懒人包2025EX（版本问题回复在此 尽量不开新帖）

[hooke007]

主要问题追踪

上游问题追踪

• （懒人包发布后）上游变动了一项色彩管理方面的默认值 mpv-player/mpv@04055e5
• 截屏在某种条件下存在变暗的问题 懒人包2025EX（版本问题回复在此 尽量不开新帖） #628 (comment)

下版本开发记录

k7fsunc 的 UAI 类模块移除 clamp 参数；
一键下载assets的所有包 #628 (comment) ；
k7fsunc 扩展MacOS支持 #628 (comment)
后备字体切换 #628 (comment)
k7fsunc 支持DRBA补帧 #628 (comment)
k7fsunc 更新首个正式版 #628 (comment)
esrgan #628 (comment)

其它

• 张小龙是傻逼 Error launching CrashSender.exe M2Team/NanaZip#497
  如果你在用微信输入法，临时切换到微软输入法即可

---

This discussion was created from the release mpv-lazy-2025EX.

上期 #610

[hui1971]

鼠标滚轮快进后，预览图显示不正确，然后播放器闪退,win11,25h2 i13500cpu,intel A380 GPU

[hooke007]

没有日志，不进行虚空诊断，如果只有这点有效信息，可以自行更换mpv程序测试，缩略图脚本上一次变动在4月前，不太可能是它自身的问题。

[CyaAllen]

这个快进闪退问题，我替换了上一个版本的mpv.exe就正常了，你可以试试；如果还是不行，就把thumbfast.lua也替换一下。

[hui1971]

是的，换了上8月份以前版本的mpv.exe就基本正常，如果Thumbfast.lua也替换的话完全正常了。

[Joctor-Z]

大佬,请问右侧的音量界面的大小在哪里设置?找了好久没找到

[hooke007]

mpv_PlayKit/portable_config/script-opts.conf

Line 482 in cd1ce43

script-opts-append = uosc-volume_size=40

怎么会找不到

[Joctor-Z]

谢谢大佬,另外还有一个请问一下这个版本跟上个版本的注册逻辑是不是有点不一样啊?上个版本注册完设置mpv为默认播放器以后,我可以单独给mp4/mkv/avi等不同格式的设置不同图标(icon),这个新版本的我搞了一晚上,设置完mp4的所有视频格式的都会跟着变,好像所有的媒体格式都变成了mpv,我在注册表里都没法一个一个改

[hooke007]

官方注册的逻辑目前是这样的

[Joctor-Z]

全部显示成mpv的图标不会很丑吗?大佬是怎么解决的?

[hooke007]

我不在意这个东西，你可以自行选择注册方式

[lingyust]

字幕位置怎么往下调？

[hooke007]

https://mpv.io/manual/master/#options-sub-pos

[lingyust]

设置了，文档里面说调整范围是0-150。调整100，就是软件默认的字幕位置，但是100以上到150，不管怎么调整，都是一个位置

[hooke007]

人家warning部分写的很清楚，你的疑问是什么

[hooke007]

你的字幕到底是什么类型，我测试常规文本没有这个限制

[hooke007]

这个版本开始追加一个精简版（移除了vapoursynth滤镜相关组件，仍然可与vsNV包无缝组合）

[HedioKojima]

使用仓库内NVScaler_RT.glsl 时报错很多 vo/gpu-next libplacebo

[HedioKojima]

[vo/gpu-next/libplacebo] shaderc output:
[vo/gpu-next/libplacebo] input:112: error: '=' : global const initializers must be constant ' const highp float'
[vo/gpu-next/libplacebo] input:113: error: '=' : global const initializers must be constant ' const highp float'
[vo/gpu-next/libplacebo] input:114: error: '=' : global const initializers must be constant ' const highp float'
[vo/gpu-next/libplacebo] input:115: error: '=' : global const initializers must be constant ' const highp float'
[vo/gpu-next/libplacebo] input:122: error: '=' : global const initializers must be constant ' const highp float'
[vo/gpu-next/libplacebo] input:123: error: '=' : global const initializers must be constant ' const highp float'
[vo/gpu-next/libplacebo] input:124: error: '=' : global const initializers must be constant ' const highp float'
[vo/gpu-next/libplacebo] input:125: error: '=' : global const initializers must be constant ' const highp float'
[vo/gpu-next/libplacebo] input:128: error: '=' : global const initializers must be constant ' const highp float'
[vo/gpu-next/libplacebo] input:129: error: '=' : global const initializers must be constant ' const highp float'
[vo/gpu-next/libplacebo]
[vo/gpu-next/libplacebo] shaderc compile status 'error' (10 errors, 0 warnings)
[vo/gpu-next/libplacebo] vk_compile_glsl(gpu, tmp, GLSL_SHADER_COMPUTE, params->glsl_shader, &comp): VK_ERROR_INITIALIZATION_FAILED (../../../../../src_packages/libplacebo/src/vulkan/gpu_pass.c:464)
[vo/gpu-next/libplacebo] compute shader source:
[vo/gpu-next/libplacebo] [  1] #version 450
[vo/gpu-next/libplacebo] [  2] #extension GL_ARB_compute_shader : enable
[vo/gpu-next/libplacebo] [  3] #extension GL_KHR_shader_subgroup_basic : enable
[vo/gpu-next/libplacebo] [  4] #extension GL_KHR_shader_subgroup_vote : enable
[vo/gpu-next/libplacebo] [  5] #extension GL_KHR_shader_subgroup_arithmetic : enable
[vo/gpu-next/libplacebo] [  6] #extension GL_KHR_shader_subgroup_ballot : enable
[vo/gpu-next/libplacebo] [  7] #extension GL_KHR_shader_subgroup_shuffle : enable
[vo/gpu-next/libplacebo] [  8] #extension GL_KHR_shader_subgroup_clustered : enable
[vo/gpu-next/libplacebo] [  9] #extension GL_KHR_shader_subgroup_quad : enable
[vo/gpu-next/libplacebo] [ 10] #extension GL_ARB_shader_image_load_store : enable
[vo/gpu-next/libplacebo] [ 11] #extension GL_ARB_texture_gather : enable
[vo/gpu-next/libplacebo] [ 12] layout(std430, push_constant) uniform PushC {
[vo/gpu-next/libplacebo] [ 13]     layout(offset=0) vec2 _4;
[vo/gpu-next/libplacebo] [ 14]     layout(offset=8) vec2 _5;
[vo/gpu-next/libplacebo] [ 15]     layout(offset=16) int _9;
[vo/gpu-next/libplacebo] [ 16]     layout(offset=20) float _a;
[vo/gpu-next/libplacebo] [ 17]     layout(offset=24) vec2 _b;
[vo/gpu-next/libplacebo] [ 18]     layout(offset=32) vec2 _c;
[vo/gpu-next/libplacebo] [ 19]     layout(offset=40) vec2 _d;
[vo/gpu-next/libplacebo] [ 20]     layout(offset=48) float _e;
[vo/gpu-next/libplacebo] [ 21]     layout(offset=56) vec2 _10;
[vo/gpu-next/libplacebo] [ 22]     layout(offset=64) vec2 _11;
[vo/gpu-next/libplacebo] [ 23]     layout(offset=72) vec2 _12;
[vo/gpu-next/libplacebo] [ 24]     layout(offset=80) vec2 _13;
[vo/gpu-next/libplacebo] [ 25]     layout(offset=88) vec2 _14;
[vo/gpu-next/libplacebo] [ 26] };
[vo/gpu-next/libplacebo] [ 27] layout(constant_id=0) const float _6 = 1;
[vo/gpu-next/libplacebo] [ 28] layout(constant_id=1) const float _8002 = 1;
[vo/gpu-next/libplacebo] [ 29] layout(constant_id=2) const float _8003 = 1;
[vo/gpu-next/libplacebo] [ 30] layout(constant_id=3) const float _4002 = 1;
[vo/gpu-next/libplacebo] [ 31] layout(constant_id=4) const float _4003 = 1;
[vo/gpu-next/libplacebo] [ 32] layout(binding=0) uniform  sampler2D _2;
[vo/gpu-next/libplacebo] [ 33] layout(binding=1) uniform  sampler2D _7;
[vo/gpu-next/libplacebo] [ 34] layout(binding=2) uniform  sampler2D _8;
[vo/gpu-next/libplacebo] [ 35] layout(binding=3, rgba16f) writeonly restrict uniform image2D _f;
[vo/gpu-next/libplacebo] [ 36] layout (local_size_x = 256, local_size_y = 1) in;
[vo/gpu-next/libplacebo] [ 37] #define out_image _f
[vo/gpu-next/libplacebo] [ 38] #define frag_pos(id) (vec2(id) + vec2(0.5))
[vo/gpu-next/libplacebo] [ 39] #define frag_map(id) (_10 * frag_pos(id))
[vo/gpu-next/libplacebo] [ 40] #define gl_FragCoord vec4(frag_pos(gl_GlobalInvocationID), 0.0, 1.0)
[vo/gpu-next/libplacebo] [ 41] #define _3_map(id) (mix(mix(_11, _12, frag_map(id).x),      mix(_13, _14, frag_map(id).x), frag_map(id).y))
[vo/gpu-next/libplacebo] [ 42] #define _3 (_3_map(gl_GlobalInvocationID))
[vo/gpu-next/libplacebo] [ 43]
[vo/gpu-next/libplacebo] [ 44] #define LUMA_raw _2
[vo/gpu-next/libplacebo] [ 45] #define LUMA_pos _3
[vo/gpu-next/libplacebo] [ 46] #define LUMA_map _3_map
[vo/gpu-next/libplacebo] [ 47] #define LUMA_size vec2(textureSize(_2, 0))
[vo/gpu-next/libplacebo] [ 48] #define LUMA_pt _4
[vo/gpu-next/libplacebo] [ 49] #define LUMA_off _5
[vo/gpu-next/libplacebo] [ 50] #define LUMA_mul _6
[vo/gpu-next/libplacebo] [ 51] #define LUMA_rot mat2(1.0, 0.0, 0.0, 1.0)
[vo/gpu-next/libplacebo] [ 52] #define LUMA_tex(pos) (_6 * vec4(textureLod(_2, pos, 0.0)))
[vo/gpu-next/libplacebo] [ 53] #define LUMA_texOff(off) (LUMA_tex(_3 + _4 * vec2(off)))
[vo/gpu-next/libplacebo] [ 54] #define LUMA_gather(pos, c) (_6 * vec4(textureGather(_2, pos, c)))
[vo/gpu-next/libplacebo] [ 55] #define HOOKED_raw LUMA_raw
[vo/gpu-next/libplacebo] [ 56] #define HOOKED_pos LUMA_pos
[vo/gpu-next/libplacebo] [ 57] #define HOOKED_size LUMA_size
[vo/gpu-next/libplacebo] [ 58] #define HOOKED_rot LUMA_rot
[vo/gpu-next/libplacebo] [ 59] #define HOOKED_off LUMA_off
[vo/gpu-next/libplacebo] [ 60] #define HOOKED_pt LUMA_pt
[vo/gpu-next/libplacebo] [ 61] #define HOOKED_map LUMA_map
[vo/gpu-next/libplacebo] [ 62] #define HOOKED_mul LUMA_mul
[vo/gpu-next/libplacebo] [ 63] #define HOOKED_tex LUMA_tex
[vo/gpu-next/libplacebo] [ 64] #define HOOKED_texOff LUMA_texOff
[vo/gpu-next/libplacebo] [ 65] #define HOOKED_gather LUMA_gather
[vo/gpu-next/libplacebo] [ 66] #define coef_scaler _7
[vo/gpu-next/libplacebo] [ 67] #define coef_scaler_tex(pos) (textureLod(_7, pos, 0.0))
[vo/gpu-next/libplacebo] [ 68] #define coef_usm _8
[vo/gpu-next/libplacebo] [ 69] #define coef_usm_tex(pos) (textureLod(_8, pos, 0.0))
[vo/gpu-next/libplacebo] [ 70] #define frame _9
[vo/gpu-next/libplacebo] [ 71] #define random _a
[vo/gpu-next/libplacebo] [ 72] #define input_size _b
[vo/gpu-next/libplacebo] [ 73] #define target_size _c
[vo/gpu-next/libplacebo] [ 74] #define tex_offset _d
[vo/gpu-next/libplacebo] [ 75] #define SHARP _e
[vo/gpu-next/libplacebo] [ 76] #define PQ 0
[vo/gpu-next/libplacebo] [ 77] vec4 _8001(vec4 color) {
[vo/gpu-next/libplacebo] [ 78] // pl_shader_linearize
[vo/gpu-next/libplacebo] [ 79] color.rgb = max(color.rgb, 0.0);
[vo/gpu-next/libplacebo] [ 80] color.rgb = _8002 * pow(color.rgb + vec3(_8003), vec3(2.4));
[vo/gpu-next/libplacebo] [ 81] return color;
[vo/gpu-next/libplacebo] [ 82] }
[vo/gpu-next/libplacebo] [ 83]
[vo/gpu-next/libplacebo] [ 84] #define linearize _8001
[vo/gpu-next/libplacebo] [ 85] vec4 _4001(vec4 color) {
[vo/gpu-next/libplacebo] [ 86] // pl_shader_delinearize
[vo/gpu-next/libplacebo] [ 87] color.rgb = max(color.rgb, 0.0);
[vo/gpu-next/libplacebo] [ 88] color.rgb = pow(_4002 * color.rgb, vec3(1.0/2.4)) - vec3(_4003);
[vo/gpu-next/libplacebo] [ 89] return color;
[vo/gpu-next/libplacebo] [ 90] }
[vo/gpu-next/libplacebo] [ 91]
[vo/gpu-next/libplacebo] [ 92] #define delinearize _4001
[vo/gpu-next/libplacebo] [ 93]
[vo/gpu-next/libplacebo] [ 94] // User variables
[vo/gpu-next/libplacebo] [ 95] #define SHARPNESS      SHARP   // Amount of sharpening. 0.0 to 1.0.
[vo/gpu-next/libplacebo] [ 96] #define NIS_HDR_MODE   PQ      // Must be set to 1 for content with PQ colorspace. 0 or 1.
[vo/gpu-next/libplacebo] [ 97]
[vo/gpu-next/libplacebo] [ 98] // Constant variables
[vo/gpu-next/libplacebo] [ 99] #define NIS_THREAD_GROUP_SIZE 256 // May be set to 128 for better performance on NVIDIA hardware, otherwise set to 256. Don't forget to modify the COMPUTE directive accordingly as well (e.g., COMPUTE 32 24 128 1).
[vo/gpu-next/libplacebo] [100] #define NIS_BLOCK_WIDTH 32
[vo/gpu-next/libplacebo] [101] #define NIS_BLOCK_HEIGHT 24
[vo/gpu-next/libplacebo] [102] #define kPhaseCount  64
[vo/gpu-next/libplacebo] [103] #define kFilterSize  6
[vo/gpu-next/libplacebo] [104] #define kSupportSize 6
[vo/gpu-next/libplacebo] [105] #define kPadSize     kSupportSize
[vo/gpu-next/libplacebo] [106] #define NIS_SCALE_INT 1
[vo/gpu-next/libplacebo] [107] #define NIS_SCALE_FLOAT 1.0f
[vo/gpu-next/libplacebo] [108] #define kTilePitch (NIS_BLOCK_WIDTH + kPadSize)
[vo/gpu-next/libplacebo] [109] #define kTileSize (kTilePitch * (NIS_BLOCK_HEIGHT + kPadSize))
[vo/gpu-next/libplacebo] [110] #define kEdgeMapPitch (NIS_BLOCK_WIDTH + 2)
[vo/gpu-next/libplacebo] [111] #define kEdgeMapSize (kEdgeMapPitch * (NIS_BLOCK_HEIGHT + 2))
[vo/gpu-next/libplacebo] [112] const float sharpen_slider = clamp(SHARPNESS, 0.0f, 1.0f) - 0.5f;
[vo/gpu-next/libplacebo] [113] const float MaxScale = (sharpen_slider >= 0.0f) ? 1.25f : 1.75f;
[vo/gpu-next/libplacebo] [114] const float MinScale = (sharpen_slider >= 0.0f) ? 1.25f : 1.0f;
[vo/gpu-next/libplacebo] [115] const float LimitScale = (sharpen_slider >= 0.0f) ? 1.25f : 1.0f;
[vo/gpu-next/libplacebo] [116] const float kDetectRatio = 2 * 1127.f / 1024.f;
[vo/gpu-next/libplacebo] [117] const float kDetectThres = (bool(NIS_HDR_MODE) ? 32.0f : 64.0f) / 1024.0f;
[vo/gpu-next/libplacebo] [118] const float kMinContrastRatio = bool(NIS_HDR_MODE) ? 1.5f : 2.0f;
[vo/gpu-next/libplacebo] [119] const float kMaxContrastRatio = bool(NIS_HDR_MODE) ? 5.0f : 10.0f;
[vo/gpu-next/libplacebo] [120] const float kSharpStartY = bool(NIS_HDR_MODE) ? 0.35f : 0.45f;
[vo/gpu-next/libplacebo] [121] const float kSharpEndY = bool(NIS_HDR_MODE) ? 0.55f : 0.9f;
[vo/gpu-next/libplacebo] [122] const float kSharpStrengthMin = max(0.0f, 0.4f + sharpen_slider * MinScale * (bool(NIS_HDR_MODE) ? 1.1f : 1.2));
[vo/gpu-next/libplacebo] [123] const float kSharpStrengthMax = ((bool(NIS_HDR_MODE) ? 2.2f : 1.6f) + sharpen_slider * MaxScale * 1.8f);
[vo/gpu-next/libplacebo] [124] const float kSharpLimitMin = max((bool(NIS_HDR_MODE) ? 0.06f :0.1f), (bool(NIS_HDR_MODE) ? 0.1f : 0.14f) + sharpen_slider * LimitScale * (bool(NIS_HDR_MODE) ? 0.28f : 0.32f));
[vo/gpu-next/libplacebo] [125] const float kSharpLimitMax = ((bool(NIS_HDR_MODE) ? 0.6f : 0.5f) + sharpen_slider * LimitScale * 0.6f);
[vo/gpu-next/libplacebo] [126] const float kRatioNorm = 1.0f / (kMaxContrastRatio - kMinContrastRatio);
[vo/gpu-next/libplacebo] [127] const float kSharpScaleY = 1.0f / (kSharpEndY - kSharpStartY);
[vo/gpu-next/libplacebo] [128] const float kSharpStrengthScale = kSharpStrengthMax - kSharpStrengthMin;
[vo/gpu-next/libplacebo] [129] const float kSharpLimitScale = kSharpLimitMax - kSharpLimitMin;
[vo/gpu-next/libplacebo] [130] const float kContrastBoost = 1.0f;
[vo/gpu-next/libplacebo] [131] const float kEps = 1.0f;
[vo/gpu-next/libplacebo] [132] #define kScaleX (HOOKED_size.x / target_size.x)
[vo/gpu-next/libplacebo] [133] #define kScaleY (HOOKED_size.y / target_size.y)
[vo/gpu-next/libplacebo] [134] #define kSrcNormX HOOKED_pt.x
[vo/gpu-next/libplacebo] [135] #define kSrcNormY HOOKED_pt.y
[vo/gpu-next/libplacebo] [136]
[vo/gpu-next/libplacebo] [137] // HLSL to GLSL macros
[vo/gpu-next/libplacebo] [138] #define saturate(x) clamp(x, 0, 1)
[vo/gpu-next/libplacebo] [139] #define lerp(a, b, x) mix(a, b, x)
[vo/gpu-next/libplacebo] [140]
[vo/gpu-next/libplacebo] [141] // CS Shared variables
[vo/gpu-next/libplacebo] [142] shared float shPixelsY[kTileSize];
[vo/gpu-next/libplacebo] [143] shared float shCoefScaler[kPhaseCount][kFilterSize];
[vo/gpu-next/libplacebo] [144] shared float shCoefUSM[kPhaseCount][kFilterSize];
[vo/gpu-next/libplacebo] [145] shared vec4 shEdgeMap[kEdgeMapSize];
[vo/gpu-next/libplacebo] [146]
[vo/gpu-next/libplacebo] [147] // Shader code
[vo/gpu-next/libplacebo] [148]
[vo/gpu-next/libplacebo] [149] vec4 GetEdgeMap(float p[4][4], int i, int j) {
[vo/gpu-next/libplacebo] [150]  const float g_0 = abs(p[0 + i][0 + j] + p[0 + i][1 + j] + p[0 + i][2 + j] - p[2 + i][0 + j] - p[2 + i][1 + j] - p[2 + i][2 + j]);
[vo/gpu-next/libplacebo] [151]  const float g_45 = abs(p[1 + i][0 + j] + p[0 + i][0 + j] + p[0 + i][1 + j] - p[2 + i][1 + j] - p[2 + i][2 + j] - p[1 + i][2 + j]);
[vo/gpu-next/libplacebo] [152]  const float g_90 = abs(p[0 + i][0 + j] + p[1 + i][0 + j] + p[2 + i][0 + j] - p[0 + i][2 + j] - p[1 + i][2 + j] - p[2 + i][2 + j]);
[vo/gpu-next/libplacebo] [153]  const float g_135 = abs(p[1 + i][0 + j] + p[2 + i][0 + j] + p[2 + i][1 + j] - p[0 + i][1 + j] - p[0 + i][2 + j] - p[1 + i][2 + j]);
[vo/gpu-next/libplacebo] [154]
[vo/gpu-next/libplacebo] [155]  const float g_0_90_max = max(g_0, g_90);
[vo/gpu-next/libplacebo] [156]  const float g_0_90_min = min(g_0, g_90);
[vo/gpu-next/libplacebo] [157]  const float g_45_135_max = max(g_45, g_135);
[vo/gpu-next/libplacebo] [158]  const float g_45_135_min = min(g_45, g_135);
[vo/gpu-next/libplacebo] [159]
[vo/gpu-next/libplacebo] [160]  float e_0_90 = 0;
[vo/gpu-next/libplacebo] [161]  float e_45_135 = 0;
[vo/gpu-next/libplacebo] [162]
[vo/gpu-next/libplacebo] [163]  if (g_0_90_max + g_45_135_max == 0)
[vo/gpu-next/libplacebo] [164]  {
[vo/gpu-next/libplacebo] [165]          return vec4(0, 0, 0, 0);
[vo/gpu-next/libplacebo] [166]  }
[vo/gpu-next/libplacebo] [167]
[vo/gpu-next/libplacebo] [168]  e_0_90 = min(g_0_90_max / (g_0_90_max + g_45_135_max), 1.0f);
[vo/gpu-next/libplacebo] [169]  e_45_135 = 1.0f - e_0_90;
[vo/gpu-next/libplacebo] [170]
[vo/gpu-next/libplacebo] [171]  bool c_0_90 = (g_0_90_max > (g_0_90_min * kDetectRatio)) && (g_0_90_max > kDetectThres) && (g_0_90_max > g_45_135_min);
[vo/gpu-next/libplacebo] [172]  bool c_45_135 = (g_45_135_max > (g_45_135_min * kDetectRatio)) && (g_45_135_max > kDetectThres) && (g_45_135_max > g_0_90_min);
[vo/gpu-next/libplacebo] [173]  bool c_g_0_90 = g_0_90_max == g_0;
[vo/gpu-next/libplacebo] [174]  bool c_g_45_135 = g_45_135_max == g_45;
[vo/gpu-next/libplacebo] [175]
[vo/gpu-next/libplacebo] [176]  float f_e_0_90 = (c_0_90 && c_45_135) ? e_0_90 : 1.0f;
[vo/gpu-next/libplacebo] [177]  float f_e_45_135 = (c_0_90 && c_45_135) ? e_45_135 : 1.0f;
[vo/gpu-next/libplacebo] [178]
[vo/gpu-next/libplacebo] [179]  float weight_0 = (c_0_90 && c_g_0_90) ? f_e_0_90 : 0.0f;
[vo/gpu-next/libplacebo] [180]  float weight_90 = (c_0_90 && !c_g_0_90) ? f_e_0_90 : 0.0f;
[vo/gpu-next/libplacebo] [181]  float weight_45 = (c_45_135 && c_g_45_135) ? f_e_45_135 : 0.0f;
[vo/gpu-next/libplacebo] [182]  float weight_135 = (c_45_135 && !c_g_45_135) ? f_e_45_135 : 0.0f;
[vo/gpu-next/libplacebo] [183]
[vo/gpu-next/libplacebo] [184]  return vec4(weight_0, weight_90, weight_45, weight_135);
[vo/gpu-next/libplacebo] [185] }
[vo/gpu-next/libplacebo] [186]
[vo/gpu-next/libplacebo] [187] void LoadFilterBanksSh(int i0, int di) {
[vo/gpu-next/libplacebo] [188]  // Load up filter banks to shared memory
[vo/gpu-next/libplacebo] [189]  // The work is spread over (kPhaseCount * 2) threads
[vo/gpu-next/libplacebo] [190]  for (int i = i0; i < kPhaseCount * 2; i += di)
[vo/gpu-next/libplacebo] [191]  {
[vo/gpu-next/libplacebo] [192]          int phase = i >> 1;
[vo/gpu-next/libplacebo] [193]          int vIdx = i & 1;
[vo/gpu-next/libplacebo] [194]
[vo/gpu-next/libplacebo] [195]          vec4 v = vec4(texelFetch(coef_scaler, ivec2(vIdx, phase), 0));
[vo/gpu-next/libplacebo] [196]          int filterOffset = vIdx * 4;
[vo/gpu-next/libplacebo] [197]          shCoefScaler[phase][filterOffset + 0] = v.x;
[vo/gpu-next/libplacebo] [198]          shCoefScaler[phase][filterOffset + 1] = v.y;
[vo/gpu-next/libplacebo] [199]          if (vIdx == 0)
[vo/gpu-next/libplacebo] [200]          {
[vo/gpu-next/libplacebo] [201]                  shCoefScaler[phase][2] = v.z;
[vo/gpu-next/libplacebo] [202]                  shCoefScaler[phase][3] = v.w;
[vo/gpu-next/libplacebo] [203]          }
[vo/gpu-next/libplacebo] [204]
[vo/gpu-next/libplacebo] [205]          v = vec4(texelFetch(coef_usm, ivec2(vIdx, phase), 0));
[vo/gpu-next/libplacebo] [206]          shCoefUSM[phase][filterOffset + 0] = v.x;
[vo/gpu-next/libplacebo] [207]          shCoefUSM[phase][filterOffset + 1] = v.y;
[vo/gpu-next/libplacebo] [208]          if (vIdx == 0)
[vo/gpu-next/libplacebo] [209]          {
[vo/gpu-next/libplacebo] [210]                  shCoefUSM[phase][2] = v.z;
[vo/gpu-next/libplacebo] [211]                  shCoefUSM[phase][3] = v.w;
[vo/gpu-next/libplacebo] [212]          }
[vo/gpu-next/libplacebo] [213]  }
[vo/gpu-next/libplacebo] [214] }
[vo/gpu-next/libplacebo] [215]
[vo/gpu-next/libplacebo] [216] float CalcLTI(float p0, float p1, float p2, float p3, float p4, float p5, int phase_index)
[vo/gpu-next/libplacebo] [217] {
[vo/gpu-next/libplacebo] [218]  const bool selector = (phase_index <= kPhaseCount / 2);
[vo/gpu-next/libplacebo] [219]  float sel = selector ? p0 : p3;
[vo/gpu-next/libplacebo] [220]  const float a_min = min(min(p1, p2), sel);
[vo/gpu-next/libplacebo] [221]  const float a_max = max(max(p1, p2), sel);
[vo/gpu-next/libplacebo] [222]  sel = selector ? p2 : p5;
[vo/gpu-next/libplacebo] [223]  const float b_min = min(min(p3, p4), sel);
[vo/gpu-next/libplacebo] [224]  const float b_max = max(max(p3, p4), sel);
[vo/gpu-next/libplacebo] [225]
[vo/gpu-next/libplacebo] [226]  const float a_cont = a_max - a_min;
[vo/gpu-next/libplacebo] [227]  const float b_cont = b_max - b_min;
[vo/gpu-next/libplacebo] [228]
[vo/gpu-next/libplacebo] [229]  const float cont_ratio = max(a_cont, b_cont) / (min(a_cont, b_cont) + kEps);
[vo/gpu-next/libplacebo] [230]  return (1.0f - saturate((cont_ratio - kMinContrastRatio) * kRatioNorm)) * kContrastBoost;
[vo/gpu-next/libplacebo] [231] }
[vo/gpu-next/libplacebo] [232]
[vo/gpu-next/libplacebo] [233] vec4 GetInterpEdgeMap(const vec4 edge[2][2], float phase_frac_x, float phase_frac_y)
[vo/gpu-next/libplacebo] [234] {
[vo/gpu-next/libplacebo] [235]  vec4 h0 = lerp(edge[0][0], edge[0][1], phase_frac_x);
[vo/gpu-next/libplacebo] [236]  vec4 h1 = lerp(edge[1][0], edge[1][1], phase_frac_x);
[vo/gpu-next/libplacebo] [237]  return lerp(h0, h1, phase_frac_y);
[vo/gpu-next/libplacebo] [238] }
[vo/gpu-next/libplacebo] [239]
[vo/gpu-next/libplacebo] [240] float EvalPoly6(const float pxl[6], int phase_int)
[vo/gpu-next/libplacebo] [241] {
[vo/gpu-next/libplacebo] [242]  float y = 0.f;
[vo/gpu-next/libplacebo] [243]  {
[vo/gpu-next/libplacebo] [244]          for (int i = 0; i < 6; ++i)
[vo/gpu-next/libplacebo] [245]          {
[vo/gpu-next/libplacebo] [246]                  y += shCoefScaler[phase_int][i] * pxl[i];
[vo/gpu-next/libplacebo] [247]          }
[vo/gpu-next/libplacebo] [248]  }
[vo/gpu-next/libplacebo] [249]  float y_usm = 0.f;
[vo/gpu-next/libplacebo] [250]  {
[vo/gpu-next/libplacebo] [251]          for (int i = 0; i < 6; ++i)
[vo/gpu-next/libplacebo] [252]          {
[vo/gpu-next/libplacebo] [253]                  y_usm += shCoefUSM[phase_int][i] * pxl[i];
[vo/gpu-next/libplacebo] [254]          }
[vo/gpu-next/libplacebo] [255]  }
[vo/gpu-next/libplacebo] [256]
[vo/gpu-next/libplacebo] [257]  // let's compute a piece-wise ramp based on luma
[vo/gpu-next/libplacebo] [258]  const float y_scale = 1.0f - saturate((y * (1.0f / NIS_SCALE_FLOAT) - kSharpStartY) * kSharpScaleY);
[vo/gpu-next/libplacebo] [259]
[vo/gpu-next/libplacebo] [260]  // scale the ramp to sharpen as a function of luma
[vo/gpu-next/libplacebo] [261]  const float y_sharpness = y_scale * kSharpStrengthScale + kSharpStrengthMin;
[vo/gpu-next/libplacebo] [262]
[vo/gpu-next/libplacebo] [263]  y_usm *= y_sharpness;
[vo/gpu-next/libplacebo] [264]
[vo/gpu-next/libplacebo] [265]  // scale the ramp to limit USM as a function of luma
[vo/gpu-next/libplacebo] [266]  const float y_sharpness_limit = (y_scale * kSharpLimitScale + kSharpLimitMin) * y;
[vo/gpu-next/libplacebo] [267]
[vo/gpu-next/libplacebo] [268]  y_usm = min(y_sharpness_limit, max(-y_sharpness_limit, y_usm));
[vo/gpu-next/libplacebo] [269]  // reduce ringing
[vo/gpu-next/libplacebo] [270]  y_usm *= CalcLTI(pxl[0], pxl[1], pxl[2], pxl[3], pxl[4], pxl[5], phase_int);
[vo/gpu-next/libplacebo] [271]
[vo/gpu-next/libplacebo] [272]  return y + y_usm;
[vo/gpu-next/libplacebo] [273] }
[vo/gpu-next/libplacebo] [274]
[vo/gpu-next/libplacebo] [275] float FilterNormal(const float p[6][6], int phase_x_frac_int, int phase_y_frac_int)
[vo/gpu-next/libplacebo] [276] {
[vo/gpu-next/libplacebo] [277]  float h_acc = 0.0f;
[vo/gpu-next/libplacebo] [278]  for (int j = 0; j < 6; ++j)
[vo/gpu-next/libplacebo] [279]  {
[vo/gpu-next/libplacebo] [280]          float v_acc = 0.0f;
[vo/gpu-next/libplacebo] [281]          for (int i = 0; i < 6; ++i)
[vo/gpu-next/libplacebo] [282]          {
[vo/gpu-next/libplacebo] [283]                  v_acc += p[i][j] * shCoefScaler[phase_y_frac_int][i];
[vo/gpu-next/libplacebo] [284]          }
[vo/gpu-next/libplacebo] [285]          h_acc += v_acc * shCoefScaler[phase_x_frac_int][j];
[vo/gpu-next/libplacebo] [286]  }
[vo/gpu-next/libplacebo] [287]
[vo/gpu-next/libplacebo] [288]  // let's return the sum unpacked -> we can accumulate it later
[vo/gpu-next/libplacebo] [289]  return h_acc;
[vo/gpu-next/libplacebo] [290] }
[vo/gpu-next/libplacebo] [291]
[vo/gpu-next/libplacebo] [292] float AddDirFilters(float p[6][6], float phase_x_frac, float phase_y_frac, int phase_x_frac_int, int phase_y_frac_int, vec4 w)
[vo/gpu-next/libplacebo] [293] {
[vo/gpu-next/libplacebo] [294]  float f = 0.f;
[vo/gpu-next/libplacebo] [295]  if (w.x > 0.0f)
[vo/gpu-next/libplacebo] [296]  {
[vo/gpu-next/libplacebo] [297]          // 0 deg filter
[vo/gpu-next/libplacebo] [298]          float interp0Deg[6];
[vo/gpu-next/libplacebo] [299]          {
[vo/gpu-next/libplacebo] [300]                  for (int i = 0; i < 6; ++i)
[vo/gpu-next/libplacebo] [301]                  {
[vo/gpu-next/libplacebo] [302]                          interp0Deg[i] = lerp(p[i][2], p[i][3], phase_x_frac);
[vo/gpu-next/libplacebo] [303]                  }
[vo/gpu-next/libplacebo] [304]          }
[vo/gpu-next/libplacebo] [305]          f += EvalPoly6(interp0Deg, phase_y_frac_int) * w.x;
[vo/gpu-next/libplacebo] [306]  }
[vo/gpu-next/libplacebo] [307]
[vo/gpu-next/libplacebo] [308]  if (w.y > 0.0f)
[vo/gpu-next/libplacebo] [309]  {
[vo/gpu-next/libplacebo] [310]          // 90 deg filter
[vo/gpu-next/libplacebo] [311]          float interp90Deg[6];
[vo/gpu-next/libplacebo] [312]          {
[vo/gpu-next/libplacebo] [313]                  for (int i = 0; i < 6; ++i)
[vo/gpu-next/libplacebo] [314]                  {
[vo/gpu-next/libplacebo] [315]                          interp90Deg[i] = lerp(p[2][i], p[3][i], phase_y_frac);
[vo/gpu-next/libplacebo] [316]                  }
[vo/gpu-next/libplacebo] [317]          }
[vo/gpu-next/libplacebo] [318]          f += EvalPoly6(interp90Deg, phase_x_frac_int) * w.y;
[vo/gpu-next/libplacebo] [319]  }
[vo/gpu-next/libplacebo] [320]  if (w.z > 0.0f)
[vo/gpu-next/libplacebo] [321]  {
[vo/gpu-next/libplacebo] [322]          //45 deg filter
[vo/gpu-next/libplacebo] [323]          float pphase_b45;
[vo/gpu-next/libplacebo] [324]          pphase_b45 = 0.5f + 0.5f * (phase_x_frac - phase_y_frac);
[vo/gpu-next/libplacebo] [325]
[vo/gpu-next/libplacebo] [326]          float temp_interp45Deg[7];
[vo/gpu-next/libplacebo] [327]          temp_interp45Deg[1] = lerp(p[2][1], p[1][2], pphase_b45);
[vo/gpu-next/libplacebo] [328]          temp_interp45Deg[3] = lerp(p[3][2], p[2][3], pphase_b45);
[vo/gpu-next/libplacebo] [329]          temp_interp45Deg[5] = lerp(p[4][3], p[3][4], pphase_b45);
[vo/gpu-next/libplacebo] [330]          {
[vo/gpu-next/libplacebo] [331]                  pphase_b45 = pphase_b45 - 0.5f;
[vo/gpu-next/libplacebo] [332]                  float a = (pphase_b45 >= 0.f) ? p[0][2] : p[2][0];
[vo/gpu-next/libplacebo] [333]                  float b = (pphase_b45 >= 0.f) ? p[1][3] : p[3][1];
[vo/gpu-next/libplacebo] [334]                  float c = (pphase_b45 >= 0.f) ? p[2][4] : p[4][2];
[vo/gpu-next/libplacebo] [335]                  float d = (pphase_b45 >= 0.f) ? p[3][5] : p[5][3];
[vo/gpu-next/libplacebo] [336]                  temp_interp45Deg[0] = lerp(p[1][1], a, abs(pphase_b45));
[vo/gpu-next/libplacebo] [337]                  temp_interp45Deg[2] = lerp(p[2][2], b, abs(pphase_b45));
[vo/gpu-next/libplacebo] [338]                  temp_interp45Deg[4] = lerp(p[3][3], c, abs(pphase_b45));
[vo/gpu-next/libplacebo] [339]                  temp_interp45Deg[6] = lerp(p[4][4], d, abs(pphase_b45));
[vo/gpu-next/libplacebo] [340]          }
[vo/gpu-next/libplacebo] [341]
[vo/gpu-next/libplacebo] [342]          float interp45Deg[6];
[vo/gpu-next/libplacebo] [343]          float pphase_p45 = phase_x_frac + phase_y_frac;
[vo/gpu-next/libplacebo] [344]          if (pphase_p45 >= 1)
[vo/gpu-next/libplacebo] [345]          {
[vo/gpu-next/libplacebo] [346]                  for (int i = 0; i < 6; i++)
[vo/gpu-next/libplacebo] [347]                  {
[vo/gpu-next/libplacebo] [348]                          interp45Deg[i] = temp_interp45Deg[i + 1];
[vo/gpu-next/libplacebo] [349]                  }
[vo/gpu-next/libplacebo] [350]                  pphase_p45 = pphase_p45 - 1;
[vo/gpu-next/libplacebo] [351]          }
[vo/gpu-next/libplacebo] [352]          else
[vo/gpu-next/libplacebo] [353]          {
[vo/gpu-next/libplacebo] [354]                  for (int i = 0; i < 6; i++)
[vo/gpu-next/libplacebo] [355]                  {
[vo/gpu-next/libplacebo] [356]                          interp45Deg[i] = temp_interp45Deg[i];
[vo/gpu-next/libplacebo] [357]                  }
[vo/gpu-next/libplacebo] [358]          }
[vo/gpu-next/libplacebo] [359]          f += EvalPoly6(interp45Deg, int(pphase_p45 * 64)) * w.z;
[vo/gpu-next/libplacebo] [360]  }
[vo/gpu-next/libplacebo] [361]
[vo/gpu-next/libplacebo] [362]  if (w.w > 0.0f)
[vo/gpu-next/libplacebo] [363]  {
[vo/gpu-next/libplacebo] [364]          //135 deg filter
[vo/gpu-next/libplacebo] [365]          float pphase_b135 = 0.5f * (phase_x_frac + phase_y_frac);
[vo/gpu-next/libplacebo] [366]
[vo/gpu-next/libplacebo] [367]          float temp_interp135Deg[7];
[vo/gpu-next/libplacebo] [368]
[vo/gpu-next/libplacebo] [369]          temp_interp135Deg[1] = lerp(p[3][1], p[4][2], pphase_b135);
[vo/gpu-next/libplacebo] [370]          temp_interp135Deg[3] = lerp(p[2][2], p[3][3], pphase_b135);
[vo/gpu-next/libplacebo] [371]          temp_interp135Deg[5] = lerp(p[1][3], p[2][4], pphase_b135);
[vo/gpu-next/libplacebo] [372]
[vo/gpu-next/libplacebo] [373]          {
[vo/gpu-next/libplacebo] [374]                  pphase_b135 = pphase_b135 - 0.5f;
[vo/gpu-next/libplacebo] [375]                  float a = (pphase_b135 >= 0.f) ? p[5][2] : p[3][0];
[vo/gpu-next/libplacebo] [376]                  float b = (pphase_b135 >= 0.f) ? p[4][3] : p[2][1];
[vo/gpu-next/libplacebo] [377]                  float c = (pphase_b135 >= 0.f) ? p[3][4] : p[1][2];
[vo/gpu-next/libplacebo] [378]                  float d = (pphase_b135 >= 0.f) ? p[2][5] : p[0][3];
[vo/gpu-next/libplacebo] [379]                  temp_interp135Deg[0] = lerp(p[4][1], a, abs(pphase_b135));
[vo/gpu-next/libplacebo] [380]                  temp_interp135Deg[2] = lerp(p[3][2], b, abs(pphase_b135));
[vo/gpu-next/libplacebo] [381]                  temp_interp135Deg[4] = lerp(p[2][3], c, abs(pphase_b135));
[vo/gpu-next/libplacebo] [382]                  temp_interp135Deg[6] = lerp(p[1][4], d, abs(pphase_b135));
[vo/gpu-next/libplacebo] [383]          }
[vo/gpu-next/libplacebo] [384]
[vo/gpu-next/libplacebo] [385]
[vo/gpu-next/libplacebo] [386]          float interp135Deg[6];
[vo/gpu-next/libplacebo] [387]          float pphase_p135 = 1 + (phase_x_frac - phase_y_frac);
[vo/gpu-next/libplacebo] [388]          if (pphase_p135 >= 1)
[vo/gpu-next/libplacebo] [389]          {
[vo/gpu-next/libplacebo] [390]                  for (int i = 0; i < 6; ++i)
[vo/gpu-next/libplacebo] [391]                  {
[vo/gpu-next/libplacebo] [392]                          interp135Deg[i] = temp_interp135Deg[i + 1];
[vo/gpu-next/libplacebo] [393]                  }
[vo/gpu-next/libplacebo] [394]                  pphase_p135 = pphase_p135 - 1;
[vo/gpu-next/libplacebo] [395]          }
[vo/gpu-next/libplacebo] [396]          else
[vo/gpu-next/libplacebo] [397]          {
[vo/gpu-next/libplacebo] [398]                  for (int i = 0; i < 6; ++i)
[vo/gpu-next/libplacebo] [399]                  {
[vo/gpu-next/libplacebo] [400]                          interp135Deg[i] = temp_interp135Deg[i];
[vo/gpu-next/libplacebo] [401]                  }
[vo/gpu-next/libplacebo] [402]          }
[vo/gpu-next/libplacebo] [403]          f += EvalPoly6(interp135Deg, int(pphase_p135 * 64)) * w.w;
[vo/gpu-next/libplacebo] [404]  }
[vo/gpu-next/libplacebo] [405]  return f;
[vo/gpu-next/libplacebo] [406] }
[vo/gpu-next/libplacebo] [407]
[vo/gpu-next/libplacebo] [408] void hook()
[vo/gpu-next/libplacebo] [409] {
[vo/gpu-next/libplacebo] [410]  uvec2 blockIdx = gl_WorkGroupID.xy;
[vo/gpu-next/libplacebo] [411]  uint threadIdx = gl_LocalInvocationID.x;
[vo/gpu-next/libplacebo] [412]
[vo/gpu-next/libplacebo] [413]  // Figure out the range of pixels from input image that would be needed to be loaded for this thread-block
[vo/gpu-next/libplacebo] [414]  int dstBlockX = int(NIS_BLOCK_WIDTH * blockIdx.x);
[vo/gpu-next/libplacebo] [415]  int dstBlockY = int(NIS_BLOCK_HEIGHT * blockIdx.y);
[vo/gpu-next/libplacebo] [416]
[vo/gpu-next/libplacebo] [417]  const int srcBlockStartX = int(floor((dstBlockX + 0.5f) * kScaleX - 0.5f));
[vo/gpu-next/libplacebo] [418]  const int srcBlockStartY = int(floor((dstBlockY + 0.5f) * kScaleY - 0.5f));
[vo/gpu-next/libplacebo] [419]  const int srcBlockEndX = int(ceil((dstBlockX + NIS_BLOCK_WIDTH + 0.5f) * kScaleX - 0.5f));
[vo/gpu-next/libplacebo] [420]  const int srcBlockEndY = int(ceil((dstBlockY + NIS_BLOCK_HEIGHT + 0.5f) * kScaleY - 0.5f));
[vo/gpu-next/libplacebo] [421]
[vo/gpu-next/libplacebo] [422]  int numTilePixelsX = srcBlockEndX - srcBlockStartX + kSupportSize - 1;
[vo/gpu-next/libplacebo] [423]  int numTilePixelsY = srcBlockEndY - srcBlockStartY + kSupportSize - 1;
[vo/gpu-next/libplacebo] [424]
[vo/gpu-next/libplacebo] [425]  // round-up load region to even size since we're loading in 2x2 batches
[vo/gpu-next/libplacebo] [426]  numTilePixelsX += numTilePixelsX & 0x1;
[vo/gpu-next/libplacebo] [427]  numTilePixelsY += numTilePixelsY & 0x1;
[vo/gpu-next/libplacebo] [428]  const int numTilePixels = numTilePixelsX * numTilePixelsY;
[vo/gpu-next/libplacebo] [429]
[vo/gpu-next/libplacebo] [430]  // calculate the equivalent values for the edge map
[vo/gpu-next/libplacebo] [431]  const int numEdgeMapPixelsX = numTilePixelsX - kSupportSize + 2;
[vo/gpu-next/libplacebo] [432]  const int numEdgeMapPixelsY = numTilePixelsY - kSupportSize + 2;
[vo/gpu-next/libplacebo] [433]  const int numEdgeMapPixels = numEdgeMapPixelsX * numEdgeMapPixelsY;
[vo/gpu-next/libplacebo] [434]
[vo/gpu-next/libplacebo] [435]  // fill in input luma tile (shPixelsY) in batches of 2x2 pixels
[vo/gpu-next/libplacebo] [436]  // we use texture gather to get extra support necessary
[vo/gpu-next/libplacebo] [437]  // to compute 2x2 edge map outputs too
[vo/gpu-next/libplacebo] [438]  {
[vo/gpu-next/libplacebo] [439]          for (uint i = threadIdx * 2; i < uint(numTilePixels) >> 1; i += NIS_THREAD_GROUP_SIZE * 2)
[vo/gpu-next/libplacebo] [440]          {
[vo/gpu-next/libplacebo] [441]                  uint py = (i / numTilePixelsX) * 2;
[vo/gpu-next/libplacebo] [442]                  uint px = i % numTilePixelsX;
[vo/gpu-next/libplacebo] [443]
[vo/gpu-next/libplacebo] [444]                  // 0.5 to be in the center of texel
[vo/gpu-next/libplacebo] [445]                  // - (kSupportSize - 1) / 2 to shift by the kernel support size
[vo/gpu-next/libplacebo] [446]                  float kShift = 0.5f - (kSupportSize - 1) / 2;
[vo/gpu-next/libplacebo] [447]
[vo/gpu-next/libplacebo] [448]                  const float tx = (srcBlockStartX + px + kShift) * kSrcNormX;
[vo/gpu-next/libplacebo] [449]                  const float ty = (srcBlockStartY + py + kShift) * kSrcNormY;
[vo/gpu-next/libplacebo] [450]
[vo/gpu-next/libplacebo] [451]                  float p[2][2];
[vo/gpu-next/libplacebo] [452] #ifdef HOOKED_gather
[vo/gpu-next/libplacebo] [453]                  {
[vo/gpu-next/libplacebo] [454]                          const vec4 sY = HOOKED_gather(vec2(tx, ty), 0);
[vo/gpu-next/libplacebo] [455]
[vo/gpu-next/libplacebo] [456]                          p[0][0] = sY.w;
[vo/gpu-next/libplacebo] [457]                          p[0][1] = sY.z;
[vo/gpu-next/libplacebo] [458]                          p[1][0] = sY.x;
[vo/gpu-next/libplacebo] [459]                          p[1][1] = sY.y;
[vo/gpu-next/libplacebo] [460]                  }
[vo/gpu-next/libplacebo] [461] #else
[vo/gpu-next/libplacebo] [462]                  for (int j = 0; j < 2; j++)
[vo/gpu-next/libplacebo] [463]                  {
[vo/gpu-next/libplacebo] [464]                          for (int k = 0; k < 2; k++)
[vo/gpu-next/libplacebo] [465]                          {
[vo/gpu-next/libplacebo] [466]                                  const float px = HOOKED_tex(vec2(tx + k * kSrcNormX, ty + j * kSrcNormY)).r;
[vo/gpu-next/libplacebo] [467]                                  p[j][k] = px;
[vo/gpu-next/libplacebo] [468]                          }
[vo/gpu-next/libplacebo] [469]                  }
[vo/gpu-next/libplacebo] [470] #endif
[vo/gpu-next/libplacebo] [471]                  const uint idx = py * kTilePitch + px;
[vo/gpu-next/libplacebo] [472]                  shPixelsY[idx] = float(p[0][0]);
[vo/gpu-next/libplacebo] [473]                  shPixelsY[idx + 1] = float(p[0][1]);
[vo/gpu-next/libplacebo] [474]                  shPixelsY[idx + kTilePitch] = float(p[1][0]);
[vo/gpu-next/libplacebo] [475]                  shPixelsY[idx + kTilePitch + 1] = float(p[1][1]);
[vo/gpu-next/libplacebo] [476]          }
[vo/gpu-next/libplacebo] [477]  }
[vo/gpu-next/libplacebo] [478]
[vo/gpu-next/libplacebo] [479]  groupMemoryBarrier();
[vo/gpu-next/libplacebo] [480]  barrier();
[vo/gpu-next/libplacebo] [481]
[vo/gpu-next/libplacebo] [482]  {
[vo/gpu-next/libplacebo] [483]          // fill in the edge map of 2x2 pixels
[vo/gpu-next/libplacebo] [484]          for (uint i = threadIdx * 2; i < uint(numEdgeMapPixels) >> 1; i += NIS_THREAD_GROUP_SIZE * 2)
[vo/gpu-next/libplacebo] [485]          {
[vo/gpu-next/libplacebo] [486]                  uint py = (i / numEdgeMapPixelsX) * 2;
[vo/gpu-next/libplacebo] [487]                  uint px = i % numEdgeMapPixelsX;
[vo/gpu-next/libplacebo] [488]
[vo/gpu-next/libplacebo] [489]                  const uint edgeMapIdx = py * kEdgeMapPitch + px;
[vo/gpu-next/libplacebo] [490]
[vo/gpu-next/libplacebo] [491]                  uint tileCornerIdx = (py+1) * kTilePitch + px + 1;
[vo/gpu-next/libplacebo] [492]                  float p[4][4];
[vo/gpu-next/libplacebo] [493]                  for (int j = 0; j < 4; j++)
[vo/gpu-next/libplacebo] [494]                  {
[vo/gpu-next/libplacebo] [495]                          for (int k = 0; k < 4; k++)
[vo/gpu-next/libplacebo] [496]                          {
[vo/gpu-next/libplacebo] [497]                                  p[j][k] = shPixelsY[tileCornerIdx + j * kTilePitch + k];
[vo/gpu-next/libplacebo] [498]                          }
[vo/gpu-next/libplacebo] [499]                  }
[vo/gpu-next/libplacebo] [500]
[vo/gpu-next/libplacebo] [501]                  shEdgeMap[edgeMapIdx] = vec4(GetEdgeMap(p, 0, 0));
[vo/gpu-next/libplacebo] [502]                  shEdgeMap[edgeMapIdx + 1] = vec4(GetEdgeMap(p, 0, 1));
[vo/gpu-next/libplacebo] [503]                  shEdgeMap[edgeMapIdx + kEdgeMapPitch] = vec4(GetEdgeMap(p, 1, 0));
[vo/gpu-next/libplacebo] [504]                  shEdgeMap[edgeMapIdx + kEdgeMapPitch + 1] = vec4(GetEdgeMap(p, 1, 1));
[vo/gpu-next/libplacebo] [505]          }
[vo/gpu-next/libplacebo] [506]  }
[vo/gpu-next/libplacebo] [507]
[vo/gpu-next/libplacebo] [508]  LoadFilterBanksSh(int(threadIdx), NIS_THREAD_GROUP_SIZE);
[vo/gpu-next/libplacebo] [509]
[vo/gpu-next/libplacebo] [510]  groupMemoryBarrier();
[vo/gpu-next/libplacebo] [511]  barrier();
[vo/gpu-next/libplacebo] [512]
[vo/gpu-next/libplacebo] [513]  // output coord within a tile
[vo/gpu-next/libplacebo] [514]  const ivec2 pos = ivec2(uint(threadIdx) % uint(NIS_BLOCK_WIDTH), uint(threadIdx) / uint(NIS_BLOCK_WIDTH));
[vo/gpu-next/libplacebo] [515]  // x coord inside the output image
[vo/gpu-next/libplacebo] [516]  const int dstX = dstBlockX + pos.x;
[vo/gpu-next/libplacebo] [517]  // x coord inside the input image
[vo/gpu-next/libplacebo] [518]  const float srcX = (0.5f + dstX) * kScaleX - 0.5f;
[vo/gpu-next/libplacebo] [519]  // nearest integer part
[vo/gpu-next/libplacebo] [520]  const int px = int(floor(srcX) - srcBlockStartX);
[vo/gpu-next/libplacebo] [521]  // fractional part
[vo/gpu-next/libplacebo] [522]  const float fx = srcX - floor(srcX);
[vo/gpu-next/libplacebo] [523]  // discretized phase
[vo/gpu-next/libplacebo] [524]  const int fx_int = int(fx * kPhaseCount);
[vo/gpu-next/libplacebo] [525]
[vo/gpu-next/libplacebo] [526]  for (int k = 0; k < NIS_BLOCK_WIDTH * NIS_BLOCK_HEIGHT / NIS_THREAD_GROUP_SIZE; ++k)
[vo/gpu-next/libplacebo] [527]  {
[vo/gpu-next/libplacebo] [528]          // y coord inside the output image
[vo/gpu-next/libplacebo] [529]          const int dstY = dstBlockY + pos.y + k * (NIS_THREAD_GROUP_SIZE / NIS_BLOCK_WIDTH);
[vo/gpu-next/libplacebo] [530]          // y coord inside the input image
[vo/gpu-next/libplacebo] [531]          const float srcY = (0.5f + dstY) * kScaleY - 0.5f;
[vo/gpu-next/libplacebo] [532]
[vo/gpu-next/libplacebo] [533]          // nearest integer part
[vo/gpu-next/libplacebo] [534]          const int py = int(floor(srcY) - srcBlockStartY);
[vo/gpu-next/libplacebo] [535]          // fractional part
[vo/gpu-next/libplacebo] [536]          const float fy = srcY - floor(srcY);
[vo/gpu-next/libplacebo] [537]          // discretized phase
[vo/gpu-next/libplacebo] [538]          const int fy_int = int(fy * kPhaseCount);
[vo/gpu-next/libplacebo] [539]
[vo/gpu-next/libplacebo] [540]          // generate weights for directional filters
[vo/gpu-next/libplacebo] [541]          const int startEdgeMapIdx = py * kEdgeMapPitch + px;
[vo/gpu-next/libplacebo] [542]          vec4 edge[2][2];
[vo/gpu-next/libplacebo] [543]          for (int i = 0; i < 2; i++)
[vo/gpu-next/libplacebo] [544]          {
[vo/gpu-next/libplacebo] [545]                  for (int j = 0; j < 2; j++)
[vo/gpu-next/libplacebo] [546]                  {
[vo/gpu-next/libplacebo] [547]                          // need to shift edge map sampling since it's a 2x2 centered inside 6x6 grid
[vo/gpu-next/libplacebo] [548]                          edge[i][j] = shEdgeMap[startEdgeMapIdx + (i * kEdgeMapPitch) + j];
[vo/gpu-next/libplacebo] [549]                  }
[vo/gpu-next/libplacebo] [550]          }
[vo/gpu-next/libplacebo] [551]          const vec4 w = GetInterpEdgeMap(edge, fx, fy) * NIS_SCALE_INT;
[vo/gpu-next/libplacebo] [552]
[vo/gpu-next/libplacebo] [553]          // load 6x6 support to regs
[vo/gpu-next/libplacebo] [554]          const int startTileIdx = py * kTilePitch + px;
[vo/gpu-next/libplacebo] [555]          float p[6][6];
[vo/gpu-next/libplacebo] [556]          {
[vo/gpu-next/libplacebo] [557]                  for (int i = 0; i < 6; ++i)
[vo/gpu-next/libplacebo] [558]                  {
[vo/gpu-next/libplacebo] [559]                          for (int j = 0; j < 6; ++j)
[vo/gpu-next/libplacebo] [560]                          {
[vo/gpu-next/libplacebo] [561]                                  p[i][j] = shPixelsY[startTileIdx + i * kTilePitch + j];
[vo/gpu-next/libplacebo] [562]                          }
[vo/gpu-next/libplacebo] [563]                  }
[vo/gpu-next/libplacebo] [564]          }
[vo/gpu-next/libplacebo] [565]
[vo/gpu-next/libplacebo] [566]          // weigth for luma
[vo/gpu-next/libplacebo] [567]          const float baseWeight = NIS_SCALE_FLOAT - w.x - w.y - w.z - w.w;
[vo/gpu-next/libplacebo] [568]
[vo/gpu-next/libplacebo] [569]          // final luma is a weighted product of directional & normal filters
[vo/gpu-next/libplacebo] [570]          float opY = 0;
[vo/gpu-next/libplacebo] [571]
[vo/gpu-next/libplacebo] [572]          // get traditional scaler filter output
[vo/gpu-next/libplacebo] [573]          opY += FilterNormal(p, fx_int, fy_int) * baseWeight;
[vo/gpu-next/libplacebo] [574]
[vo/gpu-next/libplacebo] [575]          // get directional filter bank output
[vo/gpu-next/libplacebo] [576]          opY += AddDirFilters(p, fx, fy, fx_int, fy_int, w);
[vo/gpu-next/libplacebo] [577]
[vo/gpu-next/libplacebo] [578]          // do bilinear tap for luma upscaling
[vo/gpu-next/libplacebo] [579]          vec4 op = HOOKED_tex(vec2((srcX + 0.5f) * kSrcNormX, (srcY + 0.5f) * kSrcNormY));
[vo/gpu-next/libplacebo] [580]
[vo/gpu-next/libplacebo] [581]          const float corr = opY * (1.0f / NIS_SCALE_FLOAT) - op.r;
[vo/gpu-next/libplacebo] [582]          op.x += corr;
[vo/gpu-next/libplacebo] [583]
[vo/gpu-next/libplacebo] [584]          imageStore(out_image, ivec2(dstX, dstY), op);
[vo/gpu-next/libplacebo] [585]  }
[vo/gpu-next/libplacebo] [586] }
[vo/gpu-next/libplacebo] [587]
[vo/gpu-next/libplacebo] [588] void _1() {
[vo/gpu-next/libplacebo] [589] hook();
[vo/gpu-next/libplacebo] [590]
[vo/gpu-next/libplacebo] [591] }
[vo/gpu-next/libplacebo] [592]
[vo/gpu-next/libplacebo] [593] void main() {
[vo/gpu-next/libplacebo] [594] _1();
[vo/gpu-next/libplacebo] [595] }
[vo/gpu-next/libplacebo]   Backtrace:
[vo/gpu-next/libplacebo]     #0  0x7ff6418e586f in pl_pass_create+0x81f (D:\App\mpv-Native\mpv.exe+0x119586f) (0x14119586f)
[vo/gpu-next/libplacebo]     #1  0x7ff64256e56e in pl_dispatch_finish+0x2bae (D:\App\mpv-Native\mpv.exe+0x1e1e56e) (0x141e1e56e)
[vo/gpu-next/libplacebo]     #2  0x7ff64256f72d in pl_dispatch_compute+0x16d (D:\App\mpv-Native\mpv.exe+0x1e1f72d) (0x141e1f72d)
[vo/gpu-next/libplacebo]     #3  0x7ff64192a99b in pl_mpv_user_shader_parse+0x699b (D:\App\mpv-Native\mpv.exe+0x11da99b) (0x1411da99b)
[vo/gpu-next/libplacebo]     #4  0x7ff641910e2c in pl_renderer_reset_errors+0x1aec (D:\App\mpv-Native\mpv.exe+0x11c0e2c) (0x1411c0e2c)
[vo/gpu-next/libplacebo]     #5  0x7ff641902f41 in pl_render_image+0x3d91 (D:\App\mpv-Native\mpv.exe+0x11b2f41) (0x1411b2f41)
[vo/gpu-next/libplacebo]     #6  0x7ff6418ff3b4 in pl_render_image+0x204 (D:\App\mpv-Native\mpv.exe+0x11af3b4) (0x1411af3b4)
[vo/gpu-next/libplacebo]     #7  0x7ff64190b586 in pl_render_image_mix+0x2a06 (D:\App\mpv-Native\mpv.exe+0x11bb586) (0x1411bb586)
[vo/gpu-next/libplacebo]     #8  0x7ff6417cb79d in mpv_render_context_get_info+0x17a9d (D:\App\mpv-Native\mpv.exe+0x107b79d) (0x14107b79d)
[vo/gpu-next/libplacebo]     #9  0x7ff6417b07ed in mpv_stream_cb_add_ro+0x97dcd (D:\App\mpv-Native\mpv.exe+0x10607ed) (0x1410607ed)
[vo/gpu-next/libplacebo]     #10 0x7ff9b53f37af in wcsrchr+0x14f (C:\WINDOWS\System32\ucrtbase.dll+0x37af) (0x1800037af)
[vo/gpu-next/libplacebo]     #11 0x7ff9b6ebe8d6 in BaseThreadInitThunk+0x16 (C:\WINDOWS\System32\KERNEL32.DLL+0x2e8d6) (0x18002e8d6)
[vo/gpu-next/libplacebo]     #12 0x7ff9b822c53b in RtlUserThreadStart+0x2b (C:\WINDOWS\SYSTEM32\ntdll.dll+0x8c53b) (0x18008c53b)
[vo/gpu-next/libplacebo] Failed creating render pass for dispatch
[vo/gpu-next/libplacebo] Failed executing hook, disabling

[HedioKojima]

嗯，谢谢。需要和d3d11搭配使用

[hooke007]

不应该存在这个限制，已重新修正 5764d2f

[HedioKojima]

完美了，vulkan也正常了。

[JustWantTheMoney]

请问作者之前mpv-lazy的注册怎么删除，现在显示两个mpv的默认程序。

[hooke007]

无法复现，你的这个默认程序可能不是注册产生的

[JustWantTheMoney]

刚刚又试了下，只有一个了，谢谢，麻烦了作者！

[JustWantTheMoney]

还有个问题，请问这个是默认HDR转换到SDR的吗？

[hooke007]

是

[JustWantTheMoney]

好的，谢谢解答！

[melody0709]

有没有想法搞个支持 iptv m3u 分组 回看 的界面,参考Android酷9,就想在pc上无边框看电视,不想开模拟器看电视
能支持如下格式:图标,回看参数自定义!!!要求有点过分哈

#EXTINF:-1 tvg-id="020000000000000500000000153" tvg-name="CCTV5+体育高清" tvg-logo="http://c.cc.top:7077/183.235.16.92:8081/pics/micro-picture/channelNew/02000000000000050000000000000153.png" catchup="default" catchup-source="http://183.235.162.80:6610/190000002005/ch000000000000203/index.m3u8?starttime=${(b)yyyyMMddHHmmss|UTC}&endtime=${(e)yyyyMMddHHmmss|UTC}" group-title="央视",CCTV5+体育高清
http://o.ee.top:5140/rtp/239.20.0.203:2204

[hooke007]

无

[yemijun]

sub-hdr-peak = 100
hdr-reference-white = 100
这两个有什么必要吗？全部注释掉采用默认值203不可以吗？按照这个设置在target-colorspace-hint = yes的情况下观看非HDR影片好像有些过于昏暗了。

[hooke007]

会误判输出状态

[hooke007]

gpu-next 引入了更多的选项让本就复杂的输出链更加复杂。
hdr输出相关的issue随手一搜就有 mpv-player/mpv#16933
我也不是很想在桌面平台研究并折腾hdr，多方因素下选择保守策略。

[yemijun]

windows平台是这样的。虽然win11的HDR和色彩管理已经有了很大进步但是由于微软选择完全遵守IEC 61966-2-1:1999，采用了分段转换函数而不是伽马2.2来处理SDR在HDR下的转换，导致了不少问题。好像最近上游对色彩管理策略的更改就跟这个有关。也不好说到底那种方案对，这个问题只能交给微软来彻底解决，像MacOS一样可以让用户自行选择EOTF函数。在这个问题彻底解决之前，在桌面端折腾HDR也确实意义不大。同一套配置文件在采用不同EOTF函数的显示器上的表现更是没办法预测。https://www.colour-science.org/posts/srgb-eotf-pure-gamma-22-or-piece-wise-function/

[hooke007]

我想表达的并不是这个意思。 target-colorspace-hint 还有一个auto，但实际上也有报告这玩意会在不应该启用的时候启用，至少在我看来，默认配置倾向hdr输出对windows来讲为时尚早

[yemijun]

明白了，windows上色彩管理HDR相关问题的太过复杂了，HDR方向的配置还是交给用户自己判断选择更为合适。

[Czerror]

换了个最新编译的主程序就好了

[hooke007]

没见过

[Czerror]

我是 win11 25h2 26200.6901

[hooke007]

我的win11也是及时更新随着版本走的，没发生过，以bug11的这个状态，难说到底是谁的问题

[Remonli]

你好，设置border = no 的时候，左上角提示未加载文件。
这其实无所谓，但这提示好像是一系列的，会在打开文件的时候显示一个黑色的文件名，大概1秒左右。
这样就跟mpv.conf里面设置的两秒的osd-playing-msg就重复了。先闪过一个黑色的文件名，然后显示一个紫色的文件名，这就很傻了。
我知道禁用osd-playing-msg可以，我其实倾向于用osd-playing-msg，这个好看一点。
不知道有没有办法？要禁用只能保留border吗？

[hooke007]

mpv_PlayKit/portable_config/script-opts.conf

Line 497 in eff64bc

script-opts-append = uosc-top_bar=no-border # 顶栏的显隐逻辑（仅在无边框和全屏模式下显示），默认 no-border 其它可用的值 never always

[Remonli]

这个文件里面选项好多，但我下载你release的包里面并没有这么多选项，但效果也一样是怎么做到的。

[hooke007]

你觉得文档的意思是什么

mpv_PlayKit/portable_config/script-opts.conf

Line 1 in f014406

### 文档_ https://github.com/hooke007/mpv_PlayKit/blob/main/portable_config/script-opts.conf

[Remonli]

额，我的意思是release包里的script-opt.conf 很精简，没有文档里那么多行，但实际上也实现了文档里选项的效果。
比如音量控件放在右边，size是40。 是不是uosc默认就是右边，40 ？ 但我翻了翻uosc的源码，没找到有关的默认值，我不太懂编程，
有点搞不清原理了。

[hooke007]

文档之所以是文档，自然是全面。mpv的配置千行，有必要全部写吗
https://mpv.io/manual/master/

[hooke007]

一键下载assets的所有包 https://github.com/hooke007/dotfiles/blob/main/GitHub%20Desktop/dl_gh_assets.ps1
示例：下载所有k7sfunc测试过的ai模型

PowerShell -ExecutionPolicy Bypass -File ./dl_gh_assets.ps1 -Owner "hooke007" -Repo "dotfiles" -Tag "onnx_models"

[hooke007]

• [vapoursynth] k7sfunc 扩展支持 Apple Silicon #632

clip = k7f.RIFE_COREML(clip, fps_in=container_fps, be=1)

clip = k7f.UAI_COREML(clip, model_pth="the_database_AnimeJaNaiV3L1_sharp_HD_x2_fp32_op17.onnx", fp16_qnt=False, be=1)

clip = k7f.BM3D_METAL(clip)

[mshiliu-xyb]

使用alist挂载115网盘，然后用mpv播放alist下的115网盘的某些视频，会开始有几秒钟的正常播放，然后卡住一直缓冲，怀疑过是服务器问题！使用PotPlayer播放却是正常的！更诡异的事情是，某些电影可以播放，某些不行！也不是解码什么的问题，因为一开始是可以播放几秒钟的！

[mshiliu-xyb]

#552 我的意思是有人报告过，无关版本。 上游也有报告过，没人用也没人在意

我抓包了一下，感觉是证书问题，我使用了 --no-check-certificates这个忽略证书就可以下载！但是在mpv中要如何配置yt的命令呢？

[hooke007]

mpv_PlayKit/portable_config/mpv.conf

Lines 254 to 256 in 5d7c63d

	ytdl-raw-options-append =
	# [当 --ytdl=yes 时有效] 将自定义的选项传递给ytdl，默认为空 https://github.com/ytdl-org/youtube-dl#options
	# 示例值（ cookies-from-browser=firefox ）仅为传递单个选项，多个选项最好独立追加，即写多个 --ytdl-raw-options-append 参数

你再看一下本地的mpv.conf ，示例也是有的

[mshiliu-xyb]

也不是证书的问题，yt解析好了，d暗示这个会一直连接不知原因！为什么同一个网盘不同资源结果不一样，好像和解析出来的ip有关系！

[hooke007]

如图所示是ffmpeg报告的问题。

[mshiliu-xyb]

如图所示是ffmpeg报告的问题。

解决不了，感觉莫名其妙的

[SuperBlackWednesday]

我的显卡是5070 无法开启120帧模式 下个版本能不能直接编进滤镜里 或者出个教程讲一下怎么开启？网络上的教程不行 我的5070开不了

[hooke007]

什么120帧模式？

[SuperBlackWednesday]

https://blog.csdn.net/gitblog_00800/article/details/151915253 这个能集成到播放器里吗？ 我估计现在有很多人都不再使用60Hz的显示器了

[hooke007]

一篇 ai 写的不明所以的水文（CSDN全是垃圾怎么会真的有人会看里边发的东西的。。。

[Lafowle]

兄弟不是我要吐槽，把VS濾鏡功能讀懂就知道怎麼用了

[zwo1a]

 cscale = bilinear
 scale = bilinear
 dscale = bicubic

为什么默认要用 bilinear，而不是使用 lanczos？

[hooke007]

为什么要使用lanczos

[hooke007]

先说说lazy目前的设计，用bilinear+预设启用的SGEDS着色器组合来复现原始SGSR算法，这个拆分设计的好处是可以自定义缩放核，不像原始实现是隐式的强制bilinear，当前情况bilinear作为缩放核能提供的锐度已经足够了甚至会偶尔溢出，所以我之前又局部重写了SGEDS的锐化逻辑来抑制一些边缘情况。

其次，lanczos 在各种方面看都不是一个优势选项，一定只在使用内置的算法里挑一个我也是spline36。如果单用内置算法这两能提供的锐度也够看了，但是绑上SGEDS就大概有点多余。

[hooke007]

霞鹜文楷的原版体积太膨胀了，后边切换到 https://github.com/lxgw/LxgwWenKai-Lite

[Ladersbt]

大佬我想问一下，mpv上游是从大概一个月前开始的内置截图机制有变化吗？就是我之前一直是用
s screenshot subtitles #menu: 截取 > 截屏 > 同源尺寸 - 有字幕 - 无 OSD-单帧
S screenshot video #menu: 截取 > 截屏 > 同源尺寸 - 无字幕 - 无 OSD-单帧
截图的，用您的最新版本在input uosc.conf加入这个，使用后截出来的图会比视频实际显示更暗，把mpv主程序相关换回一个月以前的版本就正常了，这是什么原因

[hooke007]

没留意

[Ladersbt]

没留意

而且就是用您发布的最新的版本里自带的截图功能也有这个问题，input uosc.conf里是有三个截图相关的，第一个是正常的，其他的两个也会出现截图比视频暗的情况

[hooke007]

我的意思是没留意上游

影响参数 screenshot-high-bit-depth = no

mpv-player/mpv#17057

[Ladersbt]

我的意思是没留意上游

影响参数 screenshot-high-bit-depth = no

mpv-player/mpv#17057

原来是这样，感谢大佬解答

[hooke007]

https://github.com/routineLife1/DRBA 验证可行性的实验性支持测试（RTX4080）

RIFE_NV(model=46)

DRBA_NV(model=2)

其它接口DML COREML未达预期性能，放弃。

[hooke007]

• DRBA 实时补帧 #641 随26年第一版增强包上线
• 94a5607

[hooke007]

这个模型 int8 量化性能反而出现了倒退，质量倒是没有可见退化，很例外。
6efaffc 完整实现了任意整数倍，2x速率略下滑，在3x及以上仍然比rife4.6快

[hooke007]

0c4fd58
非整数倍补帧也实现，但是降速严重(v2模型 x2.5， 198fps/s)，比x3甚至x4都慢，在播放情境下除了强迫症想不出有什么理由使用它

[gatorawu]

请问有什么脚本可以从MPV里调用ffmpeg 录屏并保存为mp4?

[gatorawu]

敲命令麻烦，有没有脚本把录制功能绑一个快捷键？

[hooke007]

你在说什么？你系统的录屏功能还要敲命令？？

[gatorawu]

从控制台用dump-cashe命令。用系统的录制软件就不能碰鼠标，会把菜单栏和鼠标录进去。

[hooke007]

1 那写进 input 里不就完事了
2 可以换软件

[3a311088]

我之前曾想要用https://github.com/natural-harmonia-gropius/hdr-toys的效果，放USB插電視上看，錄影會導致檔案非常龐大，問了ai，使用ffmpeg轉檔套用這些著色器似乎要改寫一些東西，看似非常麻煩，我也看不懂，ai也沒辦法改出簡單的解決辦法

[hooke007]

• [vapoursynth] k7sfunc 迁移与发布 1.0.0 #643

[hooke007]

https://github.com/hooke007/K7sfunc/releases/
pypi和手动安装包都上线了，后续k7f的更新全都在新仓库

[verygoodlee]

import k7sfunc as k7f都会报错，vsmlrt找不到滤镜，至少要下载一个它的滤镜放在vs-plugins文件夹下才能正常使用

D:\Downloads\py.3_14_0+vs.R72-portable>python -m pip install k7sfunc
Collecting k7sfunc
  Downloading k7sfunc-1.0.0-py3-none-any.whl.metadata (2.4 kB)
Downloading k7sfunc-1.0.0-py3-none-any.whl (112 kB)
Installing collected packages: k7sfunc
Successfully installed k7sfunc-1.0.0

D:\Downloads\py.3_14_0+vs.R72-portable>python
Python 3.14.0 (tags/v3.14.0:ebf955d, Oct  7 2025, 10:15:03) [MSC v.1944 64 bit (AMD64)] on win32
Type "help", "copyright", "credits" or "license" for more information.
>>> import k7sfunc as k7f
Traceback (most recent call last):
  File "<python-input-0>", line 1, in <module>
    import k7sfunc as k7f
  File "D:\Downloads\py.3_14_0+vs.R72-portable\Lib\site-packages\k7sfunc\__init__.py", line 7, in <module>
    from .mod_scale import *
  File "D:\Downloads\py.3_14_0+vs.R72-portable\Lib\site-packages\k7sfunc\mod_scale.py", line 17, in <module>
    from ._external import vsmlrt, nnedi3_resample
  File "D:\Downloads\py.3_14_0+vs.R72-portable\Lib\site-packages\k7sfunc\_external\vsmlrt.py", line 80, in <module>
    plugins_path: str = get_plugins_path()
                        ~~~~~~~~~~~~~~~~^^
  File "D:\Downloads\py.3_14_0+vs.R72-portable\Lib\site-packages\k7sfunc\_external\vsmlrt.py", line 75, in get_plugins_path
    raise RuntimeError("vsmlrt: cannot load any filters")
RuntimeError: vsmlrt: cannot load any filters

不应强制绑定vsmlrt，把导入vsmlrt的语句from ._external import vsmlrt放在要使用它的函数内部应该就可以了，在要使用到时再导入

[hooke007]

插件肯定是需要的，这个整个生态是混乱的没办法统一协议分发的。
external目前替代的是vs-scripts的目录，一是我自己基本做了大量fork修改没法和上游的版本绑定，这个时候原先的版本号检查已经没有意义了

[hooke007]

我get到你的意思了，我这本地的环境有点奇怪没有报错

[zjkd110]

如何设置点击窗口视频任意位置就能暂停播放，再点一下就能继续播放，每次暂停都需要点播放/暂停那个按钮比较麻烦。。

[zjkd110]

改了后，单击窗口视频任意位置，是窗口最大化和还原，并不是播放/暂停

[hooke007]

你要不要再仔细看看你改的东西，或者提供日志证明

[hooke007]

你的脑补能力很强，我哪一句话让你改fullscreen的

[zjkd110]

知道什么原因了，应该是pause 不是fullscreen

[zjkd110]

哈哈，终于好了，万分感谢啊！！！！！

[rasiel2k]

win10企业版 LTSC 21H2,字幕设置不起作用，全部强制微软雅黑。强制使用单一字体也不行，触发字体回退机制。

[rasiel2k]

上个版本没问题

[hooke007]

请使用纯净版复现，但就算能复现我大概也解决不了这种某个版本专属的离奇问题
也可以自行更新/后退主程序版本

[rasiel2k]

行吧，先用上个版本

[jiankeyan]

请问如果以后有新版本，如何从旧版本更新到新版本呢，是直接下载新版本覆盖旧版本，那还需不需要重新注册一遍呢？

[hooke007]

彻底删除干净然后重新解压

保证mpv.exe的位置不变即可，除非注册行为本身有更新

[jiankeyan]

懂了，感谢大大！

[jiankeyan]

忘了问下大大，如果有新版本注册行为有更新的话会说明是吧？

[hooke007]

看每个版本的发版说明

[jiankeyan]

好的，再次感谢！

[3a311088]

整個畫面明顯的閃爍，感覺是很多雜訊在整個畫面各處閃，但視覺上某些顏色會特別明顯
上上個月重置windows後發現同樣的影片、同樣的mpv設定檔、同樣的驅動程式，畫面卻很完美沒有瑕疵
相同的影片檔案在電視上播放沒有問題，而且他曾經在電腦上完美播放過，因此可以排除是片源的問題
顯示器不支援可變更新頻率、HDR、FreeSync功能從顯示器直接關掉了、視窗化遊戲最佳化也沒有開
嘗試過修改顯示器頻率為片源的整數倍或其他，59.94、60、100、120、119.88、120等，沒改善
這之後大概就裝了windows自帶的系統更新，和一些遊戲
遊戲似乎似乎有一併安裝了DirectX和一些Microsoft可轉散發套件
但這DirectX在控制台裡的程式列表找不到，沒法單獨移除

不知道問題在哪，不知道有沒有大佬能隔空抓藥

unknown_2025.12.04-15.14.mp4

猜測可能會影響的
vo = gpu-next
gpu-context = winvk
hwdec = d3d11va-copy (顯卡不支援vulkan及vulkan-copy)
deinterlace=no
dither-depth=8 (曾經改成no，沒改善)
scale = bilinear
cscale = bilinear
dscale = bicubic

[hooke007]

你的录屏没有闪烁

[hooke007]

画面噪点过多，可以把默认的着色器关掉

[3a311088]

你的录屏没有闪烁

看到這段我很訝異 我覺得閃爍的很明顯
立馬拿手機看這段影片 還真的看不出來...
我好像幹了一件類似用妄想用截圖去拍螢幕破損的蠢事
但手機錄影會有嚴重的摩爾紋 會覆蓋掉那些點點的閃爍
我在想想...

[hooke007]

录屏没有问题说明问题不在播放器

[hooke007]

esrgan在mpv上的glsl实现测试结果。

只能勉强跑最简化的srvggnet架构下的superultracompact这类极简模型

1080p-x2性能占用不算特别高但是平均帧计时基本迫近60Hz的间隔了（16.6ms）
d3d11后端编译时长半分钟，但是播放基本不丢帧；winvk后端编译时长10秒，但是播放时有稀碎丢帧。

尝试上一级的ultracompact不出意外失去可用性。
首先是在d3d11后端用了11分钟编译（主要受限于单线程），其次每一步的平均帧计时都差不多是之前的总和了

最后就是播个dvd都卡。。。

[hooke007]

开放测试 8fea026
维基也同步了portable页面的特性，可点击emoji跳转路径 https://github.com/hooke007/mpv_PlayKit/wiki/4_GLSL#esrgan-

[hooke007]

macvk后端画面直接抽搐，暂停又是正常的。
总之就是vulkan后端不正常。

[hooke007]

逆向实现onnx通用版nnedi3的结果。

nns256 性能非常糟糕，1080p x2 trt马力开足的理论帧率也不达预期，安安心心用着色器版即可

[CrayonSSR]

我用mpv打开的在线缓冲视频，不会鼠标划入时，在导航进度中生成thumbfast的预览图。

大佬，我想问一下，这是我的设置问题？还是目前的uosc中还没兼容在线视频使用thumbfast？

[hooke007]

mpv_PlayKit/portable_config/script-opts.conf

Line 440 in b5965e4

script-opts-append = thumbfast-network=no # 是否对流媒体启用，默认 no