Give frame generation and FSR the scene's own near and far planes, field of view and units.

DLSS frame generation (Streamline and the NGX path) and FSR 3.1 and 4 upscaling and frame generation were told the depth came from a camera with near 0.1 and no far plane. Ocarina's projection has a finite far plane that changes with the camera: in Hyrule Field its z row is (-1.00001, -0.200001), near 0.1 and far about 25,000; the camera before it had far about 1,000. With the far plane taken as missing, a surface 10,000 units away was read as 16,650. The planes, and whether the depth is inverted, now come from the frame's projection and the device's depth mode every frame (vk_dlaa.c proj_planes), with one log line when they change. The stored depth is already 0 near and 1 far: the -w..w clip depth the pipelines keep maps to exactly what a 0..w projection with the same planes stores, and a frame dump's scene depth spans 0.975 to 1.0.

FSR's contexts lose their infinite-depth flag and take the far plane from each dispatch. FSR frame generation now gets the camera's field of view instead of a fixed 1 radian, and the upscaler gets it too (its caller passed 0). The view's units are metres, not the centimetres FSR was told: the camera orbits Link about 4 units behind him at about 3 units a second.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
tangerineturtle1
2026-11-11 13:19:31 -05:00
parent 6a011d2adf
commit e1cb6301ab
7 changed files with 116 additions and 31 deletions

View File

@@ -3,9 +3,10 @@
// headers are in third_party/ffx_api (MIT); they are C++ in places, hence this file. Nothing of it runs unless the
// upscaler or frame generation asks for FSR 3.1.
//
// FSR 3.1 takes the same inputs vk_dlaa.c makes for DLSS: the scene's light target (linear HDR), its D32F depth
// (OpenGL-style, an infinite far plane, not inverted) and the camera motion vectors in render pixels from the current
// frame to the previous one, the convention both share, so the vectors go in unscaled.
// FSR 3.1 takes the same inputs vk_dlaa.c makes for DLSS: the scene's light target (linear HDR), its D32F depth (0
// near, 1 far, with the near and far planes of the frame's own projection, vk_dlaa.c proj_planes) and the camera
// motion vectors in render pixels from the current frame to the previous one, the convention both share, so the
// vectors go in unscaled.
#ifdef _WIN32
#include <windows.h>
#endif
@@ -191,7 +192,8 @@ Up *context_for(uint32_t w, uint32_t h, uint32_t ow, uint32_t oh) {
ffxCreateContextDescUpscale d = {};
d.header.type = FFX_API_CREATE_CONTEXT_DESC_TYPE_UPSCALE;
d.header.pNext = &be.header;
d.flags = FFX_UPSCALE_ENABLE_HIGH_DYNAMIC_RANGE | FFX_UPSCALE_ENABLE_DEPTH_INFINITE | FFX_UPSCALE_ENABLE_AUTO_EXPOSURE;
// no infinite-depth flag: the game's far plane is finite (vk_dlaa.c proj_planes) and each dispatch gives it
d.flags = FFX_UPSCALE_ENABLE_HIGH_DYNAMIC_RANGE | FFX_UPSCALE_ENABLE_AUTO_EXPOSURE;
d.maxRenderSize = {mw, mh};
d.maxUpscaleSize = {ow, oh};
d.fpMessage = message;
@@ -271,7 +273,7 @@ extern "C" int vk_ffx_trim_idle(uint64_t recent) {
}
extern "C" int vk_ffx_upscale(VkCommandBuffer cb, const VkFfxImage *color, const VkFfxImage *depth, const VkFfxImage *mv,
const VkFfxImage *out, float jx, float jy, int reset, float dt_ms, float fov_y) {
const VkFfxImage *out, float jx, float jy, int reset, float dt_ms, const VkFfxCamera *cam) {
if (!vk_ffx_load()) return 0;
uint64_t now = plat_now_ns();
trim(now);
@@ -280,7 +282,7 @@ extern "C" int vk_ffx_upscale(VkCommandBuffer cb, const VkFfxImage *color, const
// would blend this picture with whatever that upscaler last saw.
static int last; // 3 or 4: which one made the frame before
if (vk_ffx_fsr4()) {
if (vk_ffx4_upscale(cb, color, depth, mv, out, jx, jy, reset || last != 4, dt_ms, fov_y)) {
if (vk_ffx4_upscale(cb, color, depth, mv, out, jx, jy, reset || last != 4, dt_ms, cam)) {
if (last != 4) rt_log("ffx: the upscaler is %s", vk_ffx4_name());
last = 4;
return 1;
@@ -312,10 +314,11 @@ extern "C" int vk_ffx_upscale(VkCommandBuffer cb, const VkFfxImage *color, const
d.frameTimeDelta = dt_ms;
d.preExposure = 1.0f;
d.reset = reset != 0;
d.cameraNear = 0.1f; // Ocarina's projection (vk_dlaa.c ocam_vp): near 0.1, far at infinity
d.cameraFar = 1e8f;
d.cameraFovAngleVertical = fov_y > 0 ? fov_y : 1.0f;
d.viewSpaceToMetersFactor = 0.01f; // the game's units are centimetres
// the frame's camera (vk_dlaa.c proj_planes)
d.cameraNear = cam && cam->znear > 0 ? cam->znear : 0.1f;
d.cameraFar = cam && cam->zfar > 0 ? cam->zfar : 1e8f;
d.cameraFovAngleVertical = cam && cam->fov_y > 0 ? cam->fov_y : 1.0f;
d.viewSpaceToMetersFactor = cam && cam->meters > 0 ? cam->meters : 1.0f;
ffxReturnCode_t r = A.dispatch(&u->ctx, &d.header);
if (r != FFX_API_RETURN_OK) {
static int told;