[PATCH 2/3] nv50: fix viewport transform

[Christoph Bumiller]

We need to invert the viewport translate/scale parameters
when the state tracker thinks we have Y_0_TOP.

If these cases, we have do to bypass mode by setting an
identity viewport transform for x, z and inversion for y,
or p.e. clear_with_quad won't work correctly.
Clipping for xy in NDC space needs to be disabled then.
---
 src/gallium/drivers/nv50/nv50_context.h        |    1 +
 src/gallium/drivers/nv50/nv50_state_validate.c |   55 +++++++++++++++++-------
 2 files changed, 40 insertions(+), 16 deletions(-)

diff --git a/src/gallium/drivers/nv50/nv50_context.h b/src/gallium/drivers/nv50/nv50_context.h
index 9b8cc4d..f66e0f2 100644
--- a/src/gallium/drivers/nv50/nv50_context.h
+++ b/src/gallium/drivers/nv50/nv50_context.h
@@ -140,6 +140,7 @@ struct nv50_context {
 	struct pipe_scissor_state scissor;
 	struct pipe_viewport_state viewport;
 	struct pipe_framebuffer_state framebuffer;
+	int fb_invert_y;
 	struct nv50_program *vertprog;
 	struct nv50_program *fragprog;
 	struct pipe_buffer *constbuf[PIPE_SHADER_TYPES];
diff --git a/src/gallium/drivers/nv50/nv50_state_validate.c b/src/gallium/drivers/nv50/nv50_state_validate.c
index 4981b1b..b65301b 100644
--- a/src/gallium/drivers/nv50/nv50_state_validate.c
+++ b/src/gallium/drivers/nv50/nv50_state_validate.c
@@ -31,6 +31,11 @@ nv50_state_validate_fb(struct nv50_context *nv50)
 	struct pipe_framebuffer_state *fb = &nv50->framebuffer;
 	unsigned i, w, h, gw = 0;
 
+	nv50->fb_invert_y = 0;
+	if (fb->nr_cbufs && (fb->cbufs[0]->texture->tex_usage &
+			     PIPE_TEXTURE_USAGE_DISPLAY_TARGET))
+		nv50->fb_invert_y = fb->cbufs[0]->height;
+
 	for (i = 0; i < fb->nr_cbufs; i++) {
 		struct pipe_texture *pt = fb->cbufs[i]->texture;
 		struct nouveau_bo *bo = nv50_miptree(pt)->bo;
@@ -263,6 +268,7 @@ scissor_uptodate:
 
 	if (nv50->dirty & (NV50_NEW_VIEWPORT | NV50_NEW_RASTERIZER)) {
 		unsigned bypass;
+		struct pipe_viewport_state viewp;
 
 		if (!nv50->rasterizer->pipe.bypass_vs_clip_and_viewport)
 			bypass = 0;
@@ -277,25 +283,42 @@ scissor_uptodate:
 
 		so = so_new(12, 0);
 		if (!bypass) {
-			so_method(so, tesla, NV50TCL_VIEWPORT_UNK1(0), 3);
-			so_data  (so, fui(nv50->viewport.translate[0]));
-			so_data  (so, fui(nv50->viewport.translate[1]));
-			so_data  (so, fui(nv50->viewport.translate[2]));
-			so_method(so, tesla, NV50TCL_VIEWPORT_UNK0(0), 3);
-			so_data  (so, fui(nv50->viewport.scale[0]));
-			so_data  (so, fui(-nv50->viewport.scale[1]));
-			so_data  (so, fui(nv50->viewport.scale[2]));
-			so_method(so, tesla, 0x192c, 1);
-			so_data  (so, 1);
-			so_method(so, tesla, 0x0f90, 1);
-			so_data  (so, 0);
+			memcpy(&viewp, &nv50->viewport, sizeof(viewp));
+
+			if (nv50->fb_invert_y) {
+				viewp.scale[1] = -nv50->viewport.scale[1];
+				viewp.translate[1] = (float)nv50->fb_invert_y -
+					nv50->viewport.translate[1];
+			}
+			so_method(so, tesla, 0x193c, 1);
+			so_data  (so, 0x0000);
 		} else {
-			so_method(so, tesla, 0x192c, 1);
-			so_data  (so, 0);
-			so_method(so, tesla, 0x0f90, 1);
-			so_data  (so, 1);
+			viewp.scale[0] = viewp.scale[2] = 1.0f;
+			viewp.translate[0] = viewp.translate[2] = 0.0f;
+
+			if (nv50->fb_invert_y) {
+				viewp.scale[1] = -1.0f;
+				viewp.translate[1] = (float)nv50->fb_invert_y;
+			} else {
+				viewp.scale[1] = 1.0f;
+				viewp.translate[1] = 0.0f;
+			}
+			/* don't do xy-clipping in NDC space */
+			so_method(so, tesla, 0x193c, 1);
+			so_data  (so, 0x0800);
 		}
 
+		so_method(so, tesla, 0x192c, 1);
+		so_data  (so, 1);
+		so_method(so, tesla, NV50TCL_VIEWPORT_UNK1(0), 3);
+		so_data  (so, fui(viewp.translate[0]));
+		so_data  (so, fui(viewp.translate[1]));
+		so_data  (so, fui(viewp.translate[2]));
+		so_method(so, tesla, NV50TCL_VIEWPORT_UNK0(0), 3);
+		so_data  (so, fui(viewp.scale[0]));
+		so_data  (so, fui(viewp.scale[1]));
+		so_data  (so, fui(viewp.scale[2]));
+
 		so_ref(so, &nv50->state.viewport);
 		so_ref(NULL, &so);
 		nv50->state.dirty |= NV50_NEW_VIEWPORT;
-- 
1.6.3.3


--------------050205000506050205030809
Content-Type: text/plain;
 name="0003-nv50-invert-y-coordinates-in-transfers-if-needed.patch"
Content-Transfer-Encoding: 7bit
Content-Disposition: inline;
 filename*0="0003-nv50-invert-y-coordinates-in-transfers-if-needed.patch"