Merge remote-tracking branch 'qatar/master'

[ffmpeg] / libavcodec / x86 / h264_i386.h

diff --git a/libavcodec/x86/h264_i386.h b/libavcodec/x86/h264_i386.h
index 510f726ba108cf6248b4fa3a7f96abdc0096e04e..6aa2d073e3c096bb86a54f08d22730358358b9d9 100644 (file)
--- a/libavcodec/x86/h264_i386.h
+++ b/libavcodec/x86/h264_i386.h
@@ -46,17 +46,19 @@ static int decode_significance_x86(CABACContext *c, int max_coeff,
     int bit;
     x86_reg coeff_count;
     __asm__ volatile(
-        "2:                                     \n\t"
+        "3:                                     \n\t"
 
-        BRANCHLESS_GET_CABAC("%4", "(%1)", "%3",
-                             "%w3", "%5", "%k0", "%b0", "%6")
+        BRANCHLESS_GET_CABAC("%4", "(%1)", "%3", "%w3",
+                             "%5", "%k0", "%b0",
+                             "%a11(%6)", "%a12(%6)")
 
         "test $1, %4                            \n\t"
-        " jz 3f                                 \n\t"
+        " jz 4f                                 \n\t"
         "add  %10, %1                           \n\t"
 
-        BRANCHLESS_GET_CABAC("%4", "(%1)", "%3",
-                             "%w3", "%5", "%k0", "%b0", "%6")
+        BRANCHLESS_GET_CABAC("%4", "(%1)", "%3", "%w3",
+                             "%5", "%k0", "%b0",
+                             "%a11(%6)", "%a12(%6)")
 
         "sub  %10, %1                           \n\t"
         "mov  %2, %0                            \n\t"
@@ -65,25 +67,26 @@ static int decode_significance_x86(CABACContext *c, int max_coeff,
         "movl %%ecx, (%0)                       \n\t"
 
         "test $1, %4                            \n\t"
-        " jnz 4f                                \n\t"
+        " jnz 5f                                \n\t"
 
         "add"OPSIZE"  $4, %2                    \n\t"
 
-        "3:                                     \n\t"
+        "4:                                     \n\t"
         "add  $1, %1                            \n\t"
         "cmp  %8, %1                            \n\t"
-        " jb 2b                                 \n\t"
+        " jb 3b                                 \n\t"
         "mov  %2, %0                            \n\t"
         "movl %7, %%ecx                         \n\t"
         "add  %1, %%"REG_c"                     \n\t"
         "movl %%ecx, (%0)                       \n\t"
-        "4:                                     \n\t"
+        "5:                                     \n\t"
         "add  %9, %k0                           \n\t"
         "shr $2, %k0                            \n\t"
-        :"=&q"(coeff_count), "+r"(significant_coeff_ctx_base), "+m"(index),
-         "+&r"(c->low), "=&r"(bit), "+&r"(c->range),
-         "+m"(c->bytestream)
-        :"m"(minusstart), "m"(end), "m"(minusindex), "m"(last_off)
+        : "=&q"(coeff_count), "+r"(significant_coeff_ctx_base), "+m"(index),
+          "+&r"(c->low), "=&r"(bit), "+&r"(c->range)
+        : "r"(c), "m"(minusstart), "m"(end), "m"(minusindex), "m"(last_off),
+          "i"(offsetof(CABACContext, bytestream)),
+          "i"(offsetof(CABACContext, bytestream_end))
         : "%"REG_c, "memory"
     );
     return coeff_count;
@@ -99,47 +102,52 @@ static int decode_significance_8x8_x86(CABACContext *c,
     x86_reg state;
     __asm__ volatile(
         "mov %1, %6                             \n\t"
-        "2:                                     \n\t"
+        "3:                                     \n\t"
 
         "mov %10, %0                            \n\t"
         "movzbl (%0, %6), %k6                   \n\t"
         "add %9, %6                             \n\t"
 
-        BRANCHLESS_GET_CABAC("%4", "(%6)", "%3",
-                             "%w3", "%5", "%k0", "%b0", "%7")
+        BRANCHLESS_GET_CABAC("%4", "(%6)", "%3", "%w3",
+                             "%5", "%k0", "%b0",
+                             "%a12(%7)", "%a13(%7)")
 
         "mov %1, %k6                            \n\t"
         "test $1, %4                            \n\t"
-        " jz 3f                                 \n\t"
+        " jz 4f                                 \n\t"
 
         "movzbl "MANGLE(last_coeff_flag_offset_8x8)"(%k6), %k6\n\t"
         "add %11, %6                            \n\t"
 
-        BRANCHLESS_GET_CABAC("%4", "(%6)", "%3",
-                             "%w3", "%5", "%k0", "%b0", "%7")
+        BRANCHLESS_GET_CABAC("%4", "(%6)", "%3", "%w3",
+                             "%5", "%k0", "%b0",
+                             "%a12(%7)", "%a13(%7)")
 
         "mov %2, %0                             \n\t"
         "mov %1, %k6                            \n\t"
         "movl %k6, (%0)                         \n\t"
 
         "test $1, %4                            \n\t"
-        " jnz 4f                                \n\t"
+        " jnz 5f                                \n\t"
 
         "add"OPSIZE"  $4, %2                    \n\t"
 
-        "3:                                     \n\t"
+        "4:                                     \n\t"
         "addl $1, %k6                           \n\t"
         "mov %k6, %1                            \n\t"
         "cmpl $63, %k6                          \n\t"
-        " jb 2b                                 \n\t"
+        " jb 3b                                 \n\t"
         "mov %2, %0                             \n\t"
         "movl %k6, (%0)                         \n\t"
-        "4:                                     \n\t"
+        "5:                                     \n\t"
         "addl %8, %k0                           \n\t"
         "shr $2, %k0                            \n\t"
-        :"=&q"(coeff_count),"+m"(last), "+m"(index), "+&r"(c->low), "=&r"(bit),
-         "+&r"(c->range), "=&r"(state), "+m"(c->bytestream)
-        :"m"(minusindex), "m"(significant_coeff_ctx_base), "m"(sig_off), "m"(last_coeff_ctx_base)
+        : "=&q"(coeff_count), "+m"(last), "+m"(index), "+&r"(c->low),
+          "=&r"(bit), "+&r"(c->range), "=&r"(state)
+        : "r"(c), "m"(minusindex), "m"(significant_coeff_ctx_base),
+          "m"(sig_off), "m"(last_coeff_ctx_base),
+          "i"(offsetof(CABACContext, bytestream)),
+          "i"(offsetof(CABACContext, bytestream_end))
         : "%"REG_c, "memory"
     );
     return coeff_count;