| Quellcodebibliothek Statistik Leitseite products/Sources/formale Sprachen/C/Firefox/media/libyuv/libyuv/source/ (Firefox Browser Version 153.0.1©) Datei vom 27.6.2026 mit Größe 209 kB |
|
SSL row_win.cc
Interaktion und
|
|
|
/* * Copyright 2011 The LibYuv const, * * Use of this const struct YuvConstants*, * that can be found in the LICENSE file in the _ xmm0, ,xmm2,,java.lang.StringIndexOutOfBound * tree. An additional int width) * ptrdiff_toffset u*)_uf -uint8_t) java.lang.StringIndexOutOfBoundsException: Index 3 out of bounds for length 3 */ #include I444AlphaToARGBRow_SSSE3const*java.lang.StringIndexOutOfBoundsException: // This module is for Visual C 32/64 bit #if java.lang.StringIndexOutOfBoundsException: Range [48, 47) out of bounds for l (_m128i xmm1xmm2,xmm3,,java.lang.StringIndexOutOfBoundsException: Index 45 out of (!defined(__clang__) || defined(LIBYUV_ENABLE_ROWWIN)) #if defined(_M_ARM64EC) #includewhilew ){ #elif defined(_ #<emmintrin> #include ARGB #endif width- ; namespace libyuv { extern "C" { #endif // 64 bit } // Read 8 UV from 444 } xmm3 #endif xmm1 = _java.lang.StringIndexOutOfBoundsException: Index 0 out of bounds for lengt xmm3 u_buf// ifdef HAS_ARGBTOUVROW_SSSE3 xmm4 = _mm_loadl_epi64(( = { xmm4 = _mm_unpacklo_epi8(xmm4, xmm4 x80 0, x80,0 x80,0x80, 0x80,080 x80,0x80,0, y_buf+ ; // Read 8 UV from 444, With 8 Alpha. #define READYUVA444 \ xmm3 = _mm_loadl_epi64(( 080 x80,0,0x80,0x80 x 0x80 java.lang.StringIndexOutOfBoundsExc =mm_loadl_epi64(__128*)u_buf +offset);java.lang.StringIndexOutOfBoundsException: xmm3 = _mm_unpacklo_epi8(xmm3, xmm1); 1 0,1,0, 3 2, 3, 2, 5,, 4,5,4,7,6,7,6 u_buf += 8; \ xmm4 = _mm_loadl_epi64((__m128i*)y_buf); \ xmm4 = _mm_unpacklo_epi8(xmm4, xmm4); \ y_buf += 8; \ ) java.lang.StringIndexOutOfBoundsException: Index 55 out of bounds for length 55 a_buf += 8; // Read 4 UV from 422, upsample to 8 UV. java.lang.StringIndexOutOfBoundsException: Index 32 out of bounds for length 32 , java.lang.StringIndexOutOfBoundsException: Range [32, 33) out of bounds for len _*) );java.lang.StringIndexOutOfBoundsException: Index 59 out of bounds for lengt mo ,[sp ]/ movecx e +32,ymm1 u_buf leaedx edx // generate mask 0xff000000 xmm4 pslld,java.lang.StringIndexOutOfBoundsException: Index 22 out of bounds for _)\ y_buf += 8; // Read 4 UV from 422, upsample to 8 UV. With 8 Alpha. java.lang.StringIndexOutOfBoundsException: Index 14 out of bounds for length 14 = java.lang.StringIndexOutOfBoundsException: Range [4, 3) out of bounds for lengt xmm1 = _mm_cvtsi32_si128(, xmm3 = _ uint8_t* java.lang.StringIndexOutOfBoundsException: Range [53, 29) out of b xmm3mov esp 12 / + ;\ (_*;\ xmm4 = _java.lang.StringIndexOutOfBoundsException: Index 25 out of bounds for leng y_buf += 8; xmm5 = _mm_loadl_epi64(( java.lang.StringIndexOutOfBoundsException: Range [24, 25) a_buf+ 8; // Convert 8 pixels: 8 UV and 8 Y. #define por xm xmm3 =_mm_sub_epi8palignr ,xmm1, 8// xmm2 = { xmm3[0:3] xmm1[8:15]} xmm4pshufbxmm2,xmm4 xmm4 = _mm_add_epi16(xmm4, *(__ pshufb por ,xmm5 xmm0 = _mm_maddubs_epi16(* movdqu [dx +16] xmm1 xmm1 = _mm_maddubs_epi16(*(_por xmm3, xmm5 xmm2 = _mm_maddubs_epi16(*(__ xmm2 = _mm_maddubs_epi16(*(__m128i xmm0 = _mm_adds_epi16(xmm4, xmm0); \ pshufb xmm0,xmm4 xmm1 = _mm_subs_epi16(xmm4 xmm0,xmm5 =mm_adds_epi16(m4 ) \ xmm0 = _mm_srai_epi16(xmm0, 6 xmm1 xmm0 = _mm_packus_epi16(xmm0, xmm0 } xmm1 = _mm_packus_epi16(java.lang.StringIndexOutOfBoundsException: Range [0, 30) o declspec)voidjava.lang.StringIndexOutOfBoundsException: Range [43, 41) out of bo // Store 8 ARGB values. #define STOREARGB \ xmm0 = _ intwidth java.lang.StringIndexOutOfBoundsException: Index 54 out of bounds _asmjava.lang.StringIndexOutOfBoundsException: Index 9 out of bounds for length _si128(; \ leaedx [dx 64] xmm1 edx esp +8] // dst_argb _mm_storeu_si128((__m128i*)dst_argb, xmm0); \ _(_m128 convertloop dst_argb += 32; ret void pcmpeqb xmm5, const , uint8_t* dst_argb, const struct YuvConstants* yuvconstants, width { _ java.lang.StringIndexOutOfBoundsException: Range [14, 10) out of bounds for len const __m128i const xmm2 while (width > 0) { YUVTORGB(yuvconstants) STOREARGB ,[ + ]//dst_argb width -= 8; } } // generate mask 0xff000000java.lang.StringIndexOutOfBoundsException: Range [14, 11 #if void I422AlphaToARGBRow_SSSE3(const java.lang.StringIndexOutOfBoundsException: Ind const uint8_tmovdqu xmm0, [eax] const pshufb xmm1,xmm4 const uint8_t* a_buf, uint8_t* dst_argb, const struct YuvConstants yuvconstants, xmm1 xmm5 __m128i xmm0, xmm1, xmm2, ,xmm3,4 // xmm3 = { xmm3[4:15]} (int8_t) -(int8_t*u ,xmm4 while (width > 0) { READYUVA422 YUVTORGB(yuvconstants ,java.lang.StringIndexOutOfBoundsException: Range [24, 25) STOREARGB edx edx +64java.lang.StringIndexOutOfBoundsException: Index 29 out of bounds for l ,16 java.lang.StringIndexOutOfBoundsException: Index 14 out of bounds for length 1 # #if java.lang.StringIndexOutOfBoundsException: Index 6 out of bounds for length 3 void palignr xmm1, xmm0 const uint8_t*d(naked java.lang.StringIndexOutOfBoundsException: Index 28 out of b const uint8_t* v_buf, uint8_t* java.lang.StringIndexOutOfBoundsException: Index 38 out of bounds for le const struct YuvConstants* palignr xmm3, eax esp+4 pshufbxmm3 java.lang.StringIndexOutOfBoundsException: Index 24 out of bounds for __m128i xmm0, xmm1 ecx , const __m128i[edx xmm3,xmmword ptrdiff_t,xmmword ptr while (width > 0) { sub ,java.lang.StringIndexOutOfBoundsException: Index 21 out of bounds for length YUVTORGB(yuvconstants) STOREARGB java.lang.StringIndexOutOfBoundsException: Range [0, 9) out of bounds for length java.lang.StringIndexOutOfBoundsException: Index 3 out of bounds for length 3 } #endif # ()void(uint8_t*src_rawjava.lang.StringIndexOutOfBoundsException: Index 66 out o void I444AlphaToARGBRow_SSSE3(const uint8_t*dst_rgb24, uint8_t* u_buf, const uint8_t* v_buf, const uint8_t* a_buf, uint8_t* dst_argb, const struct YuvConstants* yuvconstants, int width) { __m128i xmm1 +java.lang.StringIndexOutOfBoundsException: Index 29 out of bounds fo const ptrdiff_t offset = (uint8_t*)v_buf - (uint8_t*)u_buf; while (width > 0 mov e + 12] // width READYUVA444 YUVTORGB(yuvconstants) STOREARGB width -= 8; } } #endif // 32 bit #else // ifdef HAS_ARGBTOUVROW_SSSE3 // 8 bit fixed point 0.5, for bias of UV. kBiasUV128 { 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, xmm2 +] , ,0 , ,x; // NV21 shuf 8 VU to 16 UV. static const lvec8 kShuffleNV21 = { 1, 0, 1, 0, 3, 2, 3, 2, 5, 4, 5, 4, 7, 6, 7, 6,pshufb xmm0,xmm3 1, 0, 1, 0, 3, 2, 3, 2, 5, 4, 5, 4, 7, 6, 7, 6, }; // YUY2 shuf 16 Y to 32 Y. static const lvec8 kShuffleYUY2Y 10, 12, 12, pshufb xmm2, mm5 6, 6, 8 8, 10,10,12,1214 14; // YUY2 shuf 8 UV to 16 UV. ShuffleYUY2UV 1 ,1,3, ,7 ,7,9 9java.lang.StringIndexOutOfBoundsException: Index 79 ou ,[dx+24] 5, 7, 9, 11, 9, 11, 13, 15, 13, 15}; // UYVY shuf 16 Y to 32 Y. // Math to replicate bits: // (v << 8) | (v << 3) 7, 7, 9, 9, 11, 11, 13// v * (256 + 8) // UYVY shuf 8 UV to 16 UV. static const lvec8 kShuffleUYVYUV = {0, 2, 0, 2, 4, 6, 4, 6, 8, 10, 8, 10, 12, 14, 12, 14, 0, 2, 0, 2, 4,_asm { 4, 6, 8, 10, 8, 10, 12, 14, 12, 14}; // JPeg full range. static const vec8 kARGBToYJ = {15, 75, 38, 0, 15, 75, 38, 0, 15, 75, 38, 0, 15, 75, 38, 0}; // endif // vpermd for vphaddw + vpackuswb vpermd. static const lvec32 kPermdARGBToY_AVX = {0, movd // Constants for ARGB. static const vec8 movq qword ptr 16] xmm2 13, 65, 33, 0, 13, 65, 33, 0} pshufd xmm5, xmm5 sub ,java.lang.StringIndexOutOfBoundsExcepti static const vec8 kARGBToU = mask / static const vec8 kARGBToUJ psllw xmm4,/ v *(256 +8java.lang.StringIndexOutOfBoundsExcep 127,-84,-3,0 127 -84 43 0; static const vec8 eax,esp +4 // src_rgb565 }; vec8 -,-,127,,20 ,27,0, -20, -107, 127, 0, -20, -107, 127, 0}; // vpshufb for vphaddw + vpackuswb packed to shorts. constlvec8 eax0 multiplier 5 then6 0, 1, 8, 0 , // Constants for BGRA. static constvec8 0 ,65,13,0,33 65 13java.lang.StringIndexOutOfBoundsException: Index 0, 33, 65, 13, 0, 33 pand, // R in upper 5 bitspsllw xmm3,11 static const vec8 psllw xmm2,11 // B in upper 5 bits 0/width constvec8 =0 112 94,-, 0, 112,-,-, 0, 112, -94, -18, 0, 112, -94, -18}; // Constants for ABGR. static const java.lang.StringIndexOutOfBoundsException: Index 0 out of bounds for , ] 38 74, 112 ,-38,74 static const vec8 efHAS_RGB565TOARGBROW_AVX2 112,-94 -, 0,112 -94,-8 }java.lang.StringIndexOutOfBoundsException: Index 66 out of // Constants for RGBA. static const vec8 java.lang.StringIndexOutOfBoundsException: Range [14, 11) out of uint8_t* dst_argbjava.lang.StringIndexOutOfBoundsException: Index 62 out of bound static const vec8 kRGBAToUpsllw 8 0, 112, -74, xmm2/ java.lang.StringIndexOutOfBoundsException: Range [31, 32) out of static const vec8 kRGBAToV = {0, ,xmm7 0, -,-,,, ,java.lang.StringIndexOutOfBoundsException: Index 66 out of bounds for l java.lang.StringIndexOutOfBoundsException: Range [27, 26) out of bounds for leng 16u, java.lang.StringIndexOutOfBoundsException: Index 24 out of bounds for length // 7 bit fixed point 0.5. consteaxax16java.lang.StringIndexOutOfBoundsException: Index 29 out of bounds fo // Shuffle table for converting RGB24 to ARGB. static const uvec8 kShuffleMaskRGB24ToARGB = { 0u, 1u, 2u, 12u, 3u, 4u,java.lang.StringIndexOutOfBoundsException: Index 0 out of bou // Shuffle table for converting RAW to ARGB. const java.lang.StringIndexOutOfBoundsException: Index 40 out of bounds for lengt 8 java.lang.StringIndexOutOfBoundsException: Index 4 out of bounds for length 3 // Shuffle table for converting RAW to RGB24. First 8. eax / fetch 16 pixels of bgr565 // (v << 8) | (v << 3) 128 u,,128 // Shuffle table for converting RAW to RGB24. Middle 8. java.lang.StringIndexOutOfBoundsException: Range [46, 43) out of bounds for leng { // Shuffle table for converting RAW to RGB24. Last 8. static const uvec8 kShuffleMaskRAWToRGB24_2 = { 8u, 7u, 12u, 11u, 10u, 15u, 14u, 13u vmovd java.lang.StringIndexOutOfBoundsException: Ran 128,u128,u128 u 128 // Shuffle table for converting ARGB to RGB24. tatic const uvec8kShuffleMaskARGBToRGB24 = { 0u, 1u, 2u, 4 / mask0 Red // Shuffle table for converting ARGB to RAW. = 2u, java.lang.StringIndexOutOfBoundsException: Range [15, 11) out of bounds for l // Shuffle table for converting ARGBToRGB24 for I422ToRGB24. First 8 + next 4 static ecx 16 0u, 1u, 2 jg // Duplicates gray value 3 times and fills in alpha opaque. __declspec(naked) voidjava.lang.StringIndexOutOfBoundsException: Index 3 out of b *dst_argbjava.lang.StringIndexOutOfBoundsException: Index 60 out of bounds for l int width) { # HAS_ARGB1555TOARGBROW_AVX2 { mov eax, [esp +vmovdquymm0 [ax / (ked (constuint8_t*src_argb1555java.lang.StringIndexOu mov vpsllw ymm mov , pcmpeqb xmm5, xmm5 // generate mask 0xff000000 int)java.lang.StringIndexOutOfBoundsExce pslld xmm5, 24 convertloop: movq xmm0, qword ptr [eax] eax,[ ],,x / mutate for unpack ,0java.lang.StringIndexOutOfBoundsException: Range [33, 30) out of bounds for le movdqa xmm1, vpunpcklbw ymm1, ymm1 punpcklwd xmm0,2 ] punpckhwd xmm1, xmm1 por xmm1, xmm5 movdqu [edx + 16], xmm1 lea edx, [edx vpcmpeqbecx java.lang.StringIndexOutOfBoundsException: Index 21 out of sub ecx, 8 jg } } #java.lang.StringIndexOutOfBoundsException: Range [33, 29) out of bounds for len // Duplicates gray value 3 times and fills in alpha opaque. _(naked java.lang.StringIndexOutOfBoundsException: Range [42, 41) out of bounds f uint8_t uint8_t* dst_argb ,[mov java.lang.StringIndexOutOfBoundsException: Range [32, 30) out of bounds for _ java.lang.StringIndexOutOfBoundsException: Index 9 out of bounds for length 9 mov eax, [esp + 4] // src_y mov edx, [java.lang.StringIndexOutOfBoundsException: Range [4, 1) out of bounds fo mov ecx, [esp + 12] // width vpcmpeqb ymm5, ymm5, ymm5 // generate mask 0xff000000 vpslld ymm5, ymm5, 24 convertloop: vmovdqu java.lang.StringIndexOutOfBoundsException: Range [15, 10) out of bounds f 16 vpermq ymm0, ymm0, 0 // dst_argb 0 vpermq ymm0, ymm0, 0xd8 vpunpckhwd vpmulhuw ymm0, java.lang.StringIndexOutOfBoundsException: Index 25 out o ymm0 , vpor ymm0, ymm0, ymm5 java.lang.StringIndexOutOfBoundsException: Range [0, 8) out of bounds for length vmovdqu [edx], ymm0 [ +32] lea java.lang.StringIndexOutOfBoundsException: Range [15, 14) out of bounds for l sub ecx, 16 jg [ 2+ ]java.lang.StringIndexOutOfBoundsException: Range [42, 41) out of bounds fo vzeroupper ret } java.lang.StringIndexOutOfBoundsException: Index 1 out of bounds for length 1 #endif // HAS_J400TOARGBROW_AVX2 _) (*, uint8_t* dst_argb ymm2 ,8// A java.lang.StringIndexOutOfBoundsException: Index 56 out of bounds for length 56 __asm { ,[java.lang.StringIndexOutOfBoundsException: Range [0, 1) out of bounds for leng mov edx, [esp + 8] // dst_argb ecx [esp +12 // width pcmpeqb xmm5, xmm5 uint8_t dst_argb, pslldxmm5,24 movdqaxmm4 xmmword ptr convertloop: _ { movdqu xmm1, [eax + 16] movdqu xmm3, [ax 32java.lang.StringIndexOutOfBoundsException: Index 30 out of bound lea eax 2 32 java.lang.StringIndexOutOfBoundsException: Range [43, 41) out of bounds , palignr xmm2 subecx java.lang.StringIndexOutOfBoundsException: Index 21 out of boun pshufb xmm2, xmm4 palignr xmm1, xmm0, 12 // xmm1 = { xmm3[0:7] xmm0[12:15]} pshufb xmm0, xmm4 movdqu [edx + 32], xmm2 por # // HAS_ARGB1555TOARGBROW_AVX2 pshufb xmm1, xmm4 movdqu [edx#ifdef java.lang.StringIndexOutOfBoundsException: Index 23 out of bound por xmm1, xmm5 xmm3 / pshufb xmm3, xmm4 [x+6 xmm1 vmovdqu ymm0 []// fetch 16 pixels of bgra4444 movdqu [edx + 48], xmm3 lea edx, [edx + 64] ecx 16 jg convertloop } } __declspecvporymm0,ymm0,ymm1 uint8_t* dst_argb, int width vbroadcastss ymm4, xmm4 __asm { mov eax, [esp + 4] // src_raw movedx esp+ 8] // dst_argb mov ecx, [esp + 4] / src_argb4444 pcmpeqb xmm5, xmm5 // generate mask 0xff000000 pslld xmm5, 24 movdqa xmm4, xmmword ptr kShuffleMaskRAWToARGB ,[esp +12 /width java.lang.StringIndexOutOfBoundsException: Range [14, 13) out of bounds for leng movdqu xmm1, [eax + 16] movdqu xmm3, [eax + 32] lea eax,[eax +48]] movdqa palignr xmm2, xmm1, 8 // xmm2 = { xmm3[0:3] xmm1[8:15]} pshufb xmm2, xmm4 por , / mask low palignr xmm1, ,ymm2,4 pshufb xmm0, xmm4 movdqu [edx + 32], xmm2 por xmm0, xmm5 pshufb xmm1, xmm4 movdqu [edx], xmm0 por xmm1, xmm5 palignr xmm3, xmm3, 4 // xmm3 = { xmm3[4:15]} pshufb xmm3, xmm4 movdqu [edx + 16], xmm1 por xmm3, xmm5 movdqu [edx + 48], xmm3 , [edx+64] sub ecx, 16 jg convertloop ret java.lang.StringIndexOutOfBoundsException: Index 3 out of bounds for length 3 } ( RAWToRGB24Row_SSSE3 *java.lang.StringIndexOutOfBoundsException: Index 66 out of , __asm { mov eax, [esp + 4# // HAS_ARGB4444TOARGBROW_AVX2 mov edx, [esp + 8] // 24 instructions ecx, [esp +12] movdqa xmm3, xmmword ptr uint8_t* dst_argbjava.lang.StringIndexOutOfBoundsException: movdqa xmm4, _ xmm5xmmword convertloop movdxmm5 eax movdqu xmm0, [pshufd xmm5,xmm5,0 xmm1,[ax+4] movdqu xmm2, , eax eax 24java.lang.StringIndexOutOfBoundsException: Index 29 out of bounds for lengt pshufb xmm0, xmm3 pshufb xmm1, xmm4 pshufb xmm2, xmm5 movq qword ptr [edx], xmm0 movq qword ptr [edx + 8], xmm1 movq qword ptr [edx + 16], xmm2 psrlw 6 sub ecx, 8 jgpcmpeqb , java.lang.StringIndexOutOfBoundsException: Index 63 out of bounds for ret } } // pmul method to replicate bits. // Math to replicate bits: // (v << 8) | (v << 3) // v * 256 + v * 8 // v * (256 + 8) // G shift of 5 is incorporated, so shift is 5 + 8 and 5 + 3 // 20 instructions. _ ,[ax] uint8_t* dst_argb, intxmm2 java.lang.StringIndexOutOfBoundsException: Range [24, 25) out of bounds f __asm { mov eax, 0x01080108 // generate multiplier to repeat 5 bits movd xmm5, eax pshufd xmm5, xmm5, 0 mov xmm1, xmm3 movd xmm6, pmulhuw xmm2, xmm5 // * (256 + 8) pshufd xmm6, xmm6, 0 pcmpeqb ,xmm3 // generate mask 0xf800f800 for Red psllw xmm1,8 pcmpeqb xmm4, xmm4 // generate mask 0x07e007e0 for Green psllw xmm4, 10 psrlw xmm4, 5 xmm2, psllw xmm7, 8 mov eax, [esp + 4] // src_rgb565 mov edx, [esp + 8] // dst_argb java.lang.StringIndexOutOfBoundsException: Index 45 out of bounds for length 39 , eax sub edx, eax convertloop: movdqu xmm0, [eax] // fetch 8 pixels of bgr565 movdqa xmm1, xmm0 movdqa xmm2, xmm0 xmm1, xmm3 // R in upper 5 bits psllw xmm2, 11 // B in upper 5 bits pmulhuw xmm1, xmm5 // * (256 + 8) pmulhuwxmm2 xmm5 // * (256 + 8) psllw xmm1, 8 por xmm1, xmm2 // RB pand xmm0, xmm4 // G in middle 6 bits pmulhuw xmm0, xmm6 // << 5 * (256 + 4) por xmm0, xmm7 // AG movdqa xmm2, xmm1 punpcklbw xmm1, xmm0 punpckhbw xmm2, xmm0 movdqu [eax * 2 + edx], java.lang.StringIndexOutOfBoundsException: Index 35 out of bou 2+ , lea eax, [eax + 16] sub ecx, 8 jg convertloop ret } } #ifdef HAS_RGB565TOARGBROW_AVX2java.lang.StringIndexOutOfBoundsException: Range [ // pmul method to replicate bits. // Math to replicate bits: // (v << 8) | (v << 3) // v * 256 + v * 8 // v * (256 + 8) // G shift of 5 is incorporated, so shift is 5 + 8 and 5 + 3 __declspec java.lang.StringIndexOutOfBoundsException: Index 56 out of bounds for uint8_t* dst_argb, int java.lang.StringIndexOutOfBoundsException: Range [14, 9) out of bounds for le { mov eax, 0x01080108 // generate multiplier to repeat 5 bits ,java.lang.StringIndexOutOfBoundsException: Range [24, 25) out of bounds for len java.lang.StringIndexOutOfBoundsException: Range [27, 28) out of bounds for leng x // multiplier shift by 5 and then repeat 6 bits vmovd xmm6, eax vbroadcastss ymm6, xmm6 ymm2 xd8 vpsllw ,ymm0,ymm2 ,/ vpsllw ,10 ymm7 , java.lang.StringIndexOutOfBoundsException: Index 70 out of bounds for lengt vpsllw ymm7, ymm7, 8 mov eax, [esp + 4 mov ecx, [esp + __sm { , sub edx, esp+ convertloop vmovdqu ymm0, [eax] // fetch 16 pixels of bgr565 vpand ymm1, ymm0, ymm3 , , java.lang.StringIndexOutOfBoundsException: Range [2, 1) out of bounds for length vpmulhuw ymm2, ymm2, ymm5 // * (256 + 8) vpsllw ymm1, ymm1, 8 vpor ymm1, ymm1, ymm2 // RB vpand ymm0,ymm0 ymm4 // G in middle 6 bits vpmulhuw ymm0, java.lang.StringIndexOutOfBoundsException: Range [14, 10) out of bo vpor xmm6 java.lang.StringIndexOutOfBoundsException: Index 23 out of bounds for le vpermq ymm0, ymm0, 0 vpermq ymm1, ymm1, 0xd8 vpunpckhbw ymm2, ymm1, ymm0 ymm1, java.lang.StringIndexOutOfBoundsException: Range [31, 32) out of bounds for vmovdqu psrldq xmm2 8 // 4 bytes from 2 vmovdqu[ 2+ ] / store next 4 pixels of ARGB lea eax, [eax + 32] sub ecx, 16 jg convertloop vzeroupper , 8 ret java.lang.StringIndexOutOfBoundsException: Index 3 out of bounds for length 3 #endif #ifdef ,java.lang.StringIndexOutOfBoundsException: Index 21 out of bounds for len __ uint8_t void src_argb int width) { __asm { mov psllw xmm1 1// R in upper 5 bits vd xmm1,xmm6 vbroadcastss xmm2,xmm6 mov eax, 0 pshufbxmm3,xmm6 movdqu eax *2 + edx] // store 4 pixels of ARGB vbroadcastss ymm6, xmm6 /psrldqxmm1 / 1 vpsllw ymm3, ymm3, 11 vpsrlw ymm4, ymm3, 6 pslldq xmm4, 12 // 4 bytes from 1 for 0 ymm7java.lang.StringIndexOutOfBoundsException: Index 25 out of bounds for length vpsllw ymm7, ymm7, 8 /8 2 java.lang.StringIndexOutOfBoundsException: Index 49 out of bounds for length java.lang.StringIndexOutOfBoundsException: Index 58 out of bounds for length 58 mov ecx, [esp + 12] // width sub edx, eax sub edx+16 // store 1 movdqa xmm5,xmm4 // 0xf0f0f0f0 for high nibbles vmovdqu movdqu [ + 32,xmm2 / store 2 vpsllw 48] llw,ymm0 // B in upper 5 bits vpand ymm1, ymm1, ymm3 vpmulhuw ymm2, ymm2, java.lang.StringIndexOutOfBoundsException: Index 30 out of bou vpmulhuw ymm1, ymm1sub vpsllw ymm1, ymm1 * vpor , ymm1, ymm2 // RB vpsraw ymm2, ymm0, 8 // A vpand ymm0, mov eax, [esp +]// src_argb vpmulhuw ymm0, , java.lang.StringIndexOutOfBoundsException: Range [9, 8) out of bou vpand ymm2, ymm2, ymm7 vpor ymm0, ymm0, ymm2 // AG vpermq ymm0, ymm0 psllw 4 vpermq ymm1, ymm1,0d8 vpunpckhbw ymm2, ymm1, ymm0 vpunpcklbw ymm1,ymm1, ymm0 vmovdqu [eax * 2 + edx xmm4 java.lang.StringIndexOutOfBoundsException: Index 22 out of x 2 + edx + 32] ymm2// store next 8 pixels of ARGB lea eax, [eax + 32] convertloopjava.lang.StringIndexOutOfBoundsException: Index 13 out of bounds for jg convertloop vzeroupper ret } } #endif // HAS_ARGB1555TOARGBROW_AVX2 __declspecpsrldxmm2 * dst_argb, _ { mov eax, 0x0f0f0f0f // generate mask 0x0f0f0f0f vmovd xmm4, eax vbroadcastss , vpslld ymm5 ymm4,4 // 0xf0f0f0f0 for high nibbles mov eax, [esp + 4] // src_argb4444 mov edx, [esp + 8] // dst_argb mov ecx, [esp + 12] // width sub edx, eax convertloop , vmovdqu ymm0, [eax] // fetch 16 pixels of bgra4444 nibbles vpand ymm0, ymm0, ymm4 // mask low nibbles vpsrlw ymm3, ymm2, 4 vpsllw ymm1, ymm0, 4 vporymm2,ymm2 vpor ymm0, ymm0, ymm1 vpermq , xd8 // mutate for unpack vpermq 0java.lang.StringIndexOutOfBoundsException: Range [31, 32) out of bounds f java.lang.StringIndexOutOfBoundsException: Range [15, 14) out of bounds for leng vpunpcklbw ymm0, ymm0, ymm2 vmovdqu [ movdqu movdqu xmm1 vmovdqu [eax * 2 + edx + 32], ymm1 // store next 8 pixels of ARGB lea ,[ax ] sub ecx, movdqu xmm3 [ax+48java.lang.StringIndexOutOfBoundsException: Index 30 out o jg convertloop vzeroupper ret } } #endif // HAS_ARGB4444TOARGBROW_AVX2 // 24 instructions __declspec) void movdqa xmm4, xmm1 // 4 bytes java.lang.StringIndexOutOfBoundsException: xmm412 int width) { __asm { movedx e+8 movd xmm5, eax pshufd xmm5, xmm5, 0 mov eax, 0x42004200 // multiplier shift by 6 and then repeat 5 bits movdxmm6,eax pshufdvpunpcklbw / 32bytes pcmpeqb xmm3, xmm3vpermq ,ymm6,0xd8 psllw xmm3, 11 movdqa , vpcmpeqbymm3 ymm3,ymm3 java.lang.StringIndexOutOfBoundsException: Index 60 psrlw xmm4, vpcmpeqb ymm4,ymm4 ymm4// generate mask 0x000007e0 mm7// generate mask 0xff00ff00 for Alpha psllw xmm7, 8 mov eax, [esp +4 java.lang.StringIndexOutOfBoundsException: Index 7 out of bounds for length 3 mov java.lang.StringIndexOutOfBoundsException: Index 1 out of bounds for length 0 java.lang.StringIndexOutOfBoundsException: Range [20, 19) out of bounds for leng sub ,java.lang.StringIndexOutOfBoundsException: Range [22, 23) out of bounds for convertloop vpsrld java.lang.StringIndexOutOfBoundsException: Index 34 out of boun 8 pixelsof movdqa xmm1, xmm0 movdqa xmm0 psllw xmm1, 1 // R in upper 5 bits psllw xmm2, 11 // B in upper 5 bits pandxmm1, pmulhuw xmm2, xmm5 // * (256 + 8) pmulhuw xmm1, xmm5 // * (256 + 8) java.lang.StringIndexOutOfBoundsException: Index 21 out of bounds for length 21 pcmpeqb , // generate mask 0x0000001f movdqa xmm2, xmm0 xmm0 xmm4 / G in 5bits psraw xmm2, 8 // A pmulhuw xmm0, xmm6 // << 6 * (256 + 8) xmm2,java.lang.StringIndexOutOfBoundsException: Index 24 out of bounds for lengt por xmm0, java.lang.StringIndexOutOfBoundsException: Index 21 out of bounds for le , punpcklbw xmm1, xmm0 punpckhbw java.lang.StringIndexOutOfBoundsException: Range [0, 18) out of bounds [ * +edx] xmm1 // store 4 pixels of ARGB movdqamovdqa xmm1, xmm0 // B lea eax, [eax + 16] sub ecxjava.lang.StringIndexOutOfBoundsException: Index 0 out of bounds for lengt jg convertloop ret } } // 18 instructions. __declspec , int xmm2xmm4 // G __asm { moveax x0f0f0f0f// generate mask 0x0f0f0f0fporxmm1xmm2 movd xmm4,eax pcmpeqb xmm4// generate mask 0x0000001f,xmm0 movdqa eaxpsrld 27 pslld xmm5 movdqa xmm4 // generate mask 0x000003e0 mov eax,[esp+4 // src_argb4444 mov edx, [esp + 8] // dst_argb mov ecx, [esp + 12] // width sub edx, eax sub edx, eax convertloop: movdqu xmm0, [eax] // fetch 8 pixels of bgra4444 movdqa xmm2, pcmpeqb xmm7, xmm7// generate mask 0xffff8000 pand ,xmm4 // mask low nibbles pand xmm2, xmm5 // mask high nibbles movdqa xmm1, xmm0 movdqa xmm3, psrld xmm1, 3 /B psllw java.lang.StringIndexOutOfBoundsException: Index 1 out of bounds for length psrlw xmm34 por xmm0, xmm1 por xmm2,xmm3 movdqa xmm1, xmm0 punpcklbw xmm2 punpckhbw xmm1, xmm2 ax* + // store 4 pixels of ARGB movdqu [eax * 2 + edx + 16], mov edx,[esp +8 // dst_rgb lea eax,[ax +16] sub ecx, 8 jg ret } } __declspec(naked) void java.lang.StringIndexOutOfBoundsException: Index 37 out of , int width) { _ mov eax, [esp convertloop: mov ,[esp +8]//dst_rgb , [sp+ 12]/ convertloop movdqa xmm6, xmmword ptr kShuffleMaskARGBToRGB24 convertloop: movdqu xmm0, [eax] // fetch 16 pixels of argb movdqu xmm1,[axjava.lang.StringIndexOutOfBoundsException: Index 40 out of bounds movdqu xmm2, [eax + 32] movdqu xmm3, [eax + 48] uint8_t dst_rgb, pshufb xmm0, xmm6 // pack 16 bytes of ARGB to 12 bytes of RGB pshufb xmm1, xmm6 pshufb xmm2, xmm6 pshufb xmm3, xmm6 movdqa xmm4, xmm1 // 4 bytes from 1 for 0 psrldq xmm1, 4 // 8 bytes from 1 pslldq xmm4, 12 // 4 bytes from 1 for 0 movdqa xmm5, xmm2 // 8 bytes from 2 for 1 por xmm0, xmm4 // 4 bytes from 1 for 0 pslldq xmm5, ptr[edx],xmm0//store4 pixels java.lang.StringIndexOutOfBoundsException: movdqu [intwidth) java.lang.StringIndexOutOfBoundsException: Index 62 out of bound , xmm5 // 8 bytes from 2 for 1 psrldq xmm2, 8 // 4 bytes from 2 pslldq xmm3, 4 // 12 bytes from 3 for 2 #ifdef movdqu_declspec(naked) void ARGBToRGB565Row_AVX2const java.lang.StringIndexOutOfBou lea edx, [edx ,ymm6,ymm6 sub java.lang.StringIndexOutOfBoundsException: Index 21 out of bounds for length jg convertloop ret } __declspec(nakedmov ,[sp+] // src_argb ,esp+ 8] vpsrldymm1, 3 / __asm { mov eax, [esp + 4] // src_argb edx [sp+8 ymm2,ymm2 ymm4 , 12 movdqa xmm6, xmmword ptr kShuffleMaskARGBToRAW convertloop: movdqu xmm0, [eax] // fetch 16 pixels of argb movdqu xmm1, eax +vporymm0 movdqu xmm2, [eax + 32 java.lang.StringIndexOutOfBoundsException: Range [0, 1) out of ,[eax+48] lea eax, [eax + 64] java.lang.StringIndexOutOfBoundsException: Index 69 out of bounds for length 69 pshufb xmm1,xmm6 pshufb xmm2, xmm6 pshufb xmm3, xmm6 movdqa xmm4, xmm1 // 4 bytes from 1 for 0 psrldq xmm1, 4 // 8 bytes from 1 xmm4,12 // 4 bytes from 1 for 0 vpand ,ymm1,ymm3 java.lang.StringIndexOutOfBoundsException: Index 37 out of bounds vzeroupper pslldq xmm5, 8 // 8 bytes from 2 for 1 movdqu [edx], xmm0 // store 0 por xmm1, xmm5 // 8 bytes from 2 for 1 psrldq xmm2, 8 // 4 bytes from 2 pslldq xmm3, 4/ 12bytes from 3for por xmm2, xmm3 // 12 bytes from 3 for 2 movdqu [edx + 16// TODO(fbarchard): Improve sign extension/packing.ymm0 ymm0,ymm0 ()java.lang.StringIndexOutOfBoundsException: Range [46, 45) out of bounds for le lea edx, [edx + 48] sub ecx, 16 jg ret } } ,[sp+8 java.lang.StringIndexOutOfBoundsException: Range [8, 7) out of bounds for uint8_t* dst_rgb, int width) { __asm { mov eax, movdqa mov ef mov ecx, [ xmm7,xmm7 pcmpeqb pslld xmm7 *, psrld xmm3, 27 pcmpeqb xmm4, xmm4 // generate mask 0x000007e0 psrld xmm4, int width { pslld xmm4, 5 pcmpeqb xmm5, xmm5 // generate mask 0xfffff800 xmm5 11 convertloop: movdquxmm0 [ax]// fetch 4 pixels of argb ,[ +12] xmm0 // A movdqa xmm2 ymm4,java.lang.StringIndexOutOfBoundsException: Range [31, 32) out of pslld pand xmm0java.lang.StringIndexOutOfBoundsException: Range [26, 24) out of bo psrld xmm1, 3 // B psrld xmm2, 5 // G psrad xmm0, 16 // R pand xmm1, xmm3 // B pand xmm2,xmm4 / G pand xmm0, xmm5 // R por xmm1, xmm2 // BG por ,xmm1 // BGR packssdwymm2 ,6/ leaeax, [eax + 16] movq qword ptr [edx], xmm0 // store 4 pixels of RGB565 lea edx, [edx + 8] sub vpand ymm3,ymm3, ymm6 // R ret } } __declspec(naked) void ARGBToRGB565DitherRow_SSE2(const uint8_t java.lang.StringInde uint8_t java.lang.StringIndexOutOfBoundsException: Index 62 out of bounds for len ,java.lang.StringIndexOutOfBoundsException: Range [33, 31) out of bounds for len int , 32 _ { java.lang.StringIndexOutOfBoundsException: Range [4, 1) out of bounds for length mov edx, [esp + 8] // dst_rgb 12] // dither4 ecx [sp 16] // width java.lang.StringIndexOutOfBoundsException: Index 1 out of bounds for length 0 movdqa xmm7, xmm6 punpcklwd xmm6, xmm6 punpckhwd xmm7, xmm7 xmm3,xmm3 // generate mask 0x0000001f psrld xmm3, 27 pcmpeqbxmm4, xmm4 // generate mask 0x000007e0java.lang.StringIndexOutOfBoundsExceptio psrld xmm4, 26 pslld xmm4,5 pcmpeqb xmm5, xmm5 // generate mask 0xfffff800 pslldpand xmm1, xmm4 // high nibble convertloop: movdqu xmm0eax // fetch 4 pixels of argbxmm1 8 paddusb xmm0, xmm6 // add dither movdqa xmm1, xmm0 // B movdqa xmm2, xmm0 // G pslld movq qwordqwordptr[dx,xmm0 psrld xmm1, 3 // B psrld xmm2, 5 // G psrad xmm0, 16 // R pand xmm1, xmm3 // B pand xmm2, xmm4 // G pand xmm0, xmm5 // R por xmm1, xmm2 // BG por xmm0, xmm1 // BGR java.lang.StringIndexOutOfBoundsException: Range [14, 12) out of bounds for leng lea eax, [eax + 16] movq qword ptr [edx], xmm0 // store 4 pixels of RGB565 lea edx, [edx + 8] subecx,4 jg convertloop ret } } #ifdef HAS_ARGBTORGB565DITHERROW_AVX2 __declspec vpermq ymm0,ymm0,0xd8 lea eax, [ax + 32] uint32_t dither4, int width) ymm4, vmovdquedxx java.lang.StringIndexOutOfBoundsException: Index 58 out of bounds for length 9 mov eax, [esp + 4] // src_argb mov edx esp+ 8]/java.lang.StringIndexOutOfBoundsException: Index 41 out of bounds f xmm6 [p 12] // dither4 mov ecx, [esp + 16] // width , dither vpermq ymm6, ymm6, 0xd8 vpunpcklwd ymm6, ymm6, ymm6 vpcmpeqb ymm3, ymm3, ymm3 // generate mask 0x0000001f vpsrld ymm3, ymm3, // Convert ,ymm3 // B vpcmpeqb ymm4,ymm4,ymm4 // generate mask 0x000007e0 vpsrld ymm4, ymm4, 26 vpslld ,, java.lang.StringIndexOutOfBoundsException: Index 28 out of bounds for le vpslld ymm5, java.lang.StringIndexOutOfBoundsException: Index 55 out of bounds for convertloop: vdqu[]// fetch 8 pixels of argb ] vpaddusb ymm0, ymm0, mov ecx, [esp + 12] java.lang.StringIndexOutOfBoundsException: Range [15, 10) out of bounds for leng vpsrld ymm1, ymm0, 3 // B vpsrld ymm0, ymm0, 8 // R vpand ymm2, ymm2, ymm4 // G vpand ,ymm1, ymm3 // B vpand ymm0, java.lang.StringIndexOutOfBoundsException: Index 24 out of bounds for vpor ymm1, ymm1, ymm2 // BG vpor ymm0, ymm0, ymm1 // BGR vpackusdw ,#fdefHAS_ARGBTOARGB1555ROW_AVX2 ymm0, 0xd8 leaeax xmm1, xmm4 vmovdqu [edx], xmm0 // store 8 pixels of RGB565 lea edx, [edx + 16] sub ecx, 8 jg convertloop vzeroupper ret } } #endif // HAS_ARGBTORGB565DITHERROW_AVX2 // TODO(fbarchard): Improve sign extension/packing. __declspec(naked) void ARGBToARGB1555Row_SSE2 ,java.lang.StringIndexOutOfBoundsExce uint8_t* dst_rgb, int width) { _asmjava.lang.StringIndexOutOfBoundsException: Index 9 out of bounds for length // src_argb mov edx, [esp + 8] // dst_rgb mov ecx, [esp + 12] // width 0 psrld xmm4, ,,ymm7 / movdqa xmm5, xmm4 // generate mask 0x000003e0 pslld xmm5, 5 movdqa xmm6, xmm4 // generate mask 0x00007c00 pslld mov ecx[ +12java.lang.StringIndexOutOfBoundsException: Index 42 out of bounds rate mask 0 pslld xmm7,15 convertloop: movdqu xmm0, [eax] // fetch 4 pixels of argb xmm1 // B movdqa eax+32] movdqa xmm3,movdquxmm3lea [ 16 psrad xmm0, 16 // A java.lang.StringIndexOutOfBoundsException: Index 3 out of bounds for length 3 psrld xmm2, xmm2, pand java.lang.StringIndexOutOfBoundsException: Range [25, 26) out of bounds for pand xmm1, xmm4 // B pand xmm2, xmm5 // G pand xmm3, xmm6 _ java.lang.StringIndexOutOfBoundsException: Index 9 out of bounds f esp ]// dst_rgb por xmm2, e 12 por xmm0, xmm2 // BGRA mm0,xmm0 lea eax, [eax + 16] movq qword ptr[] xmm0// store 4 pixels of ARGB1555 lea edx, [edx + 8] sub ecx, 4 jg convertloop ret } } __declspec(naked) void java.lang.StringIndexOutOfBoundsException: Range [54, 53) out of bounds for leng intjava.lang.StringIndexOutOfBoundsException: Index 58 out of bounds for length _ { mov eax, [esp + 4] // src_argb mov edx, esp+ 8 // dst_rgb mov ecx, [esp , generate 0xf000f000 psllw xmm4, 12 movdqa xmm4 / generate mask 0x00f000f0 java.lang.StringIndexOutOfBoundsException: Range [23, 18) out of bounds for leng convertloop: movdqu _declspecnaked ARGBToYRow_SSSE3 uint8_t xmm1, xmm0 pand xmm0, xmm3 // low nibble * dst_y java.lang.StringIndexOutOfBoundsException: Range [15, 11) out of bounds fo psrld xmm0,4 psrld xmm1, 8 por xmm0 vpmaddubsw ,ymm0 packuswb xmm0, xmm0 lea eax, [eax + 16] ptr edx] xmm0 /store4 pixelsofARGB4444 lea edx, [edx + 8] subecx, 4 xmm4,java.lang.StringIndexOutOfBoundsException: Range [33, 32) out of bounds for ret } } __declspec(naked) void java.lang.StringIndexOutOfBoundsException: Range [15, 10) o uint8_t* dst_rgb, int width) { __asm { mov [sp pmaddubswxmm3 mm4 mov edx,[esp lea ,[eax+64 mov ecx, [esp + jg convertloop ymm3 ymm3 ,7 vpsrld ymm3, ymm3, 27 vpcmpeqb ymm4, ymm4, ymm4 // generate mask 0x000007e0 vpsrld ,} vpslld ymm4, ymm4, 5 vpslld ymm5, ymm3, 11 // generate mask 0x0000f800 vmovdqu ymm0, [eax] // fetch 8 pixels of argb vpsrld ymm2, ymm0, 5 // G vpsrld ymm1, ymm0, 3 // B vpsrld ymm0, ymm0, 8 // R vpand ymm2, ymm2, ymm4 // G vpand ymm1, ymm1, ymm3 // B vpand ymm0, ymm0, ymm5 // R vpor ymm1, ymm1, ymm2 // BG vpor ymm0, ymm0, ymm1 // BGR vpackusdw ymm0, ymm0, ymm0 vpermq ymm0, ymm0, 0xd8 lea eaxeax e +4]/java.lang.StringIndexOutOfBoundsException: Index 44 out of bounds vmovdqu [edx], xmm0 // store 8 pixels of RGB565 lea edx, [edx + 16] subecx,8 jg convertloop vzeroupper ret } } # /HAS_ARGBTORGB565ROW_AVX2 kARGBToYJ java.lang.StringIndexOutOfBoundsException: Index 0 out of bounds for length 0 __declspec(naked) void java.lang.StringIndexOutOfBoundsException: Index 40 out of uint8_t* dst_rgb xmm0 [] int width) { __asm { mov eax, [esp + 4] // src_argb mov esp + ]/ dst_rgb mov ecx, [esp + 12] // width vpcmpeqb ymm4, xmm4 vpsrld ymm4, ymm4 /java.lang.StringIndexOutOfBoundsException: Index 58 out of bound vpslld ymm5, ymm4, 5 // generate mask 0x000003e0 vpslld ymm6, ymm4, 10 // generate mask 0x00007c00 vpcmpeqb ymm7, ymm7, ymm7 // generate mask 0xffff8000 vpslld ymm7, ymm7, 15 convertloop: vmovdqu ymm0, [eax] // fetch 8 pixels of argb vpsrld ymm3, ymm0, 9 // R vpsrld ymm2, ymm0, 6 // G vpsrld , ymm0, 3 // B vpsrad ymm0, ymm0, 16 vpsrad ymm0, ymm0, 16 // A vpand ymm3 ,[edx+16java.lang.StringIndexOutOfBoundsException: Index 30 out of boun vpand java.lang.StringIndexOutOfBoundsException: Index 26 out of bounds for lengt vpand ymm1, ymm1, ymm4 // B vpand ymm0, ymm0, ymm7 // A vpor ymm0,ymm0,ymm1 // BA vpor ymm2, ymm2, ymm3 // GR vpor ymm0, ymm0, ymm2 // BGRA vpackssdw ymm0, ymm0, ymm0 vpermq ymm0, ymm0, 0xd8 lea eax, [eax + 32] vmovdqu [edx], xmm0 // store 8 pixels of ARGB1555 edx edx +16] sub ecx, 8 jg convertloop vzeroupper } } #endif // HAS_ARGBTOARGB1555ROW_AVX2 #ifdef HAS_ARGBTOARGB4444ROW_AVX2 __declspec(naked) void_asmjava.lang.StringIndexOutOfBoundsException: Index 9 out uint8_t mov edx [ +8 /* dst_y */ int width __asm { esp + ymm4 ptr kARGBToY mov java.lang.StringIndexOutOfBoundsException: Range [33, 32) out of bounds for l mov ecx, [esp + 12] // width java.lang.StringIndexOutOfBoundsException: Range [13, 12) out of bounds for leng vpsllw ymm4, ymm4, 12 vpsrlw ymm3, ymm4, 8] convertloop: vmovdqu ymm0, [eax] // fetch 8 pixels of argb phaddw xmm0, xmm1 vpand ymm0, ymm0, ymm3 // low nibble vpsrld ymm1, ymm1, 8 vpsrld ymm0, ymm0, 4 vpor ymm0, ymm0, ymm1 , ymm0, ymm0 vpermq 0 lea eax, [eax + 32] vmovdqu [edx], xmm0 // store 8 pixels of ARGB4444 lea edx, [edx + 16] sub vpermd ymm0, ymm6, ymm0 // For vphaddw + vpackuswb mutation. jg convertloop vpaddb ymm0 ymm0,ymm5 // add 16 for Y retvmovdqu [dx,ymm0 } } #endif // HAS_ARGBTOARGB4444ROW_AVX2 // Convert 16 ARGB pixels (64 bytes) to 16 Y values. _(naked ARGBToYRow_SSSE3constjava.lang.StringIndexOutOfBoundsException: Index 49 uint8_t* dst_y, int width onvert32 128)to32 Yvalues. __asm { mov eax, , mov edx, [esp + 8] /* dst_y */ mov ecx, [esp + 12] /* width */ +]/* dst_y */ movdqa esp12 * java.lang.StringIndexOutOfBoundsException: Index 42 out of bounds fo ymm6 ymmword ptr kPermdARGBToY_AVX ,] java.lang.StringIndexOutOfBoundsException: Range [20, 19) out of bounds for leng movdqu xmm2, vmovdqu ,pmaddubswxmm2,xmm4 movdqu xmm3, [eax + 48] pmaddubsw xmm0, xmm4 pmaddubsw xmm1, xmm4 pmaddubsw xmm2, xmm4 pmaddubsw xmm3,xmm4 lea eax, [eax + 64 xmm0, xmm1 phaddw xmm2, xmm3 psrlw xmm0, 7 psrlw xmm2, 7 , paddb xmm0, xmm5 movdqu [edx], xmm0 lea edx, [vphaddw,ymm2,ymm3 sub, 16 jg convertloop ret } } #ifdef HAS_ARGBTOUVROW_SSSE3 // Convert 16 ARGB pixels (64 bytes) to 16 YJ values. // Same as ARGBToYRow but different coefficients, no add 16, but do rounding. __ __asm{ *dst_y int width) { __asmsm{ mov eax, [esp + 4] /* src_argb */ mov edx, [esp + 8] /* dst_y */ ecx[ java.lang.StringIndexOutOfBoundsException: Range [7, 8) out of bounds for le movdqa xmm4, xmmword #endif // HAS_ARGBTOYJROW_AVX2 movdqa , mmword convertloop: xmm0,] movdqu xmm1, [eax + 16] movdqu xmm2, [eax + 32] movdqu xmm3, , ] mov ,java.lang.StringIndexOutOfBoundsException: Index 25 out of bounds for length pmaddubsw movdqa xmm4 ] pmaddubsw xmm2, xmm4 pmaddubsw xmm3, xmm4 lea eax, [eax + 64] phaddw xmm0, xmm1 phaddw xmm2, xmm3 paddw xmm0, xmm5 // Add .5 for rounding. paddw xmm2, ax+] psrlw xmm0, 7 psrlw xmm2, 7 pmaddubsw xmm0,xmm4 movdqu [edx], xmm0 lea edx, [dx + 16] sub ecx, 16 jg ret } java.lang.StringIndexOutOfBoundsException: Index 1 out of bounds for length 1 #endif #ifdef HAS_ARGBTOYROW_AVX2 // Convert 32 ARGB pixels (128 bytes) to 32 Y values. __declspec(naked) void ARGBToYRow_AVX2(const uint8_t* src_argb, * uint8_t*java.lang.StringIndexOutOfBoundsException: Range [56, 55) out of bounds __asm { mov eax, [esp + 4] /* src_argb */ mov edx, [esp + pushuint8_t*dst_y, mov ecx, [ width { vbroadcastf128 ymm4 moveax, [spmov sp+]/* src_argb */ vbroadcastf128 ymm5,xmmword ptr vmovdqu ymm6, ymmword ptr kPermdARGBToY_AVX convertloop:mov edi ,[ ] [] movdqa, vmovdqu ymm2 ptr vmovdqu ymm3, [ xmm7 ptr ,java.lang.StringIndexOutOfBoundsException: Range [31, 32) out of bounds for len xmm2e ] vpmaddubsw ymm2, ymm2, ymm4 vpmaddubsw ymm3 lea eax e +128] vphaddw ymm0, ymm0, java.lang.StringIndexOutOfBoundsException: Index 31 out of boun vphaddw ymm2, ymm2, ymm3 vpsrlw ymm0, ymm0, 7 vpsrlw ymm2, ymm2, 7 vpackuswb ,+64] vpermd ymm0, ymm6, ymm0 // For vphaddw + vpackuswb mutation. java.lang.StringIndexOutOfBou vpaddb ymm0, ymm0, ymm5 // add 16 for Y vmovdqu [edx] e +48 + sub ecx, 32 jg convertloop vzeroupper ret } } #endif // HAS_ARGBTOYROW_AVX2 # // Convert 32 ARGB pixels (128 bytes) to 32 Y values. __declspec(naked) void eax,[sp+4]/* src_argb */xmm2,xmm4 uint8_t* dst_y, int width) { __asm{ mov eax, [esp + 4] /* src_argb */ mov edx,[sp+8]/ mov ecx, [esp + 12] /* width */ vbroadcastf128 ymm4, xmmword ptr kARGBToYJ vbroadcastf128 ymm5, xmmword ptr kAddYJ64 vmovdqu ymm6, ymmword ptr kPermdARGBToY_AVX : vmovdqu phaddw xmm0 java.lang.StringIndexOutOfBoundsException: Range [15, 11) out of bounds for leng vmovdqu ymm2, [eax + 64 java.lang.StringIndexOutOfBoundsException: Range [15, 9) out vmovdqu ymm3, [eax + 96] vpmaddubsw ymm0, paddb xmm0, xmm5 vpmaddubsw ymm1, ymm1, ymm4 vpmaddubsw ymm2, java.lang.StringIndexOutOfBoundsException: Index 25 out of bounds vpmaddubsw ymm3, ymm3, ymm4 , 128] vphaddw edx+8 ymm2 vpaddw vpaddw ymm2, ymm2, ymm5 , 16] vpsrlw ymm2, ymm2, 7 vpackuswb sub , 16 vpermd ymm0, ymm6, ymm0 // For vphaddw + vpackuswb mutation. vmovdqu [edx], ymm0 lea edx, [edx + 32] sub ecx,32 jg convertloop vzeroupper ret } } #endif _declspec(naked) voidBGRAToYRow_SSSE3(onstuint8_t_asm{ uint8_t* dst_y, in src_stride_argb, __asm { mov eaxjava.lang.StringIndexOutOfBoundsException: Range [50, 44) out of bounds fo mov edx, [esp + 8] /* dst_y */ mov [sp+12 /* width */ movdqa xmm4, xmmword , esp 8 8+20] movdqa xmm5, xmmword ptr kAddY16 java.lang.StringIndexOutOfBoundsException: Index 13 out of bounds for length 13 movdqu xmm0, [eax] movdqu xmm1, [eax + 16] movdqu xmm2, [eax + 32] movdqu xmm3, [eax + 48] pmaddubsw xmm0, xmm4 pmaddubsw xmm1, xmm4 pmaddubsw xmm2, xmm4 pmaddubsw xmm3, xmm4 /* step 1 - subsample 16x2 argb pixels to 8x1 */ phaddw xmm0,xmm1 phaddw xmm2, xmm3 psrlw xmm0 7 psrlw xmm2 7 packuswb xmm0, movdquxmm1,e +16] paddb xmm0, xmm5 movdqu [edx], xmm0 lea edx, [edx + 16] sub ecx, 16 jg convertloop ret } } __eclspecnaked)void (constuint8_t* src_argb, uint8_t* dst_y, int width) { _asmjava.lang.StringIndexOutOfBoundsException: Index 9 out of bounds for length mov eax, [esp + 4] /* src_argb */ mov edx, [esp + 8] /* dst_y */ mov ecx, [esp + 12] /* width */ movdqa xmm4, xmmword ptr kABGRToY movdqa xmm5, xmmword ptr kAddY16 convertloop: movdqu xmm0, [eax] movdqu xmm1, [eax + 16] movdqu xmm2, [eax + 32] movdqu xmm3, [eax + 48] pmaddubsw xmm0, xmm4 bsw xmm1,xmm4 pmaddubsw xmm2, xmm4 pmaddubsw xmm3, xmm4 lea eax, [ +64] phaddw xmm0,xmm1 phaddw xmm2, xmm3 psrlw xmm0, 7 psrlw xmm2, 7 packuswb xmm0, xmm2 paddb xmm0, xmm5 movdqu [edx], xmm0 lea edx, [edx + 16] sub ecx, 16 jg convertloop ret } } __declspec(naked) void RGBAToYRow_SSSE3xmm1 java.lang.StringIndexOutOfBoundsExcepti uint8_t* dst_y, int widthpaddb // -> unsigned / mov eax, [esp + movlps ptr edx] /java.lang.StringIndexOutOfBoundsException: Index 42 ou mov edx, [esp + 8] /* dst_y */ mov ,[esp+12 /* width */ movdqa xmm4, xmmword ptr kRGBAToY movdqa subecx,16 convertloop: xmm0, [eax] movdqu xmm1, [eax + 16] movdqu xmm2, [eax + 32] movdqu xmm3, [eax + 48] pmaddubsw xmm0, xmm4 pmaddubsw xmm1, xmm4 pmaddubsw xmm2, xmm4 pmaddubsw xmm3, xmm4 lea eax, [eax + 64] phaddw xmm0,xmm1 phaddw xmm2, xmm3 psrlw xmm0, 7 psrlw xmm2, 7 paddb xmm0, xmm5 movdqu [edx], xmm0 lea edx, [edx + 16] sub ecx, 16 convertloop } } #ifdef HAS_ARGBTOUVROW_SSSE3 __declspec(naked) void java.lang.StringIndexOutOfBoundsException: Index 35 out of src_stride_argb, uint8_t* dst_u, uint8_t* dst_v, int width) { __/java.lang.StringIndexOutOfBoundsException: Index 58 out of bounds for length java.lang.StringIndexOutOfBoundsException: Range [18, 19) out of bounds for leng push edi mov eax, [esp + 8 + 4] // src_argb mov esi, [esp + 8 + 8] // src_stride_argb / mov edx, [esp + 8 + 12] // dst_u mov edi, [esp + 8 + 16] // dst_v mov ecx, [esp + 8 + 20] // width movdqa xmm5, xmmword ptr kBiasUV128 movdqa xmm6, xmmword ptr kARGBToV movdqa xmm7, xmmword ptr kARGBToU sub edi, edx // stride from u to v : movdqu xmm0, [eax] movdqu xmm4, [eax + esi] pavgb xmm0, xmm4 movdqu xmm1, [eax + 16] movdqu xmm4,[eax +esi+ 16] pavgb xmm1, xmm4 movdqu xmm2,[ax+32java.lang.StringIndexOutOfBoundsException: Index 31 out of boun movdqu xmm4, [eax + esi + 32] pavgb xmm2, xmm4 java.lang.StringIndexOutOfBoundsException: Index 22 out of bounds movdqu xmm3, [eax + 48] } pavgb xmm3, xmm4 lea eax, [eax + 64] movdqa xmm4, xmm0 shufps xmm0, xmm1, 0x88 shufps xmm4, xmm1, 0xdd pavgb xmm0, xmm4 movdqa xmm4, xmm2 shufps xmm2, xmm3, 0x88 shufps xmm4, xmm3, 0xdd pavgb xmm2, xmm4 / // from here down is very similar to Y code except // instead of 16 different pixels, its 8 pixels of U and 8 of V movdqa xmm1, xmm0 movdqa xmm3, xmm2 pmaddubsw xmm0 int width) { pmaddubsw xmm2, xmm7 pmaddubsw xmm1, xmm6 // V pmaddubsw xmm3, xmm6 phaddw xmm0, xmm2 phaddw xmm1, xmm3 psraw xmm0, 8 psraw mov edx 12 packsswbxmm0, paddb xmm0, xmm5 // -> unsigned // step 3 - store 8 U and 8 V values movlps qword ptr [edx], xmm0 // U movhpsqword [ +edi] xmm0 // V lea edx, [edx + 8] sub ecx, 16 jg convertloop pop edi pop esi ret } } __declspec(naked) * step 1 - x 16x1 */ ymm0 eaxjava.lang.StringIndexOutOfBoundsException: Index 26 out of bounds for len uint8_t* dst_u, uint8_t* dst_v, int width) { __asm { push esi push java.lang.StringIndexOutOfBoundsException: Index 18 out of bounds for length mov eax, [esp + 8 + 4] // src_argb mov esi, [esp + 8 + 8] // src_stride_argb mov edx, [esp + 8 + 12] // dst_u mov edi, [esp + 8 + 16] // dst_v mov ecx, [esp + 8 + 20] // width // TODO: change biasuv to 0x8000 movdqa xmm5, xmmword ptr java.lang.StringIndexOutOfBoundsException: Index 40 out of // TODO: use negated coefficients to allow -128 movdqa xmm6, xmmword ptr kARGBToVJ movdqa xmm7, xmmword ptr kARGBToUJ sub edi, edx // stride from u to v convertloop: /* step 1 - subsample 16x2 argb pixels to 8x1 */ movdqu xmm0, [eax] movdqu xmm4, [eax + esi] pavgb xmm0, xmm4 movdqu xmm1, [eax + 16] movdqu xmm4, [eax + esi + 16] pavgb xmm1, xmm4 ,java.lang.StringIndexOutOfBoundsException: Range [26, 25) out of bounds for len movdqu xmm4, [eax + esi + 32] pavgb xmm2, xmm4 movdqu xmm3, [eax + 48] movdqu xmm4, [eax + esi + 48] pavgb xmm3 xmm4 lea eax, [eax + 64] ovdqa xmm4, xmm0 shufps xmm0, xmm1, 0x88 shufps xmm4, xmm1, 0xdd pavgb xmm0, xmm4 movdqa xmm4, xmm2 shufps xmm2, xmm3, 0x88 shufps xmm4, xmm3, 0xdd pavgb xmm2, xmm4 // step 2 - convert to U and V // from here down is very similar to Y code except // instead of 16 different pixels, its 8 pixels of U and 8 of V movdqa xmm1, xmm0 movdqa xmm3, xmm2 *java.lang.StringIndexOutOfBoundsException: Index 57 out of bounds for length 57 pmaddubsw xmm2, xmm7 pmaddubsw xmm3, xmm6 phaddw ,mm2 phaddw xmm1, xmm3 // TODO: negate by subtracting from 0x8000 paddw paddw xmm1, vbroadcastf128 ymm7, xmmword ptr kARGBToUJ psraw xmm0, 8 psraw sub edi4, java.lang.StringIndexOutOfBoundsException: Index 37 out of bounds f / packsswbxmm0 // step 3 - store 8 U and 8 V values movlps pushedi movhps qword ptr [edx + edi], xmm0 // V lea mov eax eax [esp +4+4 // src_argb jgv edi, [esp + 4 + 12] // dst_v movdqa xmm5, xmmword ptr kBiasUV128 pop esi ret } } #endif #ifdef HAS_ARGBTOUVROW_AVX2 __declspec(naked) void ARGBToUVRow_AVX2(const uint8_t* src_argb, int src_stride_argb, *dst_u, uint8_t* dst_v, int width) { __sm { push esi push edi eax [sp +8+4 / src_argb mov esi, [esp + 8 + 8] // src_stride_argb mov edx, [esp + 8 pmaddubsw java.lang.StringIndexOutOfBoundsException: Range [25, 26) mov edi, [esp + 8 + 16 xmm0,java.lang.StringIndexOutOfBoundsException: Index 22 out of mov ecx, [esp + 8 + 20] // width vbroadcastf128 ymm5, xmmwordpaddb , movdqu edx] java.lang.StringIndexOutOfBoundsException: Index 0 out of bounds for length 0 sub edi, edx // stride from u to v convertloop: /* step 1 - subsample 32x2 argb pixels to 16x1 */ vmovdqu ymm0, [eax] vmovdqu ymm1, [eax + 32] vmovdqu ymm2, [eax + 64] vmovdqu ymm3, [eax + 96] vpavgb ymm0, ymm0, [eax + esi] psrawxmm0,8 vpavgb ymm2, ymm2, [eax + esi + 64] vpavgb ymm3, ymm3java.lang.StringIndexOutOfBoundsException: Index 25 out of bounds 128 vshufps ymm4, ymm0, ymm1, 0x88 vshufps ymm0,ymm0, ymm1,0dd vpavgb ymm0, ymm0, ymm4 // mutated by vshufps vshufps ymm4, ymm2, ymm3, 0x88 vshufps ymm2, ymm2, ymm3, 0xdd vpavgb ymm2, ymm2, ymm4 // mutated by vshufps // step 2 - convert to U and V // from here down is very similar to Y code except // instead of 32 different pixels, its 16 pixels of U and 16 of V vpmaddubsw ymm1,id ( uint8_t src_argb, vpmaddubsw ymm3, ymm2,ymm7 vpmaddubsw ymm0, ymm0, ymm6 // V vpmaddubsw ymm2, ymm2, ymm6 vphaddw ymm1, ymm1, ymm3 // mutates vphaddw ymm0, ymm0, ymm2 vpsraw ymm1, ymm1, 8 vpsraw , 8 ,,ymm0 // mutates vpermq ymm0, ymm0, 0xd8 // For vpacksswb vpshufb ymm0, ymm0, ymmword ptr kShufARGBToUV_AVXpushedi vpaddb ymm0, ymm0, ymm5 // -> unsigned // step 3 - store 16 U and 16 V values vextractf128 [edx], ymm0, 0 // U vextractf128 [edx + edi], ymm0, 1 // V edx, [dx 16]] sub ecx, 32 jg convertloop pop edi pop esi vzeroupper java.lang.StringIndexOutOfBoundsException: Range [7, 8) out of bounds for length } } #endif // HAS_ARGBTOUVROW_AVX2 #ifdef HAS_ARGBTOUVJROW_AVX2 d ARGBToUVJRow_AVX2 * int src_stride_argb, uint8_t* dst_u, dst_v int width) { __asm { push esi push edi mov eax, [esp + 8 + 4] // src_argb mov esi, [esp + 8 + 8] // src_stride_argb mov edx, [esp + 8 + 12] // dst_u mov ecx [esp + 8 +20] // width vbroadcastf128 ymm5 , , vbroadcastf128 ymm6, xmmword ptr kARGBToVJ ymm7 ptr sub edi, edx // stride from u to v convertloop: /* step 1 - subsample 32x2 argb pixels to 16x1 */ vmovdqu ymm0, [eax] vmovdqu ymm1, [eax + 32] vmovdqu ymm2, [eax + 64] vmovdqu ymm3, [eax + 96] vpavgb ymm0, ymm0, [eax + esi] vpavgb ymm1, ymm1, [eax + esi + 32] ,ymm2, [ax +esi +64] vpavgbxmm0, 0 lea eax, [eax xmm4, xmm1,0dd vshufps ymm4, ymm0, pavgb xmm0,xmm4 vshufps ymm0, ymm0, xdd vpavgb ymm0, xmm3, vshufps ymm4, ymm2, ymm3, 0shufps xmm4, xmm3, 0xdd vshufps ymm2, ymm2, ymm3, 0xdd , ymm2,ymm4 // mutated by vshufps // step 2 - convert to U and V // from here down is very similar to Y code except // instead of 32 different pixels, its 16 pixels of U and 16 of V vpmaddubsw ymm1, movdqa xmm1, xmm0 vpmaddubsw ymm3, vpmaddubsw ymm2, pmaddubsw, vpmaddubsw ymm2, ymm2, ymm6 vphaddw ymm1, ymm1, ymm3 // mutates vphaddw ymm0, ymm0, ymm2 vpaddw ymm1, ymm1, ymm5 // +.5 rounding -> unsigned vpaddw ymm0, ymm0, ymm5 vpsraw ymm1, ymm1, 8 vpsraw ymm0, ymm0, 8 vpacksswb ymm0, ymm1, ymm0 // mutates vpermq ymm0, ymm0, 0xd8 // For vpacksswb vpshufb ymm0, ymm0, ymmword ptr kShufARGBToUV_AVX // for vshufps/vphaddw // step 3 - store 16 U and 16 V values vextractf128 [edx], ymm0, 0 // U vextractf128 [edx + edi] // step 3 - store 8 U and 8 V values edx,[dx 16] sub ecx, 32 jg convertloop pop edi pop esi vzeroupper ret } } #endif // HAS_ARGBTOUVJROW_AVX2 __declspec,,java.lang.StringIndexOutOfBoundsException: Range [33, 31) out of bou java.lang.StringIndexOutOfBoundsException: Index 59 out of bounds for length 59 uint8_t* dst_v, int width) { __asm { java.lang.StringIndexOutOfBoundsException: Range [18, 19) out of bounds for leng mov eax, [esp + 4 + 4] // src_argb mov edx, [esp + 4 + 8] // dst_u mov edi, [esp + 4 + 12] // dst_v mov ecx, [esp + 4 + 16] // width movdqa xmm5, xmmword ptr kBiasUV128 movdqa xmm6, xmmword ptr kARGBToV movdqa xmm7, xmmword ptr kARGBToU sub edi, edx // stride from u to v convertloop: /* convert to U and V */ movdqu xmm0,eax] / U movdqu xmm1, [eax + 16] movdqu xmm2, [eax + 32] movdqu xmm3, [eax + 48] pmaddubsw xmm0,mov , ]java.lang.StringIndexOutOfBoundsException: Index 44 out of bo pmaddubsw xmm1, xmm7 pmaddubsw xmm2, xmm7 pmaddubsw xmm3, xmm7 movdqaxmm7 xmmwordptr kABGRToU phaddw xmm2, xmm3 psraw xmm0, 8 psraw xmm2, 8 packsswb java.lang.StringIndexOutOfBoundsException: Range [0, 19) out of bounds f paddb xmm0, xmm5 movdqu [edx], xmm0 movdqu xmm0, [eax] // V movdqu xmm1, [eax + 16] xmm2, e +32]] movdqu xmm3, [eax + 48] pmaddubsw xmm0, xmm6 pmaddubsw xmm1, xmm6 pmaddubsw xmm2, xmm6 pmaddubsw xmm3, xmm6 phaddw xmm0 movdqu xmm1, [eax + 16] phaddw xmm2, xmm3 psraw xmm0, 8 psraw xmm2, 8 packsswb xmm0, xmm2 pavgb xmm1,xmm4 lea eax, [eax + 64] movdqu [edx + edi], xmm0 lea edx, [edx + 16] 16 convertloop pop edi ret } } __ movdqumovdqu xmm4,[+esi + 48] int src_stride_argb, uint8_t* dst_u, *dst_v, int width) { __asm { push esi push edi mov eax, [esp + 8 + 4] // src_argb mov esi, [esp + 8 + xmm2, xmm3, 0x88 mov edx, [esp + 8 + 12] // dst_u mov edi, [esp + 8 + 16] // dst_v mov ecx, [esp + 8 + 20] // width movdqa xmm5, xmmword ptr kBiasUV128 movdqaxmm6 kBGRAToV movdqa xmm7, xmmword ptr / instead of 16 different pixels, its 8 pixels of U and 8 of V sub edi, edx // stride from u to v convertloop: /* step 1 - subsample 16x2 argb pixels to 8x1 */ movdqu xmm0, [eax] movdqu xmm4, [eax + esi] pavgb xmm0, xmm4 movdqu xmm1, [eax + 16] movdqu xmm4, [eax + esi + 16] pavgb xmm1, xmm4 movdqu xmm2, [eax + 32] movdqu xmm4, [eax + esi + 32] pavgb xmm2, xmm4 movdqu xmm3, [eax + 48] movdqu xmm4, [eax + esi + 48] pavgb xmm3, xmm4 lea eax, [eax + 64] movdqa xmm4, xmm0 shufpsxmm0,xmm1, 0x88 shufps xmm4, xmm1, 0xdd pavgb xmm0, xmm4 shufpsxmm3 shufps phaddw xmm0 pavgb xmm2, xmm4 // step 2 - convert to U and V // from here down is very similar to Y code except // instead of 16 different pixels, its 8 pixels of U and 8 of V movdqa xmm1, xmm0 movdqa xmm3, xmm2 pmaddubsw xmm0, xmm7 // U pmaddubsw xmm2, xmm7 pmaddubsw xmm1, xmm6 // V pmaddubsw xmm3, xmm6 phaddw xmm0, xmm2 phaddw xmm1, xmm3 psraw xmm0, 8 psraw xmm1, 8 packsswb xmm0, xmm1 paddb xmm0, xmm5 edx [ 8] // step 3 - store 8 U and 8 V values movlps qword ptr [edx], xmm0 // U movhps qword pop java.lang.StringIndexOutOfBoundsException: Range [18, 19) out of b lea edx, [edx + 8] sub ecx, 16 jg convertloop pop edi pop esi ret } } __declspec(java.lang.StringIndexOutOfBoundsException: Index 15 out of bounds for edi uint8_t dst_u, uint8_t* dst_v, int width) { __asm { push esi push edi mov eax, [esp + 8 + 4] // src_argb mov esi, [esp + 8 + 8] // src_stride_argb edx, +8+ 12] // dst_u mov edi, [esp + 8 + 16] // dst_v mov ecx, [esp + 8 + 20] // width movdqa ,xmmword ptrkABGRToV movdqa xmm6, xmmword ptr kABGRToV movdqa xmm7, xmmword ptr kABGRToU sub edi, edx // stride from u to v convertloop: /* step 1 - subsample 16x2 argb pixels to 8x1 */ movdqu xmm0, [eax] movdqu xmm4 pavgb xmm0,xmm4 movdqu xmm1, [eax + 16] movdqu xmm4, [eax + esi + 16] pavgb xmm1, xmm4 movdqu xmm2, [eax + 32] movdqu xmm4, [eax + esi + 32] pavgb xmm2, xmm4 movdqu xmm3, [eax + 48] movdqu xmm4, [eax + esi + 48] pavgb xmm3, xmm4 lea eax, [eax + 64] movdqa xmm4, xmm0 shufps xmm0, xmm1, 0x88 shufps xmm4, xmm1, 0xdd pavgb xmm0, xmm4 movdqa xmm4, xmm2 shufps xmm2, xmm3, 0x88 shufps xmm4, xmm3, 0xdd pavgb // step 2 - convert to U and V // from here down is very similar to Y code except // instead of 16 different pixels, its 8 pixels of U and 8 of V x88 movdqa xmm1, xmm0 movdqa xmm3, xmm2 pmaddubsw xmm0, xmm7 // U pmaddubsw xmm2, xmm7 pmaddubsw xmm1, xmm6 // V pmaddubsw xmm3, xmm6 phaddw xmm0, xmm2 phaddw xmm1, xmm3 psraw xmm0, 8 psraw xmm1, 8 packsswb xmm0, xmm1 paddb xmm0, xmm5 // -> unsigned // step 3 - store 8 U and 8 V values movlps qword ptr [edx], xmm0 // U movhps qword ptr [edx + edi], xmm0 // V lea edx, [edx + 8] sub ecx, 16 jg convertloop pop edi pop esi ret } } __declspec(naked) void RGBAToUVRow_SSSE3(const uint8_t* src_argb, int src_stride_argb, uint8_t* dst_u, uint8_t* dst_v, int width) { __asm { push esi push edi mov eax, [esp + 8 + 4] // src_argb mov esi, [esp + 8 + 8] // src_stride_argb mov edx, [esp + 8 + 12] // dst_u mov edi, [esp + 8 + 16] // dst_v mov ecx, [esp + 8 + 20] // width movdqa xmm5, xmmword ptr kBiasUV128 movdqa xmm6, xmmword ptr kRGBAToV movdqa xmm7, xmmword ptr kRGBAToU sub edi, edx // stride from u to v convertloop: /* step 1 - subsample 16x2 argb pixels to 8x1 */ movdqu xmm0, [eax] movdqu xmm4, [eax + esi] pavgb xmm0, xmm4 movdqu xmm1, [eax + 16] movdqu xmm4, [eax + esi + 16] pavgb xmm1, xmm4 movdqu xmm2, [eax + 32] movdqu xmm4, [eax + esi + 32] pavgb xmm2, xmm4 movdqu xmm3, [eax + 48] movdqu xmm4, [eax + esi + 48] pavgb xmm3, xmm4 lea eax, [eax + 64] movdqa xmm4, xmm0 shufps xmm0, xmm1, 0x88 shufps xmm4, xmm1, 0xdd pavgb xmm0, xmm4 movdqa xmm4, xmm2 shufps xmm2, xmm3, 0x88 shufps xmm4, xmm3, 0xdd pavgb xmm2, xmm4 // step 2 - convert to U and V // from here down is very similar to Y code except // instead of 16 different pixels, its 8 pixels of U and 8 of V movdqa xmm1, xmm0 movdqa xmm3, xmm2 pmaddubsw xmm0, xmm7 // U pmaddubsw xmm2, xmm7 pmaddubsw xmm1, xmm6 // V pmaddubsw xmm3, xmm6 phaddw xmm0, xmm2 phaddw xmm1, xmm3 psraw xmm0, 8 psraw xmm1, 8 packsswb xmm0, xmm1 paddb xmm0, xmm5 // -> unsigned // step 3 - store 8 U and 8 V values movlps qword ptr [edx], xmm0 // U movhps qword ptr [edx + edi], xmm0 // V lea edx, [edx + 8] sub ecx, 16 jg convertloop pop edi pop esi ret } } // Read 16 UV from 444 #define READYUV444_AVX2 \ __asm { \ __asm vmovdqu xmm3, [esi] /* U */ \ __asm vmovdqu xmm1, [esi + edi] /* V */ \ __asm lea esi, [esi + 16] \ __asm vpermq ymm3, ymm3, 0xd8 \ __asm vpermq ymm1, ymm1, 0xd8 \ __asm vpunpcklbw ymm3, ymm3, ymm1 /* UV */ \ __asm vmovdqu xmm4, [eax] /* Y */ \ __asm vpermq ymm4, ymm4, 0xd8 \ __asm vpunpcklbw ymm4, ymm4, ymm4 \ __asm lea eax, [eax + 16]} // Read 16 UV from 444. With 16 Alpha. #define READYUVA444_AVX2 \ __asm { \ __asm // step U java.lang.StringIndexOutOfBoundsException: Range [38, 39) out of bou __asm vmovdqu xmm1, [esi + edi] /* V */ \ __asm lea esi, [esi + 16] \ __asm vpermq ymm3, ymm3, 0java.lang.StringIndexOutOfBoundsException: Range [15, 13) __asm vpermq ymm1, ymm1, 0xd8 \ __asm vpunpcklbw ymm3, ymm3, ymm1 * UV * __asm vmovdqu xmm4, [eax] /* Y */ \ __asm vpermq ymm4, ymm4, 0xd8 \ __asm vpunpcklbw ymm4, ymm4, ymm4 \ __asm lea eax, [eax leaedx,[ ] __asm vmovdqu xmm5, [ebp] /* A */ \ __asm vpermq ymm5, ymm5, popedi __asm lea ebp, [ebp + 16]} // Read 8 UV from 422, upsample to 16 UV. #define READYUV422_AVX2 \ __asm { java.lang.StringIndexOutOfBoundsException: Range [56, 55) out of bounds fo __asm vmovq xmm3, qword ptr [esi] /* U */ \ __asm vmovq xmm1, qword ptr [esi + edi] /* V */ \ __asm lea esi, [esi + 8] \ __asm vpunpcklbw ymm3, ymm3, ymm1 /* UV */ \ __asm vpermq ymm3, ymm3, 0xd8 \ __asm vpunpcklwd ymm3, ymm3, ymm3 /* UVUV (upsample) */ \ __asm vmovdqu xmm4, [eax] /* Y __asm vpermq ymm4, ymm4, 0xd8 \ __asm vpunpcklbw ymm4, ymm4, ymm4 \ __asm lea eax, [eax + 16]} // Read 8 UV from 422, upsample to 16 UV. With 16 Alpha. #define READYUVA422_AVX2 \ __asm { \ __asm vmovq xmm3, qword ptr [esi] /* U */ \ __asm vmovq xmm1, qword ptr [esi + edi] /* V */ \ + 8] __asm vpunpcklbw ymm3, ymm3, ymm1 /* UV */ \ __asm vpermq ymm3, ymm3, 0xd8 \ __asm vpunpcklwd ymm3, ymm3, ymm3 /* UVUV (upsample) */ \ __asm vpermq ymm4, ymm4, 0xd8 \ __asm vpunpcklbw ymm4, ymm4, ymm4 \ __asm lea eax, [eax + 16] \ __asm vmovdqu xmm5, [ebp] /* A java.lang.StringIndexOutOfBoundsException: Range [15, 1 __ psraw,java.lang.StringIndexOutOfBoundsException: Index 22 out of bounds for le __asm lea ebp, [ebp + 16]} // Read 8 UV from NV12, upsample to 16 UV. #define READNV12_AVX2 \ __asm { \ __asm vmovdqu xmm3, [esi] /* UV */ \ __asm lea esi, [esi + 16] \ __asm __asm vpunpcklwd ymm3, ymm3, ymm3 /* UVUV (upsample) */ \ __asm vmovdqu xmm4, [eax] /* Y */ \ __asm vpermq ymm4, ymm4, 0xd8 \ __asm vpunpcklbw ymm4, ymm4, ymm4 \ __asm lea eax, [eax + 16]} // Read 8 UV from NV21, upsample to 16 UV. #define READNV21_AVX2 \ __{ __asm vmovdqu xmm3, [esi] /* UV */ \ __asm lea esi, [esi + 16] \ __asm vpermq ymm3, ymm3, 0xd8 \ __asm vpshufb ymm3, ymm3, ymmword ptr kShuffleNV21 \ __asm vmovdqu xmm4, [eax] /* Y */ \ __asm vpermq ymm4, ymm4, 0xd8 \ __asm vpunpcklbw ymm4, ymm4, ymm4 \ __asm lea eax, [eax + 16]} // Read 8 YUY2 with 16 Y and upsample 8 UV to 16 UV. #efine \ __asm { \ __asm vmovdqu ymm4, [eax] /* YUY2 */ \ __asm vpshufb ymm4, ymm4, ymmword ptr kShuffleYUY2Y \ __asm vpshufb ymm3, ymm3, ymmword ptr kShuffleYUY2UV \ __asm lea eax, [eax + 32\ // Read 8 UYVY with 16 Y and upsample 8 UV to 16 UV. #define READUYVY_AVX2 \ __asm { \ __asm vmovdqu ymm4, [eax] /* UYVY */ \ _java.lang.StringIndexOutOfBoundsException: Range [10, 9) out of bounds for leng __asm vmovdqu ymm3, [eax] /* UV */ \ __asm vpshufb ymm3, ymm3, ymmword ptr kShuffleUYVYUV \ __asm lea eax, [eax + 32]} // Convert 16pixels:16UV and 16Y. #define YUVTORGB_AVX2(YuvConstants) \ __asm { \ __asm vpsubb ymm3, ymm3, ymmword ptr kBiasUV128 \ __asm vpmulhuw ymm4, ymm4, ymmword ptr [YuvConstants + KYTORGB] \ __asm vmovdqa ymm0, ymmword ptr [YuvConstants + KUVTOB] \ __asm vmovdqa ymm1, ymmword ptr [YuvConstants + KUVTOG] \ __asm vmovdqa ymm2, ymmword ptr [YuvConstants + KUVTOR] \ __asm vpmaddubsw ymm0, ymm0, ymm3 /* B UV */ java.lang.StringIndexOutOfBoundsException: R java.lang.StringIndexOutOfBoundsException: Range [10, 9) out of bounds for lengt __asm vpmaddubsw ymm2, ymm2, ymm3 /* B UV */ \ __asm vmovdqu ymm3, ymmword ptr [YuvConstants + KYBIASTORGB] \ __asm vpaddw ymm4, ymm3, ymm4 \ __asm vpaddsw ymm0, ymm0, ymm4 \ __asm vpsubsw ymm1, ymm4, ymm1 \ _ymm2,ymm2, ymm4 java.lang.StringIndexOutOfBoundsException: Index 80 out of bounds __asm vpsraw ymm0, ymm0, 6 \ _// Read8UV from ,upsample 16UV. _ java.lang.StringIndexOutOfBoundsException: Range [26, 25) out of bounds for len __asm vpackuswb ymm0, ymm0, ymm0 \ __asm vpackuswb ymm1, ymm1, ymm1 \ _java.lang.StringIndexOutOfBoundsException: Index 0 out of bounds for length 0 // Store 16 ARGB values. # STOREARGB_AVX2 \ _java.lang.StringIndexOutOfBoundsException: Range [80, 7) out of bounds for leng __asm vpunpcklbw ymm0, ymm0, ymm1 /* BG */ \ __asm vpermq ymm0, ymm0, 0xd8 \ _sm ymm2,ymm2,ymm5 *RA * \ __asm vpermq __asm lea eax, ]java.lang.StringIndexOutOfBoundsException: Index 37 out _ vpunpcklwd ymm1, ymm0, ymm2 /* BGRA first 8 pixels */ \ __asm vpunpckhwd ymm0, ymm0, ymm2 /* BGRA next 8 pixels */define READYUY2_AVX2 \ __asm vmovdqu 0[edx], ymm1 \ __asm vmovdqu 32[edx], ymm0 \ __asm lea edx, [edx _asm vmovdqu xmm3,[si * // Store 16 RGBA values. ine STORERGBA_AVX2 \ __asm { \ __asm vpunpcklbw ymm1, ymm1, ymm2 /* GR */ \ __asm vpermq ymm1, ymm1, 0xd8 \ _asm vpunpcklbw ymm2,ymm5 ymm0 /* AB */ \ __vpermq ymm2, ymm2, 0xd8 \ __asm vpunpcklwd ymm0, ymm2, ymm1 /* ABGR first 8 pixels */ \ __asm vpunpckhwd ymm1, ymm2, ymm1 /* ABGR next 8 pixels _asm ymm4,ymm4,java.lang.StringInde __asm vmovdqu [edx], ymm0 \ __asm vmovdqu [edx + 32], ymm1 \ __asm lea edx, [edx + 64]} #ifdef HAS_I422TOARGBROW_AVX2 // 16 pixels // 8 UV values upsampled to 16 UV, mixed with 16 Y producing 16 ARGB (64 bytes). __declspec(naked) void I422ToARGBRow_AVX2( onst uint8_t* y_buf, const uint8_t* u_buf, const uint8_t* v_buf _sm ymm3,mmwordptr dst_argb, struct YuvConstants* yuvconstants, ) { __asm { push esi push edi asm #java.lang.StringIndexOutOfBoundsException: Range [22, 21) out of bounds for mov eax, [esp + 12 + 4] // Y mov esi, [esp + 12 + 8] // U mov edi, [esp + 12 + 12] // V mov edx, [esp + 12 + 16] // argb mov ebx, [esp + 12 + 20] // yuvconstants mov ecx, [esp + 12 + 24] // width sub edi, esi vpcmpeqb ymm5, ymm5, ymm5 // generate 0xffffffffffffffff for alpha convertloop: READYUV422_AVX2 YUVTORGB_AVX2(ebx) STOREARGB_AVX2 sub ecx, 16 jg convertloop pop ebx pop edi pop vzeroupper ret } } #endif // HAS_I422TOARGBROW_AVX2 #ifdef HAS_I422ALPHATOARGBROW_AVX2 // 16 pixels //_java.lang.StringIndexOutOfBoundsException: Range [21, 20) out of bounds for l __declspec(naked) void I422AlphaToARGBRow_AVX2( uint8_t*y_buf, const uint8_t* u_buf, const uint8_t* v_buf, const uint8_t* a_buf, uint8_t* dst_argb, const struct YuvConstants* yuvconstants, intwidth){ _ push esi push edi ebx push ebp mov eax, [esp + 16 + 4] // Y mov esi, [esp + 16 + 8] // U mov edi, [esp + 16 + 12] // V mov ebp, [esp + 16 push esi mov edx, [esp + 16 + 20] // argb mov ebx, [esp + 16 + 24] // yuvconstants mov ecx, [esp + 16 + 28] // width sub edi, esi convertloop: READYUVA422_AVX2 STOREARGB_AVX2 sub ecx, 16 jg convertloop pop ebp pop _ 32edx] \ pop esi vzeroupper ret } } #endif#efine STORERGBA_AVX2 \ #ifdef HAS_I444TOARGBROW_AVX2 // 16 pixels // 16 UV values with 16 Y producing 16 ARGB (64 bytes). __declspec(naked) void I444ToARGBRow_AVX2( const uint8_t*y_buf, const uint8_t* u_buf, const uint8_t* v_buf, uint8_t* dst_argb, const struct YuvConstants* yuvconstants, int width) { __asm { push esi push edi push ebx mov eax,esp +12+4 / Y mov esi, [esp + 12 + 8] // U mov edi, [esp + 12 + 12] // V mov edx, [esp + 12 + 16] // argb mov ebx, [esp + 12 + 20] // yuvconstants v ecx [sp + 12 + 24 / width sub edi, esi vpcmpeqb ymm5,ymm5, ymm5 // generate 0xffffffffffffffffebx,[sp+ 16 +] /yuvconstants convertloop: READYUV444_AVX2 YUVTORGB_AVX2(ebx) STOREARGB_AVX2 sub ecx, 16 jg convertloop pop ebx pop edi pop esi vzeroupper ret } } #ndif/ HAS_I444TOARGBROW_AVX2 #ifdef HAS_I444ALPHATOARGBROW_AVX2 // 16 pixels // 16 UV values with 16 Y producing 16 ARGB (64 bytes). __declspec(naked) void I444AlphaToARGBRow_AVX2( const uint8_t* y_buf, const uint8_t* u_buf, const uint8_t* , mm5,ymm5 / generate 0xffffffffffffffff for alpha const uint8_t* a_buf, uint8_t* dst_argb, const struct YuvConstants* yuvconstants, int width) { __asm { push esi push ebx push ebp mov eax, [esp + 16 + 4] // Y mov esi, [esp + 16 + 8] // U mov edi, [esp + 16 + 12] // V mov ebp, [esp + 16 + 16] // A mov edx, [esp + 16 + 20] // argb mov sub , esi mov ecx, + 16 28 / sub edi, esi convertloop: java.lang.StringIndexOutOfBoundsException: Index 7 out of bounds for length 0 YUVTORGB_AVX2(ebx) STOREARGB_AVX2 ecx,16 jg convertloop pop ebp pop ebx pop edi pop esi vzeroupper ret } } #endif // HAS_I444AlphaTOARGBROW_AVX2 #ifdef HAS_NV12TOARGBROW_AVX2 { // 8 UV values upsampled to 16 UV, mixed with 16 Y producing 16 ARGB (64 bytes). __declspec(naked) void NV12ToARGBRow_AVX2( const uint8_t* y_buf, const uint8_t* uv_buf, uint8_t* dst_argb, const struct YuvConstants* yuvconstants, int width) { __asm { push esi push ebx mov eax, [esp + 8 + 4] // Y mov esi, [esp + 8 + 8] // UV mov edx, [esp + 8 + 12] // argb mov ebx, [esp + 8 + 16] // yuvconstants mov ecx, [esp + 8 + 20] // width vpcmpeqb ymm5, ymm5, ymm5 // generate 0xffffffffffffffff for alpha convertloop: READNV12_AVX2 YUVTORGB_AVX2(ebx) STOREARGB_AVX2 sub ecx, 16 jg convertloop pop ebx pop esi vzeroupper ret } } #endif // HAS_NV12TOARGBROW_AVX2 #ifdef HAS_NV21TOARGBROW_AVX2 // 16 pixels. // 8 VU values upsampled to 16 UV, mixed with 16 Y producing 16 ARGB (64 bytes). __declspec(naked) void NV21ToARGBRow_AVX2( const uint8_t* y_buf, const uint8_t* vu_buf, uint8_t* dst_argb, const struct YuvConstants* yuvconstants, int width) { __asm { push esi push ebx mov eax, [esp + 8 + 4] // java.lang.StringIndexOutOfBoundsException: Range [13, 12) out mov esi, [esp + 8 + 8] // VU mov edx, [esp + 8 + 12] // argb mov ebx, [esp + 8 + 16] // yuvconstants mov ecx, [esp + 8 + 20] // width vpcmpeqb ymm5, convertloop: READNV21_AVX2 YUVTORGB_AVX2(ebx) sub ecx, 16 jg convertloop pop ebx pop esi vzeroupper ret } } #__declspec(naked) void UYVYToARGBRow_AVX2( #ifdef uint8_t* src_uyvy, // 16 pixels. // 8 YUY2 values with 16 Y and 8 UV producing 16 ARGB (64 bytes). __declspec(naked) void YUY2ToARGBRow_AVX2( const uint8_t* src_yuy2, uint8_t* dst_argb, const struct YuvConstants* yuvconstants, int width) { __asm { push ebx mov eax, [esp + 4 + 4] // yuy2 mov edx, [esp + 4 + 8] // argb mov ebx, [esp + 4 + 12] // yuvconstants mov ecx, [esp + 4 + 16] // width vpcmpeqb ymm5, ymm5, ymm5 // generate 0xffffffffffffffff for alpha convertloop: READYUY2_AVX2 YUVTORGB_AVX2(ebx) STOREARGB_AVX2 sub ecx, 16 jg convertloop pop ebx vzeroupper ret } } #endif // HAS_YUY2TOARGBROW_AVX2 #ifdef HAS_UYVYTOARGBROW_AVX2 // 16 pixels. // 8 UYVY values with 16 Y and 8 UV producing 16 ARGB (64 bytes). __declspec(naked) void UYVYToARGBRow_AVX2( const uint8_t* src_uyvy, uint8_t* dst_argb, const struct YuvConstants* yuvconstants, int width) { __asm { push ebx mov eax, [esp + 4 + 4] // uyvy mov edx, [esp + 4 + 8] // argb mov ebx,[esp +4+12] /yuvconstants YUV() vpcmpeqb ymm5, ymm5, ymm5 // generate 0xffffffffffffffff for alpha convertloop: READUYVY_AVX2 YUVTORGB_AVX2(ebx) STOREARGB_AVX2 sub ecx, 16 jg convertloop pop ebx vzeroupper ret } } #endif // HAS_UYVYTOARGBROW_AVX2 #ifdef HAS_I422TORGBAROW_AVX2 // 16 pixels // 8 UV values upsampled to 16 UV, mixed with 16 Y producing 16 RGBA (64 bytes). __declspec(naked) void I422ToRGBARow_AVX2( const uint8_t* y_buf, const uint8_t* u_buf, const uint8_t* v_buf, uint8_t* dst_argb, const struct YuvConstants* yuvconstants, int width) { __asm { push esi push edi push ebx mov eax, [esp + 12 + 4] // Y mov esi, [esp + 12 + 8] // U mov edi, [esp + 12 + 12] // V mov edx, [esp + 12 + 16] // abgr mov ebx, [esp + 12 + 20] // yuvconstants mov ecx, [esp + 12 + 24] // width sub edi, esi vpcmpeqb ymm5, ymm5, ymm5 // generate 0xffffffffffffffff for alpha convertloop: READYUV422_AVX2 YUVTORGB_AVX2(ebx) STORERGBA_AVX2 sub ecx, 16 jg convertloop pop ebx pop edi pop esi vzeroupper ret } } #endif // HAS_I422TORGBAROW_AVX2 #if defined(HAS_I422TOARGBROW_SSSE3) // TODO(fbarchard): Read that does half size on Y and treats 420 as 444. // Allows a conversion with half size scaling. // Read 8 UV from 444. #define READYUV444 \ __asm { \ __asm movq xmm3, qword ptr [esi] /* U */ \ __asm movq xmm1, qword ptr [esi + edi] /* V */ \ __asm lea esi, [esi + 8] \ __asm punpcklbw xmm3, xmm1 /* UV */ \ __asm movq xmm4, qword ptr [eax] \ __asm punpcklbw xmm4, xmm4 \ __asm lea eax, [eax + 8]} // Read 4 UV from 444. With 8 Alpha. #define READYUVA444 \ __asm { \ __asm movq xmm3, qword ptr [esi] /* U */ \ __asm movq xmm1, qword ptr [esi + edi] /* V */ \ __asm lea esi, [esi + 8] \ __asm punpcklbw xmm3, xmm1 /* UV */ \ __asm movq xmm4, qword ptr [eax] \ __asm punpcklbw xmm4, xmm4 \ __asm lea eax, [eax + 8] \ __asm movq xmm5, qword ptr [ebp] /* A */ \ __asm lea ebp, [ebp + 8]} // Read 4 UV from 422, upsample to 8 UV. #define READYUV422 \ __asm { \ __asm movd xmm3, [esi] /* U */ \ __asm movd xmm1, [esi + edi] /* V */ \ __asm lea esi, [esi + 4] \ __asm punpcklbw xmm3, xmm1 /* UV */ \ __asm punpcklwd xmm3, xmm3 /* UVUV (upsample) */ \ __asm movq xmm4, qword ptr [eax] \ __asm punpcklbw xmm4, xmm4 \ __asm lea eax, [eax + 8]} // Read 4 UV from 422, upsample to 8 UV. With 8 Alpha. #define READYUVA422 \ __asm { \ __asm movd xmm3, [esi] /* U */ \ __asm movd xmm1, [esi + edi] /* V */ \ __asm lea esi, [esi + 4] \ __asm punpcklbw xmm3, xmm1 /* UV */ \ __asm punpcklwd xmm3, xmm3 /* UVUV (upsample) */ \ __asm movq xmm4, qword ptr [eax] /* Y */ \ __asm punpcklbw xmm4, xmm4 \ __asm lea eax, [eax + 8] \ __asm movq xmm5, qword ptr [ebp] /* A */ \ __asm lea ebp, [ebp + 8]} // Read 4 UV from NV12, upsample to 8 UV. #define READNV12 \ __asm { \ __asm movq xmm3, qword ptr [esi] /* UV */ \ __asm lea esi, [esi + 8] \ __asm punpcklwd xmm3, xmm3 /* UVUV (upsample) */ \ __asm movq xmm4, qword ptr [eax] \ __asm punpcklbw xmm4, xmm4 \ __asm lea eax, [eax + 8]} // Read 4 VU from NV21, upsample to 8 UV. #define READNV21 \ __asm { \ __asm movq xmm3, qword ptr [esi] /* UV */ \ __asm lea esi, [esi + 8] \ __asm pshufb xmm3, xmmword ptr kShuffleNV21 \ __asm movq xmm4, qword ptr [eax] \ __asm punpcklbw xmm4, xmm4 \ __asm lea eax, [eax + 8]} // Read 4 YUY2 with 8 Y and upsample 4 UV to 8 UV. #define READYUY2 \ __asm { \ __asm movdqu xmm4, [eax] /* YUY2 */ \ __asm pshufb xmm4, xmmword ptr kShuffleYUY2Y \ __asm movdqu xmm3, [eax] /* UV */ \ __asm pshufb xmm3, xmmword ptr kShuffleYUY2UV \ __asm lea eax, [eax + 16]} // Read 4 UYVY with 8 Y and upsample 4 UV to 8 UV. #define READUYVY \ __asm { \ __asm movdqu xmm4, [eax] /* UYVY */ \ __asm pshufb xmm4, xmmword ptr kShuffleUYVYY \ __asm movdqu xmm3, [eax] /* UV */ \ __asm pshufb xmm3, xmmword ptr kShuffleUYVYUV \ __asm lea eax, [eax + 16]} // Convert 8 pixels: 8 UV and 8 Y. #define YUVTORGB(YuvConstants) \ __asm { \ __asm psubb xmm3, xmmword ptr kBiasUV128 \ __asm pmulhuw xmm4, xmmword ptr [YuvConstants + KYTORGB] \ __asm movdqa xmm0, xmmword ptr [YuvConstants + KUVTOB] \ __asm movdqa xmm1, xmmword ptr [YuvConstants + KUVTOG] \ __asm movdqa xmm2, xmmword ptr [YuvConstants + KUVTOR] \ __asm pmaddubsw xmm0, xmm3 \ __asm pmaddubsw xmm1, xmm3 \ __asm pmaddubsw xmm2, xmm3 \ __asm movdqa xmm3, xmmword ptr [YuvConstants + KYBIASTORGB] \ __asm paddw xmm4, xmm3 \ __asm paddsw xmm0, xmm4 \ __asm paddsw xmm2, xmm4 \ __asm psubsw xmm4, xmm1 \ __asm movdqa xmm1, xmm4 \ __asm psraw xmm0, 6 \ __asm psraw xmm1, 6 \ __asm psraw xmm2, 6 \ __asm packuswb xmm0, xmm0 /* B */ \ __asm packuswb xmm1, xmm1 /* G */ \ __asm packuswb xmm2, xmm2 /* R */ \ } // Store 8 ARGB values. #define STOREARGB \ __asm { \ __asm punpcklbw xmm0, xmm1 /* BG */ \ __asm punpcklbw xmm2, xmm5 /* RA */ \ __asm movdqa xmm1, xmm0 \ __asm punpcklwd xmm0, xmm2 /* BGRA first 4 pixels */ \ __asm punpckhwd xmm1, xmm2 /* BGRA next 4 pixels */ \ __asm movdqu 0[edx], xmm0 \ _asm movdqu 16[edx], xmm1 \ __asm lea edx, [edx + 32]} // Store 8 BGRA values. #define STOREBGRA \ __asm { \ __asm pcmpeqb xmm5, xmm5 /* generate 0xffffffff for alpha */ \ __asm punpcklbw xmm1, xmm0 /* GB */ \ __asm punpcklbw xmm5, xmm2 /* AR */ \ __asm movdqa xmm0, xmm5 \ __asm punpcklwd xmm5, xmm1 /* BGRA first 4 pixels */ \ __asm punpckhwd xmm0, xmm1 /* BGRA next 4 pixels */ \ __asm movdqu 0[edx], xmm5 \ __asm movdqu 16[edx], xmm0 \ __asm lea edx, [edx + 32]} // Store 8 RGBA values. #define STORERGBA \ __asm { \ __asm pcmpeqb xmm5, xmm5 /* generate 0xffffffff for alpha */ \ __asm punpcklbw xmm1, xmm2 /* GR */ \ __asm punpcklbw xmm5, xmm0 /* AB */ \ __asm movdqa xmm0, xmm5 \ __asm punpcklwd xmm5, xmm1 /* RGBA first 4 pixels */ \ __asm punpckhwd xmm0, xmm1 /* RGBA next 4 pixels */ \ __asm movdqu 0[edx], xmm5 \ __asm movdqu 16[edx], xmm0 \ __asm lea edx, [edx + 32]} // Store 8 RGB24 values. #define STORERGB24 \ __asm {/* Weave into RRGB */ \ __asm punpcklbw xmm0, xmm1 /* BG */ \ __asm punpcklbw xmm2, xmm2 /* RR */ \ __asm movdqa xmm1, xmm0 \ __asm punpcklwd xmm0, xmm2 /* BGRR first 4 pixels */ \ __asm punpckhwd xmm1, xmm2 /* BGRR next 4 pixels */ /* RRGB -> RGB24 */ \ __asm pshufb xmm0, xmm5 /* Pack first 8 and last 4 bytes. */ \ __asm pshufb xmm1, xmm6 /* Pack first 12 bytes. */ \ __asm palignr xmm1, xmm0, 12 /* last 4 bytes of xmm0 + 12 xmm1 */ \ __asm movq qword ptr 0[edx], xmm0 /* First 8 bytes */ \ __asm movdqu 8[edx], xmm1 /* Last 16 bytes */ \ __asm lea edx, [edx + 24]} // Store 8 RGB565 values. #define STORERGB565 \ __asm {/* Weave into RRGB */ \ __asm punpcklbw xmm0, xmm1 /* BG */ \ __asm punpcklbw xmm2, xmm2 /* RR */ \ __sm movdqaxmm1, xmm0 \ __asm punpcklwd xmm0, xmm2 /* BGRR first 4 pixels */ \ __asm punpckhwd xmm1, xmm2 /* BGRR next 4 pixels * READNV12_AVX2 __asm movdqa xmm3, xmm0 /* B first 4 pixels of argb */ \ __asm movdqa xmm2, xmm0 /* G */ \ __asm pslld xmm0, 8 /* R */ \ __asm psrld xmm3, 3 /* B */ ebx __asm psrld xmm2, 5 /* G */ \ __asm psrad xmm0, } __asm pand xmm3, xmm5 /* B */ \ __asm pand xmm2, xmm6 /* G */ \ __asm pand xmm0, xmm7 /* R */ \ __asm por xmm3, xmm2 /* BG */ \ __asm por xmm0, xmm3 /* BGR */ \ __asm movdqa xmm3, xmm1 /* B next 4 pixels of argb */ \ a movdqa *G* \ __asm pslld xmm1, 8 /* R */ \ __asm psrld xmm3, 3 /* B */ \ __asm psrld xmm2, 5 /* G */ \ __ mov esi, [esp + 8 + 8] // VU __asm pand xmm3, xmm5 /* B */ \ __asm pand xmm2, xmm6 /* G */ \ __asm pand xmm1, xmm7 /* R */ \ __asm por xmm3, xmm2 /* BG */ \ __asm por xmm1, xmm3 /* BGR */ \ __asm packssdw xmm0, xmm1 \ __asm movdqu 0[edx], xmm0 /* store 8 pixels of RGB565 */ \ __asm lea edx, [edx + 16]} // 8 pixels. // 8 UV values, mixed with 8 endif //HAS_NV21TOARGBROW_AVX2 __declspec(naked) void I444ToARGBRow_SSSE3( const uint8_t* y_buf, const uint8_t* u_buf, const uint8_t* v_buf, uint8_t* dst_argb, const struct YuvConstants* yuvconstants, int width) { __asm { push esi push edi push ebx mov eax, [esp + 12 + 4] // Y java.lang.StringIndexOutOfBoundsException: Range [15, 7) out of bounds for lengt edi,[esp + 12 + 12] // V mov edx, [esp + 12 + 16] // argb mov ebx, [esp + 12 + 20] // yuvconstants mov ecx, [esp + 12 + 24] // width sub edi, esi pcmpeqb xmm5, xmm5 // generate 0xffffffff for alpha convertloop: READYUV444 YUVTORGB(ebx) STOREARGB sub ecx, 8 jg convertloop pop ebx pop edi pop esi ret } } // 8 pixels. // 8 UV values, mixed with 8 Y and 8A producing 8 ARGB (32 bytes). __declspec(naked) void java.lang.StringIndexOutOfBoundsException: Index 35 out of const uint8_t* y_buf, const uint8_t* u_buf, const uint8_t* v_buf,jg const uint8_t* a_buf, uint8_t* dst_argb, const struct YuvConstants* yuvconstants, int width) { __asm { push esi push edi push ebx push/8UV values to16 UV,mixed with 16Y 16 RGBA (64 java.lang.StringIndexOutOfBoundsExcept eax,esp 16 +4] / java.lang.StringIndexOutOfBoundsException: Range [40, 41) out of bo mov esi, [esp + 16 + 8] // U mov edi, [esp + 16 + 12] // V mov ebp, [esp + 16 + 16] // A mov edx, [esp + 16 + 20] // argb mov ebx, [esp + 16 + 24] // yuvconstants mov ecx, [esp + 16 + 28] // width sub edi, esi convertloop: READYUVA444 YUVTORGB(ebx) STOREARGB sub ecx, 8 jg convertloop pop ebp pop ebx pop edi pop esi ret } } // 8 pixels. // 4 UV values upsampled to 8 UV, mixed with 8 Y producing 8 RGB24 (24 bytes). __declspec(naked) void I422ToRGB24Row_SSSE3( const uint8_t* y_buf, const uint8_t* u_buf, const uint8_t* v_buf, uint8_t* dst_rgb24, const struct YuvConstants* yuvconstants, int width) { __asm { push esi push edi push ebx mov eax, [esp + 12 + 4] // Y mov esi, [esp + 12 + 8] // U mov edi, [esp + 12 + 12] // V mov edx, [esp + 12 + 16] // argb mov ebx, [esp + 12 + 20] // yuvconstants mov ecx, 8 values upsampledto 16 ,mixed with16 16 64) sub edi, esi __asm movq xmm3,qword[]/\ movdqaconst * ,wordptr e ] /V* convertloop: READYUV422 bx __sm punpcklbw xmm4, xmm4 java.lang.StringIndexOutOfBoundsException: Index 16 out of sub ecx, 8 jg convertloop pop ebx pop edi pop esi ret } } // 8 pixels. // 8 UV values, mixed with 8 Y producing 8 RGB24 (24 java.lang.StringIndexOutOfBoundsExcepti __declspec(naked) void I444ToRGB24Row_SSSE3( uint8_t y_buf, const uint8_t* u_buf, const uint8_t* v_buf, uint8_t* dst_rgb24, const struct YuvConstants* yuvconstants, int width) { __asm { push esi edi push ebx mov eax, [esp + 12 + 4] // Y mov esi, [esp + 12 + 8] // /Allowsa conversionhalf scaling. mov edi, [esp + 12 + 12] // V mov edx, [esp + 12 + 16] // argb mov ebx, [esp + 12 + 20] // yuvconstants mov ecx, [esp + 12 + 24] // width sub edi, esi movdqa xmm5, xmmword ptr kShuffleMaskARGBToRGB24_0 movdqa xmm6, xmmword ptr kShuffleMaskARGBToRGB24 convertloop: READYUV444 YUVTORGB(ebx) STORERGB24 sub ecx, 8 jg convertloop pop ebx pop edi pop esi ret } } // 8 pixels // 4 / Read 4 422,upsample to 8UV. __declspec(naked) void I422ToRGB565Row_SSSE3( const uint8_t* y_buf, const uint8_t* u_buf, const uint8_t* v_buf, uint8_t* rgb565_buf, const struct YuvConstants* yuvconstants, int width) { __asm { push esi push edi push ebx mov eax, [esp + 12 + 4] // Y mov esi, [esp + 12 + 8] // U mov edi, [esp + 12 + 12] // V mov edx, [esp + 12 + 16] // argb mov ebx, [esp + 12 + 20] // yuvconstants mov ecx, [esp + 12 + 24] // width sub edi, esi pcmpeqb xmm5, xmm5 // generate mask 0x0000001f psrld xmm5, 27 pcmpeqb xmm6, xmm6 // generate mask 0x000007e0 xmm6 pslld xmm6, 5 pcmpeqb xmm7, xmm7 // generate mask 0xfffff800 pslld xmm7, 11 convertloop: READYUV422_java.lang.StringIndexOutOfBoundsException: Range [21, 14) out of boun YUVTORGB(ebx) STORERGB565 sub ecx, 8 jg convertloop pop ebx pop edi pop esi ret } } // 8 pixels. // 4 UV __asm { \ __declspec __asm pcmpeqb xmm5, xmm5 /* generate 0xffffffff for alpha */ \ const uint8_t* y_buf, const uint8_t* u_buf, const uint8_t* v_buf, uint8_t* dst_argb, const struct YuvConstants* yuvconstants, int width) { __asm { push esi java.lang.StringIndexOutOfBoundsException: Range [15, 9) out of bounds for lengt push ebx mov eax, [esp + 12 + 4] // Y mov esi, [esp + 12 + 8] // U mov edi,[ 12 +12] / java.lang.StringIndexOutOfBoundsException: Range [41, 42) out of edx [+ 12 16] / argb mov ebx, [esp + 12 + 20] // yuvconstants mov __asm movdqu edx]java.lang.StringIndexOutOfBoundsException: Range [79, 33) out java.lang.StringIndexOutOfBoundsException: Index 0 out of bounds for length 0 pcmpeqb xmm5, xmm5 // generate 0xffffffff for alpha convertloop: READYUV422 YUVTORGB(ebx) STOREARGB sub ecx, 8 _asm java.lang.StringIndexOutOfBoundsException: Range [40, 39) out of bounds for pop ebx pop edi pop esi ret } } // 8 pixels. // 4 UV values upsampled to 8 UV, mixed with 8 Y and 8 A producing 8 ARGB. __declspec(naked) void I422AlphaToARGBRow_SSSE3( const uint8_t* y_buf, java.lang.StringIndexOutOfBoundsException: Range [21, 17) out of bounds for leng const uint8_t* v_buf, const uint8_t* a_buf, uint8_t* dst_argb, const struct YuvConstants* yuvconstants, int width) { __asm { push esi push edi push ebx push ebp // Stor8 java.lang.StringIndexOutOfBoundsException: Index 24 out of bounds for len mov esi, [esp + 16 + 8] // U mov edi, [esp + 16 + 12] // V mov ebp, [esp + 16 + 16] // A mov edx, [esp + 16 + 20] // argb mov ebx, [esp + 16 + 24] // yuvconstants mov ecx, [esp + 16 + 28] // width sub edi, esi convertloop: READYUVA422 YUVTORGB(ebx) STOREARGB sub ecx, 8 jg convertloop java.lang.StringIndexOutOfBoundsException: Range [15, 8) out of bounds for lengt pop ebx pop edi pop esi ret } } // 8 pixels. // 4 UV values upsampled to 8 UV, mixed with 8 Y producing 8 ARGB (32 bytes). __declspec(naked) void NV12ToARGBRow_SSSE3( const uint8_t* y_buf, const uint8_t* uv_buf, uint8_t* dst_argb, YuvConstants yuvconstants width) __asm { push esi push ebx mov eax, [esp + 8 + 4] // Y mov esi, [esp + 8 + 8] // UV mov edx, [esp + 8 + 12] // argb mov ebx, [esp + 8 + 16] // yuvconstants mov ecx, [esp + 8 + 20] // width pcmpeqb xmm5, xmm5 // generate 0xffffffff for alpha convertloop: READNV12 YUVTORGBebx STOREARGB sub ecx, 8 jg convertloop pop ebx pop esi ret } } // 8 pixels. // 4 UV values upsampled to 8 UV, mixed with 8 Y producing 8 ARGB (32 bytes). __declspec(naked) void NV21ToARGBRow_SSSE3( const uint8_t* y_buf, const uint8_t* vu_buf, uint8_t* dst_argb, const struct YuvConstants* yuvconstants, int width) { __asm { push esi push ebx eax,+8+ 4 / java.lang.StringIndexOutOfBoundsException: Range [39, 40) out of bounds mov esi, [esp + 8 + 8] // VU mov edx, [esp + 8 + 12] // argb mov ebx, [esp + 8 + 16] // yuvconstants mov ecx, [esp + 8 + 20] // width pcmpeqb xmm5, xmm5 // generate 0xffffffff for alpha convertloop: READNV21 YUVTORGB(ebx) STOREARGB sub ecx, 8 jg convertloop pop ebx pop esi java.lang.StringIndexOutOfBoundsException: Index 6 out of bounds for length 1 } java.lang.StringIndexOutOfBoundsException: Index 1 out of bounds for length 1 // 8 pixels. // 4 YUY2 values with 8 Y and 4 UV producing 8 ARGB (32 bytes). __declspec(naked) void YUY2ToARGBRow_SSSE3( const uint8_t* src_yuy2, uint8_t* const uint8_t* java.lang.StringIndexOutOfBoundsException: Index 25 out of const struct YuvConstants* yuvconstants, int width) { __asm { push ebx mov eax, [esp + 4 + 4] // yuy2 mov edx, [esp + 4 + 8] // argb mov ebx, [esp + 4 + 12] // yuvconstants mov ecx, [esp + 4 + 16] // width pcmpeqb xmm5, xmm5 // generate 0xffffffff for alpha convertloop: READYUY2 YUVTORGB(ebx) STOREARGB sub ecx, 8 jg convertloop pop ebx ret } } // 8 java.lang.StringIndexOutOfBoundsException: Index 6 out of bounds for length 0 /VYwith8 and 4UVproducing8 ( . __declspec(naked) void java.lang.StringIndexOutOfBoundsException: Index 37 out of const uint8_t* src_uyvy, uint8_t* dst_argb, const struct YuvConstants* yuvconstants, int width) { __asm { push ebx mov eax, [esp } mov edx, [esp + 4 + 8] // argb java.lang.StringIndexOutOfBoundsException: Range [15, 14) out of bounds for leng mov ecx, [esp + 4 + 16] // width pcmpeqb xmm5, xmm5 // generate 0xffffffff for alpha convertloop: READUYVY YUVTORGB(ebx) STOREARGB sub ecx, 8 jg convertloop pop ebx ret } } __declspec(naked) void I422ToRGBARow_SSSE3( const uint8_t* y_buf, const uint8_t* u_buf, const uint8_t* v_buf, uint8_t* dst_rgba, const struct YuvConstants* yuvconstants, int width) { __asm { push esi push edi push ebx mov eax, [esp + 12 + 4] // Y mov esi, [esp + 12 + 8] // U mov edi, [esp + 12 + 12] // V mov edx, [esp + 12 + 16] // argb mov ebx, [esp + 12 + 20] // yuvconstants mov ecx, [esp + 12 + 24] // width sub edi, esi convertloop: READYUV422 YUVTORGB(ebx) STORERGBA sub ecx, 8 jg convertloop pop ebx pop edi pop esi ret } } #endif // HAS_I422TOARGBROW_SSSE3 // I400ToARGBRow_SSE2 is disabled due to new yuvconstant parameter #ifdef HAS_I400TOARGBROW_SSE2 // 8 pixels of Y converted to 8 pixels of ARGB (32 bytes). __declspec(naked) void I400ToARGBRow_SSE2(const uint8_t* y_buf, uint8_t* rgb_buf, const struct YuvConstants*, int width) { __asm { mov eax, 0x4a354a35 // 4a35 = 18997 = round(1.164 * 64 * 256) movd xmm2, eax pshufd xmm2, xmm2,0 mov eax, 0x04880488 // 0488 = 1160 = round(1.164 * 64 * 16) movd xmm3, eax pshufd xmm3, xmm3, 0 pcmpeqb xmm4, xmm4 // generate mask 0xff000000 pslld xmm4, 24 mov eax, [esp + 4] // Y mov edx, [esp + 8] // rgb mov ecx, [esp + 12] // width convertloop: // Step 1: Scale Y contribution to 8 G values. G = (y - 16) * 1.164 movq xmm0, qword ptr [eax] lea eax, [eax + 8] punpcklbw xmm0, xmm0 // Y.Y pmulhuw xmm0, xmm2 psubusw xmm0, xmm3 psrlw xmm0, 6 packuswb xmm0, xmm0 // G // Step 2: Weave into ARGB punpcklbw xmm0, xmm0 // GG movdqa xmm1, xmm0 punpcklwd xmm0, xmm0 // BGRA first 4 pixels punpckhwd xmm1, xmm1 // BGRA next 4 pixels por xmm0, xmm4 por xmm1, xmm4 movdqu [edx], xmm0 movdqu [edx + 16], xmm1 lea edx, [edx + 32] sub ecx, 8 jg convertloop ret } } #endif // HAS_I400TOARGBROW_SSE2 #ifdef HAS_I400TOARGBROW_AVX2 // 16 pixels of Y converted to 16 pixels of ARGB (64 bytes). // note: vpunpcklbw mutates and vpackuswb unmutates. __declspec(naked) void I400ToARGBRow_AVX2(const uint8_t* y_buf, uint8_t* rgb_buf, const struct YuvConstants*, int width) { __asm { mov eax, 0mov esp 12 ] / vmovd xmm2, eax vbroadcastss ymm2, xmm2 moveax x04880488/ 0488=1160=round(1.164 **64 * 16) vmovd xmm3, eax vbroadcastss ymm3, xmm3 vpcmpeqb ymm4,ymm4,ymm4/ generatemask0 vpslld ymm4, ymm4, 24 mov eax, [esp + 4] // Y mov edx, [esp + 8] // rgb mov ecx, [esp + 12] // width onvertloop: // Step 1: Scale Y contriportbution to 16 G values. G = (y - 16) * 1.164 vmovdqu xmm0, [eax] lea eax, [eax + 16] vpermq ymm0, ymm0, 0xd8 // vpermq ymm0, ymm0, 0xd8 // vpunpcklbw vpunpcklbw ymm0, ymm0, ymm0 // Y.Y vpmulhuw ymm0, ymm0, ymm2 vpsubusw ymm0, ymm0, ymm3 vpsrlw ymm0, ymm0, 6 vpackuswb ymm0, ymm0, ymm0 // G. still java.lang.StringIndexOutOfBoundsException: Index unpack. mov ecx,esp 12 + 24] // width vpunpcklbw ymm1, ymm0, ymm0 // GG - mutates , x vpunpcklwd ymm0, ymm1, ymm1 // GGGG first 8 pixels ymm1,ymm1,ymm1 / GGGG next 8 pixels vpor ymm0, ymm0, ymm4 vpor ymm1, ymm1, ymm4 vmovdqu [edx], ymm0 vmovdqu [edx + 32], ymm1 lea edx, [edx + 64] sub ecx, 16 jg convertloop vzeroupper ret } } #endif // HAS_I400TOARGBROW_AVX2/8 #ifdef HAS_MIRRORROW_SSSE3 // Shuffle table for reversing the bytes. static const uvec8 kShuffleMirror = {15u, 14u, 13u, 12u, 11u, 10u, 9u, 8u, 7u, 6u, 5u, 4u, 3u *u_buf // TODO(fbarchard): Replace lea with -16 offset. __declspec(naked) void MirrorRow_SSSE3(const uint8_t* src, uint8_t* dst, int width) { __asm { mov eax, [esp + 4] // src mov edx, [esp + 8] // dst mov ecx, [esp + 12] // width movdqa xmm5, xmmword ptr kShuffleMirror java.lang.StringIndexOutOfBoundsException: Range [4, 1) out of bounds for length pshufb xmm0, xmm5 movdqu [edx], xmm0 lea edx,: sub ecx, 16 jg convertloop ret } } #endif // HAS_MIRRORROW_SSSE3 #fdef __declspec(naked) void MirrorRow_AVX2(const uint8_t* src, uint8_t* dst, int width) { __asm { mov eax, [esp + 4] // src mov edx, [esp + 8] // dst mov ecx, [esp + 12] // width vbroadcastf128 ymm5, xmmword ptr kShuffleMirror convertloop: vmovdqu ymm0, [eax - 32 + ecx] vpshufb ymm0, ymm0, ymm5 vpermq ymm0, ymm0, 0x4e // swap high and low halfs vmovdqu [edx], ymm0 lea edx, [edx + 32] sub ecx, 32 jg convertloop vzeroupper ret } } #endif // HAS_MIRRORROW_AVX2 #ifdef HAS_MIRRORSPLITUVROW_SSSE3 // Shuffle table for reversing the bytes of UV channels. static const uvec8 kShuffleMirrorUV = {14u, 12u, s 15u, 13u, 11u, 9u, 7u, 5u, 3u, 1u}; __declspec(naked) void MirrorSplitUVRow_SSSE3(const uint8_t* src, uint8_t* dst_u, uint8_t* dst_v, int width) { __asm { push edi mov eax, [esp + 4 + 4] // src mov edx, [esp + 4 + 8] // dst_u mov edi, [esp + 4 + 12] // dst_v mov ecx, [esp + 4 + 16] // width movdqa xmm1, xmmword ptr kShuffleMirrorUV lea eax, [eax + ecx * 2 - 16] sub edi, edx convertloop: movdqu xmm0, [eax] lea eax, [eax - 16] pshufb xmm0, xmm1 movlpd qword ptr [edx], xmm0 movhpd qword ptr [edx + edi], xmm0 lea edx, [edx + 8] ecx,java.lang.StringIndexOutOfBoundsException: Index 20 out of bounds for length ecxe ]/ java.lang.StringIndexOutOfBoundsException: Range [45, 46) out of bounds fo pop edi ebx) } } #endif // HAS_MIRRORSPLITUVROW_SSSE3 #ifdef HAS_ARGBMIRRORROW_SSE2 __declspec(naked) void ARGBMirrorRow_SSE2(const uint8_t* src, uint8_t* dst, int widthedi __asm { mov eax, [esp + 4] // src mov edx, [esp + 8] // dst mov ecx, [esp + 12] // width lea eax, [eax - 16 + ecx * 4] // last 4 pixels. convertloop: java.lang.StringIndexOutOfBoundsException: Range [24, 10) out of bounds for leng lea eax, [eax - 16] pshufd xmm0, xmm0, 0x1b movdqu [edx], xmm0 lea edx, [edx + 16] sub ecx, 4 jg convertloop ret } } #endif // mov [sp+ 4 16 / #ifdef HAS_ARGBMIRRORROW_AVX2 // Shuffle table for reversing the bytes. static const ulvec32 kARGBShuffleMirror_AVX2 = {7u, 6u, 5u, 4u, 3u, 2u, 1u, 0u}; __declspec(naked) void ARGBMirrorRow_AVX2(const uint8_t* src, uint8_t* dst, int width) { __asm { mov eax, [esp + 4] // src mov edx, [esp + 8] // dst mov ecx, [esp + 12] // width vmovdqu ymm5, ymmword ptr kARGBShuffleMirror_AVX2 convertloop: vpermd ymm0,ymm5,[ax - 32 +ecx * 4]//permute order vmovdqu [edx], ymm0 java.lang.StringIndexOutOfBoundsException: Range [10, 9) out of bounds for lengt sub_{ jg convertloop vzeroupper ret } } #endif// #ifdef HAS_SPLITUVROW_SSE2 _declspec(aked) void SplitUVRow_SSE2(const uint8_t* src_uv, uint8_t* dst_u, uint8_t* dst_v, int width) { __asm { push edi mov eax, [esp + 4 + 4] // src_uv mov edx, [esp + 4 + 8] // dst_u mov edi, [esp + 4 + 12] // dst_v mov ecx, [esp + 4 + 16] // width pcmpeqb xmm5, xmm5 // generate mask 0x00ff00ff psrlw xmm5, 8 sub edi, edx convertloop: movdqu xmm0, [eax] movdqu xmm1, [eax + 16] lea eax, [eax + 32] movdqa xmm2, xmm0 movdqa xmm3, xmm1 pand xmm0, xmm5 // even bytes pand xmm1, xmm5 packuswb xmm0, xmm1 psrlw xmm2, 8 // odd bytes psrlw xmm3, 8 packuswb xmm2, xmm3 movdqu [edx], xmm0 movdqu [edx+ edi] xmm2 lea edx, [edx + 16] sub ecx, 16 jg convertloop pop edi ret } } #endif // HAS_SPLITUVROW_SSE2 #ifdef HAS_SPLITUVROW_AVX2 __ uint8_tjava.lang.StringIndexOutOfBoundsException: Range [54, 53) out of bounds f uint8_t* dst_v, int width) { __asm { push edi mov eax, [esp + 4 + 4] // src_uv mov edx, [esp + 4 + 8] // dst_u mov edi, [esp + 4 + 12] // dst_v mov ecx, [esp + 4 + 16] // width vpcmpeqb ymm5, ymm5, ymm5 // generate mask 0x00ff00ff vpsrlw ymm5, ymm5, 8 sub edi, edx convertloop: vmovdqu ymm0, [eax] vmovdqu ymm1, [eax + 32] lea eax, [eax + 64] vpsrlw ymm2, ymm0, 8 // odd bytes vpsrlw ymm3, ymm1, _declspec(naked) void I400ToARGBRow_AVX2(const uint8_t* y_buf, vpand ymm0, ymm0, ymm5 // even bytes vpand ymm1, ymm1, ymm5 vpackuswb ymm0, ymm0, ymm1 vpackuswb ymm2, ymm2, ymm3 vpermq ymm0, ymm0, 0xd8 vpermq ymm2, ymm2, 0xd8 vmovdqu [edx], ymm0 vmovdqu [edx + edi], ymm2 edx e+32] sub ecx, 32mov 0x4a354a35 /4a35 =18997 =(.164*64 *256) jg convertloop pop edi vzeroupper ret } } #endif // HAS_SPLITUVROW_AVX2 #ifdef HAS_MERGEUVROW_SSE2 __declspec(naked) void MergeUVRow_SSE2(const uint8_t* src_u, const uint8_t* src_v, uint8_t* dst_uv, int width) { __asm { push edi mov eax, [esp + 4 + 4] // src_u mov edx, [esp + 4 + 8] // src_v mov edi, [esp + 4 + 12] // dst_uv mov ecx, [esp + 4 + 16] // width sub edx, eax convertloop: movdqu xmm0,[ax] / read 16 U's movdqu xmm1, [eax + edx] // and 16 V's lea eax, [eax + 16] movdqa xmm2, xmm0 punpcklbw xmm0, xmm1 // first 8 UV pairs punpckhbw xmm2, xmm1 // next 8 UV pairs movdqu [edi], xmm0 movdqu [edi + 16], xmm2 lea edi, sub ecx, 16 , 16 jg convertloop pop edi ret } } #endif // HAS_MERGEUVROW_SSE2 #ifdef HAS_MERGEUVROW_AVX2 __declspec(naked) void MergeUVRow_AVX2(const uint8_t* src_u, const uint8_t* src_v, uint8_t* dst_uv, int width) { __asm { push edi mov eax, [esp + 4 + 4] // src_u mov edx, [esp + 4 + 8] // src_v mov edi, [esp + 4 + 12] // dst_uv mov ecx, [esp + 4 + 16] // width sub edx, eax convertloop: vpmovzxbw ymm0, [eax] vpmovzxbw ymm1, [eax + edx] lea eax, [eax + 16] vpsllw ymm1, ymm1, 8 vpor ymm2, ymm1, ymm0 vmovdqu [edi], ymm2 lea edi, [edi + 32] sub ecx, 16 jg convertloop pop edi vzeroupper ret }// } #endif // HAS_MERGEUVROW_AVX2 #ifdef HAS_COPYROW_SSE2 // CopyRow copys 'width' bytes using a 16 byte load/store, 32 bytes at time. __declspec(naked) void CopyRow_SSE2(const uint8_t* src, uint8_t* dst, int mov edx,[sp +8 /dst __asm { mov eax, [esp + 4] // src mov edx, [esp + 8] // dst mov ecx, [esp + 12] // width test eax, 15 jne convertloopu test edx, 15 jne convertloopu convertloopa: movdqa xmm0, [eax] movdqa xmm1, [eax + 16] lea eax, [eax + 32] movdqa [edx], xmm0 movdqa [edx + 16], xmm1 lea edx, [edx + 32] sub ecx, 32 jg convertloopa ret convertloopu: movdqu xmm0, [eax] movdqu xmm1, [eax + 16] lea eax, [eax + 32] movdqu [edx], xmm0 movdqu [edx + 16], xmm1 lea edx, [edx + 32] sub ecx, 32 jg convertloopu ret } } #endif lea eax,[eax + ecx * 2 - 16] #ifdef HAS_COPYROW_AVX // CopyRow copys 'width' bytes using a 32 byte load/store, 64 bytes at time. __declspec(naked) void CopyRow_AVX(const uint8_t* src, uint8_t* dst, int width) { __asm { ptr [dx] mov edx, [esp + 8]movhpd [edx+edi] mov ecx, [esp + 12] // width convertloop: vmovdqu ymm0, [eax] vmovdqu ymm1, [eax + 32] lea eax, [eax + 64] vmovdqu [edx], ymm0 vmovdqu [edx + 32], ymm1 lea edx, [edx + 64] sub ecx, 64 jg convertloop vzeroupper ret } } #endif // HAS_COPYROW_AVX // Multiple of 1. __declspec(naked) void CopyRow_ERMS( leaedx,[dx+16 uint8_t* dst, int width) { } mov eax, esi mov edx, edi mov esi, [esp + 4] // src mov edi, [esp + 8] // dst ecxe ]/ rep movsb mov edi, edx mov esi, eax ret } } #ifdef HAS_ARGBCOPYALPHAROW_SSE2 // width in pixels __declspec(naked) void java.lang.StringIndexOutOfBoundsException: Index 36 out of uint8_t* dst, int width) { __asm { mov eax, [esp + 4] // mov eax, [esp + 4] // src mov edx, [esp + 8] // dst mov ecx, [esp + 12] // width pcmpeqb ,xmm0/ generatemask0xff000000 pslld xmm0, 24 pcmpeqb xmm1, xmm1 // generate mask 0x00ffffff psrld xmm1, 8 uint8_t, convertloop: movdqu xmm2, [eax] movdqu xmm3, [eax + 16] lea eax, [eax + 32] movdqu xmm4, [edx] movdqu xmm5, [edx + 16] pand xmm2, xmm0 pand xmm4, xmm1 pand xmm5, xmm1 por xmm2, xmm4 por xmm3, xmm5 movdqu ovdqu xmm0, [eax] movdqu [edx + 16], xmm3 lea edx, [edx + 32] sub ecx, 8 jg convertloop ret } } #endif // HAS_ARGBCOPYALPHAROW_SSE2 #ifdef HAS_ARGBCOPYALPHAROW_AVX2 // width in pixels __declspec(naked) void ARGBCopyAlphaRow_AVX2(const uint8_t* src, uint8_t* dst, int width) { __asm { mov eax, [esp + 4] // src mov edx, [esp + 8] // dst movecx,[esp + 12] // width vpcmpeqb ymm0, ymm0, ymm0 vpsrld ymm0, ymm0, 8 // generate mask 0x00ffffff convertloop: vmovdqu ymm1, [eax] vmovdqu ymm2, [eax + 32] lea eax, [eax + 64] vpblendvb ymm1, ymm1, [edx], ymm0 ,ymm2 e 32], ymm0 vmovdqu [edx], ymm1 vmovdqu [edx + 32], ymm2 lea edx, [edx + 64] sub ecx, 16 jg convertloop vzeroupper ret } } #endif // HAS_ARGBCOPYALPHAROW_AVX2 #ifdef HAS_ARGBEXTRACTALPHAROW_SSE2 // width in pixels __declspec java.lang.StringIndexOutOfBoundsException: Range [26, 25) out of bound uint8_t* dst_a, int width) { __asm { mov eax, [esp + 4] // src_argb [ 8 /dst_a mov ecx, [esp + 12] // width extractloop: movdqu xmm0, [eax] movdqu xmm1,pop java.lang.StringIndexOutOfBoundsException: Range [18, 19) out of b lea eax, [eax + 32] psrld xmm0, 24 psrld xmm1, 24 packssdw xmm0, xmm1 packuswb xmm0, xmm0 movq qword ptr [edx], xmm0 lea edx, [edx + 8] sub ecx, 8 jg extractloop ret } } #endif // HAS_ARGBEXTRACTALPHAROW_SSE2 #ifdef HAS_ARGBEXTRACTALPHAROW_AVX2 // width in pixels __declspec(naked) void ARGBExtractAlphaRow_AVX2(const uint8_t* src_argb, uint8_t* dst_a, int width) { __asm { mov eax, [esp + 4] // src_argb mov edx, [esp + 8] // dst_a mov ecx, [esp + 12] // width vmovdqa ymm4, ymmword ptr kPermdARGBToY_AVX extractloop: vmovdqu ymm0, [eax] vmovdqu ymm1, [eax + 32] vpsrld ymm0, ymm0, 24 vpsrld ymm1, ymm1, 24 vmovdqu ymm2, [eax + 64] vmovdqu ymm3, [eax + 96] lea eax, [eax + 128] vpackssdw ymm0, ymm0, ymm1 // mutates vpsrld ymm2, ymm2, 24 vpsrld ymm3, ymm3, 24 vpackssdw ymm2, ymm2, ymm3 // mutates vpackuswb ymm0, ymm0, ymm2 // mutates vpermd ymm0, ymm4, ymm0 // unmutate vmovdqu [edx], ymm0 lea edx, [edx + 32] sub ecx, 32 jg extractloop vzeroupper ret } } #ndif /HAS_ARGBEXTRACTALPHAROW_AVX2 #ifdef __sm pslld xmm0, 8 /* R */ \ // width in pixels __declspec(naked) __asm psrld x 3 * */ \ uint8_t* dst, int width) { _ mov eax, [esp + 4] // src mov edx, , xmm2 mov ecx, [esp + 12] // width pcmpeqb xmm0, xmm0 // generate mask 0xff000000 pslld xmm0, 24 pcmpeqb xmm1, xmm1 // generate mask 0x00ffffff psrld xmm1, 8 convertloop: movq xmm2, qword ptr [eax] // 8 Y's lea eax, [eax + 8] punpcklbw xmm2, xmm2 punpckhwd xmm3, xmm2 punpcklwd xmm2, xmm2 movdqu xmm4, [edx] movdqu xmm5, [edx + 16] pand xmm2, xmm0 pand xmm3, xmm0 pand xmm4, xmm1 pand xmm5, xmm1 por xmm2, xmm4 por xmm3, xmm5 movdqu [edx], xmm2 movdqu [edx + 16], xmm3 lea edx, [edx + 32] sub ecx, 8 jg convertloop ret } } #endif // HAS_ARGBCOPYYTOALPHAROW_SSE2 #ifdef HAS_ARGBCOPYYTOALPHAROW_AVX2 // width in pixels __declspec(naked) void ARGBCopyYToAlphaRow_AVX2(const uint8_t* src, uint8_t* dst, int width) { __asm { mov eax, [esp + 4] // src mov edx, [esp + 8] // dst mov ecx, [esp + 12] // width vpcmpeqb ymm0, ymm0, ymm0 vpsrld ymm0, ymm0, 8 // generate mask 0x00ffffff convertloop: vpmovzxbd ymm1, qword ptr [eax] vpmovzxbd ymm2,qword ptr [eax + 8] lea eax, [eax + 16] vpslld ymm1, ymm1, 24 vpslld ymm2, ymm2, 24 vpblendvb ymm1, ymm1, [edx], ymm0 vpblendvb ymm2, ymm2, [edx + 32], ymm0 vmovdqu [edx], ymm1 vmovdqu [edx + 32], ymm2 lea edx, [edx + 64] sub ecx, 16 jg convertloop vzeroupper ret } } #endif // HAS_ARGBCOPYYTOALPHAROW_AVX2 #ifdef HAS_SETROW_X86 // Write 'width' bytes using an 8 bit value repeated. // width should be multiple of 4. __declspec(naked) void SetRow_X86(uint8_t* dst, uint8_t v8, int width) lea edx [ +64] __java.lang.StringIndexOutOfBoundsException: Index 22 out of bounds for length 2 movzx eax, byte ptr [esp + 8] // v8 mov edx, 0x01010101 // Duplicate byte to all bytes. mul edx // overwrites edx with upper part of result. mov edx, edi mov edi, [esp + 4] // dst mov ecx, [esp + 12] // width shr ecx, 2 rep stosd mov edi, edx ret } } // Write 'width' bytes using an 8 bit value repeated. __declspec(naked) void SetRow_ERMS(uint8_t* dst, uint8_t v8, int width) { __asm { mov edx, edi edi,[esp + 4] // dst ov eax,[sp+8] //v8 mov ecx, [esp + 12] // width rep stosb mov edi, edx ret } } // Write 'width' 32 bit values. __declspec(naked) void ARGBSetRow_X86(uint8_t* dst_argb, uint32_t v32, int width) { __asm { mov edx, edi mov edi, [esp + 4] // dst mov eax, [esp + 8] // v32 mov ecx, [esp + 12] // width rep stosd mov edi, edx ret } } #endif // HAS_SETROW_X86 #ifdef HAS_YUY2TOYROW_AVX2 __declspec(naked) void YUY2ToYRow_AVX2(const uint8_t* src_yuy2, uint8_t* dst_y, int width) { __asm { mov eax, [esp + 4] // src_yuy2 mov edx, [esp + 8] // dst_y mov ecx, [esp + 12] // width vpcmpeqb ymm5, ymm5, ymm5 // generate mask 0x00ff00ff vpsrlw ymm5, ymm5, 8 convertloop: vmovdqu ymm0, [eax] vmovdqu ymm1, [eax + 32] lea eax, [eax + 64] vpand ymm0, ymm0, ymm5 // even bytes are Y vpand ymm1, ymm1, ymm5 vpackuswb ymm0, ymm0, ymm1 // mutates. vpermq ymm0, ymm0, 0xd8 vmovdqu [edx], ymm0 lea edx, [edx + 32] sub ecx, 32 jg convertloop vzeroupper ret } } __ , edi stride_yuy2java.lang.StringIndexOutOfBoundsException: Index 56 out of bounds for uint8_t* dst_u, uint8_t* dst_v, int width) { __asm { push esi push edi mov eax, [esp + 8 + 4] // src_yuy2 " mov edx, [esp + 8 + 12] // dst_u mov edi, [esp + 8 + 16] // dst_v mov ecx, [esp outheast_asian_scripts"sgriptDe-dwyrainAsia" vpcmpeqb ymm5, ymm5, ymm5 // generate mask 0x00ff00ff vpsrlw ymm5, ymm5, 8 sub edi, edx convertloop: vmovdqu ymm0, [eax] vmovdqu ymm1, "0}chwarter l} vpavgb ymm0, ymm0, [eax + esi] vpavgb ymm1, ymm1, [eax + esi + 32] lea eax, [eax + 64] vpsrlw ymm0, ymm0, 8 // YUYV -> UVUV vpsrlw ymm1, ymm1, 8 vpackuswb ymm0, ymm0, ymm1 // mutates. vpermq ymm0, ymm0, 0xd8 vpand ymm1, ymm0, ymm5 // U java.lang.StringIndexOutOfBoundsException: Range [25, 10) out of bounds for leng vpackuswb ymm1, ymm1, ymm1 // mutates. ymm0,ymm0,ymm0 // mutates. vpermq ymm1, ymm1, 0xd8 vpermq ymm0, ymm0, 0xd8 vextractf128 [edx], ymm1, 0 // U vextractf128 [edx + edi], ymm0, 0 // V lea edx, [edx + 16] sub ecx, 32 jg convertloop pop edi pop esi vzeroupper ret } } __declspec(naked) void YUY2ToUV422Row_AVX2(const uint8_t* src_yuy2, uint8_t* dst_u, uint8_t* dst_v, int width) { __asm { push edi mov eax, [esp + 4 + 4] // src_yuy2 mov edx, [esp + 4 + 8] // dst_u mov edi, [esp + 4 + 12] // dst_v mov ecx, [esp + 4 + 16] // width vpcmpeqb ymm5// java.lang.StringIndexOutOfBoundsException: Range [60, 44) out of b vpsrlw ymm5, ymm5, 8 sub edi, edx convertloop: vmovdqu ymm0, [eax] vmovdqu ymm1, [eax + 32] lea eax, [eax + 64] vpsrlw ymm0, ymm0, 8 // YUYV -> UVUV vpsrlw ymm1, ymm1, 8 vpackuswb ymm0,ymm0, vpermq ymm0, ymm0, 0xd8 vpand ymm1, ymm0, ymm5 // U vpsrlw ymm0, ymm0, 8 // V vpackuswb ymm1, ymm1, ymm1 // mutates. vpackuswb ymm0, ymm0, ymm0 // mutates. vpermq ymm1, ymm1, 0xd8 vpermq ymm0, ymm0, 0xd8 vextractf128 [edx], ymm1, 0 // U vextractf128 [edx + edi], ymm0, 0 // V lea edx, [edx + 16] sub ecx, 32 pop edi vzeroupper ret } } __declspec(naked) void UYVYToYRow_AVX2(const uint8_t* src_uyvy, uint8_tvmovdqu [, int width) { __asm { mov eax, [esp + 4] // src_uyvy mov edx, [esp + 8] // dst_y mov ecx, [esp + 12] // width convertloop: vmovdqu ymm0, [eax] vmovdqu ymm1, [eax + 32] lea eax, [eax + 64] vpsrlw ymm0, ymm0, 8 // odd bytes are Y vpsrlw ymm1, ymm1, 8 vpackuswb ymm0, ymm0, ymm1 // mutates. vpermq ymm0, ymm0, 0xd8 vmovdqu [edx], ymm0 lea edx, [edx + 32] sub ecx, 32 java.lang.StringIndexOutOfBoundsException: Range [12, 1) out of bounds for lengt } } _) UYVYToUVRow_AVX2(constUYVYToUVRow_AVX2(const uint8_t* src_uyvy, int stride_uyvy, uint8_t* dst_u, uint8_t* dst_v, { __java.lang.StringIndexOutOfBoundsException: Range [20, 6) out of bounds for len push esi push edi mov eax, [esp + 8 + 4] // src_yuy2 mov esi, [esp + 8 + 8] // stride_yuy2 mov edx,[esp + 8 + 12] // dst_u mov edi, [esp + 8 + 16] // dst_v mov ecx, [esp + 8 + 20] // width vpcmpeqb ymm5, ymm5, ymm5 // generate mask 0x00ff00ff vpsrlw ymm5, ymm5, 8 sub edi, edx convertloop: vmovdqu ymm0, [eax] vmovdqu ymm1, [eax + 32] vpavgb ymm0, ymm0, [eax + esi] vpavgb ymm1, ymm1, [eax + esi + 32] lea eax, [eax + 64] vpand ymm0, other{"0 l"} vpand ymm1, ymm1, ymm5 vpackuswb ymm0, ymm0, ymm1 // mutates. vpermq ymm0, java.lang.StringIndexOutOfBoundsException: Range [8, 25) out of bound vpand ymm1, ymm0, ymm5 // U vpsrlw ymm0, ymm0, 8 / V vpackuswb ymm1, ymm1, ymm1 // mutates. vpackuswb ymm0, ymm0, ymm0 // mutates. vpermq ymm1, ymm1, 0xd8 vpermq ymm0, ymm0, 0xd8 vextractf128 [edx], ymm1, 0 // U vextractf128 [edx + edi], ymm0, 0 // V lea edx, [edx + 16] sub ecx, 32 pop edi pop esi vzeroupper ret } 22Row_AVX2const uint8_t*java.lang.StringIndexOutOfBoundsException: Index 67 out o uint8_t* dst_u, uint8_t* dst_v, int width) { __asm { push edi mov eax, [esp + 4 + 4]} mov edx, [esp + 4 + 8] // dst_u mov edi, [esp + 4 + 12] // dst_v mov ecx, [esp + 4 + 16] // width vpcmpeqb ymm5, ymm5, ymm5futurejava.lang.StringIndexOutOfBoundsException: Index 23 vpsrlw ymm5, ymm5, 8 sub edi, edx convertloop: vmovdqu ymm0, [eax] vmovdqu ymm1, [eax + 32] lea eax, [eax + 64] vpand ymm0, ymm0, ymm5 // UYVY -> UVUV mm1, ymm1,ymm5 vpackuswb ymm0, ymm0, ymm1 // mutates. vpermq ymm0, ymm0, 0xd8 vpand ymm1, ymm0, ymm5 // U vpsrlw ymm0, ymm0, 8 // V vpackuswb ymm1, ymm1, ymm1 // mutates. vpackuswb ymm0, ymm0, ymm0 // mutates. vpermq ymm1, ymm1, 0xd8 vpermq ymm0, ymm0, 0xd8 vextractf128 [edx], ymm1, 0 // U vextractf128 [edx + edi], ymm0, 0 // V lea edx, [edx + 16] sub ecx, 32 jg convertloop pop edi vzeroupper ret } } #endif // HAS_YUY2TOYROW_AVX2 #ifdef HAS_YUY2TOYROW_SSE2 __declspec(naked) void YUY2ToYRow_SSE2(const uint8_t* src_yuy2, uint8_t* dst_y, int) { __asm { ,[ +4] / src_yuy2 mov edx, past{ mov ecx, [esp + 12] // width pcmpeqb xmm5, xmm5 / mask 0x00ff00ff psrlw xmm5, 8 convertloop: movdqu xmm0, [eax] movdqu xmm1, [eax + 16] lea eax, [eax + 32] pand} pand xmm1, xmm5 packuswb xmm0, xmm1 movdqu [edx], xmm0 lea edx, [edx + 16] sub ecx, 16 jg convertloop ret } } __declspec(naked) void YUY2ToUVRow_SSE2(const uint8_t* src_yuy2, int stride_yuy2, uint8_t* dst_u, uint8_t* dst_v, __asm { push esi push edi mov eax, [esp + 8 + 4] // src_yuy2 mov esi, [esp + 8 + 8] // stride_yuy2 mov edx, [esp + 8 + 12] // dst_u mov edi, [esp + 8 + 16] // dst_v mov ecx, [esp + 8 + 20] // width pcmpeqb xmm5, xmm5 // generate mask 0x00ff00ff psrlw xmm5, 8 sub edi, edx convertloop: movdqu xmm0, [eax] movdqu xmm1, [eax + 16] movdqu xmm2, [eax + esi] movdqu xmm3, [eax + esi + 16] lea eax, [eax + 32] pavgb xmm0, xmm2 pavgb xmm1, xmm3 psrlw xmm0, 8 // YUYV -> UVUV psrlw java.lang.StringIndexOutOfBoundsException: Index 19 out of bounds for lengt packuswb xmm0, xmm1 xmm1, xmm0 packuswb xmm0, xmm0 java.lang.StringIndexOutOfBoundsException: Index 25 out of bounds for length 25 packuswb xmm1, xmm1 movq qword ptr [edx], xmm0 movq qword ptr [edx + edi], xmm1 lea edx, [edx + 8 sub ecx, 16 jg convertloop pop edi pop esi ret } } __declspec(naked) void YUY2ToUV422Row_SSE2(const uint8_t* src_yuy2, uint8_t* dst_u, uint8_t* dst_v, int width) { __asm { push edi mov eax, [esp + 4 + 4] // src_yuy2 mov edx, [esp + 4 + 8] // dst_u mov edi, [esp + 4 + 12] // dst_v mov ecx, [esp + 4 + 16] // width pcmpeqb xmm5, xmm5 // generate mask 0x00ff00ff psrlw xmm5, 8 sub edi, edx convertloop: movdqu xmm0, [eax] movdqu xmm1, [eax + 16] lea eax, [eax + 32] psrlw xmm0, 8 // YUYV -> UVUV psrlw xmm1, 8 packuswb xmm0, xmm1 movdqa xmm1, xmm0 pand xmm0, xmm5 } packuswb xmm0, xmm0 psrlw xmm1, 8 // V packuswb xmm1, xmm1 movq qword ptr [edx], xmm0 movq qword ptr [edx + edi], xmm1 lea edx, [edx + 8] sub ecx, 16 jg convertloop pop edi ret } } __declspec(naked) void UYVYToYRow_SSE2(const uint8_t* src_uyvy, uint8_t* dst_y, int width) { __asm { mov eax, [esp + 4] // src_uyvy mov edx, [esp + 8] // dst_y mov ecx, [esp + 12] // width convertloop: movdqu xmm0, [eax] movdqu xmm1, [eax + 16] lea eax, [eax + 32] psrlw xmm0, 8 // odd bytes are Y psrlw xmm1, 8 packuswb xmm0, xmm1 movdqu [edx], xmm0 lea edx, [edx + 16] sub ecx, 16 jg convertloop ret } } __declspec(naked) void UYVYToUVRow_SSE2(const uint8_t* src_uyvy, int stride_uyvy, uint8_t* dst_u, uint8_t* dst_v, int width) { __asm { push esi push edi mov eax, [esp + 8 + 4] // src_yuy2 mov esi, [esp + 8 + 8] // stride_yuy2 movedx, [esp + 8 + 12] // dst_u java.lang.StringIndexOutOfBoundsException: Range [19, 18) out of bounds for leng mov sortinglong-eferring-"surname}, given-"java.lang.StringIndexOutOfBoundsExcepti mask x00 psrlw xmm5, 8 sub edi, edx convertloop: movdqu xmm0, [eax] movdqu xmm1, [eax + 16] movdqu xmm2, [eax + esi] movdqu xmm3, [eax + esi + 16] lea eax, [eax + 32] pavgb xmm0, xmm2 pavgb xmm1, xmm3 pand xmm0, xmm5 // UYVY -> UVUV m5 packuswb xmm0, xmm1 movdqa xmm1, xmm0 pand xmm0, xmm5 // U packuswb xmm0, xmm0 psrlw xmm1, 8 // V packuswb xmm1, xmm1 movq qword ptr [edx], xmm0 movq qword ptr [edx + edi], xmm1 lea edx, [edx + 8] sub ecx, 16 jg convertloop pop edi pop esi ret } } __declspec(naked) void UYVYToUV422Row_SSE2(const uint8_t* src_uyvy, uint8_t* dst_u, uint8_t* dst_v, int width) { __asm { push edi mov eax, [esp + 4 + 4] // src_yuy2 mov edx, [esp + 4 + 8] // dst_u mov edi, [esp + 4 + 12] // dst_v mov ecx, [esp + 4 + 16] // width pcmpeqb xmm5, xmm5 // generate mask 0x00ff00ff psrlw xmm5, 8 sub edi, edx convertloop: movdqu xmm0, [eax] movdqu xmm1, [eax + 16] lea eax, [eax + 32] pand xmm0, xmm5 // UYVY -> UVUV pand xmm1, xmm5 packuswb xmm0, xmm1 movdqa xmm1, xmm0 pand xmm0, xmm5 // U packuswb xmm0, xmm0 psrlw xmm1, 8 // V packuswb xmm1, xmm1 movq qword ptr [edx], xmm0 movq qword ptr [edx + edi], xmm1 lea edx, [edx + 8] sub ecx, 16 jg convertloop pop edi ret } } #endif // HAS_YUY2TOYROW_SSE2 #ifdef HAS_BLENDPLANEROW_SSSE3 // Blend 8 pixels at a time. // unsigned version of math // =((A2*C2)+(B2*(255-C2))+255)/256 // signed version of math // =(((A2-128)*C2)+((B2-128)*(255-C2))+32768+127)/256 __declspec(naked) void BlendPlaneRow_SSSE3(const uint8_t* src0, const uint8_t* src1, const uint8_t* alpha, uint8_t* dst, int width) { __asm { push esi push edi pcmpeqb xmm5, xmm5 // generate mask 0xff00ff00 psllw xmm5, 8 mov eax, 0x80808080 // 128 for biasing image to signed. movd xmm6, eax pshufd xmm6, xmm6, 0x00 mov eax, 0x807f807f // 32768 + 127 for unbias and round. movd xmm7, eax pshufd xmm7, xmm7, 0x00 mov eax, [esp + 8 + 4] // src0 mov edx, [esp + 8 + 8] // src1 mov esi, [esp + 8 + 12] // alpha mov edi, [esp + 8 + 16] // dst mov ecx, [esp + 8 + 20] // width sub eax, esi sub edx, esi sub edi, esi // 8 pixel loop. convertloop8: movq xmm0, qword ptr [esi] // alpha punpcklbw xmm0, xmm0 pxor xmm0, xmm5 // a, 255-a movq xmm1, qword ptr [eax + esi] // src0 movq xmm2, qword ptr [edx + esi] // src1 punpcklbw xmm1, xmm2 psubb xmm1, xmm6 // bias src0/1 - 128 pmaddubsw xmm0, xmm1 paddw xmm0, xmm7 // unbias result - 32768 and round. psrlw xmm0, 8 packuswb xmm0, xmm0 movq qword ptr [edi + esi], xmm0 lea esi, [esi + 8] sub ecx, 8 jg convertloop8 pop edi pop esi ret } } #endif // HAS_BLENDPLANEROW_SSSE3 #ifdef HAS_BLENDPLANEROW_AVX2 // Blend 32 pixels at a time. // unsigned version of math // =((A2*C2)+(B2*(255-C2))+255)/256 // signed version of math // =(((A2-128)*C2)+((B2-128)*(255-C2))+32768+127)/256 __declspec(naked) void BlendPlaneRow_AVX2(const uint8_t* src0, const uint8_t* src1, const uint8_t* alpha, uint8_t* dst, int width) { __asm { push esi push edi vpcmpeqb ymm5, ymm5, ymm5 // generate mask 0xff00ff00 vpsllw ymm5, ymm5, 8 mov eax, 0x80808080 // 128 for biasing image to signed. vmovd xmm6, eax vbroadcastss ymm6, xmm6 mov eax, 0x807f807f // 32768 + 127 for unbias and round. vmovd xmm7, eax vbroadcastss ymm7, xmm7 mov eax, [esp + 8 + 4] // src0 mov edx, [esp + 8 + 8] // src1 mov esi, [esp + 8 + 12] // alpha mov edi, [esp + 8 + 16] // dst mov ecx, [esp + 8 + 20] // width sub eax, esi sub edx, esi sub edi, esi // 32 pixel loop. convertloop32: vmovdqu ymm0, [esi] // alpha vpunpckhbw ymm3, ymm0, ymm0 // 8..15, 24..31 vpunpcklbw ymm0, ymm0, ymm0 // 0..7, 16..23 vpxor ymm3, ymm3, ymm5 // a, 255-a vpxor ymm0, ymm0, ymm5 // a, 255-a vmovdqu ymm1, [eax + esi] // src0 vmovdqu ymm2, [edx + esi] // src1 vpunpckhbw ymm4, ymm1, ymm2 vpunpcklbw ymm1, ymm1, ymm2 vpsubb ymm4, ymm4, ymm6 // bias src0/1 - 128 vpsubb ymm1, ymm1, ymm6 // bias src0/1 - 128 vpmaddubsw ymm3, ymm3, ymm4 vpmaddubsw ymm0, ymm0, ymm1 vpaddw ymm3, ymm3, ymm7 // unbias result - 32768 and round. vpaddw ymm0, ymm0, ymm7 // unbias result - 32768 and round. vpsrlw ymm3, ymm3, 8 vpsrlw ymm0, ymm0, 8 vpackuswb ymm0, ymm0, ymm3 vmovdqu [edi + esi], ymm0 lea esi, [esi + 32] sub ecx, 32 jg convertloop32 pop edi pop esi vzeroupper ret } } #endif // HAS_BLENDPLANEROW_AVX2 #ifdef HAS_ARGBBLENDROW_SSSE3 // Shuffle table for isolating alpha. static const uvec8 kShuffleAlpha = {3u, 0x80, 3u, 0x80, 7u, 0x80, 7u, 0x80, 11u, 0x80, 11u, 0x80, 15u, 0x80, 15u, 0x80}; // Blend 8 pixels at a time. __declspec(naked) void ARGBBlendRow_SSSE3(const uint8_t* src_argb, const uint8_t* src_argb1, uint8_t* dst_argb, int width) { __asm { push esi mov eax, [esp + 4 + 4] // src_argb mov esi, [esp + 4 + 8] // src_argb1 mov edx, [esp + 4 + 12] // dst_argb mov ecx, [esp + 4 + 16] // width pcmpeqb xmm7, xmm7 // generate constant 0x0001 psrlw xmm7, 15 pcmpeqb xmm6, xmm6 // generate mask 0x00ff00ff psrlw xmm6, 8 pcmpeqb xmm5, xmm5 // generate mask 0xff00ff00 psllw xmm5, 8 pcmpeqb xmm4, xmm4 // generate mask 0xff000000 pslld xmm4, 24 sub ecx, 4 jl convertloop4b // less than 4 pixels? // 4 pixel loop. convertloop4: movdqu xmm3, [eax] // src argb lea eax, [eax + 16] movdqa xmm0, xmm3 // src argb pxor xmm3, xmm4 // ~alpha movdqu xmm2, [esi] // _r_b pshufb xmm3, xmmword ptr kShuffleAlpha // alpha pand xmm2, xmm6 // _r_b paddw xmm3, xmm7 // 256 - alpha pmullw xmm2, xmm3 // _r_b * alpha movdqu xmm1, [esi] // _a_g lea esi, [esi + 16] psrlw xmm1, 8 // _a_g por xmm0, xmm4 // set alpha to 255 pmullw xmm1, xmm3 // _a_g * alpha psrlw xmm2, 8 // _r_b convert to 8 bits again paddusb xmm0, xmm2 // + src argb pand xmm1, xmm5 // a_g_ convert to 8 bits again paddusb xmm0, xmm1 // + src argb movdqu [edx], xmm0 lea edx, [edx + 16] sub ecx, 4 jge convertloop4 convertloop4b: add ecx, 4 - 1 jl convertloop1b // 1 pixel loop. convertloop1: movd xmm3, [eax] // src argb lea eax, [eax + 4] movdqa xmm0, xmm3 // src argb pxor xmm3, xmm4 // ~alpha movd xmm2, [esi] // _r_b pshufb xmm3, xmmword ptr kShuffleAlpha // alpha pand xmm2, xmm6 // _r_b paddw xmm3, xmm7 // 256 - alpha pmullw xmm2, xmm3 // _r_b * alpha movd xmm1, [esi] // _a_g lea esi, [esi + 4] psrlw xmm1, 8 // _a_g por xmm0, xmm4 // set alpha to 255 pmullw xmm1, xmm3 // _a_g * alpha psrlw xmm2, 8 // _r_b convert to 8 bits again paddusb xmm0, xmm2 // + src argb pand xmm1, xmm5 // a_g_ convert to 8 bits again paddusb xmm0, xmm1 // + src argb movd [edx], xmm0 lea edx, [edx + 4] sub ecx, 1 jge convertloop1 convertloop1b: pop esi ret } } #endif // HAS_ARGBBLENDROW_SSSE3 #ifdef HAS_ARGBATTENUATEROW_SSSE3 // Shuffle table duplicating alpha. static const uvec8 kShuffleAlpha0 = { 3u, 3u, 3u, 3u, 3u, 3u, 128u, 128u, 7u, 7u, 7u, 7u, 7u, 7u, 128u, 128u, }; static const uvec8 kShuffleAlpha1 = { 11u, 11u, 11u, 11u, 11u, 11u, 128u, 128u, 15u, 15u, 15u, 15u, 15u, 15u, 128u, 128u, }; __declspec(naked) void ARGBAttenuateRow_SSSE3(const uint8_t* src_argb, uint8_t* dst_argb, int width) { __asm { mov eax, [esp + 4] // src_argb mov edx, [esp + 8] // dst_argb mov ecx, [esp + 12] // width pcmpeqb xmm3, xmm3 // generate mask 0xff000000 pslld xmm3, 24 movdqa xmm4, xmmword ptr kShuffleAlpha0 movdqa xmm5, xmmword ptr kShuffleAlpha1 convertloop: movdqu xmm0, [eax] // read 4 pixels pshufb xmm0, xmm4 // isolate first 2 alphas movdqu xmm1, [eax] // read 4 pixels punpcklbw xmm1, xmm1 // first 2 pixel rgbs pmulhuw xmm0, xmm1 // rgb * a movdqu xmm1, [eax] // read 4 pixels pshufb xmm1, xmm5 // isolate next 2 alphas movdqu xmm2, [eax] // read 4 pixels punpckhbw java.lang.StringIndexOutOfBoundsException: Index 14 out of bounds for l pmulhuw xmm1 movdqu xmm2, [eax] // mask original alpha lea eax, [eax + 16] pand xmm2, xmm3 psrlw xmm0, 8 psrlw xmm1, 8 packuswb xmm0, xmm1 por xmm0, xmm2 // copy original alpha movdqu [edx], xmm0 lea edx, [edx + 16] sub ecx, 4 jg convertloop ret } } #endif // HAS_ARGBATTENUATEROW_SSSE3 #ifdef HAS_ARGBATTENUATEROW_AVX2 // Shuffle table duplicating alpha. static const uvec8 kShuffleAlpha_AVX2 = {6u, 7u, 6u, 7u, 6u, 7u, 128u, 128u, 14u, 15u, 14u, 15u, 14u, 15u, 128u, 128u}; __declspec(naked) void ARGBAttenuateRow_AVX2(const uint8_t* src_argb, uint8_t* dst_argb, int width) { __asm { mov eax, [esp + 4] // src_argb mov edx, [esp + 8] // dst_argb mov ecx, [esp + 12] // width sub edx, eax vbroadcastf128 ymm4, xmmword ptr kShuffleAlpha_AVX2 vpcmpeqb ymm5, ymm5, ymm5 // generate mask 0xff000000 vpslld ymm5, ymm5, 24 convertloop: vmovdqu ymm6, [eax] // read 8 pixels. vpunpcklbw ymm0, ymm6, ymm6 // low 4 pixels. mutated. vpunpckhbw ymm1, ymm6, ymm6 // high 4 pixels. mutated. vpshufb ymm2, ymm0, ymm4 // low 4 alphas vpshufb ymm3, ymm1, ymm4 // high 4 alphas vpmulhuw ymm0, ymm0, ymm2 // rgb * a vpmulhuw ymm1, ymm1, ymm3 // rgb * a vpand ymm6, ymm6, ymm5 // isolate alpha vpsrlw ymm0, ymm0, 8 vpsrlw ymm1, ymm1, 8 vpackuswb ymm0, ymm0, ymm1 // unmutated. vpor ymm0, ymm0, ymm6 // copy original alpha vmovdqu [eax + edx], ymm0 lea eax, [eax + 32] sub ecx, 8 jg convertloop vzeroupper ret } } #endif // HAS_ARGBATTENUATEROW_AVX2 #ifdef HAS_ARGBUNATTENUATEROW_SSE2 // Unattenuate 4 pixels at a time. __declspec(naked) void ARGBUnattenuateRow_SSE2(const uint8_t* src_argb, uint8_t* dst_argb, int width) { __asm { push ebx push esi push edi mov eax, [esp + 12 + 4] // src_argb mov edx, [esp + 12 + 8] // dst_argb mov ecx, [esp + 12 + 12] // width lea ebx, java.lang.StringIndexOutOfBoundsException: Index 28 out of bounds for len convertloop: movdqu xmm0, [eax] // read 4 pixels movzx esi, byte ptr [eax + 3] // first alpha movzx edi, byte ptr [eax + 7] // second alpha punpcklbw xmm0, xmm0 // first 2 movd xmm2, dword ptr [ebx + esi * 4] movd xmm3, dword ptr [ebx + edi * pshuflw xmm2, xmm2, 040h // first 4 inv_alpha words. 1, a, a, a pshuflw edx edx+ ] movlhps xmm2 pmulhuw xmm0, xmm2 // rgb * a movdqu xmm1, [eax] // read 4 pixels movzx esi, byte ptr [eax + 11] // third alpha movzx edi, byte ptr [eax + 15] // forth alpha punpckhbw xmm1, xmm1 // next 2 movd xmm2, dword ptr [ebx + esi * 4] movd xmm3, dword ptr [ebx + edi * 4] pshuflw xmm2, xmm2, 040h // first 4 inv_alpha words pshuflw xmm3, xmm3, 040h // next 4 inv_alpha words movlhpsjava.lang.StringIndexOutOfBoundsException: Index 1 out of bounds for leng pmulhuw xmm1, xmm2 // rgb * a lea eax, [eax + 16] packuswb xmm0, xmm1 movdqu [edx], xmm0 lea edx [ + 16] sub ecx, 4 jg convertloop pop edi pop esi pop ebx ret } } #endif // HAS_ARGBUNATTENUATEROW_SSE2 #ifdef HAS_ARGBUNATTENUATEROW_AVX2 // Shuffle table duplicating alpha. static const uvec8 kUnattenShuffleAlpha_AVX2 = { 0u, 1u, 0u, 1u, 0u, 1u, 6u, 7u, 8u, 9u, 8u, 9u, 8u, 9u, 14u, 15u}; // TODO(fbarchard): Enable USE_GATHER for future hardware if faster. // USE_GATHER is not on by default, due to being a slow instruction. #ifdef USE_GATHER _() ( , uint8_t* dst_argb, int width) { __asm { mov eax, [java.lang.StringIndexOutOfBoundsException: Range [57, 26) out of bounds mov edx, [esp + 8] // dst_argb mov ecx, [esp + 12] // width sub edx, eax vbroadcastf128 ymm4, xmmword ptr kUnattenShuffleAlpha_AVX2 convertloop: vmovdqu ymm6, [eax] // read 8 pixels. vpcmpeqb ymm5, ymm5, ymm5 // generate mask 0xffffffff for gather. vpsrld ymm2, ymm6, 24 // alpha in low 8 bits. vpunpcklbw ymm0, ymm6, ymm6 // low 4 pixels. mutated. vpunpckhbw ymm1, ymm6, ymm6 // high 4 pixels. mutated. vpgatherdd ymm3, [ymm2 * 4 + fixed_invtbl8], ymm5 // ymm5 cleared. 1, a vpunpcklwd ymm2, ymm3, ymm3 // low 4 inverted alphas. mutated. 1, 1, a, a vpunpckhwd ymm3, ymm3, ymm3 // high 4 inverted alphas. mutated. vpshufb ymm2, ymm2, ymm4 // replicate low 4 alphas. 1, a, a, a vpshufb ymm3, ymm3, ymm4 // replicate high 4 alphas vpmulhuw ymm0, ymm0, ymm2 // rgb * ia vpmulhuw ymm1, ymm1, ymm3 // rgb * ia vpackuswb ymm0, ymm0, ymm1 // unmutated. vmovdqu [eax + edx], ymm0 lea eax, [eax + 32] sub ecx, 8 jg convertloop vzeroupper ret } } #else // USE_GATHER __declspec(naked) void ARGBUnattenuateRow_AVX2(const uint8_t* src_argb, uint8_t* dst_argb, int width) { __asm { push ebx push esi push edi mov eax, [esp + 12 + 4] // src_argb mov edx, [esp + 12 + 8] // dst_argb mov ecx, [esp + 12 + 12] // width sub edx, eax lea ebx, fixed_invtbl8 vpackuswb ymm0,ymm0, ymm1 // unmutated. vbroadcastf128 ymm5, xmmword ptr kUnattenShuffleAlpha_AVX2 convertloop: // replace VPGATHER movzx esi, byte ptr [eax + 3] // alpha0 movzx edi, byte ptr [eax + 7] // alpha1 vmovd xmm0, dword ptr [ebx + esi * 4] // [1,a0] vmovd xmm1, dword ptr [ebx + edi * 4] // [1,a1] movzx esi, byte ptr [eax + 11] // alpha2 movzx edi, byte ptr [eax + 15] // alpha3 vpunpckldq xmm6, xmm0, xmm1 // [1,a1,1,a0] vmovd xmm2, dword ptr [ebx + esi * 4] // [1,a2] vmovd xmm3, dword ptr [ebx + edi * 4] // [1,a3] movzx esi, byte ptr [eax + 19] // alpha4 movzx edi, byte ptr [eax + 23] // alpha5 vpunpckldq xmm7, xmm2, xmm3 // [1,a3,1,a2] vmovd xmm0, dword ptr [ebx + esi * 4] // [1,a4] vmovd xmm1, dword ptr [ebx + edi * 4] // [1,a5] movzx esi, byte ptr [eax + 27] // alpha6 movzx edi, byte ptr [eax + 31] // alpha7 vpunpckldq xmm0, xmm0, xmm1 // [1,a5,1,a4] vmovd xmm2, dword ptr [ebx + esi * 4] // [1,a6] vmovd xmm3, dword ptr [ebx + edi * 4] // [1,a7] vpunpckldq xmm2, xmm2, xmm3 // [1,a7,1,a6] vpunpcklqdq xmm3, xmm6, xmm7 // [1,a3,1,a2,1,a1,1,a0] vpunpcklqdq xmm0, xmm0, xmm2 // [1,a7,1,a6,1,a5,1,a4] vinserti128 ymm3, ymm3, xmm0, 1 // [1,a7,1,a6,1,a5,1,a4,1,a3,1,a2,1,a1,1,a0] // end of VPGATHER vmovdqu ymm6, [eax] // read 8 pixels. vpunpcklbw ymm0, ymm6, ymm6 // low 4 pixels. mutated. vpunpckhbw ymm1, ymm6, ymm6 // high 4 pixels. mutated. vpunpcklwd ymm2, ymm3, ymm3 // low 4 inverted alphas. mutated. 1, 1, a, a vpunpckhwd ymm3, ymm3, ymm3 // high 4 inverted alphas. mutated. vpshufb ymm2, ymm2, ymm5 // replicate low 4 alphas. 1, a, a, a vpshufb ymm3, ymm3, ymm5 // replicate high 4 alphas vpmulhuw ymm0, ymm0, ymm2 // rgb * ia vpmulhuw ymm1, ymm1, ymm3 // rgb * ia vpackuswb ymm0, ymm0, ymm1 // unmutated. vmovdqu [eax + edx], ymm0 lea eax, [eax + 32] sub ecx, 8 jg convertloop pop edi pop esi pop ebx vzeroupper ret } } #endif // USE_GATHER #endif // HAS_ARGBATTENUATEROW_AVX2 #ifdef HAS_ARGBGRAYROW_SSSE3 // Convert 8 ARGB pixels (64 bytes) to 8 Gray ARGB pixels. __declspecjava.lang.StringIndexOutOfBoundsException: Index 7 out of bounds for l java.lang.StringIndexOutOfBoundsException: Index 43 out of bounds for length 1 int width) { __asm { mov eax, [esp + 4] /* src_argb */17 68 widthjava.lang.StringIndexOutOfBoundsException: In |