Fix CPU Core Count detection and Enable Parallel Shader Compilation

This does this following things:

- Default to the runtime automatic number of threads for pre-compiling shaders
- Adds a distinct automatic thread count computation for pre-compilation  (which has less other things going on
and should scale better beyond 4 cores)
- Removes the unused logical_core_count field from the CPU detection
- Changes the semantics of num_cores from maximaum addressable number of cores to actually available CPU cores
(which is also how it was actually used)
- Updates the computation of the HTT flag now that AMD no longer lies about it for its Zen processors
- Background shader compilation is *not* enabled by default

Source/Core/Common/CPUDetect.h

@@ -26,7 +26,6 @@ struct CPUInfo
 
   bool HTT = false;
   int num_cores = 0;
-  int logical_cpu_count = 0;
 
   bool bSSE = false;
   bool bSSE2 = false;

Source/Core/Common/x64CPUDetect.cpp

@@ -3,6 +3,7 @@
 
 #include <cstring>
 #include <string>
+#include <thread>
 
 #include "Common/CPUDetect.h"
 #include "Common/CommonTypes.h"
@@ -106,7 +107,6 @@ void CPUInfo::Detect()
   // Detect family and other misc stuff.
   bool ht = false;
   HTT = ht;
-  logical_cpu_count = 1;
   if (max_std_fn >= 1)
   {
     __cpuid(cpu_id, 0x00000001);
@@ -120,9 +120,13 @@ void CPUInfo::Detect()
     // Detect AMD Zen1, Zen1+ and Zen2
     if (family == 23)
       bZen1p2 = true;
-    logical_cpu_count = (cpu_id[1] >> 16) & 0xFF;
     ht = (cpu_id[3] >> 28) & 1;
 
+    // AMD CPUs before Zen faked this flag and didn't actually
+    // implement simultaneous multithreading (SMT; Intel calls it HTT)
+    // but rather some weird middle-ground between 1-2 cores
+    HTT = ht && (vendor == CPUVendor::Intel || family >= 23);
+
     if ((cpu_id[3] >> 25) & 1)
       bSSE = true;
     if ((cpu_id[3] >> 26) & 1)
@@ -200,35 +204,10 @@ void CPUInfo::Detect()
       bLongMode = true;
   }
 
-  num_cores = (logical_cpu_count == 0) ? 1 : logical_cpu_count;
+  // this should be much more reliable and easier
-
+  // than trying to get the number of cores out of the CPUID data
-  if (max_ex_fn >= 0x80000008)
+  // ourselves
-  {
+  num_cores = std::max(std::thread::hardware_concurrency(), 1u);
-    // Get number of cores. This is a bit complicated. Following AMD manual here.
-    __cpuid(cpu_id, 0x80000008);
-    int apic_id_core_id_size = (cpu_id[2] >> 12) & 0xF;
-    if (apic_id_core_id_size == 0)
-    {
-      if (ht)
-      {
-        // New mechanism for modern Intel CPUs.
-        if (vendor == CPUVendor::Intel)
-        {
-          __cpuidex(cpu_id, 0x00000004, 0x00000000);
-          int cores_x_package = ((cpu_id[0] >> 26) & 0x3F) + 1;
-          HTT = (cores_x_package < logical_cpu_count);
-          cores_x_package = ((logical_cpu_count % cores_x_package) == 0) ? cores_x_package : 1;
-          num_cores = (cores_x_package > 1) ? cores_x_package : num_cores;
-          logical_cpu_count /= cores_x_package;
-        }
-      }
-    }
-    else
-    {
-      // Use AMD's new method.
-      num_cores = (cpu_id[2] & 0xFF) + 1;
-    }
-  }
 }
 
 // Turn the CPU info into a string we can show

Source/Core/Core/Config/GraphicsSettings.cpp

@@ -85,7 +85,7 @@ const Info<ShaderCompilationMode> GFX_SHADER_COMPILATION_MODE{
     {System::GFX, "Settings", "ShaderCompilationMode"}, ShaderCompilationMode::Synchronous};
 const Info<int> GFX_SHADER_COMPILER_THREADS{{System::GFX, "Settings", "ShaderCompilerThreads"}, 1};
 const Info<int> GFX_SHADER_PRECOMPILER_THREADS{
-    {System::GFX, "Settings", "ShaderPrecompilerThreads"}, 1};
+    {System::GFX, "Settings", "ShaderPrecompilerThreads"}, -1};
 const Info<bool> GFX_SAVE_TEXTURE_CACHE_TO_STATE{
     {System::GFX, "Settings", "SaveTextureCacheToState"}, true};
 

Source/Core/VideoCommon/VideoConfig.cpp

@@ -176,6 +176,14 @@ static u32 GetNumAutoShaderCompilerThreads()
   return static_cast<u32>(std::min(std::max(cpu_info.num_cores - 3, 1), 4));
 }
 
+static u32 GetNumAutoShaderPreCompilerThreads()
+{
+  // Automatic number. We use clamp(cpus - 2, 1, infty) here.
+  // We chose this because we don't want to limit our speed-up
+  // and at the same time leave two logical cores for the dolphin UI and the rest of the OS.
+  return static_cast<u32>(std::max(cpu_info.num_cores - 2, 1));
+}
+
 u32 VideoConfig::GetShaderCompilerThreads() const
 {
   if (!backend_info.bSupportsBackgroundCompiling)
@@ -199,5 +207,5 @@ u32 VideoConfig::GetShaderPrecompilerThreads() const
   if (iShaderPrecompilerThreads >= 0)
     return static_cast<u32>(iShaderPrecompilerThreads);
   else
-    return GetNumAutoShaderCompilerThreads();
+    return GetNumAutoShaderPreCompilerThreads();
 }