcontext * ctx) r300->hw.at.cmd[R300_AT_UNKNOWN] = 0; } -static void r300AlphaFunc(GLcontext * ctx, GLenum func, GLfloat ref) +static void r300AlphaFunc(struct gl_context * ctx, GLenum func, GLfloat ref) { (void)func; (void)ref; @@ -579,7 +579,7 @@ static int translate_func(int func) return 0; } -static void r300SetDepthState(GLcontext * ctx) +static void r300SetDepthState(struct gl_context * ctx) { r300ContextPtr r300 = R300_CONTEXT(ctx); @@ -598,7 +598,7 @@ static void r300SetDepthState(GLcontext * ctx) } } -static void r300CatchStencilFallback(GLcontext *ctx) +static void r300CatchStencilFallback(struct gl_context *ctx) { r300ContextPtr rmesa = R300_CONTEXT(ctx); const unsigned back = ctx->Stencil._BackFace; @@ -616,7 +616,7 @@ static void r300CatchStencilFallback(GLcontext *ctx) } } -static void r300SetStencilState(GLcontext * ctx, GLboolean state) +static void r300SetStencilState(struct gl_context * ctx, GLboolean state) { r300ContextPtr r300 = R300_CONTEXT(ctx); GLboolean hw_stencil = GL_FALSE; @@ -641,7 +641,7 @@ static void r300SetStencilState(GLcontext * ctx, GLboolean state) } } -static void r300UpdatePolygonMode(GLcontext * ctx) +static void r300UpdatePolygonMode(struct gl_context * ctx) { r300ContextPtr r300 = R300_CONTEXT(ctx); uint32_t hw_mode = R300_GA_POLY_MODE_DISABLE; @@ -704,7 +704,7 @@ static void r300UpdatePolygonMode(GLcontext * ctx) * * \note Mesa already filters redundant calls to this function. */ -static void r300CullFace(GLcontext * ctx, GLenum mode) +static void r300CullFace(struct gl_context * ctx, GLenum mode) { (void)mode; @@ -716,7 +716,7 @@ static void r300CullFace(GLcontext * ctx, GLenum mode) * * \note Mesa already filters redundant calls to this function. */ -static void r300FrontFace(GLcontext * ctx, GLenum mode) +static void r300FrontFace(struct gl_context * ctx, GLenum mode) { (void)mode; @@ -729,7 +729,7 @@ static void r300FrontFace(GLcontext * ctx, GLenum mode) * * \note Mesa already filters redundant calls to this function. */ -static void r300DepthFunc(GLcontext * ctx, GLenum func) +static void r300DepthFunc(struct gl_context * ctx, GLenum func) { (void)func; r300SetDepthState(ctx); @@ -740,7 +740,7 @@ static void r300DepthFunc(GLcontext * ctx, GLenum func) * * \note Mesa already filters redundant calls to this function. */ -static void r300DepthMask(GLcontext * ctx, GLboolean mask) +static void r300DepthMask(struct gl_context * ctx, GLboolean mask) { (void)mask; r300SetDepthState(ctx); @@ -749,7 +749,7 @@ static void r300DepthMask(GLcontext * ctx, GLboolean mask) /** * Handle glColorMask() */ -static void r300ColorMask(GLcontext * ctx, +static void r300ColorMask(struct gl_context * ctx, GLboolean r, GLboolean g, GLboolean b, GLboolean a) { r300ContextPtr r300 = R300_CONTEXT(ctx); @@ -767,7 +767,7 @@ static void r300ColorMask(GLcontext * ctx, /* ============================================================= * Point state */ -static void r300PointSize(GLcontext * ctx, GLfloat size) +static void r300PointSize(struct gl_context * ctx, GLfloat size) { r300ContextPtr r300 = R300_CONTEXT(ctx); @@ -784,7 +784,7 @@ static void r300PointSize(GLcontext * ctx, GLfloat size) ((int)(size * 6) << R300_POINTSIZE_Y_SHIFT); } -static void r300PointParameter(GLcontext * ctx, GLenum pname, const GLfloat * param) +static void r300PointParameter(struct gl_context * ctx, GLenum pname, const GLfloat * param) { r300ContextPtr r300 = R300_CONTEXT(ctx); @@ -814,7 +814,7 @@ static void r300PointParameter(GLcontext * ctx, GLenum pname, const GLfloat * pa /* ============================================================= * Line state */ -static void r300LineWidth(GLcontext * ctx, GLfloat widthf) +static void r300LineWidth(struct gl_context * ctx, GLfloat widthf) { r300ContextPtr r300 = R300_CONTEXT(ctx); @@ -826,7 +826,7 @@ static void r300LineWidth(GLcontext * ctx, GLfloat widthf) R300_LINE_CNT_HO | R300_LINE_CNT_VE | (int)(widthf * 6.0); } -static void r300PolygonMode(GLcontext * ctx, GLenum face, GLenum mode) +static void r300PolygonMode(struct gl_context * ctx, GLenum face, GLenum mode) { (void)face; (void)mode; @@ -864,7 +864,7 @@ static int translate_stencil_op(int op) return 0; } -static void r300ShadeModel(GLcontext * ctx, GLenum mode) +static void r300ShadeModel(struct gl_context * ctx, GLenum mode) { r300ContextPtr rmesa = R300_CONTEXT(ctx); @@ -885,7 +885,7 @@ static void r300ShadeModel(GLcontext * ctx, GLenum mode) rmesa->hw.shade2.cmd[3] = 0x00000000; } -static void r300StencilFuncSeparate(GLcontext * ctx, GLenum face, +static void r300StencilFuncSeparate(struct gl_context * ctx, GLenum face, GLenum func, GLint ref, GLuint mask) { r300ContextPtr rmesa = R300_CONTEXT(ctx); @@ -932,7 +932,7 @@ static void r300StencilFuncSeparate(GLcontext * ctx, GLenum face, } } -static void r300StencilMaskSeparate(GLcontext * ctx, GLenum face, GLuint mask) +static void r300StencilMaskSeparate(struct gl_context * ctx, GLenum face, GLuint mask) { r300ContextPtr rmesa = R300_CONTEXT(ctx); const unsigned back = ctx->Stencil._BackFace; @@ -956,7 +956,7 @@ static void r300StencilMaskSeparate(GLcontext * ctx, GLenum face, GLuint mask) } } -static void r300StencilOpSeparate(GLcontext * ctx, GLenum face, +static void r300StencilOpSeparate(struct gl_context * ctx, GLenum face, GLenum fail, GLenum zfail, GLenum zpass) { r300ContextPtr rmesa = R300_CONTEXT(ctx); @@ -992,7 +992,7 @@ static void r300StencilOpSeparate(GLcontext * ctx, GLenum face, * Window position and viewport transformation */ -static void r300UpdateWindow(GLcontext * ctx) +static void r300UpdateWindow(struct gl_context * ctx) { r300ContextPtr rmesa = R300_CONTEXT(ctx); __DRIdrawable *dPriv = radeon_get_drawable(&rmesa->radeon); @@ -1028,7 +1028,7 @@ static void r300UpdateWindow(GLcontext * ctx) rmesa->hw.vpt.cmd[R300_VPT_ZOFFSET] = r300PackFloat32(tz); } -static void r300Viewport(GLcontext * ctx, GLint x, GLint y, +static void r300Viewport(struct gl_context * ctx, GLint x, GLint y, GLsizei width, GLsizei height) { /* Don't pipeline viewport changes, conflict with window offset @@ -1040,12 +1040,12 @@ static void r300Viewport(GLcontext * ctx, GLint x, GLint y, radeon_viewport(ctx, x, y, width, height); } -static void r300DepthRange(GLcontext * ctx, GLclampd nearval, GLclampd farval) +static void r300DepthRange(struct gl_context * ctx, GLclampd nearval, GLclampd farval) { r300UpdateWindow(ctx); } -void r300UpdateViewportOffset(GLcontext * ctx) +void r300UpdateViewportOffset(struct gl_context * ctx) { r300ContextPtr rmesa = R300_CONTEXT(ctx); __DRIdrawable *dPriv = radeon_get_drawable(&rmesa->radeon); @@ -1074,7 +1074,7 @@ void r300UpdateViewportOffset(GLcontext * ctx) * Update R300's own internal state parameters. * For now just STATE_R300_WINDOW_DIMENSION */ -static void r300UpdateStateParameters(GLcontext * ctx, GLuint new_state) +static void r300UpdateStateParameters(struct gl_context * ctx, GLuint new_state) { r300ContextPtr rmesa = R300_CONTEXT(ctx); struct gl_program_parameter_list *paramList; @@ -1096,7 +1096,7 @@ static void r300UpdateStateParameters(GLcontext * ctx, GLuint new_state) /* ============================================================= * Polygon state */ -static void r300PolygonOffset(GLcontext * ctx, GLfloat factor, GLfloat units) +static void r300PolygonOffset(struct gl_context * ctx, GLfloat factor, GLfloat units) { r300ContextPtr rmesa = R300_CONTEXT(ctx); GLfloat constant = units; @@ -1193,7 +1193,7 @@ static unsigned long gen_fixed_filter(unsigned long f) return f; } -static void r300SetupFragmentShaderTextures(GLcontext *ctx, int *tmu_mappings) +static void r300SetupFragmentShaderTextures(struct gl_context *ctx, int *tmu_mappings) { r300ContextPtr r300 = R300_CONTEXT(ctx); int i; @@ -1235,7 +1235,7 @@ static void r300SetupFragmentShaderTextures(GLcontext *ctx, int *tmu_mappings) R300_US_TEX_INST_0, code->tex.length); } -static void r500SetupFragmentShaderTextures(GLcontext *ctx, int *tmu_mappings) +static void r500SetupFragmentShaderTextures(struct gl_context *ctx, int *tmu_mappings) { r300ContextPtr r300 = R300_CONTEXT(ctx); int i; @@ -1280,7 +1280,7 @@ static GLuint translate_lod_bias(GLfloat bias) } -static void r300SetupTextures(GLcontext * ctx) +static void r300SetupTextures(struct gl_context * ctx) { int i, mtu; struct radeon_tex_obj *t; @@ -1427,7 +1427,7 @@ union r300_outputs_written { ((hw_tcl_on) ? (ow).vp_outputs & (1 << (vp_result)) : \ RENDERINPUTS_TEST( (ow.index_bitset), (tnl_attrib) )) -static void r300SetupRSUnit(GLcontext * ctx) +static void r300SetupRSUnit(struct gl_context * ctx) { r300ContextPtr r300 = R300_CONTEXT(ctx); union r300_outputs_written OutputsWritten; @@ -1521,7 +1521,7 @@ static void r300SetupRSUnit(GLcontext * ctx) WARN_ONCE("Don't know how to satisfy InputsRead=0x%08x\n", InputsRead); } -static void r500SetupRSUnit(GLcontext * ctx) +static void r500SetupRSUnit(struct gl_context * ctx) { r300ContextPtr r300 = R300_CONTEXT(ctx); union r300_outputs_written OutputsWritten; @@ -1681,7 +1681,7 @@ void r300VapCntl(r300ContextPtr rmesa, GLuint input_count, * * \note Mesa already filters redundant calls to this function. */ -static void r300Enable(GLcontext * ctx, GLenum cap, GLboolean state) +static void r300Enable(struct gl_context * ctx, GLenum cap, GLboolean state) { r300ContextPtr rmesa = R300_CONTEXT(ctx); if (RADEON_DEBUG & RADEON_STATE) @@ -1756,7 +1756,7 @@ static void r300Enable(GLcontext * ctx, GLenum cap, GLboolean state) */ static void r300ResetHwState(r300ContextPtr r300) { - GLcontext *ctx = r300->radeon.glCtx; + struct gl_context *ctx = r300->radeon.glCtx; int has_tcl; has_tcl = r300->options.hw_tcl_enabled; @@ -1965,7 +1965,7 @@ static void r300ResetHwState(r300ContextPtr r300) void r300UpdateShaders(r300ContextPtr rmesa) { - GLcontext *ctx = rmesa->radeon.glCtx; + struct gl_context *ctx = rmesa->radeon.glCtx; /* should only happenen once, just after context is created */ /* TODO: shouldn't we fallback to sw here? */ @@ -1994,7 +1994,7 @@ void r300UpdateShaders(r300ContextPtr rmesa) rmesa->radeon.NewGLState = 0; } -static const GLfloat *get_fragmentprogram_constant(GLcontext *ctx, GLuint index, GLfloat * buffer) +static const GLfloat *get_fragmentprogram_constant(struct gl_context *ctx, GLuint index, GLfloat * buffer) { static const GLfloat dummy[4] = { 0, 0, 0, 0 }; r300ContextPtr rmesa = R300_CONTEXT(ctx); @@ -2052,7 +2052,7 @@ static const GLfloat *get_fragmentprogram_constant(GLcontext *ctx, GLuint index, } -static void r300SetupPixelShader(GLcontext *ctx) +static void r300SetupPixelShader(struct gl_context *ctx) { r300ContextPtr rmesa = R300_CONTEXT(ctx); struct r300_fragment_program *fp = rmesa->selected_fp; @@ -2109,7 +2109,7 @@ static void r300SetupPixelShader(GLcontext *ctx) if(_nc>_p->r500fp.count)_p->r500fp.count=_nc;\ } while(0) -static void r500SetupPixelShader(GLcontext *ctx) +static void r500SetupPixelShader(struct gl_context *ctx) { r300ContextPtr rmesa = R300_CONTEXT(ctx); struct r300_fragment_program *fp = rmesa->selected_fp; @@ -2158,7 +2158,7 @@ static void r500SetupPixelShader(GLcontext *ctx) bump_r500fp_const_count(rmesa->hw.r500fp_const.cmd, fp->code.constants.Count * 4); } -void r300SetupVAP(GLcontext *ctx, GLuint InputsRead, GLuint OutputsWritten) +void r300SetupVAP(struct gl_context *ctx, GLuint InputsRead, GLuint OutputsWritten) { r300ContextPtr rmesa = R300_CONTEXT( ctx ); struct vertex_attribute *attrs = rmesa->vbuf.attribs; @@ -2218,7 +2218,7 @@ void r300SetupVAP(GLcontext *ctx, GLuint InputsRead, GLuint OutputsWritten) void r300UpdateShaderStates(r300ContextPtr rmesa) { - GLcontext *ctx; + struct gl_context *ctx; ctx = rmesa->radeon.glCtx; /* should only happenen once, just after context is created */ @@ -2241,7 +2241,7 @@ void r300UpdateShaderStates(r300ContextPtr rmesa) #define EASY_US_OUT_FMT(comps, c0, c1, c2, c3) \ (R500_OUT_FMT_##comps | R500_C0_SEL_##c0 | R500_C1_SEL_##c1 | \ R500_C2_SEL_##c2 | R500_C3_SEL_##c3) -static void r300SetupUsOutputFormat(GLcontext *ctx) +static void r300SetupUsOutputFormat(struct gl_context *ctx) { r300ContextPtr rmesa = R300_CONTEXT(ctx); uint32_t hw_format; @@ -2292,6 +2292,7 @@ static void r300SetupUsOutputFormat(GLcontext *ctx) break; default: assert(!"Unsupported format"); + hw_format = 0; break; } @@ -2303,7 +2304,7 @@ static void r300SetupUsOutputFormat(GLcontext *ctx) /** * Called by Mesa after an internal state update. */ -static void r300InvalidateState(GLcontext * ctx, GLuint new_state) +static void r300InvalidateState(struct gl_context * ctx, GLuint new_state) { r300ContextPtr r300 = R300_CONTEXT(ctx); @@ -2346,7 +2347,7 @@ void r300InitState(r300ContextPtr r300) r300ResetHwState(r300); } -static void r300RenderMode(GLcontext * ctx, GLenum mode) +static void r300RenderMode(struct gl_context * ctx, GLenum mode) { r300SwitchFallback(ctx, R300_FALLBACK_RENDER_MODE, ctx->RenderMode != GL_RENDER); } diff --git a/src/mesa/drivers/dri/r300/r300_state.h b/src/mesa/drivers/dri/r300/r300_state.h index e70f84f4e4..e3b0da4cbd 100644 --- a/src/mesa/drivers/dri/r300/r300_state.h +++ b/src/mesa/drivers/dri/r300/r300_state.h @@ -50,13 +50,13 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. r300->radeon.hw.is_dirty = GL_TRUE; \ } while(0) -void r300UpdateViewportOffset (GLcontext * ctx); -void r300UpdateDrawBuffer (GLcontext * ctx); +void r300UpdateViewportOffset (struct gl_context * ctx); +void r300UpdateDrawBuffer (struct gl_context * ctx); void r300UpdateShaders (r300ContextPtr rmesa); void r300UpdateShaderStates (r300ContextPtr rmesa); void r300InitState (r300ContextPtr r300); void r300InitStateFuncs (radeonContextPtr radeon, struct dd_function_table *functions); void r300VapCntl(r300ContextPtr rmesa, GLuint input_count, GLuint output_count, GLuint temp_count); -void r300SetupVAP(GLcontext *ctx, GLuint InputsRead, GLuint OutputsWritten); +void r300SetupVAP(struct gl_context *ctx, GLuint InputsRead, GLuint OutputsWritten); #endif /* __R300_STATE_H__ */ diff --git a/src/mesa/drivers/dri/r300/r300_swtcl.c b/src/mesa/drivers/dri/r300/r300_swtcl.c index 4dcc7cb022..4a6762ff83 100644 --- a/src/mesa/drivers/dri/r300/r300_swtcl.c +++ b/src/mesa/drivers/dri/r300/r300_swtcl.c @@ -68,7 +68,7 @@ do { \ ++num_attrs; \ } while (0) -void r300ChooseSwtclVertexFormat(GLcontext *ctx, GLuint *_InputsRead, GLuint *_OutputsWritten) +void r300ChooseSwtclVertexFormat(struct gl_context *ctx, GLuint *_InputsRead, GLuint *_OutputsWritten) { r300ContextPtr rmesa = R300_CONTEXT( ctx ); TNLcontext *tnl = TNL_CONTEXT(ctx); @@ -226,7 +226,7 @@ void r300ChooseSwtclVertexFormat(GLcontext *ctx, GLuint *_InputsRead, GLuint *_ RENDERINPUTS_COPY(rmesa->render_inputs_bitset, tnl->render_inputs_bitset); } -static void r300PrepareVertices(GLcontext *ctx) +static void r300PrepareVertices(struct gl_context *ctx) { r300ContextPtr rmesa = R300_CONTEXT(ctx); GLuint InputsRead, OutputsWritten; @@ -285,7 +285,7 @@ static GLuint reduced_prim[] = { GL_TRIANGLES, }; -static void r300RasterPrimitive( GLcontext *ctx, GLuint prim ); +static void r300RasterPrimitive( struct gl_context *ctx, GLuint prim ); /*********************************************************************** * Emit primitives as inline vertices * @@ -497,7 +497,7 @@ static void init_rast_tab( void ) /**********************************************************************/ /* Choose render functions */ /**********************************************************************/ -static void r300ChooseRenderState( GLcontext *ctx ) +static void r300ChooseRenderState( struct gl_context *ctx ) { TNLcontext *tnl = TNL_CONTEXT(ctx); r300ContextPtr rmesa = R300_CONTEXT(ctx); @@ -528,7 +528,7 @@ static void r300ChooseRenderState( GLcontext *ctx ) } } -void r300RenderStart(GLcontext *ctx) +void r300RenderStart(struct gl_context *ctx) { radeon_print(RADEON_SWRENDER, RADEON_VERBOSE, "%s\n", __func__); r300ContextPtr rmesa = R300_CONTEXT( ctx ); @@ -550,11 +550,11 @@ void r300RenderStart(GLcontext *ctx) } } -void r300RenderFinish(GLcontext *ctx) +void r300RenderFinish(struct gl_context *ctx) { } -static void r300RasterPrimitive( GLcontext *ctx, GLuint hwprim ) +static void r300RasterPrimitive( struct gl_context *ctx, GLuint hwprim ) { r300ContextPtr rmesa = R300_CONTEXT(ctx); radeon_print(RADEON_SWRENDER, RADEON_TRACE, "%s\n", __func__); @@ -565,7 +565,7 @@ static void r300RasterPrimitive( GLcontext *ctx, GLuint hwprim ) } } -void r300RenderPrimitive(GLcontext *ctx, GLenum prim) +void r300RenderPrimitive(struct gl_context *ctx, GLenum prim) { r300ContextPtr rmesa = R300_CONTEXT(ctx); @@ -578,13 +578,13 @@ void r300RenderPrimitive(GLcontext *ctx, GLenum prim) r300RasterPrimitive( ctx, reduced_prim[prim] ); } -void r300ResetLineStipple(GLcontext *ctx) +void r300ResetLineStipple(struct gl_context *ctx) { if (RADEON_DEBUG & RADEON_VERTS) fprintf(stderr, "%s\n", __func__); } -void r300InitSwtcl(GLcontext *ctx) +void r300InitSwtcl(struct gl_context *ctx) { TNLcontext *tnl = TNL_CONTEXT(ctx); r300ContextPtr rmesa = R300_CONTEXT(ctx); @@ -620,7 +620,7 @@ void r300InitSwtcl(GLcontext *ctx) _tnl_need_projected_coords( ctx, GL_FALSE ); } -void r300DestroySwtcl(GLcontext *ctx) +void r300DestroySwtcl(struct gl_context *ctx) { } @@ -656,7 +656,7 @@ static void r300EmitVbufPrim(r300ContextPtr rmesa, GLuint primitive, GLuint vert END_BATCH(); } -void r300_swtcl_flush(GLcontext *ctx, uint32_t current_offset) +void r300_swtcl_flush(struct gl_context *ctx, uint32_t current_offset) { radeon_print(RADEON_SWRENDER, RADEON_TRACE, "%s\n", __func__); r300ContextPtr rmesa = R300_CONTEXT(ctx); diff --git a/src/mesa/drivers/dri/r300/r300_swtcl.h b/src/mesa/drivers/dri/r300/r300_swtcl.h index c271d26546..51cfffc2af 100644 --- a/src/mesa/drivers/dri/r300/r300_swtcl.h +++ b/src/mesa/drivers/dri/r300/r300_swtcl.h @@ -50,16 +50,16 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #define SWTCL_OVM_TEX(n) ((n) + 6) #define SWTCL_OVM_POINT_SIZE 15 -extern void r300ChooseSwtclVertexFormat(GLcontext *ctx, GLuint *InputsRead, GLuint *OutputsWritten); +extern void r300ChooseSwtclVertexFormat(struct gl_context *ctx, GLuint *InputsRead, GLuint *OutputsWritten); -extern void r300InitSwtcl( GLcontext *ctx ); -extern void r300DestroySwtcl( GLcontext *ctx ); +extern void r300InitSwtcl( struct gl_context *ctx ); +extern void r300DestroySwtcl( struct gl_context *ctx ); -extern void r300RenderStart(GLcontext *ctx); -extern void r300RenderFinish(GLcontext *ctx); -extern void r300RenderPrimitive(GLcontext *ctx, GLenum prim); -extern void r300ResetLineStipple(GLcontext *ctx); +extern void r300RenderStart(struct gl_context *ctx); +extern void r300RenderFinish(struct gl_context *ctx); +extern void r300RenderPrimitive(struct gl_context *ctx, GLenum prim); +extern void r300ResetLineStipple(struct gl_context *ctx); -extern void r300_swtcl_flush(GLcontext *ctx, uint32_t current_offset); +extern void r300_swtcl_flush(struct gl_context *ctx, uint32_t current_offset); #endif diff --git a/src/mesa/drivers/dri/r300/r300_tex.c b/src/mesa/drivers/dri/r300/r300_tex.c index baef206bc2..a6bda0e499 100644 --- a/src/mesa/drivers/dri/r300/r300_tex.c +++ b/src/mesa/drivers/dri/r300/r300_tex.c @@ -185,7 +185,7 @@ static void r300SetTexBorderColor(radeonTexObjPtr t, const GLfloat color[4]) * next UpdateTextureState */ -static void r300TexParameter(GLcontext * ctx, GLenum target, +static void r300TexParameter(struct gl_context * ctx, GLenum target, struct gl_texture_object *texObj, GLenum pname, const GLfloat * params) { @@ -243,7 +243,7 @@ static void r300TexParameter(GLcontext * ctx, GLenum target, } } -static void r300DeleteTexture(GLcontext * ctx, struct gl_texture_object *texObj) +static void r300DeleteTexture(struct gl_context * ctx, struct gl_texture_object *texObj) { r300ContextPtr rmesa = R300_CONTEXT(ctx); radeonTexObj* t = radeon_tex_obj(texObj); @@ -284,7 +284,7 @@ static void r300DeleteTexture(GLcontext * ctx, struct gl_texture_object *texObj) * allocate the default texture objects. * Fixup MaxAnisotropy according to user preference. */ -static struct gl_texture_object *r300NewTextureObject(GLcontext * ctx, +static struct gl_texture_object *r300NewTextureObject(struct gl_context * ctx, GLuint name, GLenum target) { diff --git a/src/mesa/drivers/dri/r300/r300_tex.h b/src/mesa/drivers/dri/r300/r300_tex.h index aca44cd766..c44a39cb46 100644 --- a/src/mesa/drivers/dri/r300/r300_tex.h +++ b/src/mesa/drivers/dri/r300/r300_tex.h @@ -47,7 +47,7 @@ extern void r300SetTexOffset(__DRIcontext *pDRICtx, GLint texname, unsigned long long offset, GLint depth, GLuint pitch); -extern GLboolean r300ValidateBuffers(GLcontext * ctx); +extern GLboolean r300ValidateBuffers(struct gl_context * ctx); extern void r300InitTextureFuncs(radeonContextPtr radeon, struct dd_function_table *functions); diff --git a/src/mesa/drivers/dri/r300/r300_texstate.c b/src/mesa/drivers/dri/r300/r300_texstate.c index 9458869826..0116c5d2fa 100644 --- a/src/mesa/drivers/dri/r300/r300_texstate.c +++ b/src/mesa/drivers/dri/r300/r300_texstate.c @@ -301,7 +301,7 @@ static void setup_hardware_state(r300ContextPtr rmesa, radeonTexObj *t) * * Mostly this means populating the texture object's mipmap tree. */ -static GLboolean r300_validate_texture(GLcontext * ctx, struct gl_texture_object *texObj) +static GLboolean r300_validate_texture(struct gl_context * ctx, struct gl_texture_object *texObj) { r300ContextPtr rmesa = R300_CONTEXT(ctx); radeonTexObj *t = radeon_tex_obj(texObj); @@ -320,7 +320,7 @@ static GLboolean r300_validate_texture(GLcontext * ctx, struct gl_texture_object /** * Ensure all enabled and complete textures are uploaded along with any buffers being used. */ -GLboolean r300ValidateBuffers(GLcontext * ctx) +GLboolean r300ValidateBuffers(struct gl_context * ctx) { r300ContextPtr rmesa = R300_CONTEXT(ctx); struct radeon_renderbuffer *rrb; diff --git a/src/mesa/drivers/dri/r300/r300_vertprog.c b/src/mesa/drivers/dri/r300/r300_vertprog.c index a160128091..1daa305e3c 100644 --- a/src/mesa/drivers/dri/r300/r300_vertprog.c +++ b/src/mesa/drivers/dri/r300/r300_vertprog.c @@ -49,7 +49,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. * Write parameter array for the given vertex program into dst. * Return the total number of components written. */ -static int r300VertexProgUpdateParams(GLcontext * ctx, struct r300_vertex_program *vp, float *dst) +static int r300VertexProgUpdateParams(struct gl_context * ctx, struct r300_vertex_program *vp, float *dst) { int i; @@ -227,7 +227,7 @@ static void initialize_NV_registers(struct radeon_compiler * compiler) inst->U.I.SrcReg[0].Swizzle = RC_SWIZZLE_0000; } -static struct r300_vertex_program *build_program(GLcontext *ctx, +static struct r300_vertex_program *build_program(struct gl_context *ctx, struct r300_vertex_program_key *wanted_key, const struct gl_vertex_program *mesa_vp) { @@ -307,7 +307,7 @@ static struct r300_vertex_program *build_program(GLcontext *ctx, return vp; } -struct r300_vertex_program * r300SelectAndTranslateVertexShader(GLcontext *ctx) +struct r300_vertex_program * r300SelectAndTranslateVertexShader(struct gl_context *ctx) { r300ContextPtr r300 = R300_CONTEXT(ctx); struct r300_vertex_program_key wanted_key = { 0 }; @@ -386,7 +386,7 @@ static void r300EmitVertexProgram(r300ContextPtr r300, int dest, struct r300_ver void r300SetupVertexProgram(r300ContextPtr rmesa) { - GLcontext *ctx = rmesa->radeon.glCtx; + struct gl_context *ctx = rmesa->radeon.glCtx; struct r300_vertex_program *prog = rmesa->selected_vp; int inst_count = 0; int param_count = 0; diff --git a/src/mesa/drivers/dri/r300/r300_vertprog.h b/src/mesa/drivers/dri/r300/r300_vertprog.h index ccec896be4..ce24dcb353 100644 --- a/src/mesa/drivers/dri/r300/r300_vertprog.h +++ b/src/mesa/drivers/dri/r300/r300_vertprog.h @@ -6,6 +6,6 @@ void r300SetupVertexProgram(r300ContextPtr rmesa); -struct r300_vertex_program * r300SelectAndTranslateVertexShader(GLcontext *ctx); +struct r300_vertex_program * r300SelectAndTranslateVertexShader(struct gl_context *ctx); #endif diff --git a/src/mesa/drivers/dri/r300/radeon_context.h b/src/mesa/drivers/dri/r300/radeon_context.h index da4812d323..52b7fb91e6 100644 --- a/src/mesa/drivers/dri/r300/radeon_context.h +++ b/src/mesa/drivers/dri/r300/radeon_context.h @@ -49,12 +49,10 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "drm.h" #include "dri_util.h" -#include "radeon_screen.h" - #define FALLBACK( radeon, bit, mode ) fprintf(stderr, "%s:%s\n", __LINE__, __FILE__); /* TCL fallbacks */ -extern void radeonTclFallback(GLcontext * ctx, GLuint bit, GLboolean mode); +extern void radeonTclFallback(struct gl_context * ctx, GLuint bit, GLboolean mode); #define TCL_FALLBACK( ctx, bit, mode ) ; |