📄 i420_rgb_mmx.h
字号:
punpcklbw %%mm1, %%mm0 # r7r6r5r4 r3______ ______b7 b6b5b4b3 \n\psllw $2,%%mm2 # ________ ____g7g6 g5g4g3__ ________ \n\por %%mm2, %%mm0 # r7r6r5r4 r3__g7g6 g5g4g3b7 b6b5b4b3 \n\movq 8(%0), %%mm6 # Load 8 Y Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0 \n\movq %%mm0, (%3) # store pixel 0-3 \n\ \n\# convert rgb24 plane to rgb16 pack for pixel 0-3 \n\punpckhbw %%mm4, %%mm7 # ________ ________ g7g6g5g4 g3______ \n\punpckhbw %%mm1, %%mm5 # r7r6r5r4 r3______ ______b7 b6b5b4b3 \n\psllw $2,%%mm7 # ________ ____g7g6 g5g4g3__ ________ \n\movd 4(%1), %%mm0 # Load 4 Cb __ __ __ __ u3 u2 u1 u0 \n\por %%mm7, %%mm5 # r7r6r5r4 r3__g7g6 g5g4g3b7 b6b5b4b3 \n\movd 4(%2), %%mm1 # Load 4 Cr __ __ __ __ v3 v2 v1 v0 \n\movq %%mm5, 8(%3) # store pixel 4-7 \n\"/* * convert RGB plane to RGB 16 bits, * mm0 -> B, mm1 -> R, mm2 -> G, * mm4 -> GB, mm5 -> AR pixel 4-7, * mm6 -> GB, mm7 -> AR pixel 0-3 */#define MMX_UNPACK_16 " \n\# mask unneeded bits off \n\pand mmx_mask_f8"G", %%mm0 # b7b6b5b4 b3______ b7b6b5b4 b3______ \n\pand mmx_mask_fc"G", %%mm2 # g7g6g5g4 g3g2____ g7g6g5g4 g3g2____ \n\pand mmx_mask_f8"G", %%mm1 # r7r6r5r4 r3______ r7r6r5r4 r3______ \n\psrlw $3,%%mm0 # ______b7 b6b5b4b3 ______b7 b6b5b4b3 \n\pxor %%mm4, %%mm4 # zero mm4 \n\movq %%mm0, %%mm5 # Copy B7-B0 \n\movq %%mm2, %%mm7 # Copy G7-G0 \n\ \n\# convert rgb24 plane to rgb16 pack for pixel 0-3 \n\punpcklbw %%mm4, %%mm2 # ________ ________ g7g6g5g4 g3g2____ \n\punpcklbw %%mm1, %%mm0 # r7r6r5r4 r3______ ______b7 b6b5b4b3 \n\psllw $3,%%mm2 # ________ __g7g6g5 g4g3g2__ ________ \n\por %%mm2, %%mm0 # r7r6r5r4 r3g7g6g5 g4g3g2b7 b6b5b4b3 \n\movq 8(%0), %%mm6 # Load 8 Y Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0 \n\movq %%mm0, (%3) # store pixel 0-3 \n\ \n\# convert rgb24 plane to rgb16 pack for pixel 0-3 \n\punpckhbw %%mm4, %%mm7 # ________ ________ g7g6g5g4 g3g2____ \n\punpckhbw %%mm1, %%mm5 # r7r6r5r4 r3______ ______b7 b6b5b4b3 \n\psllw $3,%%mm7 # ________ __g7g6g5 g4g3g2__ ________ \n\movd 4(%1), %%mm0 # Load 4 Cb __ __ __ __ u3 u2 u1 u0 \n\por %%mm7, %%mm5 # r7r6r5r4 r3g7g6g5 g4g3g2b7 b6b5b4b3 \n\movd 4(%2), %%mm1 # Load 4 Cr __ __ __ __ v3 v2 v1 v0 \n\movq %%mm5, 8(%3) # store pixel 4-7 \n\"/* * convert RGB plane to RGB packed format, * mm0 -> B, mm1 -> R, mm2 -> G */#define MMX_UNPACK_32_ARGB " \n\pxor %%mm3, %%mm3 # zero mm3 \n\movq %%mm0, %%mm4 # B7 B6 B5 B4 B3 B2 B1 B0 \n\punpcklbw %%mm2, %%mm4 # G3 B3 G2 B2 G1 B1 G0 B0 \n\movq %%mm1, %%mm5 # R7 R6 R5 R4 R3 R2 R1 R0 \n\punpcklbw %%mm3, %%mm5 # 00 R3 00 R2 00 R1 00 R0 \n\movq %%mm4, %%mm6 # G3 B3 G2 B2 G1 B1 G0 B0 \n\punpcklwd %%mm5, %%mm4 # 00 R1 B1 G1 00 R0 B0 G0 \n\movq %%mm4, (%3) # Store ARGB1 ARGB0 \n\punpckhwd %%mm5, %%mm6 # 00 R3 B3 G3 00 R2 B2 G2 \n\movq %%mm6, 8(%3) # Store ARGB3 ARGB2 \n\punpckhbw %%mm2, %%mm0 # G7 B7 G6 B6 G5 B5 G4 B4 \n\punpckhbw %%mm3, %%mm1 # 00 R7 00 R6 00 R5 00 R4 \n\movq %%mm0, %%mm5 # G7 B7 G6 B6 G5 B5 G4 B4 \n\punpcklwd %%mm1, %%mm5 # 00 R5 B5 G5 00 R4 B4 G4 \n\movq %%mm5, 16(%3) # Store ARGB5 ARGB4 \n\punpckhwd %%mm1, %%mm0 # 00 R7 B7 G7 00 R6 B6 G6 \n\movq %%mm0, 24(%3) # Store ARGB7 ARGB6 \n\"#define MMX_UNPACK_32_RGBA " \n\pxor %%mm3, %%mm3 # zero mm3 \n\movq %%mm2, %%mm4 # G7 G6 G5 G4 G3 G2 G1 G0 \n\punpcklbw %%mm1, %%mm4 # R3 G3 R2 G2 R1 G1 R0 G0 \n\punpcklbw %%mm0, %%mm3 # B3 00 B2 00 B1 00 B0 00 \n\movq %%mm3, %%mm5 # R3 00 R2 00 R1 00 R0 00 \n\punpcklwd %%mm4, %%mm3 # R1 G1 B1 00 R0 G0 B0 00 \n\movq %%mm3, (%3) # Store RGBA1 RGBA0 \n\punpckhwd %%mm4, %%mm5 # R3 G3 B3 00 R2 G2 B2 00 \n\movq %%mm5, 8(%3) # Store RGBA3 RGBA2 \n\pxor %%mm6, %%mm6 # zero mm6 \n\punpckhbw %%mm1, %%mm2 # R7 G7 R6 G6 R5 G5 R4 G4 \n\punpckhbw %%mm0, %%mm6 # B7 00 B6 00 B5 00 B4 00 \n\movq %%mm6, %%mm0 # B7 00 B6 00 B5 00 B4 00 \n\punpcklwd %%mm2, %%mm6 # R5 G5 B5 00 R4 G4 B4 00 \n\movq %%mm6, 16(%3) # Store RGBA5 RGBA4 \n\punpckhwd %%mm2, %%mm0 # R7 G7 B7 00 R6 G6 B6 00 \n\movq %%mm0, 24(%3) # Store RGBA7 RGBA6 \n\"#define MMX_UNPACK_32_BGRA " \n\pxor %%mm3, %%mm3 # zero mm3 \n\movq %%mm2, %%mm4 # G7 G6 G5 G4 G3 G2 G1 G0 \n\punpcklbw %%mm0, %%mm4 # B3 G3 B2 G2 B1 G1 B0 G0 \n\punpcklbw %%mm1, %%mm3 # R3 00 R2 00 R1 00 R0 00 \n\movq %%mm3, %%mm5 # R3 00 R2 00 R1 00 R0 00 \n\punpcklwd %%mm4, %%mm3 # B1 G1 R1 00 B0 G0 R0 00 \n\movq %%mm3, (%3) # Store BGRA1 BGRA0 \n\punpckhwd %%mm4, %%mm5 # B3 G3 R3 00 B2 G2 R2 00 \n\movq %%mm5, 8(%3) # Store BGRA3 BGRA2 \n\pxor %%mm6, %%mm6 # zero mm6 \n\punpckhbw %%mm0, %%mm2 # B7 G7 B6 G6 B5 G5 B4 G4 \n\punpckhbw %%mm1, %%mm6 # R7 00 R6 00 R5 00 R4 00 \n\movq %%mm6, %%mm0 # R7 00 R6 00 R5 00 R4 00 \n\punpcklwd %%mm2, %%mm6 # B5 G5 R5 00 B4 G4 R4 00 \n\movq %%mm6, 16(%3) # Store BGRA5 BGRA4 \n\punpckhwd %%mm2, %%mm0 # B7 G7 R7 00 B6 G6 R6 00 \n\movq %%mm0, 24(%3) # Store BGRA7 BGRA6 \n\"#define MMX_UNPACK_32_ABGR " \n\pxor %%mm3, %%mm3 # zero mm3 \n\movq %%mm1, %%mm4 # R7 R6 R5 R4 R3 R2 R1 R0 \n\punpcklbw %%mm2, %%mm4 # G3 R3 G2 R2 G1 R1 G0 R0 \n\movq %%mm0, %%mm5 # B7 B6 B5 B4 B3 B2 B1 B0 \n\punpcklbw %%mm3, %%mm5 # 00 B3 00 B2 00 B1 00 B0 \n\movq %%mm4, %%mm6 # G3 R3 G2 R2 G1 R1 G0 R0 \n\punpcklwd %%mm5, %%mm4 # 00 B1 G1 R1 00 B0 G0 R0 \n\movq %%mm4, (%3) # Store ABGR1 ABGR0 \n\punpckhwd %%mm5, %%mm6 # 00 B3 G3 R3 00 B2 G2 R2 \n\movq %%mm6, 8(%3) # Store ABGR3 ABGR2 \n\punpckhbw %%mm2, %%mm1 # G7 R7 G6 R6 G5 R5 G4 R4 \n\punpckhbw %%mm3, %%mm0 # 00 B7 00 B6 00 B5 00 B4 \n\movq %%mm1, %%mm2 # G7 R7 G6 R6 G5 R5 G4 R4 \n\punpcklwd %%mm0, %%mm1 # 00 B5 G5 R5 00 B4 G4 R4 \n\movq %%mm1, 16(%3) # Store ABGR5 ABGR4 \n\punpckhwd %%mm0, %%mm2 # B7 G7 R7 00 B6 G6 R6 00 \n\movq %%mm2, 24(%3) # Store ABGR7 ABGR6 \n\"#elif defined(HAVE_MMX_INTRINSICS)/* MMX intrinsics */#include <mmintrin.h>#define MMX_CALL(MMX_INSTRUCTIONS) \ do { \ __m64 mm0, mm1, mm2, mm3, \ mm4, mm5, mm6, mm7; \ MMX_INSTRUCTIONS \ } while(0)#define MMX_END _mm_empty() #define MMX_INIT_16 \ mm0 = _mm_cvtsi32_si64(*(int*)p_u); \ mm1 = _mm_cvtsi32_si64(*(int*)p_v); \ mm4 = _mm_setzero_si64(); \ mm6 = (__m64)*(uint64_t *)p_y;#define MMX_INIT_32 \ mm0 = _mm_cvtsi32_si64(*(int*)p_u); \ *(uint16_t *)p_buffer = 0; \ mm1 = _mm_cvtsi32_si64(*(int*)p_v); \ mm4 = _mm_setzero_si64(); \ mm6 = (__m64)*(uint64_t *)p_y;#define MMX_YUV_MUL \ mm0 = _mm_unpacklo_pi8(mm0, mm4); \ mm1 = _mm_unpacklo_pi8(mm1, mm4); \ mm0 = _mm_subs_pi16(mm0, (__m64)mmx_80w); \ mm1 = _mm_subs_pi16(mm1, (__m64)mmx_80w); \ mm0 = _mm_slli_pi16(mm0, 3); \ mm1 = _mm_slli_pi16(mm1, 3); \ mm2 = mm0; \ mm3 = mm1; \ mm2 = _mm_mulhi_pi16(mm2, (__m64)mmx_U_green); \ mm3 = _mm_mulhi_pi16(mm3, (__m64)mmx_V_green); \ mm0 = _mm_mulhi_pi16(mm0, (__m64)mmx_U_blue); \ mm1 = _mm_mulhi_pi16(mm1, (__m64)mmx_V_red); \ mm2 = _mm_adds_pi16(mm2, mm3); \ \ mm6 = _mm_subs_pu8(mm6, (__m64)mmx_10w); \ mm7 = mm6; \ mm6 = _mm_and_si64(mm6, (__m64)mmx_00ffw); \ mm7 = _mm_srli_pi16(mm7, 8); \ mm6 = _mm_slli_pi16(mm6, 3); \ mm7 = _mm_slli_pi16(mm7, 3); \ mm6 = _mm_mulhi_pi16(mm6, (__m64)mmx_Y_coeff); \ mm7 = _mm_mulhi_pi16(mm7, (__m64)mmx_Y_coeff);#define MMX_YUV_ADD \ mm3 = mm0; \ mm4 = mm1; \ mm5 = mm2; \ mm0 = _mm_adds_pi16(mm0, mm6); \ mm3 = _mm_adds_pi16(mm3, mm7); \ mm1 = _mm_adds_pi16(mm1, mm6); \ mm4 = _mm_adds_pi16(mm4, mm7); \ mm2 = _mm_adds_pi16(mm2, mm6); \ mm5 = _mm_adds_pi16(mm5, mm7); \ \ mm0 = _mm_packs_pu16(mm0, mm0); \ mm1 = _mm_packs_pu16(mm1, mm1); \ mm2 = _mm_packs_pu16(mm2, mm2); \ \ mm3 = _mm_packs_pu16(mm3, mm3); \ mm4 = _mm_packs_pu16(mm4, mm4); \ mm5 = _mm_packs_pu16(mm5, mm5); \ \ mm0 = _mm_unpacklo_pi8(mm0, mm3); \ mm1 = _mm_unpacklo_pi8(mm1, mm4); \ mm2 = _mm_unpacklo_pi8(mm2, mm5);#define MMX_UNPACK_15 \ mm0 = _mm_and_si64(mm0, (__m64)mmx_mask_f8); \ mm0 = _mm_srli_pi16(mm0, 3); \ mm2 = _mm_and_si64(mm2, (__m64)mmx_mask_f8); \ mm1 = _mm_and_si64(mm1, (__m64)mmx_mask_f8); \ mm1 = _mm_srli_pi16(mm1, 1); \ mm4 = _mm_setzero_si64(); \ mm5 = mm0; \ mm7 = mm2; \ \ mm2 = _mm_unpacklo_pi8(mm2, mm4); \ mm0 = _mm_unpacklo_pi8(mm0, mm1); \ mm2 = _mm_slli_pi16(mm2, 2); \ mm0 = _mm_or_si64(mm0, mm2); \ mm6 = (__m64)*(uint64_t *)(p_y + 8); \⌨️ 快捷键说明
复制代码
Ctrl + C
搜索代码
Ctrl + F
全屏模式
F11
切换主题
Ctrl + Shift + D
显示快捷键
?
增大字号
Ctrl + =
减小字号
Ctrl + -