Improved CPU detection code.

src/global.cpp

@@ -412,6 +412,14 @@ bool x264_is_prerelease(void)
 	return (VER_X264_PRE_RELEASE);
 }
 
+/*
+ * CPUID prototype (actual function is in ASM code)
+ */
+extern "C"
+{
+	void x264_cpu_cpuid(unsigned int op, unsigned int *eax, unsigned int *ebx, unsigned int *ecx, unsigned int *edx);
+}
+
 /*
  * Detect CPU features
  */
@@ -425,7 +433,7 @@ x264_cpu_t x264_detect_cpu_features(int argc, char **argv)
 
 	x264_cpu_t features;
 	SYSTEM_INFO systemInfo;
-	int CPUInfo[4] = {-1};
+	unsigned int CPUInfo[4];
 	char CPUIdentificationString[0x40];
 	char CPUBrandString[0x40];
 
@@ -434,42 +442,49 @@ x264_cpu_t x264_detect_cpu_features(int argc, char **argv)
 	memset(CPUIdentificationString, 0, sizeof(CPUIdentificationString));
 	memset(CPUBrandString, 0, sizeof(CPUBrandString));
 	
-	__cpuid(CPUInfo, 0);
+	x264_cpu_cpuid(0, &CPUInfo[0], &CPUInfo[1], &CPUInfo[2], &CPUInfo[3]);
-	memcpy(CPUIdentificationString, &CPUInfo[1], sizeof(int));
+	memcpy(CPUIdentificationString, &CPUInfo[1], 4);
-	memcpy(CPUIdentificationString + 4, &CPUInfo[3], sizeof(int));
+	memcpy(CPUIdentificationString + 4, &CPUInfo[3], 4);
-	memcpy(CPUIdentificationString + 8, &CPUInfo[2], sizeof(int));
+	memcpy(CPUIdentificationString + 8, &CPUInfo[2], 4);
 	features.intel = (_stricmp(CPUIdentificationString, "GenuineIntel") == 0);
 	strncpy_s(features.vendor, 0x40, CPUIdentificationString, _TRUNCATE);
 
 	if(CPUInfo[0] >= 1)
 	{
-		__cpuid(CPUInfo, 1);
+		x264_cpu_cpuid(1, &CPUInfo[0], &CPUInfo[1], &CPUInfo[2], &CPUInfo[3]);
-		features.mmx = (CPUInfo[3] & 0x800000) || false;
+		features.mmx = (CPUInfo[3] & 0x800000U) || false;
-		features.sse = (CPUInfo[3] & 0x2000000) || false;
+		features.sse = (CPUInfo[3] & 0x2000000U) || false;
-		features.sse2 = (CPUInfo[3] & 0x4000000) || false;
+		features.sse2 = (CPUInfo[3] & 0x4000000U) || false;
-		features.ssse3 = (CPUInfo[2] & 0x200) || false;
+		features.ssse3 = (CPUInfo[2] & 0x200U) || false;
-		features.sse3 = (CPUInfo[2] & 0x1) || false;
+		features.sse3 = (CPUInfo[2] & 0x1U) || false;
-		features.ssse3 = (CPUInfo[2] & 0x200) || false;
+		features.ssse3 = (CPUInfo[2] & 0x200U) || false;
 		features.stepping = CPUInfo[0] & 0xf;
 		features.model = ((CPUInfo[0] >> 4) & 0xf) + (((CPUInfo[0] >> 16) & 0xf) << 4);
 		features.family = ((CPUInfo[0] >> 8) & 0xf) + ((CPUInfo[0] >> 20) & 0xff);
+		if(features.sse) features.mmx2 = true; //MMXEXT is a subset of SSE!
 	}
 
-	__cpuid(CPUInfo, 0x80000000);
+	x264_cpu_cpuid(0x80000000U, &CPUInfo[0], &CPUInfo[1], &CPUInfo[2], &CPUInfo[3]);
-	int nExIds = qMax<int>(qMin<int>(CPUInfo[0], 0x80000004), 0x80000000);
+	unsigned int nExIds = qBound(0x80000000U, CPUInfo[0], 0x80000004U);
 
-	for(int i = 0x80000002; i <= nExIds; ++i)
+	if((_stricmp(CPUIdentificationString, "AuthenticAMD") == 0) && (nExIds >= 0x80000001U))
 	{
-		__cpuid(CPUInfo, i);
+		x264_cpu_cpuid(0x80000001U, &CPUInfo[0], &CPUInfo[1], &CPUInfo[2], &CPUInfo[3]);
+		features.mmx2 = features.mmx2 || (CPUInfo[3] & 0x00400000U);
+	}
+
+	for(unsigned int i = 0x80000002U; i <= nExIds; ++i)
+	{
+		x264_cpu_cpuid(i, &CPUInfo[0], &CPUInfo[1], &CPUInfo[2], &CPUInfo[3]);
 		switch(i)
 		{
-		case 0x80000002:
+		case 0x80000002U:
 			memcpy(CPUBrandString, CPUInfo, sizeof(CPUInfo));
 			break;
-		case 0x80000003:
+		case 0x80000003U:
 			memcpy(CPUBrandString + 16, CPUInfo, sizeof(CPUInfo));
 			break;
-		case 0x80000004:
+		case 0x80000004U:
 			memcpy(CPUBrandString + 32, CPUInfo, sizeof(CPUInfo));
 			break;
 		}
@@ -517,11 +532,13 @@ x264_cpu_t x264_detect_cpu_features(int argc, char **argv)
 		{
 			if(!_stricmp("--force-cpu-no-64bit", argv[i])) { flag = true; features.x64 = false; }
 			if(!_stricmp("--force-cpu-no-mmx", argv[i])) { flag = true; features.mmx = false; }
+			if(!_stricmp("--force-cpu-no-mmx2", argv[i])) { flag = true; features.mmx2 = false; }
 			if(!_stricmp("--force-cpu-no-sse", argv[i])) { flag = true; features.sse = features.sse2 = features.sse3 = features.ssse3 = false; }
 			if(!_stricmp("--force-cpu-no-intel", argv[i])) { flag = true; features.intel = false; }
 			
 			if(!_stricmp("--force-cpu-have-64bit", argv[i])) { flag = true; features.x64 = true; }
 			if(!_stricmp("--force-cpu-have-mmx", argv[i])) { flag = true; features.mmx = true; }
+			if(!_stricmp("--force-cpu-have-mmx2", argv[i])) { flag = true; features.mmx2 = true; }
 			if(!_stricmp("--force-cpu-have-sse", argv[i])) { flag = true; features.sse = features.sse2 = features.sse3 = features.ssse3 = true; }
 			if(!_stricmp("--force-cpu-have-intel", argv[i])) { flag = true; features.intel = true; }
 		}

src/global.h

@@ -78,6 +78,7 @@ typedef struct
 	int count;
 	bool x64;
 	bool mmx;
+	bool mmx2;
 	bool sse;
 	bool sse2;
 	bool sse3;

src/main.cpp

@@ -59,13 +59,22 @@ static int x264_main(int argc, char* argv[])
 	qDebug("   CPU vendor id  :  %s (Intel: %s)", cpuFeatures.vendor, X264_BOOL(cpuFeatures.intel));
 	qDebug("CPU brand string  :  %s", cpuFeatures.brand);
 	qDebug("   CPU signature  :  Family: %d, Model: %d, Stepping: %d", cpuFeatures.family, cpuFeatures.model, cpuFeatures.stepping);
-	qDebug("CPU capabilities  :  MMX: %s, SSE: %s, SSE2: %s, SSE3: %s, SSSE3: %s, x64: %s", X264_BOOL(cpuFeatures.mmx), X264_BOOL(cpuFeatures.sse), X264_BOOL(cpuFeatures.sse2), X264_BOOL(cpuFeatures.sse3), X264_BOOL(cpuFeatures.ssse3), X264_BOOL(cpuFeatures.x64));
+	qDebug("CPU capabilities  :  MMX=%s, MMXEXT=%s, SSE=%s, SSE2=%s, SSE3=%s, SSSE3=%s, X64=%s", X264_BOOL(cpuFeatures.mmx), X264_BOOL(cpuFeatures.mmx2), X264_BOOL(cpuFeatures.sse), X264_BOOL(cpuFeatures.sse2), X264_BOOL(cpuFeatures.sse3), X264_BOOL(cpuFeatures.ssse3), X264_BOOL(cpuFeatures.x64));
 	qDebug(" Number of CPU's  :  %d\n", cpuFeatures.count);
 	
-	//Make sure this CPU can run x264
+	//Make sure this CPU can run x264 (requires MMX + MMXEXT/iSSE to run x264 with ASM enabled, additionally requires SSE1 for most x264 builds)
-	if(!(cpuFeatures.mmx && cpuFeatures.sse))
+	if(!(cpuFeatures.mmx && cpuFeatures.mmx2))
 	{
-		qFatal("Sorry, but this machine is not physically capable of running x264. Please get a CPU that supports at least the MMX and ISSE instruction sets!");
+		qFatal("Sorry, but this machine is not physically capable of running x264. Please get a CPU that supports at least the MMX and MMXEXT instruction sets!");
+	}
+	else if(!(cpuFeatures.mmx && cpuFeatures.sse))
+	{
+		qWarning("WARNING: System does not support SSE1, most x264 builds will not work !!!\n");
+		for(;;)
+		{
+			int ret = MessageBoxW(NULL, L"BIG FAT WARNING: This machine apparently does NOT support the SSE1 instruction set and thus probably will NOT be able to run x264!", L"Simple x264 Launcher", MB_ABORTRETRYIGNORE|MB_TOPMOST|MB_ICONEXCLAMATION);
+			if(ret == IDIGNORE) break; else if(ret == IDRETRY) continue; else return -1;
+		}
 	}
 
 	//Initialize Qt

x264_launcher.vcxproj

@@ -260,6 +260,9 @@ copy "$(SolutionDir)res\toolset\*.exe" "$(SolutionDir)bin\$(Configuration)\tools
   <ItemGroup>
     <ResourceCompile Include="x264_launcher.rc" />
   </ItemGroup>
+  <ItemGroup>
+    <Object Include="asm\cpu-detect.obj" />
+  </ItemGroup>
   <Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
   <ImportGroup Label="ExtensionTargets">
   </ImportGroup>

x264_launcher.vcxproj.filters

@@ -19,6 +19,9 @@
     <Filter Include="Generated Files">
       <UniqueIdentifier>{961e9f99-8107-45a2-984d-188819a67e8e}</UniqueIdentifier>
     </Filter>
+    <Filter Include="Assembly">
+      <UniqueIdentifier>{2c9c8836-2259-4412-894e-7cc32bef99de}</UniqueIdentifier>
+    </Filter>
   </ItemGroup>
   <ItemGroup>
     <None Include="ReadMe.txt" />
@@ -142,4 +145,9 @@
       <Filter>Resource Files</Filter>
     </ResourceCompile>
   </ItemGroup>
+  <ItemGroup>
+    <Object Include="asm\cpu-detect.obj">
+      <Filter>Assembly</Filter>
+    </Object>
+  </ItemGroup>
 </Project>