ShuriZma
/
xenia-canary
mirror of https://github.com/xenia-canary/xenia-canary.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

[D3D12] DXBC: Fix vector write mask usage

This commit is contained in:
Triang3l 2018-09-07 20:27:08 +03:00
parent 8ecedbb20a
commit 04544945d0
1 changed files with 12 additions and 15 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

27
src/xenia/gpu/dxbc_shader_translator.cc
View File

@ -187,7 +187,7 @@ void DxbcShaderTranslator::StartVertexShader_LoadVertexIndex() {
++stat_.uint_instruction_count; ++stat_.uint_instruction_count;
// Get bits indicating what swaps should be done. // Get bits indicating what swaps should be done.
// ubfe reg.zw, l(1, 1), l(0, 1), xe_vertex_index_endian.xx // ubfe reg.zw, l(0, 0, 1, 1).zw, l(0, 0, 0, 1).zw, xe_vertex_index_endian.xx
// ABCD | BADC | 8in16/16in32? | 8in32/16in32? // ABCD | BADC | 8in16/16in32? | 8in32/16in32?
shader_code_.push_back(ENCODE_D3D10_SB_OPCODE_TYPE(D3D11_SB_OPCODE_UBFE) | shader_code_.push_back(ENCODE_D3D10_SB_OPCODE_TYPE(D3D11_SB_OPCODE_UBFE) |
ENCODE_D3D10_SB_TOKENIZED_INSTRUCTION_LENGTH(17)); ENCODE_D3D10_SB_TOKENIZED_INSTRUCTION_LENGTH(17));
@ -196,16 +196,16 @@ void DxbcShaderTranslator::StartVertexShader_LoadVertexIndex() {
shader_code_.push_back(reg); shader_code_.push_back(reg);
shader_code_.push_back(EncodeVectorSwizzledOperand( shader_code_.push_back(EncodeVectorSwizzledOperand(
D3D10_SB_OPERAND_TYPE_IMMEDIATE32, kSwizzleXYZW, 0)); D3D10_SB_OPERAND_TYPE_IMMEDIATE32, kSwizzleXYZW, 0));
shader_code_.push_back(1);
shader_code_.push_back(1);
shader_code_.push_back(0); shader_code_.push_back(0);
shader_code_.push_back(0); shader_code_.push_back(0);
shader_code_.push_back(1);
shader_code_.push_back(1);
shader_code_.push_back(EncodeVectorSwizzledOperand( shader_code_.push_back(EncodeVectorSwizzledOperand(
D3D10_SB_OPERAND_TYPE_IMMEDIATE32, kSwizzleXYZW, 0)); D3D10_SB_OPERAND_TYPE_IMMEDIATE32, kSwizzleXYZW, 0));
shader_code_.push_back(0); shader_code_.push_back(0);
shader_code_.push_back(0);
shader_code_.push_back(0);
shader_code_.push_back(1); shader_code_.push_back(1);
shader_code_.push_back(0);
shader_code_.push_back(0);
rdef_constants_used_ |= 1ull rdef_constants_used_ |= 1ull
<< uint32_t(RdefConstantIndex::kSysVertexIndexEndian); << uint32_t(RdefConstantIndex::kSysVertexIndexEndian);
shader_code_.push_back(EncodeVectorReplicatedOperand( shader_code_.push_back(EncodeVectorReplicatedOperand(
@ -1363,13 +1363,13 @@ void DxbcShaderTranslator::StoreResult(const InstructionResult& result,
} }
// Get the write masks and data required for loading of both the swizzled part // Get the write masks and data required for loading of both the swizzled part
// and the constant (zero/one) part. // and the constant (zero/one) part. The write mask is treated also as a read
// mask in DXBC, and `mov r0.zw, r1.xyzw` actually means r0.zw = r1.zw, not
// r0.zw = r1.xy.
uint32_t swizzle_mask = 0; uint32_t swizzle_mask = 0;
uint32_t swizzle_components = 0; uint32_t swizzle_components = 0;
uint32_t swizzle_shift = 0;
uint32_t constant_mask = 0; uint32_t constant_mask = 0;
uint32_t constant_values = 0; uint32_t constant_values = 0;
uint32_t constant_shift = 0;
for (uint32_t i = 0; i < 4; ++i) { for (uint32_t i = 0; i < 4; ++i) {
if (!result.write_mask[i]) { if (!result.write_mask[i]) {
continue; continue;
@ -1379,14 +1379,11 @@ void DxbcShaderTranslator::StoreResult(const InstructionResult& result,
swizzle_mask |= 1 << i; swizzle_mask |= 1 << i;
// If replicating X, just keep zero swizzle (XXXX). // If replicating X, just keep zero swizzle (XXXX).
if (!replicate_x) { if (!replicate_x) {
swizzle_components |= uint32_t(component) << swizzle_shift; swizzle_components |= uint32_t(component) << (i * 2);
swizzle_shift += 2;
} }
} else { } else {
constant_mask |= 1 << i; constant_mask |= 1 << i;
constant_values |= (component == SwizzleSource::k1 ? 1 : 0) constant_values |= (component == SwizzleSource::k1 ? 1 : 0) << i;
<< constant_shift;
++constant_shift;
} }
} }
@ -2441,7 +2438,7 @@ void DxbcShaderTranslator::ProcessVectorAluInstruction(
shader_code_.push_back(system_temp_pv_); shader_code_.push_back(system_temp_pv_);
if (instr.vector_opcode != AluVectorOpcode::kSetpGtPush) { if (instr.vector_opcode != AluVectorOpcode::kSetpGtPush) {
// lt in DXBC, not gt. // lt in DXBC, not gt.
UseDxbcSourceOperand(dxbc_operands[1], 0b11001100); UseDxbcSourceOperand(dxbc_operands[1], 0b11000000);
} }
shader_code_.push_back(EncodeVectorSwizzledOperand( shader_code_.push_back(EncodeVectorSwizzledOperand(
D3D10_SB_OPERAND_TYPE_IMMEDIATE32, kSwizzleXYZW, 0)); D3D10_SB_OPERAND_TYPE_IMMEDIATE32, kSwizzleXYZW, 0));
@ -2450,7 +2447,7 @@ void DxbcShaderTranslator::ProcessVectorAluInstruction(
shader_code_.push_back(0); shader_code_.push_back(0);
shader_code_.push_back(0); shader_code_.push_back(0);
if (instr.vector_opcode == AluVectorOpcode::kSetpGtPush) { if (instr.vector_opcode == AluVectorOpcode::kSetpGtPush) {
UseDxbcSourceOperand(dxbc_operands[1], 0b11001100); UseDxbcSourceOperand(dxbc_operands[1], 0b11000000);
} }
++stat_.instruction_count; ++stat_.instruction_count;
++stat_.float_instruction_count; ++stat_.float_instruction_count;