perf: bypass matched OoT movie reconstruction in diagnostic mode

THIRD_PARTY_NOTICES.md

@@ -36,3 +36,5 @@ Fetched or linked at build time (not in this repository):
 | [ALVR](https://github.com/alvr-org/ALVR) client v20.14.0, downloaded (SHA-256 checked) and installed on the headset by `zeldaVR/install_vr.sh` | shows the VR picture on the Meta Quest 3 | MIT |
 | [Noto Sans CJK SC](https://github.com/notofonts/noto-cjk) (Source Han Sans) DemiLight at `165c01b`, downloaded and checked by `windows_games/ra2/scripts/languages.py` | the Simplified Chinese glyphs it draws into the player's game font | SIL OFL 1.1 |
 | [OpenCC](https://github.com/BYVoid/OpenCC), run (not included) by `windows_games/ra2/scripts/languages.py` | Traditional to Simplified Chinese | Apache-2.0 |
+
+OoT movie diagnostic: command-sequence matching adapted from [nxnative PR #1331](https://github.com/seedds/nxnative/pull/1331), `76c20edba480db753ce7092cd2677f6c4a8daa15`, MIT (`games/totk/third_party/licenses/nxnative.txt`).

games/totk/Makefile

@@ -147,7 +147,7 @@ verify-boot: runtime
 
 
 
-GFX_HEADERS := $(wildcard third_party/d2r_gfx/gfx/nvn/*.h third_party/d2r_gfx/gfx/nvn/include/*.h third_party/d2r_gfx/host/*.h zeldaVR/gfx/*.h zeldaVR/runtime/*.h)
+GFX_HEADERS := third_party/d2r_gfx/gfx/nvn/oot_movie_direct.inc third_party/d2r_gfx/gfx/nvn/oot_movie_probe.inc $(wildcard third_party/d2r_gfx/gfx/nvn/*.h third_party/d2r_gfx/gfx/nvn/include/*.h third_party/d2r_gfx/host/*.h zeldaVR/gfx/*.h zeldaVR/runtime/*.h)
 
 build/gfx/%.o: % $(GFX_HEADERS) Makefile
 	@mkdir -p "$(dir $@)"
@@ -515,6 +515,10 @@ build/totk-native-oot: $(wildcard runtime/*.[ch] runtime/*.m runtime/*.S) $(GFX_
 runtime-oot binary-oot: SHADER_LIST ?= ../oot/assets/shader_list.tsv
 binary-oot: build/totk-native-oot build/pso_manifest.tsv build/pso_title.tsv
 runtime-oot: binary-oot game-data-oot
+
+.PHONY: probe-gpu-oot-movie
+probe-gpu-oot-movie: build/totk-native-oot
+	MTL_DEBUG_LAYER=1 MTL_DEBUG_LAYER_ERROR_MODE=assert TOTK_HEADLESS=1 TOTK_OOT_MOVIE_DIRECT=1 build/totk-native-oot rewritten/oot $@
 # Production Opus-wrapper regression checks; no game assets or audio device needed.
 OPUS_TEST_WORK ?= /tmp/totk-native-opus-codec
 OPUS_PREFIX ?= /opt/homebrew/opt/opus

games/totk/runtime/gpu_path_probe.m

@@ -7,8 +7,9 @@
 #include <string.h>
 
 void runtime_gpu_path_probe(const char *which) {
- int checked = !strcmp(which, "probe-gpu-writeback") ? mtl_probe_writeback() : !strcmp(which, "probe-gpu-sampler-border") ? mtl_probe_sampler_border() : !strcmp(which, "probe-gpu-pool-retire") ? mtl_probe_pool_retire() : !strcmp(which, "probe-gpu-tex-slots") ? mtl_probe_tex_slots() : !strcmp(which, "probe-gpu-arg-epoch") ? mtl_probe_arg_epoch() : !strcmp(which, "probe-gpu-anomaly") ? mtl_probe_anomaly() : !strcmp(which, "probe-gpu-vertex-streams") ? mtl_probe_vertex_streams() : !strcmp(which, "probe-gpu-vin") ? mtl_probe_vertex_input_types() : !strcmp(which, "probe-gpu-fout") ? mtl_probe_fragment_output_types() : !strcmp(which, "probe-gpu-pipeline-archive") ? mtl_probe_pipeline_archive() : !strcmp(which, "probe-gpu-layer-merge") ? mtl_probe_layer_merge() : !strcmp(which, "probe-gpu-cube-merge") ? mtl_probe_cube_merge() : !strcmp(which, "probe-gpu-texture-retire") ? mtl_probe_texture_retire() : !strcmp(which, "probe-gpu-texels-region") ? mtl_probe_texels_region() : !strcmp(which, "probe-gpu-linear-rt") ? mtl_probe_linear_rt_upload() : !strcmp(which, "probe-gpu-storage-view") ? mtl_probe_storage_view() : !strcmp(which, "probe-gpu-compressed-alias") ? mtl_probe_compressed_alias() : !strcmp(which, "probe-gpu-upload-batch") ? mtl_probe_upload_batch() : !strcmp(which, "probe-gpu-ds-clear") ? mtl_probe_ds_clear() : !strcmp(which, "probe-gpu-ub-rename") ? mtl_probe_ub_rename() : !strcmp(which, "probe-gpu-ubo-snapshot") ? mtl_probe_ubo_snapshot() : !strcmp(which, "probe-gpu-drv-c0") ? mtl_probe_drv_c0() : !strcmp(which, "probe-gpu-copy-alias") ? mtl_probe_copy_alias() : !strcmp(which, "probe-gpu-dispatch-chain") ? mtl_probe_dispatch_chain() : !strcmp(which, "probe-gpu-srgb-view") ? mtl_probe_srgb_view() : !strcmp(which, "probe-gpu-depth-output") ? mtl_probe_depth_output() : !strcmp(which, "probe-gpu-widen") ? mtl_probe_widen() : !strcmp(which, "probe-gpu-midclear") ? mtl_probe_midclear() : !strcmp(which, "probe-gpu-winding") ? mtl_probe_winding() : !strcmp(which, "probe-gpu-tile") ? mtl_probe_tile() : mtl_probe_blockrt();
+ int checked = !strcmp(which, "probe-gpu-oot-movie") ? mtl_probe_oot_movie() : !strcmp(which, "probe-gpu-writeback") ? mtl_probe_writeback() : !strcmp(which, "probe-gpu-sampler-border") ? mtl_probe_sampler_border() : !strcmp(which, "probe-gpu-pool-retire") ? mtl_probe_pool_retire() : !strcmp(which, "probe-gpu-tex-slots") ? mtl_probe_tex_slots() : !strcmp(which, "probe-gpu-arg-epoch") ? mtl_probe_arg_epoch() : !strcmp(which, "probe-gpu-anomaly") ? mtl_probe_anomaly() : !strcmp(which, "probe-gpu-vertex-streams") ? mtl_probe_vertex_streams() : !strcmp(which, "probe-gpu-vin") ? mtl_probe_vertex_input_types() : !strcmp(which, "probe-gpu-fout") ? mtl_probe_fragment_output_types() : !strcmp(which, "probe-gpu-pipeline-archive") ? mtl_probe_pipeline_archive() : !strcmp(which, "probe-gpu-layer-merge") ? mtl_probe_layer_merge() : !strcmp(which, "probe-gpu-cube-merge") ? mtl_probe_cube_merge() : !strcmp(which, "probe-gpu-texture-retire") ? mtl_probe_texture_retire() : !strcmp(which, "probe-gpu-texels-region") ? mtl_probe_texels_region() : !strcmp(which, "probe-gpu-linear-rt") ? mtl_probe_linear_rt_upload() : !strcmp(which, "probe-gpu-storage-view") ? mtl_probe_storage_view() : !strcmp(which, "probe-gpu-compressed-alias") ? mtl_probe_compressed_alias() : !strcmp(which, "probe-gpu-upload-batch") ? mtl_probe_upload_batch() : !strcmp(which, "probe-gpu-ds-clear") ? mtl_probe_ds_clear() : !strcmp(which, "probe-gpu-ub-rename") ? mtl_probe_ub_rename() : !strcmp(which, "probe-gpu-ubo-snapshot") ? mtl_probe_ubo_snapshot() : !strcmp(which, "probe-gpu-drv-c0") ? mtl_probe_drv_c0() : !strcmp(which, "probe-gpu-copy-alias") ? mtl_probe_copy_alias() : !strcmp(which, "probe-gpu-dispatch-chain") ? mtl_probe_dispatch_chain() : !strcmp(which, "probe-gpu-srgb-view") ? mtl_probe_srgb_view() : !strcmp(which, "probe-gpu-depth-output") ? mtl_probe_depth_output() : !strcmp(which, "probe-gpu-widen") ? mtl_probe_widen() : !strcmp(which, "probe-gpu-midclear") ? mtl_probe_midclear() : !strcmp(which, "probe-gpu-winding") ? mtl_probe_winding() : !strcmp(which, "probe-gpu-tile") ? mtl_probe_tile() : mtl_probe_blockrt();
  printf("{\"status\":\"pass\",\"probe\":\"%s\",\"checked\":%d,\"scope\":\"%s\"}\n", which, checked,
+        !strcmp(which, "probe-gpu-oot-movie") ? "OoT matched movie reconstruction, fallback, RGBA/sRGB and output ownership" :
         !strcmp(which, "probe-gpu-writeback") ? "terrain exact snapshots and BotW tiny GPU render completion write-back, CPU overwrite protection and padding" :
         !strcmp(which, "probe-gpu-sampler-border") ? "production VS/FS sampler border colour switching, mixed addressing and GPU readback" :
         !strcmp(which, "probe-gpu-texels-region") ? "production texel GPU region read/write, pitch guards, packed/block formats, mip/layer/volume, overlap fallback and CPU/Metal mirror visibility" :

games/totk/runtime/loader.c

@@ -641,5 +641,5 @@ int main(int argc,char **argv) {
 #endif
  verify_identity(!strcmp(argv[2],"probe-identity-negative"));
  for(unsigned i=0;i<NM;i++){seal(&modules[i]);check_permissions(&modules[i]);}
- if(!strcmp(argv[2],"probe-gpu-writeback")||!strcmp(argv[2],"probe-gpu-sampler-border")||!strcmp(argv[2],"probe-gpu-pool-retire")||!strcmp(argv[2],"probe-gpu-tex-slots")||!strcmp(argv[2],"probe-gpu-arg-epoch")||!strcmp(argv[2],"probe-gpu-anomaly")||!strcmp(argv[2],"probe-gpu-vertex-streams")||!strcmp(argv[2],"probe-gpu-vin")||!strcmp(argv[2],"probe-gpu-fout")||!strcmp(argv[2],"probe-gpu-pipeline-archive")||!strcmp(argv[2],"probe-gpu-layer-merge")||!strcmp(argv[2],"probe-gpu-cube-merge")||!strcmp(argv[2],"probe-gpu-texture-retire")||!strcmp(argv[2],"probe-gpu-texels-region")||!strcmp(argv[2],"probe-gpu-linear-rt")||!strcmp(argv[2],"probe-gpu-storage-view")||!strcmp(argv[2],"probe-gpu-compressed-alias")||!strcmp(argv[2],"probe-gpu-upload-batch")||!strcmp(argv[2],"probe-gpu-ds-clear")||!strcmp(argv[2],"probe-gpu-ub-rename")||!strcmp(argv[2],"probe-gpu-ubo-snapshot")||!strcmp(argv[2],"probe-gpu-drv-c0")||!strcmp(argv[2],"probe-gpu-dispatch-chain")||!strcmp(argv[2],"probe-gpu-copy-alias")||!strcmp(argv[2],"probe-gpu-srgb-view")||!strcmp(argv[2],"probe-gpu-depth-output"))runtime_gpu_path_probe(argv[2]);else if(!strcmp(argv[2],"probe-memory"))runtime_memory_probe();else if(!strcmp(argv[2],"probe-gfx"))runtime_gfx_probe(0);else if(!strcmp(argv[2],"probe-vpool"))runtime_vpool_probe();else if(!strcmp(argv[2],"probe-gpu-widen")||!strcmp(argv[2],"probe-gpu-blockrt")||!strcmp(argv[2],"probe-gpu-midclear")||!strcmp(argv[2],"probe-gpu-winding")||!strcmp(argv[2],"probe-gpu-tile"))runtime_gpu_path_probe(argv[2]);else if(!strcmp(argv[2],"probe-audio"))runtime_audio_probe();else if(!strcmp(argv[2],"probe-gfx-negative"))runtime_gfx_probe(1);else if(!strcmp(argv[2],"probe"))probe();else if(!strcmp(argv[2],"probe-fs"))runtime_fs_probe();else if(!strcmp(argv[2],"seed-shaders")){if(argc<4)runtime_fail("seed-shaders needs the corpus: <modules> seed-shaders <corpus.bin> (scripts/ns2_shader_corpus.py)");runtime_shader_seed(argv[3],argc>4?argv[4]:NULL);}else if(!strcmp(argv[2],"probe-swkbd"))runtime_swkbd_probe();else if(!strcmp(argv[2],"probe-save"))runtime_save_probe();else if(!strcmp(argv[2],"boot")){if(getenv("TOTK_VERIFY_STARTUP")||getenv("TOTK_HEADLESS"))boot();else runtime_front_run(boot);}else runtime_fail("unknown mode");return 0;
+ if(!strcmp(argv[2],"probe-gpu-oot-movie")||!strcmp(argv[2],"probe-gpu-writeback")||!strcmp(argv[2],"probe-gpu-sampler-border")||!strcmp(argv[2],"probe-gpu-pool-retire")||!strcmp(argv[2],"probe-gpu-tex-slots")||!strcmp(argv[2],"probe-gpu-arg-epoch")||!strcmp(argv[2],"probe-gpu-anomaly")||!strcmp(argv[2],"probe-gpu-vertex-streams")||!strcmp(argv[2],"probe-gpu-vin")||!strcmp(argv[2],"probe-gpu-fout")||!strcmp(argv[2],"probe-gpu-pipeline-archive")||!strcmp(argv[2],"probe-gpu-layer-merge")||!strcmp(argv[2],"probe-gpu-cube-merge")||!strcmp(argv[2],"probe-gpu-texture-retire")||!strcmp(argv[2],"probe-gpu-texels-region")||!strcmp(argv[2],"probe-gpu-linear-rt")||!strcmp(argv[2],"probe-gpu-storage-view")||!strcmp(argv[2],"probe-gpu-compressed-alias")||!strcmp(argv[2],"probe-gpu-upload-batch")||!strcmp(argv[2],"probe-gpu-ds-clear")||!strcmp(argv[2],"probe-gpu-ub-rename")||!strcmp(argv[2],"probe-gpu-ubo-snapshot")||!strcmp(argv[2],"probe-gpu-drv-c0")||!strcmp(argv[2],"probe-gpu-dispatch-chain")||!strcmp(argv[2],"probe-gpu-copy-alias")||!strcmp(argv[2],"probe-gpu-srgb-view")||!strcmp(argv[2],"probe-gpu-depth-output"))runtime_gpu_path_probe(argv[2]);else if(!strcmp(argv[2],"probe-memory"))runtime_memory_probe();else if(!strcmp(argv[2],"probe-gfx"))runtime_gfx_probe(0);else if(!strcmp(argv[2],"probe-vpool"))runtime_vpool_probe();else if(!strcmp(argv[2],"probe-gpu-widen")||!strcmp(argv[2],"probe-gpu-blockrt")||!strcmp(argv[2],"probe-gpu-midclear")||!strcmp(argv[2],"probe-gpu-winding")||!strcmp(argv[2],"probe-gpu-tile"))runtime_gpu_path_probe(argv[2]);else if(!strcmp(argv[2],"probe-audio"))runtime_audio_probe();else if(!strcmp(argv[2],"probe-gfx-negative"))runtime_gfx_probe(1);else if(!strcmp(argv[2],"probe"))probe();else if(!strcmp(argv[2],"probe-fs"))runtime_fs_probe();else if(!strcmp(argv[2],"seed-shaders")){if(argc<4)runtime_fail("seed-shaders needs the corpus: <modules> seed-shaders <corpus.bin> (scripts/ns2_shader_corpus.py)");runtime_shader_seed(argv[3],argc>4?argv[4]:NULL);}else if(!strcmp(argv[2],"probe-swkbd"))runtime_swkbd_probe();else if(!strcmp(argv[2],"probe-save"))runtime_save_probe();else if(!strcmp(argv[2],"boot")){if(getenv("TOTK_VERIFY_STARTUP")||getenv("TOTK_HEADLESS"))boot();else runtime_front_run(boot);}else runtime_fail("unknown mode");return 0;
 }

games/totk/third_party/d2r_gfx/gfx/nvn/nvn_mtl.h

@@ -56,6 +56,7 @@ int mtl_probe_srgb_view(void);
 int mtl_probe_storage_view(void);
 int mtl_probe_compressed_alias(void);
 int mtl_probe_linear_rt_upload(void);
+int mtl_probe_oot_movie(void);
 int mtl_probe_texels_region(void);
 int mtl_probe_texture_retire(void);
 int mtl_probe_ds_clear(void);

games/totk/third_party/d2r_gfx/gfx/nvn/nvn_mtl.m

@@ -1212,7 +1212,12 @@ static void b2_scene_shot(int eye) {   // diagnostic: the eye's own last pass ta
         else MLOG("[vr] b2shot: this eye pass drew no presented texture"); }
 }
 void mtl_eye_capture(int eye) { g_b2_capture = eye; }
+static id<MTLTexture> g_oot_movie_decoded;
+static uint64_t g_oot_movie_sub;
+static const uint8_t *g_oot_movie_skip_end;
 void mtl_sub_begin(void) {
+    g_oot_movie_decoded = nil;
+    g_oot_movie_skip_end = NULL;
     g_eye_region = 3;   // stereo B2: a submission starts outside eye marks
     g_fc_begin0 = clock_gettime_nsec_np(CLOCK_UPTIME_RAW);
     /* The submission's autorelease pool, drained in mtl_sub_commit, opens before anything else: the guest's submit
@@ -14934,7 +14939,8 @@ stages with device visibility (blit()), as the end / start barriers of two encod
     if (g_disp_n % 20000 == 0) MLOG("[nvn] 计算派发 至 %llu 次：其中 %llu 次（%.0f%%）接在已开的计算编码器里，没有另开编码器", (unsigned long long)g_disp_n, (unsigned long long)g_disp_shared, 100.0 * (double)g_disp_shared / (double)g_disp_n);
     return g_blit;
 }
-static void dispatch(const p_dispatch *dp) {
+#include "oot_movie_direct.inc"
+static void dispatch_at(const p_dispatch *dp, const uint8_t *next) {
     zskip_clear();
     const prog_t *cp = S.cprog ? S.cprog : S.prog;   // the compute program (S.cprog); a probe that sets S.prog only, as before
     if (!cp) { RED_ONCE(27, "派发前没有绑定程序"); return; }
@@ -14945,6 +14951,9 @@ static void dispatch(const p_dispatch *dp) {
     if (cpr->missing) { static uint64_t n; if (!n++) MLOG("[shader-miss] 计算着色器还没翻译：跳过这个派发（以后同类不再提示）"); return; }
     NvShader *cs = cpr->sh[5];
     if (!cs) { RED_ONCE(28, "派发的程序没有计算着色器（或未映射）"); return; }
+    int movie = oot_movie_enabled() && dp->x > 0 && dp->y > 0 && dp->z == 1 ? oot_movie_role(cs->key) : 0;
+    if (movie == 1) g_oot_movie_decoded = nil;
+    if (movie == 2 && oot_movie_replace(cs)) return;
     /* TOTK_SKIP_CS=all | <key prefix>[,<key prefix>...] (diagnostic): those compute dispatches are not run (bisecting a
        GPU hang to its shader); TOTK_SKIP_CS_LOG=1 names each compute shader once as it is first dispatched. */
     { static char list[16][17]; static int nl = -1, all, logk; static NSMutableSet *seen;
@@ -15028,6 +15037,10 @@ first waits for the encode stage to finish that pass (rs_sync, ~5% of the replay
     if (atomic_load(&g_draw_log) > 0) MLOG("[uw] F%llu pass %llu dispatch cs %.16s grid %dx%dx%d", (unsigned long long)atomic_load(&g_cur_frame), g_stat_pass, cs->key, dp->x, dp->y, dp->z);
     enc_c_dispatch(ce, MTLSizeMake(dp->x, dp->y, dp->z), MTLSizeMake(tg[0], tg[1], tg[2]));
     if (!reuse) enc_c_end(ce);
+    if (movie == 1) {
+        oot_movie_remember(cs);
+        g_oot_movie_skip_end = oot_movie_try_network(next);
+    }
     bright_lut(cs);   // 3D 画面加亮：游戏算完调色表之后重写 | 3D picture brightening: rewrite after the game finishes computing the color lookup table
     /* TOTK_EXPO_LOG=<file> (diagnostic): after each dispatch of the exposure kernel (ae725063), copy floats 96..127 of its
        storage buffer 0 (statistics; exposure at 108..111) into a shared ring and write them with the frame number once
@@ -15045,6 +15058,7 @@ first waits for the encode stage to finish that pass (rs_sync, ~5% of the replay
 }
 // GPU：贴图 → 缓冲（行距取 SetCopyRowStride，0 = 紧排）
 // GPU: texture -> buffer (row stride taken from SetCopyRowStride, 0 = tightly packed)
+static void dispatch(const p_dispatch *dp) { dispatch_at(dp, NULL); }
 static void copy_t2b(const p_copy_b2t *c) {
     { static long on = -1; static int left = 24; if (on < 0) { const char *e = getenv("TOTK_T2T_LOG"); on = e ? atol(e) : 0; if (e && !on) on = 1; }   // with TOTK_T2T_LOG: buffer <-> texture copies too
       if (on > 0 && (long)atomic_load(&g_cur_frame) >= on && left > 0 && c->tex) { left--;
@@ -15766,7 +15780,11 @@ void mtl_exec(const uint8_t *p) {
             case OP_COPY_T2B: copy_t2b(a); break;
             case OP_COPY_T2T: copy_t2t(a); break;
             case OP_CLEAR_TEX: clear_tex(a); break;
-            case OP_DISPATCH: dispatch(a); break;
+            case OP_DISPATCH:
+                if (g_oot_movie_skip_end) {
+                    if (p == g_oot_movie_skip_end) g_oot_movie_skip_end = NULL;
+                } else dispatch_at(a, p + h.size);
+                break;
             case OP_SET_TEX_POOL: { const dpool_t *tp = ((const p_set_pool *)a)->pool; extern int nvn_tex_pool_base(const void *pool);   // hle_nvn.c: IDs past this pool's part of the table read 0
                                     S.tex_pool_n = tp ? nvn_tex_pool_base(tp) + tp->n : 0; break; }
             case OP_BARRIER: case OP_SET_SAMP_POOL: break;
@@ -18171,6 +18189,8 @@ int mtl_probe_compressed_alias(void) {
 
 /* CPU-visible linear-render-target textures contain ordinary pool data until the GPU writes them.
    Exercise both water-height formats, cold sampling, a CPU texel update, and subsequent GPU rendering. */
+#include "oot_movie_probe.inc"
+
 int mtl_probe_linear_rt_upload(void) {
     mtl_init();int checked=0;
     for(int mode=0;mode<2;mode++) {