From f4603d23c551ece65cd205a850a31a84531daf43 Mon Sep 17 00:00:00 2001 From: Fernando Sahmkow Date: Sun, 5 Jan 2020 15:53:22 -0400 Subject: Shader_IR: Setup Indexed Samplers on the IR --- src/video_core/shader/decode/texture.cpp | 66 ++++++++++++++++++++++---------- 1 file changed, 46 insertions(+), 20 deletions(-) (limited to 'src/video_core/shader/decode') diff --git a/src/video_core/shader/decode/texture.cpp b/src/video_core/shader/decode/texture.cpp index 0b567e39d..886650d9e 100644 --- a/src/video_core/shader/decode/texture.cpp +++ b/src/video_core/shader/decode/texture.cpp @@ -389,31 +389,57 @@ const Sampler* ShaderIR::GetSampler(const Tegra::Shader::Sampler& sampler, const Sampler* ShaderIR::GetBindlessSampler(Tegra::Shader::Register reg, std::optional sampler_info) { const Node sampler_register = GetRegister(reg); - const auto [base_sampler, buffer, offset] = - TrackCbuf(sampler_register, global_code, static_cast(global_code.size())); - ASSERT(base_sampler != nullptr); - if (base_sampler == nullptr) { + const auto [base_node, tracked_sampler_info] = + TrackSampler(sampler_register, global_code, static_cast(global_code.size())); + ASSERT(base_node != nullptr); + if (base_node == nullptr) { return nullptr; } - const auto info = GetSamplerInfo(sampler_info, offset, buffer); + if (const auto bindless_sampler_info = + std::get_if(&*tracked_sampler_info)) { + const u32 buffer = bindless_sampler_info->GetIndex(); + const u32 offset = bindless_sampler_info->GetOffset(); + const auto info = GetSamplerInfo(sampler_info, offset, buffer); + + // If this sampler has already been used, return the existing mapping. + const auto it = + std::find_if(used_samplers.begin(), used_samplers.end(), + [buffer = buffer, offset = offset](const Sampler& entry) { + return entry.GetBuffer() == buffer && entry.GetOffset() == offset; + }); + if (it != used_samplers.end()) { + ASSERT(it->IsBindless() && it->GetType() == info.type && + it->IsArray() == info.is_array && it->IsShadow() == info.is_shadow); + return &*it; + } - // If this sampler has already been used, return the existing mapping. - const auto it = - std::find_if(used_samplers.begin(), used_samplers.end(), - [buffer = buffer, offset = offset](const Sampler& entry) { - return entry.GetBuffer() == buffer && entry.GetOffset() == offset; - }); - if (it != used_samplers.end()) { - ASSERT(it->IsBindless() && it->GetType() == info.type && it->IsArray() == info.is_array && - it->IsShadow() == info.is_shadow); - return &*it; - } + // Otherwise create a new mapping for this sampler + const auto next_index = static_cast(used_samplers.size()); + return &used_samplers.emplace_back(next_index, offset, buffer, info.type, info.is_array, + info.is_shadow, info.is_buffer); + } else if (const auto array_sampler_info = + std::get_if(&*tracked_sampler_info)) { + const u32 base_offset = array_sampler_info->GetBaseOffset() / 4; + const auto info = GetSamplerInfo(sampler_info, base_offset); + + // If this sampler has already been used, return the existing mapping. + const auto it = std::find_if( + used_samplers.begin(), used_samplers.end(), + [base_offset](const Sampler& entry) { return entry.GetOffset() == base_offset; }); + if (it != used_samplers.end()) { + ASSERT(!it->IsBindless() && it->GetType() == info.type && + it->IsArray() == info.is_array && it->IsShadow() == info.is_shadow && + it->IsBuffer() == info.is_buffer); + return &*it; + } - // Otherwise create a new mapping for this sampler - const auto next_index = static_cast(used_samplers.size()); - return &used_samplers.emplace_back(next_index, offset, buffer, info.type, info.is_array, - info.is_shadow, info.is_buffer); + // Otherwise create a new mapping for this sampler + const auto next_index = static_cast(used_samplers.size()); + return &used_samplers.emplace_back(next_index, base_offset, info.type, info.is_array, + info.is_shadow, info.is_buffer); + } + return nullptr; } void ShaderIR::WriteTexInstructionFloat(NodeBlock& bb, Instruction instr, const Node4& components) { -- cgit v1.2.3 From 037ea431ceb93e93274fdcf9fb724819639d04fd Mon Sep 17 00:00:00 2001 From: Fernando Sahmkow Date: Sun, 5 Jan 2020 18:36:21 -0400 Subject: Shader_IR: deduce size of indexed samplers --- src/video_core/shader/decode.cpp | 36 ++++++++++++++++++++++++++++++++ src/video_core/shader/decode/texture.cpp | 9 ++++---- src/video_core/shader/node.h | 22 +++++++++++++++---- src/video_core/shader/shader_ir.h | 1 + 4 files changed, 60 insertions(+), 8 deletions(-) (limited to 'src/video_core/shader/decode') diff --git a/src/video_core/shader/decode.cpp b/src/video_core/shader/decode.cpp index 507614d59..dd2f68a3e 100644 --- a/src/video_core/shader/decode.cpp +++ b/src/video_core/shader/decode.cpp @@ -56,6 +56,29 @@ void DeduceTextureHandlerSize(VideoCore::GuestDriverProfile* gpu_driver, } } +std::optional TryDeduceSamplerSize(Sampler& sampler_to_deduce, + VideoCore::GuestDriverProfile* gpu_driver, + std::list& used_samplers) { + if (gpu_driver == nullptr) { + LOG_CRITICAL(HW_GPU, "GPU Driver profile has not been created yet"); + return std::nullopt; + } + const u32 base_offset = sampler_to_deduce.GetOffset(); + u32 max_offset{UINT_MAX}; + for (const auto& sampler : used_samplers) { + if (sampler.IsBindless()) { + continue; + } + if (sampler.GetOffset() > base_offset) { + max_offset = std::min(sampler.GetOffset(), max_offset); + } + } + if (max_offset == UINT_MAX) { + return std::nullopt; + } + return ((max_offset - base_offset) * 4) / gpu_driver->GetTextureHandlerSize(); +} + } // Anonymous namespace class ASTDecoder { @@ -342,6 +365,19 @@ void ShaderIR::PostDecode() { // Deduce texture handler size if needed auto gpu_driver = locker.AccessGuestDriverProfile(); DeduceTextureHandlerSize(gpu_driver, used_samplers); + // Deduce Indexed Samplers + if (uses_indexed_samplers) { + for (auto& sampler : used_samplers) { + if (sampler.IsIndexed()) { + auto size = TryDeduceSamplerSize(sampler, gpu_driver, used_samplers); + if (size) { + sampler.SetSize(*size); + } else { + sampler.SetSize(1); + } + } + } + } } } // namespace VideoCommon::Shader diff --git a/src/video_core/shader/decode/texture.cpp b/src/video_core/shader/decode/texture.cpp index 886650d9e..e7c38f5d6 100644 --- a/src/video_core/shader/decode/texture.cpp +++ b/src/video_core/shader/decode/texture.cpp @@ -383,7 +383,7 @@ const Sampler* ShaderIR::GetSampler(const Tegra::Shader::Sampler& sampler, // Otherwise create a new mapping for this sampler const auto next_index = static_cast(used_samplers.size()); return &used_samplers.emplace_back(next_index, offset, info.type, info.is_array, info.is_shadow, - info.is_buffer); + info.is_buffer, false); } const Sampler* ShaderIR::GetBindlessSampler(Tegra::Shader::Register reg, @@ -417,7 +417,7 @@ const Sampler* ShaderIR::GetBindlessSampler(Tegra::Shader::Register reg, // Otherwise create a new mapping for this sampler const auto next_index = static_cast(used_samplers.size()); return &used_samplers.emplace_back(next_index, offset, buffer, info.type, info.is_array, - info.is_shadow, info.is_buffer); + info.is_shadow, info.is_buffer, false); } else if (const auto array_sampler_info = std::get_if(&*tracked_sampler_info)) { const u32 base_offset = array_sampler_info->GetBaseOffset() / 4; @@ -430,14 +430,15 @@ const Sampler* ShaderIR::GetBindlessSampler(Tegra::Shader::Register reg, if (it != used_samplers.end()) { ASSERT(!it->IsBindless() && it->GetType() == info.type && it->IsArray() == info.is_array && it->IsShadow() == info.is_shadow && - it->IsBuffer() == info.is_buffer); + it->IsBuffer() == info.is_buffer && it->IsIndexed()); return &*it; } + uses_indexed_samplers = true; // Otherwise create a new mapping for this sampler const auto next_index = static_cast(used_samplers.size()); return &used_samplers.emplace_back(next_index, base_offset, info.type, info.is_array, - info.is_shadow, info.is_buffer); + info.is_shadow, info.is_buffer, true); } return nullptr; } diff --git a/src/video_core/shader/node.h b/src/video_core/shader/node.h index b370df8f9..2f29b9506 100644 --- a/src/video_core/shader/node.h +++ b/src/video_core/shader/node.h @@ -240,15 +240,15 @@ class Sampler { public: /// This constructor is for bound samplers constexpr explicit Sampler(u32 index, u32 offset, Tegra::Shader::TextureType type, - bool is_array, bool is_shadow, bool is_buffer) + bool is_array, bool is_shadow, bool is_buffer, bool is_indexed) : index{index}, offset{offset}, type{type}, is_array{is_array}, is_shadow{is_shadow}, - is_buffer{is_buffer} {} + is_buffer{is_buffer}, is_indexed{is_indexed} {} /// This constructor is for bindless samplers constexpr explicit Sampler(u32 index, u32 offset, u32 buffer, Tegra::Shader::TextureType type, - bool is_array, bool is_shadow, bool is_buffer) + bool is_array, bool is_shadow, bool is_buffer, bool is_indexed) : index{index}, offset{offset}, buffer{buffer}, type{type}, is_array{is_array}, - is_shadow{is_shadow}, is_buffer{is_buffer}, is_bindless{true} {} + is_shadow{is_shadow}, is_buffer{is_buffer}, is_bindless{true}, is_indexed{is_indexed} {} constexpr u32 GetIndex() const { return index; @@ -282,16 +282,30 @@ public: return is_bindless; } + constexpr bool IsIndexed() const { + return is_indexed; + } + + constexpr u32 Size() const { + return size; + } + + void SetSize(u32 new_size) { + size = new_size; + } + private: u32 index{}; ///< Emulated index given for the this sampler. u32 offset{}; ///< Offset in the const buffer from where the sampler is being read. u32 buffer{}; ///< Buffer where the bindless sampler is being read (unused on bound samplers). + u32 size{}; ///< Size of the sampler if indexed. Tegra::Shader::TextureType type{}; ///< The type used to sample this texture (Texture2D, etc) bool is_array{}; ///< Whether the texture is being sampled as an array texture or not. bool is_shadow{}; ///< Whether the texture is being sampled as a depth texture or not. bool is_buffer{}; ///< Whether the texture is a texture buffer without sampler. bool is_bindless{}; ///< Whether this sampler belongs to a bindless texture or not. + bool is_indexed{}; ///< Whether this sampler is an indexed array of textures. }; /// Represents a tracked bindless sampler into a direct const buffer diff --git a/src/video_core/shader/shader_ir.h b/src/video_core/shader/shader_ir.h index d85f14c97..121528346 100644 --- a/src/video_core/shader/shader_ir.h +++ b/src/video_core/shader/shader_ir.h @@ -435,6 +435,7 @@ private: bool uses_instance_id{}; bool uses_vertex_id{}; bool uses_warps{}; + bool uses_indexed_samplers{}; Tegra::Shader::Header header; }; -- cgit v1.2.3 From 7c530e06661d760eb6366724d109468423363072 Mon Sep 17 00:00:00 2001 From: Fernando Sahmkow Date: Tue, 7 Jan 2020 17:45:12 -0400 Subject: Shader_IR: Propagate bindless index into the GL compiler. --- .../renderer_opengl/gl_shader_decompiler.cpp | 2 +- src/video_core/shader/decode/texture.cpp | 40 +++++++++++++--------- src/video_core/shader/node.h | 2 +- src/video_core/shader/shader_ir.h | 5 ++- src/video_core/shader/track.cpp | 29 ++++++++++++++-- 5 files changed, 54 insertions(+), 24 deletions(-) (limited to 'src/video_core/shader/decode') diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp index 8b413ae9a..df681bdcb 100644 --- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp +++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp @@ -1123,7 +1123,7 @@ private: if (!meta->sampler.IsIndexed()) { expr += '(' + GetSampler(meta->sampler) + ", "; } else { - expr += '(' + GetSampler(meta->sampler) + "[0], "; + expr += '(' + GetSampler(meta->sampler) + '[' + Visit(meta->index).AsUint() + "], "; } expr += coord_constructors.at(count + (has_array ? 1 : 0) + (has_shadow && !separate_dc ? 1 : 0) - 1); diff --git a/src/video_core/shader/decode/texture.cpp b/src/video_core/shader/decode/texture.cpp index e7c38f5d6..31b09b18c 100644 --- a/src/video_core/shader/decode/texture.cpp +++ b/src/video_core/shader/decode/texture.cpp @@ -144,7 +144,8 @@ u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) { Node4 values; for (u32 element = 0; element < values.size(); ++element) { auto coords_copy = coords; - MetaTexture meta{sampler, {}, depth_compare, aoffi, {}, {}, {}, {}, component, element}; + MetaTexture meta{sampler, {}, depth_compare, aoffi, {}, {}, + {}, {}, component, element, {}}; values[element] = Operation(OperationCode::TextureGather, meta, std::move(coords_copy)); } @@ -167,9 +168,9 @@ u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) { const auto derivate_reg = instr.gpr20.Value(); const auto texture_type = instr.txd.texture_type.Value(); const auto coord_count = GetCoordCount(texture_type); - + Node index_var{}; const Sampler* sampler = - is_bindless ? GetBindlessSampler(base_reg, {{texture_type, is_array, false}}) + is_bindless ? GetBindlessSampler(base_reg, index_var, {{texture_type, is_array, false}}) : GetSampler(instr.sampler, {{texture_type, is_array, false}}); Node4 values; if (sampler == nullptr) { @@ -200,7 +201,7 @@ u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) { } for (u32 element = 0; element < values.size(); ++element) { - MetaTexture meta{*sampler, array_node, {}, {}, {}, derivates, {}, {}, {}, element}; + MetaTexture meta{*sampler, array_node, {}, {}, {}, derivates, {}, {}, {}, element, index_var}; values[element] = Operation(OperationCode::TextureGradient, std::move(meta), coords); } @@ -215,8 +216,9 @@ u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) { // TODO: The new commits on the texture refactor, change the way samplers work. // Sadly, not all texture instructions specify the type of texture their sampler // uses. This must be fixed at a later instance. + Node index_var{}; const Sampler* sampler = - is_bindless ? GetBindlessSampler(instr.gpr8) : GetSampler(instr.sampler); + is_bindless ? GetBindlessSampler(instr.gpr8, index_var) : GetSampler(instr.sampler); if (sampler == nullptr) { u32 indexer = 0; @@ -240,7 +242,7 @@ u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) { if (!instr.txq.IsComponentEnabled(element)) { continue; } - MetaTexture meta{*sampler, {}, {}, {}, {}, {}, {}, {}, {}, element}; + MetaTexture meta{*sampler, {}, {}, {}, {}, {}, {}, {}, {}, element, index_var}; const Node value = Operation(OperationCode::TextureQueryDimensions, meta, GetRegister(instr.gpr8.Value() + (is_bindless ? 1 : 0))); @@ -266,8 +268,9 @@ u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) { auto texture_type = instr.tmml.texture_type.Value(); const bool is_array = instr.tmml.array != 0; + Node index_var{}; const Sampler* sampler = - is_bindless ? GetBindlessSampler(instr.gpr20) : GetSampler(instr.sampler); + is_bindless ? GetBindlessSampler(instr.gpr20, index_var) : GetSampler(instr.sampler); if (sampler == nullptr) { u32 indexer = 0; @@ -309,7 +312,7 @@ u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) { continue; } auto params = coords; - MetaTexture meta{*sampler, {}, {}, {}, {}, {}, {}, {}, {}, element}; + MetaTexture meta{*sampler, {}, {}, {}, {}, {}, {}, {}, {}, element, index_var}; const Node value = Operation(OperationCode::TextureQueryLod, meta, std::move(params)); SetTemporary(bb, indexer++, value); } @@ -386,7 +389,7 @@ const Sampler* ShaderIR::GetSampler(const Tegra::Shader::Sampler& sampler, info.is_buffer, false); } -const Sampler* ShaderIR::GetBindlessSampler(Tegra::Shader::Register reg, +const Sampler* ShaderIR::GetBindlessSampler(Tegra::Shader::Register reg, Node& index_var, std::optional sampler_info) { const Node sampler_register = GetRegister(reg); const auto [base_node, tracked_sampler_info] = @@ -421,6 +424,7 @@ const Sampler* ShaderIR::GetBindlessSampler(Tegra::Shader::Register reg, } else if (const auto array_sampler_info = std::get_if(&*tracked_sampler_info)) { const u32 base_offset = array_sampler_info->GetBaseOffset() / 4; + index_var = GetCustomVariable(array_sampler_info->GetIndexVar()); const auto info = GetSamplerInfo(sampler_info, base_offset); // If this sampler has already been used, return the existing mapping. @@ -526,8 +530,9 @@ Node4 ShaderIR::GetTextureCode(Instruction instr, TextureType texture_type, "This method is not supported."); const SamplerInfo info{texture_type, is_array, is_shadow, false}; - const Sampler* sampler = - is_bindless ? GetBindlessSampler(*bindless_reg, info) : GetSampler(instr.sampler, info); + Node index_var{}; + const Sampler* sampler = is_bindless ? GetBindlessSampler(*bindless_reg, index_var, info) + : GetSampler(instr.sampler, info); Node4 values; if (sampler == nullptr) { for (u32 element = 0; element < values.size(); ++element) { @@ -575,7 +580,8 @@ Node4 ShaderIR::GetTextureCode(Instruction instr, TextureType texture_type, for (u32 element = 0; element < values.size(); ++element) { auto copy_coords = coords; - MetaTexture meta{*sampler, array, depth_compare, aoffi, {}, {}, bias, lod, {}, element}; + MetaTexture meta{*sampler, array, depth_compare, aoffi, {}, {}, bias, + lod, {}, element, index_var}; values[element] = Operation(read_method, meta, std::move(copy_coords)); } @@ -690,7 +696,8 @@ Node4 ShaderIR::GetTld4Code(Instruction instr, TextureType texture_type, bool de u64 parameter_register = instr.gpr20.Value(); const SamplerInfo info{texture_type, is_array, depth_compare, false}; - const Sampler* sampler = is_bindless ? GetBindlessSampler(parameter_register++, info) + Node index_var{}; + const Sampler* sampler = is_bindless ? GetBindlessSampler(parameter_register++, index_var, info) : GetSampler(instr.sampler, info); Node4 values; if (sampler == nullptr) { @@ -719,7 +726,8 @@ Node4 ShaderIR::GetTld4Code(Instruction instr, TextureType texture_type, bool de for (u32 element = 0; element < values.size(); ++element) { auto coords_copy = coords; MetaTexture meta{ - *sampler, GetRegister(array_register), dc, aoffi, ptp, {}, {}, {}, component, element}; + *sampler, GetRegister(array_register), dc, aoffi, ptp, {}, {}, {}, component, element, + index_var}; values[element] = Operation(OperationCode::TextureGather, meta, std::move(coords_copy)); } @@ -752,7 +760,7 @@ Node4 ShaderIR::GetTldCode(Tegra::Shader::Instruction instr) { Node4 values; for (u32 element = 0; element < values.size(); ++element) { auto coords_copy = coords; - MetaTexture meta{sampler, array_register, {}, {}, {}, {}, {}, lod, {}, element}; + MetaTexture meta{sampler, array_register, {}, {}, {}, {}, {}, lod, {}, element, {}}; values[element] = Operation(OperationCode::TexelFetch, meta, std::move(coords_copy)); } @@ -802,7 +810,7 @@ Node4 ShaderIR::GetTldsCode(Instruction instr, TextureType texture_type, bool is Node4 values; for (u32 element = 0; element < values.size(); ++element) { auto coords_copy = coords; - MetaTexture meta{sampler, array, {}, {}, {}, {}, {}, lod, {}, element}; + MetaTexture meta{sampler, array, {}, {}, {}, {}, {}, lod, {}, element, {}}; values[element] = Operation(OperationCode::TexelFetch, meta, std::move(coords_copy)); } return values; diff --git a/src/video_core/shader/node.h b/src/video_core/shader/node.h index db06767f6..d75453458 100644 --- a/src/video_core/shader/node.h +++ b/src/video_core/shader/node.h @@ -445,6 +445,7 @@ struct MetaTexture { Node lod; Node component{}; u32 element{}; + Node index{}; }; struct MetaImage { @@ -564,7 +565,6 @@ private: u32 index{}; }; - /// A 32-bits value that represents an immediate value class ImmediateNode final { public: diff --git a/src/video_core/shader/shader_ir.h b/src/video_core/shader/shader_ir.h index 2fe14e815..0421dac0c 100644 --- a/src/video_core/shader/shader_ir.h +++ b/src/video_core/shader/shader_ir.h @@ -328,7 +328,7 @@ private: std::optional sampler_info = std::nullopt); /// Accesses a texture sampler for a bindless texture. - const Sampler* GetBindlessSampler(Tegra::Shader::Register reg, + const Sampler* GetBindlessSampler(Tegra::Shader::Register reg, Node& index_var, std::optional sampler_info = std::nullopt); /// Accesses an image. @@ -394,8 +394,7 @@ private: std::tuple TrackCbuf(Node tracked, const NodeBlock& code, s64 cursor) const; - std::tuple TrackSampler(Node tracked, const NodeBlock& code, - s64 cursor) const; + std::tuple TrackSampler(Node tracked, const NodeBlock& code, s64 cursor); std::optional TrackImmediate(Node tracked, const NodeBlock& code, s64 cursor) const; diff --git a/src/video_core/shader/track.cpp b/src/video_core/shader/track.cpp index 69a677394..d449b625e 100644 --- a/src/video_core/shader/track.cpp +++ b/src/video_core/shader/track.cpp @@ -61,8 +61,19 @@ std::optional> DecoupleIndirectRead(const OperationNode& o return std::nullopt; } +bool AmendNodeCv(std::size_t amend_index, Node node) { + if (const auto operation = std::get_if(&*node)) { + operation->SetAmendIndex(amend_index); + return true; + } else if (const auto conditional = std::get_if(&*node)) { + conditional->SetAmendIndex(amend_index); + return true; + } + return false; +} + std::tuple ShaderIR::TrackSampler(Node tracked, const NodeBlock& code, - s64 cursor) const { + s64 cursor) { if (const auto cbuf = std::get_if(&*tracked)) { // Constant buffer found, test if it's an immediate const auto offset = cbuf->GetOffset(); @@ -84,9 +95,21 @@ std::tuple ShaderIR::TrackSampler(Node tracked, const NodeBl } auto [gpr, base_offset] = *pair; const auto offset_inm = std::get_if(&*base_offset); + auto gpu_driver = locker.AccessGuestDriverProfile(); + if (gpu_driver == nullptr) { + return {}; + } + const u32 bindless_cv = NewCustomVariable(); + const Node op = Operation(OperationCode::UDiv, NO_PRECISE, gpr, + Immediate(gpu_driver->GetTextureHandlerSize())); + + const Node cv_node = GetCustomVariable(bindless_cv); + Node amend_op = Operation(OperationCode::Assign, cv_node, std::move(op)); + const std::size_t amend_index = DeclareAmend(amend_op); + AmendNodeCv(amend_index, code[cursor]); // TODO Implement Bindless Index custom variable - auto track = - MakeTrackSampler(cbuf->GetIndex(), offset_inm->GetValue(), 0); + auto track = MakeTrackSampler(cbuf->GetIndex(), + offset_inm->GetValue(), bindless_cv); return {tracked, track}; } return {}; -- cgit v1.2.3 From 806f5691430b86640d64d4c5ae77c5e1dac1625a Mon Sep 17 00:00:00 2001 From: Fernando Sahmkow Date: Wed, 8 Jan 2020 15:59:21 -0400 Subject: Shader_IR: Change name of TrackSampler function so it does not confuse with the type. --- src/video_core/shader/decode/texture.cpp | 2 +- src/video_core/shader/shader_ir.h | 3 ++- src/video_core/shader/track.cpp | 12 +++++++----- 3 files changed, 10 insertions(+), 7 deletions(-) (limited to 'src/video_core/shader/decode') diff --git a/src/video_core/shader/decode/texture.cpp b/src/video_core/shader/decode/texture.cpp index 31b09b18c..6da9668fe 100644 --- a/src/video_core/shader/decode/texture.cpp +++ b/src/video_core/shader/decode/texture.cpp @@ -393,7 +393,7 @@ const Sampler* ShaderIR::GetBindlessSampler(Tegra::Shader::Register reg, Node& i std::optional sampler_info) { const Node sampler_register = GetRegister(reg); const auto [base_node, tracked_sampler_info] = - TrackSampler(sampler_register, global_code, static_cast(global_code.size())); + TrackBindlessSampler(sampler_register, global_code, static_cast(global_code.size())); ASSERT(base_node != nullptr); if (base_node == nullptr) { return nullptr; diff --git a/src/video_core/shader/shader_ir.h b/src/video_core/shader/shader_ir.h index 0421dac0c..43672b41c 100644 --- a/src/video_core/shader/shader_ir.h +++ b/src/video_core/shader/shader_ir.h @@ -394,7 +394,8 @@ private: std::tuple TrackCbuf(Node tracked, const NodeBlock& code, s64 cursor) const; - std::tuple TrackSampler(Node tracked, const NodeBlock& code, s64 cursor); + std::tuple TrackBindlessSampler(Node tracked, const NodeBlock& code, + s64 cursor); std::optional TrackImmediate(Node tracked, const NodeBlock& code, s64 cursor) const; diff --git a/src/video_core/shader/track.cpp b/src/video_core/shader/track.cpp index d449b625e..4db721f69 100644 --- a/src/video_core/shader/track.cpp +++ b/src/video_core/shader/track.cpp @@ -72,8 +72,8 @@ bool AmendNodeCv(std::size_t amend_index, Node node) { return false; } -std::tuple ShaderIR::TrackSampler(Node tracked, const NodeBlock& code, - s64 cursor) { +std::tuple ShaderIR::TrackBindlessSampler(Node tracked, const NodeBlock& code, + s64 cursor) { if (const auto cbuf = std::get_if(&*tracked)) { // Constant buffer found, test if it's an immediate const auto offset = cbuf->GetOffset(); @@ -124,11 +124,12 @@ std::tuple ShaderIR::TrackSampler(Node tracked, const NodeBl if (!source) { return {}; } - return TrackSampler(source, code, new_cursor); + return TrackBindlessSampler(source, code, new_cursor); } if (const auto operation = std::get_if(&*tracked)) { for (std::size_t i = operation->GetOperandsCount(); i > 0; --i) { - if (auto found = TrackSampler((*operation)[i - 1], code, cursor); std::get<0>(found)) { + if (auto found = TrackBindlessSampler((*operation)[i - 1], code, cursor); + std::get<0>(found)) { // Cbuf found in operand. return found; } @@ -137,7 +138,8 @@ std::tuple ShaderIR::TrackSampler(Node tracked, const NodeBl } if (const auto conditional = std::get_if(&*tracked)) { const auto& conditional_code = conditional->GetCode(); - return TrackSampler(tracked, conditional_code, static_cast(conditional_code.size())); + return TrackBindlessSampler(tracked, conditional_code, + static_cast(conditional_code.size())); } return {}; } -- cgit v1.2.3 From bb8eb15d392d69693f8cda0427669d011e23db97 Mon Sep 17 00:00:00 2001 From: Fernando Sahmkow Date: Fri, 24 Jan 2020 10:44:34 -0400 Subject: Shader_IR: Address feedback. --- src/video_core/guest_driver.h | 1 + .../renderer_opengl/gl_shader_decompiler.cpp | 6 ++--- .../renderer_vulkan/vk_shader_decompiler.cpp | 5 ++-- src/video_core/shader/const_buffer_locker.h | 2 ++ src/video_core/shader/decode.cpp | 31 +++++++++++----------- src/video_core/shader/decode/texture.cpp | 3 ++- src/video_core/shader/node.h | 14 +++++----- src/video_core/shader/shader_ir.cpp | 3 +-- src/video_core/shader/shader_ir.h | 2 +- src/video_core/shader/track.cpp | 9 +++---- 10 files changed, 40 insertions(+), 36 deletions(-) (limited to 'src/video_core/shader/decode') diff --git a/src/video_core/guest_driver.h b/src/video_core/guest_driver.h index 0a9a826b6..fc1917347 100644 --- a/src/video_core/guest_driver.h +++ b/src/video_core/guest_driver.h @@ -33,6 +33,7 @@ private: // This goes with Vulkan and OpenGL standards but Nvidia GPUs can easily // use 4 bytes instead. Thus, certain drivers may squish the size. static constexpr u32 default_texture_handler_size = 8; + u32 texture_handler_size = default_texture_handler_size; bool texture_handler_size_deduced = false; }; diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp index 2f2bb07a4..cb1a5f35c 100644 --- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp +++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp @@ -505,11 +505,11 @@ private: } void DeclareCustomVariables() { - const u32 cv_num = ir.GetCustomVariablesAmount(); - for (u32 i = 0; i < cv_num; ++i) { + const u32 num_custom_variables = ir.GetNumCustomVariables(); + for (u32 i = 0; i < num_custom_variables; ++i) { code.AddLine("float {} = 0.0f;", GetCustomVariable(i)); } - if (cv_num > 0) { + if (num_custom_variables > 0) { code.AddNewLine(); } } diff --git a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp index 130060369..36d928fab 100644 --- a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp +++ b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp @@ -589,8 +589,8 @@ private: } void DeclareCustomVariables() { - const u32 cv_num = ir.GetCustomVariablesAmount(); - for (u32 i = 0; i < cv_num; ++i) { + const u32 num_custom_variables = ir.GetNumCustomVariables(); + for (u32 i = 0; i < num_custom_variables; ++i) { const Id id = OpVariable(t_prv_float, spv::StorageClass::Private, v_float_zero); Name(id, fmt::format("custom_var_{}", i)); custom_variables.emplace(i, AddGlobalVariable(id)); @@ -1363,6 +1363,7 @@ private: } else if (const auto cv = std::get_if(&*dest)) { target = {custom_variables.at(cv->GetIndex()), Type::Float}; + } else { UNIMPLEMENTED(); } diff --git a/src/video_core/shader/const_buffer_locker.h b/src/video_core/shader/const_buffer_locker.h index fd1bb476a..d3ea11087 100644 --- a/src/video_core/shader/const_buffer_locker.h +++ b/src/video_core/shader/const_buffer_locker.h @@ -77,10 +77,12 @@ public: return bindless_samplers; } + /// Gets bound buffer used on this shader u32 GetBoundBuffer() const { return bound_buffer; } + /// Obtains access to the guest driver's profile. VideoCore::GuestDriverProfile* AccessGuestDriverProfile() const { if (engine) { return &engine->AccessGuestDriverProfile(); diff --git a/src/video_core/shader/decode.cpp b/src/video_core/shader/decode.cpp index d4a10eee5..6b697ed5d 100644 --- a/src/video_core/shader/decode.cpp +++ b/src/video_core/shader/decode.cpp @@ -35,9 +35,9 @@ constexpr bool IsSchedInstruction(u32 offset, u32 main_offset) { } void DeduceTextureHandlerSize(VideoCore::GuestDriverProfile* gpu_driver, - std::list& used_samplers) { + const std::list& used_samplers) { if (gpu_driver == nullptr) { - LOG_CRITICAL(HW_GPU, "GPU Driver profile has not been created yet"); + LOG_CRITICAL(HW_GPU, "GPU driver profile has not been created yet"); return; } if (gpu_driver->TextureHandlerSizeKnown() || used_samplers.size() <= 1) { @@ -57,9 +57,9 @@ void DeduceTextureHandlerSize(VideoCore::GuestDriverProfile* gpu_driver, } } -std::optional TryDeduceSamplerSize(Sampler& sampler_to_deduce, +std::optional TryDeduceSamplerSize(const Sampler& sampler_to_deduce, VideoCore::GuestDriverProfile* gpu_driver, - std::list& used_samplers) { + const std::list& used_samplers) { if (gpu_driver == nullptr) { LOG_CRITICAL(HW_GPU, "GPU Driver profile has not been created yet"); return std::nullopt; @@ -367,17 +367,18 @@ void ShaderIR::PostDecode() { auto gpu_driver = locker.AccessGuestDriverProfile(); DeduceTextureHandlerSize(gpu_driver, used_samplers); // Deduce Indexed Samplers - if (uses_indexed_samplers) { - for (auto& sampler : used_samplers) { - if (sampler.IsIndexed()) { - auto size = TryDeduceSamplerSize(sampler, gpu_driver, used_samplers); - if (size) { - sampler.SetSize(*size); - } else { - LOG_CRITICAL(HW_GPU, "Failed to deduce size of indexed sampler"); - sampler.SetSize(1); - } - } + if (!uses_indexed_samplers) { + return; + } + for (auto& sampler : used_samplers) { + if (!sampler.IsIndexed()) { + continue; + } + if (const auto size = TryDeduceSamplerSize(sampler, gpu_driver, used_samplers)) { + sampler.SetSize(*size); + } else { + LOG_CRITICAL(HW_GPU, "Failed to deduce size of indexed sampler"); + sampler.SetSize(1); } } } diff --git a/src/video_core/shader/decode/texture.cpp b/src/video_core/shader/decode/texture.cpp index 6da9668fe..d980535b1 100644 --- a/src/video_core/shader/decode/texture.cpp +++ b/src/video_core/shader/decode/texture.cpp @@ -201,7 +201,8 @@ u32 ShaderIR::DecodeTexture(NodeBlock& bb, u32 pc) { } for (u32 element = 0; element < values.size(); ++element) { - MetaTexture meta{*sampler, array_node, {}, {}, {}, derivates, {}, {}, {}, element, index_var}; + MetaTexture meta{*sampler, array_node, {}, {}, {}, derivates, + {}, {}, {}, element, index_var}; values[element] = Operation(OperationCode::TextureGradient, std::move(meta), coords); } diff --git a/src/video_core/shader/node.h b/src/video_core/shader/node.h index d75453458..53a551d27 100644 --- a/src/video_core/shader/node.h +++ b/src/video_core/shader/node.h @@ -291,7 +291,7 @@ public: return size; } - void SetSize(u32 new_size) { + constexpr void SetSize(u32 new_size) { size = new_size; } @@ -315,15 +315,15 @@ public: explicit ArraySamplerNode(u32 index, u32 base_offset, u32 bindless_var) : index{index}, base_offset{base_offset}, bindless_var{bindless_var} {} - u32 GetIndex() const { + constexpr u32 GetIndex() const { return index; } - u32 GetBaseOffset() const { + constexpr u32 GetBaseOffset() const { return base_offset; } - u32 GetIndexVar() const { + constexpr u32 GetIndexVar() const { return bindless_var; } @@ -338,11 +338,11 @@ class BindlessSamplerNode final { public: explicit BindlessSamplerNode(u32 index, u32 offset) : index{index}, offset{offset} {} - u32 GetIndex() const { + constexpr u32 GetIndex() const { return index; } - u32 GetOffset() const { + constexpr u32 GetOffset() const { return offset; } @@ -557,7 +557,7 @@ class CustomVarNode final { public: explicit constexpr CustomVarNode(u32 index) : index{index} {} - u32 GetIndex() const { + constexpr u32 GetIndex() const { return index; } diff --git a/src/video_core/shader/shader_ir.cpp b/src/video_core/shader/shader_ir.cpp index 94972d57f..3a5d280a9 100644 --- a/src/video_core/shader/shader_ir.cpp +++ b/src/video_core/shader/shader_ir.cpp @@ -458,8 +458,7 @@ std::size_t ShaderIR::DeclareAmend(Node new_amend) { } u32 ShaderIR::NewCustomVariable() { - const u32 id = num_custom_variables++; - return id; + return num_custom_variables++; } } // namespace VideoCommon::Shader diff --git a/src/video_core/shader/shader_ir.h b/src/video_core/shader/shader_ir.h index 43672b41c..b0851c3be 100644 --- a/src/video_core/shader/shader_ir.h +++ b/src/video_core/shader/shader_ir.h @@ -180,7 +180,7 @@ public: return amend_code[index]; } - u32 GetCustomVariablesAmount() const { + u32 GetNumCustomVariables() const { return num_custom_variables; } diff --git a/src/video_core/shader/track.cpp b/src/video_core/shader/track.cpp index 4db721f69..ea39bca54 100644 --- a/src/video_core/shader/track.cpp +++ b/src/video_core/shader/track.cpp @@ -36,7 +36,6 @@ std::pair FindOperation(const NodeBlock& code, s64 cursor, } return {}; } -} // Anonymous namespace std::optional> DecoupleIndirectRead(const OperationNode& operation) { if (operation.GetCode() != OperationCode::UAdd) { @@ -44,9 +43,7 @@ std::optional> DecoupleIndirectRead(const OperationNode& o } Node gpr{}; Node offset{}; - if (operation.GetOperandsCount() != 2) { - return std::nullopt; - } + ASSERT(operation.GetOperandsCount() == 2); for (std::size_t i = 0; i < operation.GetOperandsCount(); i++) { Node operand = operation[i]; if (std::holds_alternative(*operand)) { @@ -56,7 +53,7 @@ std::optional> DecoupleIndirectRead(const OperationNode& o } } if (offset && gpr) { - return {std::make_pair(gpr, offset)}; + return std::make_pair(gpr, offset); } return std::nullopt; } @@ -72,6 +69,8 @@ bool AmendNodeCv(std::size_t amend_index, Node node) { return false; } +} // Anonymous namespace + std::tuple ShaderIR::TrackBindlessSampler(Node tracked, const NodeBlock& code, s64 cursor) { if (const auto cbuf = std::get_if(&*tracked)) { -- cgit v1.2.3