A uniform slot the game left unbound gets a plain line naming its shader instead of a [RED]

dm/games/0400C76024710000/README.md

@@ -38,6 +38,7 @@ Measured on the M4 Pro, in a 1920x1080 window: 60 fps at the title, 44 to 52 fps
 - [Queue builder size getters](../../../docs/reference/nvn-objects.md#queue-builder-size-getters): read after `SetDefaults` as it makes its queue.
 - [A sync never fenced since its Initialize is signalled](../../../docs/reference/nvn-objects.md#a-sync-never-fenced-since-its-initialize-is-signalled).
 - [The provoking vertex is the first](../../../docs/reference/rendering.md#the-provoking-vertex-is-the-first): it sets mode 0.
+- [An unbound uniform slot reads the zero bank](../../../docs/reference/rendering.md#an-unbound-uniform-slot-reads-the-zero-bank): compute shader `9d5de89b049f5915_b` declares cb1, which the game leaves unbound, and reads it only for a workgroup count of 2,048 or more. It was a `[RED]` line; now a plain one.
 - [Touch gestures recognized from the touch points](../../../docs/reference/services.md#touch-gestures-recognized-from-the-touch-points).
 - [SDF definitions in RE Engine PAKs](../../../docs/reference/shaders.md#sdf-definitions-in-re-engine-paks).
 - [Integer outputs told from float bits](../../../docs/reference/shaders.md#integer-outputs-told-from-float-bits).
@@ -68,7 +69,6 @@ Measured on the M4 Pro, in a 1920x1080 window: 60 fps at the title, 44 to 52 fps
 ## Open
 
 - `[RED] NVN unimplemented function called: nvnCommandBufferSetFloatBlendClamp`: the driver writes method 0x10b0, which NVIDIA's published class header (`clb197.h`) lists only as a reserved software method (read through a summary of the file, not verbatim), so what it selects is not known; the game passes 1.
-- `[RED] stage 5 cb1 has no uniform buffer bound: shader 9d5de89b049f5915_b`: the shader reads slot 1 only for a workgroup count of 2,048 or more, so the unbound slot is not read; the line is still printed.
 - `[RED] queue 4: an event wait has held it 8.1 s`, once in one run, as the game left the title: the wait sat inside a frame of 11.9 s, 11.6 s of it 29 pipelines compiled at first use (the run skipped the prepare step), so the producer's submission was late, not missing. Not seen in the two runs after it.
 - The two shaders under [Shaders](#shaders).
 - How long the prepare step's 48,957 pipelines take on a first start has not been measured. A `pso_manifest.tsv` has not been recorded. Combat, later areas, a long session and iOS are not tested; the alias region above has no iPhone answer.

dm/gfx/nvn/nvn_cbmiss.h

@@ -0,0 +1,13 @@
+// A uniform (constant) buffer a shader declares that the replay cannot give it (pure; unit test tests/t_cbmiss.c). nvn_mtl.m cbuf_slot binds the all-zero bank in its place in
+// every case, so the shader reads 0; what differs is whether that is the game's own state or a binding the host lost:
+//   CB_UNBOUND  the game left the slot unbound (address 0) on this queue: its own state, the host missed no bind (docs/reference/rendering.md, "An unbound uniform slot reads
+//               the zero bank"). A plain log line for the first 8 shaders, no [RED]
+//   CB_LOST     the game bound an address that is in no memory pool the host knows: the shader reads 0 where the game bound data. [RED], with the shader and the address
+//   CB_SLOT     a slot past the 16 the replay keeps (a reflection the host cannot place). [RED]
+#pragma once
+#include <stdint.h>
+
+enum { CB_BOUND = 0, CB_UNBOUND = 1, CB_LOST = 2, CB_SLOT = 3 };
+
+static inline int nvn_cb_state(int reg, uint64_t addr, int found) { return reg < 0 || reg >= 16 ? CB_SLOT : !addr ? CB_UNBOUND : found ? CB_BOUND : CB_LOST; }
+static inline int nvn_cb_red(int state) { return state == CB_LOST || state == CB_SLOT; }

dm/gfx/nvn/nvn_mtl.m

@@ -34,6 +34,7 @@
 #include "nvn_border.h"
 #include "nvn_primitive.h"
 #include "nvn_timestamp.h"
+#include "nvn_cbmiss.h"
 #include "msl_image.h"
 #include "dmsh_chain.h"   // DMSH_CHAIN: the translation chain version of the B shader library expected by this source tree, one for every game (beside this file, tools/dmsh_chain.py --header)
 #include "game_profile.h"   // game profile constants (profile id, the profile's switches)
@@ -11075,7 +11076,9 @@ static uint64_t cbuf_slot(NvShader *sh, int nvn_stage, const nvn_rbind *b, __str
     int reg = b->reg; const p_bind_buf *u = reg < 16 ? &S.ubo[nvn_stage][reg] : NULL;
     void *mb; uint64_t mo, gpu;
     if (!u || !u->addr || !mtl_gpu_addr(u->addr, u->size ? u->size : 1, &mb, &mo, &gpu)) {   // reads 0 (the all-zero bank), as an unbound texture slot does, instead of through a null pointer
-        if (!g_red_once[16]) { char note[400] = ""; if (u && u->addr) dm_gfx_addr_note(u->addr, note, sizeof note);   // which shader, and what the address is (unbound, or in no pool)
+        if (!nvn_cb_red(nvn_cb_state(u ? reg : -1, u ? u->addr : 0, 0))) {   // left unbound by the game (address 0): its own state, the host missed no bind (nvn_cbmiss.h)
+            if (!(sh->logged & 32)) { sh->logged |= 32; static int nlog; if (nlog++ < 8) MLOG("stage %d cb%d of shader %s is not bound (address 0): it reads the zero bank, the game's own state (first 8 shaders)", nvn_stage, reg, sh->key); } }
+        else if (!g_red_once[16]) { char note[400] = ""; if (u && u->addr) dm_gfx_addr_note(u->addr, note, sizeof note);   // a binding the host lost: which shader, and where the address is
             RED_ONCE(16, "stage %d cb%d has no uniform buffer bound: shader %s, address %#llx size %#llx (frame %llu)%s", nvn_stage, reg, sh->key, u ? (unsigned long long)u->addr : 0ull, u ? (unsigned long long)u->size : 0ull,
                      (unsigned long long)atomic_load(&g_cur_frame), note); }
         *use = zero_bank(); return zero_bank_gpu(); }

dm/gfx/tests/Makefile

@@ -80,6 +80,8 @@ t_hmap: t_hmap.c ../nvn/nvn_hmap.h
 	clang -O1 -Wall -o $@ t_hmap.c
 t_cover: t_cover.c ../nvn/nvn_cover.h
 	clang -O1 -Wall -o $@ t_cover.c
+t_cbmiss: t_cbmiss.c ../nvn/nvn_cbmiss.h
+	clang -O1 -Wall -o $@ t_cbmiss.c
 t_pace: t_pace.c ../nvn/nvn_pace.h
 	clang -O1 -Wall -o $@ t_pace.c
 t_timestamp: t_timestamp.c ../nvn/nvn_timestamp.h

dm/gfx/tests/t_cbmiss.c

@@ -0,0 +1,17 @@
+// Uniform buffers the replay cannot give a shader (nvn_cbmiss.h): a slot the game left unbound reads the zero bank (its own state, no [RED]); an address in no memory pool, or a slot the
+// replay does not keep, is a binding the host lost ([RED]). The host printed "[RED] stage 5 cb1 has no uniform buffer bound" for both, without the shader, on every PRAGMATA boot
+#include "../nvn/nvn_cbmiss.h"
+#include <stdio.h>
+
+static int fail, checks;
+static void check(int ok, const char *what) { checks++; fail += !ok; printf("%-78s %s\n", what, ok ? "✓" : "✗"); }
+
+int main(void) {
+    check(nvn_cb_state(1, 0, 0) == CB_UNBOUND && !nvn_cb_red(CB_UNBOUND), "cb1 at address 0: the game's own state, no [RED]");
+    check(nvn_cb_state(1, 0x381f8c000ull, 0) == CB_LOST && nvn_cb_red(CB_LOST), "cb1 at an address in no pool: lost, [RED]");
+    check(nvn_cb_state(1, 0x381f8c000ull, 1) == CB_BOUND && !nvn_cb_red(CB_BOUND), "cb1 at a pool address: bound");
+    check(nvn_cb_state(16, 0, 0) == CB_SLOT && nvn_cb_red(CB_SLOT), "cb16 (past the replay's 16 slots): [RED]");
+    check(nvn_cb_state(-1, 0x1000, 1) == CB_SLOT, "a negative slot: [RED]");
+    printf("%d/%d checks passed\n", checks - fail, checks);
+    return fail != 0;
+}

docs/agents/symptoms.md

@@ -828,7 +828,7 @@ One line per symptom in the [subsystem references](../reference/): what you see
 - the `residency set` line shows `passes ended to land them for a read` in the tens of thousands, every one `refused … earlier submission running`. The slower the GPU, the more passes, and each pass makes it slower: a scene stays at a very low frame rate once it falls in. → [Inline updates of a running submission refused uniform shadows](../reference/rendering.md#inline-updates-of-a-running-submission-refused-uniform-shadows)
 - a scene that holds 60 fps with the GPU at about 5 ms a frame drops for seconds at a time to 35–40 fps, with the GPU at 20–26 ms, after a slow frame such as a load. The `residency set` line shows tens of thousands of refusals per 300 frames, every one `a report still to land` (before this fix they were counted as `earlier submission running`). → [Counter reports of a running submission refused uniform shadows by page](../reference/rendering.md#counter-reports-of-a-running-submission-refused-uniform-shadows-by-page)
 - one scene of a game drops to 1-2 fps and stays there, with the GPU busy 400-600 ms a frame and about ten times the render passes of the scenes before it, many of one draw each on the same target. The `residency set` line shows `passes ended to land them for a read` in the tens of thousands, nearly all `refused … earlier submission running`, with some `unreadable`. A pass dump (`DM_NVN_PASSDUMP`) names `draw_` → `ubd_conflict` as the end of those passes (`E <pass> pass ends in <function>:<line>`). → [A submission past 8,192 queued writes counted as writing everywhere](../reference/rendering.md#a-submission-past-8192-queued-writes-counted-as-writing-everywhere)
-- `stage 0 cb5 has no uniform buffer bound` (or another stage and slot, such as `stage 0 cb0`), once per run. → [An unbound uniform slot reads the zero bank](../reference/rendering.md#an-unbound-uniform-slot-reads-the-zero-bank)
+- `stage 0 cb5 of shader <key> is not bound (address 0): it reads the zero bank, the game's own state` (or another stage and slot, such as `stage 0 cb0`), a plain line for each of the first 8 shaders. Before, the same case printed `[RED] stage 0 cb5 has no uniform buffer bound`, once per run. → [An unbound uniform slot reads the zero bank](../reference/rendering.md#an-unbound-uniform-slot-reads-the-zero-bank)
 - GPU time per frame doubles with many small dispatches (Age of Calamity: GPU 9 → 18 ms a frame on the Mac; the iPhone's 15 → 51 ms, 17 fps). → [Compute dispatches share the copy encoder](../reference/rendering.md#compute-dispatches-share-the-copy-encoder)
 - with `DM_NVN_DISPATCH_OWN=1` the host crashes in the Metal driver (`CRASH sig=11 addr=0x98` in `AGX::ComputeContext<…>::prepareForEnqueue(bool)`) at a compute shader's first read through a texture view of another format; by default (dispatches sharing the copy encoder) that read can see the texture's old pixels. Metal validation reports `Command buffer already has an open command encoder` for the first. → [Compute bindings prepare texture copies before the dispatch encoder](../reference/rendering.md#compute-bindings-prepare-texture-copies-before-the-dispatch-encoder)
 - dialog boxes with holes the size of their text, and white specks where text faded. → [Fragment-ordering barriers between draws that write images](../reference/rendering.md#fragment-ordering-barriers-between-draws-that-write-images)

docs/reference/rendering.md

@@ -1360,11 +1360,11 @@ The replay on Metal (`dm/gfx/nvn/nvn_mtl.m`) and the Vulkan and OpenGL layers. E
 
 ### An unbound uniform slot reads the zero bank
 
-- **Symptom:** `stage 0 cb5 has no uniform buffer bound` (or another stage and slot, such as `stage 0 cb0`), once per run.
+- **Symptom:** `stage 0 cb5 of shader <key> is not bound (address 0): it reads the zero bank, the game's own state` (or another stage and slot, such as `stage 0 cb0`), a plain line for each of the first 8 shaders. Before, the same case printed `[RED] stage 0 cb5 has no uniform buffer bound`, once per run.
 - **Cause:** the argument buffer kept a null pointer for the unbound slot. The game draws with nothing bound there; the host misses no bind. What the Switch reads from an unbound slot is not known.
-- **Fix:** the slot now reads the all-zero bank, as an unbound texture slot reads the zero texture.
-- **Found on:** [Paper Mario: The Thousand-Year Door](../../dm/games/0100ECD018EBE000/README.md) (at the title); [Batman: Arkham Knight](../../dm/games/0100ACD0163D0000/README.md#files-and-shaders) (`stage 0 cb0`, at boot: the first command buffer draws before the game has bound any uniform buffer).
-- **Check:** a pass dump of the frame (`DM_NVN_PASSDUMP=<frame>`, frame 0 for the boot) shows the draw (`j` line) and its stage's `K` line with no `B … cb` line, and the frame's `[nvn]` call counts show whether the game called `CommandBufferBindUniformBuffer` at all.
+- **Fix:** the slot now reads the all-zero bank, as an unbound texture slot reads the zero texture. A slot at address 0 is the game's own state, so it gets a plain line naming the shader; `[RED] … has no uniform buffer bound` is kept for a binding the host lost: an address in no memory pool, or a slot past the 16 the replay keeps (`nvn_cbmiss.h`, `cbuf_slot`). The binding is the same in every case.
+- **Found on:** [Paper Mario: The Thousand-Year Door](../../dm/games/0100ECD018EBE000/README.md) (at the title); [Batman: Arkham Knight](../../dm/games/0100ACD0163D0000/README.md#files-and-shaders) (`stage 0 cb0`, at boot: the first command buffer draws before the game has bound any uniform buffer); [PRAGMATA](../../dm/games/0400C76024710000/README.md) (`stage 5 cb1`, compute shader `9d5de89b049f5915_b`, which reads the slot only for a workgroup count of 2,048 or more).
+- **Check:** `make -C dm/gfx/tests GAME=<id> t_cbmiss && dm/gfx/tests/t_cbmiss`. A pass dump of the frame (`DM_NVN_PASSDUMP=<frame>`, frame 0 for the boot) shows the draw (`j` line) and its stage's `K` line with no `B … cb` line, and the frame's `[nvn]` call counts show whether the game called `CommandBufferBindUniformBuffer` at all.
 
 ### Compute dispatches share the copy encoder