pbkit: Enable flip stall; expose ramin, print_char, depth buffer internals #628

antangelo · 2023-03-02T02:56:29Z

This PR upstreams the pbkit changes required by the pgraph test suite.

They are all combined into one PR, but can be split it if preferred.
CC: @abaire

JayFoxRox · 2023-03-04T17:40:00Z

lib/pbkit/pbkit.c

@@ -2468,7 +2468,7 @@ int pb_finished(void)
    p=pb_push1(p,NV20_TCL_PRIMITIVE_3D_WAIT_MAKESPACE,0); //wait/makespace (obtains null status)
    p=pb_push1(p,NV20_TCL_PRIMITIVE_3D_PARAMETER_A,pb_back_index); //set param=back buffer index to show up
    p=pb_push1(p,NV20_TCL_PRIMITIVE_3D_FIRE_INTERRUPT,PB_FINISHED); //subprogID PB_FINISHED: gets frame ready to show up soon
-//  p=pb_push1(p,NV20_TCL_PRIMITIVE_3D_STALL_PIPELINE,0); //stall gpu pipeline (not sure it's needed in triple buffering technic)
+    p=pb_push1(p,NV20_TCL_PRIMITIVE_3D_STALL_PIPELINE,0); //stall gpu pipeline (not sure it's needed in triple buffering technic)


Why is this needed?

If I remember correctly it was just to mimic end of frame behavior observed in DirectX based titles in order to make tooling (like the pgraph tracer) simpler.

JayFoxRox · 2023-03-04T17:40:55Z

lib/pbkit/pbkit.c

@@ -2630,6 +2631,11 @@ void pb_set_color_format(unsigned int fmt, bool swizzled) {
    assert(swizzled == false);
 }

+void pb_set_fb_size_multiplier(unsigned int multiplier) {
+    assert(multiplier > 0);
+    pb_FBSizeMultiplier = multiplier;


When would you use this?

For fullscreen antialiasing: https://github.com/abaire/nxdk_pgraph_tests/blob/e4e1dbce47d2ee7741f4ef1b219952ad6a410f78/src/main.cpp#L154

JayFoxRox · 2023-03-04T17:41:28Z

lib/pbkit/pbkit.h

@@ -159,8 +159,6 @@ void pb_create_gr_instance(int ChannelID,
                        DWORD flags3D,
                        struct s_CtxDma *pGrObject);

-// Exposed so pb_print can be overridden using a version of vsprintf that
-// supports floats.


Why was this comment removed?

JayFoxRox · 2023-03-04T17:50:07Z

lib/pbkit/pbkit.h

@@ -144,6 +144,10 @@ void pb_create_gr_ctx(   int ChannelID,
                      struct s_CtxDma *pGrObject  );
 void pb_bind_channel(struct s_CtxDma *pCtxDmaObject);

+uint8_t *pb_depth_stencil_buffer();
+DWORD pb_depth_stencil_pitch();
+DWORD pb_depth_stencil_size();


I always thought pbkit had this interface backwards. Currently pbkit is a mix of a low-level and high-level driver.

pbkit can submit pushbuffers, but it also constructs them, some using rather high level operations.
Instead, I believe that pbkit should allow submission of pushbuffers, but it should primarily be the users responsibility to construct them.
pbkit should only control the hardware so it can process commands from the pushbuffers.

More importantly, pbkit also manages buffers internally (sometimes like a singleton), but doesn't allow the user to manually create or control these buffers.
Instead, I believe that pbkit should provide the user with buffer creation / managing functions, but the user should always explicitly create these buffers and keep track of them.

This means I'd prefer to see something like pb_create_depth_stencil_buffer(size, pitch) and then the user should be responsible for creating these buffers.
So, in an ideal world, the application would never even have to query for the buffer address / pitch / size, because it created the buffer itself.

This (explicit buffer creation by application) is also required (or at least helpful) for more advanced applications which want to switch render-targets.

Totally agree, Thrimbor also agrees and has plans for a rewrite. My fork uses this as a stopgap; not sure it's worth the effort to take a half-step in the direction of providing user control if there's a major refactor pending. I also think some of this may have been added in commits I upstreamed more recently.

JayFoxRox · 2023-03-04T17:51:40Z

Commit titles don't follow our usual style (<prefix>: <Verb> [...] without any punctuation at the end)

GXTX · 2025-07-09T12:37:03Z

Status and is this patch still relevant?

abaire

This would at least need a rebase since some of the changes were upstreamed and I ended up doing a minor refactor to break pbkit.c into multiple files for ease of maintenance. Personally I don't feel strongly that it's worth the effort given that pbkit is slated for replacement.

abaire · 2025-07-09T14:30:01Z

lib/pbkit/pbkit.c

@@ -2468,7 +2468,7 @@ int pb_finished(void)
    p=pb_push1(p,NV20_TCL_PRIMITIVE_3D_WAIT_MAKESPACE,0); //wait/makespace (obtains null status)
    p=pb_push1(p,NV20_TCL_PRIMITIVE_3D_PARAMETER_A,pb_back_index); //set param=back buffer index to show up
    p=pb_push1(p,NV20_TCL_PRIMITIVE_3D_FIRE_INTERRUPT,PB_FINISHED); //subprogID PB_FINISHED: gets frame ready to show up soon
-//  p=pb_push1(p,NV20_TCL_PRIMITIVE_3D_STALL_PIPELINE,0); //stall gpu pipeline (not sure it's needed in triple buffering technic)
+    p=pb_push1(p,NV20_TCL_PRIMITIVE_3D_STALL_PIPELINE,0); //stall gpu pipeline (not sure it's needed in triple buffering technic)


If I remember correctly it was just to mimic end of frame behavior observed in DirectX based titles in order to make tooling (like the pgraph tracer) simpler.

abaire · 2025-07-09T14:30:15Z

lib/pbkit/pbkit.c

@@ -2630,6 +2631,11 @@ void pb_set_color_format(unsigned int fmt, bool swizzled) {
    assert(swizzled == false);
 }

+void pb_set_fb_size_multiplier(unsigned int multiplier) {
+    assert(multiplier > 0);
+    pb_FBSizeMultiplier = multiplier;


For fullscreen antialiasing: https://github.com/abaire/nxdk_pgraph_tests/blob/e4e1dbce47d2ee7741f4ef1b219952ad6a410f78/src/main.cpp#L154

abaire · 2025-07-09T14:33:17Z

lib/pbkit/pbkit.h

@@ -144,6 +144,10 @@ void pb_create_gr_ctx(   int ChannelID,
                      struct s_CtxDma *pGrObject  );
 void pb_bind_channel(struct s_CtxDma *pCtxDmaObject);

+uint8_t *pb_depth_stencil_buffer();
+DWORD pb_depth_stencil_pitch();
+DWORD pb_depth_stencil_size();


Totally agree, Thrimbor also agrees and has plans for a rewrite. My fork uses this as a stopgap; not sure it's worth the effort to take a half-step in the direction of providing user control if there's a major refactor pending. I also think some of this may have been added in commits I upstreamed more recently.

abaire and others added 8 commits February 26, 2023 20:09

pbkit: Expose ramin context registration.

ef762c8

pbkit: Expose depth buffer.

741fa13

Expose context registration.

ff765f7

Exposes pb_print_char to allow use of a full featured printf lib.

edaffee

Enables flip stall in pbkit.

02d7524

Adds sleep in pbkit exception handler to keep XBDM responsive.

83073d8

pbkit: Support setting FB size multiplier

c05399a

Fix brace formatting

ec99083

JayFoxRox reviewed Mar 4, 2023

View reviewed changes

thrimbor force-pushed the master branch from 8cc3520 to 88d18d6 Compare October 9, 2023 14:41

abaire reviewed Jul 9, 2025

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

pbkit: Enable flip stall; expose ramin, print_char, depth buffer internals #628

pbkit: Enable flip stall; expose ramin, print_char, depth buffer internals #628

Uh oh!

antangelo commented Mar 2, 2023 •

edited

Loading

Uh oh!

JayFoxRox Mar 4, 2023

Uh oh!

abaire Jul 9, 2025

Uh oh!

JayFoxRox Mar 4, 2023

Uh oh!

abaire Jul 9, 2025

Uh oh!

JayFoxRox Mar 4, 2023

Uh oh!

JayFoxRox Mar 4, 2023 •

edited

Loading

Uh oh!

abaire Jul 9, 2025

Uh oh!

JayFoxRox commented Mar 4, 2023 •

edited

Loading

Uh oh!

GXTX commented Jul 9, 2025

Uh oh!

abaire left a comment

Uh oh!

abaire Jul 9, 2025

Uh oh!

abaire Jul 9, 2025

Uh oh!

abaire Jul 9, 2025

Uh oh!

Uh oh!

pbkit: Enable flip stall; expose ramin, print_char, depth buffer internals #628

Are you sure you want to change the base?

pbkit: Enable flip stall; expose ramin, print_char, depth buffer internals #628

Uh oh!

Conversation

antangelo commented Mar 2, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

JayFoxRox Mar 4, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

JayFoxRox commented Mar 4, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

GXTX commented Jul 9, 2025

Uh oh!

abaire left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

antangelo commented Mar 2, 2023 •

edited

Loading

JayFoxRox Mar 4, 2023 •

edited

Loading

JayFoxRox commented Mar 4, 2023 •

edited

Loading