summaryrefslogtreecommitdiff
path: root/libavcodec/x86
diff options
context:
space:
mode:
Diffstat (limited to 'libavcodec/x86')
-rw-r--r--libavcodec/x86/cabac.h78
-rw-r--r--libavcodec/x86/h264_i386.h62
2 files changed, 77 insertions, 63 deletions
diff --git a/libavcodec/x86/cabac.h b/libavcodec/x86/cabac.h
index 2bb0be6374..0c4419b4f3 100644
--- a/libavcodec/x86/cabac.h
+++ b/libavcodec/x86/cabac.h
@@ -27,7 +27,7 @@
#include "config.h"
#if HAVE_FAST_CMOV
-#define BRANCHLESS_GET_CABAC_UPDATE(ret, statep, low, lowword, range, tmp)\
+#define BRANCHLESS_GET_CABAC_UPDATE(ret, statep, low, range, tmp)\
"mov "tmp" , %%ecx \n\t"\
"shl $17 , "tmp" \n\t"\
"cmp "low" , "tmp" \n\t"\
@@ -37,7 +37,7 @@
"xor %%ecx , "ret" \n\t"\
"sub "tmp" , "low" \n\t"
#else /* HAVE_FAST_CMOV */
-#define BRANCHLESS_GET_CABAC_UPDATE(ret, statep, low, lowword, range, tmp)\
+#define BRANCHLESS_GET_CABAC_UPDATE(ret, statep, low, range, tmp)\
"mov "tmp" , %%ecx \n\t"\
"shl $17 , "tmp" \n\t"\
"sub "low" , "tmp" \n\t"\
@@ -51,20 +51,20 @@
"xor "tmp" , "ret" \n\t"
#endif /* HAVE_FAST_CMOV */
-#define BRANCHLESS_GET_CABAC(ret, statep, low, lowword, range, tmp, tmpbyte, byte) \
+#define BRANCHLESS_GET_CABAC(ret, statep, low, lowword, range, tmp, tmpbyte, byte, end) \
"movzbl "statep" , "ret" \n\t"\
"mov "range" , "tmp" \n\t"\
"and $0xC0 , "range" \n\t"\
"movzbl "MANGLE(ff_h264_lps_range)"("ret", "range", 2), "range" \n\t"\
"sub "range" , "tmp" \n\t"\
- BRANCHLESS_GET_CABAC_UPDATE(ret, statep, low, lowword, range, tmp) \
+ BRANCHLESS_GET_CABAC_UPDATE(ret, statep, low, range, tmp) \
"movzbl " MANGLE(ff_h264_norm_shift) "("range"), %%ecx \n\t"\
"shl %%cl , "range" \n\t"\
"movzbl "MANGLE(ff_h264_mlps_state)"+128("ret"), "tmp" \n\t"\
"shl %%cl , "low" \n\t"\
"mov "tmpbyte" , "statep" \n\t"\
"test "lowword" , "lowword" \n\t"\
- " jnz 1f \n\t"\
+ " jnz 2f \n\t"\
"mov "byte" , %%"REG_c" \n\t"\
"add"OPSIZE" $2 , "byte" \n\t"\
"movzwl (%%"REG_c") , "tmp" \n\t"\
@@ -79,7 +79,7 @@
"add $7 , %%ecx \n\t"\
"shl %%cl , "tmp" \n\t"\
"add "tmp" , "low" \n\t"\
- "1: \n\t"
+ "2: \n\t"
#if HAVE_7REGS && !defined(BROKEN_RELOCATIONS) && !(defined(__i386) && defined(__clang__) && (__clang_major__<2 || (__clang_major__==2 && __clang_minor__<10)))\
@@ -91,11 +91,13 @@ static av_always_inline int get_cabac_inline_x86(CABACContext *c,
int bit, tmp;
__asm__ volatile(
- BRANCHLESS_GET_CABAC("%0", "(%5)", "%1", "%w1", "%2",
- "%3", "%b3", "%4")
- :"=&r"(bit), "+&r"(c->low), "+&r"(c->range), "=&q"(tmp),
- "+m"(c->bytestream)
- :"r"(state)
+ BRANCHLESS_GET_CABAC("%0", "(%4)", "%1", "%w1",
+ "%2", "%3", "%b3",
+ "%a6(%5)", "%a7(%5)")
+ : "=&r"(bit), "+&r"(c->low), "+&r"(c->range), "=&q"(tmp)
+ : "r"(state), "r"(c),
+ "i"(offsetof(CABACContext, bytestream)),
+ "i"(offsetof(CABACContext, bytestream_end))
: "%"REG_c, "memory"
);
return bit & 1;
@@ -107,32 +109,36 @@ static av_always_inline int get_cabac_bypass_sign_x86(CABACContext *c, int val)
{
x86_reg tmp;
__asm__ volatile(
- "movl %4, %k1 \n\t"
- "movl %2, %%eax \n\t"
- "shl $17, %k1 \n\t"
- "add %%eax, %%eax \n\t"
- "sub %k1, %%eax \n\t"
- "cltd \n\t"
- "and %%edx, %k1 \n\t"
- "add %k1, %%eax \n\t"
- "xor %%edx, %%ecx \n\t"
- "sub %%edx, %%ecx \n\t"
- "test %%ax, %%ax \n\t"
- " jnz 1f \n\t"
- "mov %3, %1 \n\t"
- "subl $0xFFFF, %%eax \n\t"
- "movzwl (%1), %%edx \n\t"
- "bswap %%edx \n\t"
- "shrl $15, %%edx \n\t"
- "add $2, %1 \n\t"
- "addl %%edx, %%eax \n\t"
- "mov %1, %3 \n\t"
- "1: \n\t"
- "movl %%eax, %2 \n\t"
+ "movl %a6(%2), %k1 \n\t"
+ "movl %a3(%2), %%eax \n\t"
+ "shl $17, %k1 \n\t"
+ "add %%eax, %%eax \n\t"
+ "sub %k1, %%eax \n\t"
+ "cltd \n\t"
+ "and %%edx, %k1 \n\t"
+ "add %k1, %%eax \n\t"
+ "xor %%edx, %%ecx \n\t"
+ "sub %%edx, %%ecx \n\t"
+ "test %%ax, %%ax \n\t"
+ "jnz 1f \n\t"
+ "mov %a4(%2), %1 \n\t"
+ "subl $0xFFFF, %%eax \n\t"
+ "movzwl (%1), %%edx \n\t"
+ "bswap %%edx \n\t"
+ "shrl $15, %%edx \n\t"
+ "add $2, %1 \n\t"
+ "addl %%edx, %%eax \n\t"
+ "mov %1, %a4(%2) \n\t"
+ "1: \n\t"
+ "movl %%eax, %a3(%2) \n\t"
- :"+c"(val), "=&r"(tmp), "+m"(c->low), "+m"(c->bytestream)
- :"m"(c->range)
- : "%eax", "%edx"
+ : "+c"(val), "=&r"(tmp)
+ : "r"(c),
+ "i"(offsetof(CABACContext, low)),
+ "i"(offsetof(CABACContext, bytestream)),
+ "i"(offsetof(CABACContext, bytestream_end)),
+ "i"(offsetof(CABACContext, range))
+ : "%eax", "%edx", "memory"
);
return val;
}
diff --git a/libavcodec/x86/h264_i386.h b/libavcodec/x86/h264_i386.h
index 510f726ba1..6aa2d073e3 100644
--- 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;
generated by cgit v1.2.3 (git 2.39.1) at 2024-07-05 17:26:08 +0000