summaryrefslogtreecommitdiffstats
path: root/src
diff options
context:
space:
mode:
authorReinUsesLisp <reinuseslisp@airmail.cc>2019-11-01 04:34:38 +0100
committerReinUsesLisp <reinuseslisp@airmail.cc>2019-12-10 03:25:21 +0100
commit6233b1db089686e1ea0dde68278e9086981d6f76 (patch)
tree42199ac76ce7f34d543d77a5a70cc3e5f46e2576 /src
parentMerge pull request #3205 from ReinUsesLisp/vk-device (diff)
downloadyuzu-6233b1db089686e1ea0dde68278e9086981d6f76.tar
yuzu-6233b1db089686e1ea0dde68278e9086981d6f76.tar.gz
yuzu-6233b1db089686e1ea0dde68278e9086981d6f76.tar.bz2
yuzu-6233b1db089686e1ea0dde68278e9086981d6f76.tar.lz
yuzu-6233b1db089686e1ea0dde68278e9086981d6f76.tar.xz
yuzu-6233b1db089686e1ea0dde68278e9086981d6f76.tar.zst
yuzu-6233b1db089686e1ea0dde68278e9086981d6f76.zip
Diffstat (limited to 'src')
-rw-r--r--src/video_core/engines/shader_bytecode.h3
-rw-r--r--src/video_core/shader/decode/memory.cpp34
-rw-r--r--src/video_core/shader/node.h20
-rw-r--r--src/video_core/shader/track.cpp1
4 files changed, 38 insertions, 20 deletions
diff --git a/src/video_core/engines/shader_bytecode.h b/src/video_core/engines/shader_bytecode.h
index 8b7dcbe9d..7703a76a3 100644
--- a/src/video_core/engines/shader_bytecode.h
+++ b/src/video_core/engines/shader_bytecode.h
@@ -98,10 +98,11 @@ union Attribute {
BitField<20, 10, u64> immediate;
BitField<22, 2, u64> element;
BitField<24, 6, Index> index;
+ BitField<31, 1, u64> patch;
BitField<47, 3, AttributeSize> size;
bool IsPhysical() const {
- return element == 0 && static_cast<u64>(index.Value()) == 0;
+ return patch == 0 && element == 0 && static_cast<u64>(index.Value()) == 0;
}
} fmt20;
diff --git a/src/video_core/shader/decode/memory.cpp b/src/video_core/shader/decode/memory.cpp
index 335d78146..78e92f52e 100644
--- a/src/video_core/shader/decode/memory.cpp
+++ b/src/video_core/shader/decode/memory.cpp
@@ -21,6 +21,7 @@ using Tegra::Shader::OpCode;
using Tegra::Shader::Register;
namespace {
+
u32 GetUniformTypeElementsCount(Tegra::Shader::UniformType uniform_type) {
switch (uniform_type) {
case Tegra::Shader::UniformType::Single:
@@ -35,6 +36,7 @@ u32 GetUniformTypeElementsCount(Tegra::Shader::UniformType uniform_type) {
return 1;
}
}
+
} // Anonymous namespace
u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
@@ -196,28 +198,28 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
UNIMPLEMENTED_IF_MSG((instr.attribute.fmt20.immediate.Value() % sizeof(u32)) != 0,
"Unaligned attribute loads are not supported");
- u64 next_element = instr.attribute.fmt20.element;
- auto next_index = static_cast<u64>(instr.attribute.fmt20.index.Value());
+ u64 element = instr.attribute.fmt20.element;
+ auto index = static_cast<u64>(instr.attribute.fmt20.index.Value());
- const auto StoreNextElement = [&](u32 reg_offset) {
- const auto dest = GetOutputAttribute(static_cast<Attribute::Index>(next_index),
- next_element, GetRegister(instr.gpr39));
+ const u32 num_words = static_cast<u32>(instr.attribute.fmt20.size.Value()) + 1;
+ for (u32 reg_offset = 0; reg_offset < num_words; ++reg_offset) {
+ Node dest;
+ if (instr.attribute.fmt20.patch) {
+ const u32 offset = static_cast<u32>(index) * 4 + static_cast<u32>(element);
+ dest = MakeNode<PatchNode>(offset);
+ } else {
+ dest = GetOutputAttribute(static_cast<Attribute::Index>(index), element,
+ GetRegister(instr.gpr39));
+ }
const auto src = GetRegister(instr.gpr0.Value() + reg_offset);
bb.push_back(Operation(OperationCode::Assign, dest, src));
- // Load the next attribute element into the following register. If the element
- // to load goes beyond the vec4 size, load the first element of the next
- // attribute.
- next_element = (next_element + 1) % 4;
- next_index = next_index + (next_element == 0 ? 1 : 0);
- };
-
- const u32 num_words = static_cast<u32>(instr.attribute.fmt20.size.Value()) + 1;
- for (u32 reg_offset = 0; reg_offset < num_words; ++reg_offset) {
- StoreNextElement(reg_offset);
+ // Load the next attribute element into the following register. If the element to load
+ // goes beyond the vec4 size, load the first element of the next attribute.
+ element = (element + 1) % 4;
+ index = index + (element == 0 ? 1 : 0);
}
-
break;
}
case OpCode::Id::ST_L:
diff --git a/src/video_core/shader/node.h b/src/video_core/shader/node.h
index b2576bdd6..8f4811c88 100644
--- a/src/video_core/shader/node.h
+++ b/src/video_core/shader/node.h
@@ -213,13 +213,14 @@ class PredicateNode;
class AbufNode;
class CbufNode;
class LmemNode;
+class PatchNode;
class SmemNode;
class GmemNode;
class CommentNode;
-using NodeData =
- std::variant<OperationNode, ConditionalNode, GprNode, ImmediateNode, InternalFlagNode,
- PredicateNode, AbufNode, CbufNode, LmemNode, SmemNode, GmemNode, CommentNode>;
+using NodeData = std::variant<OperationNode, ConditionalNode, GprNode, ImmediateNode,
+ InternalFlagNode, PredicateNode, AbufNode, PatchNode, CbufNode,
+ LmemNode, SmemNode, GmemNode, CommentNode>;
using Node = std::shared_ptr<NodeData>;
using Node4 = std::array<Node, 4>;
using NodeBlock = std::vector<Node>;
@@ -542,6 +543,19 @@ private:
u32 element{};
};
+/// Patch memory (used to communicate tessellation stages).
+class PatchNode final {
+public:
+ explicit PatchNode(u32 offset) : offset{offset} {}
+
+ u32 GetOffset() const {
+ return offset;
+ }
+
+private:
+ u32 offset{};
+};
+
/// Constant buffer node, usually mapped to uniform buffers in GLSL
class CbufNode final {
public:
diff --git a/src/video_core/shader/track.cpp b/src/video_core/shader/track.cpp
index 55f5949e4..165c79330 100644
--- a/src/video_core/shader/track.cpp
+++ b/src/video_core/shader/track.cpp
@@ -7,6 +7,7 @@
#include <variant>
#include "common/common_types.h"
+#include "video_core/shader/node.h"
#include "video_core/shader/shader_ir.h"
namespace VideoCommon::Shader {