Merge pull request #1057 from aroulin/shader-dph-dphi

Shader: Implement DPH and DPHI in interpreter/JIT
This commit is contained in:
bunnei 2015-08-23 17:47:09 -04:00
commit 387bd3a1e4
3 changed files with 44 additions and 3 deletions

View File

@ -197,12 +197,19 @@ void RunInterpreter(UnitState<Debug>& state) {
case OpCode::Id::DP3: case OpCode::Id::DP3:
case OpCode::Id::DP4: case OpCode::Id::DP4:
case OpCode::Id::DPH:
case OpCode::Id::DPHI:
{ {
Record<DebugDataRecord::SRC1>(state.debug, iteration, src1); Record<DebugDataRecord::SRC1>(state.debug, iteration, src1);
Record<DebugDataRecord::SRC2>(state.debug, iteration, src2); Record<DebugDataRecord::SRC2>(state.debug, iteration, src2);
Record<DebugDataRecord::DEST_IN>(state.debug, iteration, dest); Record<DebugDataRecord::DEST_IN>(state.debug, iteration, dest);
OpCode::Id opcode = instr.opcode.Value().EffectiveOpCode();
if (opcode == OpCode::Id::DPH || opcode == OpCode::Id::DPHI)
src1[3] = float24::FromFloat32(1.0f);
float24 dot = float24::FromFloat32(0.f); float24 dot = float24::FromFloat32(0.f);
int num_components = (instr.opcode.Value() == OpCode::Id::DP3) ? 3 : 4; int num_components = (opcode == OpCode::Id::DP3) ? 3 : 4;
for (int i = 0; i < num_components; ++i) for (int i = 0; i < num_components; ++i)
dot = dot + src1[i] * src2[i]; dot = dot + src1[i] * src2[i];

View File

@ -23,7 +23,7 @@ const JitFunction instr_table[64] = {
&JitCompiler::Compile_ADD, // add &JitCompiler::Compile_ADD, // add
&JitCompiler::Compile_DP3, // dp3 &JitCompiler::Compile_DP3, // dp3
&JitCompiler::Compile_DP4, // dp4 &JitCompiler::Compile_DP4, // dp4
nullptr, // dph &JitCompiler::Compile_DPH, // dph
nullptr, // unknown nullptr, // unknown
&JitCompiler::Compile_EX2, // ex2 &JitCompiler::Compile_EX2, // ex2
&JitCompiler::Compile_LG2, // lg2 &JitCompiler::Compile_LG2, // lg2
@ -44,7 +44,7 @@ const JitFunction instr_table[64] = {
nullptr, // unknown nullptr, // unknown
nullptr, // unknown nullptr, // unknown
nullptr, // unknown nullptr, // unknown
nullptr, // dphi &JitCompiler::Compile_DPH, // dphi
nullptr, // unknown nullptr, // unknown
&JitCompiler::Compile_SGE, // sgei &JitCompiler::Compile_SGE, // sgei
&JitCompiler::Compile_SLT, // slti &JitCompiler::Compile_SLT, // slti
@ -347,6 +347,39 @@ void JitCompiler::Compile_DP4(Instruction instr) {
Compile_DestEnable(instr, SRC1); Compile_DestEnable(instr, SRC1);
} }
void JitCompiler::Compile_DPH(Instruction instr) {
if (instr.opcode.Value().EffectiveOpCode() == OpCode::Id::DPHI) {
Compile_SwizzleSrc(instr, 1, instr.common.src1i, SRC1);
Compile_SwizzleSrc(instr, 2, instr.common.src2i, SRC2);
} else {
Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1);
Compile_SwizzleSrc(instr, 2, instr.common.src2, SRC2);
}
if (Common::GetCPUCaps().sse4_1) {
// Set 4th component to 1.0
BLENDPS(SRC1, R(ONE), 0x8); // 0b1000
DPPS(SRC1, R(SRC2), 0xff);
} else {
// Reverse to set the 4th component to 1.0
SHUFPS(SRC1, R(SRC1), _MM_SHUFFLE(0, 1, 2, 3));
MOVSS(SRC1, R(ONE));
SHUFPS(SRC1, R(SRC1), _MM_SHUFFLE(0, 1, 2, 3));
MULPS(SRC1, R(SRC2));
MOVAPS(SRC2, R(SRC1));
SHUFPS(SRC1, R(SRC1), _MM_SHUFFLE(2, 3, 0, 1)); // XYZW -> ZWXY
ADDPS(SRC1, R(SRC2));
MOVAPS(SRC2, R(SRC1));
SHUFPS(SRC1, R(SRC1), _MM_SHUFFLE(0, 1, 2, 3)); // XYZW -> WZYX
ADDPS(SRC1, R(SRC2));
}
Compile_DestEnable(instr, SRC1);
}
void JitCompiler::Compile_EX2(Instruction instr) { void JitCompiler::Compile_EX2(Instruction instr) {
Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1); Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1);
MOVSS(XMM0, R(SRC1)); MOVSS(XMM0, R(SRC1));

View File

@ -37,6 +37,7 @@ public:
void Compile_ADD(Instruction instr); void Compile_ADD(Instruction instr);
void Compile_DP3(Instruction instr); void Compile_DP3(Instruction instr);
void Compile_DP4(Instruction instr); void Compile_DP4(Instruction instr);
void Compile_DPH(Instruction instr);
void Compile_EX2(Instruction instr); void Compile_EX2(Instruction instr);
void Compile_LG2(Instruction instr); void Compile_LG2(Instruction instr);
void Compile_MUL(Instruction instr); void Compile_MUL(Instruction instr);