nv30, nv40: non-trivially unify nv[34]0_state_fb.c
authorLuca Barbieri <luca@luca-barbieri.com>
Sat, 20 Feb 2010 23:59:30 +0000 (00:59 +0100)
committerYounes Manton <younes.m@gmail.com>
Mon, 15 Mar 2010 04:03:02 +0000 (00:03 -0400)
The files are significantly different due to:
1. nv30 support 2 render targets, nv40 4
2. z-buffer pitch is set differently
3. nv30 has a limitation of colour_bits >= zeta_bits. This may not
   actually exist in the driver though
4. nv30 points color0 at depth in the depth-only case
5. nv30 sets NV34TCL_VIEWPORT_TX_ORIGIN to 0. This is probably
   unnecessary

This patch attempts to unify the two files and preserve the existing
behavior.

src/gallium/drivers/nv30/Makefile
src/gallium/drivers/nv30/nv30_context.h
src/gallium/drivers/nv40/Makefile
src/gallium/drivers/nv40/nv40_context.h
src/gallium/drivers/nv40/nv40_state_fb.c [deleted file]
src/gallium/drivers/nvfx/Makefile
src/gallium/drivers/nvfx/nvfx_context.h
src/gallium/drivers/nvfx/nvfx_state_emit.c
src/gallium/drivers/nvfx/nvfx_state_fb.c [moved from src/gallium/drivers/nv30/nv30_state_fb.c with 53% similarity]

index 6ee0d8f..b5728a3 100644 (file)
@@ -10,7 +10,6 @@ C_SOURCES = \
        nv30_fragtex.c \
        nv30_screen.c \
        nv30_state.c \
-       nv30_state_fb.c \
        nv30_vbo.c \
        nv30_vertprog.c
 
index 37def47..c203425 100644 (file)
@@ -22,7 +22,6 @@ extern void nv30_fragtex_bind(struct nvfx_context *);
 /* nv30_state.c and friends */
 extern struct nvfx_state_entry nv30_state_fragprog;
 extern struct nvfx_state_entry nv30_state_vertprog;
-extern struct nvfx_state_entry nv30_state_framebuffer;
 extern struct nvfx_state_entry nv30_state_fragtex;
 extern struct nvfx_state_entry nv30_state_vbo;
 
index 85d6ce9..f78e81a 100644 (file)
@@ -10,7 +10,6 @@ C_SOURCES = \
        nv40_fragtex.c \
        nv40_screen.c \
        nv40_state.c \
-       nv40_state_fb.c \
        nv40_vbo.c \
        nv40_vertprog.c
 
index a3000ee..3840134 100644 (file)
@@ -26,7 +26,6 @@ extern void nv40_fragtex_bind(struct nvfx_context *);
 /* nv40_state.c and friends */
 extern struct nvfx_state_entry nv40_state_fragprog;
 extern struct nvfx_state_entry nv40_state_vertprog;
-extern struct nvfx_state_entry nv40_state_framebuffer;
 extern struct nvfx_state_entry nv40_state_fragtex;
 extern struct nvfx_state_entry nv40_state_vbo;
 extern struct nvfx_state_entry nv40_state_vtxfmt;
diff --git a/src/gallium/drivers/nv40/nv40_state_fb.c b/src/gallium/drivers/nv40/nv40_state_fb.c
deleted file mode 100644 (file)
index 95735e4..0000000
+++ /dev/null
@@ -1,175 +0,0 @@
-#include "nv40_context.h"
-#include "nouveau/nouveau_util.h"
-
-static struct pipe_buffer *
-nv40_do_surface_buffer(struct pipe_surface *surface)
-{
-       struct nvfx_miptree *mt = (struct nvfx_miptree *)surface->texture;
-       return mt->buffer;
-}
-
-#define nv40_surface_buffer(ps) nouveau_bo(nv40_do_surface_buffer(ps))
-
-static boolean
-nv40_state_framebuffer_validate(struct nvfx_context *nvfx)
-{
-       struct nouveau_channel *chan = nvfx->screen->base.channel;
-       struct nouveau_grobj *eng3d = nvfx->screen->eng3d;
-       struct pipe_framebuffer_state *fb = &nvfx->framebuffer;
-       struct nv04_surface *rt[4], *zeta;
-       uint32_t rt_enable, rt_format;
-       int i, colour_format = 0, zeta_format = 0;
-       struct nouveau_stateobj *so = so_new(18, 24, 10);
-       unsigned rt_flags = NOUVEAU_BO_RDWR | NOUVEAU_BO_VRAM;
-       unsigned w = fb->width;
-       unsigned h = fb->height;
-
-       rt_enable = 0;
-       for (i = 0; i < fb->nr_cbufs; i++) {
-               if (colour_format) {
-                       assert(colour_format == fb->cbufs[i]->format);
-               } else {
-                       colour_format = fb->cbufs[i]->format;
-                       rt_enable |= (NV40TCL_RT_ENABLE_COLOR0 << i);
-                       rt[i] = (struct nv04_surface *)fb->cbufs[i];
-               }
-       }
-
-       if (rt_enable & (NV40TCL_RT_ENABLE_COLOR1 | NV40TCL_RT_ENABLE_COLOR2 |
-                        NV40TCL_RT_ENABLE_COLOR3))
-               rt_enable |= NV40TCL_RT_ENABLE_MRT;
-
-       if (fb->zsbuf) {
-               zeta_format = fb->zsbuf->format;
-               zeta = (struct nv04_surface *)fb->zsbuf;
-       }
-
-       if (!(rt[0]->base.texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) {
-               assert(!(fb->width & (fb->width - 1)) && !(fb->height & (fb->height - 1)));
-               for (i = 1; i < fb->nr_cbufs; i++)
-                       assert(!(rt[i]->base.texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR));
-
-               rt_format = NV34TCL_RT_FORMAT_TYPE_SWIZZLED |
-                           log2i(fb->width) << NV34TCL_RT_FORMAT_LOG2_WIDTH_SHIFT |
-                           log2i(fb->height) << NV34TCL_RT_FORMAT_LOG2_HEIGHT_SHIFT;
-       }
-       else
-               rt_format = NV34TCL_RT_FORMAT_TYPE_LINEAR;
-
-       switch (colour_format) {
-       case PIPE_FORMAT_B8G8R8X8_UNORM:
-               rt_format |= NV34TCL_RT_FORMAT_COLOR_X8R8G8B8;
-               break;
-       case PIPE_FORMAT_B8G8R8A8_UNORM:
-       case 0:
-               rt_format |= NV34TCL_RT_FORMAT_COLOR_A8R8G8B8;
-               break;
-       case PIPE_FORMAT_B5G6R5_UNORM:
-               rt_format |= NV34TCL_RT_FORMAT_COLOR_R5G6B5;
-               break;
-       default:
-               assert(0);
-       }
-
-       switch (zeta_format) {
-       case PIPE_FORMAT_Z16_UNORM:
-               rt_format |= NV34TCL_RT_FORMAT_ZETA_Z16;
-               break;
-       case PIPE_FORMAT_S8Z24_UNORM:
-       case PIPE_FORMAT_X8Z24_UNORM:
-       case 0:
-               rt_format |= NV34TCL_RT_FORMAT_ZETA_Z24S8;
-               break;
-       default:
-               assert(0);
-       }
-
-       if (rt_enable & NV40TCL_RT_ENABLE_COLOR0) {
-               so_method(so, eng3d, NV34TCL_DMA_COLOR0, 1);
-               so_reloc (so, nv40_surface_buffer(&rt[0]->base), 0,
-                             rt_flags | NOUVEAU_BO_OR,
-                             chan->vram->handle, chan->gart->handle);
-               so_method(so, eng3d, NV34TCL_COLOR0_PITCH, 2);
-               so_data  (so, rt[0]->pitch);
-               so_reloc (so, nv40_surface_buffer(&rt[0]->base),
-                             rt[0]->base.offset, rt_flags | NOUVEAU_BO_LOW,
-                             0, 0);
-       }
-
-       if (rt_enable & NV40TCL_RT_ENABLE_COLOR1) {
-               so_method(so, eng3d, NV34TCL_DMA_COLOR1, 1);
-               so_reloc (so, nv40_surface_buffer(&rt[1]->base), 0,
-                             rt_flags | NOUVEAU_BO_OR,
-                             chan->vram->handle, chan->gart->handle);
-               so_method(so, eng3d, NV34TCL_COLOR1_OFFSET, 2);
-               so_reloc (so, nv40_surface_buffer(&rt[1]->base),
-                             rt[1]->base.offset, rt_flags | NOUVEAU_BO_LOW,
-                             0, 0);
-               so_data  (so, rt[1]->pitch);
-       }
-
-       if (rt_enable & NV40TCL_RT_ENABLE_COLOR2) {
-               so_method(so, eng3d, NV40TCL_DMA_COLOR2, 1);
-               so_reloc (so, nv40_surface_buffer(&rt[2]->base), 0,
-                             rt_flags | NOUVEAU_BO_OR,
-                             chan->vram->handle, chan->gart->handle);
-               so_method(so, eng3d, NV40TCL_COLOR2_OFFSET, 1);
-               so_reloc (so, nv40_surface_buffer(&rt[2]->base),
-                             rt[2]->base.offset, rt_flags | NOUVEAU_BO_LOW,
-                             0, 0);
-               so_method(so, eng3d, NV40TCL_COLOR2_PITCH, 1);
-               so_data  (so, rt[2]->pitch);
-       }
-
-       if (rt_enable & NV40TCL_RT_ENABLE_COLOR3) {
-               so_method(so, eng3d, NV40TCL_DMA_COLOR3, 1);
-               so_reloc (so, nv40_surface_buffer(&rt[3]->base), 0,
-                             rt_flags | NOUVEAU_BO_OR,
-                             chan->vram->handle, chan->gart->handle);
-               so_method(so, eng3d, NV40TCL_COLOR3_OFFSET, 1);
-               so_reloc (so, nv40_surface_buffer(&rt[3]->base),
-                             rt[3]->base.offset, rt_flags | NOUVEAU_BO_LOW,
-                             0, 0);
-               so_method(so, eng3d, NV40TCL_COLOR3_PITCH, 1);
-               so_data  (so, rt[3]->pitch);
-       }
-
-       if (zeta_format) {
-               so_method(so, eng3d, NV34TCL_DMA_ZETA, 1);
-               so_reloc (so, nv40_surface_buffer(&zeta->base), 0,
-                             rt_flags | NOUVEAU_BO_OR,
-                             chan->vram->handle, chan->gart->handle);
-               so_method(so, eng3d, NV34TCL_ZETA_OFFSET, 1);
-               so_reloc (so, nv40_surface_buffer(&zeta->base),
-                             zeta->base.offset, rt_flags | NOUVEAU_BO_LOW, 0, 0);
-               so_method(so, eng3d, NV40TCL_ZETA_PITCH, 1);
-               so_data  (so, zeta->pitch);
-       }
-
-       so_method(so, eng3d, NV40TCL_RT_ENABLE, 1);
-       so_data  (so, rt_enable);
-       so_method(so, eng3d, NV34TCL_RT_HORIZ, 3);
-       so_data  (so, (w << 16) | 0);
-       so_data  (so, (h << 16) | 0);
-       so_data  (so, rt_format);
-       so_method(so, eng3d, NV34TCL_VIEWPORT_HORIZ, 2);
-       so_data  (so, (w << 16) | 0);
-       so_data  (so, (h << 16) | 0);
-       so_method(so, eng3d, NV34TCL_VIEWPORT_CLIP_HORIZ(0), 2);
-       so_data  (so, ((w - 1) << 16) | 0);
-       so_data  (so, ((h - 1) << 16) | 0);
-       so_method(so, eng3d, 0x1d88, 1);
-       so_data  (so, (1 << 12) | h);
-
-       so_ref(so, &nvfx->state.hw[NVFX_STATE_FB]);
-       so_ref(NULL, &so);
-       return TRUE;
-}
-
-struct nvfx_state_entry nv40_state_framebuffer = {
-       .validate = nv40_state_framebuffer_validate,
-       .dirty = {
-               .pipe = NVFX_NEW_FB,
-               .hw = NVFX_STATE_FB
-       }
-};
index 2f198d9..2f80681 100644 (file)
@@ -9,6 +9,7 @@ C_SOURCES = \
        nvfx_miptree.c \
        nvfx_query.c \
        nvfx_state_blend.c \
+       nvfx_state_fb.c \
        nvfx_state_rasterizer.c \
        nvfx_state_scissor.c \
         nvfx_state_stipple.c \
index b6e6cf8..28daa1d 100644 (file)
@@ -183,6 +183,7 @@ struct nvfx_state_entry {
 
 extern struct nvfx_state_entry nvfx_state_blend;
 extern struct nvfx_state_entry nvfx_state_blend_colour;
+extern struct nvfx_state_entry nvfx_state_framebuffer;
 extern struct nvfx_state_entry nvfx_state_rasterizer;
 extern struct nvfx_state_entry nvfx_state_scissor;
 extern struct nvfx_state_entry nvfx_state_sr;
index b86cb44..d3088e4 100644 (file)
@@ -5,7 +5,7 @@
 
 #define RENDER_STATES(name, nvxx, vbo) \
 static struct nvfx_state_entry *name##_render_states[] = { \
-       &nvxx##_state_framebuffer, \
+       &nvfx_state_framebuffer, \
        &nvfx_state_rasterizer, \
        &nvfx_state_scissor, \
        &nvfx_state_stipple, \
similarity index 53%
rename from src/gallium/drivers/nv30/nv30_state_fb.c
rename to src/gallium/drivers/nvfx/nvfx_state_fb.c
index e9e215d..dd64ba4 100644 (file)
@@ -1,22 +1,36 @@
-#include "nv30_context.h"
+#include "nvfx_context.h"
 #include "nouveau/nouveau_util.h"
 
+static struct pipe_buffer *
+nvfx_do_surface_buffer(struct pipe_surface *surface)
+{
+       struct nvfx_miptree *mt = (struct nvfx_miptree *)surface->texture;
+       return mt->buffer;
+}
+
+#define nvfx_surface_buffer(ps) nouveau_bo(nvfx_do_surface_buffer(ps))
+
 static boolean
-nv30_state_framebuffer_validate(struct nvfx_context *nvfx)
+nvfx_state_framebuffer_validate(struct nvfx_context *nvfx)
 {
        struct pipe_framebuffer_state *fb = &nvfx->framebuffer;
        struct nouveau_channel *chan = nvfx->screen->base.channel;
        struct nouveau_grobj *eng3d = nvfx->screen->eng3d;
-       struct nv04_surface *rt[2], *zeta = NULL;
+       struct nv04_surface *rt[4], *zeta = NULL;
        uint32_t rt_enable = 0, rt_format = 0;
-       int i, colour_format = 0, zeta_format = 0, depth_only = 0;
-       struct nouveau_stateobj *so = so_new(12, 18, 10);
+       int i, colour_format = 0, zeta_format = 0;
+       int depth_only = 0;
+       struct nouveau_stateobj *so = so_new(18, 24, 10);
        unsigned rt_flags = NOUVEAU_BO_RDWR | NOUVEAU_BO_VRAM;
        unsigned w = fb->width;
        unsigned h = fb->height;
-       struct nvfx_miptree *nv30mt;
        int colour_bits = 32, zeta_bits = 32;
 
+       if(!nvfx->is_nv4x)
+               assert(fb->nr_cbufs <= 2);
+       else
+               assert(fb->nr_cbufs <= 4);
+
        for (i = 0; i < fb->nr_cbufs; i++) {
                if (colour_format) {
                        assert(colour_format == fb->cbufs[i]->format);
@@ -27,7 +41,8 @@ nv30_state_framebuffer_validate(struct nvfx_context *nvfx)
                }
        }
 
-       if (rt_enable & NV34TCL_RT_ENABLE_COLOR1)
+       if (rt_enable & (NV34TCL_RT_ENABLE_COLOR1 |
+                        NV40TCL_RT_ENABLE_COLOR2 | NV40TCL_RT_ENABLE_COLOR3))
                rt_enable |= NV34TCL_RT_ENABLE_MRT;
 
        if (fb->zsbuf) {
@@ -35,7 +50,8 @@ nv30_state_framebuffer_validate(struct nvfx_context *nvfx)
                zeta = (struct nv04_surface *)fb->zsbuf;
        }
 
-       if (rt_enable & (NV34TCL_RT_ENABLE_COLOR0|NV34TCL_RT_ENABLE_COLOR1)) {
+       if (rt_enable & (NV34TCL_RT_ENABLE_COLOR0 | NV34TCL_RT_ENABLE_COLOR1 |
+               NV40TCL_RT_ENABLE_COLOR2 | NV40TCL_RT_ENABLE_COLOR3)) {
                /* Render to at least a colour buffer */
                if (!(rt[0]->base.texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) {
                        assert(!(fb->width & (fb->width - 1)) && !(fb->height & (fb->height - 1)));
@@ -95,50 +111,91 @@ nv30_state_framebuffer_validate(struct nvfx_context *nvfx)
                assert(0);
        }
 
-       if (colour_bits > zeta_bits) {
+       if ((!nvfx->is_nv4x) && colour_bits > zeta_bits) {
+               /* TODO: does this limitation really exist?
+                  TODO: can it be worked around somehow? */
                return FALSE;
        }
 
-       if (depth_only || (rt_enable & NV34TCL_RT_ENABLE_COLOR0)) {
+       if ((rt_enable & NV34TCL_RT_ENABLE_COLOR0)
+               || ((!nvfx->is_nv4x) && depth_only)) {
                struct nv04_surface *rt0 = (depth_only ? zeta : rt[0]);
                uint32_t pitch = rt0->pitch;
 
-               if (zeta) {
-                       pitch |= (zeta->pitch << 16);
-               } else {
-                       pitch |= (pitch << 16);
+               if(!nvfx->is_nv4x)
+               {
+                       if (zeta) {
+                               pitch |= (zeta->pitch << 16);
+                       } else {
+                               pitch |= (pitch << 16);
+                       }
                }
 
-               nv30mt = (struct nvfx_miptree *) rt0->base.texture;
                so_method(so, eng3d, NV34TCL_DMA_COLOR0, 1);
-               so_reloc (so, nouveau_bo(nv30mt->buffer), 0, rt_flags | NOUVEAU_BO_OR,
+               so_reloc (so, nvfx_surface_buffer(&rt0->base), 0,
+                             rt_flags | NOUVEAU_BO_OR,
                              chan->vram->handle, chan->gart->handle);
                so_method(so, eng3d, NV34TCL_COLOR0_PITCH, 2);
                so_data  (so, pitch);
-               so_reloc (so, nouveau_bo(nv30mt->buffer), rt0->base.offset,
-                             rt_flags | NOUVEAU_BO_LOW, 0, 0);
+               so_reloc (so, nvfx_surface_buffer(&rt[0]->base),
+                             rt0->base.offset, rt_flags | NOUVEAU_BO_LOW,
+                             0, 0);
        }
 
        if (rt_enable & NV34TCL_RT_ENABLE_COLOR1) {
-               nv30mt = (struct nvfx_miptree *)rt[1]->base.texture;
                so_method(so, eng3d, NV34TCL_DMA_COLOR1, 1);
-               so_reloc (so, nouveau_bo(nv30mt->buffer), 0, rt_flags | NOUVEAU_BO_OR,
+               so_reloc (so, nvfx_surface_buffer(&rt[1]->base), 0,
+                             rt_flags | NOUVEAU_BO_OR,
                              chan->vram->handle, chan->gart->handle);
                so_method(so, eng3d, NV34TCL_COLOR1_OFFSET, 2);
-               so_reloc (so, nouveau_bo(nv30mt->buffer), rt[1]->base.offset,
-                             rt_flags | NOUVEAU_BO_LOW, 0, 0);
+               so_reloc (so, nvfx_surface_buffer(&rt[1]->base),
+                             rt[1]->base.offset, rt_flags | NOUVEAU_BO_LOW,
+                             0, 0);
                so_data  (so, rt[1]->pitch);
        }
 
+       if(nvfx->is_nv4x)
+       {
+               if (rt_enable & NV40TCL_RT_ENABLE_COLOR2) {
+                       so_method(so, eng3d, NV40TCL_DMA_COLOR2, 1);
+                       so_reloc (so, nvfx_surface_buffer(&rt[2]->base), 0,
+                                     rt_flags | NOUVEAU_BO_OR,
+                                     chan->vram->handle, chan->gart->handle);
+                       so_method(so, eng3d, NV40TCL_COLOR2_OFFSET, 1);
+                       so_reloc (so, nvfx_surface_buffer(&rt[2]->base),
+                                     rt[2]->base.offset, rt_flags | NOUVEAU_BO_LOW,
+                                     0, 0);
+                       so_method(so, eng3d, NV40TCL_COLOR2_PITCH, 1);
+                       so_data  (so, rt[2]->pitch);
+               }
+
+               if (rt_enable & NV40TCL_RT_ENABLE_COLOR3) {
+                       so_method(so, eng3d, NV40TCL_DMA_COLOR3, 1);
+                       so_reloc (so, nvfx_surface_buffer(&rt[3]->base), 0,
+                                     rt_flags | NOUVEAU_BO_OR,
+                                     chan->vram->handle, chan->gart->handle);
+                       so_method(so, eng3d, NV40TCL_COLOR3_OFFSET, 1);
+                       so_reloc (so, nvfx_surface_buffer(&rt[3]->base),
+                                     rt[3]->base.offset, rt_flags | NOUVEAU_BO_LOW,
+                                     0, 0);
+                       so_method(so, eng3d, NV40TCL_COLOR3_PITCH, 1);
+                       so_data  (so, rt[3]->pitch);
+               }
+       }
+
        if (zeta_format) {
-               nv30mt = (struct nvfx_miptree *)zeta->base.texture;
                so_method(so, eng3d, NV34TCL_DMA_ZETA, 1);
-               so_reloc (so, nouveau_bo(nv30mt->buffer), 0, rt_flags | NOUVEAU_BO_OR,
+               so_reloc (so, nvfx_surface_buffer(&zeta->base), 0,
+                             rt_flags | NOUVEAU_BO_OR,
                              chan->vram->handle, chan->gart->handle);
                so_method(so, eng3d, NV34TCL_ZETA_OFFSET, 1);
-               so_reloc (so, nouveau_bo(nv30mt->buffer), zeta->base.offset,
-                             rt_flags | NOUVEAU_BO_LOW, 0, 0);
-               /* TODO: allocate LMA depth buffer */
+               /* TODO: reverse engineer LMA */
+               so_reloc (so, nvfx_surface_buffer(&zeta->base),
+                             zeta->base.offset, rt_flags | NOUVEAU_BO_LOW, 0, 0);
+               if(nvfx->is_nv4x) {
+                       so_method(so, eng3d, NV40TCL_ZETA_PITCH, 1);
+                       so_data  (so, zeta->pitch);
+               }
        }
 
        so_method(so, eng3d, NV34TCL_RT_ENABLE, 1);
@@ -155,17 +212,21 @@ nv30_state_framebuffer_validate(struct nvfx_context *nvfx)
        so_data  (so, ((h - 1) << 16) | 0);
        so_method(so, eng3d, 0x1d88, 1);
        so_data  (so, (1 << 12) | h);
-       /* Wonder why this is needed, context should all be set to zero on init */
-       so_method(so, eng3d, NV34TCL_VIEWPORT_TX_ORIGIN, 1);
-       so_data  (so, 0);
+
+       if(!nvfx->is_nv4x) {
+               /* Wonder why this is needed, context should all be set to zero on init */
+               /* TODO: we can most likely remove this, after putting it in context init */
+               so_method(so, eng3d, NV34TCL_VIEWPORT_TX_ORIGIN, 1);
+               so_data  (so, 0);
+       }
 
        so_ref(so, &nvfx->state.hw[NVFX_STATE_FB]);
        so_ref(NULL, &so);
        return TRUE;
 }
 
-struct nvfx_state_entry nv30_state_framebuffer = {
-       .validate = nv30_state_framebuffer_validate,
+struct nvfx_state_entry nvfx_state_framebuffer = {
+       .validate = nvfx_state_framebuffer_validate,
        .dirty = {
                .pipe = NVFX_NEW_FB,
                .hw = NVFX_STATE_FB