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') 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