diff options
| author | Roland Scheidegger <rscheidegger_lists@hispeed.ch> | 2012-04-27 22:19:39 +0200 |
|---|---|---|
| committer | Ronald S. Bultje <rsbultje@gmail.com> | 2012-04-28 09:43:25 -0700 |
| commit | 9b9df1cdff149db5bbe6726b236934c5b5fbe21d (patch) | |
| tree | b9c6d51cc923f003b05a51af7337ad048c426982 /libavcodec/x86/h264_i386.h | |
| parent | 14e9ffc1e41424a530c83310611979c0d246417b (diff) | |
| download | ffmpeg-9b9df1cdff149db5bbe6726b236934c5b5fbe21d.tar.gz | |
h264: new assembly version of get_cabac for x86_64 with PIC
This adds a hand-optimized assembly version for get_cabac much like the
existing one, but it works if the table offsets are RIP-relative.
Compared to the non-RIP-relative version this adds 2 lea instructions
and it needs one extra register. get_cabac() gets about 40% faster, for
an overall speedup of about 5%.
Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
Diffstat (limited to 'libavcodec/x86/h264_i386.h')
| -rw-r--r-- | libavcodec/x86/h264_i386.h | 53 |
1 files changed, 37 insertions, 16 deletions
diff --git a/libavcodec/x86/h264_i386.h b/libavcodec/x86/h264_i386.h index add795e285..10ea32e0b8 100644 --- a/libavcodec/x86/h264_i386.h +++ b/libavcodec/x86/h264_i386.h @@ -36,7 +36,7 @@ //FIXME use some macros to avoid duplicating get_cabac (cannot be done yet //as that would make optimization work hard) -#if HAVE_7REGS && !defined(BROKEN_RELOCATIONS) +#if HAVE_7REGS static int decode_significance_x86(CABACContext *c, int max_coeff, uint8_t *significant_coeff_ctx_base, int *index, x86_reg last_off){ @@ -46,20 +46,29 @@ static int decode_significance_x86(CABACContext *c, int max_coeff, int bit; x86_reg coeff_count; +#ifdef BROKEN_RELOCATIONS + void *tables; + + __asm__ volatile( + "lea "MANGLE(ff_h264_cabac_tables)", %0 \n\t" + : "=&r"(tables) + ); +#endif + __asm__ volatile( "3: \n\t" - BRANCHLESS_GET_CABAC("%4", "(%1)", "%3", "%w3", - "%5", "%k0", "%b0", - "%a11(%6)", "%a12(%6)", "%a13", "%a14", "%a15") + BRANCHLESS_GET_CABAC("%4", "%q4", "(%1)", "%3", "%w3", + "%5", "%q5", "%k0", "%b0", + "%a11(%6)", "%a12(%6)", "%a13", "%a14", "%a15", "%16") "test $1, %4 \n\t" " jz 4f \n\t" "add %10, %1 \n\t" - BRANCHLESS_GET_CABAC("%4", "(%1)", "%3", "%w3", - "%5", "%k0", "%b0", - "%a11(%6)", "%a12(%6)", "%a13", "%a14", "%a15") + BRANCHLESS_GET_CABAC("%4", "%q4", "(%1)", "%3", "%w3", + "%5", "%q5", "%k0", "%b0", + "%a11(%6)", "%a12(%6)", "%a13", "%a14", "%a15", "%16") "sub %10, %1 \n\t" "mov %2, %0 \n\t" @@ -90,7 +99,7 @@ static int decode_significance_x86(CABACContext *c, int max_coeff, "i"(offsetof(CABACContext, bytestream_end)), "i"(H264_NORM_SHIFT_OFFSET), "i"(H264_LPS_RANGE_OFFSET), - "i"(H264_MLPS_STATE_OFFSET) + "i"(H264_MLPS_STATE_OFFSET) TABLES_ARG : "%"REG_c, "memory" ); return coeff_count; @@ -105,6 +114,15 @@ static int decode_significance_8x8_x86(CABACContext *c, x86_reg last=0; x86_reg state; +#ifdef BROKEN_RELOCATIONS + void *tables; + + __asm__ volatile( + "lea "MANGLE(ff_h264_cabac_tables)", %0 \n\t" + : "=&r"(tables) + ); +#endif + __asm__ volatile( "mov %1, %6 \n\t" "3: \n\t" @@ -113,21 +131,24 @@ static int decode_significance_8x8_x86(CABACContext *c, "movzbl (%0, %6), %k6 \n\t" "add %9, %6 \n\t" - BRANCHLESS_GET_CABAC("%4", "(%6)", "%3", "%w3", - "%5", "%k0", "%b0", - "%a12(%7)", "%a13(%7)", "%a14", "%a15", "%a16") + BRANCHLESS_GET_CABAC("%4", "%q4", "(%6)", "%3", "%w3", + "%5", "%q5", "%k0", "%b0", + "%a12(%7)", "%a13(%7)", "%a14", "%a15", "%a16", "%18") "mov %1, %k6 \n\t" "test $1, %4 \n\t" " jz 4f \n\t" +#ifdef BROKEN_RELOCATIONS + "movzbl %a17(%18, %q6), %k6\n\t" +#else "movzbl "MANGLE(ff_h264_cabac_tables)"+%a17(%k6), %k6\n\t" - +#endif "add %11, %6 \n\t" - BRANCHLESS_GET_CABAC("%4", "(%6)", "%3", "%w3", - "%5", "%k0", "%b0", - "%a12(%7)", "%a13(%7)", "%a14", "%a15", "%a16") + BRANCHLESS_GET_CABAC("%4", "%q4", "(%6)", "%3", "%w3", + "%5", "%q5", "%k0", "%b0", + "%a12(%7)", "%a13(%7)", "%a14", "%a15", "%a16", "%18") "mov %2, %0 \n\t" "mov %1, %k6 \n\t" @@ -157,7 +178,7 @@ static int decode_significance_8x8_x86(CABACContext *c, "i"(H264_NORM_SHIFT_OFFSET), "i"(H264_LPS_RANGE_OFFSET), "i"(H264_MLPS_STATE_OFFSET), - "i"(H264_LAST_COEFF_FLAG_OFFSET_8x8_OFFSET) + "i"(H264_LAST_COEFF_FLAG_OFFSET_8x8_OFFSET) TABLES_ARG : "%"REG_c, "memory" ); return coeff_count; |