MacKwnoX/NX2navi
1
0
Fork 0
mirror of https://github.com/tangerineturtle1/NX2navi.git synced 2026-11-15 19:19:58 -04:00
Code Issues Packages Projects Releases Wiki Activity

Log how long the memory report took on the presenting thread.

Browse Source 
The report runs at the present every 3600 frames in every run (600 with trace = perf) and walks the texture entries, the pipelines and the C runtime heap's summary; the 600-frame frames of the 2026-11-13 session carried 10-17 ms nothing else accounted for, and the report's own comment said it took well under a millisecond. A line after it now gives the time of the vram lines, the holes line and the owners' lines with what follows them, so the next session shows whether the report needs to move off that thread.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
chrg6
2026-11-13 12:42:19 -05:00
parent 7b0dbeb18d
commit b81a9561ca
1 changed files with 17 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

21
runtime/vk_memory.c
View File

@@ -988,8 +988,21 @@ void vk_mem_census(uint64_t frame) {
__atomic_store_n(&span_frame, frame, __ATOMIC_RELAXED);
if (dmabuf_on && frame % 60 == 0) span_gc(frame, 0);
static uint64_t window_start;
if ((rt_trace & RT_TRACE_PERF) && frame % 600 == 0) vram_log(frame), mirror_pool_log(frame);
if (frame && frame % ((rt_trace & RT_TRACE_PERF) ? 600 : HOLES_PERIOD) == 0) holes_log(frame), owner_log(frame);
// The report runs here, on the presenting thread, so its time is timed: the 600-frame frames of a session with
// NX2NAVI_TRACE=perf carried 10-17 ms nothing else accounted for (2026-11-13), and the walks of the texture
// entries, the pipelines and the C runtime heap's summary all happen in it.
uint64_t r0 = qpc(), rv = 0;
if ((rt_trace & RT_TRACE_PERF) && frame % 600 == 0) vram_log(frame), mirror_pool_log(frame), rv = qpc() - r0;
if (frame && frame % ((rt_trace & RT_TRACE_PERF) ? 600 : HOLES_PERIOD) == 0) {
uint64_t h0 = qpc();
holes_log(frame);
uint64_t h1 = qpc();
owner_log(frame);
uint64_t h2 = qpc();
rt_log("mem: the report of frame %llu took %.2f ms on the presenting thread: vram lines %.2f ms, holes %.2f ms, "
"owners and the rest %.2f ms", (unsigned long long)frame, qpc_ms(h2 - r0), qpc_ms(rv), qpc_ms(h1 - h0),
qpc_ms(h2 - h1));
}
if (vk_texture_compact_due()) compact_look(); // NX2NAVI_TEX_COMPACT: 0 at once with the setting off
if ((rt_trace & RT_TRACE_PERF) && frame % 60 == 0) outside_watch(frame), up_ts_log(frame), vk_gtl_log(frame, 60);
if ((rt_trace & RT_TRACE_PERF) && !vk_mirror_on) dirty_frame(frame);
@@ -5641,8 +5654,8 @@ static void beyond_log(VkCtx *c, uint64_t frame, const VkPhysicalDeviceMemoryBud
// with DLSS's share as its statistics report it. Then the host-visible allocations (staging above all) and the
// process's system memory with the game's own heap. Then the released textures' reuse (vk_texture_reuse_log).
// A category that only climbs from line to line over a long session is a leak. The figures are sums the owners keep
// already, one driver query, VMA's budget query and one walk of the texture entries (vk_texture_census): well under a
// millisecond once a minute, at the present that logs.
// already, one driver query, VMA's budget query and one walk of the texture entries (vk_texture_census), at the present
// that logs; vk_mem_census logs what the report took.
static void owner_log(uint64_t frame) {
VkCtx *c = vk_ctx();
if (!c || !c->vma) return;