From: Martin Storsjö Date: Mon, 22 Apr 2013 09:23:47 +0000 (+0300) Subject: x86: Factorize duplicated inline assembly snippets X-Git-Tag: v10_alpha1~1721 X-Git-Url: http://review.tizen.org/git/?a=commitdiff_plain;h=6a8561dbd7c078eb75985f7011ad1ad3fda9e223;p=platform%2Fupstream%2Flibav.git x86: Factorize duplicated inline assembly snippets Signed-off-by: Diego Biurrun --- diff --git a/libavcodec/x86/dsputil_mmx.c b/libavcodec/x86/dsputil_mmx.c index 1d96efc..b095b46 100644 --- a/libavcodec/x86/dsputil_mmx.c +++ b/libavcodec/x86/dsputil_mmx.c @@ -109,60 +109,6 @@ void ff_put_no_rnd_mpeg4_qpel8_v_lowpass_mmxext(uint8_t *dst, uint8_t *src, #if HAVE_INLINE_ASM -#define JUMPALIGN() __asm__ volatile (".p2align 3"::) -#define MOVQ_ZERO(regd) __asm__ volatile ("pxor %%"#regd", %%"#regd ::) - -#define MOVQ_BFE(regd) \ - __asm__ volatile ( \ - "pcmpeqd %%"#regd", %%"#regd" \n\t" \ - "paddb %%"#regd", %%"#regd" \n\t" ::) - -#ifndef PIC -#define MOVQ_BONE(regd) __asm__ volatile ("movq %0, %%"#regd" \n\t" :: "m"(ff_bone)) -#define MOVQ_WTWO(regd) __asm__ volatile ("movq %0, %%"#regd" \n\t" :: "m"(ff_wtwo)) -#else -// for shared library it's better to use this way for accessing constants -// pcmpeqd -> -1 -#define MOVQ_BONE(regd) \ - __asm__ volatile ( \ - "pcmpeqd %%"#regd", %%"#regd" \n\t" \ - "psrlw $15, %%"#regd" \n\t" \ - "packuswb %%"#regd", %%"#regd" \n\t" ::) - -#define MOVQ_WTWO(regd) \ - __asm__ volatile ( \ - "pcmpeqd %%"#regd", %%"#regd" \n\t" \ - "psrlw $15, %%"#regd" \n\t" \ - "psllw $1, %%"#regd" \n\t"::) - -#endif - -// using regr as temporary and for the output result -// first argument is unmodifed and second is trashed -// regfe is supposed to contain 0xfefefefefefefefe -#define PAVGB_MMX(rega, regb, regr, regfe) \ - "movq "#rega", "#regr" \n\t" \ - "por "#regb", "#regr" \n\t" \ - "pxor "#rega", "#regb" \n\t" \ - "pand "#regfe", "#regb" \n\t" \ - "psrlq $1, "#regb" \n\t" \ - "psubb "#regb", "#regr" \n\t" - -// mm6 is supposed to contain 0xfefefefefefefefe -#define PAVGBP_MMX(rega, regb, regr, regc, regd, regp) \ - "movq "#rega", "#regr" \n\t" \ - "movq "#regc", "#regp" \n\t" \ - "por "#regb", "#regr" \n\t" \ - "por "#regd", "#regp" \n\t" \ - "pxor "#rega", "#regb" \n\t" \ - "pxor "#regc", "#regd" \n\t" \ - "pand %%mm6, "#regb" \n\t" \ - "pand %%mm6, "#regd" \n\t" \ - "psrlq $1, "#regd" \n\t" \ - "psrlq $1, "#regb" \n\t" \ - "psubb "#regb", "#regr" \n\t" \ - "psubb "#regd", "#regp" \n\t" - /***********************************/ /* MMX rounding */ diff --git a/libavcodec/x86/dsputil_mmx.h b/libavcodec/x86/dsputil_mmx.h index c3e42c2..ca80e07 100644 --- a/libavcodec/x86/dsputil_mmx.h +++ b/libavcodec/x86/dsputil_mmx.h @@ -72,6 +72,82 @@ extern const double ff_pd_2[2]; "pcmpeqd %%" #regd ", %%" #regd " \n\t" \ "psrlw $15, %%" #regd ::) +#define JUMPALIGN() __asm__ volatile (".p2align 3"::) +#define MOVQ_ZERO(regd) __asm__ volatile ("pxor %%"#regd", %%"#regd ::) + +#define MOVQ_BFE(regd) \ + __asm__ volatile ( \ + "pcmpeqd %%"#regd", %%"#regd" \n\t" \ + "paddb %%"#regd", %%"#regd" \n\t" ::) + +#ifndef PIC +#define MOVQ_BONE(regd) __asm__ volatile ("movq %0, %%"#regd" \n\t" :: "m"(ff_bone)) +#define MOVQ_WTWO(regd) __asm__ volatile ("movq %0, %%"#regd" \n\t" :: "m"(ff_wtwo)) +#else +// for shared library it's better to use this way for accessing constants +// pcmpeqd -> -1 +#define MOVQ_BONE(regd) \ + __asm__ volatile ( \ + "pcmpeqd %%"#regd", %%"#regd" \n\t" \ + "psrlw $15, %%"#regd" \n\t" \ + "packuswb %%"#regd", %%"#regd" \n\t" ::) + +#define MOVQ_WTWO(regd) \ + __asm__ volatile ( \ + "pcmpeqd %%"#regd", %%"#regd" \n\t" \ + "psrlw $15, %%"#regd" \n\t" \ + "psllw $1, %%"#regd" \n\t"::) + +#endif + +// using regr as temporary and for the output result +// first argument is unmodifed and second is trashed +// regfe is supposed to contain 0xfefefefefefefefe +#define PAVGB_MMX_NO_RND(rega, regb, regr, regfe) \ + "movq "#rega", "#regr" \n\t" \ + "pand "#regb", "#regr" \n\t" \ + "pxor "#rega", "#regb" \n\t" \ + "pand "#regfe", "#regb" \n\t" \ + "psrlq $1, "#regb" \n\t" \ + "paddb "#regb", "#regr" \n\t" + +#define PAVGB_MMX(rega, regb, regr, regfe) \ + "movq "#rega", "#regr" \n\t" \ + "por "#regb", "#regr" \n\t" \ + "pxor "#rega", "#regb" \n\t" \ + "pand "#regfe", "#regb" \n\t" \ + "psrlq $1, "#regb" \n\t" \ + "psubb "#regb", "#regr" \n\t" + +// mm6 is supposed to contain 0xfefefefefefefefe +#define PAVGBP_MMX_NO_RND(rega, regb, regr, regc, regd, regp) \ + "movq "#rega", "#regr" \n\t" \ + "movq "#regc", "#regp" \n\t" \ + "pand "#regb", "#regr" \n\t" \ + "pand "#regd", "#regp" \n\t" \ + "pxor "#rega", "#regb" \n\t" \ + "pxor "#regc", "#regd" \n\t" \ + "pand %%mm6, "#regb" \n\t" \ + "pand %%mm6, "#regd" \n\t" \ + "psrlq $1, "#regb" \n\t" \ + "psrlq $1, "#regd" \n\t" \ + "paddb "#regb", "#regr" \n\t" \ + "paddb "#regd", "#regp" \n\t" + +#define PAVGBP_MMX(rega, regb, regr, regc, regd, regp) \ + "movq "#rega", "#regr" \n\t" \ + "movq "#regc", "#regp" \n\t" \ + "por "#regb", "#regr" \n\t" \ + "por "#regd", "#regp" \n\t" \ + "pxor "#rega", "#regb" \n\t" \ + "pxor "#regc", "#regd" \n\t" \ + "pand %%mm6, "#regb" \n\t" \ + "pand %%mm6, "#regd" \n\t" \ + "psrlq $1, "#regd" \n\t" \ + "psrlq $1, "#regb" \n\t" \ + "psubb "#regb", "#regr" \n\t" \ + "psubb "#regd", "#regp" \n\t" + void ff_dsputilenc_init_mmx(DSPContext* c, AVCodecContext *avctx); void ff_dsputil_init_pix_mmx(DSPContext* c, AVCodecContext *avctx); diff --git a/libavcodec/x86/hpeldsp_init.c b/libavcodec/x86/hpeldsp_init.c index 512852f..342bcbb 100644 --- a/libavcodec/x86/hpeldsp_init.c +++ b/libavcodec/x86/hpeldsp_init.c @@ -77,82 +77,6 @@ void ff_avg_pixels8_xy2_3dnow(uint8_t *block, const uint8_t *pixels, #if HAVE_INLINE_ASM -#define JUMPALIGN() __asm__ volatile (".p2align 3"::) -#define MOVQ_ZERO(regd) __asm__ volatile ("pxor %%"#regd", %%"#regd ::) - -#define MOVQ_BFE(regd) \ - __asm__ volatile ( \ - "pcmpeqd %%"#regd", %%"#regd" \n\t" \ - "paddb %%"#regd", %%"#regd" \n\t" ::) - -#ifndef PIC -#define MOVQ_BONE(regd) __asm__ volatile ("movq %0, %%"#regd" \n\t" :: "m"(ff_bone)) -#define MOVQ_WTWO(regd) __asm__ volatile ("movq %0, %%"#regd" \n\t" :: "m"(ff_wtwo)) -#else -// for shared library it's better to use this way for accessing constants -// pcmpeqd -> -1 -#define MOVQ_BONE(regd) \ - __asm__ volatile ( \ - "pcmpeqd %%"#regd", %%"#regd" \n\t" \ - "psrlw $15, %%"#regd" \n\t" \ - "packuswb %%"#regd", %%"#regd" \n\t" ::) - -#define MOVQ_WTWO(regd) \ - __asm__ volatile ( \ - "pcmpeqd %%"#regd", %%"#regd" \n\t" \ - "psrlw $15, %%"#regd" \n\t" \ - "psllw $1, %%"#regd" \n\t"::) - -#endif - -// using regr as temporary and for the output result -// first argument is unmodifed and second is trashed -// regfe is supposed to contain 0xfefefefefefefefe -#define PAVGB_MMX_NO_RND(rega, regb, regr, regfe) \ - "movq "#rega", "#regr" \n\t" \ - "pand "#regb", "#regr" \n\t" \ - "pxor "#rega", "#regb" \n\t" \ - "pand "#regfe", "#regb" \n\t" \ - "psrlq $1, "#regb" \n\t" \ - "paddb "#regb", "#regr" \n\t" - -#define PAVGB_MMX(rega, regb, regr, regfe) \ - "movq "#rega", "#regr" \n\t" \ - "por "#regb", "#regr" \n\t" \ - "pxor "#rega", "#regb" \n\t" \ - "pand "#regfe", "#regb" \n\t" \ - "psrlq $1, "#regb" \n\t" \ - "psubb "#regb", "#regr" \n\t" - -// mm6 is supposed to contain 0xfefefefefefefefe -#define PAVGBP_MMX_NO_RND(rega, regb, regr, regc, regd, regp) \ - "movq "#rega", "#regr" \n\t" \ - "movq "#regc", "#regp" \n\t" \ - "pand "#regb", "#regr" \n\t" \ - "pand "#regd", "#regp" \n\t" \ - "pxor "#rega", "#regb" \n\t" \ - "pxor "#regc", "#regd" \n\t" \ - "pand %%mm6, "#regb" \n\t" \ - "pand %%mm6, "#regd" \n\t" \ - "psrlq $1, "#regb" \n\t" \ - "psrlq $1, "#regd" \n\t" \ - "paddb "#regb", "#regr" \n\t" \ - "paddb "#regd", "#regp" \n\t" - -#define PAVGBP_MMX(rega, regb, regr, regc, regd, regp) \ - "movq "#rega", "#regr" \n\t" \ - "movq "#regc", "#regp" \n\t" \ - "por "#regb", "#regr" \n\t" \ - "por "#regd", "#regp" \n\t" \ - "pxor "#rega", "#regb" \n\t" \ - "pxor "#regc", "#regd" \n\t" \ - "pand %%mm6, "#regb" \n\t" \ - "pand %%mm6, "#regd" \n\t" \ - "psrlq $1, "#regd" \n\t" \ - "psrlq $1, "#regb" \n\t" \ - "psubb "#regb", "#regr" \n\t" \ - "psubb "#regd", "#regp" \n\t" - /***********************************/ /* MMX no rounding */ #define NO_RND 1