avutil: add AVX-512 flags

diff --git a/Changelog b/Changelog
index ee488761281188927552b922afd9c9b3335249ee..f44cec679f4937c3baf84d415d8c803476f40bf8 100644 (file)
--- a/Changelog
+++ b/Changelog
@@ -27,6 +27,7 @@ version <next>:
 - video setrange filter
 - nsp demuxer
 - support LibreSSL (via libtls)
+- AVX-512/ZMM support added
 
 
 version 3.4:

diff --git a/doc/APIchanges b/doc/APIchanges
index da444ffb7c0d9e0cb2dffb017f9d275b7040c06b..df79758e863f3ed11d9a87399d284eaba1298a42 100644 (file)
--- a/doc/APIchanges
+++ b/doc/APIchanges
@@ -15,6 +15,9 @@ libavutil:     2017-10-21
 
 API changes, most recent first:
 
+2017-12-xx - xxxxxxx - lavu 56.7.100 - cpu.h
+  AVX-512 flags added.
+
 2017-xx-xx - xxxxxxx - lavc 58.8.100 - avcodec.h
   The MediaCodec decoders now support AVCodecContext.hw_device_ctx.
 

diff --git a/libavutil/cpu.c b/libavutil/cpu.c
index c8401b825809b831ac26d024ac4876131bc6a7ce..6548cc3042aa85e3d49bfda2bea0e1ff1e1c96b0 100644 (file)
--- a/libavutil/cpu.c
+++ b/libavutil/cpu.c
@@ -80,7 +80,8 @@ void av_force_cpu_flags(int arg){
                     AV_CPU_FLAG_XOP      |
                     AV_CPU_FLAG_FMA3     |
                     AV_CPU_FLAG_FMA4     |
-                    AV_CPU_FLAG_AVX2     ))
+                    AV_CPU_FLAG_AVX2     |
+                    AV_CPU_FLAG_AVX512   ))
         && !(arg & AV_CPU_FLAG_MMX)) {
         av_log(NULL, AV_LOG_WARNING, "MMX implied by specified flags\n");
         arg |= AV_CPU_FLAG_MMX;
@@ -126,6 +127,7 @@ int av_parse_cpu_flags(const char *s)
 #define CPUFLAG_AVX2     (AV_CPU_FLAG_AVX2     | CPUFLAG_AVX)
 #define CPUFLAG_BMI2     (AV_CPU_FLAG_BMI2     | AV_CPU_FLAG_BMI1)
 #define CPUFLAG_AESNI    (AV_CPU_FLAG_AESNI    | CPUFLAG_SSE42)
+#define CPUFLAG_AVX512   (AV_CPU_FLAG_AVX512   | CPUFLAG_AVX2)
     static const AVOption cpuflags_opts[] = {
         { "flags"   , NULL, 0, AV_OPT_TYPE_FLAGS, { .i64 = 0 }, INT64_MIN, INT64_MAX, .unit = "flags" },
 #if   ARCH_PPC
@@ -154,6 +156,7 @@ int av_parse_cpu_flags(const char *s)
         { "3dnowext", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = CPUFLAG_3DNOWEXT     },    .unit = "flags" },
         { "cmov",     NULL, 0, AV_OPT_TYPE_CONST, { .i64 = AV_CPU_FLAG_CMOV     },    .unit = "flags" },
         { "aesni"   , NULL, 0, AV_OPT_TYPE_CONST, { .i64 = CPUFLAG_AESNI        },    .unit = "flags" },
+        { "avx512"  , NULL, 0, AV_OPT_TYPE_CONST, { .i64 = CPUFLAG_AVX512       },    .unit = "flags" },
 #elif ARCH_ARM
         { "armv5te",  NULL, 0, AV_OPT_TYPE_CONST, { .i64 = AV_CPU_FLAG_ARMV5TE  },    .unit = "flags" },
         { "armv6",    NULL, 0, AV_OPT_TYPE_CONST, { .i64 = AV_CPU_FLAG_ARMV6    },    .unit = "flags" },
@@ -216,6 +219,7 @@ int av_parse_cpu_caps(unsigned *flags, const char *s)
         { "3dnowext", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = AV_CPU_FLAG_3DNOWEXT },    .unit = "flags" },
         { "cmov",     NULL, 0, AV_OPT_TYPE_CONST, { .i64 = AV_CPU_FLAG_CMOV     },    .unit = "flags" },
         { "aesni",    NULL, 0, AV_OPT_TYPE_CONST, { .i64 = AV_CPU_FLAG_AESNI    },    .unit = "flags" },
+        { "avx512"  , NULL, 0, AV_OPT_TYPE_CONST, { .i64 = AV_CPU_FLAG_AVX512   },    .unit = "flags" },
 
 #define CPU_FLAG_P2 AV_CPU_FLAG_CMOV | AV_CPU_FLAG_MMX
 #define CPU_FLAG_P3 CPU_FLAG_P2 | AV_CPU_FLAG_MMX2 | AV_CPU_FLAG_SSE

diff --git a/libavutil/cpu.h b/libavutil/cpu.h
index 9e5d40affef65f12b465568cea13d14b201940c9..8bb9eb606bf2a2e30b879a2751764be1b12ce4f0 100644 (file)
--- a/libavutil/cpu.h
+++ b/libavutil/cpu.h
@@ -55,6 +55,7 @@
 #define AV_CPU_FLAG_FMA3        0x10000 ///< Haswell FMA3 functions
 #define AV_CPU_FLAG_BMI1        0x20000 ///< Bit Manipulation Instruction Set 1
 #define AV_CPU_FLAG_BMI2        0x40000 ///< Bit Manipulation Instruction Set 2
+#define AV_CPU_FLAG_AVX512     0x100000 ///< AVX-512 functions: requires OS support even if YMM/ZMM registers aren't used
 
 #define AV_CPU_FLAG_ALTIVEC      0x0001 ///< standard
 #define AV_CPU_FLAG_VSX          0x0002 ///< ISA 2.06

diff --git a/libavutil/tests/cpu.c b/libavutil/tests/cpu.c
index f02a54cbbb1469412a9dbbf5fa43179b3140f0dd..ce45b715a0d77e1e7fc4e90a69070f8f863a0d5a 100644 (file)
--- a/libavutil/tests/cpu.c
+++ b/libavutil/tests/cpu.c
@@ -73,6 +73,7 @@ static const struct {
     { AV_CPU_FLAG_BMI1,      "bmi1"       },
     { AV_CPU_FLAG_BMI2,      "bmi2"       },
     { AV_CPU_FLAG_AESNI,     "aesni"      },
+    { AV_CPU_FLAG_AVX512,    "avx512"     },
 #endif
     { 0 }
 };

diff --git a/libavutil/version.h b/libavutil/version.h
index 9ae9768d4d80fbe8f4a5300988840667aaf3563d..d81ec6fa7bf9eee07269b3a952f55c108703ca18 100644 (file)
--- a/libavutil/version.h
+++ b/libavutil/version.h
@@ -79,7 +79,7 @@
  */
 
 #define LIBAVUTIL_VERSION_MAJOR  56
-#define LIBAVUTIL_VERSION_MINOR   6
+#define LIBAVUTIL_VERSION_MINOR   7
 #define LIBAVUTIL_VERSION_MICRO 100
 
 #define LIBAVUTIL_VERSION_INT   AV_VERSION_INT(LIBAVUTIL_VERSION_MAJOR, \

diff --git a/libavutil/x86/cpu.h b/libavutil/x86/cpu.h
index 309b8e746c56e0b20a4f63e537673629c20042b5..7f4e5d08bb249329b53d013fb45b43280dcf44b2 100644 (file)
--- a/libavutil/x86/cpu.h
+++ b/libavutil/x86/cpu.h
@@ -50,6 +50,7 @@
 #define X86_FMA4(flags)             CPUEXT(flags, FMA4)
 #define X86_AVX2(flags)             CPUEXT(flags, AVX2)
 #define X86_AESNI(flags)            CPUEXT(flags, AESNI)
+#define X86_AVX512(flags)           CPUEXT(flags, AVX512)
 
 #define EXTERNAL_AMD3DNOW(flags)    CPUEXT_SUFFIX(flags, _EXTERNAL, AMD3DNOW)
 #define EXTERNAL_AMD3DNOWEXT(flags) CPUEXT_SUFFIX(flags, _EXTERNAL, AMD3DNOWEXT)
@@ -79,6 +80,7 @@
 #define EXTERNAL_AVX2_FAST(flags)   CPUEXT_SUFFIX_FAST2(flags, _EXTERNAL, AVX2, AVX)
 #define EXTERNAL_AVX2_SLOW(flags)   CPUEXT_SUFFIX_SLOW2(flags, _EXTERNAL, AVX2, AVX)
 #define EXTERNAL_AESNI(flags)       CPUEXT_SUFFIX(flags, _EXTERNAL, AESNI)
+#define EXTERNAL_AVX512(flags)      CPUEXT_SUFFIX(flags, _EXTERNAL, AVX512)
 
 #define INLINE_AMD3DNOW(flags)      CPUEXT_SUFFIX(flags, _INLINE, AMD3DNOW)
 #define INLINE_AMD3DNOWEXT(flags)   CPUEXT_SUFFIX(flags, _INLINE, AMD3DNOWEXT)