[g3dvl] move zscan into shaders
[mesa.git] / src / gallium / drivers / nvfx / nvfx_state_fb.c
index 28bbd36c2e81ae0d0936b6f3014ea85f6360e51f..f9fed94044a3efa14a8eb53a78ca129dd3f07895 100644 (file)
@@ -5,21 +5,23 @@
 static inline boolean
 nvfx_surface_linear_renderable(struct pipe_surface* surf)
 {
+       /* TODO: precompute this in nvfx_surface creation */
        return (surf->texture->flags & NVFX_RESOURCE_FLAG_LINEAR)
-               && !(surf->offset & 63)
+               && !(((struct nvfx_surface*)surf)->offset & 63)
                && !(((struct nvfx_surface*)surf)->pitch & 63);
 }
 
 static inline boolean
 nvfx_surface_swizzled_renderable(struct pipe_framebuffer_state* fb, struct pipe_surface* surf)
 {
-       /* TODO: return FALSE if we have a format not supporting swizzled rendering (e.g. r8); currently those are not supported at all */
+       /* TODO: precompute this in nvfx_surface creation */
        return !((struct nvfx_miptree*)surf->texture)->linear_pitch
-               && (surf->texture->target != PIPE_TEXTURE_3D || u_minify(surf->texture->depth0, surf->level) <= 1)
-               && !(surf->offset & 127)
+               && (surf->texture->target != PIPE_TEXTURE_3D || u_minify(surf->texture->depth0, surf->u.tex.level) <= 1)
+               && !(((struct nvfx_surface*)surf)->offset & 127)
                && (surf->width == fb->width)
                && (surf->height == fb->height)
-               && !((struct nvfx_surface*)surf)->temp;
+               && !((struct nvfx_surface*)surf)->temp
+               && (surf->format == PIPE_FORMAT_B8G8R8A8_UNORM || surf->format == PIPE_FORMAT_B8G8R8X8_UNORM || surf->format == PIPE_FORMAT_B5G6R5_UNORM);
 }
 
 static boolean
@@ -29,7 +31,7 @@ nvfx_surface_get_render_target(struct pipe_surface* surf, int all_swizzled, stru
        if(!ns->temp)
        {
                target->bo = ((struct nvfx_miptree*)surf->texture)->base.bo;
-               target->offset = surf->offset;
+               target->offset = ns->offset;
                target->pitch = align(ns->pitch, 64);
                assert(target->pitch);
                return FALSE;
@@ -52,7 +54,7 @@ nvfx_framebuffer_prepare(struct nvfx_context *nvfx)
        int all_swizzled = 1;
 
        if(!nvfx->is_nv4x)
-               assert(fb->nr_cbufs <= 2);
+               assert(fb->nr_cbufs <= 1);
        else
                assert(fb->nr_cbufs <= 4);
 
@@ -94,16 +96,17 @@ nvfx_framebuffer_validate(struct nvfx_context *nvfx, unsigned prepare_result)
 {
        struct pipe_framebuffer_state *fb = &nvfx->framebuffer;
        struct nouveau_channel *chan = nvfx->screen->base.channel;
+       struct nouveau_grobj *eng3d = nvfx->screen->eng3d;
        uint32_t rt_enable, rt_format;
        int i;
        unsigned rt_flags = NOUVEAU_BO_RDWR | NOUVEAU_BO_VRAM;
        unsigned w = fb->width;
        unsigned h = fb->height;
 
-       rt_enable = (NV34TCL_RT_ENABLE_COLOR0 << fb->nr_cbufs) - 1;
-       if (rt_enable & (NV34TCL_RT_ENABLE_COLOR1 |
-                        NV40TCL_RT_ENABLE_COLOR2 | NV40TCL_RT_ENABLE_COLOR3))
-               rt_enable |= NV34TCL_RT_ENABLE_MRT;
+       rt_enable = (NV30_3D_RT_ENABLE_COLOR0 << fb->nr_cbufs) - 1;
+       if (rt_enable & (NV30_3D_RT_ENABLE_COLOR1 |
+                        NV40_3D_RT_ENABLE_COLOR2 | NV40_3D_RT_ENABLE_COLOR3))
+               rt_enable |= NV30_3D_RT_ENABLE_MRT;
 
        nvfx->state.render_temps = 0;
 
@@ -111,7 +114,9 @@ nvfx_framebuffer_validate(struct nvfx_context *nvfx, unsigned prepare_result)
                nvfx->state.render_temps |= nvfx_surface_get_render_target(fb->cbufs[i], prepare_result, &nvfx->hw_rt[i]) << i;
 
        for(; i < 4; ++i)
-               nvfx->hw_rt[i].bo = 0;
+               nvfx->hw_rt[i].bo = NULL;
+
+       nvfx->hw_zeta.bo = NULL;
 
        if (fb->zsbuf) {
                nvfx->state.render_temps |= nvfx_surface_get_render_target(fb->zsbuf, prepare_result, &nvfx->hw_zeta) << 7;
@@ -123,55 +128,69 @@ nvfx_framebuffer_validate(struct nvfx_context *nvfx, unsigned prepare_result)
        if (prepare_result) {
                assert(!(fb->width & (fb->width - 1)) && !(fb->height & (fb->height - 1)));
 
-               rt_format = NV34TCL_RT_FORMAT_TYPE_SWIZZLED |
-                       (util_logbase2(fb->width) << NV34TCL_RT_FORMAT_LOG2_WIDTH_SHIFT) |
-                       (util_logbase2(fb->height) << NV34TCL_RT_FORMAT_LOG2_HEIGHT_SHIFT);
+               rt_format = NV30_3D_RT_FORMAT_TYPE_SWIZZLED |
+                       (util_logbase2(fb->width) << NV30_3D_RT_FORMAT_LOG2_WIDTH__SHIFT) |
+                       (util_logbase2(fb->height) << NV30_3D_RT_FORMAT_LOG2_HEIGHT__SHIFT);
        } else
-               rt_format = NV34TCL_RT_FORMAT_TYPE_LINEAR;
+               rt_format = NV30_3D_RT_FORMAT_TYPE_LINEAR;
 
        if(fb->nr_cbufs > 0) {
                switch (fb->cbufs[0]->format) {
                case PIPE_FORMAT_B8G8R8X8_UNORM:
-                       rt_format |= NV34TCL_RT_FORMAT_COLOR_X8R8G8B8;
+                       rt_format |= NV30_3D_RT_FORMAT_COLOR_X8R8G8B8;
                        break;
                case PIPE_FORMAT_B8G8R8A8_UNORM:
                case 0:
-                       rt_format |= NV34TCL_RT_FORMAT_COLOR_A8R8G8B8;
+                       rt_format |= NV30_3D_RT_FORMAT_COLOR_A8R8G8B8;
+                       break;
+               case PIPE_FORMAT_R8G8B8X8_UNORM:
+                       rt_format |= NV30_3D_RT_FORMAT_COLOR_X8B8G8R8;
+                       break;
+               case PIPE_FORMAT_R8G8B8A8_UNORM:
+                       rt_format |= NV30_3D_RT_FORMAT_COLOR_A8B8G8R8;
                        break;
                case PIPE_FORMAT_B5G6R5_UNORM:
-                       rt_format |= NV34TCL_RT_FORMAT_COLOR_R5G6B5;
+                       rt_format |= NV30_3D_RT_FORMAT_COLOR_R5G6B5;
+                       break;
+               case PIPE_FORMAT_R32G32B32A32_FLOAT:
+                       rt_format |= NV30_3D_RT_FORMAT_COLOR_A32B32G32R32_FLOAT;
+                       break;
+               case PIPE_FORMAT_R16G16B16A16_FLOAT:
+                       rt_format |= NV30_3D_RT_FORMAT_COLOR_A16B16G16R16_FLOAT;
                        break;
                default:
                        assert(0);
                }
        } else if(fb->zsbuf && util_format_get_blocksize(fb->zsbuf->format) == 2)
-               rt_format |= NV34TCL_RT_FORMAT_COLOR_R5G6B5;
+               rt_format |= NV30_3D_RT_FORMAT_COLOR_R5G6B5;
        else
-               rt_format |= NV34TCL_RT_FORMAT_COLOR_A8R8G8B8;
+               rt_format |= NV30_3D_RT_FORMAT_COLOR_A8R8G8B8;
 
        if(fb->zsbuf) {
                switch (fb->zsbuf->format) {
                case PIPE_FORMAT_Z16_UNORM:
-                       rt_format |= NV34TCL_RT_FORMAT_ZETA_Z16;
+                       rt_format |= NV30_3D_RT_FORMAT_ZETA_Z16;
                        break;
                case PIPE_FORMAT_S8_USCALED_Z24_UNORM:
                case PIPE_FORMAT_X8Z24_UNORM:
                case 0:
-                       rt_format |= NV34TCL_RT_FORMAT_ZETA_Z24S8;
+                       rt_format |= NV30_3D_RT_FORMAT_ZETA_Z24S8;
                        break;
                default:
                        assert(0);
                }
        } else if(fb->nr_cbufs && util_format_get_blocksize(fb->cbufs[0]->format) == 2)
-               rt_format |= NV34TCL_RT_FORMAT_ZETA_Z16;
+               rt_format |= NV30_3D_RT_FORMAT_ZETA_Z16;
        else
-               rt_format |= NV34TCL_RT_FORMAT_ZETA_Z24S8;
+               rt_format |= NV30_3D_RT_FORMAT_ZETA_Z24S8;
+
+       MARK_RING(chan, 42, 10);
 
-       if ((rt_enable & NV34TCL_RT_ENABLE_COLOR0) || fb->zsbuf) {
+       if ((rt_enable & NV30_3D_RT_ENABLE_COLOR0) || fb->zsbuf) {
                struct nvfx_render_target *rt0 = &nvfx->hw_rt[0];
                uint32_t pitch;
 
-               if(!(rt_enable & NV34TCL_RT_ENABLE_COLOR0))
+               if(!(rt_enable & NV30_3D_RT_ENABLE_COLOR0))
                        rt0 = &nvfx->hw_zeta;
 
                pitch = rt0->pitch;
@@ -186,23 +205,23 @@ nvfx_framebuffer_validate(struct nvfx_context *nvfx, unsigned prepare_result)
 
                //printf("rendering to bo %p [%i] at offset %i with pitch %i\n", rt0->bo, rt0->bo->handle, rt0->offset, pitch);
 
-               OUT_RING(chan, RING_3D(NV34TCL_DMA_COLOR0, 1));
+               BEGIN_RING(chan, eng3d, NV30_3D_DMA_COLOR0, 1);
                OUT_RELOC(chan, rt0->bo, 0,
                              rt_flags | NOUVEAU_BO_OR,
                              chan->vram->handle, chan->gart->handle);
-               OUT_RING(chan, RING_3D(NV34TCL_COLOR0_PITCH, 2));
+               BEGIN_RING(chan, eng3d, NV30_3D_COLOR0_PITCH, 2);
                OUT_RING(chan, pitch);
                OUT_RELOC(chan, rt0->bo,
                              rt0->offset, rt_flags | NOUVEAU_BO_LOW,
                              0, 0);
        }
 
-       if (rt_enable & NV34TCL_RT_ENABLE_COLOR1) {
-               OUT_RING(chan, RING_3D(NV34TCL_DMA_COLOR1, 1));
+       if (rt_enable & NV30_3D_RT_ENABLE_COLOR1) {
+               BEGIN_RING(chan, eng3d, NV30_3D_DMA_COLOR1, 1);
                OUT_RELOC(chan, nvfx->hw_rt[1].bo, 0,
                              rt_flags | NOUVEAU_BO_OR,
                              chan->vram->handle, chan->gart->handle);
-               OUT_RING(chan, RING_3D(NV34TCL_COLOR1_OFFSET, 2));
+               BEGIN_RING(chan, eng3d, NV30_3D_COLOR1_OFFSET, 2);
                OUT_RELOC(chan, nvfx->hw_rt[1].bo,
                                nvfx->hw_rt[1].offset, rt_flags | NOUVEAU_BO_LOW,
                              0, 0);
@@ -211,69 +230,72 @@ nvfx_framebuffer_validate(struct nvfx_context *nvfx, unsigned prepare_result)
 
        if(nvfx->is_nv4x)
        {
-               if (rt_enable & NV40TCL_RT_ENABLE_COLOR2) {
-                       OUT_RING(chan, RING_3D(NV40TCL_DMA_COLOR2, 1));
+               if (rt_enable & NV40_3D_RT_ENABLE_COLOR2) {
+                       BEGIN_RING(chan, eng3d, NV40_3D_DMA_COLOR2, 1);
                        OUT_RELOC(chan, nvfx->hw_rt[2].bo, 0,
                                      rt_flags | NOUVEAU_BO_OR,
                                      chan->vram->handle, chan->gart->handle);
-                       OUT_RING(chan, RING_3D(NV40TCL_COLOR2_OFFSET, 1));
+                       BEGIN_RING(chan, eng3d, NV40_3D_COLOR2_OFFSET, 1);
                        OUT_RELOC(chan, nvfx->hw_rt[2].bo,
                                      nvfx->hw_rt[2].offset, rt_flags | NOUVEAU_BO_LOW,
                                      0, 0);
-                       OUT_RING(chan, RING_3D(NV40TCL_COLOR2_PITCH, 1));
+                       BEGIN_RING(chan, eng3d, NV40_3D_COLOR2_PITCH, 1);
                        OUT_RING(chan, nvfx->hw_rt[2].pitch);
                }
 
-               if (rt_enable & NV40TCL_RT_ENABLE_COLOR3) {
-                       OUT_RING(chan, RING_3D(NV40TCL_DMA_COLOR3, 1));
+               if (rt_enable & NV40_3D_RT_ENABLE_COLOR3) {
+                       BEGIN_RING(chan, eng3d, NV40_3D_DMA_COLOR3, 1);
                        OUT_RELOC(chan, nvfx->hw_rt[3].bo, 0,
                                      rt_flags | NOUVEAU_BO_OR,
                                      chan->vram->handle, chan->gart->handle);
-                       OUT_RING(chan, RING_3D(NV40TCL_COLOR3_OFFSET, 1));
+                       BEGIN_RING(chan, eng3d, NV40_3D_COLOR3_OFFSET, 1);
                        OUT_RELOC(chan, nvfx->hw_rt[3].bo,
                                        nvfx->hw_rt[3].offset, rt_flags | NOUVEAU_BO_LOW,
                                      0, 0);
-                       OUT_RING(chan, RING_3D(NV40TCL_COLOR3_PITCH, 1));
+                       BEGIN_RING(chan, eng3d, NV40_3D_COLOR3_PITCH, 1);
                        OUT_RING(chan, nvfx->hw_rt[3].pitch);
                }
        }
 
        if (fb->zsbuf) {
-               OUT_RING(chan, RING_3D(NV34TCL_DMA_ZETA, 1));
+               BEGIN_RING(chan, eng3d, NV30_3D_DMA_ZETA, 1);
                OUT_RELOC(chan, nvfx->hw_zeta.bo, 0,
                              rt_flags | NOUVEAU_BO_OR,
                              chan->vram->handle, chan->gart->handle);
-               OUT_RING(chan, RING_3D(NV34TCL_ZETA_OFFSET, 1));
+               BEGIN_RING(chan, eng3d, NV30_3D_ZETA_OFFSET, 1);
                /* TODO: reverse engineer LMA */
                OUT_RELOC(chan, nvfx->hw_zeta.bo,
                             nvfx->hw_zeta.offset, rt_flags | NOUVEAU_BO_LOW, 0, 0);
                if(nvfx->is_nv4x) {
-                       OUT_RING(chan, RING_3D(NV40TCL_ZETA_PITCH, 1));
+                       BEGIN_RING(chan, eng3d, NV40_3D_ZETA_PITCH, 1);
                        OUT_RING(chan, nvfx->hw_zeta.pitch);
                }
        }
+       else if(nvfx->is_nv4x) {
+               BEGIN_RING(chan, eng3d, NV40_3D_ZETA_PITCH, 1);
+               OUT_RING(chan, 64);
+       }
 
-       OUT_RING(chan, RING_3D(NV34TCL_RT_ENABLE, 1));
+       BEGIN_RING(chan, eng3d, NV30_3D_RT_ENABLE, 1);
        OUT_RING(chan, rt_enable);
-       OUT_RING(chan, RING_3D(NV34TCL_RT_HORIZ, 3));
+       BEGIN_RING(chan, eng3d, NV30_3D_RT_HORIZ, 3);
        OUT_RING(chan, (w << 16) | 0);
        OUT_RING(chan, (h << 16) | 0);
        OUT_RING(chan, rt_format);
-       OUT_RING(chan, RING_3D(NV34TCL_VIEWPORT_HORIZ, 2));
+       BEGIN_RING(chan, eng3d, NV30_3D_VIEWPORT_HORIZ, 2);
        OUT_RING(chan, (w << 16) | 0);
        OUT_RING(chan, (h << 16) | 0);
-       OUT_RING(chan, RING_3D(NV34TCL_VIEWPORT_CLIP_HORIZ(0), 2));
+       BEGIN_RING(chan, eng3d, NV30_3D_VIEWPORT_CLIP_HORIZ(0), 2);
        OUT_RING(chan, ((w - 1) << 16) | 0);
        OUT_RING(chan, ((h - 1) << 16) | 0);
-       OUT_RING(chan, RING_3D(0x1d88, 1));
-       OUT_RING(chan, (1 << 12) | h);
 
        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 */
-               OUT_RING(chan, RING_3D(NV34TCL_VIEWPORT_TX_ORIGIN, 1));
+               BEGIN_RING(chan, eng3d, NV30_3D_VIEWPORT_TX_ORIGIN, 1);
                OUT_RING(chan, 0);
        }
+       nvfx->relocs_needed &=~ NVFX_RELOCATE_FRAMEBUFFER;
 }
 
 void
@@ -286,21 +308,22 @@ nvfx_framebuffer_relocate(struct nvfx_context *nvfx)
 
 #define DO_(var, pfx, name) \
        if(var.bo) { \
-               OUT_RELOC(chan, var.bo, RING_3D(pfx##TCL_DMA_##name, 1), rt_flags, 0, 0); \
+               OUT_RELOC(chan, var.bo, RING_3D(pfx##_3D_DMA_##name, 1), rt_flags, 0, 0); \
                OUT_RELOC(chan, var.bo, 0, \
                        rt_flags | NOUVEAU_BO_OR, \
                        chan->vram->handle, chan->gart->handle); \
-               OUT_RELOC(chan, var.bo, RING_3D(pfx##TCL_##name##_OFFSET, 1), rt_flags, 0, 0); \
+               OUT_RELOC(chan, var.bo, RING_3D(pfx##_3D_##name##_OFFSET, 1), rt_flags, 0, 0); \
                OUT_RELOC(chan, var.bo, \
                        var.offset, rt_flags | NOUVEAU_BO_LOW, \
                        0, 0); \
        }
 
 #define DO(pfx, num) DO_(nvfx->hw_rt[num], pfx, COLOR##num)
-       DO(NV34, 0);
-       DO(NV34, 1);
+       DO(NV30, 0);
+       DO(NV30, 1);
        DO(NV40, 2);
        DO(NV40, 3);
 
-       DO_(nvfx->hw_zeta, NV34, ZETA);
+       DO_(nvfx->hw_zeta, NV30, ZETA);
+       nvfx->relocs_needed &=~ NVFX_RELOCATE_FRAMEBUFFER;
 }