From 477d616f7df3b609afcb67d69d9570098cc00029 Mon Sep 17 00:00:00 2001
From: ReinUsesLisp <reinuseslisp@airmail.cc>
Date: Mon, 28 Jan 2019 18:11:23 -0300
Subject: [PATCH] shader_ir: Unify constant buffer offset values
Constant buffer values on the shader IR were using different offsets if
the access direct or indirect. cbuf34 has a non-multiplied offset while
cbuf36 does. On shader decoding this commit multiplies it by four on
cbuf34 queries.
---
src/video_core/engines/shader_bytecode.h | 8 ++++++++
src/video_core/renderer_opengl/gl_rasterizer.cpp | 2 +-
src/video_core/renderer_opengl/gl_shader_decompiler.cpp | 5 +++--
src/video_core/shader/decode/arithmetic.cpp | 2 +-
src/video_core/shader/decode/arithmetic_half.cpp | 2 +-
src/video_core/shader/decode/arithmetic_integer.cpp | 2 +-
src/video_core/shader/decode/conversion.cpp | 6 +++---
src/video_core/shader/decode/ffma.cpp | 4 ++--
src/video_core/shader/decode/float_set.cpp | 2 +-
src/video_core/shader/decode/float_set_predicate.cpp | 2 +-
src/video_core/shader/decode/hfma2.cpp | 7 ++++---
src/video_core/shader/decode/integer_set.cpp | 2 +-
src/video_core/shader/decode/integer_set_predicate.cpp | 2 +-
src/video_core/shader/decode/memory.cpp | 6 +++---
src/video_core/shader/decode/shift.cpp | 2 +-
src/video_core/shader/decode/xmad.cpp | 5 +++--
src/video_core/shader/shader_ir.h | 2 +-
17 files changed, 36 insertions(+), 25 deletions(-)
diff --git a/src/video_core/engines/shader_bytecode.h b/src/video_core/engines/shader_bytecode.h
index 9989825f8..713b01c9f 100644
--- a/src/video_core/engines/shader_bytecode.h
+++ b/src/video_core/engines/shader_bytecode.h
@@ -1248,11 +1248,19 @@ union Instruction {
union {
BitField<20, 14, u64> offset;
BitField<34, 5, u64> index;
+
+ u64 GetOffset() const {
+ return offset * 4;
+ }
} cbuf34;
union {
BitField<20, 16, s64> offset;
BitField<36, 5, u64> index;
+
+ s64 GetOffset() const {
+ return offset;
+ }
} cbuf36;
// Unsure about the size of this one.
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp
index ee313cb2f..aed6843d4 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp
@@ -957,7 +957,7 @@ void RasterizerOpenGL::SetupConstBuffers(Tegra::Engines::Maxwell3D::Regs::Shader
}
} else {
// Buffer is accessed directly, upload just what we use
- size = used_buffer.GetSize() * sizeof(float);
+ size = used_buffer.GetSize();
}
// Align the actual size so it ends up being a multiple of vec4 to meet the OpenGL std140
diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
index 004245431..36035d0d2 100644
--- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
@@ -543,8 +543,9 @@ private:
if (const auto immediate = std::get_if<ImmediateNode>(offset)) {
// Direct access
const u32 offset_imm = immediate->GetValue();
- return fmt::format("{}[{}][{}]", GetConstBuffer(cbuf->GetIndex()), offset_imm / 4,
- offset_imm % 4);
+ ASSERT_MSG(offset_imm % 4 == 0, "Unaligned cbuf direct access");
+ return fmt::format("{}[{}][{}]", GetConstBuffer(cbuf->GetIndex()),
+ offset_imm / (4 * 4), (offset_imm / 4) % 4);
} else if (std::holds_alternative<OperationNode>(*offset)) {
// Indirect access
diff --git a/src/video_core/shader/decode/arithmetic.cpp b/src/video_core/shader/decode/arithmetic.cpp
index e7847f614..51b8d55d4 100644
--- a/src/video_core/shader/decode/arithmetic.cpp
+++ b/src/video_core/shader/decode/arithmetic.cpp
@@ -25,7 +25,7 @@ u32 ShaderIR::DecodeArithmetic(BasicBlock& bb, const BasicBlock& code, u32 pc) {
} else if (instr.is_b_gpr) {
return GetRegister(instr.gpr20);
} else {
- return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset);
+ return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.GetOffset());
}
}();
diff --git a/src/video_core/shader/decode/arithmetic_half.cpp b/src/video_core/shader/decode/arithmetic_half.cpp
index a237dcb92..37eef2bf2 100644
--- a/src/video_core/shader/decode/arithmetic_half.cpp
+++ b/src/video_core/shader/decode/arithmetic_half.cpp
@@ -35,7 +35,7 @@ u32 ShaderIR::DecodeArithmeticHalf(BasicBlock& bb, const BasicBlock& code, u32 p
switch (opcode->get().GetId()) {
case OpCode::Id::HADD2_C:
case OpCode::Id::HMUL2_C:
- return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset);
+ return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.GetOffset());
case OpCode::Id::HADD2_R:
case OpCode::Id::HMUL2_R:
return GetRegister(instr.gpr20);
diff --git a/src/video_core/shader/decode/arithmetic_integer.cpp b/src/video_core/shader/decode/arithmetic_integer.cpp
index 4a8cc1a1c..cc9a76a19 100644
--- a/src/video_core/shader/decode/arithmetic_integer.cpp
+++ b/src/video_core/shader/decode/arithmetic_integer.cpp
@@ -26,7 +26,7 @@ u32 ShaderIR::DecodeArithmeticInteger(BasicBlock& bb, const BasicBlock& code, u3
} else if (instr.is_b_gpr) {
return GetRegister(instr.gpr20);
} else {
- return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset);
+ return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.GetOffset());
}
}();
diff --git a/src/video_core/shader/decode/conversion.cpp b/src/video_core/shader/decode/conversion.cpp
index ee18d3a99..728a393a1 100644
--- a/src/video_core/shader/decode/conversion.cpp
+++ b/src/video_core/shader/decode/conversion.cpp
@@ -48,7 +48,7 @@ u32 ShaderIR::DecodeConversion(BasicBlock& bb, const BasicBlock& code, u32 pc) {
if (instr.is_b_gpr) {
return GetRegister(instr.gpr20);
} else {
- return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset);
+ return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.GetOffset());
}
}();
const bool input_signed = instr.conversion.is_input_signed;
@@ -72,7 +72,7 @@ u32 ShaderIR::DecodeConversion(BasicBlock& bb, const BasicBlock& code, u32 pc) {
if (instr.is_b_gpr) {
return GetRegister(instr.gpr20);
} else {
- return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset);
+ return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.GetOffset());
}
}();
@@ -110,7 +110,7 @@ u32 ShaderIR::DecodeConversion(BasicBlock& bb, const BasicBlock& code, u32 pc) {
if (instr.is_b_gpr) {
return GetRegister(instr.gpr20);
} else {
- return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset);
+ return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.GetOffset());
}
}();
diff --git a/src/video_core/shader/decode/ffma.cpp b/src/video_core/shader/decode/ffma.cpp
index be8dc2230..52f39d3ff 100644
--- a/src/video_core/shader/decode/ffma.cpp
+++ b/src/video_core/shader/decode/ffma.cpp
@@ -27,14 +27,14 @@ u32 ShaderIR::DecodeFfma(BasicBlock& bb, const BasicBlock& code, u32 pc) {
auto [op_b, op_c] = [&]() -> std::tuple<Node, Node> {
switch (opcode->get().GetId()) {
case OpCode::Id::FFMA_CR: {
- return {GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset),
+ return {GetConstBuffer(instr.cbuf34.index, instr.cbuf34.GetOffset()),
GetRegister(instr.gpr39)};
}
case OpCode::Id::FFMA_RR:
return {GetRegister(instr.gpr20), GetRegister(instr.gpr39)};
case OpCode::Id::FFMA_RC: {
return {GetRegister(instr.gpr39),
- GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset)};
+ GetConstBuffer(instr.cbuf34.index, instr.cbuf34.GetOffset())};
}
case OpCode::Id::FFMA_IMM:
return {GetImmediate19(instr), GetRegister(instr.gpr39)};
diff --git a/src/video_core/shader/decode/float_set.cpp b/src/video_core/shader/decode/float_set.cpp
index ba846f1bd..9f9da2278 100644
--- a/src/video_core/shader/decode/float_set.cpp
+++ b/src/video_core/shader/decode/float_set.cpp
@@ -25,7 +25,7 @@ u32 ShaderIR::DecodeFloatSet(BasicBlock& bb, const BasicBlock& code, u32 pc) {
} else if (instr.is_b_gpr) {
return GetRegister(instr.gpr20);
} else {
- return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset);
+ return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.GetOffset());
}
}();
diff --git a/src/video_core/shader/decode/float_set_predicate.cpp b/src/video_core/shader/decode/float_set_predicate.cpp
index e88b04d18..dd3aef6f2 100644
--- a/src/video_core/shader/decode/float_set_predicate.cpp
+++ b/src/video_core/shader/decode/float_set_predicate.cpp
@@ -25,7 +25,7 @@ u32 ShaderIR::DecodeFloatSetPredicate(BasicBlock& bb, const BasicBlock& code, u3
} else if (instr.is_b_gpr) {
return GetRegister(instr.gpr20);
} else {
- return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset);
+ return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.GetOffset());
}
}();
op_b = GetOperandAbsNegFloat(op_b, instr.fsetp.abs_b, false);
diff --git a/src/video_core/shader/decode/hfma2.cpp b/src/video_core/shader/decode/hfma2.cpp
index 4a6b945f9..43a0a9e10 100644
--- a/src/video_core/shader/decode/hfma2.cpp
+++ b/src/video_core/shader/decode/hfma2.cpp
@@ -39,13 +39,14 @@ u32 ShaderIR::DecodeHfma2(BasicBlock& bb, const BasicBlock& code, u32 pc) {
neg_b = instr.hfma2.negate_b;
neg_c = instr.hfma2.negate_c;
return {instr.hfma2.saturate, instr.hfma2.type_b,
- GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset), instr.hfma2.type_reg39,
- GetRegister(instr.gpr39)};
+ GetConstBuffer(instr.cbuf34.index, instr.cbuf34.GetOffset()),
+ instr.hfma2.type_reg39, GetRegister(instr.gpr39)};
case OpCode::Id::HFMA2_RC:
neg_b = instr.hfma2.negate_b;
neg_c = instr.hfma2.negate_c;
return {instr.hfma2.saturate, instr.hfma2.type_reg39, GetRegister(instr.gpr39),
- instr.hfma2.type_b, GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset)};
+ instr.hfma2.type_b,
+ GetConstBuffer(instr.cbuf34.index, instr.cbuf34.GetOffset())};
case OpCode::Id::HFMA2_RR:
neg_b = instr.hfma2.rr.negate_b;
neg_c = instr.hfma2.rr.negate_c;
diff --git a/src/video_core/shader/decode/integer_set.cpp b/src/video_core/shader/decode/integer_set.cpp
index 85e67b03b..16eb3985f 100644
--- a/src/video_core/shader/decode/integer_set.cpp
+++ b/src/video_core/shader/decode/integer_set.cpp
@@ -23,7 +23,7 @@ u32 ShaderIR::DecodeIntegerSet(BasicBlock& bb, const BasicBlock& code, u32 pc) {
} else if (instr.is_b_gpr) {
return GetRegister(instr.gpr20);
} else {
- return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset);
+ return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.GetOffset());
}
}();
diff --git a/src/video_core/shader/decode/integer_set_predicate.cpp b/src/video_core/shader/decode/integer_set_predicate.cpp
index c8b105a08..daf97174b 100644
--- a/src/video_core/shader/decode/integer_set_predicate.cpp
+++ b/src/video_core/shader/decode/integer_set_predicate.cpp
@@ -25,7 +25,7 @@ u32 ShaderIR::DecodeIntegerSetPredicate(BasicBlock& bb, const BasicBlock& code,
} else if (instr.is_b_gpr) {
return GetRegister(instr.gpr20);
} else {
- return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset);
+ return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.GetOffset());
}
}();
diff --git a/src/video_core/shader/decode/memory.cpp b/src/video_core/shader/decode/memory.cpp
index 04cb386b7..4d075f088 100644
--- a/src/video_core/shader/decode/memory.cpp
+++ b/src/video_core/shader/decode/memory.cpp
@@ -80,7 +80,7 @@ u32 ShaderIR::DecodeMemory(BasicBlock& bb, const BasicBlock& code, u32 pc) {
Node index = GetRegister(instr.gpr8);
const Node op_a =
- GetConstBufferIndirect(instr.cbuf36.index, instr.cbuf36.offset + 0, index);
+ GetConstBufferIndirect(instr.cbuf36.index, instr.cbuf36.GetOffset() + 0, index);
switch (instr.ld_c.type.Value()) {
case Tegra::Shader::UniformType::Single:
@@ -89,7 +89,7 @@ u32 ShaderIR::DecodeMemory(BasicBlock& bb, const BasicBlock& code, u32 pc) {
case Tegra::Shader::UniformType::Double: {
const Node op_b =
- GetConstBufferIndirect(instr.cbuf36.index, instr.cbuf36.offset + 4, index);
+ GetConstBufferIndirect(instr.cbuf36.index, instr.cbuf36.GetOffset() + 4, index);
SetTemporal(bb, 0, op_a);
SetTemporal(bb, 1, op_b);
@@ -142,7 +142,7 @@ u32 ShaderIR::DecodeMemory(BasicBlock& bb, const BasicBlock& code, u32 pc) {
ASSERT(cbuf != nullptr);
const auto cbuf_offset_imm = std::get_if<ImmediateNode>(cbuf->GetOffset());
ASSERT(cbuf_offset_imm != nullptr);
- const auto cbuf_offset = cbuf_offset_imm->GetValue() * 4;
+ const auto cbuf_offset = cbuf_offset_imm->GetValue();
bb.push_back(Comment(
fmt::format("Base address is c[0x{:x}][0x{:x}]", cbuf->GetIndex(), cbuf_offset)));
diff --git a/src/video_core/shader/decode/shift.cpp b/src/video_core/shader/decode/shift.cpp
index 85026bb37..6623f8ff9 100644
--- a/src/video_core/shader/decode/shift.cpp
+++ b/src/video_core/shader/decode/shift.cpp
@@ -23,7 +23,7 @@ u32 ShaderIR::DecodeShift(BasicBlock& bb, const BasicBlock& code, u32 pc) {
} else if (instr.is_b_gpr) {
return GetRegister(instr.gpr20);
} else {
- return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset);
+ return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.GetOffset());
}
}();
diff --git a/src/video_core/shader/decode/xmad.cpp b/src/video_core/shader/decode/xmad.cpp
index 0cd9cd1cc..9cb864500 100644
--- a/src/video_core/shader/decode/xmad.cpp
+++ b/src/video_core/shader/decode/xmad.cpp
@@ -32,13 +32,14 @@ u32 ShaderIR::DecodeXmad(BasicBlock& bb, const BasicBlock& code, u32 pc) {
auto [is_merge, op_b, op_c] = [&]() -> std::tuple<bool, Node, Node> {
switch (opcode->get().GetId()) {
case OpCode::Id::XMAD_CR:
- return {instr.xmad.merge_56, GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset),
+ return {instr.xmad.merge_56,
+ GetConstBuffer(instr.cbuf34.index, instr.cbuf34.GetOffset()),
GetRegister(instr.gpr39)};
case OpCode::Id::XMAD_RR:
return {instr.xmad.merge_37, GetRegister(instr.gpr20), GetRegister(instr.gpr39)};
case OpCode::Id::XMAD_RC:
return {false, GetRegister(instr.gpr39),
- GetConstBuffer(instr.cbuf34.index, instr.cbuf34.offset)};
+ GetConstBuffer(instr.cbuf34.index, instr.cbuf34.GetOffset())};
case OpCode::Id::XMAD_IMM:
return {instr.xmad.merge_37, Immediate(static_cast<u32>(instr.xmad.imm20_16)),
GetRegister(instr.gpr39)};
diff --git a/src/video_core/shader/shader_ir.h b/src/video_core/shader/shader_ir.h
index c4ecb2e3c..6e42e3dfb 100644
--- a/src/video_core/shader/shader_ir.h
+++ b/src/video_core/shader/shader_ir.h
@@ -249,7 +249,7 @@ public:
}
u32 GetSize() const {
- return max_offset + 1;
+ return max_offset + sizeof(float);
}
private: