diff options
Diffstat (limited to 'src/gallium/drivers/nv50')
20 files changed, 1510 insertions, 1455 deletions
diff --git a/src/gallium/drivers/nv50/nv50_3d.xml.h b/src/gallium/drivers/nv50/nv50_3d.xml.h index 0f5a77de9a1..9dff8b2dd13 100644 --- a/src/gallium/drivers/nv50/nv50_3d.xml.h +++ b/src/gallium/drivers/nv50/nv50_3d.xml.h @@ -1,5 +1,5 @@ -#ifndef NV50_3D_XML -#define NV50_3D_XML +#ifndef RNNDB_NV50_3D_XML +#define RNNDB_NV50_3D_XML /* Autogenerated file, DO NOT EDIT manually! @@ -8,15 +8,15 @@ http://0x04.net/cgit/index.cgi/rules-ng-ng git clone git://0x04.net/rules-ng-ng The rules-ng-ng source files this header was generated from are: -- nv50_3d.xml ( 64479 bytes, from 2011-02-27 17:58:08) -- copyright.xml ( 6452 bytes, from 2010-12-15 23:45:18) -- nv_defs.xml ( 4437 bytes, from 2010-12-15 23:45:18) -- nv50_defs.xml ( 4487 bytes, from 2010-12-15 23:45:18) -- nv_3ddefs.xml ( 16394 bytes, from 2010-12-15 23:45:18) -- nv_object.xml ( 12191 bytes, from 2011-02-27 17:58:08) -- nvchipsets.xml ( 3074 bytes, from 2011-02-27 17:58:08) - -Copyright (C) 2006-2011 by the following authors: +- rnndb/nv50_3d.xml ( 65226 bytes, from 2012-01-28 13:46:30) +- ./rnndb/copyright.xml ( 6452 bytes, from 2011-08-11 18:25:12) +- ./rnndb/nv_defs.xml ( 4437 bytes, from 2011-08-11 18:25:12) +- ./rnndb/nv50_defs.xml ( 5468 bytes, from 2011-08-11 18:25:12) +- ./rnndb/nvchipsets.xml ( 3617 bytes, from 2011-08-11 18:25:12) +- ./rnndb/nv_3ddefs.xml ( 16394 bytes, from 2011-08-11 18:25:12) +- ./rnndb/nv_object.xml ( 12672 bytes, from 2011-08-11 18:25:12) + +Copyright (C) 2006-2012 by the following authors: - Artur Huillet <[email protected]> (ahuillet) - Ben Skeggs (darktama, darktama_) - B. R. <[email protected]> (koala_br) @@ -74,7 +74,7 @@ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ -#define NV50_3D_SERIALIZE 0x00000110 + #define NV50_3D_DMA_NOTIFY 0x00000180 @@ -592,9 +592,9 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #define NV50_3D_CLIPID_ADDRESS_LOW 0x00000fd0 -#define NV50_3D_MAP_SEMANTIC_5 0x00000fd4 -#define NV50_3D_MAP_SEMANTIC_5_VIEWPORT_ID__MASK 0x000000ff -#define NV50_3D_MAP_SEMANTIC_5_VIEWPORT_ID__SHIFT 0 +#define NV50_3D_SEMANTIC_VIEWPORT 0x00000fd4 +#define NV50_3D_SEMANTIC_VIEWPORT_VIEWPORT_ID__MASK 0x000000ff +#define NV50_3D_SEMANTIC_VIEWPORT_VIEWPORT_ID__SHIFT 0 #define NV50_3D_UNK0FD8 0x00000fd8 #define NV50_3D_UNK0FD8_UNK0 0x00000001 @@ -1184,7 +1184,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #define NV50_3D_BIND_TSC2__ESIZE 0x00000008 #define NV50_3D_BIND_TSC2__LEN 0x00000003 #define NV50_3D_BIND_TSC2_VALID 0x00000001 -#define NV50_3D_BIND_TSC2_SAMPLER__MASK 0x000000f0 +#define NV50_3D_BIND_TSC2_SAMPLER__MASK 0x00000010 #define NV50_3D_BIND_TSC2_SAMPLER__SHIFT 4 #define NV50_3D_BIND_TSC2_TSC__MASK 0x001ff000 #define NV50_3D_BIND_TSC2_TSC__SHIFT 12 @@ -1193,7 +1193,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #define NV50_3D_BIND_TIC2__ESIZE 0x00000008 #define NV50_3D_BIND_TIC2__LEN 0x00000003 #define NV50_3D_BIND_TIC2_VALID 0x00000001 -#define NV50_3D_BIND_TIC2_TEXTURE__MASK 0x000001fe +#define NV50_3D_BIND_TIC2_TEXTURE__MASK 0x00000002 #define NV50_3D_BIND_TIC2_TEXTURE__SHIFT 1 #define NV50_3D_BIND_TIC2_TIC__MASK 0x7ffffe00 #define NV50_3D_BIND_TIC2_TIC__SHIFT 9 @@ -1217,15 +1217,15 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #define NV50_3D_CLIPID_FILL_RECT_VERT_HIGH__MASK 0xffff0000 #define NV50_3D_CLIPID_FILL_RECT_VERT_HIGH__SHIFT 16 -#define NV50_3D_VP_CLIP_DISTANCE_ENABLE 0x00001510 -#define NV50_3D_VP_CLIP_DISTANCE_ENABLE_0 0x00000001 -#define NV50_3D_VP_CLIP_DISTANCE_ENABLE_1 0x00000002 -#define NV50_3D_VP_CLIP_DISTANCE_ENABLE_2 0x00000004 -#define NV50_3D_VP_CLIP_DISTANCE_ENABLE_3 0x00000008 -#define NV50_3D_VP_CLIP_DISTANCE_ENABLE_4 0x00000010 -#define NV50_3D_VP_CLIP_DISTANCE_ENABLE_5 0x00000020 -#define NV50_3D_VP_CLIP_DISTANCE_ENABLE_6 0x00000040 -#define NV50_3D_VP_CLIP_DISTANCE_ENABLE_7 0x00000080 +#define NV50_3D_CLIP_DISTANCE_ENABLE 0x00001510 +#define NV50_3D_CLIP_DISTANCE_ENABLE_0 0x00000001 +#define NV50_3D_CLIP_DISTANCE_ENABLE_1 0x00000002 +#define NV50_3D_CLIP_DISTANCE_ENABLE_2 0x00000004 +#define NV50_3D_CLIP_DISTANCE_ENABLE_3 0x00000008 +#define NV50_3D_CLIP_DISTANCE_ENABLE_4 0x00000010 +#define NV50_3D_CLIP_DISTANCE_ENABLE_5 0x00000020 +#define NV50_3D_CLIP_DISTANCE_ENABLE_6 0x00000040 +#define NV50_3D_CLIP_DISTANCE_ENABLE_7 0x00000080 #define NV50_3D_SAMPLECNT_ENABLE 0x00001514 @@ -1391,9 +1391,11 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #define NV50_3D_MULTISAMPLE_MODE_MS8 0x00000003 #define NV50_3D_MULTISAMPLE_MODE_MS8_ALT 0x00000004 #define NV50_3D_MULTISAMPLE_MODE_MS2_ALT 0x00000005 +#define NV50_3D_MULTISAMPLE_MODE_UNK6 0x00000006 #define NV50_3D_MULTISAMPLE_MODE_MS4_CS4 0x00000008 #define NV50_3D_MULTISAMPLE_MODE_MS4_CS12 0x00000009 #define NV50_3D_MULTISAMPLE_MODE_MS8_CS8 0x0000000a +#define NV50_3D_MULTISAMPLE_MODE_MS8_CS24 0x0000000b #define NV50_3D_VERTEX_BEGIN_D3D 0x000015d4 #define NV50_3D_VERTEX_BEGIN_D3D_PRIMITIVE__MASK 0x0fffffff @@ -1440,7 +1442,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #define NV50_3D_VERTEX_END_GL_UNK0 0x00000001 #define NVA0_3D_VERTEX_END_GL_UNK1 0x00000002 -#define NV50_3D_EDGEFLAG_ENABLE 0x000015e4 +#define NV50_3D_EDGEFLAG 0x000015e4 #define NV50_3D_VB_ELEMENT_U32 0x000015e8 @@ -1666,34 +1668,34 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #define NV50_3D_GP_VIEWPORT_ID_ENABLE 0x00001900 -#define NV50_3D_MAP_SEMANTIC_0 0x00001904 -#define NV50_3D_MAP_SEMANTIC_0_FFC0_ID__MASK 0x000000ff -#define NV50_3D_MAP_SEMANTIC_0_FFC0_ID__SHIFT 0 -#define NV50_3D_MAP_SEMANTIC_0_BFC0_ID__MASK 0x0000ff00 -#define NV50_3D_MAP_SEMANTIC_0_BFC0_ID__SHIFT 8 -#define NV50_3D_MAP_SEMANTIC_0_COLR_NR__MASK 0x00ff0000 -#define NV50_3D_MAP_SEMANTIC_0_COLR_NR__SHIFT 16 -#define NV50_3D_MAP_SEMANTIC_0_CLMP_EN 0x01000000 - -#define NV50_3D_MAP_SEMANTIC_1 0x00001908 -#define NV50_3D_MAP_SEMANTIC_1_CLIP_START__MASK 0x000000ff -#define NV50_3D_MAP_SEMANTIC_1_CLIP_START__SHIFT 0 -#define NV50_3D_MAP_SEMANTIC_1_CLIP_NUM__MASK 0x00000f00 -#define NV50_3D_MAP_SEMANTIC_1_CLIP_NUM__SHIFT 8 - -#define NV50_3D_MAP_SEMANTIC_2 0x0000190c -#define NV50_3D_MAP_SEMANTIC_2_LAYER_ID__MASK 0x000000ff -#define NV50_3D_MAP_SEMANTIC_2_LAYER_ID__SHIFT 0 - -#define NV50_3D_MAP_SEMANTIC_3 0x00001910 -#define NV50_3D_MAP_SEMANTIC_3_PTSZ_EN__MASK 0x00000001 -#define NV50_3D_MAP_SEMANTIC_3_PTSZ_EN__SHIFT 0 -#define NV50_3D_MAP_SEMANTIC_3_PTSZ_ID__MASK 0x00000ff0 -#define NV50_3D_MAP_SEMANTIC_3_PTSZ_ID__SHIFT 4 - -#define NV50_3D_MAP_SEMANTIC_4 0x00001914 -#define NV50_3D_MAP_SEMANTIC_4_PRIM_ID__MASK 0x000000ff -#define NV50_3D_MAP_SEMANTIC_4_PRIM_ID__SHIFT 0 +#define NV50_3D_SEMANTIC_COLOR 0x00001904 +#define NV50_3D_SEMANTIC_COLOR_FFC0_ID__MASK 0x000000ff +#define NV50_3D_SEMANTIC_COLOR_FFC0_ID__SHIFT 0 +#define NV50_3D_SEMANTIC_COLOR_BFC0_ID__MASK 0x0000ff00 +#define NV50_3D_SEMANTIC_COLOR_BFC0_ID__SHIFT 8 +#define NV50_3D_SEMANTIC_COLOR_COLR_NR__MASK 0x00ff0000 +#define NV50_3D_SEMANTIC_COLOR_COLR_NR__SHIFT 16 +#define NV50_3D_SEMANTIC_COLOR_CLMP_EN 0x01000000 + +#define NV50_3D_SEMANTIC_CLIP 0x00001908 +#define NV50_3D_SEMANTIC_CLIP_CLIP_START__MASK 0x000000ff +#define NV50_3D_SEMANTIC_CLIP_CLIP_START__SHIFT 0 +#define NV50_3D_SEMANTIC_CLIP_CLIP_NUM__MASK 0x00000f00 +#define NV50_3D_SEMANTIC_CLIP_CLIP_NUM__SHIFT 8 + +#define NV50_3D_SEMANTIC_LAYER 0x0000190c +#define NV50_3D_SEMANTIC_LAYER_LAYER_ID__MASK 0x000000ff +#define NV50_3D_SEMANTIC_LAYER_LAYER_ID__SHIFT 0 + +#define NV50_3D_SEMANTIC_PTSZ 0x00001910 +#define NV50_3D_SEMANTIC_PTSZ_PTSZ_EN__MASK 0x00000001 +#define NV50_3D_SEMANTIC_PTSZ_PTSZ_EN__SHIFT 0 +#define NV50_3D_SEMANTIC_PTSZ_PTSZ_ID__MASK 0x00000ff0 +#define NV50_3D_SEMANTIC_PTSZ_PTSZ_ID__SHIFT 4 + +#define NV50_3D_SEMANTIC_PRIM_ID 0x00001914 +#define NV50_3D_SEMANTIC_PRIM_ID_PRIM_ID__MASK 0x000000ff +#define NV50_3D_SEMANTIC_PRIM_ID_PRIM_ID__SHIFT 0 #define NV50_3D_CULL_FACE_ENABLE 0x00001918 @@ -1729,15 +1731,39 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #define NV50_3D_VIEW_VOLUME_CLIP_CTRL_UNK12_UNK1 0x00001000 #define NV84_3D_VIEW_VOLUME_CLIP_CTRL_UNK12_UNK2 0x00002000 -#define NV50_3D_UNK1940 0x00001940 -#define NV50_3D_UNK1940_0 0x00000001 -#define NV50_3D_UNK1940_1 0x00000010 -#define NV50_3D_UNK1940_2 0x00000100 -#define NV50_3D_UNK1940_3 0x00001000 -#define NV50_3D_UNK1940_4 0x00010000 -#define NV50_3D_UNK1940_5 0x00100000 -#define NV50_3D_UNK1940_6 0x01000000 -#define NV50_3D_UNK1940_7 0x10000000 +#define NV50_3D_CLIP_DISTANCE_MODE 0x00001940 +#define NV50_3D_CLIP_DISTANCE_MODE_0__MASK 0x00000001 +#define NV50_3D_CLIP_DISTANCE_MODE_0__SHIFT 0 +#define NV50_3D_CLIP_DISTANCE_MODE_0_CLIP 0x00000000 +#define NV50_3D_CLIP_DISTANCE_MODE_0_CULL 0x00000001 +#define NV50_3D_CLIP_DISTANCE_MODE_1__MASK 0x00000010 +#define NV50_3D_CLIP_DISTANCE_MODE_1__SHIFT 4 +#define NV50_3D_CLIP_DISTANCE_MODE_1_CLIP 0x00000000 +#define NV50_3D_CLIP_DISTANCE_MODE_1_CULL 0x00000010 +#define NV50_3D_CLIP_DISTANCE_MODE_2__MASK 0x00000100 +#define NV50_3D_CLIP_DISTANCE_MODE_2__SHIFT 8 +#define NV50_3D_CLIP_DISTANCE_MODE_2_CLIP 0x00000000 +#define NV50_3D_CLIP_DISTANCE_MODE_2_CULL 0x00000100 +#define NV50_3D_CLIP_DISTANCE_MODE_3__MASK 0x00001000 +#define NV50_3D_CLIP_DISTANCE_MODE_3__SHIFT 12 +#define NV50_3D_CLIP_DISTANCE_MODE_3_CLIP 0x00000000 +#define NV50_3D_CLIP_DISTANCE_MODE_3_CULL 0x00001000 +#define NV50_3D_CLIP_DISTANCE_MODE_4__MASK 0x00010000 +#define NV50_3D_CLIP_DISTANCE_MODE_4__SHIFT 16 +#define NV50_3D_CLIP_DISTANCE_MODE_4_CLIP 0x00000000 +#define NV50_3D_CLIP_DISTANCE_MODE_4_CULL 0x00010000 +#define NV50_3D_CLIP_DISTANCE_MODE_5__MASK 0x00100000 +#define NV50_3D_CLIP_DISTANCE_MODE_5__SHIFT 20 +#define NV50_3D_CLIP_DISTANCE_MODE_5_CLIP 0x00000000 +#define NV50_3D_CLIP_DISTANCE_MODE_5_CULL 0x00100000 +#define NV50_3D_CLIP_DISTANCE_MODE_6__MASK 0x01000000 +#define NV50_3D_CLIP_DISTANCE_MODE_6__SHIFT 24 +#define NV50_3D_CLIP_DISTANCE_MODE_6_CLIP 0x00000000 +#define NV50_3D_CLIP_DISTANCE_MODE_6_CULL 0x01000000 +#define NV50_3D_CLIP_DISTANCE_MODE_7__MASK 0x10000000 +#define NV50_3D_CLIP_DISTANCE_MODE_7__SHIFT 28 +#define NV50_3D_CLIP_DISTANCE_MODE_7_CLIP 0x00000000 +#define NV50_3D_CLIP_DISTANCE_MODE_7_CULL 0x10000000 #define NVA3_3D_UNK1944 0x00001944 @@ -2048,7 +2074,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #define NVA3_3D_IBLEND__ESIZE 0x00000020 #define NVA3_3D_IBLEND__LEN 0x00000008 -#define NVA3_3D_IBLEND_UNK00(i0) (0x00001e00 + 0x20*(i0)) +#define NVA3_3D_IBLEND_SEPARATE_ALPHA(i0) (0x00001e00 + 0x20*(i0)) #define NVA3_3D_IBLEND_EQUATION_RGB(i0) (0x00001e04 + 0x20*(i0)) #define NVA3_3D_IBLEND_EQUATION_RGB_FUNC_ADD 0x00008006 @@ -2081,4 +2107,4 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #define NVA3_3D_VERTEX_ARRAY_LIMIT_LOW_ALT__LEN 0x00000020 -#endif /* NV50_3D_XML */ +#endif /* RNNDB_NV50_3D_XML */ diff --git a/src/gallium/drivers/nv50/nv50_context.c b/src/gallium/drivers/nv50/nv50_context.c index 3b0f8f07d1c..88eb4bb246c 100644 --- a/src/gallium/drivers/nv50/nv50_context.c +++ b/src/gallium/drivers/nv50/nv50_context.c @@ -33,40 +33,36 @@ static void nv50_flush(struct pipe_context *pipe, struct pipe_fence_handle **fence) { - struct nouveau_screen *screen = &nv50_context(pipe)->screen->base; + struct nouveau_screen *screen = nouveau_screen(pipe->screen); if (fence) nouveau_fence_ref(screen->fence.current, (struct nouveau_fence **)fence); - /* Try to emit before firing to avoid having to flush again right after - * in case we have to wait on this fence. - */ - nouveau_fence_emit(screen->fence.current); - - FIRE_RING(screen->channel); + PUSH_KICK(screen->pushbuf); } static void nv50_texture_barrier(struct pipe_context *pipe) { - struct nouveau_channel *chan = nv50_context(pipe)->screen->base.channel; + struct nouveau_pushbuf *push = nv50_context(pipe)->base.pushbuf; - BEGIN_RING(chan, RING_3D(SERIALIZE), 1); - OUT_RING (chan, 0); - BEGIN_RING(chan, RING_3D(TEX_CACHE_CTL), 1); - OUT_RING (chan, 0x20); + BEGIN_NV04(push, SUBC_3D(NV50_GRAPH_SERIALIZE), 1); + PUSH_DATA (push, 0); + BEGIN_NV04(push, NV50_3D(TEX_CACHE_CTL), 1); + PUSH_DATA (push, 0x20); } void -nv50_default_flush_notify(struct nouveau_channel *chan) +nv50_default_kick_notify(struct nouveau_pushbuf *push) { - struct nv50_screen *screen = chan->user_private; - - if (!screen) - return; + struct nv50_screen *screen = push->user_priv; - nouveau_fence_update(&screen->base, TRUE); - nouveau_fence_next(&screen->base); + if (screen) { + nouveau_fence_next(&screen->base); + nouveau_fence_update(&screen->base, TRUE); + if (screen->cur_ctx) + screen->cur_ctx->state.flushed = TRUE; + } } static void @@ -74,8 +70,8 @@ nv50_context_unreference_resources(struct nv50_context *nv50) { unsigned s, i; - for (i = 0; i < NV50_BUFCTX_COUNT; ++i) - nv50_bufctx_reset(nv50, i); + nouveau_bufctx_del(&nv50->bufctx_3d); + nouveau_bufctx_del(&nv50->bufctx); for (i = 0; i < nv50->num_vtxbufs; ++i) pipe_resource_reference(&nv50->vtxbuf[i].buffer, NULL); @@ -96,13 +92,18 @@ nv50_destroy(struct pipe_context *pipe) { struct nv50_context *nv50 = nv50_context(pipe); + if (nv50_context_screen(nv50)->cur_ctx == nv50) { + nv50->base.pushbuf->kick_notify = NULL; + nv50_context_screen(nv50)->cur_ctx = NULL; + nouveau_pushbuf_bufctx(nv50->base.pushbuf, NULL); + } + /* need to flush before destroying the bufctx */ + nouveau_pushbuf_kick(nv50->base.pushbuf, nv50->base.pushbuf->channel); + nv50_context_unreference_resources(nv50); draw_destroy(nv50->draw); - if (nv50->screen->cur_ctx == nv50) - nv50->screen->cur_ctx = NULL; - FREE(nv50); } @@ -112,17 +113,28 @@ nv50_create(struct pipe_screen *pscreen, void *priv) struct nv50_screen *screen = nv50_screen(pscreen); struct nv50_context *nv50; struct pipe_context *pipe; + int ret; + uint32_t flags; nv50 = CALLOC_STRUCT(nv50_context); if (!nv50) return NULL; pipe = &nv50->base.pipe; - nv50->screen = screen; + nv50->base.pushbuf = screen->base.pushbuf; + + ret = nouveau_bufctx_new(screen->base.client, NV50_BIND_COUNT, + &nv50->bufctx_3d); + if (!ret) + ret = nouveau_bufctx_new(screen->base.client, 2, &nv50->bufctx); + if (ret) + goto out_err; + nv50->base.screen = &screen->base; nv50->base.copy_data = nv50_m2mf_copy_linear; nv50->base.push_data = nv50_sifc_linear_u8; + nv50->screen = screen; pipe->screen = pscreen; pipe->priv = priv; @@ -134,9 +146,10 @@ nv50_create(struct pipe_screen *pscreen, void *priv) pipe->flush = nv50_flush; pipe->texture_barrier = nv50_texture_barrier; - if (!screen->cur_ctx) + if (!screen->cur_ctx) { screen->cur_ctx = nv50; - screen->base.channel->flush_notify = nv50_default_flush_notify; + nouveau_pushbuf_bufctx(screen->base.pushbuf, nv50->bufctx); + } nv50_init_query_functions(nv50); nv50_init_surface_functions(nv50); @@ -149,72 +162,41 @@ nv50_create(struct pipe_screen *pscreen, void *priv) nouveau_context_init_vdec(&nv50->base); - return pipe; -} + flags = NOUVEAU_BO_VRAM | NOUVEAU_BO_RD; -struct resident { - struct nv04_resource *res; - uint32_t flags; -}; + BCTX_REFN_bo(nv50->bufctx_3d, SCREEN, flags, screen->code); + BCTX_REFN_bo(nv50->bufctx_3d, SCREEN, flags, screen->uniforms); + BCTX_REFN_bo(nv50->bufctx_3d, SCREEN, flags, screen->txc); + BCTX_REFN_bo(nv50->bufctx_3d, SCREEN, flags, screen->stack_bo); -void -nv50_bufctx_add_resident(struct nv50_context *nv50, int ctx, - struct nv04_resource *resource, uint32_t flags) -{ - struct resident rsd = { resource, flags }; + flags = NOUVEAU_BO_GART | NOUVEAU_BO_WR; - if (!resource->bo) - return; - nv50->residents_size += sizeof(struct resident); + BCTX_REFN_bo(nv50->bufctx_3d, SCREEN, flags, screen->fence.bo); + BCTX_REFN_bo(nv50->bufctx, FENCE, flags, screen->fence.bo); - /* We don't need to reference the resource here, it will be referenced - * in the context/state, and bufctx will be reset when state changes. - */ - util_dynarray_append(&nv50->residents[ctx], struct resident, rsd); -} + return pipe; -void -nv50_bufctx_del_resident(struct nv50_context *nv50, int ctx, - struct nv04_resource *resource) -{ - struct resident *rsd, *top; - unsigned i; - - for (i = 0; i < nv50->residents[ctx].size / sizeof(struct resident); ++i) { - rsd = util_dynarray_element(&nv50->residents[ctx], struct resident, i); - - if (rsd->res == resource) { - top = util_dynarray_pop_ptr(&nv50->residents[ctx], struct resident); - if (rsd != top) - *rsd = *top; - nv50->residents_size -= sizeof(struct resident); - break; - } +out_err: + if (nv50) { + if (nv50->bufctx_3d) + nouveau_bufctx_del(&nv50->bufctx_3d); + if (nv50->bufctx) + nouveau_bufctx_del(&nv50->bufctx); + FREE(nv50); } + return NULL; } void -nv50_bufctx_emit_relocs(struct nv50_context *nv50) +nv50_bufctx_fence(struct nouveau_bufctx *bufctx, boolean on_flush) { - struct resident *rsd; - struct util_dynarray *array; - unsigned ctx, i, n; - - n = nv50->residents_size / sizeof(struct resident); - n += NV50_SCREEN_RESIDENT_BO_COUNT; - - MARK_RING(nv50->screen->base.channel, 0, n); - - for (ctx = 0; ctx < NV50_BUFCTX_COUNT; ++ctx) { - array = &nv50->residents[ctx]; - - n = array->size / sizeof(struct resident); - for (i = 0; i < n; ++i) { - rsd = util_dynarray_element(array, struct resident, i); - - nv50_resource_validate(rsd->res, rsd->flags); - } + struct nouveau_list *list = on_flush ? &bufctx->current : &bufctx->pending; + struct nouveau_list *it; + + for (it = list->next; it != list; it = it->next) { + struct nouveau_bufref *ref = (struct nouveau_bufref *)it; + struct nv04_resource *res = ref->priv; + if (res) + nv50_resource_validate(res, (unsigned)ref->priv_data); } - - nv50_screen_make_buffers_resident(nv50->screen); } diff --git a/src/gallium/drivers/nv50/nv50_context.h b/src/gallium/drivers/nv50/nv50_context.h index 2bf634a8237..0dd6c16bb48 100644 --- a/src/gallium/drivers/nv50/nv50_context.h +++ b/src/gallium/drivers/nv50/nv50_context.h @@ -46,12 +46,20 @@ #define NV50_NEW_CONSTBUF (1 << 18) #define NV50_NEW_TEXTURES (1 << 19) #define NV50_NEW_SAMPLERS (1 << 20) - -#define NV50_BUFCTX_CONSTANT 0 -#define NV50_BUFCTX_FRAME 1 -#define NV50_BUFCTX_VERTEX 2 -#define NV50_BUFCTX_TEXTURES 3 -#define NV50_BUFCTX_COUNT 4 +#define NV50_NEW_CONTEXT (1 << 31) + +#define NV50_BIND_FB 0 +#define NV50_BIND_VERTEX 1 +#define NV50_BIND_VERTEX_TMP 2 +#define NV50_BIND_INDEX 3 +#define NV50_BIND_TEXTURES 4 +#define NV50_BIND_CB(s, i) (5 + 16 * (s) + (i)) +#define NV50_BIND_SCREEN 53 +#define NV50_BIND_TLS 54 +#define NV50_BIND_COUNT 55 +#define NV50_BIND_2D 0 +#define NV50_BIND_M2MF 0 +#define NV50_BIND_FENCE 1 #define NV50_CB_TMP 123 /* fixed constant buffer binding points - low indices for user's constbufs */ @@ -60,13 +68,14 @@ #define NV50_CB_PFP 125 #define NV50_CB_AUX 127 + struct nv50_context { struct nouveau_context base; struct nv50_screen *screen; - struct util_dynarray residents[NV50_BUFCTX_COUNT]; - unsigned residents_size; + struct nouveau_bufctx *bufctx_3d; + struct nouveau_bufctx *bufctx; uint32_t dirty; @@ -79,6 +88,9 @@ struct nv50_context { int32_t index_bias; boolean prim_restart; boolean point_sprite; + boolean rt_serialize; + boolean flushed; + uint8_t tls_required; uint8_t num_vtxbufs; uint8_t num_vtxelts; uint8_t num_textures[3]; @@ -97,6 +109,7 @@ struct nv50_context { struct pipe_resource *constbuf[3][16]; uint16_t constbuf_dirty[3]; + uint16_t constbuf_valid[3]; struct pipe_vertex_buffer vtxbuf[PIPE_MAX_ATTRIBS]; unsigned num_vtxbufs; @@ -132,22 +145,19 @@ nv50_context(struct pipe_context *pipe) return (struct nv50_context *)pipe; } +static INLINE struct nv50_screen * +nv50_context_screen(struct nv50_context *nv50) +{ + return nv50_screen(&nv50->base.screen->base); +} + + /* nv50_context.c */ struct pipe_context *nv50_create(struct pipe_screen *, void *); -void nv50_default_flush_notify(struct nouveau_channel *); +void nv50_bufctx_fence(struct nouveau_bufctx *, boolean on_flush); -void nv50_bufctx_emit_relocs(struct nv50_context *); -void nv50_bufctx_add_resident(struct nv50_context *, int ctx, - struct nv04_resource *, uint32_t flags); -void nv50_bufctx_del_resident(struct nv50_context *, int ctx, - struct nv04_resource *); -static INLINE void -nv50_bufctx_reset(struct nv50_context *nv50, int ctx) -{ - nv50->residents_size -= nv50->residents[ctx].size; - util_dynarray_resize(&nv50->residents[ctx], 0); -} +void nv50_default_kick_notify(struct nouveau_pushbuf *); /* nv50_draw.c */ extern struct draw_stage *nv50_draw_render_stage(struct nv50_context *); @@ -194,7 +204,7 @@ nv50_create_sampler_view(struct pipe_context *, /* nv50_transfer.c */ void -nv50_m2mf_transfer_rect(struct pipe_screen *pscreen, +nv50_m2mf_transfer_rect(struct nv50_context *, const struct nv50_m2mf_rect *dst, const struct nv50_m2mf_rect *src, uint32_t nblocksx, uint32_t nblocksy); diff --git a/src/gallium/drivers/nv50/nv50_miptree.c b/src/gallium/drivers/nv50/nv50_miptree.c index 76b60592bfb..bfadbc5ab28 100644 --- a/src/gallium/drivers/nv50/nv50_miptree.c +++ b/src/gallium/drivers/nv50/nv50_miptree.c @@ -31,7 +31,7 @@ static INLINE uint32_t nv50_tex_choose_tile_dims(unsigned nx, unsigned ny, unsigned nz) { - return nvc0_tex_choose_tile_dims(nx, ny * 2, nz) >> 4; + return nvc0_tex_choose_tile_dims(nx, ny * 2, nz); } static uint32_t @@ -41,58 +41,60 @@ nv50_mt_choose_storage_type(struct nv50_miptree *mt, boolean compressed) uint32_t tile_flags; - if (mt->base.base.bind & PIPE_BIND_CURSOR) - return NOUVEAU_BO_TILE_SCANOUT; + if (unlikely(mt->base.base.flags & NOUVEAU_RESOURCE_FLAG_LINEAR)) + return 0; + if (unlikely(mt->base.base.bind & PIPE_BIND_CURSOR)) + return 0; switch (mt->base.base.format) { case PIPE_FORMAT_Z16_UNORM: - tile_flags = 0x6c00 + (ms << 8); + tile_flags = 0x6c + (ms << 8); break; case PIPE_FORMAT_S8_UINT_Z24_UNORM: - tile_flags = 0x1800 + (ms << 8); + tile_flags = 0x18 + (ms << 8); break; case PIPE_FORMAT_Z24X8_UNORM: case PIPE_FORMAT_Z24_UNORM_S8_UINT: - tile_flags = 0x22800 + (ms << 8); + tile_flags = 0x128 + (ms << 8); break; case PIPE_FORMAT_Z32_FLOAT: - tile_flags = 0x4000 + (ms << 8); + tile_flags = 0x40 + (ms << 8); break; case PIPE_FORMAT_Z32_FLOAT_S8X24_UINT: - tile_flags = 0x6000 + (ms << 8); + tile_flags = 0x60 + (ms << 8); break; default: switch (util_format_get_blocksizebits(mt->base.base.format)) { case 128: assert(ms < 3); - tile_flags = 0x7400; + tile_flags = 0x74; break; case 64: switch (ms) { - case 2: tile_flags = 0x17c00; break; - case 3: tile_flags = 0x17d00; break; + case 2: tile_flags = 0xfc; break; + case 3: tile_flags = 0xfd; break; default: - tile_flags = 0x7000; + tile_flags = 0x70; break; } break; case 32: if (mt->base.base.bind & PIPE_BIND_SCANOUT) { assert(ms == 0); - tile_flags = 0x7a00; + tile_flags = 0x7a; } else { switch (ms) { - case 2: tile_flags = 0x17800; break; - case 3: tile_flags = 0x17900; break; + case 2: tile_flags = 0xf8; break; + case 3: tile_flags = 0xf9; break; default: - tile_flags = 0x7000; + tile_flags = 0x70; break; } } break; case 16: case 8: - tile_flags = 0x7000; + tile_flags = 0x70; break; default: return 0; @@ -101,14 +103,8 @@ nv50_mt_choose_storage_type(struct nv50_miptree *mt, boolean compressed) tile_flags = 0; } - if (mt->base.base.bind & (PIPE_BIND_SCANOUT | PIPE_BIND_CURSOR)) - tile_flags |= NOUVEAU_BO_TILE_SCANOUT; - if (!compressed) - tile_flags &= ~0x30000; - - if (unlikely(mt->base.base.flags & NOUVEAU_RESOURCE_FLAG_LINEAR)) - tile_flags &= ~0x3ff00; + tile_flags &= ~0x180; return tile_flags; } @@ -118,7 +114,7 @@ nv50_miptree_destroy(struct pipe_screen *pscreen, struct pipe_resource *pt) { struct nv50_miptree *mt = nv50_miptree(pt); - nouveau_screen_bo_release(pscreen, mt->base.bo); + nouveau_bo_ref(NULL, &mt->base.bo); FREE(mt); } @@ -260,7 +256,8 @@ nv50_miptree_create(struct pipe_screen *pscreen, struct nv50_miptree *mt = CALLOC_STRUCT(nv50_miptree); struct pipe_resource *pt = &mt->base.base; int ret; - uint32_t tile_flags; + union nouveau_bo_config bo_config; + uint32_t bo_flags; if (!mt) return NULL; @@ -270,30 +267,34 @@ nv50_miptree_create(struct pipe_screen *pscreen, pipe_reference_init(&pt->reference, 1); pt->screen = pscreen; - tile_flags = nv50_mt_choose_storage_type(mt, TRUE); + bo_config.nv50.memtype = nv50_mt_choose_storage_type(mt, TRUE); if (!nv50_miptree_init_ms_mode(mt)) { FREE(mt); return NULL; } - if (tile_flags & NOUVEAU_BO_TILE_LAYOUT_MASK) { + if (bo_config.nv50.memtype != 0) { nv50_miptree_init_layout_tiled(mt); } else if (!nv50_miptree_init_layout_linear(mt)) { FREE(mt); return NULL; } + bo_config.nv50.tile_mode = mt->level[0].tile_mode; + + bo_flags = NOUVEAU_BO_VRAM | NOUVEAU_BO_NOSNOOP; + if (mt->base.base.bind & (PIPE_BIND_CURSOR | PIPE_BIND_DISPLAY_TARGET)) + bo_flags |= NOUVEAU_BO_CONTIG; - ret = nouveau_bo_new_tile(dev, NOUVEAU_BO_VRAM, 4096, - mt->total_size, - mt->level[0].tile_mode, tile_flags, - &mt->base.bo); + ret = nouveau_bo_new(dev, bo_flags, 4096, mt->total_size, &bo_config, + &mt->base.bo); if (ret) { FREE(mt); return NULL; } mt->base.domain = NOUVEAU_BO_VRAM; + mt->base.address = mt->base.bo->offset; return pt; } @@ -323,6 +324,8 @@ nv50_miptree_from_handle(struct pipe_screen *pscreen, FREE(mt); return NULL; } + mt->base.domain = NOUVEAU_BO_VRAM; + mt->base.address = mt->base.bo->offset; mt->base.base = *templ; mt->base.vtbl = &nv50_miptree_vtbl; @@ -330,7 +333,7 @@ nv50_miptree_from_handle(struct pipe_screen *pscreen, mt->base.base.screen = pscreen; mt->level[0].pitch = stride; mt->level[0].offset = 0; - mt->level[0].tile_mode = mt->base.bo->tile_mode; + mt->level[0].tile_mode = mt->base.bo->config.nv50.tile_mode; /* no need to adjust bo reference count */ return &mt->base.base; diff --git a/src/gallium/drivers/nv50/nv50_program.c b/src/gallium/drivers/nv50/nv50_program.c index c141a91eec2..776cecaeb7e 100644 --- a/src/gallium/drivers/nv50/nv50_program.c +++ b/src/gallium/drivers/nv50/nv50_program.c @@ -489,7 +489,8 @@ nv50_fragprog_prepare(struct nv50_translation_info *ti) ++nintp; } - p->fp.colors = 4 << NV50_3D_MAP_SEMANTIC_0_FFC0_ID__SHIFT; /* after HPOS */ + /* after HPOS */ + p->fp.colors = 4 << NV50_3D_SEMANTIC_COLOR_FFC0_ID__SHIFT; for (i = 0; i < p->in_nr; ++i) { int j = p->in[i].id; @@ -680,8 +681,8 @@ nv50_program_destroy(struct nv50_context *nv50, struct nv50_program *p) const struct pipe_shader_state pipe = p->pipe; const ubyte type = p->type; - if (p->res) - nouveau_resource_free(&p->res); + if (p->mem) + nouveau_heap_free(&p->mem); if (p->code) FREE(p->code); diff --git a/src/gallium/drivers/nv50/nv50_program.h b/src/gallium/drivers/nv50/nv50_program.h index 993e1691ab7..17aee97a024 100644 --- a/src/gallium/drivers/nv50/nv50_program.h +++ b/src/gallium/drivers/nv50/nv50_program.h @@ -87,7 +87,7 @@ struct nv50_program { void *fixups; unsigned num_fixups; - struct nouveau_resource *res; + struct nouveau_heap *mem; }; #define NV50_INTERP_LINEAR (1 << 0) diff --git a/src/gallium/drivers/nv50/nv50_push.c b/src/gallium/drivers/nv50/nv50_push.c index e8ad1ddd38a..04e32b7e8b9 100644 --- a/src/gallium/drivers/nv50/nv50_push.c +++ b/src/gallium/drivers/nv50/nv50_push.c @@ -11,7 +11,7 @@ #include "nv50_3d.xml.h" struct push_context { - struct nouveau_channel *chan; + struct nouveau_pushbuf *push; void *idxbuf; @@ -74,20 +74,20 @@ emit_vertices_i08(struct push_context *ctx, unsigned start, unsigned count) size = ctx->vertex_words * nr; - BEGIN_RING_NI(ctx->chan, RING_3D(VERTEX_DATA), size); + BEGIN_NI04(ctx->push, NV50_3D(VERTEX_DATA), size); ctx->translate->run_elts8(ctx->translate, elts, nr, ctx->instance_id, - ctx->chan->cur); + ctx->push->cur); - ctx->chan->cur += size; + ctx->push->cur += size; count -= nr; elts += nr; if (nr != push) { count--; elts++; - BEGIN_RING(ctx->chan, RING_3D(VB_ELEMENT_U32), 1); - OUT_RING (ctx->chan, ctx->restart_index); + BEGIN_NV04(ctx->push, NV50_3D(VB_ELEMENT_U32), 1); + PUSH_DATA (ctx->push, ctx->restart_index); } } } @@ -107,20 +107,20 @@ emit_vertices_i16(struct push_context *ctx, unsigned start, unsigned count) size = ctx->vertex_words * nr; - BEGIN_RING_NI(ctx->chan, RING_3D(VERTEX_DATA), size); + BEGIN_NI04(ctx->push, NV50_3D(VERTEX_DATA), size); ctx->translate->run_elts16(ctx->translate, elts, nr, ctx->instance_id, - ctx->chan->cur); + ctx->push->cur); - ctx->chan->cur += size; + ctx->push->cur += size; count -= nr; elts += nr; if (nr != push) { count--; elts++; - BEGIN_RING(ctx->chan, RING_3D(VB_ELEMENT_U32), 1); - OUT_RING (ctx->chan, ctx->restart_index); + BEGIN_NV04(ctx->push, NV50_3D(VB_ELEMENT_U32), 1); + PUSH_DATA (ctx->push, ctx->restart_index); } } } @@ -140,20 +140,20 @@ emit_vertices_i32(struct push_context *ctx, unsigned start, unsigned count) size = ctx->vertex_words * nr; - BEGIN_RING_NI(ctx->chan, RING_3D(VERTEX_DATA), size); + BEGIN_NI04(ctx->push, NV50_3D(VERTEX_DATA), size); ctx->translate->run_elts(ctx->translate, elts, nr, ctx->instance_id, - ctx->chan->cur); + ctx->push->cur); - ctx->chan->cur += size; + ctx->push->cur += size; count -= nr; elts += nr; if (nr != push) { count--; elts++; - BEGIN_RING(ctx->chan, RING_3D(VB_ELEMENT_U32), 1); - OUT_RING (ctx->chan, ctx->restart_index); + BEGIN_NV04(ctx->push, NV50_3D(VB_ELEMENT_U32), 1); + PUSH_DATA (ctx->push, ctx->restart_index); } } } @@ -165,11 +165,11 @@ emit_vertices_seq(struct push_context *ctx, unsigned start, unsigned count) unsigned push = MIN2(count, ctx->packet_vertex_limit); unsigned size = ctx->vertex_words * push; - BEGIN_RING_NI(ctx->chan, RING_3D(VERTEX_DATA), size); + BEGIN_NI04(ctx->push, NV50_3D(VERTEX_DATA), size); ctx->translate->run(ctx->translate, start, push, ctx->instance_id, - ctx->chan->cur); - ctx->chan->cur += size; + ctx->push->cur); + ctx->push->cur += size; count -= push; start += push; } @@ -213,7 +213,7 @@ nv50_push_vbo(struct nv50_context *nv50, const struct pipe_draw_info *info) unsigned inst = info->instance_count; boolean apply_bias = info->indexed && info->index_bias; - ctx.chan = nv50->screen->base.channel; + ctx.push = nv50->base.pushbuf; ctx.translate = nv50->vertex->translate; ctx.packet_vertex_limit = nv50->vertex->packet_vertex_limit; ctx.vertex_words = nv50->vertex->vertex_size; @@ -252,19 +252,19 @@ nv50_push_vbo(struct nv50_context *nv50, const struct pipe_draw_info *info) ctx.prim = nv50_prim_gl(info->mode); if (info->primitive_restart) { - BEGIN_RING(ctx.chan, RING_3D(PRIM_RESTART_ENABLE), 2); - OUT_RING (ctx.chan, 1); - OUT_RING (ctx.chan, info->restart_index); + BEGIN_NV04(ctx.push, NV50_3D(PRIM_RESTART_ENABLE), 2); + PUSH_DATA (ctx.push, 1); + PUSH_DATA (ctx.push, info->restart_index); } else if (nv50->state.prim_restart) { - BEGIN_RING(ctx.chan, RING_3D(PRIM_RESTART_ENABLE), 1); - OUT_RING (ctx.chan, 0); + BEGIN_NV04(ctx.push, NV50_3D(PRIM_RESTART_ENABLE), 1); + PUSH_DATA (ctx.push, 0); } nv50->state.prim_restart = info->primitive_restart; while (inst--) { - BEGIN_RING(ctx.chan, RING_3D(VERTEX_BEGIN_GL), 1); - OUT_RING (ctx.chan, ctx.prim); + BEGIN_NV04(ctx.push, NV50_3D(VERTEX_BEGIN_GL), 1); + PUSH_DATA (ctx.push, ctx.prim); switch (index_size) { case 0: emit_vertices_seq(&ctx, info->start, info->count); @@ -282,8 +282,8 @@ nv50_push_vbo(struct nv50_context *nv50, const struct pipe_draw_info *info) assert(0); break; } - BEGIN_RING(ctx.chan, RING_3D(VERTEX_END_GL), 1); - OUT_RING (ctx.chan, 0); + BEGIN_NV04(ctx.push, NV50_3D(VERTEX_END_GL), 1); + PUSH_DATA (ctx.push, 0); ctx.instance_id++; ctx.prim |= NV50_3D_VERTEX_BEGIN_GL_INSTANCE_NEXT; diff --git a/src/gallium/drivers/nv50/nv50_query.c b/src/gallium/drivers/nv50/nv50_query.c index 59d315aa475..220b166f985 100644 --- a/src/gallium/drivers/nv50/nv50_query.c +++ b/src/gallium/drivers/nv50/nv50_query.c @@ -22,6 +22,8 @@ * Authors: Christoph Bumiller */ +#define NV50_PUSH_EXPLICIT_SPACE_CHECKING + #include "nv50_context.h" #include "nouveau/nv_object.xml.h" @@ -64,7 +66,8 @@ nv50_query_allocate(struct nv50_context *nv50, struct nv50_query *q, int size) if (q->ready) nouveau_mm_free(q->mm); else - nouveau_fence_work(screen->base.fence.current, nouveau_mm_free_work, q->mm); + nouveau_fence_work(screen->base.fence.current, nouveau_mm_free_work, + q->mm); } } if (size) { @@ -73,14 +76,12 @@ nv50_query_allocate(struct nv50_context *nv50, struct nv50_query *q, int size) return FALSE; q->offset = q->base; - ret = nouveau_bo_map_range(q->bo, q->base, size, NOUVEAU_BO_RD | - NOUVEAU_BO_NOSYNC); + ret = nouveau_bo_map(q->bo, 0, screen->base.client); if (ret) { nv50_query_allocate(nv50, q, 0); return FALSE; } - q->data = q->bo->map; - nouveau_bo_unmap(q->bo); + q->data = (uint32_t *)((uint8_t *)q->bo->map + q->base); } return TRUE; } @@ -121,24 +122,25 @@ nv50_query_create(struct pipe_context *pipe, unsigned type) } static void -nv50_query_get(struct nouveau_channel *chan, struct nv50_query *q, +nv50_query_get(struct nouveau_pushbuf *push, struct nv50_query *q, unsigned offset, uint32_t get) { offset += q->offset; - MARK_RING (chan, 5, 2); - BEGIN_RING(chan, RING_3D(QUERY_ADDRESS_HIGH), 4); - OUT_RELOCh(chan, q->bo, offset, NOUVEAU_BO_GART | NOUVEAU_BO_WR); - OUT_RELOCl(chan, q->bo, offset, NOUVEAU_BO_GART | NOUVEAU_BO_WR); - OUT_RING (chan, q->sequence); - OUT_RING (chan, get); + PUSH_SPACE(push, 5); + PUSH_REFN (push, q->bo, NOUVEAU_BO_GART | NOUVEAU_BO_WR); + BEGIN_NV04(push, NV50_3D(QUERY_ADDRESS_HIGH), 4); + PUSH_DATAh(push, q->bo->offset + offset); + PUSH_DATA (push, q->bo->offset + offset); + PUSH_DATA (push, q->sequence); + PUSH_DATA (push, get); } static void nv50_query_begin(struct pipe_context *pipe, struct pipe_query *pq) { struct nv50_context *nv50 = nv50_context(pipe); - struct nouveau_channel *chan = nv50->screen->base.channel; + struct nouveau_pushbuf *push = nv50->base.pushbuf; struct nv50_query *q = nv50_query(pq); /* For occlusion queries we have to change the storage, because a previous @@ -161,27 +163,31 @@ nv50_query_begin(struct pipe_context *pipe, struct pipe_query *pq) switch (q->type) { case PIPE_QUERY_OCCLUSION_COUNTER: - BEGIN_RING(chan, RING_3D(COUNTER_RESET), 1); - OUT_RING (chan, NV50_3D_COUNTER_RESET_SAMPLECNT); - BEGIN_RING(chan, RING_3D(SAMPLECNT_ENABLE), 1); - OUT_RING (chan, 1); + PUSH_SPACE(push, 4); + BEGIN_NV04(push, NV50_3D(COUNTER_RESET), 1); + PUSH_DATA (push, NV50_3D_COUNTER_RESET_SAMPLECNT); + BEGIN_NV04(push, NV50_3D(SAMPLECNT_ENABLE), 1); + PUSH_DATA (push, 1); break; case PIPE_QUERY_PRIMITIVES_GENERATED: /* store before & after instead ? */ - BEGIN_RING(chan, RING_3D(COUNTER_RESET), 1); - OUT_RING (chan, NV50_3D_COUNTER_RESET_GENERATED_PRIMITIVES); + PUSH_SPACE(push, 2); + BEGIN_NV04(push, NV50_3D(COUNTER_RESET), 1); + PUSH_DATA (push, NV50_3D_COUNTER_RESET_GENERATED_PRIMITIVES); break; case PIPE_QUERY_PRIMITIVES_EMITTED: - BEGIN_RING(chan, RING_3D(COUNTER_RESET), 1); - OUT_RING (chan, NV50_3D_COUNTER_RESET_TRANSFORM_FEEDBACK); + PUSH_SPACE(push, 2); + BEGIN_NV04(push, NV50_3D(COUNTER_RESET), 1); + PUSH_DATA (push, NV50_3D_COUNTER_RESET_TRANSFORM_FEEDBACK); break; case PIPE_QUERY_SO_STATISTICS: - BEGIN_RING_NI(chan, RING_3D(COUNTER_RESET), 2); - OUT_RING (chan, NV50_3D_COUNTER_RESET_TRANSFORM_FEEDBACK); - OUT_RING (chan, NV50_3D_COUNTER_RESET_GENERATED_PRIMITIVES); + PUSH_SPACE(push, 3); + BEGIN_NI04(push, NV50_3D(COUNTER_RESET), 2); + PUSH_DATA (push, NV50_3D_COUNTER_RESET_TRANSFORM_FEEDBACK); + PUSH_DATA (push, NV50_3D_COUNTER_RESET_GENERATED_PRIMITIVES); break; case PIPE_QUERY_TIMESTAMP_DISJOINT: case PIPE_QUERY_TIME_ELAPSED: - nv50_query_get(chan, q, 0x10, 0x00005002); + nv50_query_get(push, q, 0x10, 0x00005002); break; default: break; @@ -193,31 +199,32 @@ static void nv50_query_end(struct pipe_context *pipe, struct pipe_query *pq) { struct nv50_context *nv50 = nv50_context(pipe); - struct nouveau_channel *chan = nv50->screen->base.channel; + struct nouveau_pushbuf *push = nv50->base.pushbuf; struct nv50_query *q = nv50_query(pq); switch (q->type) { case PIPE_QUERY_OCCLUSION_COUNTER: - nv50_query_get(chan, q, 0, 0x0100f002); - BEGIN_RING(chan, RING_3D(SAMPLECNT_ENABLE), 1); - OUT_RING (chan, 0); + nv50_query_get(push, q, 0, 0x0100f002); + PUSH_SPACE(push, 2); + BEGIN_NV04(push, NV50_3D(SAMPLECNT_ENABLE), 1); + PUSH_DATA (push, 0); break; case PIPE_QUERY_PRIMITIVES_GENERATED: - nv50_query_get(chan, q, 0, 0x06805002); + nv50_query_get(push, q, 0, 0x06805002); break; case PIPE_QUERY_PRIMITIVES_EMITTED: - nv50_query_get(chan, q, 0, 0x05805002); + nv50_query_get(push, q, 0, 0x05805002); break; case PIPE_QUERY_SO_STATISTICS: - nv50_query_get(chan, q, 0x00, 0x05805002); - nv50_query_get(chan, q, 0x10, 0x06805002); + nv50_query_get(push, q, 0x00, 0x05805002); + nv50_query_get(push, q, 0x10, 0x06805002); break; case PIPE_QUERY_TIMESTAMP_DISJOINT: case PIPE_QUERY_TIME_ELAPSED: - nv50_query_get(chan, q, 0, 0x00005002); + nv50_query_get(push, q, 0, 0x00005002); break; case PIPE_QUERY_GPU_FINISHED: - nv50_query_get(chan, q, 0, 0x1000f010); + nv50_query_get(push, q, 0, 0x1000f010); break; default: assert(0); @@ -231,45 +238,33 @@ nv50_query_ready(struct nv50_query *q) return q->ready || (!q->is64bit && (q->data[0] == q->sequence)); } -static INLINE boolean -nv50_query_wait(struct nv50_query *q) -{ - int ret = nouveau_bo_map(q->bo, NOUVEAU_BO_RD); - if (ret) - return FALSE; - nouveau_bo_unmap(q->bo); - return TRUE; -} - static boolean nv50_query_result(struct pipe_context *pipe, struct pipe_query *pq, boolean wait, union pipe_query_result *result) { + struct nv50_context *nv50 = nv50_context(pipe); struct nv50_query *q = nv50_query(pq); - uint64_t *res64 = (uint64_t*)result; - boolean *res8 = (boolean*)result; + uint64_t *res64 = (uint64_t *)result; + boolean *res8 = (boolean *)result; uint64_t *data64 = (uint64_t *)q->data; - if (q->type == PIPE_QUERY_GPU_FINISHED) { - res8[0] = nv50_query_ready(q); - return TRUE; - } - if (!q->ready) /* update ? */ q->ready = nv50_query_ready(q); if (!q->ready) { - struct nouveau_channel *chan = nv50_context(pipe)->screen->base.channel; if (!wait) { - if (nouveau_bo_pending(q->bo) & NOUVEAU_BO_WR) /* for daft apps */ - FIRE_RING(chan); + /* for broken apps that spin on GL_QUERY_RESULT_AVAILABLE */ + PUSH_KICK(nv50->base.pushbuf); return FALSE; } - if (!nv50_query_wait(q)) + if (nouveau_bo_wait(q->bo, NOUVEAU_BO_RD, nv50->screen->base.client)) return FALSE; } q->ready = TRUE; switch (q->type) { + case PIPE_QUERY_GPU_FINISHED: + res8[0] = TRUE; + break; case PIPE_QUERY_OCCLUSION_COUNTER: /* u32 sequence, u32 count, u64 time */ res64[0] = q->data[1]; break; @@ -300,27 +295,28 @@ nv50_render_condition(struct pipe_context *pipe, struct pipe_query *pq, uint mode) { struct nv50_context *nv50 = nv50_context(pipe); - struct nouveau_channel *chan = nv50->screen->base.channel; + struct nouveau_pushbuf *push = nv50->base.pushbuf; struct nv50_query *q; + PUSH_SPACE(push, 6); + if (!pq) { - BEGIN_RING(chan, RING_3D(COND_MODE), 1); - OUT_RING (chan, NV50_3D_COND_MODE_ALWAYS); + BEGIN_NV04(push, NV50_3D(COND_MODE), 1); + PUSH_DATA (push, NV50_3D_COND_MODE_ALWAYS); return; } q = nv50_query(pq); if (mode == PIPE_RENDER_COND_WAIT || mode == PIPE_RENDER_COND_BY_REGION_WAIT) { - BEGIN_RING(chan, RING_3D_(NV50_GRAPH_WAIT_FOR_IDLE), 1); - OUT_RING (chan, 0); + BEGIN_NV04(push, SUBC_3D(NV50_GRAPH_SERIALIZE), 1); + PUSH_DATA (push, 0); } - MARK_RING (chan, 4, 2); - BEGIN_RING(chan, RING_3D(COND_ADDRESS_HIGH), 3); - OUT_RELOCh(chan, q->bo, q->offset, NOUVEAU_BO_GART | NOUVEAU_BO_RD); - OUT_RELOCl(chan, q->bo, q->offset, NOUVEAU_BO_GART | NOUVEAU_BO_RD); - OUT_RING (chan, NV50_3D_COND_MODE_RES_NON_ZERO); + BEGIN_NV04(push, NV50_3D(COND_ADDRESS_HIGH), 3); + PUSH_DATAh(push, q->bo->offset + q->offset); + PUSH_DATA (push, q->bo->offset + q->offset); + PUSH_DATA (push, NV50_3D_COND_MODE_RES_NON_ZERO); } void diff --git a/src/gallium/drivers/nv50/nv50_resource.h b/src/gallium/drivers/nv50/nv50_resource.h index 50cc2afdbe1..5d5596ea6e3 100644 --- a/src/gallium/drivers/nv50/nv50_resource.h +++ b/src/gallium/drivers/nv50/nv50_resource.h @@ -4,10 +4,9 @@ #include "util/u_transfer.h" #include "util/u_double_list.h" -#define NOUVEAU_NVC0 + #include "nouveau/nouveau_winsys.h" #include "nouveau/nouveau_buffer.h" -#undef NOUVEAU_NVC0 #ifndef __NVC0_RESOURCE_H__ /* make sure we don't use these in nvc0: */ @@ -19,12 +18,12 @@ nv50_screen_init_resource_functions(struct pipe_screen *pscreen); #define NV50_TILE_SHIFT_X(m) 6 -#define NV50_TILE_SHIFT_Y(m) ((((m) >> 0) & 0xf) + 2) -#define NV50_TILE_SHIFT_Z(m) ((((m) >> 4) & 0xf) + 0) +#define NV50_TILE_SHIFT_Y(m) ((((m) >> 4) & 0xf) + 2) +#define NV50_TILE_SHIFT_Z(m) ((((m) >> 8) & 0xf) + 0) #define NV50_TILE_SIZE_X(m) 64 -#define NV50_TILE_SIZE_Y(m) ( 4 << (((m) >> 0) & 0xf)) -#define NV50_TILE_SIZE_Z(m) ( 1 << (((m) >> 4) & 0xf)) +#define NV50_TILE_SIZE_Y(m) ( 4 << (((m) >> 4) & 0xf)) +#define NV50_TILE_SIZE_Z(m) ( 1 << (((m) >> 8) & 0xf)) #define NV50_TILE_SIZE_2D(m) (NV50_TILE_SIZE_X(m) << NV50_TILE_SHIFT_Y(m)) diff --git a/src/gallium/drivers/nv50/nv50_screen.c b/src/gallium/drivers/nv50/nv50_screen.c index 27566e24108..8a3c552784f 100644 --- a/src/gallium/drivers/nv50/nv50_screen.c +++ b/src/gallium/drivers/nv50/nv50_screen.c @@ -33,9 +33,6 @@ # define NOUVEAU_GETPARAM_GRAPH_UNITS 13 #endif -extern int nouveau_device_get_param(struct nouveau_device *dev, - uint64_t param, uint64_t *value); - static boolean nv50_screen_is_format_supported(struct pipe_screen *pscreen, enum pipe_format format, @@ -53,7 +50,7 @@ nv50_screen_is_format_supported(struct pipe_screen *pscreen, switch (format) { case PIPE_FORMAT_Z16_UNORM: - if (nv50_screen(pscreen)->tesla->grclass < NVA0_3D) + if (nv50_screen(pscreen)->tesla->oclass < NVA0_3D_CLASS) return FALSE; break; case PIPE_FORMAT_R8G8B8A8_UNORM: @@ -100,7 +97,7 @@ nv50_screen_get_param(struct pipe_screen *pscreen, enum pipe_cap param) return 1; case PIPE_CAP_STREAM_OUTPUT_PAUSE_RESUME: case PIPE_CAP_SEAMLESS_CUBE_MAP: - return nv50_screen(pscreen)->tesla->grclass >= NVA0_3D; + return nv50_screen(pscreen)->tesla->oclass >= NVA0_3D_CLASS; case PIPE_CAP_SEAMLESS_CUBE_MAP_PER_TEXTURE: return 0; case PIPE_CAP_TWO_SIDED_STENCIL: @@ -131,7 +128,7 @@ nv50_screen_get_param(struct pipe_screen *pscreen, enum pipe_cap param) case PIPE_CAP_INDEP_BLEND_ENABLE: return 1; case PIPE_CAP_INDEP_BLEND_FUNC: - return nv50_screen(pscreen)->tesla->grclass >= NVA3_3D; + return nv50_screen(pscreen)->tesla->oclass >= NVA3_3D_CLASS; case PIPE_CAP_TGSI_FS_COORD_ORIGIN_UPPER_LEFT: case PIPE_CAP_TGSI_FS_COORD_PIXEL_CENTER_HALF_INTEGER: return 1; @@ -241,8 +238,9 @@ nv50_screen_destroy(struct pipe_screen *pscreen) nouveau_fence_wait(screen->base.fence.current); nouveau_fence_ref (NULL, &screen->base.fence.current); } - if (screen->base.channel) - screen->base.channel->user_private = NULL; + if (screen->base.pushbuf) + screen->base.pushbuf->user_priv = NULL; + if (screen->blitctx) FREE(screen->blitctx); @@ -253,20 +251,17 @@ nv50_screen_destroy(struct pipe_screen *pscreen) nouveau_bo_ref(NULL, &screen->uniforms); nouveau_bo_ref(NULL, &screen->fence.bo); - nouveau_resource_destroy(&screen->vp_code_heap); - nouveau_resource_destroy(&screen->gp_code_heap); - nouveau_resource_destroy(&screen->fp_code_heap); + nouveau_heap_destroy(&screen->vp_code_heap); + nouveau_heap_destroy(&screen->gp_code_heap); + nouveau_heap_destroy(&screen->fp_code_heap); if (screen->tic.entries) FREE(screen->tic.entries); - nouveau_mm_destroy(screen->mm_VRAM_fe0); - - nouveau_grobj_free(&screen->tesla); - nouveau_grobj_free(&screen->eng2d); - nouveau_grobj_free(&screen->m2mf); - - nouveau_notifier_free(&screen->sync); + nouveau_object_del(&screen->tesla); + nouveau_object_del(&screen->eng2d); + nouveau_object_del(&screen->m2mf); + nouveau_object_del(&screen->sync); nouveau_screen_fini(&screen->base); @@ -277,18 +272,16 @@ static void nv50_screen_fence_emit(struct pipe_screen *pscreen, u32 *sequence) { struct nv50_screen *screen = nv50_screen(pscreen); - struct nouveau_channel *chan = screen->base.channel; - - MARK_RING (chan, 5, 2); + struct nouveau_pushbuf *push = screen->base.pushbuf; /* we need to do it after possible flush in MARK_RING */ *sequence = ++screen->base.fence.sequence; - BEGIN_RING(chan, RING_3D(QUERY_ADDRESS_HIGH), 4); - OUT_RELOCh(chan, screen->fence.bo, 0, NOUVEAU_BO_WR); - OUT_RELOCl(chan, screen->fence.bo, 0, NOUVEAU_BO_WR); - OUT_RING (chan, *sequence); - OUT_RING (chan, NV50_3D_QUERY_GET_MODE_WRITE_UNK0 | + PUSH_DATA (push, NV50_FIFO_PKHDR(NV50_3D(QUERY_ADDRESS_HIGH), 4)); + PUSH_DATAh(push, screen->fence.bo->offset); + PUSH_DATA (push, screen->fence.bo->offset); + PUSH_DATA (push, *sequence); + PUSH_DATA (push, NV50_3D_QUERY_GET_MODE_WRITE_UNK0 | NV50_3D_QUERY_GET_UNK4 | NV50_3D_QUERY_GET_UNIT_CROP | NV50_3D_QUERY_GET_TYPE_QUERY | @@ -299,8 +292,205 @@ nv50_screen_fence_emit(struct pipe_screen *pscreen, u32 *sequence) static u32 nv50_screen_fence_update(struct pipe_screen *pscreen) { - struct nv50_screen *screen = nv50_screen(pscreen); - return screen->fence.map[0]; + return nv50_screen(pscreen)->fence.map[0]; +} + +static int +nv50_screen_init_hwctx(struct nv50_screen *screen, unsigned tls_space) +{ + struct nouveau_pushbuf *push = screen->base.pushbuf; + struct nv04_fifo *fifo; + unsigned i; + + fifo = (struct nv04_fifo *)screen->base.channel->data; + + BEGIN_NV04(push, SUBC_M2MF(NV01_SUBCHAN_OBJECT), 1); + PUSH_DATA (push, screen->m2mf->handle); + BEGIN_NV04(push, SUBC_M2MF(NV03_M2MF_DMA_NOTIFY), 3); + PUSH_DATA (push, screen->sync->handle); + PUSH_DATA (push, fifo->vram); + PUSH_DATA (push, fifo->vram); + + BEGIN_NV04(push, SUBC_2D(NV01_SUBCHAN_OBJECT), 1); + PUSH_DATA (push, screen->eng2d->handle); + BEGIN_NV04(push, NV50_2D(DMA_NOTIFY), 4); + PUSH_DATA (push, screen->sync->handle); + PUSH_DATA (push, fifo->vram); + PUSH_DATA (push, fifo->vram); + PUSH_DATA (push, fifo->vram); + BEGIN_NV04(push, NV50_2D(OPERATION), 1); + PUSH_DATA (push, NV50_2D_OPERATION_SRCCOPY); + BEGIN_NV04(push, NV50_2D(CLIP_ENABLE), 1); + PUSH_DATA (push, 0); + BEGIN_NV04(push, NV50_2D(COLOR_KEY_ENABLE), 1); + PUSH_DATA (push, 0); + BEGIN_NV04(push, SUBC_2D(0x0888), 1); + PUSH_DATA (push, 1); + + BEGIN_NV04(push, SUBC_3D(NV01_SUBCHAN_OBJECT), 1); + PUSH_DATA (push, screen->tesla->handle); + + BEGIN_NV04(push, NV50_3D(COND_MODE), 1); + PUSH_DATA (push, NV50_3D_COND_MODE_ALWAYS); + + BEGIN_NV04(push, NV50_3D(DMA_NOTIFY), 1); + PUSH_DATA (push, screen->sync->handle); + BEGIN_NV04(push, NV50_3D(DMA_ZETA), 11); + for (i = 0; i < 11; ++i) + PUSH_DATA(push, fifo->vram); + BEGIN_NV04(push, NV50_3D(DMA_COLOR(0)), NV50_3D_DMA_COLOR__LEN); + for (i = 0; i < NV50_3D_DMA_COLOR__LEN; ++i) + PUSH_DATA(push, fifo->vram); + + BEGIN_NV04(push, NV50_3D(REG_MODE), 1); + PUSH_DATA (push, NV50_3D_REG_MODE_STRIPED); + BEGIN_NV04(push, NV50_3D(UNK1400_LANES), 1); + PUSH_DATA (push, 0xf); + + BEGIN_NV04(push, NV50_3D(RT_CONTROL), 1); + PUSH_DATA (push, 1); + + BEGIN_NV04(push, NV50_3D(CSAA_ENABLE), 1); + PUSH_DATA (push, 0); + BEGIN_NV04(push, NV50_3D(MULTISAMPLE_ENABLE), 1); + PUSH_DATA (push, 0); + BEGIN_NV04(push, NV50_3D(MULTISAMPLE_MODE), 1); + PUSH_DATA (push, NV50_3D_MULTISAMPLE_MODE_MS1); + BEGIN_NV04(push, NV50_3D(MULTISAMPLE_CTRL), 1); + PUSH_DATA (push, 0); + BEGIN_NV04(push, NV50_3D(LINE_LAST_PIXEL), 1); + PUSH_DATA (push, 0); + BEGIN_NV04(push, NV50_3D(BLEND_SEPARATE_ALPHA), 1); + PUSH_DATA (push, 1); + + if (screen->tesla->oclass >= NVA0_3D_CLASS) { + BEGIN_NV04(push, SUBC_3D(NVA0_3D_TEX_MISC), 1); + PUSH_DATA (push, NVA0_3D_TEX_MISC_SEAMLESS_CUBE_MAP); + } + + BEGIN_NV04(push, NV50_3D(SCREEN_Y_CONTROL), 1); + PUSH_DATA (push, 0); + BEGIN_NV04(push, NV50_3D(WINDOW_OFFSET_X), 2); + PUSH_DATA (push, 0); + PUSH_DATA (push, 0); + BEGIN_NV04(push, NV50_3D(ZCULL_REGION), 1); + PUSH_DATA (push, 0x3f); + + BEGIN_NV04(push, NV50_3D(VP_ADDRESS_HIGH), 2); + PUSH_DATAh(push, screen->code->offset + (0 << NV50_CODE_BO_SIZE_LOG2)); + PUSH_DATA (push, screen->code->offset + (0 << NV50_CODE_BO_SIZE_LOG2)); + + BEGIN_NV04(push, NV50_3D(FP_ADDRESS_HIGH), 2); + PUSH_DATAh(push, screen->code->offset + (1 << NV50_CODE_BO_SIZE_LOG2)); + PUSH_DATA (push, screen->code->offset + (1 << NV50_CODE_BO_SIZE_LOG2)); + + BEGIN_NV04(push, NV50_3D(GP_ADDRESS_HIGH), 2); + PUSH_DATAh(push, screen->code->offset + (2 << NV50_CODE_BO_SIZE_LOG2)); + PUSH_DATA (push, screen->code->offset + (2 << NV50_CODE_BO_SIZE_LOG2)); + + BEGIN_NV04(push, NV50_3D(LOCAL_ADDRESS_HIGH), 3); + PUSH_DATAh(push, screen->tls_bo->offset); + PUSH_DATA (push, screen->tls_bo->offset); + PUSH_DATA (push, util_logbase2(tls_space / 8)); + + BEGIN_NV04(push, NV50_3D(STACK_ADDRESS_HIGH), 3); + PUSH_DATAh(push, screen->stack_bo->offset); + PUSH_DATA (push, screen->stack_bo->offset); + PUSH_DATA (push, 4); + + BEGIN_NV04(push, NV50_3D(CB_DEF_ADDRESS_HIGH), 3); + PUSH_DATAh(push, screen->uniforms->offset + (0 << 16)); + PUSH_DATA (push, screen->uniforms->offset + (0 << 16)); + PUSH_DATA (push, (NV50_CB_PVP << 16) | 0x0000); + + BEGIN_NV04(push, NV50_3D(CB_DEF_ADDRESS_HIGH), 3); + PUSH_DATAh(push, screen->uniforms->offset + (1 << 16)); + PUSH_DATA (push, screen->uniforms->offset + (1 << 16)); + PUSH_DATA (push, (NV50_CB_PGP << 16) | 0x0000); + + BEGIN_NV04(push, NV50_3D(CB_DEF_ADDRESS_HIGH), 3); + PUSH_DATAh(push, screen->uniforms->offset + (2 << 16)); + PUSH_DATA (push, screen->uniforms->offset + (2 << 16)); + PUSH_DATA (push, (NV50_CB_PFP << 16) | 0x0000); + + BEGIN_NV04(push, NV50_3D(CB_DEF_ADDRESS_HIGH), 3); + PUSH_DATAh(push, screen->uniforms->offset + (3 << 16)); + PUSH_DATA (push, screen->uniforms->offset + (3 << 16)); + PUSH_DATA (push, (NV50_CB_AUX << 16) | 0x0200); + + BEGIN_NI04(push, NV50_3D(SET_PROGRAM_CB), 6); + PUSH_DATA (push, (NV50_CB_PVP << 12) | 0x001); + PUSH_DATA (push, (NV50_CB_PGP << 12) | 0x021); + PUSH_DATA (push, (NV50_CB_PFP << 12) | 0x031); + PUSH_DATA (push, (NV50_CB_AUX << 12) | 0xf01); + PUSH_DATA (push, (NV50_CB_AUX << 12) | 0xf21); + PUSH_DATA (push, (NV50_CB_AUX << 12) | 0xf31); + + /* max TIC (bits 4:8) & TSC bindings, per program type */ + for (i = 0; i < 3; ++i) { + BEGIN_NV04(push, NV50_3D(TEX_LIMITS(i)), 1); + PUSH_DATA (push, 0x54); + } + + BEGIN_NV04(push, NV50_3D(TIC_ADDRESS_HIGH), 3); + PUSH_DATAh(push, screen->txc->offset); + PUSH_DATA (push, screen->txc->offset); + PUSH_DATA (push, NV50_TIC_MAX_ENTRIES - 1); + + BEGIN_NV04(push, NV50_3D(TSC_ADDRESS_HIGH), 3); + PUSH_DATAh(push, screen->txc->offset + 65536); + PUSH_DATA (push, screen->txc->offset + 65536); + PUSH_DATA (push, NV50_TSC_MAX_ENTRIES - 1); + + BEGIN_NV04(push, NV50_3D(LINKED_TSC), 1); + PUSH_DATA (push, 0); + + BEGIN_NV04(push, NV50_3D(CLIP_RECTS_EN), 1); + PUSH_DATA (push, 0); + BEGIN_NV04(push, NV50_3D(CLIP_RECTS_MODE), 1); + PUSH_DATA (push, NV50_3D_CLIP_RECTS_MODE_INSIDE_ANY); + BEGIN_NV04(push, NV50_3D(CLIP_RECT_HORIZ(0)), 8 * 2); + for (i = 0; i < 8 * 2; ++i) + PUSH_DATA(push, 0); + BEGIN_NV04(push, NV50_3D(CLIPID_ENABLE), 1); + PUSH_DATA (push, 0); + + BEGIN_NV04(push, NV50_3D(VIEWPORT_TRANSFORM_EN), 1); + PUSH_DATA (push, 1); + BEGIN_NV04(push, NV50_3D(DEPTH_RANGE_NEAR(0)), 2); + PUSH_DATAf(push, 0.0f); + PUSH_DATAf(push, 1.0f); + + BEGIN_NV04(push, NV50_3D(VIEW_VOLUME_CLIP_CTRL), 1); +#ifdef NV50_SCISSORS_CLIPPING + PUSH_DATA (push, 0x0000); +#else + PUSH_DATA (push, 0x1080); +#endif + + BEGIN_NV04(push, NV50_3D(CLEAR_FLAGS), 1); + PUSH_DATA (push, NV50_3D_CLEAR_FLAGS_CLEAR_RECT_VIEWPORT); + + /* We use scissors instead of exact view volume clipping, + * so they're always enabled. + */ + BEGIN_NV04(push, NV50_3D(SCISSOR_ENABLE(0)), 3); + PUSH_DATA (push, 1); + PUSH_DATA (push, 8192 << 16); + PUSH_DATA (push, 8192 << 16); + + BEGIN_NV04(push, NV50_3D(RASTERIZE_ENABLE), 1); + PUSH_DATA (push, 1); + BEGIN_NV04(push, NV50_3D(POINT_RASTER_RULES), 1); + PUSH_DATA (push, NV50_3D_POINT_RASTER_RULES_OGL); + BEGIN_NV04(push, NV50_3D(FRAG_COLOR_CLAMP_EN), 1); + PUSH_DATA (push, 0x11111111); + BEGIN_NV04(push, NV50_3D(EDGEFLAG), 1); + PUSH_DATA (push, 1); + + PUSH_KICK (push); + + return 0; } #define FAIL_SCREEN_INIT(str, err) \ @@ -314,13 +504,12 @@ struct pipe_screen * nv50_screen_create(struct nouveau_device *dev) { struct nv50_screen *screen; - struct nouveau_channel *chan; struct pipe_screen *pscreen; + struct nouveau_object *chan; uint64_t value; uint32_t tesla_class; unsigned stack_size, max_warps, tls_space; int ret; - unsigned i, base; screen = CALLOC_STRUCT(nv50_screen); if (!screen) @@ -333,8 +522,10 @@ nv50_screen_create(struct nouveau_device *dev) if (ret) FAIL_SCREEN_INIT("nouveau_screen_init failed: %d\n", ret); + screen->base.pushbuf->user_priv = screen; + screen->base.pushbuf->rsvd_kick = 5; + chan = screen->base.channel; - chan->user_private = screen; pscreen->destroy = nv50_screen_destroy; pscreen->context_create = nv50_create; @@ -348,68 +539,52 @@ nv50_screen_create(struct nouveau_device *dev) nouveau_screen_init_vdec(&screen->base); ret = nouveau_bo_new(dev, NOUVEAU_BO_GART | NOUVEAU_BO_MAP, 0, 4096, - &screen->fence.bo); + NULL, &screen->fence.bo); if (ret) goto fail; - nouveau_bo_map(screen->fence.bo, NOUVEAU_BO_RDWR); + nouveau_bo_map(screen->fence.bo, 0, NULL); screen->fence.map = screen->fence.bo->map; - nouveau_bo_unmap(screen->fence.bo); screen->base.fence.emit = nv50_screen_fence_emit; screen->base.fence.update = nv50_screen_fence_update; - ret = nouveau_notifier_alloc(chan, 0xbeef0301, 1, &screen->sync); + ret = nouveau_object_new(chan, 0xbeef0301, NOUVEAU_NOTIFIER_CLASS, + &(struct nv04_notify){ .length = 32 }, + sizeof(struct nv04_notify), &screen->sync); if (ret) FAIL_SCREEN_INIT("Error allocating notifier: %d\n", ret); - ret = nouveau_grobj_alloc(chan, 0xbeef5039, NV50_M2MF, &screen->m2mf); + + ret = nouveau_object_new(chan, 0xbeef5039, NV50_M2MF_CLASS, + NULL, 0, &screen->m2mf); if (ret) FAIL_SCREEN_INIT("Error allocating PGRAPH context for M2MF: %d\n", ret); - BIND_RING (chan, screen->m2mf, NV50_SUBCH_MF); - BEGIN_RING(chan, RING_MF_(NV04_M2MF_DMA_NOTIFY), 3); - OUT_RING (chan, screen->sync->handle); - OUT_RING (chan, chan->vram->handle); - OUT_RING (chan, chan->vram->handle); - ret = nouveau_grobj_alloc(chan, 0xbeef502d, NV50_2D, &screen->eng2d); + ret = nouveau_object_new(chan, 0xbeef502d, NV50_2D_CLASS, + NULL, 0, &screen->eng2d); if (ret) FAIL_SCREEN_INIT("Error allocating PGRAPH context for 2D: %d\n", ret); - BIND_RING (chan, screen->eng2d, NV50_SUBCH_2D); - BEGIN_RING(chan, RING_2D(DMA_NOTIFY), 4); - OUT_RING (chan, screen->sync->handle); - OUT_RING (chan, chan->vram->handle); - OUT_RING (chan, chan->vram->handle); - OUT_RING (chan, chan->vram->handle); - BEGIN_RING(chan, RING_2D(OPERATION), 1); - OUT_RING (chan, NV50_2D_OPERATION_SRCCOPY); - BEGIN_RING(chan, RING_2D(CLIP_ENABLE), 1); - OUT_RING (chan, 0); - BEGIN_RING(chan, RING_2D(COLOR_KEY_ENABLE), 1); - OUT_RING (chan, 0); - BEGIN_RING(chan, RING_2D_(0x0888), 1); - OUT_RING (chan, 1); - switch (dev->chipset & 0xf0) { case 0x50: - tesla_class = NV50_3D; + tesla_class = NV50_3D_CLASS; break; case 0x80: case 0x90: - tesla_class = NV84_3D; + tesla_class = NV84_3D_CLASS; break; case 0xa0: switch (dev->chipset) { case 0xa0: case 0xaa: case 0xac: - tesla_class = NVA0_3D; + tesla_class = NVA0_3D_CLASS; break; case 0xaf: - tesla_class = NVAF_3D; + tesla_class = NVAF_3D_CLASS; break; default: - tesla_class = NVA3_3D; + tesla_class = NVA3_3D_CLASS; break; } break; @@ -418,98 +593,33 @@ nv50_screen_create(struct nouveau_device *dev) break; } - ret = nouveau_grobj_alloc(chan, 0xbeef5097, tesla_class, &screen->tesla); + ret = nouveau_object_new(chan, 0xbeef5097, tesla_class, + NULL, 0, &screen->tesla); if (ret) FAIL_SCREEN_INIT("Error allocating PGRAPH context for 3D: %d\n", ret); - BIND_RING (chan, screen->tesla, NV50_SUBCH_3D); - - BEGIN_RING(chan, RING_3D(COND_MODE), 1); - OUT_RING (chan, NV50_3D_COND_MODE_ALWAYS); - - BEGIN_RING(chan, RING_3D(DMA_NOTIFY), 1); - OUT_RING (chan, screen->sync->handle); - BEGIN_RING(chan, RING_3D(DMA_ZETA), 11); - for (i = 0; i < 11; ++i) - OUT_RING(chan, chan->vram->handle); - BEGIN_RING(chan, RING_3D(DMA_COLOR(0)), NV50_3D_DMA_COLOR__LEN); - for (i = 0; i < NV50_3D_DMA_COLOR__LEN; ++i) - OUT_RING(chan, chan->vram->handle); - - BEGIN_RING(chan, RING_3D(REG_MODE), 1); - OUT_RING (chan, NV50_3D_REG_MODE_STRIPED); - BEGIN_RING(chan, RING_3D(UNK1400_LANES), 1); - OUT_RING (chan, 0xf); - - BEGIN_RING(chan, RING_3D(RT_CONTROL), 1); - OUT_RING (chan, 1); - - BEGIN_RING(chan, RING_3D(CSAA_ENABLE), 1); - OUT_RING (chan, 0); - BEGIN_RING(chan, RING_3D(MULTISAMPLE_ENABLE), 1); - OUT_RING (chan, 0); - BEGIN_RING(chan, RING_3D(MULTISAMPLE_MODE), 1); - OUT_RING (chan, NV50_3D_MULTISAMPLE_MODE_MS1); - BEGIN_RING(chan, RING_3D(MULTISAMPLE_CTRL), 1); - OUT_RING (chan, 0); - BEGIN_RING(chan, RING_3D(LINE_LAST_PIXEL), 1); - OUT_RING (chan, 0); - BEGIN_RING(chan, RING_3D(BLEND_SEPARATE_ALPHA), 1); - OUT_RING (chan, 1); - - if (tesla_class >= NVA0_3D) { - BEGIN_RING(chan, RING_3D_(NVA0_3D_TEX_MISC), 1); - OUT_RING (chan, NVA0_3D_TEX_MISC_SEAMLESS_CUBE_MAP); - } - - BEGIN_RING(chan, RING_3D(SCREEN_Y_CONTROL), 1); - OUT_RING (chan, 0); - BEGIN_RING(chan, RING_3D(WINDOW_OFFSET_X), 2); - OUT_RING (chan, 0); - OUT_RING (chan, 0); - BEGIN_RING(chan, RING_3D(ZCULL_REGION), 1); /* deactivate ZCULL */ - OUT_RING (chan, 0x3f); ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM, 1 << 16, - 3 << NV50_CODE_BO_SIZE_LOG2, &screen->code); + 3 << NV50_CODE_BO_SIZE_LOG2, NULL, &screen->code); if (ret) goto fail; - nouveau_resource_init(&screen->vp_code_heap, 0, 1 << NV50_CODE_BO_SIZE_LOG2); - nouveau_resource_init(&screen->gp_code_heap, 0, 1 << NV50_CODE_BO_SIZE_LOG2); - nouveau_resource_init(&screen->fp_code_heap, 0, 1 << NV50_CODE_BO_SIZE_LOG2); - - base = 1 << NV50_CODE_BO_SIZE_LOG2; - - BEGIN_RING(chan, RING_3D(VP_ADDRESS_HIGH), 2); - OUT_RELOCh(chan, screen->code, base * 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD); - OUT_RELOCl(chan, screen->code, base * 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD); - - BEGIN_RING(chan, RING_3D(FP_ADDRESS_HIGH), 2); - OUT_RELOCh(chan, screen->code, base * 1, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD); - OUT_RELOCl(chan, screen->code, base * 1, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD); - - BEGIN_RING(chan, RING_3D(GP_ADDRESS_HIGH), 2); - OUT_RELOCh(chan, screen->code, base * 2, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD); - OUT_RELOCl(chan, screen->code, base * 2, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD); + nouveau_heap_init(&screen->vp_code_heap, 0, 1 << NV50_CODE_BO_SIZE_LOG2); + nouveau_heap_init(&screen->gp_code_heap, 0, 1 << NV50_CODE_BO_SIZE_LOG2); + nouveau_heap_init(&screen->fp_code_heap, 0, 1 << NV50_CODE_BO_SIZE_LOG2); - nouveau_device_get_param(dev, NOUVEAU_GETPARAM_GRAPH_UNITS, &value); + nouveau_getparam(dev, NOUVEAU_GETPARAM_GRAPH_UNITS, &value); max_warps = util_bitcount(value & 0xffff); max_warps *= util_bitcount((value >> 24) & 0xf) * 32; stack_size = max_warps * 64 * 8; - ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM, 1 << 16, stack_size, + ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM, 1 << 16, stack_size, NULL, &screen->stack_bo); if (ret) FAIL_SCREEN_INIT("Failed to allocate stack bo: %d\n", ret); - BEGIN_RING(chan, RING_3D(STACK_ADDRESS_HIGH), 3); - OUT_RELOCh(chan, screen->stack_bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RDWR); - OUT_RELOCl(chan, screen->stack_bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RDWR); - OUT_RING (chan, 4); - tls_space = NV50_CAP_MAX_PROGRAM_TEMPS * 16; screen->tls_size = tls_space * max_warps * 32; @@ -518,126 +628,32 @@ nv50_screen_create(struct nouveau_device *dev) debug_printf("max_warps = %i, tls_size = %"PRIu64" KiB\n", max_warps, screen->tls_size >> 10); - ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM, 1 << 16, screen->tls_size, + ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM, 1 << 16, screen->tls_size, NULL, &screen->tls_bo); if (ret) FAIL_SCREEN_INIT("Failed to allocate stack bo: %d\n", ret); - BEGIN_RING(chan, RING_3D(LOCAL_ADDRESS_HIGH), 3); - OUT_RELOCh(chan, screen->tls_bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RDWR); - OUT_RELOCl(chan, screen->tls_bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RDWR); - OUT_RING (chan, util_logbase2(tls_space / 8)); - ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM, 1 << 16, 4 << 16, + ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM, 1 << 16, 4 << 16, NULL, &screen->uniforms); if (ret) goto fail; - BEGIN_RING(chan, RING_3D(CB_DEF_ADDRESS_HIGH), 3); - OUT_RELOCh(chan, screen->uniforms, 0 << 16, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD); - OUT_RELOCl(chan, screen->uniforms, 0 << 16, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD); - OUT_RING (chan, (NV50_CB_PVP << 16) | 0x0000); - - BEGIN_RING(chan, RING_3D(CB_DEF_ADDRESS_HIGH), 3); - OUT_RELOCh(chan, screen->uniforms, 1 << 16, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD); - OUT_RELOCl(chan, screen->uniforms, 1 << 16, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD); - OUT_RING (chan, (NV50_CB_PGP << 16) | 0x0000); - - BEGIN_RING(chan, RING_3D(CB_DEF_ADDRESS_HIGH), 3); - OUT_RELOCh(chan, screen->uniforms, 2 << 16, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD); - OUT_RELOCl(chan, screen->uniforms, 2 << 16, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD); - OUT_RING (chan, (NV50_CB_PFP << 16) | 0x0000); - - BEGIN_RING(chan, RING_3D(CB_DEF_ADDRESS_HIGH), 3); - OUT_RELOCh(chan, screen->uniforms, 3 << 16, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD); - OUT_RELOCl(chan, screen->uniforms, 3 << 16, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD); - OUT_RING (chan, (NV50_CB_AUX << 16) | 0x0200); - - BEGIN_RING_NI(chan, RING_3D(SET_PROGRAM_CB), 6); - OUT_RING (chan, (NV50_CB_PVP << 12) | 0x001); - OUT_RING (chan, (NV50_CB_PGP << 12) | 0x021); - OUT_RING (chan, (NV50_CB_PFP << 12) | 0x031); - OUT_RING (chan, (NV50_CB_AUX << 12) | 0xf01); - OUT_RING (chan, (NV50_CB_AUX << 12) | 0xf21); - OUT_RING (chan, (NV50_CB_AUX << 12) | 0xf31); - - ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM, 1 << 16, 3 << 16, + ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM, 1 << 16, 3 << 16, NULL, &screen->txc); if (ret) FAIL_SCREEN_INIT("Could not allocate TIC/TSC bo: %d\n", ret); - /* max TIC (bits 4:8) & TSC bindings, per program type */ - for (i = 0; i < 3; ++i) { - BEGIN_RING(chan, RING_3D(TEX_LIMITS(i)), 1); - OUT_RING (chan, 0x54); - } - - BEGIN_RING(chan, RING_3D(TIC_ADDRESS_HIGH), 3); - OUT_RELOCh(chan, screen->txc, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD); - OUT_RELOCl(chan, screen->txc, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD); - OUT_RING (chan, NV50_TIC_MAX_ENTRIES - 1); - - BEGIN_RING(chan, RING_3D(TSC_ADDRESS_HIGH), 3); - OUT_RELOCh(chan, screen->txc, 65536, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD); - OUT_RELOCl(chan, screen->txc, 65536, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD); - OUT_RING (chan, NV50_TSC_MAX_ENTRIES - 1); - - BEGIN_RING(chan, RING_3D(LINKED_TSC), 1); - OUT_RING (chan, 0); - - BEGIN_RING(chan, RING_3D(CLIP_RECTS_EN), 1); - OUT_RING (chan, 0); - BEGIN_RING(chan, RING_3D(CLIP_RECTS_MODE), 1); - OUT_RING (chan, NV50_3D_CLIP_RECTS_MODE_INSIDE_ANY); - BEGIN_RING(chan, RING_3D(CLIP_RECT_HORIZ(0)), 8 * 2); - for (i = 0; i < 8 * 2; ++i) - OUT_RING(chan, 0); - BEGIN_RING(chan, RING_3D(CLIPID_ENABLE), 1); - OUT_RING (chan, 0); - - BEGIN_RING(chan, RING_3D(VIEWPORT_TRANSFORM_EN), 1); - OUT_RING (chan, 1); - BEGIN_RING(chan, RING_3D(DEPTH_RANGE_NEAR(0)), 2); - OUT_RINGf (chan, 0.0f); - OUT_RINGf (chan, 1.0f); - - BEGIN_RING(chan, RING_3D(VIEW_VOLUME_CLIP_CTRL), 1); -#ifdef NV50_SCISSORS_CLIPPING - OUT_RING (chan, 0x0000); -#else - OUT_RING (chan, 0x1080); -#endif - - BEGIN_RING(chan, RING_3D(CLEAR_FLAGS), 1); - OUT_RING (chan, NV50_3D_CLEAR_FLAGS_CLEAR_RECT_VIEWPORT); - - /* We use scissors instead of exact view volume clipping, - * so they're always enabled. - */ - BEGIN_RING(chan, RING_3D(SCISSOR_ENABLE(0)), 3); - OUT_RING (chan, 1); - OUT_RING (chan, 8192 << 16); - OUT_RING (chan, 8192 << 16); - - BEGIN_RING(chan, RING_3D(RASTERIZE_ENABLE), 1); - OUT_RING (chan, 1); - BEGIN_RING(chan, RING_3D(POINT_RASTER_RULES), 1); - OUT_RING (chan, NV50_3D_POINT_RASTER_RULES_OGL); - BEGIN_RING(chan, RING_3D(FRAG_COLOR_CLAMP_EN), 1); - OUT_RING (chan, 0x11111111); - BEGIN_RING(chan, RING_3D(EDGEFLAG_ENABLE), 1); - OUT_RING (chan, 1); - - FIRE_RING (chan); - screen->tic.entries = CALLOC(4096, sizeof(void *)); screen->tsc.entries = screen->tic.entries + 2048; - screen->mm_VRAM_fe0 = nouveau_mm_create(dev, NOUVEAU_BO_VRAM, 0xfe0); if (!nv50_blitctx_create(screen)) goto fail; + if (nv50_screen_init_hwctx(screen, tls_space)) + goto fail; + nouveau_fence_new(&screen->base, &screen->base.fence.current, FALSE); return pscreen; @@ -647,21 +663,6 @@ fail: return NULL; } -void -nv50_screen_make_buffers_resident(struct nv50_screen *screen) -{ - struct nouveau_channel *chan = screen->base.channel; - - const unsigned flags = NOUVEAU_BO_VRAM | NOUVEAU_BO_RD; - - MARK_RING(chan, 0, 5); - nouveau_bo_validate(chan, screen->code, flags); - nouveau_bo_validate(chan, screen->uniforms, flags); - nouveau_bo_validate(chan, screen->txc, flags); - nouveau_bo_validate(chan, screen->tls_bo, flags); - nouveau_bo_validate(chan, screen->stack_bo, flags); -} - int nv50_screen_tic_alloc(struct nv50_screen *screen, void *entry) { diff --git a/src/gallium/drivers/nv50/nv50_screen.h b/src/gallium/drivers/nv50/nv50_screen.h index cb7e956cd84..b7803cfed80 100644 --- a/src/gallium/drivers/nv50/nv50_screen.h +++ b/src/gallium/drivers/nv50/nv50_screen.h @@ -1,11 +1,11 @@ #ifndef __NV50_SCREEN_H__ #define __NV50_SCREEN_H__ -#define NOUVEAU_NVC0 #include "nouveau/nouveau_screen.h" #include "nouveau/nouveau_fence.h" #include "nouveau/nouveau_mm.h" -#undef NOUVEAU_NVC0 +#include "nouveau/nouveau_heap.h" + #include "nv50_winsys.h" #include "nv50_stateobj.h" @@ -36,9 +36,9 @@ struct nv50_screen { uint64_t tls_size; - struct nouveau_resource *vp_code_heap; - struct nouveau_resource *gp_code_heap; - struct nouveau_resource *fp_code_heap; + struct nouveau_heap *vp_code_heap; + struct nouveau_heap *gp_code_heap; + struct nouveau_heap *fp_code_heap; struct nv50_blitctx *blitctx; @@ -59,13 +59,11 @@ struct nv50_screen { struct nouveau_bo *bo; } fence; - struct nouveau_notifier *sync; - - struct nouveau_mman *mm_VRAM_fe0; + struct nouveau_object *sync; - struct nouveau_grobj *tesla; - struct nouveau_grobj *eng2d; - struct nouveau_grobj *m2mf; + struct nouveau_object *tesla; + struct nouveau_object *eng2d; + struct nouveau_object *m2mf; }; static INLINE struct nv50_screen * @@ -76,8 +74,6 @@ nv50_screen(struct pipe_screen *screen) boolean nv50_blitctx_create(struct nv50_screen *); -void nv50_screen_make_buffers_resident(struct nv50_screen *); - int nv50_screen_tic_alloc(struct nv50_screen *, void *); int nv50_screen_tsc_alloc(struct nv50_screen *, void *); @@ -88,7 +84,6 @@ nv50_resource_fence(struct nv04_resource *res, uint32_t flags) if (res->mm) { nouveau_fence_ref(screen->base.fence.current, &res->fence); - if (flags & NOUVEAU_BO_WR) nouveau_fence_ref(screen->base.fence.current, &res->fence_wr); } @@ -97,11 +92,7 @@ nv50_resource_fence(struct nv04_resource *res, uint32_t flags) static INLINE void nv50_resource_validate(struct nv04_resource *res, uint32_t flags) { - struct nv50_screen *screen = nv50_screen(res->base.screen); - if (likely(res->bo)) { - nouveau_bo_validate(screen->base.channel, res->bo, flags); - if (flags & NOUVEAU_BO_WR) res->status |= NOUVEAU_BUFFER_STATUS_GPU_WRITING; if (flags & NOUVEAU_BO_RD) diff --git a/src/gallium/drivers/nv50/nv50_shader_state.c b/src/gallium/drivers/nv50/nv50_shader_state.c index d3cdb697960..07d4e6bb0c5 100644 --- a/src/gallium/drivers/nv50/nv50_shader_state.c +++ b/src/gallium/drivers/nv50/nv50_shader_state.c @@ -31,7 +31,7 @@ void nv50_constbufs_validate(struct nv50_context *nv50) { - struct nouveau_channel *chan = nv50->screen->base.channel; + struct nouveau_pushbuf *push = nv50->base.pushbuf; unsigned s; for (s = 0; s < 3; ++s) { @@ -58,8 +58,8 @@ nv50_constbufs_validate(struct nv50_context *nv50) res = nv04_resource(nv50->constbuf[s][i]); if (!res) { if (i != 0) { - BEGIN_RING(chan, RING_3D(SET_PROGRAM_CB), 1); - OUT_RING (chan, (i << 8) | p | 0); + BEGIN_NV04(push, NV50_3D(SET_PROGRAM_CB), 1); + PUSH_DATA (push, (i << 8) | p | 0); } continue; } @@ -78,43 +78,35 @@ nv50_constbufs_validate(struct nv50_context *nv50) if (!nouveau_resource_mapped_by_gpu(&res->base)) { nouveau_buffer_migrate(&nv50->base, res, NOUVEAU_BO_VRAM); - BEGIN_RING(chan, RING_3D(CODE_CB_FLUSH), 1); - OUT_RING (chan, 0); + BEGIN_NV04(push, NV50_3D(CODE_CB_FLUSH), 1); + PUSH_DATA (push, 0); } - MARK_RING (chan, 6, 2); - BEGIN_RING(chan, RING_3D(CB_DEF_ADDRESS_HIGH), 3); - OUT_RESRCh(chan, res, 0, NOUVEAU_BO_RD); - OUT_RESRCl(chan, res, 0, NOUVEAU_BO_RD); - OUT_RING (chan, (b << 16) | (res->base.width0 & 0xffff)); - BEGIN_RING(chan, RING_3D(SET_PROGRAM_CB), 1); - OUT_RING (chan, (b << 12) | (i << 8) | p | 1); + BEGIN_NV04(push, NV50_3D(CB_DEF_ADDRESS_HIGH), 3); + PUSH_DATAh(push, res->address); + PUSH_DATA (push, res->address); + PUSH_DATA (push, (b << 16) | (res->base.width0 & 0xffff)); + BEGIN_NV04(push, NV50_3D(SET_PROGRAM_CB), 1); + PUSH_DATA (push, (b << 12) | (i << 8) | p | 1); bo = res->bo; - - nv50_bufctx_add_resident(nv50, NV50_BUFCTX_CONSTANT, res, - res->domain | NOUVEAU_BO_RD); } - if (words) { - MARK_RING(chan, 8, 1); - - nouveau_bo_validate(chan, bo, res->domain | NOUVEAU_BO_WR); - } + if (bo != nv50->screen->uniforms) + BCTX_REFN(nv50->bufctx_3d, CB(s, i), res, RD); while (words) { - unsigned nr = AVAIL_RING(chan); + unsigned nr; - if (nr < 16) { - FIRE_RING(chan); - nouveau_bo_validate(chan, bo, res->domain | NOUVEAU_BO_WR); - continue; - } + if (!PUSH_SPACE(push, 16)) + break; + nr = PUSH_AVAIL(push); + assert(nr >= 16); nr = MIN2(MIN2(nr - 3, words), NV04_PFIFO_MAX_PACKET_LEN); - BEGIN_RING(chan, RING_3D(CB_ADDR), 1); - OUT_RING (chan, (start << 8) | b); - BEGIN_RING_NI(chan, RING_3D(CB_DATA(0)), nr); - OUT_RINGp (chan, &res->data[start * 4], nr); + BEGIN_NV04(push, NV50_3D(CB_ADDR), 1); + PUSH_DATA (push, (start << 8) | b); + BEGIN_NI04(push, NV50_3D(CB_DATA(0)), nr); + PUSH_DATAp(push, &res->data[start * 4], nr); start += nr; words -= nr; @@ -126,7 +118,7 @@ nv50_constbufs_validate(struct nv50_context *nv50) static boolean nv50_program_validate(struct nv50_context *nv50, struct nv50_program *prog) { - struct nouveau_resource *heap; + struct nouveau_heap *heap; int ret; unsigned size; @@ -135,7 +127,7 @@ nv50_program_validate(struct nv50_context *nv50, struct nv50_program *prog) if (!prog->translated) return FALSE; } else - if (prog->res) + if (prog->mem) return TRUE; if (prog->type == PIPE_SHADER_FRAGMENT) heap = nv50->screen->fp_code_heap; @@ -146,12 +138,12 @@ nv50_program_validate(struct nv50_context *nv50, struct nv50_program *prog) size = align(prog->code_size, 0x100); - ret = nouveau_resource_alloc(heap, size, prog, &prog->res); + ret = nouveau_heap_alloc(heap, size, prog, &prog->mem); if (ret) { NOUVEAU_ERR("out of code space for shader type %i\n", prog->type); return FALSE; } - prog->code_base = prog->res->start; + prog->code_base = prog->mem->start; nv50_relocate_program(prog, prog->code_base, 0); @@ -159,80 +151,99 @@ nv50_program_validate(struct nv50_context *nv50, struct nv50_program *prog) (prog->type << NV50_CODE_BO_SIZE_LOG2) + prog->code_base, NOUVEAU_BO_VRAM, prog->code_size, prog->code); - BEGIN_RING(nv50->screen->base.channel, RING_3D(CODE_CB_FLUSH), 1); - OUT_RING (nv50->screen->base.channel, 0); + BEGIN_NV04(nv50->base.pushbuf, NV50_3D(CODE_CB_FLUSH), 1); + PUSH_DATA (nv50->base.pushbuf, 0); return TRUE; } +static INLINE void +nv50_program_update_context_state(struct nv50_context *nv50, + struct nv50_program *prog, int stage) +{ + const unsigned flags = NOUVEAU_BO_VRAM | NOUVEAU_BO_RDWR; + + if (prog && prog->uses_lmem) { + if (!nv50->state.tls_required) + BCTX_REFN_bo(nv50->bufctx_3d, TLS, flags, nv50->screen->tls_bo); + nv50->state.tls_required |= 1 << stage; + } else { + if (nv50->state.tls_required == (1 << stage)) + nouveau_bufctx_reset(nv50->bufctx_3d, NV50_BIND_TLS); + nv50->state.tls_required &= ~(1 << stage); + } +} + void nv50_vertprog_validate(struct nv50_context *nv50) { - struct nouveau_channel *chan = nv50->screen->base.channel; + struct nouveau_pushbuf *push = nv50->base.pushbuf; struct nv50_program *vp = nv50->vertprog; if (!nv50_program_validate(nv50, vp)) return; - - BEGIN_RING(chan, RING_3D(VP_ATTR_EN(0)), 2); - OUT_RING (chan, vp->vp.attrs[0]); - OUT_RING (chan, vp->vp.attrs[1]); - BEGIN_RING(chan, RING_3D(VP_REG_ALLOC_RESULT), 1); - OUT_RING (chan, vp->max_out); - BEGIN_RING(chan, RING_3D(VP_REG_ALLOC_TEMP), 1); - OUT_RING (chan, vp->max_gpr); - BEGIN_RING(chan, RING_3D(VP_START_ID), 1); - OUT_RING (chan, vp->code_base); + nv50_program_update_context_state(nv50, vp, 0); + + BEGIN_NV04(push, NV50_3D(VP_ATTR_EN(0)), 2); + PUSH_DATA (push, vp->vp.attrs[0]); + PUSH_DATA (push, vp->vp.attrs[1]); + BEGIN_NV04(push, NV50_3D(VP_REG_ALLOC_RESULT), 1); + PUSH_DATA (push, vp->max_out); + BEGIN_NV04(push, NV50_3D(VP_REG_ALLOC_TEMP), 1); + PUSH_DATA (push, vp->max_gpr); + BEGIN_NV04(push, NV50_3D(VP_START_ID), 1); + PUSH_DATA (push, vp->code_base); } void nv50_fragprog_validate(struct nv50_context *nv50) { - struct nouveau_channel *chan = nv50->screen->base.channel; + struct nouveau_pushbuf *push = nv50->base.pushbuf; struct nv50_program *fp = nv50->fragprog; if (!nv50_program_validate(nv50, fp)) return; - - BEGIN_RING(chan, RING_3D(FP_REG_ALLOC_TEMP), 1); - OUT_RING (chan, fp->max_gpr); - BEGIN_RING(chan, RING_3D(FP_RESULT_COUNT), 1); - OUT_RING (chan, fp->max_out); - BEGIN_RING(chan, RING_3D(FP_CONTROL), 1); - OUT_RING (chan, fp->fp.flags[0]); - BEGIN_RING(chan, RING_3D(FP_CTRL_UNK196C), 1); - OUT_RING (chan, fp->fp.flags[1]); - BEGIN_RING(chan, RING_3D(FP_START_ID), 1); - OUT_RING (chan, fp->code_base); + nv50_program_update_context_state(nv50, fp, 1); + + BEGIN_NV04(push, NV50_3D(FP_REG_ALLOC_TEMP), 1); + PUSH_DATA (push, fp->max_gpr); + BEGIN_NV04(push, NV50_3D(FP_RESULT_COUNT), 1); + PUSH_DATA (push, fp->max_out); + BEGIN_NV04(push, NV50_3D(FP_CONTROL), 1); + PUSH_DATA (push, fp->fp.flags[0]); + BEGIN_NV04(push, NV50_3D(FP_CTRL_UNK196C), 1); + PUSH_DATA (push, fp->fp.flags[1]); + BEGIN_NV04(push, NV50_3D(FP_START_ID), 1); + PUSH_DATA (push, fp->code_base); } void nv50_gmtyprog_validate(struct nv50_context *nv50) { - struct nouveau_channel *chan = nv50->screen->base.channel; + struct nouveau_pushbuf *push = nv50->base.pushbuf; struct nv50_program *gp = nv50->gmtyprog; - if (!gp) /* GP_ENABLE is updated in linkage validation */ - return; - if (!nv50_program_validate(nv50, gp)) - return; + if (gp) { + BEGIN_NV04(push, NV50_3D(GP_REG_ALLOC_TEMP), 1); + PUSH_DATA (push, gp->max_gpr); + BEGIN_NV04(push, NV50_3D(GP_REG_ALLOC_RESULT), 1); + PUSH_DATA (push, gp->max_out); + BEGIN_NV04(push, NV50_3D(GP_OUTPUT_PRIMITIVE_TYPE), 1); + PUSH_DATA (push, gp->gp.prim_type); + BEGIN_NV04(push, NV50_3D(GP_VERTEX_OUTPUT_COUNT), 1); + PUSH_DATA (push, gp->gp.vert_count); + BEGIN_NV04(push, NV50_3D(GP_START_ID), 1); + PUSH_DATA (push, gp->code_base); + } + nv50_program_update_context_state(nv50, gp, 2); - BEGIN_RING(chan, RING_3D(GP_REG_ALLOC_TEMP), 1); - OUT_RING (chan, gp->max_gpr); - BEGIN_RING(chan, RING_3D(GP_REG_ALLOC_RESULT), 1); - OUT_RING (chan, gp->max_out); - BEGIN_RING(chan, RING_3D(GP_OUTPUT_PRIMITIVE_TYPE), 1); - OUT_RING (chan, gp->gp.prim_type); - BEGIN_RING(chan, RING_3D(GP_VERTEX_OUTPUT_COUNT), 1); - OUT_RING (chan, gp->gp.vert_count); - BEGIN_RING(chan, RING_3D(GP_START_ID), 1); - OUT_RING (chan, gp->code_base); + /* GP_ENABLE is updated in linkage validation */ } static void nv50_sprite_coords_validate(struct nv50_context *nv50) { - struct nouveau_channel *chan = nv50->screen->base.channel; + struct nouveau_pushbuf *push = nv50->base.pushbuf; uint32_t pntc[8], mode; struct nv50_program *fp = nv50->fragprog; unsigned i, c; @@ -240,9 +251,9 @@ nv50_sprite_coords_validate(struct nv50_context *nv50) if (!nv50->rast->pipe.point_quad_rasterization) { if (nv50->state.point_sprite) { - BEGIN_RING(chan, RING_3D(POINT_COORD_REPLACE_MAP(0)), 8); + BEGIN_NV04(push, NV50_3D(POINT_COORD_REPLACE_MAP(0)), 8); for (i = 0; i < 8; ++i) - OUT_RING(chan, 0); + PUSH_DATA(push, 0); nv50->state.point_sprite = FALSE; } @@ -278,43 +289,43 @@ nv50_sprite_coords_validate(struct nv50_context *nv50) else mode = 0x10; - BEGIN_RING(chan, RING_3D(POINT_SPRITE_CTRL), 1); - OUT_RING (chan, mode); + BEGIN_NV04(push, NV50_3D(POINT_SPRITE_CTRL), 1); + PUSH_DATA (push, mode); - BEGIN_RING(chan, RING_3D(POINT_COORD_REPLACE_MAP(0)), 8); - OUT_RINGp (chan, pntc, 8); + BEGIN_NV04(push, NV50_3D(POINT_COORD_REPLACE_MAP(0)), 8); + PUSH_DATAp(push, pntc, 8); } /* Validate state derived from shaders and the rasterizer cso. */ void nv50_validate_derived_rs(struct nv50_context *nv50) { - struct nouveau_channel *chan = nv50->screen->base.channel; + struct nouveau_pushbuf *push = nv50->base.pushbuf; uint32_t color, psize; nv50_sprite_coords_validate(nv50); if (nv50->dirty & NV50_NEW_FRAGPROG) return; - psize = nv50->state.semantic_psize & ~NV50_3D_MAP_SEMANTIC_3_PTSZ_EN__MASK; - color = nv50->state.semantic_color & ~NV50_3D_MAP_SEMANTIC_0_CLMP_EN; + psize = nv50->state.semantic_psize & ~NV50_3D_SEMANTIC_PTSZ_PTSZ_EN__MASK; + color = nv50->state.semantic_color & ~NV50_3D_SEMANTIC_COLOR_CLMP_EN; if (nv50->rast->pipe.clamp_vertex_color) - color |= NV50_3D_MAP_SEMANTIC_0_CLMP_EN; + color |= NV50_3D_SEMANTIC_COLOR_CLMP_EN; if (color != nv50->state.semantic_color) { nv50->state.semantic_color = color; - BEGIN_RING(chan, RING_3D(MAP_SEMANTIC_0), 1); - OUT_RING (chan, color); + BEGIN_NV04(push, NV50_3D(SEMANTIC_COLOR), 1); + PUSH_DATA (push, color); } if (nv50->rast->pipe.point_size_per_vertex) - psize |= NV50_3D_MAP_SEMANTIC_3_PTSZ_EN__MASK; + psize |= NV50_3D_SEMANTIC_PTSZ_PTSZ_EN__MASK; if (psize != nv50->state.semantic_psize) { nv50->state.semantic_psize = psize; - BEGIN_RING(chan, RING_3D(MAP_SEMANTIC_3), 1); - OUT_RING (chan, psize); + BEGIN_NV04(push, NV50_3D(SEMANTIC_PTSZ), 1); + PUSH_DATA (push, psize); } } @@ -348,7 +359,7 @@ nv50_vec4_map(uint8_t *map, int mid, uint32_t lin[4], void nv50_fp_linkage_validate(struct nv50_context *nv50) { - struct nouveau_channel *chan = nv50->screen->base.channel; + struct nouveau_pushbuf *push = nv50->base.pushbuf; struct nv50_program *vp = nv50->gmtyprog ? nv50->gmtyprog : nv50->vertprog; struct nv50_program *fp = nv50->fragprog; struct nv50_varying dummy; @@ -409,48 +420,48 @@ nv50_fp_linkage_validate(struct nv50_context *nv50) } if (nv50->rast->pipe.clamp_vertex_color) - colors |= NV50_3D_MAP_SEMANTIC_0_CLMP_EN; + colors |= NV50_3D_SEMANTIC_COLOR_CLMP_EN; n = (m + 3) / 4; assert(m <= 64); if (unlikely(nv50->gmtyprog)) { - BEGIN_RING(chan, RING_3D(GP_RESULT_MAP_SIZE), 1); - OUT_RING (chan, m); - BEGIN_RING(chan, RING_3D(GP_RESULT_MAP(0)), n); - OUT_RINGp (chan, map, n); + BEGIN_NV04(push, NV50_3D(GP_RESULT_MAP_SIZE), 1); + PUSH_DATA (push, m); + BEGIN_NV04(push, NV50_3D(GP_RESULT_MAP(0)), n); + PUSH_DATAp(push, map, n); } else { - BEGIN_RING(chan, RING_3D(VP_GP_BUILTIN_ATTR_EN), 1); - OUT_RING (chan, vp->vp.attrs[2]); + BEGIN_NV04(push, NV50_3D(VP_GP_BUILTIN_ATTR_EN), 1); + PUSH_DATA (push, vp->vp.attrs[2]); - BEGIN_RING(chan, RING_3D(MAP_SEMANTIC_4), 1); - OUT_RING (chan, primid); + BEGIN_NV04(push, NV50_3D(SEMANTIC_PRIM_ID), 1); + PUSH_DATA (push, primid); - BEGIN_RING(chan, RING_3D(VP_RESULT_MAP_SIZE), 1); - OUT_RING (chan, m); - BEGIN_RING(chan, RING_3D(VP_RESULT_MAP(0)), n); - OUT_RINGp (chan, map, n); + BEGIN_NV04(push, NV50_3D(VP_RESULT_MAP_SIZE), 1); + PUSH_DATA (push, m); + BEGIN_NV04(push, NV50_3D(VP_RESULT_MAP(0)), n); + PUSH_DATAp(push, map, n); } - BEGIN_RING(chan, RING_3D(MAP_SEMANTIC_0), 4); - OUT_RING (chan, colors); - OUT_RING (chan, (vp->vp.clpd_nr << 8) | 4); - OUT_RING (chan, 0); - OUT_RING (chan, psiz); + BEGIN_NV04(push, NV50_3D(SEMANTIC_COLOR), 4); + PUSH_DATA (push, colors); + PUSH_DATA (push, (vp->vp.clpd_nr << 8) | 4); + PUSH_DATA (push, 0); + PUSH_DATA (push, psiz); - BEGIN_RING(chan, RING_3D(FP_INTERPOLANT_CTRL), 1); - OUT_RING (chan, interp); + BEGIN_NV04(push, NV50_3D(FP_INTERPOLANT_CTRL), 1); + PUSH_DATA (push, interp); nv50->state.interpolant_ctrl = interp; nv50->state.semantic_color = colors; nv50->state.semantic_psize = psiz; - BEGIN_RING(chan, RING_3D(NOPERSPECTIVE_BITMAP(0)), 4); - OUT_RINGp (chan, lin, 4); + BEGIN_NV04(push, NV50_3D(NOPERSPECTIVE_BITMAP(0)), 4); + PUSH_DATAp(push, lin, 4); - BEGIN_RING(chan, RING_3D(GP_ENABLE), 1); - OUT_RING (chan, nv50->gmtyprog ? 1 : 0); + BEGIN_NV04(push, NV50_3D(GP_ENABLE), 1); + PUSH_DATA (push, nv50->gmtyprog ? 1 : 0); } static int @@ -486,7 +497,7 @@ nv50_vp_gp_mapping(uint8_t *map, int m, void nv50_gp_linkage_validate(struct nv50_context *nv50) { - struct nouveau_channel *chan = nv50->screen->base.channel; + struct nouveau_pushbuf *push = nv50->base.pushbuf; struct nv50_program *vp = nv50->vertprog; struct nv50_program *gp = nv50->gmtyprog; int m = 0; @@ -501,11 +512,11 @@ nv50_gp_linkage_validate(struct nv50_context *nv50) n = (m + 3) / 4; - BEGIN_RING(chan, RING_3D(VP_GP_BUILTIN_ATTR_EN), 1); - OUT_RING (chan, vp->vp.attrs[2] | gp->vp.attrs[2]); + BEGIN_NV04(push, NV50_3D(VP_GP_BUILTIN_ATTR_EN), 1); + PUSH_DATA (push, vp->vp.attrs[2] | gp->vp.attrs[2]); - BEGIN_RING(chan, RING_3D(VP_RESULT_MAP_SIZE), 1); - OUT_RING (chan, m); - BEGIN_RING(chan, RING_3D(VP_RESULT_MAP(0)), n); - OUT_RINGp (chan, map, n); + BEGIN_NV04(push, NV50_3D(VP_RESULT_MAP_SIZE), 1); + PUSH_DATA (push, m); + BEGIN_NV04(push, NV50_3D(VP_RESULT_MAP(0)), n); + PUSH_DATAp(push, map, n); } diff --git a/src/gallium/drivers/nv50/nv50_state.c b/src/gallium/drivers/nv50/nv50_state.c index 0f46afb882b..e1e1af039bb 100644 --- a/src/gallium/drivers/nv50/nv50_state.c +++ b/src/gallium/drivers/nv50/nv50_state.c @@ -121,7 +121,7 @@ nv50_blend_state_create(struct pipe_context *pipe, boolean emit_common_func = cso->rt[0].blend_enable; uint32_t ms; - if (nv50_context(pipe)->screen->tesla->grclass >= NVA3_3D) { + if (nv50_context(pipe)->screen->tesla->oclass >= NVA3_3D_CLASS) { SB_BEGIN_3D(so, BLEND_INDEPENDENT, 1); SB_DATA (so, cso->independent_blend_enable); } @@ -142,7 +142,7 @@ nv50_blend_state_create(struct pipe_context *pipe, emit_common_func = TRUE; } - if (nv50_context(pipe)->screen->tesla->grclass >= NVA3_3D) { + if (nv50_context(pipe)->screen->tesla->oclass >= NVA3_3D_CLASS) { emit_common_func = FALSE; for (i = 0; i < 8; ++i) { @@ -628,7 +628,7 @@ nv50_stage_set_sampler_views(struct nv50_context *nv50, int s, nv50->num_textures[s] = nr; - nv50_bufctx_reset(nv50, NV50_BUFCTX_TEXTURES); + nouveau_bufctx_reset(nv50->bufctx_3d, NV50_BIND_TEXTURES); nv50->dirty |= NV50_NEW_TEXTURES; } @@ -741,13 +741,15 @@ nv50_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index, { struct nv50_context *nv50 = nv50_context(pipe); - if (nv50->constbuf[shader][index]) - nv50_bufctx_del_resident(nv50, NV50_BUFCTX_CONSTANT, - nv04_resource(nv50->constbuf[shader][index])); - pipe_resource_reference(&nv50->constbuf[shader][index], res); nv50->constbuf_dirty[shader] |= 1 << index; + if (res) + nv50->constbuf_valid[shader] |= 1 << index; + else + nv50->constbuf_valid[shader] &= ~(1 << index); + + nouveau_bufctx_reset(nv50->bufctx_3d, NV50_BIND_CB(shader, index)); nv50->dirty |= NV50_NEW_CONSTBUF; } @@ -802,6 +804,8 @@ nv50_set_framebuffer_state(struct pipe_context *pipe, { struct nv50_context *nv50 = nv50_context(pipe); + nouveau_bufctx_reset(nv50->bufctx_3d, NV50_BIND_FB); + nv50->framebuffer = *fb; nv50->dirty |= NV50_NEW_FRAMEBUFFER; } @@ -852,7 +856,7 @@ nv50_set_vertex_buffers(struct pipe_context *pipe, memcpy(nv50->vtxbuf, vb, sizeof(*vb) * count); nv50->num_vtxbufs = count; - nv50_bufctx_reset(nv50, NV50_BUFCTX_VERTEX); + nouveau_bufctx_reset(nv50->bufctx_3d, NV50_BIND_VERTEX); nv50->dirty |= NV50_NEW_ARRAYS; } @@ -863,10 +867,15 @@ nv50_set_index_buffer(struct pipe_context *pipe, { struct nv50_context *nv50 = nv50_context(pipe); - if (ib) { - pipe_resource_reference(&nv50->idxbuf.buffer, ib->buffer); + if (nv50->idxbuf.buffer) + nouveau_bufctx_reset(nv50->bufctx_3d, NV50_BIND_INDEX); - memcpy(&nv50->idxbuf, ib, sizeof(nv50->idxbuf)); + if (ib && ib->buffer) { + pipe_resource_reference(&nv50->idxbuf.buffer, ib->buffer); + nv50->idxbuf.offset = ib->offset; + nv50->idxbuf.index_size = ib->index_size; + if (nouveau_resource_mapped_by_gpu(ib->buffer)) + BCTX_REFN(nv50->bufctx_3d, INDEX, nv04_resource(ib->buffer), RD); } else { pipe_resource_reference(&nv50->idxbuf.buffer, NULL); } diff --git a/src/gallium/drivers/nv50/nv50_state_validate.c b/src/gallium/drivers/nv50/nv50_state_validate.c index c8a1c4ecb37..7af992076c7 100644 --- a/src/gallium/drivers/nv50/nv50_state_validate.c +++ b/src/gallium/drivers/nv50/nv50_state_validate.c @@ -5,48 +5,44 @@ static void nv50_validate_fb(struct nv50_context *nv50) { - struct nouveau_channel *chan = nv50->screen->base.channel; + struct nouveau_pushbuf *push = nv50->base.pushbuf; struct pipe_framebuffer_state *fb = &nv50->framebuffer; unsigned i; unsigned ms_mode = NV50_3D_MULTISAMPLE_MODE_MS1; - boolean serialize = FALSE; - nv50_bufctx_reset(nv50, NV50_BUFCTX_FRAME); + nouveau_bufctx_reset(nv50->bufctx_3d, NV50_BIND_FB); - BEGIN_RING(chan, RING_3D(RT_CONTROL), 1); - OUT_RING (chan, (076543210 << 4) | fb->nr_cbufs); - BEGIN_RING(chan, RING_3D(SCREEN_SCISSOR_HORIZ), 2); - OUT_RING (chan, fb->width << 16); - OUT_RING (chan, fb->height << 16); - - MARK_RING(chan, 9 * fb->nr_cbufs, 2 * fb->nr_cbufs); + BEGIN_NV04(push, NV50_3D(RT_CONTROL), 1); + PUSH_DATA (push, (076543210 << 4) | fb->nr_cbufs); + BEGIN_NV04(push, NV50_3D(SCREEN_SCISSOR_HORIZ), 2); + PUSH_DATA (push, fb->width << 16); + PUSH_DATA (push, fb->height << 16); for (i = 0; i < fb->nr_cbufs; ++i) { struct nv50_miptree *mt = nv50_miptree(fb->cbufs[i]->texture); struct nv50_surface *sf = nv50_surface(fb->cbufs[i]); struct nouveau_bo *bo = mt->base.bo; - uint32_t offset = sf->offset; - - BEGIN_RING(chan, RING_3D(RT_ADDRESS_HIGH(i)), 5); - OUT_RELOCh(chan, bo, offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_RDWR); - OUT_RELOCl(chan, bo, offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_RDWR); - OUT_RING (chan, nv50_format_table[sf->base.format].rt); - if (likely(nouveau_bo_tile_layout(bo))) { - OUT_RING (chan, mt->level[sf->base.u.tex.level].tile_mode << 4); - OUT_RING (chan, mt->layer_stride >> 2); - BEGIN_RING(chan, RING_3D(RT_HORIZ(i)), 2); - OUT_RING (chan, sf->width); - OUT_RING (chan, sf->height); - BEGIN_RING(chan, RING_3D(RT_ARRAY_MODE), 1); - OUT_RING (chan, sf->depth); + + BEGIN_NV04(push, NV50_3D(RT_ADDRESS_HIGH(i)), 5); + PUSH_DATAh(push, bo->offset + sf->offset); + PUSH_DATA (push, bo->offset + sf->offset); + PUSH_DATA (push, nv50_format_table[sf->base.format].rt); + if (likely(nouveau_bo_memtype(bo))) { + PUSH_DATA (push, mt->level[sf->base.u.tex.level].tile_mode); + PUSH_DATA (push, mt->layer_stride >> 2); + BEGIN_NV04(push, NV50_3D(RT_HORIZ(i)), 2); + PUSH_DATA (push, sf->width); + PUSH_DATA (push, sf->height); + BEGIN_NV04(push, NV50_3D(RT_ARRAY_MODE), 1); + PUSH_DATA (push, sf->depth); } else { - OUT_RING (chan, 0); - OUT_RING (chan, 0); - BEGIN_RING(chan, RING_3D(RT_HORIZ(i)), 2); - OUT_RING (chan, NV50_3D_RT_HORIZ_LINEAR | mt->level[0].pitch); - OUT_RING (chan, sf->height); - BEGIN_RING(chan, RING_3D(RT_ARRAY_MODE), 1); - OUT_RING (chan, 0); + PUSH_DATA (push, 0); + PUSH_DATA (push, 0); + BEGIN_NV04(push, NV50_3D(RT_HORIZ(i)), 2); + PUSH_DATA (push, NV50_3D_RT_HORIZ_LINEAR | mt->level[0].pitch); + PUSH_DATA (push, sf->height); + BEGIN_NV04(push, NV50_3D(RT_ARRAY_MODE), 1); + PUSH_DATA (push, 0); assert(!fb->zsbuf); assert(!mt->ms_mode); @@ -55,13 +51,12 @@ nv50_validate_fb(struct nv50_context *nv50) ms_mode = mt->ms_mode; if (mt->base.status & NOUVEAU_BUFFER_STATUS_GPU_READING) - serialize = TRUE; + nv50->state.rt_serialize = TRUE; mt->base.status |= NOUVEAU_BUFFER_STATUS_GPU_WRITING; mt->base.status &= NOUVEAU_BUFFER_STATUS_GPU_READING; /* only register for writing, otherwise we'd always serialize here */ - nv50_bufctx_add_resident(nv50, NV50_BUFCTX_FRAME, &mt->base, - NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); + BCTX_REFN(nv50->bufctx_3d, FB, &mt->base, WR); } if (fb->zsbuf) { @@ -69,87 +64,79 @@ nv50_validate_fb(struct nv50_context *nv50) struct nv50_surface *sf = nv50_surface(fb->zsbuf); struct nouveau_bo *bo = mt->base.bo; int unk = mt->base.base.target == PIPE_TEXTURE_2D; - uint32_t offset = sf->offset; - - MARK_RING (chan, 12, 2); - BEGIN_RING(chan, RING_3D(ZETA_ADDRESS_HIGH), 5); - OUT_RELOCh(chan, bo, offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_RDWR); - OUT_RELOCl(chan, bo, offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_RDWR); - OUT_RING (chan, nv50_format_table[fb->zsbuf->format].rt); - OUT_RING (chan, mt->level[sf->base.u.tex.level].tile_mode << 4); - OUT_RING (chan, mt->layer_stride >> 2); - BEGIN_RING(chan, RING_3D(ZETA_ENABLE), 1); - OUT_RING (chan, 1); - BEGIN_RING(chan, RING_3D(ZETA_HORIZ), 3); - OUT_RING (chan, sf->width); - OUT_RING (chan, sf->height); - OUT_RING (chan, (unk << 16) | sf->depth); + + BEGIN_NV04(push, NV50_3D(ZETA_ADDRESS_HIGH), 5); + PUSH_DATAh(push, bo->offset + sf->offset); + PUSH_DATA (push, bo->offset + sf->offset); + PUSH_DATA (push, nv50_format_table[fb->zsbuf->format].rt); + PUSH_DATA (push, mt->level[sf->base.u.tex.level].tile_mode); + PUSH_DATA (push, mt->layer_stride >> 2); + BEGIN_NV04(push, NV50_3D(ZETA_ENABLE), 1); + PUSH_DATA (push, 1); + BEGIN_NV04(push, NV50_3D(ZETA_HORIZ), 3); + PUSH_DATA (push, sf->width); + PUSH_DATA (push, sf->height); + PUSH_DATA (push, (unk << 16) | sf->depth); ms_mode = mt->ms_mode; if (mt->base.status & NOUVEAU_BUFFER_STATUS_GPU_READING) - serialize = TRUE; + nv50->state.rt_serialize = TRUE; mt->base.status |= NOUVEAU_BUFFER_STATUS_GPU_WRITING; mt->base.status &= NOUVEAU_BUFFER_STATUS_GPU_READING; - nv50_bufctx_add_resident(nv50, NV50_BUFCTX_FRAME, &mt->base, - NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); + BCTX_REFN(nv50->bufctx_3d, FB, &mt->base, WR); } else { - BEGIN_RING(chan, RING_3D(ZETA_ENABLE), 1); - OUT_RING (chan, 0); + BEGIN_NV04(push, NV50_3D(ZETA_ENABLE), 1); + PUSH_DATA (push, 0); } - BEGIN_RING(chan, RING_3D(MULTISAMPLE_MODE), 1); - OUT_RING (chan, ms_mode); + BEGIN_NV04(push, NV50_3D(MULTISAMPLE_MODE), 1); + PUSH_DATA (push, ms_mode); - BEGIN_RING(chan, RING_3D(VIEWPORT_HORIZ(0)), 2); - OUT_RING (chan, fb->width << 16); - OUT_RING (chan, fb->height << 16); - - if (serialize) { - BEGIN_RING(chan, RING_3D(SERIALIZE), 1); - OUT_RING (chan, 0); - } + BEGIN_NV04(push, NV50_3D(VIEWPORT_HORIZ(0)), 2); + PUSH_DATA (push, fb->width << 16); + PUSH_DATA (push, fb->height << 16); } static void nv50_validate_blend_colour(struct nv50_context *nv50) { - struct nouveau_channel *chan = nv50->screen->base.channel; + struct nouveau_pushbuf *push = nv50->base.pushbuf; - BEGIN_RING(chan, RING_3D(BLEND_COLOR(0)), 4); - OUT_RINGf (chan, nv50->blend_colour.color[0]); - OUT_RINGf (chan, nv50->blend_colour.color[1]); - OUT_RINGf (chan, nv50->blend_colour.color[2]); - OUT_RINGf (chan, nv50->blend_colour.color[3]); + BEGIN_NV04(push, NV50_3D(BLEND_COLOR(0)), 4); + PUSH_DATAf(push, nv50->blend_colour.color[0]); + PUSH_DATAf(push, nv50->blend_colour.color[1]); + PUSH_DATAf(push, nv50->blend_colour.color[2]); + PUSH_DATAf(push, nv50->blend_colour.color[3]); } static void nv50_validate_stencil_ref(struct nv50_context *nv50) { - struct nouveau_channel *chan = nv50->screen->base.channel; + struct nouveau_pushbuf *push = nv50->base.pushbuf; - BEGIN_RING(chan, RING_3D(STENCIL_FRONT_FUNC_REF), 1); - OUT_RING (chan, nv50->stencil_ref.ref_value[0]); - BEGIN_RING(chan, RING_3D(STENCIL_BACK_FUNC_REF), 1); - OUT_RING (chan, nv50->stencil_ref.ref_value[1]); + BEGIN_NV04(push, NV50_3D(STENCIL_FRONT_FUNC_REF), 1); + PUSH_DATA (push, nv50->stencil_ref.ref_value[0]); + BEGIN_NV04(push, NV50_3D(STENCIL_BACK_FUNC_REF), 1); + PUSH_DATA (push, nv50->stencil_ref.ref_value[1]); } static void nv50_validate_stipple(struct nv50_context *nv50) { - struct nouveau_channel *chan = nv50->screen->base.channel; + struct nouveau_pushbuf *push = nv50->base.pushbuf; unsigned i; - BEGIN_RING(chan, RING_3D(POLYGON_STIPPLE_PATTERN(0)), 32); + BEGIN_NV04(push, NV50_3D(POLYGON_STIPPLE_PATTERN(0)), 32); for (i = 0; i < 32; ++i) - OUT_RING(chan, util_bswap32(nv50->stipple.stipple[i])); + PUSH_DATA(push, util_bswap32(nv50->stipple.stipple[i])); } static void nv50_validate_scissor(struct nv50_context *nv50) { - struct nouveau_channel *chan = nv50->screen->base.channel; + struct nouveau_pushbuf *push = nv50->base.pushbuf; struct pipe_scissor_state *s = &nv50->scissor; #ifdef NV50_SCISSORS_CLIPPING struct pipe_viewport_state *vp = &nv50->viewport; @@ -178,38 +165,38 @@ nv50_validate_scissor(struct nv50_context *nv50) miny = MAX2(miny, (int)(vp->translate[1] - fabsf(vp->scale[1]))); maxy = MIN2(maxy, (int)(vp->translate[1] + fabsf(vp->scale[1]))); - BEGIN_RING(chan, RING_3D(SCISSOR_HORIZ(0)), 2); - OUT_RING (chan, (maxx << 16) | minx); - OUT_RING (chan, (maxy << 16) | miny); + BEGIN_NV04(push, NV50_3D(SCISSOR_HORIZ(0)), 2); + PUSH_DATA (push, (maxx << 16) | minx); + PUSH_DATA (push, (maxy << 16) | miny); #else - BEGIN_RING(chan, RING_3D(SCISSOR_HORIZ(0)), 2); - OUT_RING (chan, (s->maxx << 16) | s->minx); - OUT_RING (chan, (s->maxy << 16) | s->miny); + BEGIN_NV04(push, NV50_3D(SCISSOR_HORIZ(0)), 2); + PUSH_DATA (push, (s->maxx << 16) | s->minx); + PUSH_DATA (push, (s->maxy << 16) | s->miny); #endif } static void nv50_validate_viewport(struct nv50_context *nv50) { - struct nouveau_channel *chan = nv50->screen->base.channel; + struct nouveau_pushbuf *push = nv50->base.pushbuf; float zmin, zmax; - BEGIN_RING(chan, RING_3D(VIEWPORT_TRANSLATE_X(0)), 3); - OUT_RINGf (chan, nv50->viewport.translate[0]); - OUT_RINGf (chan, nv50->viewport.translate[1]); - OUT_RINGf (chan, nv50->viewport.translate[2]); - BEGIN_RING(chan, RING_3D(VIEWPORT_SCALE_X(0)), 3); - OUT_RINGf (chan, nv50->viewport.scale[0]); - OUT_RINGf (chan, nv50->viewport.scale[1]); - OUT_RINGf (chan, nv50->viewport.scale[2]); + BEGIN_NV04(push, NV50_3D(VIEWPORT_TRANSLATE_X(0)), 3); + PUSH_DATAf(push, nv50->viewport.translate[0]); + PUSH_DATAf(push, nv50->viewport.translate[1]); + PUSH_DATAf(push, nv50->viewport.translate[2]); + BEGIN_NV04(push, NV50_3D(VIEWPORT_SCALE_X(0)), 3); + PUSH_DATAf(push, nv50->viewport.scale[0]); + PUSH_DATAf(push, nv50->viewport.scale[1]); + PUSH_DATAf(push, nv50->viewport.scale[2]); zmin = nv50->viewport.translate[2] - fabsf(nv50->viewport.scale[2]); zmax = nv50->viewport.translate[2] + fabsf(nv50->viewport.scale[2]); #ifdef NV50_SCISSORS_CLIPPING - BEGIN_RING(chan, RING_3D(DEPTH_RANGE_NEAR(0)), 2); - OUT_RINGf (chan, zmin); - OUT_RINGf (chan, zmax); + BEGIN_NV04(push, NV50_3D(DEPTH_RANGE_NEAR(0)), 2); + PUSH_DATAf(push, zmin); + PUSH_DATAf(push, zmax); #endif } @@ -234,15 +221,15 @@ nv50_check_program_ucps(struct nv50_context *nv50, static void nv50_validate_clip(struct nv50_context *nv50) { - struct nouveau_channel *chan = nv50->screen->base.channel; + struct nouveau_pushbuf *push = nv50->base.pushbuf; struct nv50_program *vp; uint8_t clip_enable; if (nv50->dirty & NV50_NEW_CLIP) { - BEGIN_RING(chan, RING_3D(CB_ADDR), 1); - OUT_RING (chan, (0 << 8) | NV50_CB_AUX); - BEGIN_RING_NI(chan, RING_3D(CB_DATA(0)), PIPE_MAX_CLIP_PLANES * 4); - OUT_RINGp (chan, &nv50->clip.ucp[0][0], PIPE_MAX_CLIP_PLANES * 4); + BEGIN_NV04(push, NV50_3D(CB_ADDR), 1); + PUSH_DATA (push, (0 << 8) | NV50_CB_AUX); + BEGIN_NI04(push, NV50_3D(CB_DATA(0)), PIPE_MAX_CLIP_PLANES * 4); + PUSH_DATAp(push, &nv50->clip.ucp[0][0], PIPE_MAX_CLIP_PLANES * 4); } vp = nv50->gmtyprog; @@ -251,8 +238,8 @@ nv50_validate_clip(struct nv50_context *nv50) clip_enable = nv50->rast->pipe.clip_plane_enable; - BEGIN_RING(chan, RING_3D(VP_CLIP_DISTANCE_ENABLE), 1); - OUT_RING (chan, clip_enable); + BEGIN_NV04(push, NV50_3D(CLIP_DISTANCE_ENABLE), 1); + PUSH_DATA (push, clip_enable); if (clip_enable) nv50_check_program_ucps(nv50, vp, clip_enable); @@ -261,34 +248,34 @@ nv50_validate_clip(struct nv50_context *nv50) static void nv50_validate_blend(struct nv50_context *nv50) { - struct nouveau_channel *chan = nv50->screen->base.channel; + struct nouveau_pushbuf *push = nv50->base.pushbuf; - WAIT_RING(chan, nv50->blend->size); - OUT_RINGp(chan, nv50->blend->state, nv50->blend->size); + PUSH_SPACE(push, nv50->blend->size); + PUSH_DATAp(push, nv50->blend->state, nv50->blend->size); } static void nv50_validate_zsa(struct nv50_context *nv50) { - struct nouveau_channel *chan = nv50->screen->base.channel; + struct nouveau_pushbuf *push = nv50->base.pushbuf; - WAIT_RING(chan, nv50->zsa->size); - OUT_RINGp(chan, nv50->zsa->state, nv50->zsa->size); + PUSH_SPACE(push, nv50->zsa->size); + PUSH_DATAp(push, nv50->zsa->state, nv50->zsa->size); } static void nv50_validate_rasterizer(struct nv50_context *nv50) { - struct nouveau_channel *chan = nv50->screen->base.channel; + struct nouveau_pushbuf *push = nv50->base.pushbuf; - WAIT_RING(chan, nv50->rast->size); - OUT_RINGp(chan, nv50->rast->state, nv50->rast->size); + PUSH_SPACE(push, nv50->rast->size); + PUSH_DATAp(push, nv50->rast->state, nv50->rast->size); } static void nv50_validate_sample_mask(struct nv50_context *nv50) { - struct nouveau_channel *chan = nv50->screen->base.channel; + struct nouveau_pushbuf *push = nv50->base.pushbuf; unsigned mask[4] = { @@ -298,11 +285,11 @@ nv50_validate_sample_mask(struct nv50_context *nv50) nv50->sample_mask & 0xffff }; - BEGIN_RING(chan, RING_3D(MSAA_MASK(0)), 4); - OUT_RING (chan, mask[0]); - OUT_RING (chan, mask[1]); - OUT_RING (chan, mask[2]); - OUT_RING (chan, mask[3]); + BEGIN_NV04(push, NV50_3D(MSAA_MASK(0)), 4); + PUSH_DATA (push, mask[0]); + PUSH_DATA (push, mask[1]); + PUSH_DATA (push, mask[2]); + PUSH_DATA (push, mask[3]); } static void @@ -378,6 +365,7 @@ boolean nv50_state_validate(struct nv50_context *nv50, uint32_t mask, unsigned words) { uint32_t state_mask; + int ret; unsigned i; if (nv50->screen->cur_ctx != nv50) @@ -393,11 +381,21 @@ nv50_state_validate(struct nv50_context *nv50, uint32_t mask, unsigned words) validate->func(nv50); } nv50->dirty &= ~state_mask; - } - MARK_RING(nv50->screen->base.channel, words, 0); + if (nv50->state.rt_serialize) { + nv50->state.rt_serialize = FALSE; + BEGIN_NV04(nv50->base.pushbuf, SUBC_3D(NV50_GRAPH_SERIALIZE), 1); + PUSH_DATA (nv50->base.pushbuf, 0); + } - nv50_bufctx_emit_relocs(nv50); + nv50_bufctx_fence(nv50->bufctx_3d, FALSE); + } + nouveau_pushbuf_bufctx(nv50->base.pushbuf, nv50->bufctx_3d); + ret = nouveau_pushbuf_validate(nv50->base.pushbuf); - return TRUE; + if (unlikely(nv50->state.flushed)) { + nv50->state.flushed = FALSE; + nv50_bufctx_fence(nv50->bufctx_3d, TRUE); + } + return !ret; } diff --git a/src/gallium/drivers/nv50/nv50_stateobj.h b/src/gallium/drivers/nv50/nv50_stateobj.h index c443dfbfec6..b056e19ea50 100644 --- a/src/gallium/drivers/nv50/nv50_stateobj.h +++ b/src/gallium/drivers/nv50/nv50_stateobj.h @@ -6,13 +6,11 @@ #define NV50_SCISSORS_CLIPPING -#define SB_BEGIN_3D(so, m, s) \ - (so)->state[(so)->size++] = \ - ((s) << 18) | (NV50_SUBCH_3D << 13) | NV50_3D_##m +#define SB_BEGIN_3D(so, m, s) \ + (so)->state[(so)->size++] = NV50_FIFO_PKHDR(NV50_3D(m), s) -#define SB_BEGIN_3D_(so, m, s) \ - (so)->state[(so)->size++] = \ - ((s) << 18) | (NV50_SUBCH_3D << 13) | m +#define SB_BEGIN_3D_(so, m, s) \ + (so)->state[(so)->size++] = NV50_FIFO_PKHDR(SUBC_3D(m), s) #define SB_DATA(so, u) (so)->state[(so)->size++] = (u) diff --git a/src/gallium/drivers/nv50/nv50_surface.c b/src/gallium/drivers/nv50/nv50_surface.c index 28e34bd9ef6..0d29dcc8a2a 100644 --- a/src/gallium/drivers/nv50/nv50_surface.c +++ b/src/gallium/drivers/nv50/nv50_surface.c @@ -70,14 +70,13 @@ nv50_2d_format(enum pipe_format format) } static int -nv50_2d_texture_set(struct nouveau_channel *chan, int dst, +nv50_2d_texture_set(struct nouveau_pushbuf *push, int dst, struct nv50_miptree *mt, unsigned level, unsigned layer) { struct nouveau_bo *bo = mt->base.bo; uint32_t width, height, depth; uint32_t format; uint32_t mthd = dst ? NV50_2D_DST_FORMAT : NV50_2D_SRC_FORMAT; - uint32_t flags = mt->base.domain | (dst ? NOUVEAU_BO_WR : NOUVEAU_BO_RD); uint32_t offset = mt->level[level].offset; format = nv50_2d_format(mt->base.base.format); @@ -99,44 +98,44 @@ nv50_2d_texture_set(struct nouveau_channel *chan, int dst, depth = u_minify(mt->base.base.depth0, level); } - if (!(bo->tile_flags & NOUVEAU_BO_TILE_LAYOUT_MASK)) { - BEGIN_RING(chan, RING_2D_(mthd), 2); - OUT_RING (chan, format); - OUT_RING (chan, 1); - BEGIN_RING(chan, RING_2D_(mthd + 0x14), 5); - OUT_RING (chan, mt->level[level].pitch); - OUT_RING (chan, width); - OUT_RING (chan, height); - OUT_RELOCh(chan, bo, offset, flags); - OUT_RELOCl(chan, bo, offset, flags); + if (!nouveau_bo_memtype(bo)) { + BEGIN_NV04(push, SUBC_2D(mthd), 2); + PUSH_DATA (push, format); + PUSH_DATA (push, 1); + BEGIN_NV04(push, SUBC_2D(mthd + 0x14), 5); + PUSH_DATA (push, mt->level[level].pitch); + PUSH_DATA (push, width); + PUSH_DATA (push, height); + PUSH_DATAh(push, bo->offset + offset); + PUSH_DATA (push, bo->offset + offset); } else { - BEGIN_RING(chan, RING_2D_(mthd), 5); - OUT_RING (chan, format); - OUT_RING (chan, 0); - OUT_RING (chan, mt->level[level].tile_mode << 4); - OUT_RING (chan, depth); - OUT_RING (chan, layer); - BEGIN_RING(chan, RING_2D_(mthd + 0x18), 4); - OUT_RING (chan, width); - OUT_RING (chan, height); - OUT_RELOCh(chan, bo, offset, flags); - OUT_RELOCl(chan, bo, offset, flags); + BEGIN_NV04(push, SUBC_2D(mthd), 5); + PUSH_DATA (push, format); + PUSH_DATA (push, 0); + PUSH_DATA (push, mt->level[level].tile_mode); + PUSH_DATA (push, depth); + PUSH_DATA (push, layer); + BEGIN_NV04(push, SUBC_2D(mthd + 0x18), 4); + PUSH_DATA (push, width); + PUSH_DATA (push, height); + PUSH_DATAh(push, bo->offset + offset); + PUSH_DATA (push, bo->offset + offset); } #if 0 if (dst) { - BEGIN_RING(chan, RING_2D_(NV50_2D_CLIP_X), 4); - OUT_RING (chan, 0); - OUT_RING (chan, 0); - OUT_RING (chan, width); - OUT_RING (chan, height); + BEGIN_NV04(push, SUBC_2D(NV50_2D_CLIP_X), 4); + PUSH_DATA (push, 0); + PUSH_DATA (push, 0); + PUSH_DATA (push, width); + PUSH_DATA (push, height); } #endif return 0; } static int -nv50_2d_texture_do_copy(struct nouveau_channel *chan, +nv50_2d_texture_do_copy(struct nouveau_pushbuf *push, struct nv50_miptree *dst, unsigned dst_level, unsigned dx, unsigned dy, unsigned dz, struct nv50_miptree *src, unsigned src_level, @@ -150,16 +149,16 @@ nv50_2d_texture_do_copy(struct nouveau_channel *chan, int ret; uint32_t ctrl; - +#if 0 ret = MARK_RING(chan, 2 * 16 + 32, 4); if (ret) return ret; - - ret = nv50_2d_texture_set(chan, 1, dst, dst_level, dz); +#endif + ret = nv50_2d_texture_set(push, 1, dst, dst_level, dz); if (ret) return ret; - ret = nv50_2d_texture_set(chan, 0, src, src_level, sz); + ret = nv50_2d_texture_set(push, 0, src, src_level, sz); if (ret) return ret; @@ -168,23 +167,23 @@ nv50_2d_texture_do_copy(struct nouveau_channel *chan, ctrl = 0x11; /* 0/1 = CENTER/CORNER, 00/10 = POINT/BILINEAR */ - BEGIN_RING(chan, RING_2D(BLIT_CONTROL), 1); - OUT_RING (chan, ctrl); - BEGIN_RING(chan, RING_2D(BLIT_DST_X), 4); - OUT_RING (chan, dx << dst->ms_x); - OUT_RING (chan, dy << dst->ms_y); - OUT_RING (chan, w << dst->ms_x); - OUT_RING (chan, h << dst->ms_y); - BEGIN_RING(chan, RING_2D(BLIT_DU_DX_FRACT), 4); - OUT_RING (chan, duvdxy[2 + ((int)src->ms_x - (int)dst->ms_x)] & 0xf0000000); - OUT_RING (chan, duvdxy[2 + ((int)src->ms_x - (int)dst->ms_x)] & 0x0000000f); - OUT_RING (chan, duvdxy[2 + ((int)src->ms_y - (int)dst->ms_y)] & 0xf0000000); - OUT_RING (chan, duvdxy[2 + ((int)src->ms_y - (int)dst->ms_y)] & 0x0000000f); - BEGIN_RING(chan, RING_2D(BLIT_SRC_X_FRACT), 4); - OUT_RING (chan, 0); - OUT_RING (chan, sx << src->ms_x); - OUT_RING (chan, 0); - OUT_RING (chan, sy << src->ms_y); + BEGIN_NV04(push, NV50_2D(BLIT_CONTROL), 1); + PUSH_DATA (push, ctrl); + BEGIN_NV04(push, NV50_2D(BLIT_DST_X), 4); + PUSH_DATA (push, dx << dst->ms_x); + PUSH_DATA (push, dy << dst->ms_y); + PUSH_DATA (push, w << dst->ms_x); + PUSH_DATA (push, h << dst->ms_y); + BEGIN_NV04(push, NV50_2D(BLIT_DU_DX_FRACT), 4); + PUSH_DATA (push, duvdxy[2 + ((int)src->ms_x - (int)dst->ms_x)] & 0xf0000000); + PUSH_DATA (push, duvdxy[2 + ((int)src->ms_x - (int)dst->ms_x)] & 0x0000000f); + PUSH_DATA (push, duvdxy[2 + ((int)src->ms_y - (int)dst->ms_y)] & 0xf0000000); + PUSH_DATA (push, duvdxy[2 + ((int)src->ms_y - (int)dst->ms_y)] & 0x0000000f); + BEGIN_NV04(push, NV50_2D(BLIT_SRC_X_FRACT), 4); + PUSH_DATA (push, 0); + PUSH_DATA (push, sx << src->ms_x); + PUSH_DATA (push, 0); + PUSH_DATA (push, sy << src->ms_y); return 0; } @@ -196,7 +195,7 @@ nv50_resource_copy_region(struct pipe_context *pipe, struct pipe_resource *src, unsigned src_level, const struct pipe_box *src_box) { - struct nv50_screen *screen = nv50_context(pipe)->screen; + struct nv50_context *nv50 = nv50_context(pipe); int ret; boolean m2mf; unsigned dst_layer = dstz, src_layer = src_box->z; @@ -227,7 +226,7 @@ nv50_resource_copy_region(struct pipe_context *pipe, src_box->x, src_box->y, src_box->z); for (i = 0; i < src_box->depth; ++i) { - nv50_m2mf_transfer_rect(&screen->base.base, &drect, &srect, nx, ny); + nv50_m2mf_transfer_rect(nv50, &drect, &srect, nx, ny); if (nv50_miptree(dst)->layout_3d) drect.z++; @@ -246,16 +245,22 @@ nv50_resource_copy_region(struct pipe_context *pipe, (nv50_2d_format_faithful(src->format) && nv50_2d_format_faithful(dst->format))); + BCTX_REFN(nv50->bufctx, 2D, nv04_resource(src), RD); + BCTX_REFN(nv50->bufctx, 2D, nv04_resource(dst), WR); + nouveau_pushbuf_bufctx(nv50->base.pushbuf, nv50->bufctx); + nouveau_pushbuf_validate(nv50->base.pushbuf); + for (; dst_layer < dstz + src_box->depth; ++dst_layer, ++src_layer) { - ret = nv50_2d_texture_do_copy(screen->base.channel, + ret = nv50_2d_texture_do_copy(nv50->base.pushbuf, nv50_miptree(dst), dst_level, dstx, dsty, dst_layer, nv50_miptree(src), src_level, src_box->x, src_box->y, src_layer, src_box->width, src_box->height); if (ret) - return; + break; } + nouveau_bufctx_reset(nv50->bufctx, NV50_BIND_2D); } static void @@ -266,51 +271,50 @@ nv50_clear_render_target(struct pipe_context *pipe, unsigned width, unsigned height) { struct nv50_context *nv50 = nv50_context(pipe); - struct nv50_screen *screen = nv50->screen; - struct nouveau_channel *chan = screen->base.channel; + struct nouveau_pushbuf *push = nv50->base.pushbuf; struct nv50_miptree *mt = nv50_miptree(dst->texture); struct nv50_surface *sf = nv50_surface(dst); struct nouveau_bo *bo = mt->base.bo; - BEGIN_RING(chan, RING_3D(CLEAR_COLOR(0)), 4); - OUT_RINGf (chan, color->f[0]); - OUT_RINGf (chan, color->f[1]); - OUT_RINGf (chan, color->f[2]); - OUT_RINGf (chan, color->f[3]); - + BEGIN_NV04(push, NV50_3D(CLEAR_COLOR(0)), 4); + PUSH_DATAf(push, color->f[0]); + PUSH_DATAf(push, color->f[1]); + PUSH_DATAf(push, color->f[2]); + PUSH_DATAf(push, color->f[3]); +#if 0 if (MARK_RING(chan, 18, 2)) return; - - BEGIN_RING(chan, RING_3D(RT_CONTROL), 1); - OUT_RING (chan, 1); - BEGIN_RING(chan, RING_3D(RT_ADDRESS_HIGH(0)), 5); - OUT_RELOCh(chan, bo, sf->offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); - OUT_RELOCl(chan, bo, sf->offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); - OUT_RING (chan, nv50_format_table[dst->format].rt); - OUT_RING (chan, mt->level[sf->base.u.tex.level].tile_mode << 4); - OUT_RING (chan, 0); - BEGIN_RING(chan, RING_3D(RT_HORIZ(0)), 2); - if (nouveau_bo_tile_layout(bo)) - OUT_RING(chan, sf->width); +#endif + BEGIN_NV04(push, NV50_3D(RT_CONTROL), 1); + PUSH_DATA (push, 1); + BEGIN_NV04(push, NV50_3D(RT_ADDRESS_HIGH(0)), 5); + PUSH_DATAh(push, bo->offset + sf->offset); + PUSH_DATA (push, bo->offset + sf->offset); + PUSH_DATA (push, nv50_format_table[dst->format].rt); + PUSH_DATA (push, mt->level[sf->base.u.tex.level].tile_mode); + PUSH_DATA (push, 0); + BEGIN_NV04(push, NV50_3D(RT_HORIZ(0)), 2); + if (nouveau_bo_memtype(bo)) + PUSH_DATA(push, sf->width); else - OUT_RING(chan, NV50_3D_RT_HORIZ_LINEAR | mt->level[0].pitch); - OUT_RING (chan, sf->height); - BEGIN_RING(chan, RING_3D(RT_ARRAY_MODE), 1); - OUT_RING (chan, 1); - - if (!nouveau_bo_tile_layout(bo)) { - BEGIN_RING(chan, RING_3D(ZETA_ENABLE), 1); - OUT_RING (chan, 0); + PUSH_DATA(push, NV50_3D_RT_HORIZ_LINEAR | mt->level[0].pitch); + PUSH_DATA (push, sf->height); + BEGIN_NV04(push, NV50_3D(RT_ARRAY_MODE), 1); + PUSH_DATA (push, 1); + + if (!nouveau_bo_memtype(bo)) { + BEGIN_NV04(push, NV50_3D(ZETA_ENABLE), 1); + PUSH_DATA (push, 0); } /* NOTE: only works with D3D clear flag (5097/0x143c bit 4) */ - BEGIN_RING(chan, RING_3D(VIEWPORT_HORIZ(0)), 2); - OUT_RING (chan, (width << 16) | dstx); - OUT_RING (chan, (height << 16) | dsty); + BEGIN_NV04(push, NV50_3D(VIEWPORT_HORIZ(0)), 2); + PUSH_DATA (push, (width << 16) | dstx); + PUSH_DATA (push, (height << 16) | dsty); - BEGIN_RING(chan, RING_3D(CLEAR_BUFFERS), 1); - OUT_RING (chan, 0x3c); + BEGIN_NV04(push, NV50_3D(CLEAR_BUFFERS), 1); + PUSH_DATA (push, 0x3c); nv50->dirty |= NV50_NEW_FRAMEBUFFER; } @@ -325,49 +329,48 @@ nv50_clear_depth_stencil(struct pipe_context *pipe, unsigned width, unsigned height) { struct nv50_context *nv50 = nv50_context(pipe); - struct nv50_screen *screen = nv50->screen; - struct nouveau_channel *chan = screen->base.channel; + struct nouveau_pushbuf *push = nv50->base.pushbuf; struct nv50_miptree *mt = nv50_miptree(dst->texture); struct nv50_surface *sf = nv50_surface(dst); struct nouveau_bo *bo = mt->base.bo; uint32_t mode = 0; - assert(nouveau_bo_tile_layout(bo)); /* ZETA cannot be linear */ + assert(nouveau_bo_memtype(bo)); /* ZETA cannot be linear */ if (clear_flags & PIPE_CLEAR_DEPTH) { - BEGIN_RING(chan, RING_3D(CLEAR_DEPTH), 1); - OUT_RINGf (chan, depth); + BEGIN_NV04(push, NV50_3D(CLEAR_DEPTH), 1); + PUSH_DATAf(push, depth); mode |= NV50_3D_CLEAR_BUFFERS_Z; } if (clear_flags & PIPE_CLEAR_STENCIL) { - BEGIN_RING(chan, RING_3D(CLEAR_STENCIL), 1); - OUT_RING (chan, stencil & 0xff); + BEGIN_NV04(push, NV50_3D(CLEAR_STENCIL), 1); + PUSH_DATA (push, stencil & 0xff); mode |= NV50_3D_CLEAR_BUFFERS_S; } - +#if 0 if (MARK_RING(chan, 17, 2)) return; - - BEGIN_RING(chan, RING_3D(ZETA_ADDRESS_HIGH), 5); - OUT_RELOCh(chan, bo, sf->offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); - OUT_RELOCl(chan, bo, sf->offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); - OUT_RING (chan, nv50_format_table[dst->format].rt); - OUT_RING (chan, mt->level[sf->base.u.tex.level].tile_mode << 4); - OUT_RING (chan, 0); - BEGIN_RING(chan, RING_3D(ZETA_ENABLE), 1); - OUT_RING (chan, 1); - BEGIN_RING(chan, RING_3D(ZETA_HORIZ), 3); - OUT_RING (chan, sf->width); - OUT_RING (chan, sf->height); - OUT_RING (chan, (1 << 16) | 1); - - BEGIN_RING(chan, RING_3D(VIEWPORT_HORIZ(0)), 2); - OUT_RING (chan, (width << 16) | dstx); - OUT_RING (chan, (height << 16) | dsty); - - BEGIN_RING(chan, RING_3D(CLEAR_BUFFERS), 1); - OUT_RING (chan, mode); +#endif + BEGIN_NV04(push, NV50_3D(ZETA_ADDRESS_HIGH), 5); + PUSH_DATAh(push, bo->offset + sf->offset); + PUSH_DATA (push, bo->offset + sf->offset); + PUSH_DATA (push, nv50_format_table[dst->format].rt); + PUSH_DATA (push, mt->level[sf->base.u.tex.level].tile_mode); + PUSH_DATA (push, 0); + BEGIN_NV04(push, NV50_3D(ZETA_ENABLE), 1); + PUSH_DATA (push, 1); + BEGIN_NV04(push, NV50_3D(ZETA_HORIZ), 3); + PUSH_DATA (push, sf->width); + PUSH_DATA (push, sf->height); + PUSH_DATA (push, (1 << 16) | 1); + + BEGIN_NV04(push, NV50_3D(VIEWPORT_HORIZ(0)), 2); + PUSH_DATA (push, (width << 16) | dstx); + PUSH_DATA (push, (height << 16) | dsty); + + BEGIN_NV04(push, NV50_3D(CLEAR_BUFFERS), 1); + PUSH_DATA (push, mode); nv50->dirty |= NV50_NEW_FRAMEBUFFER; } @@ -378,7 +381,7 @@ nv50_clear(struct pipe_context *pipe, unsigned buffers, double depth, unsigned stencil) { struct nv50_context *nv50 = nv50_context(pipe); - struct nouveau_channel *chan = nv50->screen->base.channel; + struct nouveau_pushbuf *push = nv50->base.pushbuf; struct pipe_framebuffer_state *fb = &nv50->framebuffer; unsigned i; uint32_t mode = 0; @@ -388,34 +391,34 @@ nv50_clear(struct pipe_context *pipe, unsigned buffers, return; if (buffers & PIPE_CLEAR_COLOR && fb->nr_cbufs) { - BEGIN_RING(chan, RING_3D(CLEAR_COLOR(0)), 4); - OUT_RINGf (chan, color->f[0]); - OUT_RINGf (chan, color->f[1]); - OUT_RINGf (chan, color->f[2]); - OUT_RINGf (chan, color->f[3]); + BEGIN_NV04(push, NV50_3D(CLEAR_COLOR(0)), 4); + PUSH_DATAf(push, color->f[0]); + PUSH_DATAf(push, color->f[1]); + PUSH_DATAf(push, color->f[2]); + PUSH_DATAf(push, color->f[3]); mode = NV50_3D_CLEAR_BUFFERS_R | NV50_3D_CLEAR_BUFFERS_G | NV50_3D_CLEAR_BUFFERS_B | NV50_3D_CLEAR_BUFFERS_A; } if (buffers & PIPE_CLEAR_DEPTH) { - BEGIN_RING(chan, RING_3D(CLEAR_DEPTH), 1); - OUT_RING (chan, fui(depth)); + BEGIN_NV04(push, NV50_3D(CLEAR_DEPTH), 1); + PUSH_DATA (push, fui(depth)); mode |= NV50_3D_CLEAR_BUFFERS_Z; } if (buffers & PIPE_CLEAR_STENCIL) { - BEGIN_RING(chan, RING_3D(CLEAR_STENCIL), 1); - OUT_RING (chan, stencil & 0xff); + BEGIN_NV04(push, NV50_3D(CLEAR_STENCIL), 1); + PUSH_DATA (push, stencil & 0xff); mode |= NV50_3D_CLEAR_BUFFERS_S; } - BEGIN_RING(chan, RING_3D(CLEAR_BUFFERS), 1); - OUT_RING (chan, mode); + BEGIN_NV04(push, NV50_3D(CLEAR_BUFFERS), 1); + PUSH_DATA (push, mode); for (i = 1; i < fb->nr_cbufs; i++) { - BEGIN_RING(chan, RING_3D(CLEAR_BUFFERS), 1); - OUT_RING (chan, (i << 6) | 0x3c); + BEGIN_NV04(push, NV50_3D(CLEAR_BUFFERS), 1); + PUSH_DATA (push, (i << 6) | 0x3c); } } @@ -709,50 +712,50 @@ nv50_blit_set_src(struct nv50_context *nv50, static void nv50_blitctx_prepare_state(struct nv50_blitctx *blit) { - struct nouveau_channel *chan = blit->screen->base.channel; + struct nouveau_pushbuf *push = blit->screen->base.pushbuf; /* blend state */ - BEGIN_RING(chan, RING_3D(COLOR_MASK(0)), 1); - OUT_RING (chan, blit->color_mask); - BEGIN_RING(chan, RING_3D(BLEND_ENABLE(0)), 1); - OUT_RING (chan, 0); - BEGIN_RING(chan, RING_3D(LOGIC_OP_ENABLE), 1); - OUT_RING (chan, 0); + BEGIN_NV04(push, NV50_3D(COLOR_MASK(0)), 1); + PUSH_DATA (push, blit->color_mask); + BEGIN_NV04(push, NV50_3D(BLEND_ENABLE(0)), 1); + PUSH_DATA (push, 0); + BEGIN_NV04(push, NV50_3D(LOGIC_OP_ENABLE), 1); + PUSH_DATA (push, 0); /* rasterizer state */ #ifndef NV50_SCISSORS_CLIPPING - BEGIN_RING(chan, RING_3D(SCISSOR_ENABLE(0)), 1); - OUT_RING (chan, 1); + BEGIN_NV04(push, NV50_3D(SCISSOR_ENABLE(0)), 1); + PUSH_DATA (push, 1); #endif - BEGIN_RING(chan, RING_3D(VERTEX_TWO_SIDE_ENABLE), 1); - OUT_RING (chan, 0); - BEGIN_RING(chan, RING_3D(FRAG_COLOR_CLAMP_EN), 1); - OUT_RING (chan, 0); - BEGIN_RING(chan, RING_3D(MULTISAMPLE_ENABLE), 1); - OUT_RING (chan, 0); - BEGIN_RING(chan, RING_3D(MSAA_MASK(0)), 4); - OUT_RING (chan, 0xffff); - OUT_RING (chan, 0xffff); - OUT_RING (chan, 0xffff); - OUT_RING (chan, 0xffff); - BEGIN_RING(chan, RING_3D(POLYGON_MODE_FRONT), 3); - OUT_RING (chan, NV50_3D_POLYGON_MODE_FRONT_FILL); - OUT_RING (chan, NV50_3D_POLYGON_MODE_BACK_FILL); - OUT_RING (chan, 0); - BEGIN_RING(chan, RING_3D(CULL_FACE_ENABLE), 1); - OUT_RING (chan, 0); - BEGIN_RING(chan, RING_3D(POLYGON_STIPPLE_ENABLE), 1); - OUT_RING (chan, 0); - BEGIN_RING(chan, RING_3D(POLYGON_OFFSET_FILL_ENABLE), 1); - OUT_RING (chan, 0); + BEGIN_NV04(push, NV50_3D(VERTEX_TWO_SIDE_ENABLE), 1); + PUSH_DATA (push, 0); + BEGIN_NV04(push, NV50_3D(FRAG_COLOR_CLAMP_EN), 1); + PUSH_DATA (push, 0); + BEGIN_NV04(push, NV50_3D(MULTISAMPLE_ENABLE), 1); + PUSH_DATA (push, 0); + BEGIN_NV04(push, NV50_3D(MSAA_MASK(0)), 4); + PUSH_DATA (push, 0xffff); + PUSH_DATA (push, 0xffff); + PUSH_DATA (push, 0xffff); + PUSH_DATA (push, 0xffff); + BEGIN_NV04(push, NV50_3D(POLYGON_MODE_FRONT), 3); + PUSH_DATA (push, NV50_3D_POLYGON_MODE_FRONT_FILL); + PUSH_DATA (push, NV50_3D_POLYGON_MODE_BACK_FILL); + PUSH_DATA (push, 0); + BEGIN_NV04(push, NV50_3D(CULL_FACE_ENABLE), 1); + PUSH_DATA (push, 0); + BEGIN_NV04(push, NV50_3D(POLYGON_STIPPLE_ENABLE), 1); + PUSH_DATA (push, 0); + BEGIN_NV04(push, NV50_3D(POLYGON_OFFSET_FILL_ENABLE), 1); + PUSH_DATA (push, 0); /* zsa state */ - BEGIN_RING(chan, RING_3D(DEPTH_TEST_ENABLE), 1); - OUT_RING (chan, 0); - BEGIN_RING(chan, RING_3D(STENCIL_ENABLE), 1); - OUT_RING (chan, 0); - BEGIN_RING(chan, RING_3D(ALPHA_TEST_ENABLE), 1); - OUT_RING (chan, 0); + BEGIN_NV04(push, NV50_3D(DEPTH_TEST_ENABLE), 1); + PUSH_DATA (push, 0); + BEGIN_NV04(push, NV50_3D(STENCIL_ENABLE), 1); + PUSH_DATA (push, 0); + BEGIN_NV04(push, NV50_3D(ALPHA_TEST_ENABLE), 1); + PUSH_DATA (push, 0); } static void @@ -834,7 +837,7 @@ nv50_resource_resolve(struct pipe_context *pipe, struct nv50_context *nv50 = nv50_context(pipe); struct nv50_screen *screen = nv50->screen; struct nv50_blitctx *blit = screen->blitctx; - struct nouveau_channel *chan = screen->base.channel; + struct nouveau_pushbuf *push = nv50->base.pushbuf; struct pipe_resource *src = info->src.res; struct pipe_resource *dst = info->dst.res; float x0, x1, y0, y1, z; @@ -873,12 +876,12 @@ nv50_resource_resolve(struct pipe_context *pipe, z = (float)info->src.layer; - BEGIN_RING(chan, RING_3D(FP_START_ID), 1); - OUT_RING (chan, + BEGIN_NV04(push, NV50_3D(FP_START_ID), 1); + PUSH_DATA (push, blit->fp.code_base + blit->fp_offset); - BEGIN_RING(chan, RING_3D(VIEWPORT_TRANSFORM_EN), 1); - OUT_RING (chan, 0); + BEGIN_NV04(push, NV50_3D(VIEWPORT_TRANSFORM_EN), 1); + PUSH_DATA (push, 0); /* Draw a large triangle in screen coordinates covering the whole * render target, with scissors defining the destination region. @@ -886,40 +889,40 @@ nv50_resource_resolve(struct pipe_context *pipe, * arranged in a way to yield the desired offset and scale. */ - BEGIN_RING(chan, RING_3D(SCISSOR_HORIZ(0)), 2); - OUT_RING (chan, (info->dst.x1 << 16) | info->dst.x0); - OUT_RING (chan, (info->dst.y1 << 16) | info->dst.y0); - - BEGIN_RING(chan, RING_3D(VERTEX_BEGIN_GL), 1); - OUT_RING (chan, NV50_3D_VERTEX_BEGIN_GL_PRIMITIVE_TRIANGLES); - BEGIN_RING(chan, RING_3D(VTX_ATTR_3F_X(1)), 3); - OUT_RINGf (chan, x0); - OUT_RINGf (chan, y0); - OUT_RINGf (chan, z); - BEGIN_RING(chan, RING_3D(VTX_ATTR_2F_X(0)), 2); - OUT_RINGf (chan, 0.0f); - OUT_RINGf (chan, 0.0f); - BEGIN_RING(chan, RING_3D(VTX_ATTR_3F_X(1)), 3); - OUT_RINGf (chan, x1); - OUT_RINGf (chan, y0); - OUT_RINGf (chan, z); - BEGIN_RING(chan, RING_3D(VTX_ATTR_2F_X(0)), 2); - OUT_RINGf (chan, 16384 << nv50_miptree(dst)->ms_x); - OUT_RINGf (chan, 0.0f); - BEGIN_RING(chan, RING_3D(VTX_ATTR_3F_X(1)), 3); - OUT_RINGf (chan, x0); - OUT_RINGf (chan, y1); - OUT_RINGf (chan, z); - BEGIN_RING(chan, RING_3D(VTX_ATTR_2F_X(0)), 2); - OUT_RINGf (chan, 0.0f); - OUT_RINGf (chan, 16384 << nv50_miptree(dst)->ms_y); - BEGIN_RING(chan, RING_3D(VERTEX_END_GL), 1); - OUT_RING (chan, 0); + BEGIN_NV04(push, NV50_3D(SCISSOR_HORIZ(0)), 2); + PUSH_DATA (push, (info->dst.x1 << 16) | info->dst.x0); + PUSH_DATA (push, (info->dst.y1 << 16) | info->dst.y0); + + BEGIN_NV04(push, NV50_3D(VERTEX_BEGIN_GL), 1); + PUSH_DATA (push, NV50_3D_VERTEX_BEGIN_GL_PRIMITIVE_TRIANGLES); + BEGIN_NV04(push, NV50_3D(VTX_ATTR_3F_X(1)), 3); + PUSH_DATAf(push, x0); + PUSH_DATAf(push, y0); + PUSH_DATAf(push, z); + BEGIN_NV04(push, NV50_3D(VTX_ATTR_2F_X(0)), 2); + PUSH_DATAf(push, 0.0f); + PUSH_DATAf(push, 0.0f); + BEGIN_NV04(push, NV50_3D(VTX_ATTR_3F_X(1)), 3); + PUSH_DATAf(push, x1); + PUSH_DATAf(push, y0); + PUSH_DATAf(push, z); + BEGIN_NV04(push, NV50_3D(VTX_ATTR_2F_X(0)), 2); + PUSH_DATAf(push, 16384 << nv50_miptree(dst)->ms_x); + PUSH_DATAf(push, 0.0f); + BEGIN_NV04(push, NV50_3D(VTX_ATTR_3F_X(1)), 3); + PUSH_DATAf(push, x0); + PUSH_DATAf(push, y1); + PUSH_DATAf(push, z); + BEGIN_NV04(push, NV50_3D(VTX_ATTR_2F_X(0)), 2); + PUSH_DATAf(push, 0.0f); + PUSH_DATAf(push, 16384 << nv50_miptree(dst)->ms_y); + BEGIN_NV04(push, NV50_3D(VERTEX_END_GL), 1); + PUSH_DATA (push, 0); /* re-enable normally constant state */ - BEGIN_RING(chan, RING_3D(VIEWPORT_TRANSFORM_EN), 1); - OUT_RING (chan, 1); + BEGIN_NV04(push, NV50_3D(VIEWPORT_TRANSFORM_EN), 1); + PUSH_DATA (push, 1); nv50_blitctx_post_blit(nv50, blit); } diff --git a/src/gallium/drivers/nv50/nv50_tex.c b/src/gallium/drivers/nv50/nv50_tex.c index 1be4e9aff19..4a769d5b810 100644 --- a/src/gallium/drivers/nv50/nv50_tex.c +++ b/src/gallium/drivers/nv50/nv50_tex.c @@ -75,6 +75,7 @@ nv50_create_sampler_view(struct pipe_context *pipe, const struct pipe_sampler_view *templ) { const struct util_format_description *desc; + uint64_t addr; uint32_t *tic; uint32_t swz[4]; uint32_t depth; @@ -115,15 +116,25 @@ nv50_create_sampler_view(struct pipe_context *pipe, (swz[2] << NV50_TIC_0_MAPB__SHIFT) | (swz[3] << NV50_TIC_0_MAPA__SHIFT); - tic[1] = /* mt->base.bo->offset; */ 0; - tic[2] = /* mt->base.bo->offset >> 32 */ 0; + addr = mt->base.address; + + if (mt->base.base.target == PIPE_TEXTURE_1D_ARRAY || + mt->base.base.target == PIPE_TEXTURE_2D_ARRAY) { + addr += view->pipe.u.tex.first_layer * mt->layer_stride; + depth = view->pipe.u.tex.last_layer - view->pipe.u.tex.first_layer + 1; + } else { + depth = mt->base.base.depth0; + } + + tic[1] = addr; + tic[2] = (addr >> 32) & 0xff; tic[2] |= 0x10001000 | NV50_TIC_2_NO_BORDER; if (desc->colorspace == UTIL_FORMAT_COLORSPACE_SRGB) tic[2] |= NV50_TIC_2_COLORSPACE_SRGB; - if (unlikely(!nouveau_bo_tile_layout(nv04_resource(texture)->bo))) { + if (unlikely(!nouveau_bo_memtype(nv04_resource(texture)->bo))) { nv50_init_tic_entry_linear(tic, texture); return &view->pipe; } @@ -132,16 +143,8 @@ nv50_create_sampler_view(struct pipe_context *pipe, tic[2] |= NV50_TIC_2_NORMALIZED_COORDS; tic[2] |= - ((mt->base.bo->tile_mode & 0x0f) << (22 - 0)) | - ((mt->base.bo->tile_mode & 0xf0) << (25 - 4)); - - depth = MAX2(mt->base.base.array_size, mt->base.base.depth0); - - if (mt->base.base.target == PIPE_TEXTURE_1D_ARRAY || - mt->base.base.target == PIPE_TEXTURE_2D_ARRAY) { - tic[1] = view->pipe.u.tex.first_layer * mt->layer_stride; - depth = view->pipe.u.tex.last_layer - view->pipe.u.tex.first_layer + 1; - } + ((mt->level[0].tile_mode & 0x0f0) << (22 - 4)) | + ((mt->level[0].tile_mode & 0xf00) << (25 - 8)); switch (mt->base.base.target) { case PIPE_TEXTURE_1D: @@ -196,7 +199,7 @@ nv50_create_sampler_view(struct pipe_context *pipe, static boolean nv50_validate_tic(struct nv50_context *nv50, int s) { - struct nouveau_channel *chan = nv50->screen->base.channel; + struct nouveau_pushbuf *push = nv50->base.pushbuf; struct nouveau_bo *txc = nv50->screen->txc; unsigned i; boolean need_flush = FALSE; @@ -206,53 +209,46 @@ nv50_validate_tic(struct nv50_context *nv50, int s) struct nv04_resource *res; if (!tic) { - BEGIN_RING(chan, RING_3D(BIND_TIC(s)), 1); - OUT_RING (chan, (i << 1) | 0); + BEGIN_NV04(push, NV50_3D(BIND_TIC(s)), 1); + PUSH_DATA (push, (i << 1) | 0); continue; } res = &nv50_miptree(tic->pipe.texture)->base; if (tic->id < 0) { - uint32_t offset = tic->tic[1]; - tic->id = nv50_screen_tic_alloc(nv50->screen, tic); - MARK_RING (chan, 24 + 8, 4); - BEGIN_RING(chan, RING_2D(DST_FORMAT), 2); - OUT_RING (chan, NV50_SURFACE_FORMAT_R8_UNORM); - OUT_RING (chan, 1); - BEGIN_RING(chan, RING_2D(DST_PITCH), 5); - OUT_RING (chan, 262144); - OUT_RING (chan, 65536); - OUT_RING (chan, 1); - OUT_RELOCh(chan, txc, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); - OUT_RELOCl(chan, txc, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); - BEGIN_RING(chan, RING_2D(SIFC_BITMAP_ENABLE), 2); - OUT_RING (chan, 0); - OUT_RING (chan, NV50_SURFACE_FORMAT_R8_UNORM); - BEGIN_RING(chan, RING_2D(SIFC_WIDTH), 10); - OUT_RING (chan, 32); - OUT_RING (chan, 1); - OUT_RING (chan, 0); - OUT_RING (chan, 1); - OUT_RING (chan, 0); - OUT_RING (chan, 1); - OUT_RING (chan, 0); - OUT_RING (chan, tic->id * 32); - OUT_RING (chan, 0); - OUT_RING (chan, 0); - BEGIN_RING_NI(chan, RING_2D(SIFC_DATA), 8); - OUT_RING (chan, tic->tic[0]); - OUT_RELOCl(chan, res->bo, offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD); - OUT_RELOC (chan, res->bo, offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | - NOUVEAU_BO_HIGH | NOUVEAU_BO_OR, tic->tic[2], tic->tic[2]); - OUT_RINGp (chan, &tic->tic[3], 5); + BEGIN_NV04(push, NV50_2D(DST_FORMAT), 2); + PUSH_DATA (push, NV50_SURFACE_FORMAT_R8_UNORM); + PUSH_DATA (push, 1); + BEGIN_NV04(push, NV50_2D(DST_PITCH), 5); + PUSH_DATA (push, 262144); + PUSH_DATA (push, 65536); + PUSH_DATA (push, 1); + PUSH_DATAh(push, txc->offset); + PUSH_DATA (push, txc->offset); + BEGIN_NV04(push, NV50_2D(SIFC_BITMAP_ENABLE), 2); + PUSH_DATA (push, 0); + PUSH_DATA (push, NV50_SURFACE_FORMAT_R8_UNORM); + BEGIN_NV04(push, NV50_2D(SIFC_WIDTH), 10); + PUSH_DATA (push, 32); + PUSH_DATA (push, 1); + PUSH_DATA (push, 0); + PUSH_DATA (push, 1); + PUSH_DATA (push, 0); + PUSH_DATA (push, 1); + PUSH_DATA (push, 0); + PUSH_DATA (push, tic->id * 32); + PUSH_DATA (push, 0); + PUSH_DATA (push, 0); + BEGIN_NI04(push, NV50_2D(SIFC_DATA), 8); + PUSH_DATAp(push, &tic->tic[0], 8); need_flush = TRUE; } else if (res->status & NOUVEAU_BUFFER_STATUS_GPU_WRITING) { - BEGIN_RING(chan, RING_3D(TEX_CACHE_CTL), 1); - OUT_RING (chan, 0x20); //(tic->id << 4) | 1); + BEGIN_NV04(push, NV50_3D(TEX_CACHE_CTL), 1); + PUSH_DATA (push, 0x20); } nv50->screen->tic.lock[tic->id / 32] |= 1 << (tic->id % 32); @@ -260,15 +256,14 @@ nv50_validate_tic(struct nv50_context *nv50, int s) res->status &= NOUVEAU_BUFFER_STATUS_GPU_WRITING; res->status |= NOUVEAU_BUFFER_STATUS_GPU_READING; - nv50_bufctx_add_resident(nv50, NV50_BUFCTX_TEXTURES, res, - NOUVEAU_BO_VRAM | NOUVEAU_BO_RD); + BCTX_REFN(nv50->bufctx_3d, TEXTURES, res, RD); - BEGIN_RING(chan, RING_3D(BIND_TIC(s)), 1); - OUT_RING (chan, (tic->id << 9) | (i << 1) | 1); + BEGIN_NV04(push, NV50_3D(BIND_TIC(s)), 1); + PUSH_DATA (push, (tic->id << 9) | (i << 1) | 1); } for (; i < nv50->state.num_textures[s]; ++i) { - BEGIN_RING(chan, RING_3D(BIND_TIC(s)), 1); - OUT_RING (chan, (i << 1) | 0); + BEGIN_NV04(push, NV50_3D(BIND_TIC(s)), 1); + PUSH_DATA (push, (i << 1) | 0); } nv50->state.num_textures[s] = nv50->num_textures[s]; @@ -283,15 +278,15 @@ void nv50_validate_textures(struct nv50_context *nv50) need_flush |= nv50_validate_tic(nv50, 2); if (need_flush) { - BEGIN_RING(nv50->screen->base.channel, RING_3D(TIC_FLUSH), 1); - OUT_RING (nv50->screen->base.channel, 0); + BEGIN_NV04(nv50->base.pushbuf, NV50_3D(TIC_FLUSH), 1); + PUSH_DATA (nv50->base.pushbuf, 0); } } static boolean nv50_validate_tsc(struct nv50_context *nv50, int s) { - struct nouveau_channel *chan = nv50->screen->base.channel; + struct nouveau_pushbuf *push = nv50->base.pushbuf; unsigned i; boolean need_flush = FALSE; @@ -299,8 +294,8 @@ nv50_validate_tsc(struct nv50_context *nv50, int s) struct nv50_tsc_entry *tsc = nv50_tsc_entry(nv50->samplers[s][i]); if (!tsc) { - BEGIN_RING(chan, RING_3D(BIND_TSC(s)), 1); - OUT_RING (chan, (i << 4) | 0); + BEGIN_NV04(push, NV50_3D(BIND_TSC(s)), 1); + PUSH_DATA (push, (i << 4) | 0); continue; } if (tsc->id < 0) { @@ -313,12 +308,12 @@ nv50_validate_tsc(struct nv50_context *nv50, int s) } nv50->screen->tsc.lock[tsc->id / 32] |= 1 << (tsc->id % 32); - BEGIN_RING(chan, RING_3D(BIND_TSC(s)), 1); - OUT_RING (chan, (tsc->id << 12) | (i << 4) | 1); + BEGIN_NV04(push, NV50_3D(BIND_TSC(s)), 1); + PUSH_DATA (push, (tsc->id << 12) | (i << 4) | 1); } for (; i < nv50->state.num_samplers[s]; ++i) { - BEGIN_RING(chan, RING_3D(BIND_TSC(s)), 1); - OUT_RING (chan, (i << 4) | 0); + BEGIN_NV04(push, NV50_3D(BIND_TSC(s)), 1); + PUSH_DATA (push, (i << 4) | 0); } nv50->state.num_samplers[s] = nv50->num_samplers[s]; @@ -333,7 +328,7 @@ void nv50_validate_samplers(struct nv50_context *nv50) need_flush |= nv50_validate_tsc(nv50, 2); if (need_flush) { - BEGIN_RING(nv50->screen->base.channel, RING_3D(TSC_FLUSH), 1); - OUT_RING (nv50->screen->base.channel, 0); + BEGIN_NV04(nv50->base.pushbuf, NV50_3D(TSC_FLUSH), 1); + PUSH_DATA (nv50->base.pushbuf, 0); } } diff --git a/src/gallium/drivers/nv50/nv50_transfer.c b/src/gallium/drivers/nv50/nv50_transfer.c index 6f860e73348..b960dc0d191 100644 --- a/src/gallium/drivers/nv50/nv50_transfer.c +++ b/src/gallium/drivers/nv50/nv50_transfer.c @@ -50,12 +50,13 @@ nv50_m2mf_rect_setup(struct nv50_m2mf_rect *rect, } void -nv50_m2mf_transfer_rect(struct pipe_screen *pscreen, +nv50_m2mf_transfer_rect(struct nv50_context *nv50, const struct nv50_m2mf_rect *dst, const struct nv50_m2mf_rect *src, uint32_t nblocksx, uint32_t nblocksy) { - struct nouveau_channel *chan = nouveau_screen(pscreen)->channel; + struct nouveau_pushbuf *push = nv50->base.pushbuf; + struct nouveau_bufctx *bctx = nv50->bufctx; const int cpp = dst->cpp; uint32_t src_ofst = src->base; uint32_t dst_ofst = dst->base; @@ -65,76 +66,81 @@ nv50_m2mf_transfer_rect(struct pipe_screen *pscreen, assert(dst->cpp == src->cpp); - if (nouveau_bo_tile_layout(src->bo)) { - BEGIN_RING(chan, RING_MF(LINEAR_IN), 6); - OUT_RING (chan, 0); - OUT_RING (chan, src->tile_mode << 4); - OUT_RING (chan, src->width * cpp); - OUT_RING (chan, src->height); - OUT_RING (chan, src->depth); - OUT_RING (chan, src->z); + nouveau_bufctx_refn(bctx, 0, src->bo, src->domain | NOUVEAU_BO_RD); + nouveau_bufctx_refn(bctx, 0, dst->bo, dst->domain | NOUVEAU_BO_WR); + nouveau_pushbuf_bufctx(push, bctx); + nouveau_pushbuf_validate(push); + + if (nouveau_bo_memtype(src->bo)) { + BEGIN_NV04(push, NV50_M2MF(LINEAR_IN), 6); + PUSH_DATA (push, 0); + PUSH_DATA (push, src->tile_mode); + PUSH_DATA (push, src->width * cpp); + PUSH_DATA (push, src->height); + PUSH_DATA (push, src->depth); + PUSH_DATA (push, src->z); } else { src_ofst += src->y * src->pitch + src->x * cpp; - BEGIN_RING(chan, RING_MF(LINEAR_IN), 1); - OUT_RING (chan, 1); - BEGIN_RING(chan, RING_MF_(NV04_M2MF_PITCH_IN), 1); - OUT_RING (chan, src->pitch); + BEGIN_NV04(push, NV50_M2MF(LINEAR_IN), 1); + PUSH_DATA (push, 1); + BEGIN_NV04(push, SUBC_M2MF(NV03_M2MF_PITCH_IN), 1); + PUSH_DATA (push, src->pitch); } - if (nouveau_bo_tile_layout(dst->bo)) { - BEGIN_RING(chan, RING_MF(LINEAR_OUT), 6); - OUT_RING (chan, 0); - OUT_RING (chan, dst->tile_mode << 4); - OUT_RING (chan, dst->width * cpp); - OUT_RING (chan, dst->height); - OUT_RING (chan, dst->depth); - OUT_RING (chan, dst->z); + if (nouveau_bo_memtype(dst->bo)) { + BEGIN_NV04(push, NV50_M2MF(LINEAR_OUT), 6); + PUSH_DATA (push, 0); + PUSH_DATA (push, dst->tile_mode); + PUSH_DATA (push, dst->width * cpp); + PUSH_DATA (push, dst->height); + PUSH_DATA (push, dst->depth); + PUSH_DATA (push, dst->z); } else { dst_ofst += dst->y * dst->pitch + dst->x * cpp; - BEGIN_RING(chan, RING_MF(LINEAR_OUT), 1); - OUT_RING (chan, 1); - BEGIN_RING(chan, RING_MF_(NV04_M2MF_PITCH_OUT), 1); - OUT_RING (chan, dst->pitch); + BEGIN_NV04(push, NV50_M2MF(LINEAR_OUT), 1); + PUSH_DATA (push, 1); + BEGIN_NV04(push, SUBC_M2MF(NV03_M2MF_PITCH_OUT), 1); + PUSH_DATA (push, dst->pitch); } while (height) { int line_count = height > 2047 ? 2047 : height; - MARK_RING (chan, 17, 4); + BEGIN_NV04(push, NV50_M2MF(OFFSET_IN_HIGH), 2); + PUSH_DATAh(push, src->bo->offset + src_ofst); + PUSH_DATAh(push, dst->bo->offset + dst_ofst); - BEGIN_RING(chan, RING_MF(OFFSET_IN_HIGH), 2); - OUT_RELOCh(chan, src->bo, src_ofst, src->domain | NOUVEAU_BO_RD); - OUT_RELOCh(chan, dst->bo, dst_ofst, dst->domain | NOUVEAU_BO_WR); + BEGIN_NV04(push, SUBC_M2MF(NV03_M2MF_OFFSET_IN), 2); + PUSH_DATA (push, src->bo->offset + src_ofst); + PUSH_DATA (push, dst->bo->offset + dst_ofst); - BEGIN_RING(chan, RING_MF_(NV04_M2MF_OFFSET_IN), 2); - OUT_RELOCl(chan, src->bo, src_ofst, src->domain | NOUVEAU_BO_RD); - OUT_RELOCl(chan, dst->bo, dst_ofst, dst->domain | NOUVEAU_BO_WR); - - if (nouveau_bo_tile_layout(src->bo)) { - BEGIN_RING(chan, RING_MF(TILING_POSITION_IN), 1); - OUT_RING (chan, (sy << 16) | (src->x * cpp)); + if (nouveau_bo_memtype(src->bo)) { + BEGIN_NV04(push, NV50_M2MF(TILING_POSITION_IN), 1); + PUSH_DATA (push, (sy << 16) | (src->x * cpp)); } else { src_ofst += line_count * src->pitch; } - if (nouveau_bo_tile_layout(dst->bo)) { - BEGIN_RING(chan, RING_MF(TILING_POSITION_OUT), 1); - OUT_RING (chan, (dy << 16) | (dst->x * cpp)); + if (nouveau_bo_memtype(dst->bo)) { + BEGIN_NV04(push, NV50_M2MF(TILING_POSITION_OUT), 1); + PUSH_DATA (push, (dy << 16) | (dst->x * cpp)); } else { dst_ofst += line_count * dst->pitch; } - BEGIN_RING(chan, RING_MF_(NV04_M2MF_LINE_LENGTH_IN), 4); - OUT_RING (chan, nblocksx * cpp); - OUT_RING (chan, line_count); - OUT_RING (chan, (1 << 8) | (1 << 0)); - OUT_RING (chan, 0); + BEGIN_NV04(push, SUBC_M2MF(NV03_M2MF_LINE_LENGTH_IN), 4); + PUSH_DATA (push, nblocksx * cpp); + PUSH_DATA (push, line_count); + PUSH_DATA (push, (1 << 8) | (1 << 0)); + PUSH_DATA (push, 0); height -= line_count; sy += line_count; dy += line_count; } + + nouveau_bufctx_reset(bctx, 0); } void @@ -142,55 +148,60 @@ nv50_sifc_linear_u8(struct nouveau_context *nv, struct nouveau_bo *dst, unsigned offset, unsigned domain, unsigned size, const void *data) { - struct nouveau_channel *chan = nv->screen->channel; + struct nv50_context *nv50 = nv50_context(&nv->pipe); + struct nouveau_pushbuf *push = nv50->base.pushbuf; uint32_t *src = (uint32_t *)data; unsigned count = (size + 3) / 4; unsigned xcoord = offset & 0xff; + nouveau_bufctx_refn(nv50->bufctx, 0, dst, domain | NOUVEAU_BO_WR); + nouveau_pushbuf_bufctx(push, nv50->bufctx); + nouveau_pushbuf_validate(push); + offset &= ~0xff; - MARK_RING (chan, 23, 4); - BEGIN_RING(chan, RING_2D(DST_FORMAT), 2); - OUT_RING (chan, NV50_SURFACE_FORMAT_R8_UNORM); - OUT_RING (chan, 1); - BEGIN_RING(chan, RING_2D(DST_PITCH), 5); - OUT_RING (chan, 262144); - OUT_RING (chan, 65536); - OUT_RING (chan, 1); - OUT_RELOCh(chan, dst, offset, domain | NOUVEAU_BO_WR); - OUT_RELOCl(chan, dst, offset, domain | NOUVEAU_BO_WR); - BEGIN_RING(chan, RING_2D(SIFC_BITMAP_ENABLE), 2); - OUT_RING (chan, 0); - OUT_RING (chan, NV50_SURFACE_FORMAT_R8_UNORM); - BEGIN_RING(chan, RING_2D(SIFC_WIDTH), 10); - OUT_RING (chan, size); - OUT_RING (chan, 1); - OUT_RING (chan, 0); - OUT_RING (chan, 1); - OUT_RING (chan, 0); - OUT_RING (chan, 1); - OUT_RING (chan, 0); - OUT_RING (chan, xcoord); - OUT_RING (chan, 0); - OUT_RING (chan, 0); + BEGIN_NV04(push, NV50_2D(DST_FORMAT), 2); + PUSH_DATA (push, NV50_SURFACE_FORMAT_R8_UNORM); + PUSH_DATA (push, 1); + BEGIN_NV04(push, NV50_2D(DST_PITCH), 5); + PUSH_DATA (push, 262144); + PUSH_DATA (push, 65536); + PUSH_DATA (push, 1); + PUSH_DATAh(push, dst->offset + offset); + PUSH_DATA (push, dst->offset + offset); + BEGIN_NV04(push, NV50_2D(SIFC_BITMAP_ENABLE), 2); + PUSH_DATA (push, 0); + PUSH_DATA (push, NV50_SURFACE_FORMAT_R8_UNORM); + BEGIN_NV04(push, NV50_2D(SIFC_WIDTH), 10); + PUSH_DATA (push, size); + PUSH_DATA (push, 1); + PUSH_DATA (push, 0); + PUSH_DATA (push, 1); + PUSH_DATA (push, 0); + PUSH_DATA (push, 1); + PUSH_DATA (push, 0); + PUSH_DATA (push, xcoord); + PUSH_DATA (push, 0); + PUSH_DATA (push, 0); while (count) { - unsigned nr = AVAIL_RING(chan); + unsigned nr; - if (nr < 9) { - FIRE_RING(chan); - nouveau_bo_validate(chan, dst, NOUVEAU_BO_WR); - continue; - } + if (!PUSH_SPACE(push, 16)) + break; + nr = PUSH_AVAIL(push); + assert(nr >= 16); nr = MIN2(count, nr - 1); nr = MIN2(nr, NV04_PFIFO_MAX_PACKET_LEN); - BEGIN_RING_NI(chan, RING_2D(SIFC_DATA), nr); - OUT_RINGp (chan, src, nr); + BEGIN_NI04(push, NV50_2D(SIFC_DATA), nr); + PUSH_DATAp(push, src, nr); src += nr; count -= nr; } + + nouveau_bufctx_reset(nv50->bufctx, 0); } void @@ -199,33 +210,40 @@ nv50_m2mf_copy_linear(struct nouveau_context *nv, struct nouveau_bo *src, unsigned srcoff, unsigned srcdom, unsigned size) { - struct nouveau_channel *chan = nv->screen->channel; + struct nouveau_pushbuf *push = nv->pushbuf; + struct nouveau_bufctx *bctx = nv50_context(&nv->pipe)->bufctx; + + nouveau_bufctx_refn(bctx, 0, src, srcdom | NOUVEAU_BO_RD); + nouveau_bufctx_refn(bctx, 0, dst, dstdom | NOUVEAU_BO_WR); + nouveau_pushbuf_bufctx(push, bctx); + nouveau_pushbuf_validate(push); - BEGIN_RING(chan, RING_MF(LINEAR_IN), 1); - OUT_RING (chan, 1); - BEGIN_RING(chan, RING_MF(LINEAR_OUT), 1); - OUT_RING (chan, 1); + BEGIN_NV04(push, NV50_M2MF(LINEAR_IN), 1); + PUSH_DATA (push, 1); + BEGIN_NV04(push, NV50_M2MF(LINEAR_OUT), 1); + PUSH_DATA (push, 1); while (size) { unsigned bytes = MIN2(size, 1 << 17); - MARK_RING (chan, 11, 4); - BEGIN_RING(chan, RING_MF(OFFSET_IN_HIGH), 2); - OUT_RELOCh(chan, src, srcoff, srcdom | NOUVEAU_BO_RD); - OUT_RELOCh(chan, dst, dstoff, dstdom | NOUVEAU_BO_WR); - BEGIN_RING(chan, RING_MF_(NV04_M2MF_OFFSET_IN), 2); - OUT_RELOCl(chan, src, srcoff, srcdom | NOUVEAU_BO_RD); - OUT_RELOCl(chan, dst, dstoff, dstdom | NOUVEAU_BO_WR); - BEGIN_RING(chan, RING_MF_(NV04_M2MF_LINE_LENGTH_IN), 4); - OUT_RING (chan, bytes); - OUT_RING (chan, 1); - OUT_RING (chan, (1 << 8) | (1 << 0)); - OUT_RING (chan, 0); + BEGIN_NV04(push, NV50_M2MF(OFFSET_IN_HIGH), 2); + PUSH_DATAh(push, src->offset + srcoff); + PUSH_DATAh(push, dst->offset + dstoff); + BEGIN_NV04(push, SUBC_M2MF(NV03_M2MF_OFFSET_IN), 2); + PUSH_DATA (push, src->offset + srcoff); + PUSH_DATA (push, dst->offset + dstoff); + BEGIN_NV04(push, SUBC_M2MF(NV03_M2MF_LINE_LENGTH_IN), 4); + PUSH_DATA (push, bytes); + PUSH_DATA (push, 1); + PUSH_DATA (push, (1 << 8) | (1 << 0)); + PUSH_DATA (push, 0); srcoff += bytes; dstoff += bytes; size -= bytes; } + + nouveau_bufctx_reset(bctx, 0); } struct pipe_transfer * @@ -236,7 +254,6 @@ nv50_miptree_transfer_new(struct pipe_context *pctx, const struct pipe_box *box) { struct nv50_context *nv50 = nv50_context(pctx); - struct pipe_screen *pscreen = pctx->screen; struct nouveau_device *dev = nv50->screen->base.device; const struct nv50_miptree *mt = nv50_miptree(res); struct nv50_transfer *tx; @@ -272,7 +289,7 @@ nv50_miptree_transfer_new(struct pipe_context *pctx, size = tx->base.layer_stride; ret = nouveau_bo_new(dev, NOUVEAU_BO_GART | NOUVEAU_BO_MAP, 0, - size * tx->base.box.depth, &tx->rect[1].bo); + size * tx->base.box.depth, NULL, &tx->rect[1].bo); if (ret) { FREE(tx); return NULL; @@ -290,7 +307,7 @@ nv50_miptree_transfer_new(struct pipe_context *pctx, unsigned z = tx->rect[0].z; unsigned i; for (i = 0; i < box->depth; ++i) { - nv50_m2mf_transfer_rect(pscreen, &tx->rect[1], &tx->rect[0], + nv50_m2mf_transfer_rect(nv50, &tx->rect[1], &tx->rect[0], tx->nblocksx, tx->nblocksy); if (mt->layout_3d) tx->rect[0].z++; @@ -310,14 +327,14 @@ void nv50_miptree_transfer_del(struct pipe_context *pctx, struct pipe_transfer *transfer) { - struct pipe_screen *pscreen = pctx->screen; + struct nv50_context *nv50 = nv50_context(pctx); struct nv50_transfer *tx = (struct nv50_transfer *)transfer; struct nv50_miptree *mt = nv50_miptree(tx->base.resource); unsigned i; if (tx->base.usage & PIPE_TRANSFER_WRITE) { for (i = 0; i < tx->base.box.depth; ++i) { - nv50_m2mf_transfer_rect(pscreen, &tx->rect[0], &tx->rect[1], + nv50_m2mf_transfer_rect(nv50, &tx->rect[0], &tx->rect[1], tx->nblocksx, tx->nblocksy); if (mt->layout_3d) tx->rect[0].z++; @@ -337,6 +354,7 @@ void * nv50_miptree_transfer_map(struct pipe_context *pctx, struct pipe_transfer *transfer) { + struct nv50_screen *screen = nv50_screen(pctx->screen); struct nv50_transfer *tx = (struct nv50_transfer *)transfer; int ret; unsigned flags = 0; @@ -349,7 +367,7 @@ nv50_miptree_transfer_map(struct pipe_context *pctx, if (transfer->usage & PIPE_TRANSFER_WRITE) flags |= NOUVEAU_BO_WR; - ret = nouveau_bo_map(tx->rect[1].bo, flags); + ret = nouveau_bo_map(tx->rect[1].bo, flags, screen->base.client); if (ret) return NULL; return tx->rect[1].bo->map; @@ -359,9 +377,7 @@ void nv50_miptree_transfer_unmap(struct pipe_context *pctx, struct pipe_transfer *transfer) { - struct nv50_transfer *tx = (struct nv50_transfer *)transfer; - - nouveau_bo_unmap(tx->rect[1].bo); + /* nothing to do */ } void @@ -370,30 +386,36 @@ nv50_cb_push(struct nouveau_context *nv, unsigned base, unsigned size, unsigned offset, unsigned words, const uint32_t *data) { - struct nouveau_channel *chan = nv->screen->channel; + struct nouveau_pushbuf *push = nv->pushbuf; + struct nouveau_bufctx *bctx = nv50_context(&nv->pipe)->bufctx; assert(!(offset & 3)); size = align(size, 0x100); + nouveau_bufctx_refn(bctx, 0, bo, NOUVEAU_BO_WR | domain); + nouveau_pushbuf_bufctx(push, bctx); + nouveau_pushbuf_validate(push); + while (words) { unsigned nr; - MARK_RING(chan, 24, 2); - nr = AVAIL_RING(chan); + nr = PUSH_AVAIL(push); nr = MIN2(nr - 7, words); nr = MIN2(nr, NV04_PFIFO_MAX_PACKET_LEN - 1); - BEGIN_RING(chan, RING_3D(CB_DEF_ADDRESS_HIGH), 3); - OUT_RELOCh(chan, bo, base, domain | NOUVEAU_BO_WR); - OUT_RELOCl(chan, bo, base, domain | NOUVEAU_BO_WR); - OUT_RING (chan, (NV50_CB_TMP << 16) | (size & 0xffff)); - BEGIN_RING(chan, RING_3D(CB_ADDR), 1); - OUT_RING (chan, (offset << 6) | NV50_CB_TMP); - BEGIN_RING_NI(chan, RING_3D(CB_DATA(0)), nr); - OUT_RINGp (chan, data, nr); + BEGIN_NV04(push, NV50_3D(CB_DEF_ADDRESS_HIGH), 3); + PUSH_DATAh(push, bo->offset + base); + PUSH_DATA (push, bo->offset + base); + PUSH_DATA (push, (NV50_CB_TMP << 16) | (size & 0xffff)); + BEGIN_NV04(push, NV50_3D(CB_ADDR), 1); + PUSH_DATA (push, (offset << 6) | NV50_CB_TMP); + BEGIN_NI04(push, NV50_3D(CB_DATA(0)), nr); + PUSH_DATAp(push, data, nr); words -= nr; data += nr; offset += nr * 4; } + + nouveau_bufctx_reset(bctx, 0); } diff --git a/src/gallium/drivers/nv50/nv50_vbo.c b/src/gallium/drivers/nv50/nv50_vbo.c index cb3ea382f3d..9be43ef15d4 100644 --- a/src/gallium/drivers/nv50/nv50_vbo.c +++ b/src/gallium/drivers/nv50/nv50_vbo.c @@ -124,7 +124,7 @@ nv50_emit_vtxattr(struct nv50_context *nv50, struct pipe_vertex_buffer *vb, struct pipe_vertex_element *ve, unsigned attr) { const void *data; - struct nouveau_channel *chan = nv50->screen->base.channel; + struct nouveau_pushbuf *push = nv50->base.pushbuf; struct nv04_resource *res = nv04_resource(vb->buffer); float v[4]; const unsigned nc = util_format_get_nr_components(ve->src_format); @@ -136,30 +136,30 @@ nv50_emit_vtxattr(struct nv50_context *nv50, struct pipe_vertex_buffer *vb, switch (nc) { case 4: - BEGIN_RING(chan, RING_3D(VTX_ATTR_4F_X(attr)), 4); - OUT_RINGf (chan, v[0]); - OUT_RINGf (chan, v[1]); - OUT_RINGf (chan, v[2]); - OUT_RINGf (chan, v[3]); + BEGIN_NV04(push, NV50_3D(VTX_ATTR_4F_X(attr)), 4); + PUSH_DATAf(push, v[0]); + PUSH_DATAf(push, v[1]); + PUSH_DATAf(push, v[2]); + PUSH_DATAf(push, v[3]); break; case 3: - BEGIN_RING(chan, RING_3D(VTX_ATTR_3F_X(attr)), 3); - OUT_RINGf (chan, v[0]); - OUT_RINGf (chan, v[1]); - OUT_RINGf (chan, v[2]); + BEGIN_NV04(push, NV50_3D(VTX_ATTR_3F_X(attr)), 3); + PUSH_DATAf(push, v[0]); + PUSH_DATAf(push, v[1]); + PUSH_DATAf(push, v[2]); break; case 2: - BEGIN_RING(chan, RING_3D(VTX_ATTR_2F_X(attr)), 2); - OUT_RINGf (chan, v[0]); - OUT_RINGf (chan, v[1]); + BEGIN_NV04(push, NV50_3D(VTX_ATTR_2F_X(attr)), 2); + PUSH_DATAf(push, v[0]); + PUSH_DATAf(push, v[1]); break; case 1: if (attr == nv50->vertprog->vp.edgeflag) { - BEGIN_RING(chan, RING_3D(EDGEFLAG_ENABLE), 1); - OUT_RING (chan, v[0] ? 1 : 0); + BEGIN_NV04(push, NV50_3D(EDGEFLAG), 1); + PUSH_DATA (push, v[0] ? 1 : 0); } - BEGIN_RING(chan, RING_3D(VTX_ATTR_1F(attr)), 1); - OUT_RINGf (chan, v[0]); + BEGIN_NV04(push, NV50_3D(VTX_ATTR_1F(attr)), 1); + PUSH_DATAf(push, v[0]); break; default: assert(0); @@ -193,7 +193,7 @@ nv50_prevalidate_vbufs(struct nv50_context *nv50) nv50->vbo_fifo = nv50->vbo_user = 0; - nv50_bufctx_reset(nv50, NV50_BUFCTX_VERTEX); + nouveau_bufctx_reset(nv50->bufctx_3d, NV50_BIND_VERTEX); for (i = 0; i < nv50->num_vtxbufs; ++i) { vb = &nv50->vtxbuf[i]; @@ -211,33 +211,37 @@ nv50_prevalidate_vbufs(struct nv50_context *nv50) nv50->vbo_user |= 1 << i; assert(vb->stride > vb->buffer_offset); nv50_vbuf_range(nv50, i, &base, &size); - nouveau_user_buffer_upload(buf, base, size); + nouveau_user_buffer_upload(&nv50->base, buf, base, size); } else { nouveau_buffer_migrate(&nv50->base, buf, NOUVEAU_BO_GART); } nv50->base.vbo_dirty = TRUE; } } - nv50_bufctx_add_resident(nv50, NV50_BUFCTX_VERTEX, buf, NOUVEAU_BO_RD); + BCTX_REFN(nv50->bufctx_3d, VERTEX, buf, RD); } } static void nv50_update_user_vbufs(struct nv50_context *nv50) { - struct nouveau_channel *chan = nv50->screen->base.channel; + struct nouveau_pushbuf *push = nv50->base.pushbuf; uint32_t base, offset, size; int i; uint32_t written = 0; + nouveau_bufctx_reset(nv50->bufctx_3d, NV50_BIND_VERTEX); + for (i = 0; i < nv50->vertex->num_elements; ++i) { struct pipe_vertex_element *ve = &nv50->vertex->element[i].pipe; const int b = ve->vertex_buffer_index; struct pipe_vertex_buffer *vb = &nv50->vtxbuf[b]; struct nv04_resource *buf = nv04_resource(vb->buffer); - if (!(nv50->vbo_user & (1 << b))) + if (!(nv50->vbo_user & (1 << b))) { + BCTX_REFN(nv50->bufctx_3d, VERTEX, buf, RD); continue; + } if (!vb->stride) { nv50_emit_vtxattr(nv50, vb, ve, i); @@ -247,17 +251,18 @@ nv50_update_user_vbufs(struct nv50_context *nv50) if (!(written & (1 << b))) { written |= 1 << b; - nouveau_user_buffer_upload(buf, base, size); + nouveau_user_buffer_upload(&nv50->base, buf, base, size); } offset = vb->buffer_offset + ve->src_offset; - MARK_RING (chan, 6, 4); - BEGIN_RING(chan, RING_3D(VERTEX_ARRAY_LIMIT_HIGH(i)), 2); - OUT_RESRCh(chan, buf, base + size - 1, NOUVEAU_BO_RD); - OUT_RESRCl(chan, buf, base + size - 1, NOUVEAU_BO_RD); - BEGIN_RING(chan, RING_3D(VERTEX_ARRAY_START_HIGH(i)), 2); - OUT_RESRCh(chan, buf, offset, NOUVEAU_BO_RD); - OUT_RESRCl(chan, buf, offset, NOUVEAU_BO_RD); + BEGIN_NV04(push, NV50_3D(VERTEX_ARRAY_LIMIT_HIGH(i)), 2); + PUSH_DATAh(push, buf->address + base + size - 1); + PUSH_DATA (push, buf->address + base + size - 1); + BEGIN_NV04(push, NV50_3D(VERTEX_ARRAY_START_HIGH(i)), 2); + PUSH_DATAh(push, buf->address + offset); + PUSH_DATA (push, buf->address + offset); + + BCTX_REFN(nv50->bufctx_3d, VERTEX, buf, RD); } nv50->base.vbo_dirty = TRUE; } @@ -278,7 +283,7 @@ nv50_release_user_vbufs(struct nv50_context *nv50) void nv50_vertex_arrays_validate(struct nv50_context *nv50) { - struct nouveau_channel *chan = nv50->screen->base.channel; + struct nouveau_pushbuf *push = nv50->base.pushbuf; struct nv50_vertex_stateobj *vertex = nv50->vertex; struct pipe_vertex_buffer *vb; struct nv50_vertex_element *ve; @@ -291,15 +296,15 @@ nv50_vertex_arrays_validate(struct nv50_context *nv50) nv50_prevalidate_vbufs(nv50); } - BEGIN_RING(chan, RING_3D(VERTEX_ARRAY_ATTRIB(0)), vertex->num_elements); + BEGIN_NV04(push, NV50_3D(VERTEX_ARRAY_ATTRIB(0)), vertex->num_elements); for (i = 0; i < vertex->num_elements; ++i) { ve = &vertex->element[i]; vb = &nv50->vtxbuf[ve->pipe.vertex_buffer_index]; if (likely(vb->stride) || nv50->vbo_fifo) { - OUT_RING(chan, ve->state); + PUSH_DATA(push, ve->state); } else { - OUT_RING(chan, ve->state | NV50_3D_VERTEX_ARRAY_ATTRIB_CONST); + PUSH_DATA(push, ve->state | NV50_3D_VERTEX_ARRAY_ATTRIB_CONST); nv50->vbo_fifo &= ~(1 << i); } } @@ -313,15 +318,15 @@ nv50_vertex_arrays_validate(struct nv50_context *nv50) if (unlikely(ve->pipe.instance_divisor)) { if (!(nv50->state.instance_elts & (1 << i))) { - BEGIN_RING(chan, RING_3D(VERTEX_ARRAY_PER_INSTANCE(i)), 1); - OUT_RING (chan, 1); + BEGIN_NV04(push, NV50_3D(VERTEX_ARRAY_PER_INSTANCE(i)), 1); + PUSH_DATA (push, 1); } - BEGIN_RING(chan, RING_3D(VERTEX_ARRAY_DIVISOR(i)), 1); - OUT_RING (chan, ve->pipe.instance_divisor); + BEGIN_NV04(push, NV50_3D(VERTEX_ARRAY_DIVISOR(i)), 1); + PUSH_DATA (push, ve->pipe.instance_divisor); } else if (unlikely(nv50->state.instance_elts & (1 << i))) { - BEGIN_RING(chan, RING_3D(VERTEX_ARRAY_PER_INSTANCE(i)), 1); - OUT_RING (chan, 0); + BEGIN_NV04(push, NV50_3D(VERTEX_ARRAY_PER_INSTANCE(i)), 1); + PUSH_DATA (push, 0); } res = nv04_resource(vb->buffer); @@ -329,33 +334,32 @@ nv50_vertex_arrays_validate(struct nv50_context *nv50) if (nv50->vbo_fifo || unlikely(vb->stride == 0)) { if (!nv50->vbo_fifo) nv50_emit_vtxattr(nv50, vb, &ve->pipe, i); - BEGIN_RING(chan, RING_3D(VERTEX_ARRAY_FETCH(i)), 1); - OUT_RING (chan, 0); + BEGIN_NV04(push, NV50_3D(VERTEX_ARRAY_FETCH(i)), 1); + PUSH_DATA (push, 0); continue; } size = vb->buffer->width0; offset = ve->pipe.src_offset + vb->buffer_offset; - MARK_RING (chan, 8, 4); - BEGIN_RING(chan, RING_3D(VERTEX_ARRAY_FETCH(i)), 1); - OUT_RING (chan, NV50_3D_VERTEX_ARRAY_FETCH_ENABLE | vb->stride); - BEGIN_RING(chan, RING_3D(VERTEX_ARRAY_LIMIT_HIGH(i)), 2); - OUT_RESRCh(chan, res, size - 1, NOUVEAU_BO_RD); - OUT_RESRCl(chan, res, size - 1, NOUVEAU_BO_RD); - BEGIN_RING(chan, RING_3D(VERTEX_ARRAY_START_HIGH(i)), 2); - OUT_RESRCh(chan, res, offset, NOUVEAU_BO_RD); - OUT_RESRCl(chan, res, offset, NOUVEAU_BO_RD); + BEGIN_NV04(push, NV50_3D(VERTEX_ARRAY_FETCH(i)), 1); + PUSH_DATA (push, NV50_3D_VERTEX_ARRAY_FETCH_ENABLE | vb->stride); + BEGIN_NV04(push, NV50_3D(VERTEX_ARRAY_LIMIT_HIGH(i)), 2); + PUSH_DATAh(push, res->address + size - 1); + PUSH_DATA (push, res->address + size - 1); + BEGIN_NV04(push, NV50_3D(VERTEX_ARRAY_START_HIGH(i)), 2); + PUSH_DATAh(push, res->address + offset); + PUSH_DATA (push, res->address + offset); } for (; i < nv50->state.num_vtxelts; ++i) { - BEGIN_RING(chan, RING_3D(VERTEX_ARRAY_ATTRIB(i)), 1); - OUT_RING (chan, NV50_3D_VERTEX_ATTRIB_INACTIVE); + BEGIN_NV04(push, NV50_3D(VERTEX_ARRAY_ATTRIB(i)), 1); + PUSH_DATA (push, NV50_3D_VERTEX_ATTRIB_INACTIVE); if (unlikely(nv50->state.instance_elts & (1 << i))) { - BEGIN_RING(chan, RING_3D(VERTEX_ARRAY_PER_INSTANCE(i)), 1); - OUT_RING (chan, 0); + BEGIN_NV04(push, NV50_3D(VERTEX_ARRAY_PER_INSTANCE(i)), 1); + PUSH_DATA (push, 0); } - BEGIN_RING(chan, RING_3D(VERTEX_ARRAY_FETCH(i)), 1); - OUT_RING (chan, 0); + BEGIN_NV04(push, NV50_3D(VERTEX_ARRAY_FETCH(i)), 1); + PUSH_DATA (push, 0); } nv50->state.num_vtxelts = vertex->num_elements; @@ -390,64 +394,54 @@ nv50_prim_gl(unsigned prim) } static void -nv50_draw_vbo_flush_notify(struct nouveau_channel *chan) -{ - struct nv50_screen *screen = chan->user_private; - - nouveau_fence_update(&screen->base, TRUE); - - nv50_bufctx_emit_relocs(screen->cur_ctx); -} - -static void nv50_draw_arrays(struct nv50_context *nv50, unsigned mode, unsigned start, unsigned count, unsigned instance_count) { - struct nouveau_channel *chan = nv50->screen->base.channel; + struct nouveau_pushbuf *push = nv50->base.pushbuf; unsigned prim; if (nv50->state.index_bias) { - BEGIN_RING(chan, RING_3D(VB_ELEMENT_BASE), 1); - OUT_RING (chan, 0); + BEGIN_NV04(push, NV50_3D(VB_ELEMENT_BASE), 1); + PUSH_DATA (push, 0); nv50->state.index_bias = 0; } prim = nv50_prim_gl(mode); while (instance_count--) { - BEGIN_RING(chan, RING_3D(VERTEX_BEGIN_GL), 1); - OUT_RING (chan, prim); - BEGIN_RING(chan, RING_3D(VERTEX_BUFFER_FIRST), 2); - OUT_RING (chan, start); - OUT_RING (chan, count); - BEGIN_RING(chan, RING_3D(VERTEX_END_GL), 1); - OUT_RING (chan, 0); + BEGIN_NV04(push, NV50_3D(VERTEX_BEGIN_GL), 1); + PUSH_DATA (push, prim); + BEGIN_NV04(push, NV50_3D(VERTEX_BUFFER_FIRST), 2); + PUSH_DATA (push, start); + PUSH_DATA (push, count); + BEGIN_NV04(push, NV50_3D(VERTEX_END_GL), 1); + PUSH_DATA (push, 0); prim |= NV50_3D_VERTEX_BEGIN_GL_INSTANCE_NEXT; } } static void -nv50_draw_elements_inline_u08(struct nouveau_channel *chan, uint8_t *map, +nv50_draw_elements_inline_u08(struct nouveau_pushbuf *push, uint8_t *map, unsigned start, unsigned count) { map += start; if (count & 3) { unsigned i; - BEGIN_RING_NI(chan, RING_3D(VB_ELEMENT_U32), count & 3); + BEGIN_NI04(push, NV50_3D(VB_ELEMENT_U32), count & 3); for (i = 0; i < (count & 3); ++i) - OUT_RING(chan, *map++); + PUSH_DATA(push, *map++); count &= ~3; } while (count) { unsigned i, nr = MIN2(count, NV04_PFIFO_MAX_PACKET_LEN * 4) / 4; - BEGIN_RING_NI(chan, RING_3D(VB_ELEMENT_U8), nr); + BEGIN_NI04(push, NV50_3D(VB_ELEMENT_U8), nr); for (i = 0; i < nr; ++i) { - OUT_RING(chan, - (map[3] << 24) | (map[2] << 16) | (map[1] << 8) | map[0]); + PUSH_DATA(push, + (map[3] << 24) | (map[2] << 16) | (map[1] << 8) | map[0]); map += 4; } count -= nr * 4; @@ -455,22 +449,22 @@ nv50_draw_elements_inline_u08(struct nouveau_channel *chan, uint8_t *map, } static void -nv50_draw_elements_inline_u16(struct nouveau_channel *chan, uint16_t *map, +nv50_draw_elements_inline_u16(struct nouveau_pushbuf *push, uint16_t *map, unsigned start, unsigned count) { map += start; if (count & 1) { count &= ~1; - BEGIN_RING(chan, RING_3D(VB_ELEMENT_U32), 1); - OUT_RING (chan, *map++); + BEGIN_NV04(push, NV50_3D(VB_ELEMENT_U32), 1); + PUSH_DATA (push, *map++); } while (count) { unsigned i, nr = MIN2(count, NV04_PFIFO_MAX_PACKET_LEN * 2) / 2; - BEGIN_RING_NI(chan, RING_3D(VB_ELEMENT_U16), nr); + BEGIN_NI04(push, NV50_3D(VB_ELEMENT_U16), nr); for (i = 0; i < nr; ++i) { - OUT_RING(chan, (map[1] << 16) | map[0]); + PUSH_DATA(push, (map[1] << 16) | map[0]); map += 2; } count -= nr * 2; @@ -478,7 +472,7 @@ nv50_draw_elements_inline_u16(struct nouveau_channel *chan, uint16_t *map, } static void -nv50_draw_elements_inline_u32(struct nouveau_channel *chan, uint32_t *map, +nv50_draw_elements_inline_u32(struct nouveau_pushbuf *push, uint32_t *map, unsigned start, unsigned count) { map += start; @@ -486,8 +480,8 @@ nv50_draw_elements_inline_u32(struct nouveau_channel *chan, uint32_t *map, while (count) { const unsigned nr = MIN2(count, NV04_PFIFO_MAX_PACKET_LEN); - BEGIN_RING_NI(chan, RING_3D(VB_ELEMENT_U32), nr); - OUT_RINGp (chan, map, nr); + BEGIN_NI04(push, NV50_3D(VB_ELEMENT_U32), nr); + PUSH_DATAp(push, map, nr); map += nr; count -= nr; @@ -495,22 +489,22 @@ nv50_draw_elements_inline_u32(struct nouveau_channel *chan, uint32_t *map, } static void -nv50_draw_elements_inline_u32_short(struct nouveau_channel *chan, uint32_t *map, +nv50_draw_elements_inline_u32_short(struct nouveau_pushbuf *push, uint32_t *map, unsigned start, unsigned count) { map += start; if (count & 1) { count--; - BEGIN_RING(chan, RING_3D(VB_ELEMENT_U32), 1); - OUT_RING (chan, *map++); + BEGIN_NV04(push, NV50_3D(VB_ELEMENT_U32), 1); + PUSH_DATA (push, *map++); } while (count) { unsigned i, nr = MIN2(count, NV04_PFIFO_MAX_PACKET_LEN * 2) / 2; - BEGIN_RING_NI(chan, RING_3D(VB_ELEMENT_U16), nr); + BEGIN_NI04(push, NV50_3D(VB_ELEMENT_U16), nr); for (i = 0; i < nr; ++i) { - OUT_RING(chan, (map[1] << 16) | map[0]); + PUSH_DATA(push, (map[1] << 16) | map[0]); map += 2; } count -= nr * 2; @@ -522,121 +516,116 @@ nv50_draw_elements(struct nv50_context *nv50, boolean shorten, unsigned mode, unsigned start, unsigned count, unsigned instance_count, int32_t index_bias) { - struct nouveau_channel *chan = nv50->screen->base.channel; + struct nouveau_pushbuf *push = nv50->base.pushbuf; void *data; + struct nv04_resource *buf = nv04_resource(nv50->idxbuf.buffer); unsigned prim; const unsigned index_size = nv50->idxbuf.index_size; prim = nv50_prim_gl(mode); if (index_bias != nv50->state.index_bias) { - BEGIN_RING(chan, RING_3D(VB_ELEMENT_BASE), 1); - OUT_RING (chan, index_bias); + BEGIN_NV04(push, NV50_3D(VB_ELEMENT_BASE), 1); + PUSH_DATA (push, index_bias); nv50->state.index_bias = index_bias; } if (nouveau_resource_mapped_by_gpu(nv50->idxbuf.buffer)) { - struct nv04_resource *res = nv04_resource(nv50->idxbuf.buffer); + unsigned pb_start; + unsigned pb_bytes; + const unsigned base = buf->offset; start += nv50->idxbuf.offset >> (index_size >> 1); while (instance_count--) { - BEGIN_RING(chan, RING_3D(VERTEX_BEGIN_GL), 1); - OUT_RING (chan, mode); + BEGIN_NV04(push, NV50_3D(VERTEX_BEGIN_GL), 1); + PUSH_DATA (push, prim); + + nouveau_pushbuf_space(push, 8, 0, 1); switch (index_size) { case 4: - { - WAIT_RING (chan, 2); - BEGIN_RING(chan, RING_3D(VB_ELEMENT_U32) | 0x30000, 0); - OUT_RING (chan, count); - nouveau_pushbuf_submit(chan, res->bo, res->offset + start * 4, - count * 4); - } + BEGIN_NL50(push, NV50_3D(VB_ELEMENT_U32), count); + nouveau_pushbuf_data(push, buf->bo, base + start * 4, count * 4); break; case 2: - { - unsigned pb_start = (start & ~1); - unsigned pb_words = (((start + count + 1) & ~1) - pb_start) >> 1; - - BEGIN_RING(chan, RING_3D(VB_ELEMENT_U16_SETUP), 1); - OUT_RING (chan, (start << 31) | count); - WAIT_RING (chan, 2); - BEGIN_RING(chan, RING_3D(VB_ELEMENT_U16) | 0x30000, 0); - OUT_RING (chan, pb_words); - nouveau_pushbuf_submit(chan, res->bo, res->offset + pb_start * 2, - pb_words * 4); - BEGIN_RING(chan, RING_3D(VB_ELEMENT_U16_SETUP), 1); - OUT_RING (chan, 0); - break; - } - case 1: - { - unsigned pb_start = (start & ~3); - unsigned pb_words = (((start + count + 3) & ~3) - pb_start) >> 1; - - BEGIN_RING(chan, RING_3D(VB_ELEMENT_U8_SETUP), 1); - OUT_RING (chan, (start << 30) | count); - WAIT_RING (chan, 2); - BEGIN_RING(chan, RING_3D(VB_ELEMENT_U8) | 0x30000, 0); - OUT_RING (chan, pb_words); - nouveau_pushbuf_submit(chan, res->bo, res->offset + pb_start, - pb_words * 4); - BEGIN_RING(chan, RING_3D(VB_ELEMENT_U8_SETUP), 1); - OUT_RING (chan, 0); + pb_start = (start & ~1) * 2; + pb_bytes = ((start + count + 1) & ~1) * 2 - pb_start; + + BEGIN_NV04(push, NV50_3D(VB_ELEMENT_U16_SETUP), 1); + PUSH_DATA (push, (start << 31) | count); + BEGIN_NL50(push, NV50_3D(VB_ELEMENT_U16), pb_bytes / 4); + nouveau_pushbuf_data(push, buf->bo, base + pb_start, pb_bytes); + BEGIN_NV04(push, NV50_3D(VB_ELEMENT_U16_SETUP), 1); + PUSH_DATA (push, 0); break; - } default: - assert(0); - return; + assert(index_size == 1); + pb_start = start & ~3; + pb_bytes = ((start + count + 3) & ~3) - pb_start; + + BEGIN_NV04(push, NV50_3D(VB_ELEMENT_U8_SETUP), 1); + PUSH_DATA (push, (start << 30) | count); + BEGIN_NL50(push, NV50_3D(VB_ELEMENT_U8), pb_bytes / 4); + nouveau_pushbuf_data(push, buf->bo, base + pb_start, pb_bytes); + BEGIN_NV04(push, NV50_3D(VB_ELEMENT_U8_SETUP), 1); + PUSH_DATA (push, 0); + break; } - BEGIN_RING(chan, RING_3D(VERTEX_END_GL), 1); - OUT_RING (chan, 0); + BEGIN_NV04(push, NV50_3D(VERTEX_END_GL), 1); + PUSH_DATA (push, 0); - nv50_resource_fence(res, NOUVEAU_BO_RD); - - mode |= NV50_3D_VERTEX_BEGIN_GL_INSTANCE_NEXT; + prim |= NV50_3D_VERTEX_BEGIN_GL_INSTANCE_NEXT; } } else { - data = nouveau_resource_map_offset(&nv50->base, - nv04_resource(nv50->idxbuf.buffer), + data = nouveau_resource_map_offset(&nv50->base, buf, nv50->idxbuf.offset, NOUVEAU_BO_RD); if (!data) return; while (instance_count--) { - BEGIN_RING(chan, RING_3D(VERTEX_BEGIN_GL), 1); - OUT_RING (chan, prim); + BEGIN_NV04(push, NV50_3D(VERTEX_BEGIN_GL), 1); + PUSH_DATA (push, prim); switch (index_size) { case 1: - nv50_draw_elements_inline_u08(chan, data, start, count); + nv50_draw_elements_inline_u08(push, data, start, count); break; case 2: - nv50_draw_elements_inline_u16(chan, data, start, count); + nv50_draw_elements_inline_u16(push, data, start, count); break; case 4: if (shorten) - nv50_draw_elements_inline_u32_short(chan, data, start, count); + nv50_draw_elements_inline_u32_short(push, data, start, count); else - nv50_draw_elements_inline_u32(chan, data, start, count); + nv50_draw_elements_inline_u32(push, data, start, count); break; default: assert(0); return; } - BEGIN_RING(chan, RING_3D(VERTEX_END_GL), 1); - OUT_RING (chan, 0); + BEGIN_NV04(push, NV50_3D(VERTEX_END_GL), 1); + PUSH_DATA (push, 0); prim |= NV50_3D_VERTEX_BEGIN_GL_INSTANCE_NEXT; } } } +static void +nv50_draw_vbo_kick_notify(struct nouveau_pushbuf *chan) +{ + struct nv50_screen *screen = chan->user_priv; + + nouveau_fence_update(&screen->base, TRUE); + + nv50_bufctx_fence(screen->cur_ctx->bufctx_3d, TRUE); +} + void nv50_draw_vbo(struct pipe_context *pipe, const struct pipe_draw_info *info) { struct nv50_context *nv50 = nv50_context(pipe); - struct nouveau_channel *chan = nv50->screen->base.channel; + struct nouveau_pushbuf *push = nv50->base.pushbuf; /* For picking only a few vertices from a large user buffer, push is better, * if index count is larger and we expect repeated vertices, suggest upload. @@ -656,24 +645,25 @@ nv50_draw_vbo(struct pipe_context *pipe, const struct pipe_draw_info *info) nv50_state_validate(nv50, ~0, 8); /* 8 as minimum, we use flush_notify */ - chan->flush_notify = nv50_draw_vbo_flush_notify; + push->kick_notify = nv50_draw_vbo_kick_notify; if (nv50->vbo_fifo) { nv50_push_vbo(nv50, info); - chan->flush_notify = nv50_default_flush_notify; + push->kick_notify = nv50_default_kick_notify; + nouveau_pushbuf_bufctx(push, NULL); return; } if (nv50->state.instance_base != info->start_instance) { nv50->state.instance_base = info->start_instance; /* NOTE: this does not affect the shader input, should it ? */ - BEGIN_RING(chan, RING_3D(VB_INSTANCE_BASE), 1); - OUT_RING (chan, info->start_instance); + BEGIN_NV04(push, NV50_3D(VB_INSTANCE_BASE), 1); + PUSH_DATA (push, info->start_instance); } if (nv50->base.vbo_dirty) { - BEGIN_RING(chan, RING_3D(VERTEX_ARRAY_FLUSH), 1); - OUT_RING (chan, 0); + BEGIN_NV04(push, NV50_3D(VERTEX_ARRAY_FLUSH), 1); + PUSH_DATA (push, 0); nv50->base.vbo_dirty = FALSE; } @@ -688,21 +678,21 @@ nv50_draw_vbo(struct pipe_context *pipe, const struct pipe_draw_info *info) if (info->primitive_restart != nv50->state.prim_restart) { if (info->primitive_restart) { - BEGIN_RING(chan, RING_3D(PRIM_RESTART_ENABLE), 2); - OUT_RING (chan, 1); - OUT_RING (chan, info->restart_index); + BEGIN_NV04(push, NV50_3D(PRIM_RESTART_ENABLE), 2); + PUSH_DATA (push, 1); + PUSH_DATA (push, info->restart_index); if (info->restart_index > 65535) shorten = FALSE; } else { - BEGIN_RING(chan, RING_3D(PRIM_RESTART_ENABLE), 1); - OUT_RING (chan, 0); + BEGIN_NV04(push, NV50_3D(PRIM_RESTART_ENABLE), 1); + PUSH_DATA (push, 0); } nv50->state.prim_restart = info->primitive_restart; } else if (info->primitive_restart) { - BEGIN_RING(chan, RING_3D(PRIM_RESTART_INDEX), 1); - OUT_RING (chan, info->restart_index); + BEGIN_NV04(push, NV50_3D(PRIM_RESTART_INDEX), 1); + PUSH_DATA (push, info->restart_index); if (info->restart_index > 65535) shorten = FALSE; @@ -712,7 +702,9 @@ nv50_draw_vbo(struct pipe_context *pipe, const struct pipe_draw_info *info) info->mode, info->start, info->count, info->instance_count, info->index_bias); } - chan->flush_notify = nv50_default_flush_notify; + push->kick_notify = nv50_default_kick_notify; nv50_release_user_vbufs(nv50); + + nouveau_pushbuf_bufctx(push, NULL); } diff --git a/src/gallium/drivers/nv50/nv50_winsys.h b/src/gallium/drivers/nv50/nv50_winsys.h index 694c78a2fbd..b36898dabe6 100644 --- a/src/gallium/drivers/nv50/nv50_winsys.h +++ b/src/gallium/drivers/nv50/nv50_winsys.h @@ -7,100 +7,118 @@ #include "pipe/p_defines.h" -#include "nouveau/nouveau_bo.h" -#include "nouveau/nouveau_channel.h" -#include "nouveau/nouveau_grobj.h" -#include "nouveau/nouveau_device.h" -#include "nouveau/nouveau_resource.h" -#include "nouveau/nouveau_pushbuf.h" -#include "nouveau/nouveau_reloc.h" -#include "nouveau/nouveau_notifier.h" - +#include "nouveau/nouveau_winsys.h" #include "nouveau/nouveau_buffer.h" + #ifndef NV04_PFIFO_MAX_PACKET_LEN #define NV04_PFIFO_MAX_PACKET_LEN 2047 #endif -#define NV50_SUBCH_3D 5 -#define NV50_SUBCH_2D 6 -#define NV50_SUBCH_MF 7 -#define NV50_MF_(n) NV50_M2MF_##n +static INLINE void +nv50_add_bufctx_resident_bo(struct nouveau_bufctx *bufctx, int bin, + unsigned flags, struct nouveau_bo *bo) +{ + nouveau_bufctx_refn(bufctx, bin, bo, flags)->priv = NULL; +} + +static INLINE void +nv50_add_bufctx_resident(struct nouveau_bufctx *bufctx, int bin, + struct nv04_resource *res, unsigned flags) +{ + struct nouveau_bufref *ref = + nouveau_bufctx_refn(bufctx, bin, res->bo, flags | res->domain); + ref->priv = res; + ref->priv_data = flags; +} + +#define BCTX_REFN_bo(ctx, bin, fl, bo) \ + nv50_add_bufctx_resident_bo(ctx, NV50_BIND_##bin, fl, bo); + +#define BCTX_REFN(bctx, bin, res, acc) \ + nv50_add_bufctx_resident(bctx, NV50_BIND_##bin, res, NOUVEAU_BO_##acc) + +static INLINE void +PUSH_REFN(struct nouveau_pushbuf *push, struct nouveau_bo *bo, uint32_t flags) +{ + struct nouveau_pushbuf_refn ref = { bo, flags }; + nouveau_pushbuf_refn(push, &ref, 1); +} + + +#define SUBC_3D(m) 3, (m) +#define NV50_3D(n) SUBC_3D(NV50_3D_##n) -#define RING_3D(n) ((NV50_SUBCH_3D << 13) | NV50_3D_##n) -#define RING_2D(n) ((NV50_SUBCH_2D << 13) | NV50_2D_##n) -#define RING_MF(n) ((NV50_SUBCH_MF << 13) | NV50_MF_(n)) +#define SUBC_2D(m) 4, (m) +#define NV50_2D(n) SUBC_2D(NV50_2D_##n) -#define RING_3D_(m) ((NV50_SUBCH_3D << 13) | (m)) -#define RING_2D_(m) ((NV50_SUBCH_2D << 13) | (m)) -#define RING_MF_(m) ((NV50_SUBCH_MF << 13) | (m)) +#define SUBC_M2MF(m) 5, (m) +#define NV50_M2MF(n) SUBC_M2MF(NV50_M2MF_##n) -#define RING_GR(gr, m) (((gr)->subc << 13) | (m)) +#define SUBC_COMPUTE(m) 6, (m) +#define NV50_COMPUTE(n) SUBC_COMPUTE(NV50_COMPUTE_##n) -int nouveau_pushbuf_flush(struct nouveau_channel *, unsigned min); -static inline uint32_t -nouveau_bo_tile_layout(const struct nouveau_bo *bo) +static INLINE uint32_t +NV50_FIFO_PKHDR(int subc, int mthd, unsigned size) { - return bo->tile_flags & NOUVEAU_BO_TILE_LAYOUT_MASK; + return 0x00000000 | (size << 18) | (subc << 13) | mthd; } -static INLINE void -nouveau_bo_validate(struct nouveau_channel *chan, - struct nouveau_bo *bo, unsigned flags) +static INLINE uint32_t +NV50_FIFO_PKHDR_NI(int subc, int mthd, unsigned size) { - nouveau_reloc_emit(chan, NULL, 0, NULL, bo, 0, 0, flags, 0, 0); + return 0x40000000 | (size << 18) | (subc << 13) | mthd; } -/* incremental methods */ -static INLINE void -BEGIN_RING(struct nouveau_channel *chan, uint32_t mthd, unsigned size) +static INLINE uint32_t +NV50_FIFO_PKHDR_L(int subc, int mthd) { - WAIT_RING(chan, size + 1); - OUT_RING (chan, (size << 18) | mthd); + return 0x00030000 | (subc << 13) | mthd; } -/* non-incremental */ -static INLINE void -BEGIN_RING_NI(struct nouveau_channel *chan, uint32_t mthd, unsigned size) + +static INLINE uint32_t +nouveau_bo_memtype(const struct nouveau_bo *bo) { - WAIT_RING(chan, size + 1); - OUT_RING (chan, (0x2 << 29) | (size << 18) | mthd); + return bo->config.nv50.memtype; } -static INLINE int -OUT_RESRCh(struct nouveau_channel *chan, struct nv04_resource *res, - unsigned delta, unsigned flags) + +static INLINE void +PUSH_DATAh(struct nouveau_pushbuf *push, uint64_t data) { - return OUT_RELOCh(chan, res->bo, res->offset + delta, res->domain | flags); + *push->cur++ = (uint32_t)(data >> 32); } -static INLINE int -OUT_RESRCl(struct nouveau_channel *chan, struct nv04_resource *res, - unsigned delta, unsigned flags) +static INLINE void +BEGIN_NV04(struct nouveau_pushbuf *push, int subc, int mthd, unsigned size) { - if (flags & NOUVEAU_BO_WR) - res->status |= NOUVEAU_BUFFER_STATUS_GPU_WRITING; - return OUT_RELOCl(chan, res->bo, res->offset + delta, res->domain | flags); +#ifndef NV50_PUSH_EXPLICIT_SPACE_CHECKING + PUSH_SPACE(push, size + 1); +#endif + PUSH_DATA (push, NV50_FIFO_PKHDR(subc, mthd, size)); } static INLINE void -BIND_RING(struct nouveau_channel *chan, struct nouveau_grobj *gr, unsigned s) +BEGIN_NI04(struct nouveau_pushbuf *push, int subc, int mthd, unsigned size) { - struct nouveau_subchannel *subc = &gr->channel->subc[s]; - - assert(s < 8); - if (subc->gr) { - assert(subc->gr->bound != NOUVEAU_GROBJ_BOUND_EXPLICIT); - subc->gr->bound = NOUVEAU_GROBJ_UNBOUND; - } - subc->gr = gr; - subc->gr->subc = s; - subc->gr->bound = NOUVEAU_GROBJ_BOUND_EXPLICIT; - - BEGIN_RING(chan, RING_GR(gr, 0x0000), 1); - OUT_RING (chan, gr->handle); +#ifndef NV50_PUSH_EXPLICIT_SPACE_CHECKING + PUSH_SPACE(push, size + 1); +#endif + PUSH_DATA (push, NV50_FIFO_PKHDR_NI(subc, mthd, size)); } +/* long, non-incremental, nv50-only */ +static INLINE void +BEGIN_NL50(struct nouveau_pushbuf *push, int subc, int mthd, uint32_t size) +{ +#ifndef NV50_PUSH_EXPLICIT_SPACE_CHECKING + PUSH_SPACE(push, 2); #endif + PUSH_DATA (push, NV50_FIFO_PKHDR_L(subc, mthd)); + PUSH_DATA (push, size); +} + +#endif /* __NV50_WINSYS_H__ */ |