diff options
Diffstat (limited to 'src/video_core')
-rw-r--r-- | src/video_core/engines/shader_bytecode.h | 2 | ||||
-rw-r--r-- | src/video_core/shader/async_shaders.cpp | 17 | ||||
-rw-r--r-- | src/video_core/shader/decode/arithmetic.cpp | 3 |
3 files changed, 13 insertions, 9 deletions
diff --git a/src/video_core/engines/shader_bytecode.h b/src/video_core/engines/shader_bytecode.h index d374b73cf..a3c05d1b0 100644 --- a/src/video_core/engines/shader_bytecode.h +++ b/src/video_core/engines/shader_bytecode.h @@ -1893,6 +1893,7 @@ public: ICMP_IMM, FCMP_RR, FCMP_RC, + FCMP_IMMR, MUFU, // Multi-Function Operator RRO_C, // Range Reduction Operator RRO_R, @@ -2205,6 +2206,7 @@ private: INST("0111110-0-------", Id::HSET2_IMM, Type::HalfSet, "HSET2_IMM"), INST("010110111010----", Id::FCMP_RR, Type::Arithmetic, "FCMP_RR"), INST("010010111010----", Id::FCMP_RC, Type::Arithmetic, "FCMP_RC"), + INST("0011011-1010----", Id::FCMP_IMMR, Type::Arithmetic, "FCMP_IMMR"), INST("0101000010000---", Id::MUFU, Type::Arithmetic, "MUFU"), INST("0100110010010---", Id::RRO_C, Type::Arithmetic, "RRO_C"), INST("0101110010010---", Id::RRO_R, Type::Arithmetic, "RRO_R"), diff --git a/src/video_core/shader/async_shaders.cpp b/src/video_core/shader/async_shaders.cpp index aabd62c5c..39cc3b869 100644 --- a/src/video_core/shader/async_shaders.cpp +++ b/src/video_core/shader/async_shaders.cpp @@ -20,14 +20,15 @@ AsyncShaders::~AsyncShaders() { } void AsyncShaders::AllocateWorkers() { - // Max worker threads we should allow - constexpr u32 MAX_THREADS = 4; - // Deduce how many threads we can use - const u32 threads_used = std::thread::hardware_concurrency() / 4; - // Always allow at least 1 thread regardless of our settings - const auto max_worker_count = std::max(1U, threads_used); - // Don't use more than MAX_THREADS - const auto num_workers = std::min(max_worker_count, MAX_THREADS); + // Use at least one thread + u32 num_workers = 1; + + // Deduce how many more threads we can use + const u32 thread_count = std::thread::hardware_concurrency(); + if (thread_count >= 8) { + // Increase async workers by 1 for every 2 threads >= 8 + num_workers += 1 + (thread_count - 8) / 2; + } // If we already have workers queued, ignore if (num_workers == worker_threads.size()) { diff --git a/src/video_core/shader/decode/arithmetic.cpp b/src/video_core/shader/decode/arithmetic.cpp index 4db329fa5..afef5948d 100644 --- a/src/video_core/shader/decode/arithmetic.cpp +++ b/src/video_core/shader/decode/arithmetic.cpp @@ -137,7 +137,8 @@ u32 ShaderIR::DecodeArithmetic(NodeBlock& bb, u32 pc) { break; } case OpCode::Id::FCMP_RR: - case OpCode::Id::FCMP_RC: { + case OpCode::Id::FCMP_RC: + case OpCode::Id::FCMP_IMMR: { UNIMPLEMENTED_IF(instr.fcmp.ftz == 0); Node op_c = GetRegister(instr.gpr39); Node comp = GetPredicateComparisonFloat(instr.fcmp.cond, std::move(op_c), Immediate(0.0f)); |