Update FFmpeg code to n3.2-65-gee56777

media/ffvpx/libavutil/x86/asm.h

@@ -28,46 +28,46 @@ typedef struct xmm_reg { uint64_t a, b; } xmm_reg;
 typedef struct ymm_reg { uint64_t a, b, c, d; } ymm_reg;
 
 #if ARCH_X86_64
-#    define OPSIZE "q"
-#    define REG_a "rax"
-#    define REG_b "rbx"
-#    define REG_c "rcx"
-#    define REG_d "rdx"
-#    define REG_D "rdi"
-#    define REG_S "rsi"
-#    define PTR_SIZE "8"
+#    define FF_OPSIZE "q"
+#    define FF_REG_a "rax"
+#    define FF_REG_b "rbx"
+#    define FF_REG_c "rcx"
+#    define FF_REG_d "rdx"
+#    define FF_REG_D "rdi"
+#    define FF_REG_S "rsi"
+#    define FF_PTR_SIZE "8"
 typedef int64_t x86_reg;
 
-/* REG_SP is defined in Solaris sys headers, so use REG_sp */
-#    define REG_sp "rsp"
-#    define REG_BP "rbp"
-#    define REGBP   rbp
-#    define REGa    rax
-#    define REGb    rbx
-#    define REGc    rcx
-#    define REGd    rdx
-#    define REGSP   rsp
+/* FF_REG_SP is defined in Solaris sys headers, so use FF_REG_sp */
+#    define FF_REG_sp "rsp"
+#    define FF_REG_BP "rbp"
+#    define FF_REGBP   rbp
+#    define FF_REGa    rax
+#    define FF_REGb    rbx
+#    define FF_REGc    rcx
+#    define FF_REGd    rdx
+#    define FF_REGSP   rsp
 
 #elif ARCH_X86_32
 
-#    define OPSIZE "l"
-#    define REG_a "eax"
-#    define REG_b "ebx"
-#    define REG_c "ecx"
-#    define REG_d "edx"
-#    define REG_D "edi"
-#    define REG_S "esi"
-#    define PTR_SIZE "4"
+#    define FF_OPSIZE "l"
+#    define FF_REG_a "eax"
+#    define FF_REG_b "ebx"
+#    define FF_REG_c "ecx"
+#    define FF_REG_d "edx"
+#    define FF_REG_D "edi"
+#    define FF_REG_S "esi"
+#    define FF_PTR_SIZE "4"
 typedef int32_t x86_reg;
 
-#    define REG_sp "esp"
-#    define REG_BP "ebp"
-#    define REGBP   ebp
-#    define REGa    eax
-#    define REGb    ebx
-#    define REGc    ecx
-#    define REGd    edx
-#    define REGSP   esp
+#    define FF_REG_sp "esp"
+#    define FF_REG_BP "ebp"
+#    define FF_REGBP   ebp
+#    define FF_REGa    eax
+#    define FF_REGb    ebx
+#    define FF_REGc    ecx
+#    define FF_REGd    edx
+#    define FF_REGSP   esp
 #else
 typedef int x86_reg;
 #endif

media/ffvpx/libavutil/x86/cpu.c

@@ -41,9 +41,9 @@
 /* ebx saving is necessary for PIC. gcc seems unable to see it alone */
 #define cpuid(index, eax, ebx, ecx, edx)                        \
     __asm__ volatile (                                          \
-        "mov    %%"REG_b", %%"REG_S" \n\t"                      \
+        "mov    %%"FF_REG_b", %%"FF_REG_S" \n\t"                \
         "cpuid                       \n\t"                      \
-        "xchg   %%"REG_b", %%"REG_S                             \
+        "xchg   %%"FF_REG_b", %%"FF_REG_S                       \
         : "=a" (eax), "=S" (ebx), "=c" (ecx), "=d" (edx)        \
         : "0" (index), "2"(0))
 

media/ffvpx/libavutil/x86/emms.h

@@ -31,6 +31,8 @@ void avpriv_emms_yasm(void);
  * Empty mmx state.
  * this must be called between any dsp function and float/double code.
  * for example sin(); dsp->idct_put(); emms_c(); cos()
+ * Note, *alloc() and *free() also use float code in some libc implementations
+ * thus this also applies to them or any function using them.
  */
 static av_always_inline void emms_c(void)
 {

media/ffvpx/libavutil/x86/x86util.asm

@@ -30,7 +30,10 @@
 %include "libavutil/x86/x86inc.asm"
 
 %macro SBUTTERFLY 4
-%if avx_enabled == 0
+%ifidn %1, dqqq
+    vperm2i128  m%4, m%2, m%3, q0301
+    vinserti128 m%2, m%2, xm%3, 1
+%elif avx_enabled == 0
     mova      m%4, m%2
     punpckl%1 m%2, m%3
     punpckh%1 m%4, m%3
@@ -193,6 +196,70 @@
 %endif
 %endmacro
 
+%macro TRANSPOSE16x16W 18-19
+; in:  m0..m15, unless %19 in which case m6 is in %17
+; out: m0..m15, unless %19 in which case m4 is in %18
+; spills into %17 and %18
+%if %0 < 19
+    mova       %17, m%7
+%endif
+
+    SBUTTERFLY dqqq, %1,  %9, %7
+    SBUTTERFLY dqqq, %2, %10, %7
+    SBUTTERFLY dqqq, %3, %11, %7
+    SBUTTERFLY dqqq, %4, %12, %7
+    SBUTTERFLY dqqq, %5, %13, %7
+    SBUTTERFLY dqqq, %6, %14, %7
+    mova       %18, m%14
+    mova       m%7, %17
+    SBUTTERFLY dqqq, %7, %15, %14
+    SBUTTERFLY dqqq, %8, %16, %14
+
+    SBUTTERFLY  wd,  %1,  %2, %14
+    SBUTTERFLY  wd,  %3,  %4, %14
+    SBUTTERFLY  wd,  %5,  %6, %14
+    SBUTTERFLY  wd,  %7,  %8, %14
+    SBUTTERFLY  wd,  %9, %10, %14
+    SBUTTERFLY  wd, %11, %12, %14
+    mova       %17, m%12
+    mova      m%14, %18
+    SBUTTERFLY  wd, %13, %14, %12
+    SBUTTERFLY  wd, %15, %16, %12
+
+    SBUTTERFLY  dq,  %1,  %3, %12
+    SBUTTERFLY  dq,  %2,  %4, %12
+    SBUTTERFLY  dq,  %5,  %7, %12
+    SBUTTERFLY  dq,  %6,  %8, %12
+    SBUTTERFLY  dq,  %9, %11, %12
+    mova       %18, m%11
+    mova      m%12, %17
+    SBUTTERFLY  dq, %10, %12, %11
+    SBUTTERFLY  dq, %13, %15, %11
+    SBUTTERFLY  dq, %14, %16, %11
+
+    SBUTTERFLY qdq,  %1,  %5, %11
+    SBUTTERFLY qdq,  %2,  %6, %11
+    SBUTTERFLY qdq,  %3,  %7, %11
+    SBUTTERFLY qdq,  %4,  %8, %11
+
+    SWAP        %2, %5
+    SWAP        %4, %7
+
+    SBUTTERFLY qdq,  %9, %13, %11
+    SBUTTERFLY qdq, %10, %14, %11
+    mova      m%11, %18
+    mova       %18, m%5
+    SBUTTERFLY qdq, %11, %15, %5
+    SBUTTERFLY qdq, %12, %16, %5
+
+%if %0 < 19
+    mova       m%5, %18
+%endif
+
+    SWAP       %10, %13
+    SWAP       %12, %15
+%endmacro
+
 ; PABSW macro assumes %1 != %2, while ABS1/2 macros work in-place
 %macro PABSW 2
 %if cpuflag(ssse3)