From 62c36a4ef0a37fe83bb8f8680f928970bead545b Mon Sep 17 00:00:00 2001 From: Tony Wasserka Date: Thu, 14 Aug 2014 23:28:55 +0200 Subject: Pica/VertexShader: Fix a bug in the bitfield definitions and add the "negate" field for swizzlers. --- src/video_core/vertex_shader.cpp | 28 +++++++++------ src/video_core/vertex_shader.h | 78 ++++++++++++++++++++++++++++++++++++++-- 2 files changed, 92 insertions(+), 14 deletions(-) diff --git a/src/video_core/vertex_shader.cpp b/src/video_core/vertex_shader.cpp index 8df14b51..cdecbff3 100644 --- a/src/video_core/vertex_shader.cpp +++ b/src/video_core/vertex_shader.cpp @@ -65,26 +65,32 @@ static void ProcessShaderCode(VertexShaderState& state) { const Instruction& instr = *(const Instruction*)state.program_counter; state.debug.max_offset = std::max(state.debug.max_offset, 1 + (state.program_counter - shader_memory)); - const float24* src1_ = (instr.common.src1 < 0x10) ? state.input_register_table[instr.common.src1] - : (instr.common.src1 < 0x20) ? &state.temporary_registers[instr.common.src1-0x10].x - : (instr.common.src1 < 0x80) ? &shader_uniforms.f[instr.common.src1-0x20].x - : nullptr; - const float24* src2_ = (instr.common.src2 < 0x10) ? state.input_register_table[instr.common.src2] - : &state.temporary_registers[instr.common.src2-0x10].x; - // TODO: Unsure about the limit values - float24* dest = (instr.common.dest <= 0x1C) ? state.output_register_table[instr.common.dest] - : (instr.common.dest <= 0x3C) ? nullptr - : (instr.common.dest <= 0x7C) ? &state.temporary_registers[(instr.common.dest-0x40)/4][instr.common.dest%4] + const float24* src1_ = (instr.common.src1 < 0x10) ? state.input_register_table[instr.common.src1.GetIndex()] + : (instr.common.src1 < 0x20) ? &state.temporary_registers[instr.common.src1.GetIndex()].x + : (instr.common.src1 < 0x80) ? &shader_uniforms.f[instr.common.src1.GetIndex()].x : nullptr; + const float24* src2_ = (instr.common.src2 < 0x10) ? state.input_register_table[instr.common.src2.GetIndex()] + : &state.temporary_registers[instr.common.src2.GetIndex()].x; + float24* dest = (instr.common.dest < 0x08) ? state.output_register_table[4*instr.common.dest.GetIndex()] + : (instr.common.dest < 0x10) ? nullptr + : (instr.common.dest < 0x20) ? &state.temporary_registers[instr.common.dest.GetIndex()][0] + : nullptr; const SwizzlePattern& swizzle = *(SwizzlePattern*)&swizzle_data[instr.common.operand_desc_id]; + const bool negate_src1 = swizzle.negate; - const float24 src1[4] = { + float24 src1[4] = { src1_[(int)swizzle.GetSelectorSrc1(0)], src1_[(int)swizzle.GetSelectorSrc1(1)], src1_[(int)swizzle.GetSelectorSrc1(2)], src1_[(int)swizzle.GetSelectorSrc1(3)], }; + if (negate_src1) { + src1[0] = src1[0] * float24::FromFloat32(-1); + src1[1] = src1[1] * float24::FromFloat32(-1); + src1[2] = src1[2] * float24::FromFloat32(-1); + src1[3] = src1[3] * float24::FromFloat32(-1); + } const float24 src2[4] = { src2_[(int)swizzle.GetSelectorSrc2(0)], src2_[(int)swizzle.GetSelectorSrc2(1)], diff --git a/src/video_core/vertex_shader.h b/src/video_core/vertex_shader.h index 1b71e367..f0a8a5b6 100644 --- a/src/video_core/vertex_shader.h +++ b/src/video_core/vertex_shader.h @@ -117,9 +117,78 @@ union Instruction { // while "dest" addresses individual floats. union { BitField<0x00, 0x5, u32> operand_desc_id; - BitField<0x07, 0x5, u32> src2; - BitField<0x0c, 0x7, u32> src1; - BitField<0x13, 0x7, u32> dest; + + template + struct SourceRegister : BitFieldType { + enum RegisterType { + Input, + Temporary, + FloatUniform + }; + + RegisterType GetRegisterType() const { + if (BitFieldType::Value() < 0x10) + return Input; + else if (BitFieldType::Value() < 0x20) + return Temporary; + else + return FloatUniform; + } + + int GetIndex() const { + if (GetRegisterType() == Input) + return BitFieldType::Value(); + else if (GetRegisterType() == Temporary) + return BitFieldType::Value() - 0x10; + else if (GetRegisterType() == FloatUniform) + return BitFieldType::Value() - 0x20; + } + + std::string GetRegisterName() const { + std::map type = { + { Input, "i" }, + { Temporary, "t" }, + { FloatUniform, "f" }, + }; + return type[GetRegisterType()] + std::to_string(GetIndex()); + } + }; + + SourceRegister> src2; + SourceRegister> src1; + + struct : BitField<0x15, 0x5, u32> + { + enum RegisterType { + Output, + Temporary, + Unknown + }; + RegisterType GetRegisterType() const { + if (Value() < 0x8) + return Output; + else if (Value() < 0x10) + return Unknown; + else + return Temporary; + } + int GetIndex() const { + if (GetRegisterType() == Output) + return Value(); + else if (GetRegisterType() == Temporary) + return Value() - 0x10; + else + return Value(); + } + std::string GetRegisterName() const { + std::map type = { + { Output, "o" }, + { Temporary, "t" }, + { Unknown, "u" } + }; + return type[GetRegisterType()] + std::to_string(GetIndex()); + } + } dest; } common; // Format used for flow control instructions ("if") @@ -128,6 +197,7 @@ union Instruction { BitField<0x0a, 0xc, u32> offset_words; } flow_control; }; +static_assert(std::is_standard_layout::value, "Structure is not using standard layout!"); union SwizzlePattern { u32 hex; @@ -185,6 +255,8 @@ union SwizzlePattern { // Components of "dest" that should be written to: LSB=dest.w, MSB=dest.x BitField< 0, 4, u32> dest_mask; + BitField< 4, 1, u32> negate; // negates src1 + BitField< 5, 2, Selector> src1_selector_3; BitField< 7, 2, Selector> src1_selector_2; BitField< 9, 2, Selector> src1_selector_1; -- cgit v1.2.3