mirror of
https://github.com/tangerineturtle1/NX2navi.git
synced 2026-11-15 19:19:58 -04:00
a1fcbb0b07
The driver keeps the prepared shaders' libraries in its shader cache, and under Windows that cache has a size limit set in NVIDIA's settings. A limit smaller than the set drops the oldest libraries as new ones come in (a 2048 MB cache could not hold them on the reporter's RTX 5070 Ti), so every start builds them again in the background and play stutters until that is done. The cold lines of the library warm now say so on NVIDIA under Windows, most plainly when the cache is cold right after a warm that finished on the same driver; the build kit's notes and the example file say it too. tests/warm_hint_check.py checks which hint each case gets. From pull request #403 by tarballz; two conflicts resolved (an include and the build kit's notes). Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
73 lines
3.9 KiB
C
73 lines
3.9 KiB
C
// runtime/vk_warm_hint.h: what the library warm's log adds when the driver's cache of stage libraries is cold
|
|
// (runtime/vk_pipeline.c lw_main), by GPU vendor, by platform and by why it is cold. Built and run by
|
|
// tests/warm_hint_check.py; prints each failed check and exits 1 on any.
|
|
#include <stdio.h>
|
|
#include <string.h>
|
|
|
|
#include "vk_warm_hint.h"
|
|
|
|
static int fails;
|
|
#define CHECK(c, ...) do { if (!(c)) { fails++; printf("FAIL line %d: ", __LINE__); printf(__VA_ARGS__); printf("\n"); } } while (0)
|
|
|
|
enum { NVIDIA = 0x10de, AMD = 0x1002, INTEL = 0x8086 };
|
|
enum { LINUX = 0, WINDOWS = 1 };
|
|
|
|
static int has(const char *s, const char *part) { return s && strstr(s, part) != NULL; }
|
|
|
|
int main(void) {
|
|
// Windows, cold at a start with no finished warm before it (a driver update, or the build's preparation step,
|
|
// which writes no shader-warm.txt): on NVIDIA the driver's Shader Cache Size limit is named as another cause,
|
|
// with the setting.
|
|
const char *first = lw_cold_hint(NVIDIA, LW_COLD_FIRST, WINDOWS);
|
|
CHECK(has(first, "Shader Cache Size") && has(first, "Unlimited"), "NVIDIA first cold: %s", first);
|
|
CHECK(first && first[0] == ' ', "a hint starts with a space, as it follows a sentence: \"%s\"", first);
|
|
CHECK(!strcmp(lw_cold_hint(AMD, LW_COLD_FIRST, WINDOWS), ""), "AMD first cold: no NVIDIA setting");
|
|
CHECK(!strcmp(lw_cold_hint(INTEL, LW_COLD_FIRST, WINDOWS), ""), "Intel first cold: no NVIDIA setting");
|
|
|
|
// Windows, cold right after a finished warm on the same driver: on NVIDIA that is the cache size limit, said
|
|
// plainly.
|
|
const char *again = lw_cold_hint(NVIDIA, LW_COLD_AFTER_WARM, WINDOWS);
|
|
CHECK(has(again, "Shader Cache Size") && has(again, "Unlimited"), "NVIDIA cold after a warm: %s", again);
|
|
CHECK(strcmp(again, first), "cold after a warm says more than a first cold");
|
|
CHECK(!strcmp(lw_cold_hint(AMD, LW_COLD_AFTER_WARM, WINDOWS), ""), "AMD cold after a warm: no NVIDIA setting");
|
|
|
|
// Warming given up after two cold starts in a row: every vendor, on every platform, is told how to warm again
|
|
// once the cache can hold the libraries; NVIDIA on Windows also names the setting that lets it.
|
|
for (int os = LINUX; os <= WINDOWS; os++)
|
|
for (unsigned v = 0; v < 3; v++) {
|
|
unsigned vendor = v == 0 ? NVIDIA : v == 1 ? AMD : INTEL;
|
|
const char *gave = lw_cold_hint(vendor, LW_COLD_GAVE_UP, os);
|
|
CHECK(has(gave, "lib_warm = force") && has(gave, "shader-warm.txt"), "os %d vendor %04x gave up: %s", os,
|
|
vendor, gave);
|
|
}
|
|
CHECK(has(lw_cold_hint(NVIDIA, LW_COLD_GAVE_UP, WINDOWS), "Shader Cache Size"),
|
|
"NVIDIA gave up: names the setting");
|
|
CHECK(!has(lw_cold_hint(AMD, LW_COLD_GAVE_UP, WINDOWS), "Shader Cache Size"), "AMD gave up: no NVIDIA setting");
|
|
|
|
// Elsewhere there is no NVIDIA app and no Shader Cache Size setting: the driver's cache is set otherwise there,
|
|
// so no case names them.
|
|
for (int why = LW_COLD_FIRST; why <= LW_WARM_FORCED; why++) {
|
|
const char *h = lw_cold_hint(NVIDIA, why, LINUX);
|
|
CHECK(!has(h, "NVIDIA app") && !has(h, "Shader Cache Size"), "Linux NVIDIA case %d: %s", why, h);
|
|
}
|
|
CHECK(!strcmp(lw_cold_hint(NVIDIA, LW_COLD_FIRST, LINUX), ""), "Linux NVIDIA first cold: nothing to add");
|
|
|
|
// A warm lib_warm = force started on a cache that is not cold: no cause to name, on any vendor or platform.
|
|
for (int os = LINUX; os <= WINDOWS; os++)
|
|
for (unsigned v = 0; v < 3; v++) {
|
|
unsigned vendor = v == 0 ? NVIDIA : v == 1 ? AMD : INTEL;
|
|
CHECK(!strcmp(lw_cold_hint(vendor, LW_WARM_FORCED, os), ""), "os %d vendor %04x forced warm: no hint", os,
|
|
vendor);
|
|
}
|
|
|
|
// The game passes the platform it was built for.
|
|
#ifdef _WIN32
|
|
CHECK(LW_ON_WINDOWS == 1, "a Windows build says so");
|
|
#else
|
|
CHECK(LW_ON_WINDOWS == 0, "a build elsewhere says so");
|
|
#endif
|
|
|
|
if (!fails) printf("warm hint: all checks passed\n");
|
|
return fails != 0;
|
|
}
|