From f5f004bc5a40e5a5de62bd5e2818ab5de7f6bf21 Mon Sep 17 00:00:00 2001 From: Mans Rullgard Date: Thu, 28 Jul 2011 20:21:33 +0100 Subject: x86: cabac: don't load/store context values in asm Inspection of compiled code shows gcc handles these fine on its own. Benchmarking also shows no measurable speed difference. Removing the remaining cases in get_cabac_bypass_sign_x86() does cause more substantial changes to the compiled code with uncertain impact. Signed-off-by: Mans Rullgard --- libavcodec/x86/cabac.h | 12 +++--------- 1 file changed, 3 insertions(+), 9 deletions(-) (limited to 'libavcodec/x86/cabac.h') diff --git a/libavcodec/x86/cabac.h b/libavcodec/x86/cabac.h index 1ad74ff3e0..ae3f4b6e9e 100644 --- a/libavcodec/x86/cabac.h +++ b/libavcodec/x86/cabac.h @@ -87,19 +87,13 @@ static av_always_inline int get_cabac_inline_x86(CABACContext *c, uint8_t *const state) { - int bit, low, range, tmp; + int bit, tmp; __asm__ volatile( - "movl %a6(%5), %2 \n\t" - "movl %a7(%5), %1 \n\t" BRANCHLESS_GET_CABAC("%0", "%5", "(%4)", "%1", "%w1", "%2", - "%3", "%b3", "%a8") - "movl %2, %a6(%5) \n\t" - "movl %1, %a7(%5) \n\t" - - :"=&r"(bit), "=&r"(low), "=&r"(range), "=&q"(tmp) + "%3", "%b3", "%a6") + :"=&r"(bit), "+&r"(c->low), "+&r"(c->range), "=&q"(tmp) :"r"(state), "r"(c), - "i"(offsetof(CABACContext, range)), "i"(offsetof(CABACContext, low)), "i"(offsetof(CABACContext, bytestream)) : "%"REG_c, "memory" ); -- cgit v1.2.3