CPU features detection fix

ilyakurdyukov · May 17, 2020 · 57cfb77 · 57cfb77
1 parent 02c3e54
commit 57cfb77
Show file tree

Hide file tree

Showing 5 changed files with 17 additions and 13 deletions.
diff --git a/README.md b/README.md
@@ -26,7 +26,11 @@ Without multithreading and SIMD optimizations it runs slower than native code.
 `-t, --threads n` Set the number of CPU threads to use  
 `-o, --optimize` Option for libjpeg to produce smaller output file  
 `-v, --verbose n` Print libjpeg debug output  
-`-i, --info n` Print quantsmooth debug output: 0 - silent, 8 - processing time, 15 - all (default)  
+`-i, --info n` Print quantsmooth debug output (default is 15)  
+Use the sum of flags: 0 - silent, 1/2/4 - various information, 8 - processing time, 16 - SIMD type.
+`-p, --cpu n` Use to lower the SIMD type if CPU detection fails:  
+0 - auto, 1 - scalar, 2 - SSE2, 3 - AVX2, 4 - AVX512.
+(`x86` build selects between modes 1-3, `x86_64` from 2-4)
 
 - The processing time includes only the smoothing algorithm, jpeg reading and writing time is not included.
 - More iterations can make the result look like CG art, can make the photos look unnatural.

diff --git a/libjpegqs.c b/libjpegqs.c
@@ -67,17 +67,19 @@ JPEGQS_ATTR int do_quantsmooth QS_ARGS {
 		get_cpuid(0, 0, cpuid); m = cpuid[0];
 		if (m < 1) break;
 		get_cpuid(1, 0, cpuid);
+		if (!(cpuid[3] & (1 << 25))) break; // SSE
 		if (!(cpuid[3] & (1 << 26))) break; // SSE2
 		type = 2;
 		// VirtualBox clears FMA, even if AVX2 is set
 		// if (!(cpuid[2] & (1 << 12))) break; // FMA
 		if (!(cpuid[2] & (1 << 27))) break; // OSXSAVE
+		if (!(cpuid[2] & (1 << 28))) break; // AVX
 		xcr0 = ~xgetbv(0);
 		if (m < 7) break;
 		get_cpuid(7, 0, cpuid);
-		if (!(cpuid[1] & (1 << 5)) && xcr0 & 6) break; // AVX2
+		if (!(cpuid[1] & (1 << 5)) || xcr0 & 6) break; // AVX2
 		type = 3;
-		if (!(cpuid[1] & (1 << 16)) && xcr0 & 0xe6) break; // AVX512F
+		if (!(cpuid[1] & (1 << 16)) || xcr0 & 0xe6) break; // AVX512F
 		type = 4;
 	} while (0);
 

diff --git a/libjpegqs.h b/libjpegqs.h
@@ -35,7 +35,7 @@ enum {
 #define JPEGQS_ATTR
 #endif
 
-#define JPEGQS_VERSION "2020-03-23"
+#define JPEGQS_VERSION "2020-05-17"
 #define JPEGQS_COPYRIGHT "Copyright (C) 2020 Ilya Kurdyukov"
 
 typedef struct {

diff --git a/quantsmooth.c b/quantsmooth.c
@@ -436,8 +436,13 @@ int main(int argc, char **argv) {
 "  -t, --threads n   Set the number of CPU threads to use\n"
 "  -o, --optimize    Option for libjpeg to produce smaller output file\n"
 "  -v, --verbose n   Print libjpeg debug output\n"
-"  -i, --info n      Print quantsmooth debug output:\n"
-"                      0 - silent, 8 - processing time, 15 - all (default)\n"
+"  -i, --info n      Print quantsmooth debug output (default is 15)\n"
+"                      Use the sum of flags: 0 - silent,\n"
+"                      1/2/4 - various information,\n"
+"                      8 - processing time, 16 - SIMD type.\n"
+"  -p, --cpu n       Use to lower the SIMD type if CPU detection fails:\n"
+"                      0 - auto, 1 - scalar, 2 - SSE2, 3 - AVX2, 4 - AVX512.\n"
+"                      (x86 build selects between modes 1-3, x86_64 from 2-4)\n"
 "\n", progname);
 		return 1;
 	}

diff --git a/quantsmooth.h b/quantsmooth.h
@@ -30,13 +30,6 @@
 #define omp_get_thread_num() 0
 #endif
 
-#if defined(__clang__) && defined(_OPENMP)
-// Clang (9.0.0) shows this warning about OpenMP loop counter variables,
-// but there is no warnings if OpenMP is disabled,
-// so it look like a bug in the compiler.
-#pragma GCC diagnostic ignored "-Wuninitialized"
-#endif
-
 #if !defined(TRANSCODE_ONLY) && !defined(JPEG_INTERNALS)
 // declarations needed from jpegint.h