Update SwiftShader to April code dump.
April code dump from Transgaming. Adds new shader compiler.
diff --git a/src/Shader/VertexProgram.cpp b/src/Shader/VertexProgram.cpp
index 41153bf..a292a8e 100644
--- a/src/Shader/VertexProgram.cpp
+++ b/src/Shader/VertexProgram.cpp
@@ -1,6 +1,6 @@
// SwiftShader Software Renderer
//
-// Copyright(c) 2005-2011 TransGaming Inc.
+// Copyright(c) 2005-2012 TransGaming Inc.
//
// All rights reserved. No part of this software may be copied, distributed, transmitted,
// transcribed, stored in a retrieval system, translated into any human or computer
@@ -22,12 +22,13 @@
namespace sw
{
- VertexProgram::VertexProgram(const VertexProcessor::State &state, const VertexShader *vertexShader) : VertexRoutine(state), vertexShader(vertexShader)
+ VertexProgram::VertexProgram(const VertexProcessor::State &state, const VertexShader *shader) : VertexRoutine(state, shader)
{
- returns = false;
ifDepth = 0;
loopRepDepth = 0;
breakDepth = 0;
+ currentLabel = -1;
+ whileTest = false;
for(int i = 0; i < 2048; i++)
{
@@ -52,7 +53,7 @@
if(!state.preTransformed)
{
- shader(r);
+ program(r);
}
else
{
@@ -60,296 +61,320 @@
}
}
- Color4f VertexProgram::readConstant(Registers &r, const Src &src, int offset)
+ void VertexProgram::program(Registers &r)
{
- Color4f c;
+ // shader->print("VertexShader-%0.8X.txt", state.shaderID);
- int i = src.index + offset;
- bool relative = src.relative;
-
- if(!relative)
- {
- c.r = c.g = c.b = c.a = *Pointer<Float4>(r.data + OFFSET(DrawData,vs.c[i]));
-
- c.r = c.r.xxxx;
- c.g = c.g.yyyy;
- c.b = c.b.zzzz;
- c.a = c.a.wwww;
-
- if(localShaderConstants) // Constant may be known at compile time
- {
- for(int j = 0; j < vertexShader->getLength(); j++)
- {
- const ShaderInstruction &instruction = *vertexShader->getInstruction(j);
-
- if(instruction.getOpcode() == ShaderOperation::OPCODE_DEF)
- {
- if(instruction.getDestinationParameter().index == i)
- {
- c.r = Float4(instruction.getSourceParameter(0).value);
- c.g = Float4(instruction.getSourceParameter(1).value);
- c.b = Float4(instruction.getSourceParameter(2).value);
- c.a = Float4(instruction.getSourceParameter(3).value);
-
- break;
- }
- }
- }
- }
- }
- else if(src.relativeType == Src::PARAMETER_LOOP)
- {
- Int loopCounter = r.aL[r.loopDepth];
-
- c.r = c.g = c.b = c.a = *Pointer<Float4>(r.data + OFFSET(DrawData,vs.c[i]) + loopCounter * 16);
-
- c.r = c.r.xxxx;
- c.g = c.g.yyyy;
- c.b = c.b.zzzz;
- c.a = c.a.wwww;
- }
- else
- {
- Int index0;
- Int index1;
- Int index2;
- Int index3;
-
- Float4 a0_;
-
- switch(src.relativeSwizzle & 0x03)
- {
- case 0: a0_ = r.a0.x; break;
- case 1: a0_ = r.a0.y; break;
- case 2: a0_ = r.a0.z; break;
- case 3: a0_ = r.a0.w; break;
- }
-
- index0 = i + RoundInt(Float(a0_.x));
- index1 = i + RoundInt(Float(a0_.y));
- index2 = i + RoundInt(Float(a0_.z));
- index3 = i + RoundInt(Float(a0_.w));
-
- // Clamp to constant register range, c[256] = {0, 0, 0, 0}
- index0 = IfThenElse(UInt(index0) > UInt(256), Int(256), index0);
- index1 = IfThenElse(UInt(index1) > UInt(256), Int(256), index1);
- index2 = IfThenElse(UInt(index2) > UInt(256), Int(256), index2);
- index3 = IfThenElse(UInt(index3) > UInt(256), Int(256), index3);
-
- c.x = *Pointer<Float4>(r.data + OFFSET(DrawData,vs.c) + index0 * 16, 16);
- c.y = *Pointer<Float4>(r.data + OFFSET(DrawData,vs.c) + index1 * 16, 16);
- c.z = *Pointer<Float4>(r.data + OFFSET(DrawData,vs.c) + index2 * 16, 16);
- c.w = *Pointer<Float4>(r.data + OFFSET(DrawData,vs.c) + index3 * 16, 16);
-
- transpose4x4(c.x, c.y, c.z, c.w);
- }
-
- return c;
- }
-
- void VertexProgram::shader(Registers &r)
- {
- // vertexShader->print("VertexShader-%0.16llX.txt", state.shaderHash);
-
- unsigned short version = vertexShader->getVersion();
+ unsigned short version = shader->getVersion();
r.enableIndex = 0;
r.stackIndex = 0;
-
- for(int i = 0; i < vertexShader->getLength(); i++)
+
+ // Create all call site return blocks up front
+ for(int i = 0; i < shader->getLength(); i++)
{
- const ShaderInstruction *instruction = vertexShader->getInstruction(i);
- Op::Opcode opcode = instruction->getOpcode();
+ const Shader::Instruction *instruction = shader->getInstruction(i);
+ Shader::Opcode opcode = instruction->opcode;
- // #ifndef NDEBUG // FIXME: Centralize debug output control
- // vertexShader->printInstruction(i, "debug.txt");
- // #endif
+ if(opcode == Shader::OPCODE_CALL || opcode == Shader::OPCODE_CALLNZ)
+ {
+ const Dst &dst = instruction->dst;
- if(opcode == Op::OPCODE_DCL || opcode == Op::OPCODE_DEF || opcode == Op::OPCODE_DEFI || opcode == Op::OPCODE_DEFB)
+ ASSERT(callRetBlock[dst.label].size() == dst.callSite);
+ callRetBlock[dst.label].push_back(Nucleus::createBasicBlock());
+ }
+ }
+
+ for(int i = 0; i < shader->getLength(); i++)
+ {
+ const Shader::Instruction *instruction = shader->getInstruction(i);
+ Shader::Opcode opcode = instruction->opcode;
+
+ if(opcode == Shader::OPCODE_DCL || opcode == Shader::OPCODE_DEF || opcode == Shader::OPCODE_DEFI || opcode == Shader::OPCODE_DEFB)
{
continue;
}
- Dst dest = instruction->getDestinationParameter();
- Src src0 = instruction->getSourceParameter(0);
- Src src1 = instruction->getSourceParameter(1);
- Src src2 = instruction->getSourceParameter(2);
- Src src3 = instruction->getSourceParameter(3);
+ Dst dst = instruction->dst;
+ Src src0 = instruction->src[0];
+ Src src1 = instruction->src[1];
+ Src src2 = instruction->src[2];
- bool predicate = instruction->isPredicate();
- int size = vertexShader->size(opcode);
- Usage usage = instruction->getUsage();
- unsigned char usageIndex = instruction->getUsageIndex();
- Control control = instruction->getControl();
- bool integer = dest.type == Dst::PARAMETER_ADDR;
- bool pp = dest.partialPrecision;
+ bool predicate = instruction->predicate;
+ int size = shader->size(opcode);
+ Usage usage = instruction->usage;
+ unsigned char usageIndex = instruction->usageIndex;
+ Control control = instruction->control;
+ bool integer = dst.type == Shader::PARAMETER_ADDR;
+ bool pp = dst.partialPrecision;
- Color4f d;
- Color4f s0;
- Color4f s1;
- Color4f s2;
- Color4f s3;
+ Vector4f d;
+ Vector4f s0;
+ Vector4f s1;
+ Vector4f s2;
- if(src0.type != Src::PARAMETER_VOID) s0 = reg(r, src0);
- if(src1.type != Src::PARAMETER_VOID) s1 = reg(r, src1);
- if(src2.type != Src::PARAMETER_VOID) s2 = reg(r, src2);
- if(src3.type != Src::PARAMETER_VOID) s3 = reg(r, src3);
+ if(src0.type != Shader::PARAMETER_VOID) s0 = reg(r, src0);
+ if(src1.type != Shader::PARAMETER_VOID) s1 = reg(r, src1);
+ if(src2.type != Shader::PARAMETER_VOID) s2 = reg(r, src2);
switch(opcode)
{
- case Op::OPCODE_VS_1_0: break;
- case Op::OPCODE_VS_1_1: break;
- case Op::OPCODE_VS_2_0: break;
- case Op::OPCODE_VS_2_x: break;
- case Op::OPCODE_VS_2_sw: break;
- case Op::OPCODE_VS_3_0: break;
- case Op::OPCODE_VS_3_sw: break;
- case Op::OPCODE_DCL: break;
- case Op::OPCODE_DEF: break;
- case Op::OPCODE_DEFI: break;
- case Op::OPCODE_DEFB: break;
- case Op::OPCODE_NOP: break;
- case Op::OPCODE_ABS: abs(d, s0); break;
- case Op::OPCODE_ADD: add(d, s0, s1); break;
- case Op::OPCODE_CRS: crs(d, s0, s1); break;
- case Op::OPCODE_DP3: dp3(d, s0, s1); break;
- case Op::OPCODE_DP4: dp4(d, s0, s1); break;
- case Op::OPCODE_DST: dst(d, s0, s1); break;
- case Op::OPCODE_EXP: exp(d, s0, pp); break;
- case Op::OPCODE_EXPP: expp(d, s0, version); break;
- case Op::OPCODE_FRC: frc(d, s0); break;
- case Op::OPCODE_LIT: lit(d, s0); break;
- case Op::OPCODE_LOG: log(d, s0, pp); break;
- case Op::OPCODE_LOGP: logp(d, s0, version); break;
- case Op::OPCODE_LRP: lrp(d, s0, s1, s2); break;
- case Op::OPCODE_M3X2: M3X2(r, d, s0, src1); break;
- case Op::OPCODE_M3X3: M3X3(r, d, s0, src1); break;
- case Op::OPCODE_M3X4: M3X4(r, d, s0, src1); break;
- case Op::OPCODE_M4X3: M4X3(r, d, s0, src1); break;
- case Op::OPCODE_M4X4: M4X4(r, d, s0, src1); break;
- case Op::OPCODE_MAD: mad(d, s0, s1, s2); break;
- case Op::OPCODE_MAX: max(d, s0, s1); break;
- case Op::OPCODE_MIN: min(d, s0, s1); break;
- case Op::OPCODE_MOV: mov(d, s0, integer); break;
- case Op::OPCODE_MOVA: mov(d, s0); break;
- case Op::OPCODE_MUL: mul(d, s0, s1); break;
- case Op::OPCODE_NRM: nrm(d, s0, pp); break;
- case Op::OPCODE_POW: pow(d, s0, s1, pp); break;
- case Op::OPCODE_RCP: rcp(d, s0, pp); break;
- case Op::OPCODE_RSQ: rsq(d, s0, pp); break;
- case Op::OPCODE_SGE: sge(d, s0, s1); break;
- case Op::OPCODE_SGN: sgn(d, s0); break;
- case Op::OPCODE_SINCOS: sincos(d, s0, pp); break;
- case Op::OPCODE_SLT: slt(d, s0, s1); break;
- case Op::OPCODE_SUB: sub(d, s0, s1); break;
- case Op::OPCODE_BREAK: BREAK(r); break;
- case Op::OPCODE_BREAKC: BREAKC(r, s0, s1, control); break;
- case Op::OPCODE_BREAKP: BREAKP(r, src0); break;
- case Op::OPCODE_CALL: CALL(r, dest.index); break;
- case Op::OPCODE_CALLNZ: CALLNZ(r, dest.index, src0); break;
- case Op::OPCODE_ELSE: ELSE(r); break;
- case Op::OPCODE_ENDIF: ENDIF(r); break;
- case Op::OPCODE_ENDLOOP: ENDLOOP(r); break;
- case Op::OPCODE_ENDREP: ENDREP(r); break;
- case Op::OPCODE_IF: IF(r, src0); break;
- case Op::OPCODE_IFC: IFC(r, s0, s1, control); break;
- case Op::OPCODE_LABEL: LABEL(dest.index); break;
- case Op::OPCODE_LOOP: LOOP(r, src1); break;
- case Op::OPCODE_REP: REP(r, src0); break;
- case Op::OPCODE_RET: RET(r); break;
- case Op::OPCODE_SETP: setp(d, s0, s1, control); break;
- case Op::OPCODE_TEXLDL: TEXLDL(r, d, s0, src1); break;
- case Op::OPCODE_END: break;
+ case Shader::OPCODE_VS_1_0: break;
+ case Shader::OPCODE_VS_1_1: break;
+ case Shader::OPCODE_VS_2_0: break;
+ case Shader::OPCODE_VS_2_x: break;
+ case Shader::OPCODE_VS_2_sw: break;
+ case Shader::OPCODE_VS_3_0: break;
+ case Shader::OPCODE_VS_3_sw: break;
+ case Shader::OPCODE_DCL: break;
+ case Shader::OPCODE_DEF: break;
+ case Shader::OPCODE_DEFI: break;
+ case Shader::OPCODE_DEFB: break;
+ case Shader::OPCODE_NOP: break;
+ case Shader::OPCODE_ABS: abs(d, s0); break;
+ case Shader::OPCODE_ADD: add(d, s0, s1); break;
+ case Shader::OPCODE_CRS: crs(d, s0, s1); break;
+ case Shader::OPCODE_FORWARD1: forward1(d, s0, s1, s2); break;
+ case Shader::OPCODE_FORWARD2: forward2(d, s0, s1, s2); break;
+ case Shader::OPCODE_FORWARD3: forward3(d, s0, s1, s2); break;
+ case Shader::OPCODE_FORWARD4: forward4(d, s0, s1, s2); break;
+ case Shader::OPCODE_REFLECT1: reflect1(d, s0, s1); break;
+ case Shader::OPCODE_REFLECT2: reflect2(d, s0, s1); break;
+ case Shader::OPCODE_REFLECT3: reflect3(d, s0, s1); break;
+ case Shader::OPCODE_REFLECT4: reflect4(d, s0, s1); break;
+ case Shader::OPCODE_REFRACT1: refract1(d, s0, s1, s2.x); break;
+ case Shader::OPCODE_REFRACT2: refract2(d, s0, s1, s2.x); break;
+ case Shader::OPCODE_REFRACT3: refract3(d, s0, s1, s2.x); break;
+ case Shader::OPCODE_REFRACT4: refract4(d, s0, s1, s2.x); break;
+ case Shader::OPCODE_DP1: dp1(d, s0, s1); break;
+ case Shader::OPCODE_DP2: dp2(d, s0, s1); break;
+ case Shader::OPCODE_DP3: dp3(d, s0, s1); break;
+ case Shader::OPCODE_DP4: dp4(d, s0, s1); break;
+ case Shader::OPCODE_ATT: att(d, s0, s1); break;
+ case Shader::OPCODE_EXP2X: exp2x(d, s0, pp); break;
+ case Shader::OPCODE_EXP2: exp2(d, s0, pp); break;
+ case Shader::OPCODE_EXPP: expp(d, s0, version); break;
+ case Shader::OPCODE_EXP: exp(d, s0, pp); break;
+ case Shader::OPCODE_FRC: frc(d, s0); break;
+ case Shader::OPCODE_TRUNC: trunc(d, s0); break;
+ case Shader::OPCODE_FLOOR: floor(d, s0); break;
+ case Shader::OPCODE_CEIL: ceil(d, s0); break;
+ case Shader::OPCODE_LIT: lit(d, s0); break;
+ case Shader::OPCODE_LOG2X: log2x(d, s0, pp); break;
+ case Shader::OPCODE_LOG2: log2(d, s0, pp); break;
+ case Shader::OPCODE_LOGP: logp(d, s0, version); break;
+ case Shader::OPCODE_LOG: log(d, s0, pp); break;
+ case Shader::OPCODE_LRP: lrp(d, s0, s1, s2); break;
+ case Shader::OPCODE_STEP: step(d, s0, s1); break;
+ case Shader::OPCODE_SMOOTH: smooth(d, s0, s1, s2); break;
+ case Shader::OPCODE_M3X2: M3X2(r, d, s0, src1); break;
+ case Shader::OPCODE_M3X3: M3X3(r, d, s0, src1); break;
+ case Shader::OPCODE_M3X4: M3X4(r, d, s0, src1); break;
+ case Shader::OPCODE_M4X3: M4X3(r, d, s0, src1); break;
+ case Shader::OPCODE_M4X4: M4X4(r, d, s0, src1); break;
+ case Shader::OPCODE_MAD: mad(d, s0, s1, s2); break;
+ case Shader::OPCODE_MAX: max(d, s0, s1); break;
+ case Shader::OPCODE_MIN: min(d, s0, s1); break;
+ case Shader::OPCODE_MOV: mov(d, s0, integer); break;
+ case Shader::OPCODE_MOVA: mov(d, s0); break;
+ case Shader::OPCODE_F2B: f2b(d, s0); break;
+ case Shader::OPCODE_B2F: b2f(d, s0); break;
+ case Shader::OPCODE_MUL: mul(d, s0, s1); break;
+ case Shader::OPCODE_NRM2: nrm2(d, s0, pp); break;
+ case Shader::OPCODE_NRM3: nrm3(d, s0, pp); break;
+ case Shader::OPCODE_NRM4: nrm4(d, s0, pp); break;
+ case Shader::OPCODE_POWX: powx(d, s0, s1, pp); break;
+ case Shader::OPCODE_POW: pow(d, s0, s1, pp); break;
+ case Shader::OPCODE_RCPX: rcpx(d, s0, pp); break;
+ case Shader::OPCODE_DIV: div(d, s0, s1); break;
+ case Shader::OPCODE_MOD: mod(d, s0, s1); break;
+ case Shader::OPCODE_RSQX: rsqx(d, s0, pp); break;
+ case Shader::OPCODE_SQRT: sqrt(d, s0, pp); break;
+ case Shader::OPCODE_RSQ: rsq(d, s0, pp); break;
+ case Shader::OPCODE_LEN2: len2(d.x, s0, pp); break;
+ case Shader::OPCODE_LEN3: len3(d.x, s0, pp); break;
+ case Shader::OPCODE_LEN4: len4(d.x, s0, pp); break;
+ case Shader::OPCODE_DIST1: dist1(d.x, s0, s1, pp); break;
+ case Shader::OPCODE_DIST2: dist2(d.x, s0, s1, pp); break;
+ case Shader::OPCODE_DIST3: dist3(d.x, s0, s1, pp); break;
+ case Shader::OPCODE_DIST4: dist4(d.x, s0, s1, pp); break;
+ case Shader::OPCODE_SGE: step(d, s1, s0); break;
+ case Shader::OPCODE_SGN: sgn(d, s0); break;
+ case Shader::OPCODE_SINCOS: sincos(d, s0, pp); break;
+ case Shader::OPCODE_COS: cos(d, s0, pp); break;
+ case Shader::OPCODE_SIN: sin(d, s0, pp); break;
+ case Shader::OPCODE_TAN: tan(d, s0); break;
+ case Shader::OPCODE_ACOS: acos(d, s0); break;
+ case Shader::OPCODE_ASIN: asin(d, s0); break;
+ case Shader::OPCODE_ATAN: atan(d, s0); break;
+ case Shader::OPCODE_ATAN2: atan2(d, s0, s1); break;
+ case Shader::OPCODE_SLT: slt(d, s0, s1); break;
+ case Shader::OPCODE_SUB: sub(d, s0, s1); break;
+ case Shader::OPCODE_BREAK: BREAK(r); break;
+ case Shader::OPCODE_BREAKC: BREAKC(r, s0, s1, control); break;
+ case Shader::OPCODE_BREAKP: BREAKP(r, src0); break;
+ case Shader::OPCODE_CONTINUE: CONTINUE(r); break;
+ case Shader::OPCODE_TEST: TEST(); break;
+ case Shader::OPCODE_CALL: CALL(r, dst.label, dst.callSite); break;
+ case Shader::OPCODE_CALLNZ: CALLNZ(r, dst.label, dst.callSite, src0); break;
+ case Shader::OPCODE_ELSE: ELSE(r); break;
+ case Shader::OPCODE_ENDIF: ENDIF(r); break;
+ case Shader::OPCODE_ENDLOOP: ENDLOOP(r); break;
+ case Shader::OPCODE_ENDREP: ENDREP(r); break;
+ case Shader::OPCODE_ENDWHILE: ENDWHILE(r); break;
+ case Shader::OPCODE_IF: IF(r, src0); break;
+ case Shader::OPCODE_IFC: IFC(r, s0, s1, control); break;
+ case Shader::OPCODE_LABEL: LABEL(dst.index); break;
+ case Shader::OPCODE_LOOP: LOOP(r, src1); break;
+ case Shader::OPCODE_REP: REP(r, src0); break;
+ case Shader::OPCODE_WHILE: WHILE(r, src0); break;
+ case Shader::OPCODE_RET: RET(r); break;
+ case Shader::OPCODE_LEAVE: LEAVE(r); break;
+ case Shader::OPCODE_CMP: cmp(d, s0, s1, control); break;
+ case Shader::OPCODE_ICMP: icmp(d, s0, s1, control); break;
+ case Shader::OPCODE_SELECT: select(d, s0, s1, s2); break;
+ case Shader::OPCODE_EXTRACT: extract(d.x, s0, s1.x); break;
+ case Shader::OPCODE_INSERT: insert(d, s0, s1.x, s2.x); break;
+ case Shader::OPCODE_ALL: all(d.x, s0); break;
+ case Shader::OPCODE_ANY: any(d.x, s0); break;
+ case Shader::OPCODE_NOT: not(d, s0); break;
+ case Shader::OPCODE_OR: or(d.x, s0.x, s1.x); break;
+ case Shader::OPCODE_XOR: xor(d.x, s0.x, s1.x); break;
+ case Shader::OPCODE_AND: and(d.x, s0.x, s1.x); break;
+ case Shader::OPCODE_TEXLDL: TEXLDL(r, d, s0, src1); break;
+ case Shader::OPCODE_TEX: TEX(r, d, s0, src1); break;
+ case Shader::OPCODE_END: break;
default:
ASSERT(false);
}
- if(dest.type != Dst::PARAMETER_VOID && dest.type != Dst::PARAMETER_LABEL)
+ if(dst.type != Shader::PARAMETER_VOID && dst.type != Shader::PARAMETER_LABEL && opcode != Shader::OPCODE_NOP)
{
- if(dest.saturate)
+ if(dst.integer)
{
- if(dest.x) d.r = Max(d.r, Float4(0.0f, 0.0f, 0.0f, 0.0f));
- if(dest.y) d.g = Max(d.g, Float4(0.0f, 0.0f, 0.0f, 0.0f));
- if(dest.z) d.b = Max(d.b, Float4(0.0f, 0.0f, 0.0f, 0.0f));
- if(dest.w) d.a = Max(d.a, Float4(0.0f, 0.0f, 0.0f, 0.0f));
-
- if(dest.x) d.r = Min(d.r, Float4(1.0f, 1.0f, 1.0f, 1.0f));
- if(dest.y) d.g = Min(d.g, Float4(1.0f, 1.0f, 1.0f, 1.0f));
- if(dest.z) d.b = Min(d.b, Float4(1.0f, 1.0f, 1.0f, 1.0f));
- if(dest.w) d.a = Min(d.a, Float4(1.0f, 1.0f, 1.0f, 1.0f));
+ switch(opcode)
+ {
+ case Shader::OPCODE_DIV:
+ if(dst.x) d.x = Trunc(d.x);
+ if(dst.y) d.y = Trunc(d.y);
+ if(dst.z) d.z = Trunc(d.z);
+ if(dst.w) d.w = Trunc(d.w);
+ break;
+ default:
+ break; // No truncation to integer required when arguments are integer
+ }
}
- if(vertexShader->containsDynamicBranching())
+ if(dst.saturate)
{
- Color4f pDst; // FIXME: Rename
+ if(dst.x) d.x = Max(d.x, Float4(0.0f));
+ if(dst.y) d.y = Max(d.y, Float4(0.0f));
+ if(dst.z) d.z = Max(d.z, Float4(0.0f));
+ if(dst.w) d.w = Max(d.w, Float4(0.0f));
- switch(dest.type)
+ if(dst.x) d.x = Min(d.x, Float4(1.0f));
+ if(dst.y) d.y = Min(d.y, Float4(1.0f));
+ if(dst.z) d.z = Min(d.z, Float4(1.0f));
+ if(dst.w) d.w = Min(d.w, Float4(1.0f));
+ }
+
+ if(shader->containsDynamicBranching())
+ {
+ Vector4f pDst; // FIXME: Rename
+
+ switch(dst.type)
{
- case Dst::PARAMETER_VOID: break;
- case Dst::PARAMETER_TEMP: pDst = r.r[dest.index]; break;
- case Dst::PARAMETER_ADDR: pDst = r.a0; break;
- case Dst::PARAMETER_RASTOUT:
- switch(dest.index)
+ case Shader::PARAMETER_VOID: break;
+ case Shader::PARAMETER_TEMP:
+ if(dst.rel.type == Shader::PARAMETER_VOID)
+ {
+ if(dst.x) pDst.x = r.r[dst.index].x;
+ if(dst.y) pDst.y = r.r[dst.index].y;
+ if(dst.z) pDst.z = r.r[dst.index].z;
+ if(dst.w) pDst.w = r.r[dst.index].w;
+ }
+ else
+ {
+ Int a = relativeAddress(r, dst);
+
+ if(dst.x) pDst.x = r.r[dst.index + a].x;
+ if(dst.y) pDst.y = r.r[dst.index + a].y;
+ if(dst.z) pDst.z = r.r[dst.index + a].z;
+ if(dst.w) pDst.w = r.r[dst.index + a].w;
+ }
+ break;
+ case Shader::PARAMETER_ADDR: pDst = r.a0; break;
+ case Shader::PARAMETER_RASTOUT:
+ switch(dst.index)
{
case 0:
- if(dest.x) pDst.x = r.ox[Pos];
- if(dest.y) pDst.y = r.oy[Pos];
- if(dest.z) pDst.z = r.oz[Pos];
- if(dest.w) pDst.w = r.ow[Pos];
+ if(dst.x) pDst.x = r.o[Pos].x;
+ if(dst.y) pDst.y = r.o[Pos].y;
+ if(dst.z) pDst.z = r.o[Pos].z;
+ if(dst.w) pDst.w = r.o[Pos].w;
break;
case 1:
- pDst.x = r.ox[Fog];
+ pDst.x = r.o[Fog].x;
break;
case 2:
- pDst.x = r.oy[Pts];
+ pDst.x = r.o[Pts].y;
break;
default:
ASSERT(false);
}
break;
- case Dst::PARAMETER_ATTROUT:
- if(dest.x) pDst.x = r.ox[D0 + dest.index];
- if(dest.y) pDst.y = r.oy[D0 + dest.index];
- if(dest.z) pDst.z = r.oz[D0 + dest.index];
- if(dest.w) pDst.w = r.ow[D0 + dest.index];
+ case Shader::PARAMETER_ATTROUT:
+ if(dst.x) pDst.x = r.o[D0 + dst.index].x;
+ if(dst.y) pDst.y = r.o[D0 + dst.index].y;
+ if(dst.z) pDst.z = r.o[D0 + dst.index].z;
+ if(dst.w) pDst.w = r.o[D0 + dst.index].w;
break;
- case Dst::PARAMETER_TEXCRDOUT:
- // case Dst::PARAMETER_OUTPUT:
+ case Shader::PARAMETER_TEXCRDOUT:
+ // case Shader::PARAMETER_OUTPUT:
if(version < 0x0300)
{
- if(dest.x) pDst.x = r.ox[T0 + dest.index];
- if(dest.y) pDst.y = r.oy[T0 + dest.index];
- if(dest.z) pDst.z = r.oz[T0 + dest.index];
- if(dest.w) pDst.w = r.ow[T0 + dest.index];
+ if(dst.x) pDst.x = r.o[T0 + dst.index].x;
+ if(dst.y) pDst.y = r.o[T0 + dst.index].y;
+ if(dst.z) pDst.z = r.o[T0 + dst.index].z;
+ if(dst.w) pDst.w = r.o[T0 + dst.index].w;
}
else
{
- if(!dest.relative)
+ if(dst.rel.type == Shader::PARAMETER_VOID) // Not relative
{
- if(dest.x) pDst.x = r.ox[dest.index];
- if(dest.y) pDst.y = r.oy[dest.index];
- if(dest.z) pDst.z = r.oz[dest.index];
- if(dest.w) pDst.w = r.ow[dest.index];
+ if(dst.x) pDst.x = r.o[dst.index].x;
+ if(dst.y) pDst.y = r.o[dst.index].y;
+ if(dst.z) pDst.z = r.o[dst.index].z;
+ if(dst.w) pDst.w = r.o[dst.index].w;
}
- else
+ else if(dst.rel.type == Shader::PARAMETER_LOOP)
{
Int aL = r.aL[r.loopDepth];
- if(dest.x) pDst.x = r.ox[dest.index + aL];
- if(dest.y) pDst.y = r.oy[dest.index + aL];
- if(dest.z) pDst.z = r.oz[dest.index + aL];
- if(dest.w) pDst.w = r.ow[dest.index + aL];
+ if(dst.x) pDst.x = r.o[dst.index + aL].x;
+ if(dst.y) pDst.y = r.o[dst.index + aL].y;
+ if(dst.z) pDst.z = r.o[dst.index + aL].z;
+ if(dst.w) pDst.w = r.o[dst.index + aL].w;
+ }
+ else
+ {
+ Int a = relativeAddress(r, dst);
+
+ if(dst.x) pDst.x = r.o[dst.index + a].x;
+ if(dst.y) pDst.y = r.o[dst.index + a].y;
+ if(dst.z) pDst.z = r.o[dst.index + a].z;
+ if(dst.w) pDst.w = r.o[dst.index + a].w;
}
}
break;
- case Dst::PARAMETER_LABEL: break;
- case Dst::PARAMETER_PREDICATE: pDst = r.p0; break;
- case Dst::PARAMETER_INPUT: break;
+ case Shader::PARAMETER_LABEL: break;
+ case Shader::PARAMETER_PREDICATE: pDst = r.p0; break;
+ case Shader::PARAMETER_INPUT: break;
default:
ASSERT(false);
}
- Int4 enable = r.enableStack[r.enableIndex] & r.enableBreak;
+ Int4 enable = enableMask(r, instruction);
Int4 xEnable = enable;
Int4 yEnable = enable;
@@ -358,119 +383,140 @@
if(predicate)
{
- unsigned char pSwizzle = instruction->getPredicateSwizzle();
+ unsigned char pSwizzle = instruction->predicateSwizzle;
Float4 xPredicate = r.p0[(pSwizzle >> 0) & 0x03];
Float4 yPredicate = r.p0[(pSwizzle >> 2) & 0x03];
Float4 zPredicate = r.p0[(pSwizzle >> 4) & 0x03];
Float4 wPredicate = r.p0[(pSwizzle >> 6) & 0x03];
- if(!instruction->isPredicateNot())
+ if(!instruction->predicateNot)
{
- if(dest.x) xEnable = xEnable & As<Int4>(xPredicate);
- if(dest.y) yEnable = yEnable & As<Int4>(yPredicate);
- if(dest.z) zEnable = zEnable & As<Int4>(zPredicate);
- if(dest.w) wEnable = wEnable & As<Int4>(wPredicate);
+ if(dst.x) xEnable = xEnable & As<Int4>(xPredicate);
+ if(dst.y) yEnable = yEnable & As<Int4>(yPredicate);
+ if(dst.z) zEnable = zEnable & As<Int4>(zPredicate);
+ if(dst.w) wEnable = wEnable & As<Int4>(wPredicate);
}
else
{
- if(dest.x) xEnable = xEnable & ~As<Int4>(xPredicate);
- if(dest.y) yEnable = yEnable & ~As<Int4>(yPredicate);
- if(dest.z) zEnable = zEnable & ~As<Int4>(zPredicate);
- if(dest.w) wEnable = wEnable & ~As<Int4>(wPredicate);
+ if(dst.x) xEnable = xEnable & ~As<Int4>(xPredicate);
+ if(dst.y) yEnable = yEnable & ~As<Int4>(yPredicate);
+ if(dst.z) zEnable = zEnable & ~As<Int4>(zPredicate);
+ if(dst.w) wEnable = wEnable & ~As<Int4>(wPredicate);
}
}
- if(dest.x) d.x = As<Float4>(As<Int4>(d.x) & xEnable);
- if(dest.y) d.y = As<Float4>(As<Int4>(d.y) & yEnable);
- if(dest.z) d.z = As<Float4>(As<Int4>(d.z) & zEnable);
- if(dest.w) d.w = As<Float4>(As<Int4>(d.w) & wEnable);
+ if(dst.x) d.x = As<Float4>(As<Int4>(d.x) & xEnable);
+ if(dst.y) d.y = As<Float4>(As<Int4>(d.y) & yEnable);
+ if(dst.z) d.z = As<Float4>(As<Int4>(d.z) & zEnable);
+ if(dst.w) d.w = As<Float4>(As<Int4>(d.w) & wEnable);
- if(dest.x) d.x = As<Float4>(As<Int4>(d.x) | (As<Int4>(pDst.x) & ~xEnable));
- if(dest.y) d.y = As<Float4>(As<Int4>(d.y) | (As<Int4>(pDst.y) & ~yEnable));
- if(dest.z) d.z = As<Float4>(As<Int4>(d.z) | (As<Int4>(pDst.z) & ~zEnable));
- if(dest.w) d.w = As<Float4>(As<Int4>(d.w) | (As<Int4>(pDst.w) & ~wEnable));
+ if(dst.x) d.x = As<Float4>(As<Int4>(d.x) | (As<Int4>(pDst.x) & ~xEnable));
+ if(dst.y) d.y = As<Float4>(As<Int4>(d.y) | (As<Int4>(pDst.y) & ~yEnable));
+ if(dst.z) d.z = As<Float4>(As<Int4>(d.z) | (As<Int4>(pDst.z) & ~zEnable));
+ if(dst.w) d.w = As<Float4>(As<Int4>(d.w) | (As<Int4>(pDst.w) & ~wEnable));
}
- switch(dest.type)
+ switch(dst.type)
{
- case Dst::PARAMETER_VOID:
+ case Shader::PARAMETER_VOID:
break;
- case Dst::PARAMETER_TEMP:
- if(dest.x) r.r[dest.index].x = d.x;
- if(dest.y) r.r[dest.index].y = d.y;
- if(dest.z) r.r[dest.index].z = d.z;
- if(dest.w) r.r[dest.index].w = d.w;
+ case Shader::PARAMETER_TEMP:
+ if(dst.rel.type == Shader::PARAMETER_VOID)
+ {
+ if(dst.x) r.r[dst.index].x = d.x;
+ if(dst.y) r.r[dst.index].y = d.y;
+ if(dst.z) r.r[dst.index].z = d.z;
+ if(dst.w) r.r[dst.index].w = d.w;
+ }
+ else
+ {
+ Int a = relativeAddress(r, dst);
+
+ if(dst.x) r.r[dst.index + a].x = d.x;
+ if(dst.y) r.r[dst.index + a].y = d.y;
+ if(dst.z) r.r[dst.index + a].z = d.z;
+ if(dst.w) r.r[dst.index + a].w = d.w;
+ }
break;
- case Dst::PARAMETER_ADDR:
- if(dest.x) r.a0.x = d.x;
- if(dest.y) r.a0.y = d.y;
- if(dest.z) r.a0.z = d.z;
- if(dest.w) r.a0.w = d.w;
+ case Shader::PARAMETER_ADDR:
+ if(dst.x) r.a0.x = d.x;
+ if(dst.y) r.a0.y = d.y;
+ if(dst.z) r.a0.z = d.z;
+ if(dst.w) r.a0.w = d.w;
break;
- case Dst::PARAMETER_RASTOUT:
- switch(dest.index)
+ case Shader::PARAMETER_RASTOUT:
+ switch(dst.index)
{
case 0:
- if(dest.x) r.ox[Pos] = d.x;
- if(dest.y) r.oy[Pos] = d.y;
- if(dest.z) r.oz[Pos] = d.z;
- if(dest.w) r.ow[Pos] = d.w;
+ if(dst.x) r.o[Pos].x = d.x;
+ if(dst.y) r.o[Pos].y = d.y;
+ if(dst.z) r.o[Pos].z = d.z;
+ if(dst.w) r.o[Pos].w = d.w;
break;
case 1:
- r.ox[Fog] = d.x;
+ r.o[Fog].x = d.x;
break;
case 2:
- r.oy[Pts] = d.x;
+ r.o[Pts].y = d.x;
break;
default: ASSERT(false);
}
break;
- case Dst::PARAMETER_ATTROUT:
- if(dest.x) r.ox[D0 + dest.index] = d.x;
- if(dest.y) r.oy[D0 + dest.index] = d.y;
- if(dest.z) r.oz[D0 + dest.index] = d.z;
- if(dest.w) r.ow[D0 + dest.index] = d.w;
+ case Shader::PARAMETER_ATTROUT:
+ if(dst.x) r.o[D0 + dst.index].x = d.x;
+ if(dst.y) r.o[D0 + dst.index].y = d.y;
+ if(dst.z) r.o[D0 + dst.index].z = d.z;
+ if(dst.w) r.o[D0 + dst.index].w = d.w;
break;
- case Dst::PARAMETER_TEXCRDOUT:
- // case Dst::PARAMETER_OUTPUT:
+ case Shader::PARAMETER_TEXCRDOUT:
+ // case Shader::PARAMETER_OUTPUT:
if(version < 0x0300)
{
- if(dest.x) r.ox[T0 + dest.index] = d.x;
- if(dest.y) r.oy[T0 + dest.index] = d.y;
- if(dest.z) r.oz[T0 + dest.index] = d.z;
- if(dest.w) r.ow[T0 + dest.index] = d.w;
+ if(dst.x) r.o[T0 + dst.index].x = d.x;
+ if(dst.y) r.o[T0 + dst.index].y = d.y;
+ if(dst.z) r.o[T0 + dst.index].z = d.z;
+ if(dst.w) r.o[T0 + dst.index].w = d.w;
}
else
{
- if(!dest.relative)
+ if(dst.rel.type == Shader::PARAMETER_VOID) // Not relative
{
- if(dest.x) r.ox[dest.index] = d.x;
- if(dest.y) r.oy[dest.index] = d.y;
- if(dest.z) r.oz[dest.index] = d.z;
- if(dest.w) r.ow[dest.index] = d.w;
+ if(dst.x) r.o[dst.index].x = d.x;
+ if(dst.y) r.o[dst.index].y = d.y;
+ if(dst.z) r.o[dst.index].z = d.z;
+ if(dst.w) r.o[dst.index].w = d.w;
}
- else
+ else if(dst.rel.type == Shader::PARAMETER_LOOP)
{
Int aL = r.aL[r.loopDepth];
- if(dest.x) r.ox[dest.index + aL] = d.x;
- if(dest.y) r.oy[dest.index + aL] = d.y;
- if(dest.z) r.oz[dest.index + aL] = d.z;
- if(dest.w) r.ow[dest.index + aL] = d.w;
+ if(dst.x) r.o[dst.index + aL].x = d.x;
+ if(dst.y) r.o[dst.index + aL].y = d.y;
+ if(dst.z) r.o[dst.index + aL].z = d.z;
+ if(dst.w) r.o[dst.index + aL].w = d.w;
+ }
+ else
+ {
+ Int a = relativeAddress(r, dst);
+
+ if(dst.x) r.o[dst.index + a].x = d.x;
+ if(dst.y) r.o[dst.index + a].y = d.y;
+ if(dst.z) r.o[dst.index + a].z = d.z;
+ if(dst.w) r.o[dst.index + a].w = d.w;
}
}
break;
- case Dst::PARAMETER_LABEL: break;
- case Dst::PARAMETER_PREDICATE: r.p0 = d; break;
- case Dst::PARAMETER_INPUT: break;
+ case Shader::PARAMETER_LABEL: break;
+ case Shader::PARAMETER_PREDICATE: r.p0 = d; break;
+ case Shader::PARAMETER_INPUT: break;
default:
ASSERT(false);
}
}
}
- if(returns)
+ if(currentLabel != -1)
{
Nucleus::setInsertBlock(returnBlock);
}
@@ -478,40 +524,40 @@
void VertexProgram::passThrough(Registers &r)
{
- if(vertexShader)
+ if(shader)
{
for(int i = 0; i < 12; i++)
{
- unsigned char usage = vertexShader->output[i][0].usage;
- unsigned char index = vertexShader->output[i][0].index;
+ unsigned char usage = shader->output[i][0].usage;
+ unsigned char index = shader->output[i][0].index;
switch(usage)
{
case 0xFF:
continue;
- case ShaderOperation::USAGE_PSIZE:
- r.oy[i] = r.v[i].x;
+ case Shader::USAGE_PSIZE:
+ r.o[i].y = r.v[i].x;
break;
- case ShaderOperation::USAGE_TEXCOORD:
- r.ox[i] = r.v[i].x;
- r.oy[i] = r.v[i].y;
- r.oz[i] = r.v[i].z;
- r.ow[i] = r.v[i].w;
+ case Shader::USAGE_TEXCOORD:
+ r.o[i].x = r.v[i].x;
+ r.o[i].y = r.v[i].y;
+ r.o[i].z = r.v[i].z;
+ r.o[i].w = r.v[i].w;
break;
- case ShaderOperation::USAGE_POSITION:
- r.ox[i] = r.v[i].x;
- r.oy[i] = r.v[i].y;
- r.oz[i] = r.v[i].z;
- r.ow[i] = r.v[i].w;
+ case Shader::USAGE_POSITION:
+ r.o[i].x = r.v[i].x;
+ r.o[i].y = r.v[i].y;
+ r.o[i].z = r.v[i].z;
+ r.o[i].w = r.v[i].w;
break;
- case ShaderOperation::USAGE_COLOR:
- r.ox[i] = r.v[i].x;
- r.oy[i] = r.v[i].y;
- r.oz[i] = r.v[i].z;
- r.ow[i] = r.v[i].w;
+ case Shader::USAGE_COLOR:
+ r.o[i].x = r.v[i].x;
+ r.o[i].y = r.v[i].y;
+ r.o[i].z = r.v[i].z;
+ r.o[i].w = r.v[i].w;
break;
- case ShaderOperation::USAGE_FOG:
- r.ox[i] = r.v[i].x;
+ case Shader::USAGE_FOG:
+ r.o[i].x = r.v[i].x;
break;
default:
ASSERT(false);
@@ -520,60 +566,99 @@
}
else
{
- r.ox[Pos] = r.v[PositionT].x;
- r.oy[Pos] = r.v[PositionT].y;
- r.oz[Pos] = r.v[PositionT].z;
- r.ow[Pos] = r.v[PositionT].w;
+ r.o[Pos].x = r.v[PositionT].x;
+ r.o[Pos].y = r.v[PositionT].y;
+ r.o[Pos].z = r.v[PositionT].z;
+ r.o[Pos].w = r.v[PositionT].w;
for(int i = 0; i < 2; i++)
{
- r.ox[D0 + i] = r.v[Color0 + i].x;
- r.oy[D0 + i] = r.v[Color0 + i].y;
- r.oz[D0 + i] = r.v[Color0 + i].z;
- r.ow[D0 + i] = r.v[Color0 + i].w;
+ r.o[D0 + i].x = r.v[Color0 + i].x;
+ r.o[D0 + i].y = r.v[Color0 + i].y;
+ r.o[D0 + i].z = r.v[Color0 + i].z;
+ r.o[D0 + i].w = r.v[Color0 + i].w;
}
for(int i = 0; i < 8; i++)
{
- r.ox[T0 + i] = r.v[TexCoord0 + i].x;
- r.oy[T0 + i] = r.v[TexCoord0 + i].y;
- r.oz[T0 + i] = r.v[TexCoord0 + i].z;
- r.ow[T0 + i] = r.v[TexCoord0 + i].w;
+ r.o[T0 + i].x = r.v[TexCoord0 + i].x;
+ r.o[T0 + i].y = r.v[TexCoord0 + i].y;
+ r.o[T0 + i].z = r.v[TexCoord0 + i].z;
+ r.o[T0 + i].w = r.v[TexCoord0 + i].w;
}
- r.oy[Pts] = r.v[PSize].x;
+ r.o[Pts].y = r.v[PSize].x;
}
}
- Color4f VertexProgram::reg(Registers &r, const Src &src, int offset)
+ Vector4f VertexProgram::reg(Registers &r, const Src &src, int offset)
{
int i = src.index + offset;
- Color4f reg;
+ Vector4f reg;
- if(src.type == Src::PARAMETER_CONST)
- {
- reg = readConstant(r, src, offset);
- }
-
switch(src.type)
{
- case Src::PARAMETER_TEMP: reg = r.r[i]; break;
- case Src::PARAMETER_CONST: break;
- case Src::PARAMETER_INPUT: reg = r.v[i]; break;
- case Src::PARAMETER_VOID: return r.r[0]; // Dummy
- case Src::PARAMETER_FLOATLITERAL: return r.r[0]; // Dummy
- case Src::PARAMETER_ADDR: reg = r.a0; break;
- case Src::PARAMETER_CONSTBOOL: return r.r[0]; // Dummy
- case Src::PARAMETER_CONSTINT: return r.r[0]; // Dummy
- case Src::PARAMETER_LOOP: return r.r[0]; // Dummy
- case Src::PARAMETER_PREDICATE: return r.r[0]; // Dummy
- case Src::PARAMETER_SAMPLER: return r.r[0]; // Dummy
+ case Shader::PARAMETER_TEMP:
+ if(src.rel.type == Shader::PARAMETER_VOID)
+ {
+ reg = r.r[i];
+ }
+ else
+ {
+ reg = r.r[i + relativeAddress(r, src)];
+ }
+ break;
+ case Shader::PARAMETER_CONST:
+ reg = readConstant(r, src, offset);
+ break;
+ case Shader::PARAMETER_INPUT:
+ if(src.rel.type == Shader::PARAMETER_VOID)
+ {
+ reg = r.v[i];
+ }
+ else
+ {
+ reg = r.v[i + relativeAddress(r, src)];
+ }
+ break;
+ case Shader::PARAMETER_VOID: return r.r[0]; // Dummy
+ case Shader::PARAMETER_FLOAT4LITERAL:
+ reg.x = Float4(src.value[0]);
+ reg.y = Float4(src.value[1]);
+ reg.z = Float4(src.value[2]);
+ reg.w = Float4(src.value[3]);
+ break;
+ case Shader::PARAMETER_ADDR: reg = r.a0; break;
+ case Shader::PARAMETER_CONSTBOOL: return r.r[0]; // Dummy
+ case Shader::PARAMETER_CONSTINT: return r.r[0]; // Dummy
+ case Shader::PARAMETER_LOOP: return r.r[0]; // Dummy
+ case Shader::PARAMETER_PREDICATE: return r.r[0]; // Dummy
+ case Shader::PARAMETER_SAMPLER:
+ if(src.rel.type == Shader::PARAMETER_VOID)
+ {
+ reg.x = As<Float4>(Int4(i));
+ }
+ else if(src.rel.type == Shader::PARAMETER_TEMP)
+ {
+ reg.x = As<Float4>(Int4(i) + RoundInt(r.r[src.rel.index].x));
+ }
+ return reg;
+ case Shader::PARAMETER_OUTPUT:
+ if(src.rel.type == Shader::PARAMETER_VOID)
+ {
+ reg = r.o[i];
+ }
+ else
+ {
+ reg = r.o[i + relativeAddress(r, src)];
+ }
+ break;
default:
ASSERT(false);
}
- Color4f mod;
+ Vector4f mod;
mod.x = reg[(src.swizzle >> 0) & 0x03];
mod.y = reg[(src.swizzle >> 2) & 0x03];
@@ -582,54 +667,54 @@
switch(src.modifier)
{
- case Src::MODIFIER_NONE:
+ case Shader::MODIFIER_NONE:
break;
- case Src::MODIFIER_NEGATE:
+ case Shader::MODIFIER_NEGATE:
mod.x = -mod.x;
mod.y = -mod.y;
mod.z = -mod.z;
mod.w = -mod.w;
break;
- case Src::MODIFIER_BIAS:
+ case Shader::MODIFIER_BIAS:
ASSERT(false); // NOTE: Unimplemented
break;
- case Src::MODIFIER_BIAS_NEGATE:
+ case Shader::MODIFIER_BIAS_NEGATE:
ASSERT(false); // NOTE: Unimplemented
break;
- case Src::MODIFIER_SIGN:
+ case Shader::MODIFIER_SIGN:
ASSERT(false); // NOTE: Unimplemented
break;
- case Src::MODIFIER_SIGN_NEGATE:
+ case Shader::MODIFIER_SIGN_NEGATE:
ASSERT(false); // NOTE: Unimplemented
break;
- case Src::MODIFIER_COMPLEMENT:
+ case Shader::MODIFIER_COMPLEMENT:
ASSERT(false); // NOTE: Unimplemented
break;
- case Src::MODIFIER_X2:
+ case Shader::MODIFIER_X2:
ASSERT(false); // NOTE: Unimplemented
break;
- case Src::MODIFIER_X2_NEGATE:
+ case Shader::MODIFIER_X2_NEGATE:
ASSERT(false); // NOTE: Unimplemented
break;
- case Src::MODIFIER_DZ:
+ case Shader::MODIFIER_DZ:
ASSERT(false); // NOTE: Unimplemented
break;
- case Src::MODIFIER_DW:
+ case Shader::MODIFIER_DW:
ASSERT(false); // NOTE: Unimplemented
break;
- case Src::MODIFIER_ABS:
+ case Shader::MODIFIER_ABS:
mod.x = Abs(mod.x);
mod.y = Abs(mod.y);
mod.z = Abs(mod.z);
mod.w = Abs(mod.w);
break;
- case Src::MODIFIER_ABS_NEGATE:
+ case Shader::MODIFIER_ABS_NEGATE:
mod.x = -Abs(mod.x);
mod.y = -Abs(mod.y);
mod.z = -Abs(mod.z);
mod.w = -Abs(mod.w);
break;
- case Src::MODIFIER_NOT:
+ case Shader::MODIFIER_NOT:
UNIMPLEMENTED();
break;
default:
@@ -639,32 +724,177 @@
return mod;
}
- void VertexProgram::M3X2(Registers &r, Color4f &dst, Color4f &src0, Src &src1)
+ Vector4f VertexProgram::readConstant(Registers &r, const Src &src, int offset)
{
- Color4f row0 = reg(r, src1, 0);
- Color4f row1 = reg(r, src1, 1);
+ Vector4f c;
+
+ int i = src.index + offset;
+
+ if(src.rel.type == Shader::PARAMETER_VOID) // Not relative
+ {
+ c.x = c.y = c.z = c.w = *Pointer<Float4>(r.data + OFFSET(DrawData,vs.c[i]));
+
+ c.x = c.x.xxxx;
+ c.y = c.y.yyyy;
+ c.z = c.z.zzzz;
+ c.w = c.w.wwww;
+
+ if(localShaderConstants) // Constant may be known at compile time
+ {
+ for(int j = 0; j < shader->getLength(); j++)
+ {
+ const Shader::Instruction &instruction = *shader->getInstruction(j);
+
+ if(instruction.opcode == Shader::OPCODE_DEF)
+ {
+ if(instruction.dst.index == i)
+ {
+ c.x = Float4(instruction.src[0].value[0]);
+ c.y = Float4(instruction.src[0].value[1]);
+ c.z = Float4(instruction.src[0].value[2]);
+ c.w = Float4(instruction.src[0].value[3]);
+
+ break;
+ }
+ }
+ }
+ }
+ }
+ else if(src.rel.type == Shader::PARAMETER_LOOP)
+ {
+ Int loopCounter = r.aL[r.loopDepth];
+
+ c.x = c.y = c.z = c.w = *Pointer<Float4>(r.data + OFFSET(DrawData,vs.c[i]) + loopCounter * 16);
+
+ c.x = c.x.xxxx;
+ c.y = c.y.yyyy;
+ c.z = c.z.zzzz;
+ c.w = c.w.wwww;
+ }
+ else
+ {
+ if(src.rel.deterministic)
+ {
+ Int a = relativeAddress(r, src);
+
+ c.x = c.y = c.z = c.w = *Pointer<Float4>(r.data + OFFSET(DrawData,vs.c[i]) + a * 16);
+
+ c.x = c.x.xxxx;
+ c.y = c.y.yyyy;
+ c.z = c.z.zzzz;
+ c.w = c.w.wwww;
+ }
+ else
+ {
+ int component = src.rel.swizzle & 0x03;
+ Float4 a;
+
+ switch(src.rel.type)
+ {
+ case Shader::PARAMETER_ADDR: a = r.a0[component]; break;
+ case Shader::PARAMETER_TEMP: a = r.r[src.rel.index][component]; break;
+ case Shader::PARAMETER_INPUT: a = r.v[src.rel.index][component]; break;
+ case Shader::PARAMETER_OUTPUT: a = r.o[src.rel.index][component]; break;
+ case Shader::PARAMETER_CONST: a = Float4(*Pointer<Float>(r.data + OFFSET(DrawData,vs.c[src.rel.index][component]))); break;
+ default: ASSERT(false);
+ }
+
+ Int4 index = Int4(i) + RoundInt(a) * Int4(src.rel.scale);
+
+ index = Min(As<UInt4>(index), UInt4(256)); // Clamp to constant register range, c[256] = {0, 0, 0, 0}
+
+ Int index0 = Extract(index, 0);
+ Int index1 = Extract(index, 1);
+ Int index2 = Extract(index, 2);
+ Int index3 = Extract(index, 3);
+
+ c.x = *Pointer<Float4>(r.data + OFFSET(DrawData,vs.c) + index0 * 16, 16);
+ c.y = *Pointer<Float4>(r.data + OFFSET(DrawData,vs.c) + index1 * 16, 16);
+ c.z = *Pointer<Float4>(r.data + OFFSET(DrawData,vs.c) + index2 * 16, 16);
+ c.w = *Pointer<Float4>(r.data + OFFSET(DrawData,vs.c) + index3 * 16, 16);
+
+ transpose4x4(c.x, c.y, c.z, c.w);
+ }
+ }
+
+ return c;
+ }
+
+ Int VertexProgram::relativeAddress(Registers &r, const Shader::Parameter &var)
+ {
+ ASSERT(var.rel.deterministic);
+
+ if(var.rel.type == Shader::PARAMETER_TEMP)
+ {
+ return RoundInt(Extract(r.r[var.rel.index].x, 0)) * var.rel.scale;
+ }
+ else if(var.rel.type == Shader::PARAMETER_INPUT)
+ {
+ return RoundInt(Extract(r.v[var.rel.index].x, 0)) * var.rel.scale;
+ }
+ else if(var.rel.type == Shader::PARAMETER_OUTPUT)
+ {
+ return RoundInt(Extract(r.o[var.rel.index].x, 0)) * var.rel.scale;
+ }
+ else if(var.rel.type == Shader::PARAMETER_CONST)
+ {
+ RValue<Float4> c = *Pointer<Float4>(r.data + OFFSET(DrawData,vs.c[var.rel.index]));
+
+ return RoundInt(Extract(c, 0)) * var.rel.scale;
+ }
+ else ASSERT(false);
+
+ return 0;
+ }
+
+ Int4 VertexProgram::enableMask(Registers &r, const Shader::Instruction *instruction)
+ {
+ Int4 enable = instruction->analysisBranch ? Int4(r.enableStack[r.enableIndex]) : Int4(0xFFFFFFFF);
+
+ if(shader->containsBreakInstruction() && !whileTest && instruction->analysisBreak)
+ {
+ enable &= r.enableBreak;
+ }
+
+ if(shader->containsContinueInstruction() && !whileTest && instruction->analysisContinue)
+ {
+ enable &= r.enableContinue;
+ }
+
+ if(shader->containsLeaveInstruction() && instruction->analysisLeave)
+ {
+ enable &= r.enableLeave;
+ }
+
+ return enable;
+ }
+
+ void VertexProgram::M3X2(Registers &r, Vector4f &dst, Vector4f &src0, Src &src1)
+ {
+ Vector4f row0 = reg(r, src1, 0);
+ Vector4f row1 = reg(r, src1, 1);
dst.x = dot3(src0, row0);
dst.y = dot3(src0, row1);
}
- void VertexProgram::M3X3(Registers &r, Color4f &dst, Color4f &src0, Src &src1)
+ void VertexProgram::M3X3(Registers &r, Vector4f &dst, Vector4f &src0, Src &src1)
{
- Color4f row0 = reg(r, src1, 0);
- Color4f row1 = reg(r, src1, 1);
- Color4f row2 = reg(r, src1, 2);
+ Vector4f row0 = reg(r, src1, 0);
+ Vector4f row1 = reg(r, src1, 1);
+ Vector4f row2 = reg(r, src1, 2);
dst.x = dot3(src0, row0);
dst.y = dot3(src0, row1);
dst.z = dot3(src0, row2);
}
- void VertexProgram::M3X4(Registers &r, Color4f &dst, Color4f &src0, Src &src1)
+ void VertexProgram::M3X4(Registers &r, Vector4f &dst, Vector4f &src0, Src &src1)
{
- Color4f row0 = reg(r, src1, 0);
- Color4f row1 = reg(r, src1, 1);
- Color4f row2 = reg(r, src1, 2);
- Color4f row3 = reg(r, src1, 3);
+ Vector4f row0 = reg(r, src1, 0);
+ Vector4f row1 = reg(r, src1, 1);
+ Vector4f row2 = reg(r, src1, 2);
+ Vector4f row3 = reg(r, src1, 3);
dst.x = dot3(src0, row0);
dst.y = dot3(src0, row1);
@@ -672,23 +902,23 @@
dst.w = dot3(src0, row3);
}
- void VertexProgram::M4X3(Registers &r, Color4f &dst, Color4f &src0, Src &src1)
+ void VertexProgram::M4X3(Registers &r, Vector4f &dst, Vector4f &src0, Src &src1)
{
- Color4f row0 = reg(r, src1, 0);
- Color4f row1 = reg(r, src1, 1);
- Color4f row2 = reg(r, src1, 2);
+ Vector4f row0 = reg(r, src1, 0);
+ Vector4f row1 = reg(r, src1, 1);
+ Vector4f row2 = reg(r, src1, 2);
dst.x = dot4(src0, row0);
dst.y = dot4(src0, row1);
dst.z = dot4(src0, row2);
}
- void VertexProgram::M4X4(Registers &r, Color4f &dst, Color4f &src0, Src &src1)
+ void VertexProgram::M4X4(Registers &r, Vector4f &dst, Vector4f &src0, Src &src1)
{
- Color4f row0 = reg(r, src1, 0);
- Color4f row1 = reg(r, src1, 1);
- Color4f row2 = reg(r, src1, 2);
- Color4f row3 = reg(r, src1, 3);
+ Vector4f row0 = reg(r, src1, 0);
+ Vector4f row1 = reg(r, src1, 1);
+ Vector4f row2 = reg(r, src1, 2);
+ Vector4f row3 = reg(r, src1, 3);
dst.x = dot4(src0, row0);
dst.y = dot4(src0, row1);
@@ -703,6 +933,7 @@
if(breakDepth == 0)
{
+ r.enableIndex = r.enableIndex - breakDepth;
Nucleus::createBr(endBlock);
}
else
@@ -710,49 +941,47 @@
r.enableBreak = r.enableBreak & ~r.enableStack[r.enableIndex];
Bool allBreak = SignMask(r.enableBreak) == 0x0;
+ r.enableIndex = r.enableIndex - breakDepth;
branch(allBreak, endBlock, deadBlock);
}
Nucleus::setInsertBlock(deadBlock);
+ r.enableIndex = r.enableIndex + breakDepth;
}
- void VertexProgram::BREAKC(Registers &r, Color4f &src0, Color4f &src1, Control control)
+ void VertexProgram::BREAKC(Registers &r, Vector4f &src0, Vector4f &src1, Control control)
{
Int4 condition;
switch(control)
{
- case Op::CONTROL_GT: condition = CmpNLE(src0.x, src1.x); break;
- case Op::CONTROL_EQ: condition = CmpEQ(src0.x, src1.x); break;
- case Op::CONTROL_GE: condition = CmpNLT(src0.x, src1.x); break;
- case Op::CONTROL_LT: condition = CmpLT(src0.x, src1.x); break;
- case Op::CONTROL_NE: condition = CmpNEQ(src0.x, src1.x); break;
- case Op::CONTROL_LE: condition = CmpLE(src0.x, src1.x); break;
+ case Shader::CONTROL_GT: condition = CmpNLE(src0.x, src1.x); break;
+ case Shader::CONTROL_EQ: condition = CmpEQ(src0.x, src1.x); break;
+ case Shader::CONTROL_GE: condition = CmpNLT(src0.x, src1.x); break;
+ case Shader::CONTROL_LT: condition = CmpLT(src0.x, src1.x); break;
+ case Shader::CONTROL_NE: condition = CmpNEQ(src0.x, src1.x); break;
+ case Shader::CONTROL_LE: condition = CmpLE(src0.x, src1.x); break;
default:
ASSERT(false);
}
- condition &= r.enableStack[r.enableIndex];
-
- llvm::BasicBlock *continueBlock = Nucleus::createBasicBlock();
- llvm::BasicBlock *endBlock = loopRepEndBlock[loopRepDepth - 1];
-
- r.enableBreak = r.enableBreak & ~condition;
- Bool allBreak = SignMask(r.enableBreak) == 0x0;
-
- branch(allBreak, endBlock, continueBlock);
- Nucleus::setInsertBlock(continueBlock);
+ BREAK(r, condition);
}
void VertexProgram::BREAKP(Registers &r, const Src &predicateRegister) // FIXME: Factor out parts common with BREAKC
{
Int4 condition = As<Int4>(r.p0[predicateRegister.swizzle & 0x3]);
- if(predicateRegister.modifier == Src::MODIFIER_NOT)
+ if(predicateRegister.modifier == Shader::MODIFIER_NOT)
{
condition = ~condition;
}
+ BREAK(r, condition);
+ }
+
+ void VertexProgram::BREAK(Registers &r, Int4 &condition)
+ {
condition &= r.enableStack[r.enableIndex];
llvm::BasicBlock *continueBlock = Nucleus::createBasicBlock();
@@ -761,44 +990,61 @@
r.enableBreak = r.enableBreak & ~condition;
Bool allBreak = SignMask(r.enableBreak) == 0x0;
+ r.enableIndex = r.enableIndex - breakDepth;
branch(allBreak, endBlock, continueBlock);
+
Nucleus::setInsertBlock(continueBlock);
+ r.enableIndex = r.enableIndex + breakDepth;
}
- void VertexProgram::CALL(Registers &r, int labelIndex)
+ void VertexProgram::CONTINUE(Registers &r)
+ {
+ r.enableContinue = r.enableContinue & ~r.enableStack[r.enableIndex];
+ }
+
+ void VertexProgram::TEST()
+ {
+ whileTest = true;
+ }
+
+ void VertexProgram::CALL(Registers &r, int labelIndex, int callSiteIndex)
{
if(!labelBlock[labelIndex])
{
labelBlock[labelIndex] = Nucleus::createBasicBlock();
}
- llvm::BasicBlock *retBlock = Nucleus::createBasicBlock();
- callRetBlock.push_back(retBlock);
+ if(callRetBlock[labelIndex].size() > 1)
+ {
+ r.callStack[r.stackIndex++] = UInt(callSiteIndex);
+ }
- r.callStack[r.stackIndex++] = UInt((unsigned int)callRetBlock.size() - 1); // FIXME
+ Int4 restoreLeave = r.enableLeave;
Nucleus::createBr(labelBlock[labelIndex]);
- Nucleus::setInsertBlock(retBlock);
+ Nucleus::setInsertBlock(callRetBlock[labelIndex][callSiteIndex]);
+
+ r.enableLeave = restoreLeave;
}
- void VertexProgram::CALLNZ(Registers &r, int labelIndex, const Src &src)
+ void VertexProgram::CALLNZ(Registers &r, int labelIndex, int callSiteIndex, const Src &src)
{
- if(src.type == Src::PARAMETER_CONSTBOOL)
+ if(src.type == Shader::PARAMETER_CONSTBOOL)
{
- CALLNZb(r, labelIndex, src);
+ CALLNZb(r, labelIndex, callSiteIndex, src);
}
- else if(src.type == Src::PARAMETER_PREDICATE)
+ else if(src.type == Shader::PARAMETER_PREDICATE)
{
- CALLNZp(r, labelIndex, src);
+ CALLNZp(r, labelIndex, callSiteIndex, src);
}
else ASSERT(false);
}
- void VertexProgram::CALLNZb(Registers &r, int labelIndex, const Src &boolRegister)
+ void VertexProgram::CALLNZb(Registers &r, int labelIndex, int callSiteIndex, const Src &boolRegister)
{
Bool condition = (*Pointer<Byte>(r.data + OFFSET(DrawData,vs.b[boolRegister.index])) != Byte(0)); // FIXME
- if(boolRegister.modifier == Src::MODIFIER_NOT)
+ if(boolRegister.modifier == Shader::MODIFIER_NOT)
{
condition = !condition;
}
@@ -808,20 +1054,24 @@
labelBlock[labelIndex] = Nucleus::createBasicBlock();
}
- llvm::BasicBlock *retBlock = Nucleus::createBasicBlock();
- callRetBlock.push_back(retBlock);
+ if(callRetBlock[labelIndex].size() > 1)
+ {
+ r.callStack[r.stackIndex++] = UInt(callSiteIndex);
+ }
- r.callStack[r.stackIndex++] = UInt((int)callRetBlock.size() - 1); // FIXME
+ Int4 restoreLeave = r.enableLeave;
- branch(condition, labelBlock[labelIndex], retBlock);
- Nucleus::setInsertBlock(retBlock);
+ branch(condition, labelBlock[labelIndex], callRetBlock[labelIndex][callSiteIndex]);
+ Nucleus::setInsertBlock(callRetBlock[labelIndex][callSiteIndex]);
+
+ r.enableLeave = restoreLeave;
}
- void VertexProgram::CALLNZp(Registers &r, int labelIndex, const Src &predicateRegister)
+ void VertexProgram::CALLNZp(Registers &r, int labelIndex, int callSiteIndex, const Src &predicateRegister)
{
Int4 condition = As<Int4>(r.p0[predicateRegister.swizzle & 0x3]);
- if(predicateRegister.modifier == Src::MODIFIER_NOT)
+ if(predicateRegister.modifier == Shader::MODIFIER_NOT)
{
condition = ~condition;
}
@@ -833,20 +1083,21 @@
labelBlock[labelIndex] = Nucleus::createBasicBlock();
}
- llvm::BasicBlock *retBlock = Nucleus::createBasicBlock();
- callRetBlock.push_back(retBlock);
-
- r.callStack[r.stackIndex++] = UInt((int)callRetBlock.size() - 1); // FIXME
+ if(callRetBlock[labelIndex].size() > 1)
+ {
+ r.callStack[r.stackIndex++] = UInt(callSiteIndex);
+ }
r.enableIndex++;
r.enableStack[r.enableIndex] = condition;
+ Int4 restoreLeave = r.enableLeave;
- Bool notAllFalse = SignMask(condition & r.enableBreak) != 0;
-
- branch(notAllFalse, labelBlock[labelIndex], retBlock);
- Nucleus::setInsertBlock(retBlock);
+ Bool notAllFalse = SignMask(condition) != 0;
+ branch(notAllFalse, labelBlock[labelIndex], callRetBlock[labelIndex][callSiteIndex]);
+ Nucleus::setInsertBlock(callRetBlock[labelIndex][callSiteIndex]);
r.enableIndex--;
+ r.enableLeave = restoreLeave;
}
void VertexProgram::ELSE(Registers &r)
@@ -859,7 +1110,7 @@
if(isConditionalIf[ifDepth])
{
Int4 condition = ~r.enableStack[r.enableIndex] & r.enableStack[r.enableIndex - 1];
- Bool notAllFalse = SignMask(condition & r.enableBreak) != 0;
+ Bool notAllFalse = SignMask(condition) != 0;
branch(notAllFalse, falseBlock, endBlock);
@@ -892,20 +1143,6 @@
}
}
- void VertexProgram::ENDREP(Registers &r)
- {
- loopRepDepth--;
-
- llvm::BasicBlock *testBlock = loopRepTestBlock[loopRepDepth];
- llvm::BasicBlock *endBlock = loopRepEndBlock[loopRepDepth];
-
- Nucleus::createBr(testBlock);
- Nucleus::setInsertBlock(endBlock);
-
- r.loopDepth--;
- r.enableBreak = Int4(0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF);
- }
-
void VertexProgram::ENDLOOP(Registers &r)
{
loopRepDepth--;
@@ -922,17 +1159,50 @@
r.enableBreak = Int4(0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF);
}
+ void VertexProgram::ENDREP(Registers &r)
+ {
+ loopRepDepth--;
+
+ llvm::BasicBlock *testBlock = loopRepTestBlock[loopRepDepth];
+ llvm::BasicBlock *endBlock = loopRepEndBlock[loopRepDepth];
+
+ Nucleus::createBr(testBlock);
+ Nucleus::setInsertBlock(endBlock);
+
+ r.loopDepth--;
+ r.enableBreak = Int4(0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF);
+ }
+
+ void VertexProgram::ENDWHILE(Registers &r)
+ {
+ loopRepDepth--;
+
+ llvm::BasicBlock *testBlock = loopRepTestBlock[loopRepDepth];
+ llvm::BasicBlock *endBlock = loopRepEndBlock[loopRepDepth];
+
+ Nucleus::createBr(testBlock);
+ Nucleus::setInsertBlock(endBlock);
+
+ r.enableIndex--;
+ r.enableBreak = Int4(0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF);
+ whileTest = false;
+ }
+
void VertexProgram::IF(Registers &r, const Src &src)
{
- if(src.type == Src::PARAMETER_CONSTBOOL)
+ if(src.type == Shader::PARAMETER_CONSTBOOL)
{
IFb(r, src);
}
- else if(src.type == Src::PARAMETER_PREDICATE)
+ else if(src.type == Shader::PARAMETER_PREDICATE)
{
IFp(r, src);
}
- else ASSERT(false);
+ else
+ {
+ Int4 condition = As<Int4>(reg(r, src).x);
+ IF(r, condition);
+ }
}
void VertexProgram::IFb(Registers &r, const Src &boolRegister)
@@ -941,9 +1211,9 @@
Bool condition = (*Pointer<Byte>(r.data + OFFSET(DrawData,vs.b[boolRegister.index])) != Byte(0)); // FIXME
- if(boolRegister.modifier == Src::MODIFIER_NOT)
+ if(boolRegister.modifier == Shader::MODIFIER_NOT)
{
- condition = !condition;
+ condition = !condition;
}
llvm::BasicBlock *trueBlock = Nucleus::createBasicBlock();
@@ -957,50 +1227,39 @@
ifDepth++;
}
- void VertexProgram::IFp(Registers &r, const Src &predicateRegister) // FIXME: Factor out parts common with IFC
+ void VertexProgram::IFp(Registers &r, const Src &predicateRegister)
{
Int4 condition = As<Int4>(r.p0[predicateRegister.swizzle & 0x3]);
- if(predicateRegister.modifier == Src::MODIFIER_NOT)
+ if(predicateRegister.modifier == Shader::MODIFIER_NOT)
{
condition = ~condition;
}
- condition &= r.enableStack[r.enableIndex];
-
- r.enableIndex++;
- r.enableStack[r.enableIndex] = condition;
-
- llvm::BasicBlock *trueBlock = Nucleus::createBasicBlock();
- llvm::BasicBlock *falseBlock = Nucleus::createBasicBlock();
-
- Bool notAllFalse = SignMask(condition & r.enableBreak) != 0;
-
- branch(notAllFalse, trueBlock, falseBlock);
-
- isConditionalIf[ifDepth] = true;
- ifFalseBlock[ifDepth] = falseBlock;
-
- ifDepth++;
- breakDepth++;
+ IF(r, condition);
}
- void VertexProgram::IFC(Registers &r, Color4f &src0, Color4f &src1, Control control)
+ void VertexProgram::IFC(Registers &r, Vector4f &src0, Vector4f &src1, Control control)
{
Int4 condition;
switch(control)
{
- case Op::CONTROL_GT: condition = CmpNLE(src0.x, src1.x); break;
- case Op::CONTROL_EQ: condition = CmpEQ(src0.x, src1.x); break;
- case Op::CONTROL_GE: condition = CmpNLT(src0.x, src1.x); break;
- case Op::CONTROL_LT: condition = CmpLT(src0.x, src1.x); break;
- case Op::CONTROL_NE: condition = CmpNEQ(src0.x, src1.x); break;
- case Op::CONTROL_LE: condition = CmpLE(src0.x, src1.x); break;
+ case Shader::CONTROL_GT: condition = CmpNLE(src0.x, src1.x); break;
+ case Shader::CONTROL_EQ: condition = CmpEQ(src0.x, src1.x); break;
+ case Shader::CONTROL_GE: condition = CmpNLT(src0.x, src1.x); break;
+ case Shader::CONTROL_LT: condition = CmpLT(src0.x, src1.x); break;
+ case Shader::CONTROL_NE: condition = CmpNEQ(src0.x, src1.x); break;
+ case Shader::CONTROL_LE: condition = CmpLE(src0.x, src1.x); break;
default:
ASSERT(false);
}
+ IF(r, condition);
+ }
+
+ void VertexProgram::IF(Registers &r, Int4 &condition)
+ {
condition &= r.enableStack[r.enableIndex];
r.enableIndex++;
@@ -1009,7 +1268,7 @@
llvm::BasicBlock *trueBlock = Nucleus::createBasicBlock();
llvm::BasicBlock *falseBlock = Nucleus::createBasicBlock();
- Bool notAllFalse = SignMask(condition & r.enableBreak) != 0;
+ Bool notAllFalse = SignMask(condition) != 0;
branch(notAllFalse, trueBlock, falseBlock);
@@ -1022,7 +1281,13 @@
void VertexProgram::LABEL(int labelIndex)
{
+ if(!labelBlock[labelIndex])
+ {
+ labelBlock[labelIndex] = Nucleus::createBasicBlock();
+ }
+
Nucleus::setInsertBlock(labelBlock[labelIndex]);
+ currentLabel = labelIndex;
}
void VertexProgram::LOOP(Registers &r, const Src &integerRegister)
@@ -1086,27 +1351,73 @@
breakDepth = 0;
}
+ void VertexProgram::WHILE(Registers &r, const Src &temporaryRegister)
+ {
+ r.enableIndex++;
+
+ llvm::BasicBlock *loopBlock = Nucleus::createBasicBlock();
+ llvm::BasicBlock *testBlock = Nucleus::createBasicBlock();
+ llvm::BasicBlock *endBlock = Nucleus::createBasicBlock();
+
+ loopRepTestBlock[loopRepDepth] = testBlock;
+ loopRepEndBlock[loopRepDepth] = endBlock;
+
+ Int4 restoreBreak = r.enableBreak;
+ Int4 restoreContinue = r.enableContinue;
+
+ // FIXME: jump(testBlock)
+ Nucleus::createBr(testBlock);
+ Nucleus::setInsertBlock(testBlock);
+ r.enableContinue = restoreContinue;
+
+ Vector4f &src = reg(r, temporaryRegister);
+ Int4 condition = As<Int4>(src.x);
+ condition &= r.enableStack[r.enableIndex - 1];
+ r.enableStack[r.enableIndex] = condition;
+
+ Bool notAllFalse = SignMask(condition) != 0;
+ branch(notAllFalse, loopBlock, endBlock);
+
+ Nucleus::setInsertBlock(endBlock);
+ r.enableBreak = restoreBreak;
+
+ Nucleus::setInsertBlock(loopBlock);
+
+ loopRepDepth++;
+ breakDepth = 0;
+ }
+
void VertexProgram::RET(Registers &r)
{
- if(!returns)
+ if(currentLabel == -1)
{
returnBlock = Nucleus::createBasicBlock();
Nucleus::createBr(returnBlock);
-
- returns = true;
}
else
{
- // FIXME: Encapsulate
- UInt index = r.callStack[--r.stackIndex];
-
llvm::BasicBlock *unreachableBlock = Nucleus::createBasicBlock();
- llvm::Value *value = Nucleus::createLoad(index.address);
- llvm::Value *switchInst = Nucleus::createSwitch(value, unreachableBlock, (int)callRetBlock.size());
- for(unsigned int i = 0; i < callRetBlock.size(); i++)
+ if(callRetBlock[currentLabel].size() > 1) // Pop the return destination from the call stack
{
- Nucleus::addSwitchCase(switchInst, i, callRetBlock[i]);
+ // FIXME: Encapsulate
+ UInt index = r.callStack[--r.stackIndex];
+
+ llvm::Value *value = Nucleus::createLoad(index.address);
+ llvm::Value *switchInst = Nucleus::createSwitch(value, unreachableBlock, (int)callRetBlock[currentLabel].size());
+
+ for(unsigned int i = 0; i < callRetBlock[currentLabel].size(); i++)
+ {
+ Nucleus::addSwitchCase(switchInst, i, callRetBlock[currentLabel][i]);
+ }
+ }
+ else if(callRetBlock[currentLabel].size() == 1) // Jump directly to the unique return destination
+ {
+ Nucleus::createBr(callRetBlock[currentLabel][0]);
+ }
+ else // Function isn't called
+ {
+ Nucleus::createBr(unreachableBlock);
}
Nucleus::setInsertBlock(unreachableBlock);
@@ -1114,17 +1425,60 @@
}
}
- void VertexProgram::TEXLDL(Registers &r, Color4f &dst, Color4f &src0, const Src &src1)
+ void VertexProgram::LEAVE(Registers &r)
{
- Pointer<Byte> texture = r.data + OFFSET(DrawData,mipmap[16]) + src1.index * sizeof(Texture);
+ r.enableLeave = r.enableLeave & ~r.enableStack[r.enableIndex];
- Color4f tmp;
+ // FIXME: Return from function if all instances left
+ // FIXME: Use enableLeave in other control-flow constructs
+ }
- sampler[src1.index]->sampleTexture(texture, tmp, src0.x, src0.y, src0.z, src0.w, src0, src0, false, false, true);
+ void VertexProgram::TEXLDL(Registers &r, Vector4f &dst, Vector4f &src0, const Src &src1)
+ {
+ Vector4f tmp;
+ sampleTexture(r, tmp, src1, src0.x, src0.y, src0.z, src0.w);
dst.x = tmp[(src1.swizzle >> 0) & 0x3];
dst.y = tmp[(src1.swizzle >> 2) & 0x3];
dst.z = tmp[(src1.swizzle >> 4) & 0x3];
dst.w = tmp[(src1.swizzle >> 6) & 0x3];
}
+
+ void VertexProgram::TEX(Registers &r, Vector4f &dst, Vector4f &src0, const Src &src1)
+ {
+ Float4 lod = Float4(0.0f);
+ Vector4f tmp;
+ sampleTexture(r, tmp, src1, src0.x, src0.y, src0.z, lod);
+
+ dst.x = tmp[(src1.swizzle >> 0) & 0x3];
+ dst.y = tmp[(src1.swizzle >> 2) & 0x3];
+ dst.z = tmp[(src1.swizzle >> 4) & 0x3];
+ dst.w = tmp[(src1.swizzle >> 6) & 0x3];
+ }
+
+ void VertexProgram::sampleTexture(Registers &r, Vector4f &c, const Src &s, Float4 &u, Float4 &v, Float4 &w, Float4 &q)
+ {
+ if(s.type == Shader::PARAMETER_SAMPLER && s.rel.type == Shader::PARAMETER_VOID)
+ {
+ Pointer<Byte> texture = r.data + OFFSET(DrawData,mipmap[16]) + s.index * sizeof(Texture);
+ sampler[s.index]->sampleTexture(texture, c, u, v, w, q, r.a0, r.a0, false, false, true);
+ }
+ else
+ {
+ Int index = As<Int>(Float(reg(r, s).x.x));
+
+ for(int i = 0; i < 16; i++)
+ {
+ if(shader->usesSampler(i))
+ {
+ If(index == i)
+ {
+ Pointer<Byte> texture = r.data + OFFSET(DrawData,mipmap[16]) + i * sizeof(Texture);
+ sampler[i]->sampleTexture(texture, c, u, v, w, q, r.a0, r.a0, false, false, true);
+ // FIXME: When the sampler states are the same, we could use one sampler and just index the texture
+ }
+ }
+ }
+ }
+ }
}
