From fba2eabe13b8a3f8c1396c5949db3daab0192156 Mon Sep 17 00:00:00 2001 From: Christoph Bumiller Date: Sat, 10 Oct 2009 13:13:16 +0200 Subject: nv50: use SIFC for TIC, TSC upload Add proper flushes for TIC and TSC and remove the costly 2D.0110 flush in nv50_flush. Correct TIC and TSC bo sizes. --- src/gallium/drivers/nv50/nv50_vbo.c | 4 ---- 1 file changed, 4 deletions(-) (limited to 'src/gallium/drivers/nv50/nv50_vbo.c') diff --git a/src/gallium/drivers/nv50/nv50_vbo.c b/src/gallium/drivers/nv50/nv50_vbo.c index eeed148c7b..8b0fbf0e76 100644 --- a/src/gallium/drivers/nv50/nv50_vbo.c +++ b/src/gallium/drivers/nv50/nv50_vbo.c @@ -139,10 +139,6 @@ nv50_draw_arrays(struct pipe_context *pipe, unsigned mode, unsigned start, OUT_RING (chan, 0); BEGIN_RING(chan, tesla, 0x142c, 1); OUT_RING (chan, 0); - BEGIN_RING(chan, tesla, 0x1440, 1); - OUT_RING (chan, 0); - BEGIN_RING(chan, tesla, 0x1334, 1); - OUT_RING (chan, 0); BEGIN_RING(chan, tesla, NV50TCL_VERTEX_BEGIN, 1); OUT_RING (chan, nv50_prim(mode)); -- cgit v1.2.3 From c0e80cf0e97cec526bb2ff0f94d9142e33374c20 Mon Sep 17 00:00:00 2001 From: Christoph Bumiller Date: Wed, 14 Oct 2009 21:23:29 +0200 Subject: nv50: submit user vbo data through the fifo Requesting a new real buffer from the kernel and copying all the data is wasteful e.g. if only a few (but widely spread) vertices are accessed. --- src/gallium/drivers/nv50/nv50_context.h | 3 + src/gallium/drivers/nv50/nv50_vbo.c | 409 ++++++++++++++++++++++++++++++-- 2 files changed, 394 insertions(+), 18 deletions(-) (limited to 'src/gallium/drivers/nv50/nv50_vbo.c') diff --git a/src/gallium/drivers/nv50/nv50_context.h b/src/gallium/drivers/nv50/nv50_context.h index 75cb65d9a2..33667e8765 100644 --- a/src/gallium/drivers/nv50/nv50_context.h +++ b/src/gallium/drivers/nv50/nv50_context.h @@ -121,6 +121,7 @@ struct nv50_state { struct nouveau_stateobj *vtxfmt; struct nouveau_stateobj *vtxbuf; struct nouveau_stateobj *vtxattr; + unsigned vtxelt_nr; }; struct nv50_context { @@ -153,6 +154,8 @@ struct nv50_context { unsigned sampler_nr; struct nv50_miptree *miptree[PIPE_MAX_SAMPLERS]; unsigned miptree_nr; + + uint16_t vbo_fifo; }; static INLINE struct nv50_context * diff --git a/src/gallium/drivers/nv50/nv50_vbo.c b/src/gallium/drivers/nv50/nv50_vbo.c index 8b0fbf0e76..db54380241 100644 --- a/src/gallium/drivers/nv50/nv50_vbo.c +++ b/src/gallium/drivers/nv50/nv50_vbo.c @@ -26,6 +26,18 @@ #include "nv50_context.h" +static boolean +nv50_push_elements_u08(struct nv50_context *, uint8_t *, unsigned); + +static boolean +nv50_push_elements_u16(struct nv50_context *, uint16_t *, unsigned); + +static boolean +nv50_push_elements_u32(struct nv50_context *, uint32_t *, unsigned); + +static boolean +nv50_push_arrays(struct nv50_context *, unsigned, unsigned); + static INLINE unsigned nv50_prim(unsigned mode) { @@ -132,6 +144,7 @@ nv50_draw_arrays(struct pipe_context *pipe, unsigned mode, unsigned start, struct nv50_context *nv50 = nv50_context(pipe); struct nouveau_channel *chan = nv50->screen->tesla->channel; struct nouveau_grobj *tesla = nv50->screen->tesla; + boolean ret; nv50_state_validate(nv50); @@ -142,17 +155,22 @@ nv50_draw_arrays(struct pipe_context *pipe, unsigned mode, unsigned start, BEGIN_RING(chan, tesla, NV50TCL_VERTEX_BEGIN, 1); OUT_RING (chan, nv50_prim(mode)); - BEGIN_RING(chan, tesla, NV50TCL_VERTEX_BUFFER_FIRST, 2); - OUT_RING (chan, start); - OUT_RING (chan, count); + + if (nv50->vbo_fifo) + ret = nv50_push_arrays(nv50, start, count); + else { + BEGIN_RING(chan, tesla, NV50TCL_VERTEX_BUFFER_FIRST, 2); + OUT_RING (chan, start); + OUT_RING (chan, count); + ret = TRUE; + } BEGIN_RING(chan, tesla, NV50TCL_VERTEX_END, 1); OUT_RING (chan, 0); - pipe->flush(pipe, 0, NULL); - return TRUE; + return ret; } -static INLINE void +static INLINE boolean nv50_draw_elements_inline_u08(struct nv50_context *nv50, uint8_t *map, unsigned start, unsigned count) { @@ -161,6 +179,9 @@ nv50_draw_elements_inline_u08(struct nv50_context *nv50, uint8_t *map, map += start; + if (nv50->vbo_fifo) + return nv50_push_elements_u08(nv50, map, count); + if (count & 1) { BEGIN_RING(chan, tesla, 0x15e8, 1); OUT_RING (chan, map[0]); @@ -179,9 +200,10 @@ nv50_draw_elements_inline_u08(struct nv50_context *nv50, uint8_t *map, count -= nr; map += nr; } + return TRUE; } -static INLINE void +static INLINE boolean nv50_draw_elements_inline_u16(struct nv50_context *nv50, uint16_t *map, unsigned start, unsigned count) { @@ -190,6 +212,9 @@ nv50_draw_elements_inline_u16(struct nv50_context *nv50, uint16_t *map, map += start; + if (nv50->vbo_fifo) + return nv50_push_elements_u16(nv50, map, count); + if (count & 1) { BEGIN_RING(chan, tesla, 0x15e8, 1); OUT_RING (chan, map[0]); @@ -208,9 +233,10 @@ nv50_draw_elements_inline_u16(struct nv50_context *nv50, uint16_t *map, count -= nr; map += nr; } + return TRUE; } -static INLINE void +static INLINE boolean nv50_draw_elements_inline_u32(struct nv50_context *nv50, uint32_t *map, unsigned start, unsigned count) { @@ -219,6 +245,9 @@ nv50_draw_elements_inline_u32(struct nv50_context *nv50, uint32_t *map, map += start; + if (nv50->vbo_fifo) + return nv50_push_elements_u32(nv50, map, count); + while (count) { unsigned nr = count > 2047 ? 2047 : count; @@ -228,6 +257,7 @@ nv50_draw_elements_inline_u32(struct nv50_context *nv50, uint32_t *map, count -= nr; map += nr; } + return TRUE; } boolean @@ -240,6 +270,7 @@ nv50_draw_elements(struct pipe_context *pipe, struct nouveau_grobj *tesla = nv50->screen->tesla; struct pipe_screen *pscreen = pipe->screen; void *map; + boolean ret; map = pipe_buffer_map(pscreen, indexBuffer, PIPE_BUFFER_USAGE_CPU_READ); @@ -254,23 +285,25 @@ nv50_draw_elements(struct pipe_context *pipe, OUT_RING (chan, nv50_prim(mode)); switch (indexSize) { case 1: - nv50_draw_elements_inline_u08(nv50, map, start, count); + ret = nv50_draw_elements_inline_u08(nv50, map, start, count); break; case 2: - nv50_draw_elements_inline_u16(nv50, map, start, count); + ret = nv50_draw_elements_inline_u16(nv50, map, start, count); break; case 4: - nv50_draw_elements_inline_u32(nv50, map, start, count); + ret = nv50_draw_elements_inline_u32(nv50, map, start, count); break; default: assert(0); + ret = FALSE; + break; } BEGIN_RING(chan, tesla, NV50TCL_VERTEX_END, 1); OUT_RING (chan, 0); pipe_buffer_unmap(pscreen, indexBuffer); - pipe->flush(pipe, 0, NULL); - return TRUE; + + return ret; } static INLINE boolean @@ -337,17 +370,24 @@ nv50_vbo_validate(struct nv50_context *nv50) { struct nouveau_grobj *tesla = nv50->screen->tesla; struct nouveau_stateobj *vtxbuf, *vtxfmt, *vtxattr; - unsigned i; + unsigned i, n_ve; /* don't validate if Gallium took away our buffers */ if (nv50->vtxbuf_nr == 0) return; + nv50->vbo_fifo = 0; + + for (i = 0; i < nv50->vtxbuf_nr; ++i) + if (nv50->vtxbuf[i].stride && + !(nv50->vtxbuf[i].buffer->usage & PIPE_BUFFER_USAGE_VERTEX)) + nv50->vbo_fifo = 0xffff; + + n_ve = MAX2(nv50->vtxelt_nr, nv50->state.vtxelt_nr); vtxattr = NULL; - vtxbuf = so_new(nv50->vtxelt_nr * 7, nv50->vtxelt_nr * 4); - vtxfmt = so_new(nv50->vtxelt_nr + 1, 0); - so_method(vtxfmt, tesla, NV50TCL_VERTEX_ARRAY_ATTRIB(0), - nv50->vtxelt_nr); + vtxbuf = so_new(n_ve * 7, nv50->vtxelt_nr * 4); + vtxfmt = so_new(n_ve + 1, 0); + so_method(vtxfmt, tesla, NV50TCL_VERTEX_ARRAY_ATTRIB(0), n_ve); for (i = 0; i < nv50->vtxelt_nr; i++) { struct pipe_vertex_element *ve = &nv50->vtxelt[i]; @@ -363,10 +403,19 @@ nv50_vbo_validate(struct nv50_context *nv50) so_method(vtxbuf, tesla, NV50TCL_VERTEX_ARRAY_FORMAT(i), 1); so_data (vtxbuf, 0); + + nv50->vbo_fifo &= ~(1 << i); continue; } so_data(vtxfmt, hw | i); + if (nv50->vbo_fifo) { + so_method(vtxbuf, tesla, + NV50TCL_VERTEX_ARRAY_FORMAT(i), 1); + so_data (vtxbuf, 0); + continue; + } + so_method(vtxbuf, tesla, NV50TCL_VERTEX_ARRAY_FORMAT(i), 3); so_data (vtxbuf, 0x20000000 | vb->stride); so_reloc (vtxbuf, bo, vb->buffer_offset + @@ -385,6 +434,13 @@ nv50_vbo_validate(struct nv50_context *nv50) NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0); } + for (; i < n_ve; ++i) { + so_data (vtxfmt, 0x7e080010); + + so_method(vtxbuf, tesla, NV50TCL_VERTEX_ARRAY_FORMAT(i), 1); + so_data (vtxbuf, 0); + } + nv50->state.vtxelt_nr = nv50->vtxelt_nr; so_ref (vtxfmt, &nv50->state.vtxfmt); so_ref (vtxbuf, &nv50->state.vtxbuf); @@ -394,3 +450,320 @@ nv50_vbo_validate(struct nv50_context *nv50) so_ref (NULL, &vtxattr); } +typedef void (*pfn_push)(struct nouveau_channel *, void *); + +struct nv50_vbo_emitctx +{ + pfn_push push[16]; + void *map[16]; + unsigned stride[16]; + unsigned nr_ve; + unsigned vtx_dwords; + unsigned vtx_max; +}; + +static INLINE void +emit_vtx_next(struct nouveau_channel *chan, struct nv50_vbo_emitctx *emit) +{ + unsigned i; + + for (i = 0; i < emit->nr_ve; ++i) { + emit->push[i](chan, emit->map[i]); + emit->map[i] += emit->stride[i]; + } +} + +static INLINE void +emit_vtx(struct nouveau_channel *chan, struct nv50_vbo_emitctx *emit, + uint32_t vi) +{ + unsigned i; + + for (i = 0; i < emit->nr_ve; ++i) + emit->push[i](chan, emit->map[i] + emit->stride[i] * vi); +} + +static INLINE boolean +nv50_map_vbufs(struct nv50_context *nv50) +{ + int i; + + for (i = 0; i < nv50->vtxbuf_nr; ++i) { + struct pipe_vertex_buffer *vb = &nv50->vtxbuf[i]; + unsigned size, delta; + + if (nouveau_bo(vb->buffer)->map) + continue; + + size = vb->stride * (vb->max_index + 1); + delta = vb->buffer_offset; + + if (!size) + size = vb->buffer->size - vb->buffer_offset; + + if (nouveau_bo_map_range(nouveau_bo(vb->buffer), + delta, size, NOUVEAU_BO_RD)) + break; + } + + if (i == nv50->vtxbuf_nr) + return TRUE; + for (; i >= 0; --i) + nouveau_bo_unmap(nouveau_bo(nv50->vtxbuf[i].buffer)); + return FALSE; +} + +static INLINE void +nv50_unmap_vbufs(struct nv50_context *nv50) +{ + unsigned i; + + for (i = 0; i < nv50->vtxbuf_nr; ++i) + if (nouveau_bo(nv50->vtxbuf[i].buffer)->map) + nouveau_bo_unmap(nouveau_bo(nv50->vtxbuf[i].buffer)); +} + +static void +emit_b32_1(struct nouveau_channel *chan, void *data) +{ + uint32_t *v = data; + + OUT_RING(chan, v[0]); +} + +static void +emit_b32_2(struct nouveau_channel *chan, void *data) +{ + uint32_t *v = data; + + OUT_RING(chan, v[0]); + OUT_RING(chan, v[1]); +} + +static void +emit_b32_3(struct nouveau_channel *chan, void *data) +{ + uint32_t *v = data; + + OUT_RING(chan, v[0]); + OUT_RING(chan, v[1]); + OUT_RING(chan, v[2]); +} + +static void +emit_b32_4(struct nouveau_channel *chan, void *data) +{ + uint32_t *v = data; + + OUT_RING(chan, v[0]); + OUT_RING(chan, v[1]); + OUT_RING(chan, v[2]); + OUT_RING(chan, v[3]); +} + +static void +emit_b16_1(struct nouveau_channel *chan, void *data) +{ + uint16_t *v = data; + + OUT_RING(chan, v[0]); +} + +static void +emit_b16_3(struct nouveau_channel *chan, void *data) +{ + uint16_t *v = data; + + OUT_RING(chan, (v[1] << 16) | v[0]); + OUT_RING(chan, v[2]); +} + +static void +emit_b08_1(struct nouveau_channel *chan, void *data) +{ + uint8_t *v = data; + + OUT_RING(chan, v[0]); +} + +static void +emit_b08_3(struct nouveau_channel *chan, void *data) +{ + uint8_t *v = data; + + OUT_RING(chan, (v[2] << 16) | (v[1] << 8) | v[0]); +} + +static boolean +emit_prepare(struct nv50_context *nv50, struct nv50_vbo_emitctx *emit, + unsigned start) +{ + unsigned i; + + if (nv50_map_vbufs(nv50) == FALSE) + return FALSE; + + emit->nr_ve = 0; + emit->vtx_dwords = 0; + + for (i = 0; i < nv50->vtxelt_nr; ++i) { + struct pipe_vertex_element *ve; + struct pipe_vertex_buffer *vb; + unsigned n, type, size; + + ve = &nv50->vtxelt[i]; + vb = &nv50->vtxbuf[ve->vertex_buffer_index]; + if (!(nv50->vbo_fifo & (1 << i))) + continue; + n = emit->nr_ve++; + + emit->stride[n] = vb->stride; + emit->map[n] = nouveau_bo(vb->buffer)->map + + (start * vb->stride + ve->src_offset); + + type = pf_type(ve->src_format); + size = pf_size_x(ve->src_format) << pf_exp2(ve->src_format); + + assert(ve->nr_components > 0 && ve->nr_components <= 4); + + /* It shouldn't be necessary to push the implicit 1s + * for case 3 and size 8 cases 1, 2, 3. + */ + switch (size) { + default: + NOUVEAU_ERR("unsupported vtxelt size: %u\n", size); + return FALSE; + case 32: + switch (ve->nr_components) { + case 1: emit->push[n] = emit_b32_1; break; + case 2: emit->push[n] = emit_b32_2; break; + case 3: emit->push[n] = emit_b32_3; break; + case 4: emit->push[n] = emit_b32_4; break; + } + emit->vtx_dwords += ve->nr_components; + break; + case 16: + switch (ve->nr_components) { + case 1: emit->push[n] = emit_b16_1; break; + case 2: emit->push[n] = emit_b32_1; break; + case 3: emit->push[n] = emit_b16_3; break; + case 4: emit->push[n] = emit_b32_2; break; + } + emit->vtx_dwords += (ve->nr_components + 1) >> 1; + break; + case 8: + switch (ve->nr_components) { + case 1: emit->push[n] = emit_b08_1; break; + case 2: emit->push[n] = emit_b16_1; break; + case 3: emit->push[n] = emit_b08_3; break; + case 4: emit->push[n] = emit_b32_1; break; + } + emit->vtx_dwords += 1; + break; + } + } + + emit->vtx_max = 512 / emit->vtx_dwords; + + return TRUE; +} + +static boolean +nv50_push_arrays(struct nv50_context *nv50, unsigned start, unsigned count) +{ + struct nouveau_channel *chan = nv50->screen->base.channel; + struct nouveau_grobj *tesla = nv50->screen->tesla; + struct nv50_vbo_emitctx emit; + + if (emit_prepare(nv50, &emit, start) == FALSE) + return FALSE; + + while (count) { + unsigned i, dw, nr = MIN2(count, emit.vtx_max); + dw = nr * emit.vtx_dwords; + + BEGIN_RING(chan, tesla, NV50TCL_VERTEX_DATA | 0x40000000, dw); + for (i = 0; i < nr; ++i) + emit_vtx_next(chan, &emit); + + count -= nr; + } + nv50_unmap_vbufs(nv50); + + return TRUE; +} + +static boolean +nv50_push_elements_u32(struct nv50_context *nv50, uint32_t *map, unsigned count) +{ + struct nouveau_channel *chan = nv50->screen->base.channel; + struct nouveau_grobj *tesla = nv50->screen->tesla; + struct nv50_vbo_emitctx emit; + + if (emit_prepare(nv50, &emit, 0) == FALSE) + return FALSE; + + while (count) { + unsigned i, dw, nr = MIN2(count, emit.vtx_max); + dw = nr * emit.vtx_dwords; + + BEGIN_RING(chan, tesla, NV50TCL_VERTEX_DATA | 0x40000000, dw); + for (i = 0; i < nr; ++i) + emit_vtx(chan, &emit, *map++); + + count -= nr; + } + nv50_unmap_vbufs(nv50); + + return TRUE; +} + +static boolean +nv50_push_elements_u16(struct nv50_context *nv50, uint16_t *map, unsigned count) +{ + struct nouveau_channel *chan = nv50->screen->base.channel; + struct nouveau_grobj *tesla = nv50->screen->tesla; + struct nv50_vbo_emitctx emit; + + if (emit_prepare(nv50, &emit, 0) == FALSE) + return FALSE; + + while (count) { + unsigned i, dw, nr = MIN2(count, emit.vtx_max); + dw = nr * emit.vtx_dwords; + + BEGIN_RING(chan, tesla, NV50TCL_VERTEX_DATA | 0x40000000, dw); + for (i = 0; i < nr; ++i) + emit_vtx(chan, &emit, *map++); + + count -= nr; + } + nv50_unmap_vbufs(nv50); + + return TRUE; +} + +static boolean +nv50_push_elements_u08(struct nv50_context *nv50, uint8_t *map, unsigned count) +{ + struct nouveau_channel *chan = nv50->screen->base.channel; + struct nouveau_grobj *tesla = nv50->screen->tesla; + struct nv50_vbo_emitctx emit; + + if (emit_prepare(nv50, &emit, 0) == FALSE) + return FALSE; + + while (count) { + unsigned i, dw, nr = MIN2(count, emit.vtx_max); + dw = nr * emit.vtx_dwords; + + BEGIN_RING(chan, tesla, NV50TCL_VERTEX_DATA | 0x40000000, dw); + for (i = 0; i < nr; ++i) + emit_vtx(chan, &emit, *map++); + + count -= nr; + } + nv50_unmap_vbufs(nv50); + + return TRUE; +} -- cgit v1.2.3 From 72befaaae5fd9555e1f6ccbd6a74c0d640fc2929 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Thu, 3 Dec 2009 12:25:49 +0100 Subject: Remove pf_swizzle_* internal macros. --- src/gallium/drivers/nv50/nv50_vbo.c | 4 +++- src/gallium/drivers/r300/r300_state_inlines.h | 13 +++++++++---- src/gallium/include/pipe/p_format.h | 5 ----- src/mesa/state_tracker/st_format.c | 11 +++++++---- 4 files changed, 19 insertions(+), 14 deletions(-) (limited to 'src/gallium/drivers/nv50/nv50_vbo.c') diff --git a/src/gallium/drivers/nv50/nv50_vbo.c b/src/gallium/drivers/nv50/nv50_vbo.c index db54380241..4b8783899e 100644 --- a/src/gallium/drivers/nv50/nv50_vbo.c +++ b/src/gallium/drivers/nv50/nv50_vbo.c @@ -24,6 +24,8 @@ #include "pipe/p_state.h" #include "pipe/p_inlines.h" +#include "util/u_format.h" + #include "nv50_context.h" static boolean @@ -131,7 +133,7 @@ nv50_vbo_vtxelt_to_hw(struct pipe_vertex_element *ve) return 0x24e80000; } - if (pf_swizzle_x(pf) == 2) /* BGRA */ + if (util_format_description(pf)->swizzle[0] == UTIL_FORMAT_SWIZZLE_Z) /* BGRA */ hw_size |= (1 << 31); /* no real swizzle bits :-( */ return (hw_type | hw_size); diff --git a/src/gallium/drivers/r300/r300_state_inlines.h b/src/gallium/drivers/r300/r300_state_inlines.h index e6c1cb54da..ab720f366d 100644 --- a/src/gallium/drivers/r300/r300_state_inlines.h +++ b/src/gallium/drivers/r300/r300_state_inlines.h @@ -28,6 +28,8 @@ #include "pipe/p_format.h" +#include "util/u_format.h" + #include "r300_reg.h" /* Some maths. These should probably find their way to u_math, if needed. */ @@ -534,6 +536,9 @@ r300_translate_vertex_data_type(enum pipe_format format) { static INLINE uint16_t r300_translate_vertex_data_swizzle(enum pipe_format format) { + const struct util_format_description *desc = util_format_description(format); + + assert(format); if (pf_layout(format) != PIPE_FORMAT_LAYOUT_RGBAZS) { debug_printf("r300: Bad format %s in %s:%d\n", @@ -541,10 +546,10 @@ r300_translate_vertex_data_swizzle(enum pipe_format format) { return 0; } - return ((pf_swizzle_x(format) << R300_SWIZZLE_SELECT_X_SHIFT) | - (pf_swizzle_y(format) << R300_SWIZZLE_SELECT_Y_SHIFT) | - (pf_swizzle_z(format) << R300_SWIZZLE_SELECT_Z_SHIFT) | - (pf_swizzle_w(format) << R300_SWIZZLE_SELECT_W_SHIFT) | + return ((desc->swizzle[0] << R300_SWIZZLE_SELECT_X_SHIFT) | + (desc->swizzle[1] << R300_SWIZZLE_SELECT_Y_SHIFT) | + (desc->swizzle[2] << R300_SWIZZLE_SELECT_Z_SHIFT) | + (desc->swizzle[3] << R300_SWIZZLE_SELECT_W_SHIFT) | (0xf << R300_WRITE_ENA_SHIFT)); } diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h index 52fdab4570..cc7a8ab8df 100644 --- a/src/gallium/include/pipe/p_format.h +++ b/src/gallium/include/pipe/p_format.h @@ -112,11 +112,6 @@ static INLINE uint pf_get(pipe_format_rgbazs_t f, uint shift, uint mask) return (f >> shift) & mask; } -#define pf_swizzle_x(f) pf_get(f, 2, 0x7) /**< PIPE_FORMAT_COMP_ */ -#define pf_swizzle_y(f) pf_get(f, 5, 0x7) /**< PIPE_FORMAT_COMP_ */ -#define pf_swizzle_z(f) pf_get(f, 8, 0x7) /**< PIPE_FORMAT_COMP_ */ -#define pf_swizzle_w(f) pf_get(f, 11, 0x7) /**< PIPE_FORMAT_COMP_ */ -#define pf_swizzle_xyzw(f,i) pf_get(f, 2+((i)*3), 0x7) #define pf_size_x(f) pf_get(f, 14, 0x7) /**< Size of X */ #define pf_size_y(f) pf_get(f, 17, 0x7) /**< Size of Y */ #define pf_size_z(f) pf_get(f, 20, 0x7) /**< Size of Z */ diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index c492d77530..c9f020c40f 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -81,8 +81,11 @@ GLboolean st_get_format_info(enum pipe_format format, struct pipe_format_info *pinfo) { if (pf_layout(format) == PIPE_FORMAT_LAYOUT_RGBAZS) { + const struct util_format_description *desc = util_format_description(format); pipe_format_rgbazs_t info; + assert(desc); + info = format; #if 0 @@ -133,10 +136,10 @@ st_get_format_info(enum pipe_format format, struct pipe_format_info *pinfo) pinfo->size = format_size( info ) / 8; /* Luminance & Intensity bits */ - if( pf_swizzle_x(info) == PIPE_FORMAT_COMP_R && - pf_swizzle_y(info) == PIPE_FORMAT_COMP_R && - pf_swizzle_z(info) == PIPE_FORMAT_COMP_R ) { - if( pf_swizzle_w(info) == PIPE_FORMAT_COMP_R ) { + if (desc->swizzle[0] == UTIL_FORMAT_SWIZZLE_X && + desc->swizzle[1] == UTIL_FORMAT_SWIZZLE_X && + desc->swizzle[2] == UTIL_FORMAT_SWIZZLE_X) { + if (desc->swizzle[3] == UTIL_FORMAT_SWIZZLE_X) { pinfo->intensity_bits = pinfo->red_bits; } else { -- cgit v1.2.3 From 6fd8b9b550713302566bb4c28e49c219870ccfec Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Tue, 8 Dec 2009 15:30:10 +0100 Subject: Remove remaining pipe format utility functions. Depricate pf_type(), pf_size_*(), pf_layout() and pf_exp2(). Map depricated PIPE_FORMAT_TYPE to new UTIL_FORMAT_ values: UNKNOWN = TYPE_VOID UNORM = TYPE_UNSIGNED + LAYOUT_ARITH SNORM = TYPE_SIGNED + LAYOUT_ARITH FIXED = TYPE_FIXED FLOAT = TYPE_FLOAT USCALED = TYPE_UNSIGNED + LAYOUT_ARRAY SSCALED = TYPE_SIGNED + LAYOUT_ARRAY SRGB = TYPE_COLORSPACE_SRGB --- src/gallium/auxiliary/util/u_pack_color.h | 3 +- src/gallium/drivers/nv50/nv50_tex.c | 8 ++- src/gallium/drivers/nv50/nv50_vbo.c | 68 +++++++++++++++++------- src/gallium/drivers/r300/r300_state_inlines.h | 62 ++++++++++----------- src/gallium/include/pipe/p_format.h | 18 ------- src/gallium/state_trackers/wgl/stw_framebuffer.c | 4 +- src/gallium/state_trackers/wgl/stw_pixelformat.c | 2 - src/mesa/state_tracker/st_format.c | 65 +++++++++++----------- 8 files changed, 122 insertions(+), 108 deletions(-) (limited to 'src/gallium/drivers/nv50/nv50_vbo.c') diff --git a/src/gallium/auxiliary/util/u_pack_color.h b/src/gallium/auxiliary/util/u_pack_color.h index 9dacc6d83d..30cc7969e9 100644 --- a/src/gallium/auxiliary/util/u_pack_color.h +++ b/src/gallium/auxiliary/util/u_pack_color.h @@ -37,6 +37,7 @@ #include "pipe/p_compiler.h" #include "pipe/p_format.h" +#include "util/u_format.h" #include "util/u_math.h" @@ -307,7 +308,7 @@ util_pack_color(const float rgba[4], enum pipe_format format, void *dest) ubyte b = 0; ubyte a = 0; - if (pf_size_x(format) <= 8) { + if (util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 0) <= 8) { /* format uses 8-bit components or less */ r = float_to_ubyte(rgba[0]); g = float_to_ubyte(rgba[1]); diff --git a/src/gallium/drivers/nv50/nv50_tex.c b/src/gallium/drivers/nv50/nv50_tex.c index 417d367942..55a601deb8 100644 --- a/src/gallium/drivers/nv50/nv50_tex.c +++ b/src/gallium/drivers/nv50/nv50_tex.c @@ -25,6 +25,8 @@ #include "nouveau/nouveau_stateobj.h" +#include "util/u_format.h" + #define _MIXED(pf, t0, t1, t2, t3, cr, cg, cb, ca, f) \ { \ PIPE_FORMAT_##pf, \ @@ -89,6 +91,7 @@ nv50_tex_construct(struct nv50_context *nv50, struct nouveau_stateobj *so, { unsigned i; uint32_t mode; + const struct util_format_description *desc; for (i = 0; i < NV50_TEX_FORMAT_LIST_SIZE; i++) if (nv50_tex_format_list[i].pf == mt->base.base.format) @@ -106,7 +109,10 @@ nv50_tex_construct(struct nv50_context *nv50, struct nouveau_stateobj *so, mode |= ((mt->base.bo->tile_mode & 0x0f) << 22) | ((mt->base.bo->tile_mode & 0xf0) << 21); - if (pf_type(mt->base.base.format) == PIPE_FORMAT_TYPE_SRGB) + desc = util_format_description(mt->base.base.format); + assert(desc); + + if (desc->colorspace == UTIL_FORMAT_COLORSPACE_SRGB) mode |= 0x0400; switch (mt->base.base.target) { diff --git a/src/gallium/drivers/nv50/nv50_vbo.c b/src/gallium/drivers/nv50/nv50_vbo.c index 4b8783899e..f3472f266e 100644 --- a/src/gallium/drivers/nv50/nv50_vbo.c +++ b/src/gallium/drivers/nv50/nv50_vbo.c @@ -64,19 +64,36 @@ nv50_prim(unsigned mode) } static INLINE uint32_t -nv50_vbo_type_to_hw(unsigned type) +nv50_vbo_type_to_hw(enum pipe_format format) { - switch (type) { - case PIPE_FORMAT_TYPE_FLOAT: + const struct util_format_description *desc; + + desc = util_format_description(format); + assert(desc); + + switch (desc->type) { + case UTIL_FORMAT_TYPE_FLOAT: return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_FLOAT; - case PIPE_FORMAT_TYPE_UNORM: - return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_UNORM; - case PIPE_FORMAT_TYPE_SNORM: - return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_SNORM; - case PIPE_FORMAT_TYPE_USCALED: - return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_USCALED; - case PIPE_FORMAT_TYPE_SSCALED: - return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_SSCALED; + case UTIL_FORMAT_TYPE_UNSIGNED: + switch (desc->layout) { + case UTIL_FORMAT_LAYOUT_ARITH: + return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_UNORM; + case UTIL_FORMAT_LAYOUT_ARRAY: + return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_USCALED; + default: + return 0; + } + break; + case UTIL_FORMAT_TYPE_SIGNED: + switch (desc->layout) { + case UTIL_FORMAT_LAYOUT_ARITH: + return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_SNORM; + case UTIL_FORMAT_LAYOUT_ARRAY: + return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_SSCALED; + default: + return 0; + } + break; /* case PIPE_FORMAT_TYPE_UINT: return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_UINT; @@ -122,9 +139,15 @@ nv50_vbo_vtxelt_to_hw(struct pipe_vertex_element *ve) { uint32_t hw_type, hw_size; enum pipe_format pf = ve->src_format; - unsigned size = pf_size_x(pf) << pf_exp2(pf); + const struct util_format_description *desc; + unsigned size; + + desc = util_format_description(pf); + assert(desc); - hw_type = nv50_vbo_type_to_hw(pf_type(pf)); + size = util_format_get_component_bits(pf, UTIL_FORMAT_COLORSPACE_RGB, 0); + + hw_type = nv50_vbo_type_to_hw(pf); hw_size = nv50_vbo_size_to_hw(size, ve->nr_components); if (!hw_type || !hw_size) { @@ -133,7 +156,7 @@ nv50_vbo_vtxelt_to_hw(struct pipe_vertex_element *ve) return 0x24e80000; } - if (util_format_description(pf)->swizzle[0] == UTIL_FORMAT_SWIZZLE_Z) /* BGRA */ + if (desc->swizzle[0] == UTIL_FORMAT_SWIZZLE_Z) /* BGRA */ hw_size |= (1 << 31); /* no real swizzle bits :-( */ return (hw_type | hw_size); @@ -321,9 +344,13 @@ nv50_vbo_static_attrib(struct nv50_context *nv50, unsigned attrib, float *v; int ret; enum pipe_format pf = ve->src_format; + const struct util_format_description *desc; + + desc = util_format_description(pf); + assert(desc); - if ((pf_type(pf) != PIPE_FORMAT_TYPE_FLOAT) || - (pf_size_x(pf) << pf_exp2(pf)) != 32) + if ((desc->type != UTIL_FORMAT_TYPE_FLOAT) || + util_format_get_component_bits(pf, UTIL_FORMAT_COLORSPACE_RGB, 0) != 32) return FALSE; ret = nouveau_bo_map(bo, NOUVEAU_BO_RD); @@ -611,7 +638,8 @@ emit_prepare(struct nv50_context *nv50, struct nv50_vbo_emitctx *emit, for (i = 0; i < nv50->vtxelt_nr; ++i) { struct pipe_vertex_element *ve; struct pipe_vertex_buffer *vb; - unsigned n, type, size; + unsigned n, size; + const struct util_format_description *desc; ve = &nv50->vtxelt[i]; vb = &nv50->vtxbuf[ve->vertex_buffer_index]; @@ -623,8 +651,10 @@ emit_prepare(struct nv50_context *nv50, struct nv50_vbo_emitctx *emit, emit->map[n] = nouveau_bo(vb->buffer)->map + (start * vb->stride + ve->src_offset); - type = pf_type(ve->src_format); - size = pf_size_x(ve->src_format) << pf_exp2(ve->src_format); + desc = util_format_description(ve->src_format); + assert(desc); + + size = util_format_get_component_bits(ve->src_format, UTIL_FORMAT_COLORSPACE_RGB, 0); assert(ve->nr_components > 0 && ve->nr_components <= 4); diff --git a/src/gallium/drivers/r300/r300_state_inlines.h b/src/gallium/drivers/r300/r300_state_inlines.h index ab720f366d..5c58655da1 100644 --- a/src/gallium/drivers/r300/r300_state_inlines.h +++ b/src/gallium/drivers/r300/r300_state_inlines.h @@ -445,20 +445,22 @@ static INLINE uint32_t r300_translate_gb_pipes(int pipe_count) static INLINE unsigned pf_component_count(enum pipe_format format) { unsigned count = 0; - if (pf_layout(format) != PIPE_FORMAT_LAYOUT_RGBAZS) { - return count; + if (util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 0)) { + count++; } - - if (pf_size_x(format)) { + if (util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 1)) { + count++; + } + if (util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 2)) { count++; } - if (pf_size_y(format)) { + if (util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 3)) { count++; } - if (pf_size_z(format)) { + if (util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_ZS, 0)) { count++; } - if (pf_size_w(format)) { + if (util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_ZS, 1)) { count++; } @@ -469,19 +471,23 @@ static INLINE unsigned pf_component_count(enum pipe_format format) { static INLINE uint16_t r300_translate_vertex_data_type(enum pipe_format format) { uint32_t result = 0; + const struct util_format_description *desc; unsigned components = pf_component_count(format); - if (pf_layout(format) != PIPE_FORMAT_LAYOUT_RGBAZS) { + desc = util_format_description(format); + + if (desc->layout != UTIL_FORMAT_LAYOUT_ARITH && + desc->layout != UTIL_FORMAT_LAYOUT_ARRAY) { debug_printf("r300: Bad format %s in %s:%d\n", pf_name(format), __FUNCTION__, __LINE__); assert(0); } - switch (pf_type(format)) { + switch (desc->type) { /* Half-floats, floats, doubles */ - case PIPE_FORMAT_TYPE_FLOAT: - switch (pf_size_x(format)) { - case 4: + case UTIL_FORMAT_TYPE_FLOAT: + switch (util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 0)) { + case 32: result = R300_DATA_TYPE_FLOAT_1 + (components - 1); break; default: @@ -490,19 +496,15 @@ r300_translate_vertex_data_type(enum pipe_format format) { assert(0); } break; - /* Normalized unsigned ints */ - case PIPE_FORMAT_TYPE_UNORM: - /* Normalized signed ints */ - case PIPE_FORMAT_TYPE_SNORM: - /* Non-normalized unsigned ints */ - case PIPE_FORMAT_TYPE_USCALED: - /* Non-normalized signed ints */ - case PIPE_FORMAT_TYPE_SSCALED: - switch (pf_size_x(format)) { - case 1: + /* Unsigned ints */ + case UTIL_FORMAT_TYPE_UNSIGNED: + /* Signed ints */ + case UTIL_FORMAT_TYPE_SIGNED: + switch (util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 0)) { + case 8: result = R300_DATA_TYPE_BYTE; break; - case 2: + case 16: if (components > 2) { result = R300_DATA_TYPE_SHORT_4; } else { @@ -512,8 +514,8 @@ r300_translate_vertex_data_type(enum pipe_format format) { default: debug_printf("r300: Bad format %s in %s:%d\n", pf_name(format), __FUNCTION__, __LINE__); - debug_printf("r300: pf_size_x(format) == %d\n", - pf_size_x(format)); + debug_printf("r300: util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 0) == %d\n", + util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 0)); assert(0); } break; @@ -523,12 +525,11 @@ r300_translate_vertex_data_type(enum pipe_format format) { assert(0); } - if (pf_type(format) == PIPE_FORMAT_TYPE_SSCALED) { + if (desc->type == UTIL_FORMAT_TYPE_SIGNED) { result |= R300_SIGNED; - } else if (pf_type(format) == PIPE_FORMAT_TYPE_UNORM) { + } + if (desc->layout == UTIL_FORMAT_LAYOUT_ARITH) { result |= R300_NORMALIZE; - } else if (pf_type(format) == PIPE_FORMAT_TYPE_SNORM) { - result |= (R300_SIGNED | R300_NORMALIZE); } return result; @@ -540,7 +541,8 @@ r300_translate_vertex_data_swizzle(enum pipe_format format) { assert(format); - if (pf_layout(format) != PIPE_FORMAT_LAYOUT_RGBAZS) { + if (desc->layout != UTIL_FORMAT_LAYOUT_ARITH && + desc->layout != UTIL_FORMAT_LAYOUT_ARRAY) { debug_printf("r300: Bad format %s in %s:%d\n", pf_name(format), __FUNCTION__, __LINE__); return 0; diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h index cc7a8ab8df..77134f664b 100644 --- a/src/gallium/include/pipe/p_format.h +++ b/src/gallium/include/pipe/p_format.h @@ -54,11 +54,6 @@ extern "C" { #define PIPE_FORMAT_LAYOUT_DXT 2 /**< XXX temporary? */ #define PIPE_FORMAT_LAYOUT_MIXED 3 -static INLINE uint pf_layout(uint f) /**< PIPE_FORMAT_LAYOUT_ */ -{ - return f & 0x3; -} - /** * RGBAZS Format Layout. */ @@ -107,19 +102,6 @@ static INLINE uint pf_layout(uint f) /**< PIPE_FORMAT_LAYOUT_ */ */ typedef uint pipe_format_rgbazs_t; -static INLINE uint pf_get(pipe_format_rgbazs_t f, uint shift, uint mask) -{ - return (f >> shift) & mask; -} - -#define pf_size_x(f) pf_get(f, 14, 0x7) /**< Size of X */ -#define pf_size_y(f) pf_get(f, 17, 0x7) /**< Size of Y */ -#define pf_size_z(f) pf_get(f, 20, 0x7) /**< Size of Z */ -#define pf_size_w(f) pf_get(f, 23, 0x7) /**< Size of W */ -#define pf_size_xyzw(f,i) pf_get(f, 14+((i)*3), 0x7) -#define pf_exp2(f) pf_get(f, 26, 0x7) /**< Scale size by 2 ^ exp2 */ -#define pf_type(f) pf_get(f, 29, 0x7) /**< PIPE_FORMAT_TYPE_ */ - /** * Helper macro to encode the above structure into a 32-bit value. */ diff --git a/src/gallium/state_trackers/wgl/stw_framebuffer.c b/src/gallium/state_trackers/wgl/stw_framebuffer.c index 5c3444777a..3d029f9174 100644 --- a/src/gallium/state_trackers/wgl/stw_framebuffer.c +++ b/src/gallium/state_trackers/wgl/stw_framebuffer.c @@ -268,9 +268,7 @@ stw_framebuffer_allocate( enum pipe_format colorFormat, depthFormat, stencilFormat; colorFormat = pfi->color_format; - - assert(pf_layout( pfi->depth_stencil_format ) == PIPE_FORMAT_LAYOUT_RGBAZS ); - + if(util_format_get_component_bits(pfi->depth_stencil_format, UTIL_FORMAT_COLORSPACE_ZS, 0)) depthFormat = pfi->depth_stencil_format; else diff --git a/src/gallium/state_trackers/wgl/stw_pixelformat.c b/src/gallium/state_trackers/wgl/stw_pixelformat.c index 5ac833ced6..54cc361412 100644 --- a/src/gallium/state_trackers/wgl/stw_pixelformat.c +++ b/src/gallium/state_trackers/wgl/stw_pixelformat.c @@ -133,12 +133,10 @@ stw_pixelformat_add( if(stw_dev->pixelformat_extended_count >= STW_MAX_PIXELFORMATS) return; - assert(pf_layout( color->format ) == PIPE_FORMAT_LAYOUT_RGBAZS ); assert(util_format_get_component_bits(color->format, UTIL_FORMAT_COLORSPACE_RGB, 0) == color->bits.red); assert(util_format_get_component_bits(color->format, UTIL_FORMAT_COLORSPACE_RGB, 1) == color->bits.green); assert(util_format_get_component_bits(color->format, UTIL_FORMAT_COLORSPACE_RGB, 2) == color->bits.blue); assert(util_format_get_component_bits(color->format, UTIL_FORMAT_COLORSPACE_RGB, 3) == color->bits.alpha); - assert(pf_layout( depth->format ) == PIPE_FORMAT_LAYOUT_RGBAZS ); assert(util_format_get_component_bits(depth->format, UTIL_FORMAT_COLORSPACE_ZS, 0) == depth->bits.depth); assert(util_format_get_component_bits(depth->format, UTIL_FORMAT_COLORSPACE_ZS, 1) == depth->bits.stencil); diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index c9f020c40f..329ae03db2 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -48,30 +48,28 @@ static GLuint -format_max_bits( - pipe_format_rgbazs_t info ) +format_max_bits(enum pipe_format format) { - GLuint size = util_format_get_component_bits((enum pipe_format)info, UTIL_FORMAT_COLORSPACE_RGB, 0); + GLuint size = util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 0); - size = MAX2(size, util_format_get_component_bits((enum pipe_format)info, UTIL_FORMAT_COLORSPACE_RGB, 1)); - size = MAX2(size, util_format_get_component_bits((enum pipe_format)info, UTIL_FORMAT_COLORSPACE_RGB, 2)); - size = MAX2(size, util_format_get_component_bits((enum pipe_format)info, UTIL_FORMAT_COLORSPACE_RGB, 3)); - size = MAX2(size, util_format_get_component_bits((enum pipe_format)info, UTIL_FORMAT_COLORSPACE_ZS, 0)); - size = MAX2(size, util_format_get_component_bits((enum pipe_format)info, UTIL_FORMAT_COLORSPACE_ZS, 1)); + size = MAX2(size, util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 1)); + size = MAX2(size, util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 2)); + size = MAX2(size, util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 3)); + size = MAX2(size, util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_ZS, 0)); + size = MAX2(size, util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_ZS, 1)); return size; } static GLuint -format_size( - pipe_format_rgbazs_t info ) +format_size(enum pipe_format format) { return - util_format_get_component_bits((enum pipe_format)info, UTIL_FORMAT_COLORSPACE_RGB, 0) + - util_format_get_component_bits((enum pipe_format)info, UTIL_FORMAT_COLORSPACE_RGB, 1) + - util_format_get_component_bits((enum pipe_format)info, UTIL_FORMAT_COLORSPACE_RGB, 2) + - util_format_get_component_bits((enum pipe_format)info, UTIL_FORMAT_COLORSPACE_RGB, 3) + - util_format_get_component_bits((enum pipe_format)info, UTIL_FORMAT_COLORSPACE_ZS, 0) + - util_format_get_component_bits((enum pipe_format)info, UTIL_FORMAT_COLORSPACE_ZS, 1); + util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 0) + + util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 1) + + util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 2) + + util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 3) + + util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_ZS, 0) + + util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_ZS, 1); } /* @@ -80,14 +78,13 @@ format_size( GLboolean st_get_format_info(enum pipe_format format, struct pipe_format_info *pinfo) { - if (pf_layout(format) == PIPE_FORMAT_LAYOUT_RGBAZS) { - const struct util_format_description *desc = util_format_description(format); - pipe_format_rgbazs_t info; + const struct util_format_description *desc; - assert(desc); - - info = format; + desc = util_format_description(format); + assert(desc); + if (desc->layout == UTIL_FORMAT_LAYOUT_ARITH || + desc->layout == UTIL_FORMAT_LAYOUT_ARRAY) { #if 0 printf("%s\n", pf_name( format ) ); #endif @@ -100,22 +97,22 @@ st_get_format_info(enum pipe_format format, struct pipe_format_info *pinfo) pinfo->datatype = GL_UNSIGNED_INT_24_8; } else { - const GLuint size = format_max_bits( info ); + const GLuint size = format_max_bits(format); if (size == 8) { - if (pf_type(info) == PIPE_FORMAT_TYPE_UNORM) + if (desc->type == UTIL_FORMAT_TYPE_UNSIGNED) pinfo->datatype = GL_UNSIGNED_BYTE; else pinfo->datatype = GL_BYTE; } else if (size == 16) { - if (pf_type(info) == PIPE_FORMAT_TYPE_UNORM) + if (desc->type == UTIL_FORMAT_TYPE_UNSIGNED) pinfo->datatype = GL_UNSIGNED_SHORT; else pinfo->datatype = GL_SHORT; } else { assert( size <= 32 ); - if (pf_type(info) == PIPE_FORMAT_TYPE_UNORM) + if (desc->type == UTIL_FORMAT_TYPE_UNSIGNED) pinfo->datatype = GL_UNSIGNED_INT; else pinfo->datatype = GL_INT; @@ -123,17 +120,17 @@ st_get_format_info(enum pipe_format format, struct pipe_format_info *pinfo) } /* Component bits */ - pinfo->red_bits = util_format_get_component_bits((enum pipe_format)info, UTIL_FORMAT_COLORSPACE_RGB, 0); - pinfo->green_bits = util_format_get_component_bits((enum pipe_format)info, UTIL_FORMAT_COLORSPACE_RGB, 1); - pinfo->blue_bits = util_format_get_component_bits((enum pipe_format)info, UTIL_FORMAT_COLORSPACE_RGB, 2); - pinfo->alpha_bits = util_format_get_component_bits((enum pipe_format)info, UTIL_FORMAT_COLORSPACE_RGB, 3); - pinfo->depth_bits = util_format_get_component_bits((enum pipe_format)info, UTIL_FORMAT_COLORSPACE_ZS, 0); - pinfo->stencil_bits = util_format_get_component_bits((enum pipe_format)info, UTIL_FORMAT_COLORSPACE_ZS, 1); + pinfo->red_bits = util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 0); + pinfo->green_bits = util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 1); + pinfo->blue_bits = util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 2); + pinfo->alpha_bits = util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 3); + pinfo->depth_bits = util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_ZS, 0); + pinfo->stencil_bits = util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_ZS, 1); pinfo->luminance_bits = 0; pinfo->intensity_bits = 0; /* Format size */ - pinfo->size = format_size( info ) / 8; + pinfo->size = format_size(format) / 8; /* Luminance & Intensity bits */ if (desc->swizzle[0] == UTIL_FORMAT_SWIZZLE_X && @@ -150,7 +147,7 @@ st_get_format_info(enum pipe_format format, struct pipe_format_info *pinfo) pinfo->mesa_format = st_pipe_format_to_mesa_format(format); } - else if (pf_layout(format) == PIPE_FORMAT_LAYOUT_YCBCR) { + else if (desc->layout == UTIL_FORMAT_LAYOUT_YUV) { pinfo->mesa_format = MESA_FORMAT_YCBCR; pinfo->datatype = GL_UNSIGNED_SHORT; pinfo->size = 2; /* two bytes per "texel" */ -- cgit v1.2.3 From 876a785a182d7987786377ff0a44ee40628254f3 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Tue, 8 Dec 2009 19:58:13 +0100 Subject: Format layout cannot be used to distinguish scaled/normalised formats. --- src/gallium/drivers/nv50/nv50_vbo.c | 18 ++++-------------- src/gallium/drivers/r300/r300_state_inlines.h | 2 +- 2 files changed, 5 insertions(+), 15 deletions(-) (limited to 'src/gallium/drivers/nv50/nv50_vbo.c') diff --git a/src/gallium/drivers/nv50/nv50_vbo.c b/src/gallium/drivers/nv50/nv50_vbo.c index f3472f266e..d1c9f3f590 100644 --- a/src/gallium/drivers/nv50/nv50_vbo.c +++ b/src/gallium/drivers/nv50/nv50_vbo.c @@ -75,25 +75,15 @@ nv50_vbo_type_to_hw(enum pipe_format format) case UTIL_FORMAT_TYPE_FLOAT: return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_FLOAT; case UTIL_FORMAT_TYPE_UNSIGNED: - switch (desc->layout) { - case UTIL_FORMAT_LAYOUT_ARITH: + if (desc->channel[0].normalized) { return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_UNORM; - case UTIL_FORMAT_LAYOUT_ARRAY: - return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_USCALED; - default: - return 0; } - break; + return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_USCALED; case UTIL_FORMAT_TYPE_SIGNED: - switch (desc->layout) { - case UTIL_FORMAT_LAYOUT_ARITH: + if (desc->channel[0].normalized) { return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_SNORM; - case UTIL_FORMAT_LAYOUT_ARRAY: - return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_SSCALED; - default: - return 0; } - break; + return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_SSCALED; /* case PIPE_FORMAT_TYPE_UINT: return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_UINT; diff --git a/src/gallium/drivers/r300/r300_state_inlines.h b/src/gallium/drivers/r300/r300_state_inlines.h index 5c58655da1..7cd1f87630 100644 --- a/src/gallium/drivers/r300/r300_state_inlines.h +++ b/src/gallium/drivers/r300/r300_state_inlines.h @@ -528,7 +528,7 @@ r300_translate_vertex_data_type(enum pipe_format format) { if (desc->type == UTIL_FORMAT_TYPE_SIGNED) { result |= R300_SIGNED; } - if (desc->layout == UTIL_FORMAT_LAYOUT_ARITH) { + if (desc->channel[0].normalized) { result |= R300_NORMALIZE; } -- cgit v1.2.3 From 3af2ddbe943f0fe0d6b0ba9e627cbb82d0cc79f2 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Thu, 17 Dec 2009 21:25:47 +0100 Subject: s/desc->type/desc->channel[0].type/ --- src/gallium/drivers/nv50/nv50_vbo.c | 4 ++-- src/gallium/drivers/r300/r300_state_inlines.h | 4 ++-- src/mesa/state_tracker/st_format.c | 6 +++--- 3 files changed, 7 insertions(+), 7 deletions(-) (limited to 'src/gallium/drivers/nv50/nv50_vbo.c') diff --git a/src/gallium/drivers/nv50/nv50_vbo.c b/src/gallium/drivers/nv50/nv50_vbo.c index d1c9f3f590..f7fa0659e8 100644 --- a/src/gallium/drivers/nv50/nv50_vbo.c +++ b/src/gallium/drivers/nv50/nv50_vbo.c @@ -71,7 +71,7 @@ nv50_vbo_type_to_hw(enum pipe_format format) desc = util_format_description(format); assert(desc); - switch (desc->type) { + switch (desc->channel[0].type) { case UTIL_FORMAT_TYPE_FLOAT: return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_FLOAT; case UTIL_FORMAT_TYPE_UNSIGNED: @@ -339,7 +339,7 @@ nv50_vbo_static_attrib(struct nv50_context *nv50, unsigned attrib, desc = util_format_description(pf); assert(desc); - if ((desc->type != UTIL_FORMAT_TYPE_FLOAT) || + if ((desc->channel[0].type != UTIL_FORMAT_TYPE_FLOAT) || util_format_get_component_bits(pf, UTIL_FORMAT_COLORSPACE_RGB, 0) != 32) return FALSE; diff --git a/src/gallium/drivers/r300/r300_state_inlines.h b/src/gallium/drivers/r300/r300_state_inlines.h index 7cd1f87630..dbe42edd91 100644 --- a/src/gallium/drivers/r300/r300_state_inlines.h +++ b/src/gallium/drivers/r300/r300_state_inlines.h @@ -483,7 +483,7 @@ r300_translate_vertex_data_type(enum pipe_format format) { assert(0); } - switch (desc->type) { + switch (desc->channel[0].type) { /* Half-floats, floats, doubles */ case UTIL_FORMAT_TYPE_FLOAT: switch (util_format_get_component_bits(format, UTIL_FORMAT_COLORSPACE_RGB, 0)) { @@ -525,7 +525,7 @@ r300_translate_vertex_data_type(enum pipe_format format) { assert(0); } - if (desc->type == UTIL_FORMAT_TYPE_SIGNED) { + if (desc->channel[0].type == UTIL_FORMAT_TYPE_SIGNED) { result |= R300_SIGNED; } if (desc->channel[0].normalized) { diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index 329ae03db2..94da0f61e0 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -99,20 +99,20 @@ st_get_format_info(enum pipe_format format, struct pipe_format_info *pinfo) else { const GLuint size = format_max_bits(format); if (size == 8) { - if (desc->type == UTIL_FORMAT_TYPE_UNSIGNED) + if (desc->channel[0].type == UTIL_FORMAT_TYPE_UNSIGNED) pinfo->datatype = GL_UNSIGNED_BYTE; else pinfo->datatype = GL_BYTE; } else if (size == 16) { - if (desc->type == UTIL_FORMAT_TYPE_UNSIGNED) + if (desc->channel[0].type == UTIL_FORMAT_TYPE_UNSIGNED) pinfo->datatype = GL_UNSIGNED_SHORT; else pinfo->datatype = GL_SHORT; } else { assert( size <= 32 ); - if (desc->type == UTIL_FORMAT_TYPE_UNSIGNED) + if (desc->channel[0].type == UTIL_FORMAT_TYPE_UNSIGNED) pinfo->datatype = GL_UNSIGNED_INT; else pinfo->datatype = GL_INT; -- cgit v1.2.3 From 03f212b0d85fed5dec9a855fb6d079e5fdb60ac9 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 21 Dec 2009 22:47:21 +0000 Subject: gallium: propogate draw retval changes into more drivers --- src/gallium/drivers/i915/i915_context.c | 18 ++++---- src/gallium/drivers/llvmpipe/lp_draw_arrays.c | 18 ++++---- src/gallium/drivers/llvmpipe/lp_state.h | 6 +-- src/gallium/drivers/nv04/nv04_context.h | 4 +- src/gallium/drivers/nv04/nv04_vbo.c | 10 ++--- src/gallium/drivers/nv10/nv10_context.h | 4 +- src/gallium/drivers/nv10/nv10_vbo.c | 10 ++--- src/gallium/drivers/nv20/nv20_context.h | 4 +- src/gallium/drivers/nv20/nv20_vbo.c | 7 ++- src/gallium/drivers/nv30/nv30_context.h | 4 +- src/gallium/drivers/nv30/nv30_vbo.c | 16 +++---- src/gallium/drivers/nv40/nv40_context.h | 6 +-- src/gallium/drivers/nv40/nv40_draw.c | 6 +-- src/gallium/drivers/nv40/nv40_vbo.c | 23 +++++----- src/gallium/drivers/nv50/nv50_context.h | 4 +- src/gallium/drivers/nv50/nv50_vbo.c | 14 +++--- src/gallium/drivers/r300/r300_render.c | 62 +++++++++++++-------------- src/gallium/drivers/r300/r300_render.h | 60 +++++++++++++------------- src/gallium/drivers/softpipe/sp_draw_arrays.c | 18 ++++---- src/gallium/drivers/softpipe/sp_state.h | 16 +++---- 20 files changed, 146 insertions(+), 164 deletions(-) (limited to 'src/gallium/drivers/nv50/nv50_vbo.c') diff --git a/src/gallium/drivers/i915/i915_context.c b/src/gallium/drivers/i915/i915_context.c index 94c8aee30f..753f37e095 100644 --- a/src/gallium/drivers/i915/i915_context.c +++ b/src/gallium/drivers/i915/i915_context.c @@ -45,7 +45,7 @@ */ -static boolean +static void i915_draw_range_elements(struct pipe_context *pipe, struct pipe_buffer *indexBuffer, unsigned indexSize, @@ -106,27 +106,25 @@ i915_draw_range_elements(struct pipe_context *pipe, pipe_buffer_unmap(pipe->screen, indexBuffer); draw_set_mapped_element_buffer_range(draw, 0, start, start + count - 1, NULL); } - - return TRUE; } -static boolean +static void i915_draw_elements(struct pipe_context *pipe, struct pipe_buffer *indexBuffer, unsigned indexSize, unsigned prim, unsigned start, unsigned count) { - return i915_draw_range_elements(pipe, indexBuffer, - indexSize, - 0, 0xffffffff, - prim, start, count); + i915_draw_range_elements(pipe, indexBuffer, + indexSize, + 0, 0xffffffff, + prim, start, count); } -static boolean +static void i915_draw_arrays(struct pipe_context *pipe, unsigned prim, unsigned start, unsigned count) { - return i915_draw_elements(pipe, NULL, 0, prim, start, count); + i915_draw_elements(pipe, NULL, 0, prim, start, count); } diff --git a/src/gallium/drivers/llvmpipe/lp_draw_arrays.c b/src/gallium/drivers/llvmpipe/lp_draw_arrays.c index 0aa13a1fc6..57ce5b17a2 100644 --- a/src/gallium/drivers/llvmpipe/lp_draw_arrays.c +++ b/src/gallium/drivers/llvmpipe/lp_draw_arrays.c @@ -45,11 +45,11 @@ -boolean +void llvmpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, unsigned start, unsigned count) { - return llvmpipe_draw_elements(pipe, NULL, 0, mode, start, count); + llvmpipe_draw_elements(pipe, NULL, 0, mode, start, count); } @@ -58,7 +58,7 @@ llvmpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, * Basically, map the vertex buffers (and drawing surfaces), then hand off * the drawing to the 'draw' module. */ -boolean +void llvmpipe_draw_range_elements(struct pipe_context *pipe, struct pipe_buffer *indexBuffer, unsigned indexSize, @@ -116,21 +116,19 @@ llvmpipe_draw_range_elements(struct pipe_context *pipe, /* Note: leave drawing surfaces mapped */ lp->dirty_render_cache = TRUE; - - return TRUE; } -boolean +void llvmpipe_draw_elements(struct pipe_context *pipe, struct pipe_buffer *indexBuffer, unsigned indexSize, unsigned mode, unsigned start, unsigned count) { - return llvmpipe_draw_range_elements( pipe, indexBuffer, - indexSize, - 0, 0xffffffff, - mode, start, count ); + llvmpipe_draw_range_elements( pipe, indexBuffer, + indexSize, + 0, 0xffffffff, + mode, start, count ); } diff --git a/src/gallium/drivers/llvmpipe/lp_state.h b/src/gallium/drivers/llvmpipe/lp_state.h index d1c74ab07b..2857853586 100644 --- a/src/gallium/drivers/llvmpipe/lp_state.h +++ b/src/gallium/drivers/llvmpipe/lp_state.h @@ -197,14 +197,14 @@ void llvmpipe_update_fs(struct llvmpipe_context *lp); void llvmpipe_update_derived( struct llvmpipe_context *llvmpipe ); -boolean llvmpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, +void llvmpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, unsigned start, unsigned count); -boolean llvmpipe_draw_elements(struct pipe_context *pipe, +void llvmpipe_draw_elements(struct pipe_context *pipe, struct pipe_buffer *indexBuffer, unsigned indexSize, unsigned mode, unsigned start, unsigned count); -boolean +void llvmpipe_draw_range_elements(struct pipe_context *pipe, struct pipe_buffer *indexBuffer, unsigned indexSize, diff --git a/src/gallium/drivers/nv04/nv04_context.h b/src/gallium/drivers/nv04/nv04_context.h index 55326c787a..5951115293 100644 --- a/src/gallium/drivers/nv04/nv04_context.h +++ b/src/gallium/drivers/nv04/nv04_context.h @@ -141,9 +141,9 @@ extern void nv04_emit_hw_state(struct nv04_context *nv04); extern void nv04_state_tex_update(struct nv04_context *nv04); /* nv04_vbo.c */ -extern boolean nv04_draw_arrays(struct pipe_context *, unsigned mode, +extern void nv04_draw_arrays(struct pipe_context *, unsigned mode, unsigned start, unsigned count); -extern boolean nv04_draw_elements( struct pipe_context *pipe, +extern void nv04_draw_elements( struct pipe_context *pipe, struct pipe_buffer *indexBuffer, unsigned indexSize, unsigned prim, unsigned start, unsigned count); diff --git a/src/gallium/drivers/nv04/nv04_vbo.c b/src/gallium/drivers/nv04/nv04_vbo.c index e3167814f2..704fae8c6c 100644 --- a/src/gallium/drivers/nv04/nv04_vbo.c +++ b/src/gallium/drivers/nv04/nv04_vbo.c @@ -9,7 +9,7 @@ #include "nouveau/nouveau_channel.h" #include "nouveau/nouveau_pushbuf.h" -boolean nv04_draw_elements( struct pipe_context *pipe, +void nv04_draw_elements( struct pipe_context *pipe, struct pipe_buffer *indexBuffer, unsigned indexSize, unsigned prim, unsigned start, unsigned count) @@ -65,15 +65,13 @@ boolean nv04_draw_elements( struct pipe_context *pipe, pipe_buffer_unmap(pscreen, indexBuffer); draw_set_mapped_element_buffer(draw, 0, NULL); } - - return TRUE; } -boolean nv04_draw_arrays( struct pipe_context *pipe, - unsigned prim, unsigned start, unsigned count) +void nv04_draw_arrays( struct pipe_context *pipe, + unsigned prim, unsigned start, unsigned count) { printf("coucou in draw arrays\n"); - return nv04_draw_elements(pipe, NULL, 0, prim, start, count); + nv04_draw_elements(pipe, NULL, 0, prim, start, count); } diff --git a/src/gallium/drivers/nv10/nv10_context.h b/src/gallium/drivers/nv10/nv10_context.h index 36a6aa7a74..3f829fd106 100644 --- a/src/gallium/drivers/nv10/nv10_context.h +++ b/src/gallium/drivers/nv10/nv10_context.h @@ -144,9 +144,9 @@ extern void nv10_emit_hw_state(struct nv10_context *nv10); extern void nv10_state_tex_update(struct nv10_context *nv10); /* nv10_vbo.c */ -extern boolean nv10_draw_arrays(struct pipe_context *, unsigned mode, +extern void nv10_draw_arrays(struct pipe_context *, unsigned mode, unsigned start, unsigned count); -extern boolean nv10_draw_elements( struct pipe_context *pipe, +extern void nv10_draw_elements( struct pipe_context *pipe, struct pipe_buffer *indexBuffer, unsigned indexSize, unsigned prim, unsigned start, unsigned count); diff --git a/src/gallium/drivers/nv10/nv10_vbo.c b/src/gallium/drivers/nv10/nv10_vbo.c index 441a4f75f3..230f2e6d39 100644 --- a/src/gallium/drivers/nv10/nv10_vbo.c +++ b/src/gallium/drivers/nv10/nv10_vbo.c @@ -9,7 +9,7 @@ #include "nouveau/nouveau_channel.h" #include "nouveau/nouveau_pushbuf.h" -boolean nv10_draw_elements( struct pipe_context *pipe, +void nv10_draw_elements( struct pipe_context *pipe, struct pipe_buffer *indexBuffer, unsigned indexSize, unsigned prim, unsigned start, unsigned count) @@ -64,14 +64,12 @@ boolean nv10_draw_elements( struct pipe_context *pipe, pipe_buffer_unmap(pscreen, indexBuffer); draw_set_mapped_element_buffer(draw, 0, NULL); } - - return TRUE; } -boolean nv10_draw_arrays( struct pipe_context *pipe, - unsigned prim, unsigned start, unsigned count) +void nv10_draw_arrays( struct pipe_context *pipe, + unsigned prim, unsigned start, unsigned count) { - return nv10_draw_elements(pipe, NULL, 0, prim, start, count); + nv10_draw_elements(pipe, NULL, 0, prim, start, count); } diff --git a/src/gallium/drivers/nv20/nv20_context.h b/src/gallium/drivers/nv20/nv20_context.h index a4eaa95660..c88a1bd9bd 100644 --- a/src/gallium/drivers/nv20/nv20_context.h +++ b/src/gallium/drivers/nv20/nv20_context.h @@ -143,9 +143,9 @@ extern void nv20_emit_hw_state(struct nv20_context *nv20); extern void nv20_state_tex_update(struct nv20_context *nv20); /* nv20_vbo.c */ -extern boolean nv20_draw_arrays(struct pipe_context *, unsigned mode, +extern void nv20_draw_arrays(struct pipe_context *, unsigned mode, unsigned start, unsigned count); -extern boolean nv20_draw_elements( struct pipe_context *pipe, +extern void nv20_draw_elements( struct pipe_context *pipe, struct pipe_buffer *indexBuffer, unsigned indexSize, unsigned prim, unsigned start, unsigned count); diff --git a/src/gallium/drivers/nv20/nv20_vbo.c b/src/gallium/drivers/nv20/nv20_vbo.c index 84d7db6c5e..2f5e12233f 100644 --- a/src/gallium/drivers/nv20/nv20_vbo.c +++ b/src/gallium/drivers/nv20/nv20_vbo.c @@ -9,7 +9,7 @@ #include "nouveau/nouveau_channel.h" #include "nouveau/nouveau_pushbuf.h" -boolean nv20_draw_elements( struct pipe_context *pipe, +void nv20_draw_elements( struct pipe_context *pipe, struct pipe_buffer *indexBuffer, unsigned indexSize, unsigned prim, unsigned start, unsigned count) @@ -67,13 +67,12 @@ boolean nv20_draw_elements( struct pipe_context *pipe, } draw_flush(nv20->draw); - return TRUE; } -boolean nv20_draw_arrays( struct pipe_context *pipe, +void nv20_draw_arrays( struct pipe_context *pipe, unsigned prim, unsigned start, unsigned count) { - return nv20_draw_elements(pipe, NULL, 0, prim, start, count); + nv20_draw_elements(pipe, NULL, 0, prim, start, count); } diff --git a/src/gallium/drivers/nv30/nv30_context.h b/src/gallium/drivers/nv30/nv30_context.h index 6f44b1c7fe..17f8660590 100644 --- a/src/gallium/drivers/nv30/nv30_context.h +++ b/src/gallium/drivers/nv30/nv30_context.h @@ -199,9 +199,9 @@ extern struct nv30_state_entry nv30_state_fragtex; extern struct nv30_state_entry nv30_state_vbo; /* nv30_vbo.c */ -extern boolean nv30_draw_arrays(struct pipe_context *, unsigned mode, +extern void nv30_draw_arrays(struct pipe_context *, unsigned mode, unsigned start, unsigned count); -extern boolean nv30_draw_elements(struct pipe_context *pipe, +extern void nv30_draw_elements(struct pipe_context *pipe, struct pipe_buffer *indexBuffer, unsigned indexSize, unsigned mode, unsigned start, diff --git a/src/gallium/drivers/nv30/nv30_vbo.c b/src/gallium/drivers/nv30/nv30_vbo.c index 189656ec81..d359f71f43 100644 --- a/src/gallium/drivers/nv30/nv30_vbo.c +++ b/src/gallium/drivers/nv30/nv30_vbo.c @@ -163,7 +163,7 @@ nv30_vbo_static_attrib(struct nv30_context *nv30, struct nouveau_stateobj *so, return TRUE; } -boolean +void nv30_draw_arrays(struct pipe_context *pipe, unsigned mode, unsigned start, unsigned count) { @@ -175,7 +175,7 @@ nv30_draw_arrays(struct pipe_context *pipe, if (FORCE_SWTNL || !nv30_state_validate(nv30)) { /*return nv30_draw_elements_swtnl(pipe, NULL, 0, mode, start, count);*/ - return FALSE; + return; } while (count) { @@ -362,7 +362,7 @@ nv30_draw_elements_u32(struct nv30_context *nv30, void *ib, } } -static boolean +static void nv30_draw_elements_inline(struct pipe_context *pipe, struct pipe_buffer *ib, unsigned ib_size, unsigned mode, unsigned start, unsigned count) @@ -393,10 +393,9 @@ nv30_draw_elements_inline(struct pipe_context *pipe, } pipe_buffer_unmap(pscreen, ib); - return TRUE; } -static boolean +static void nv30_draw_elements_vbo(struct pipe_context *pipe, unsigned mode, unsigned start, unsigned count) { @@ -445,11 +444,9 @@ nv30_draw_elements_vbo(struct pipe_context *pipe, count -= vc; start = restart; } - - return TRUE; } -boolean +void nv30_draw_elements(struct pipe_context *pipe, struct pipe_buffer *indexBuffer, unsigned indexSize, unsigned mode, unsigned start, unsigned count) @@ -461,7 +458,7 @@ nv30_draw_elements(struct pipe_context *pipe, if (FORCE_SWTNL || !nv30_state_validate(nv30)) { /*return nv30_draw_elements_swtnl(pipe, NULL, 0, mode, start, count);*/ - return FALSE; + return; } if (idxbuf) { @@ -472,7 +469,6 @@ nv30_draw_elements(struct pipe_context *pipe, } pipe->flush(pipe, 0, NULL); - return TRUE; } static boolean diff --git a/src/gallium/drivers/nv40/nv40_context.h b/src/gallium/drivers/nv40/nv40_context.h index cf33b64a86..d12b9d88cb 100644 --- a/src/gallium/drivers/nv40/nv40_context.h +++ b/src/gallium/drivers/nv40/nv40_context.h @@ -184,7 +184,7 @@ extern void nv40_screen_init_miptree_functions(struct pipe_screen *pscreen); /* nv40_draw.c */ extern struct draw_stage *nv40_draw_render_stage(struct nv40_context *nv40); -extern boolean nv40_draw_elements_swtnl(struct pipe_context *pipe, +extern void nv40_draw_elements_swtnl(struct pipe_context *pipe, struct pipe_buffer *idxbuf, unsigned ib_size, unsigned mode, unsigned start, unsigned count); @@ -220,9 +220,9 @@ extern struct nv40_state_entry nv40_state_vbo; extern struct nv40_state_entry nv40_state_vtxfmt; /* nv40_vbo.c */ -extern boolean nv40_draw_arrays(struct pipe_context *, unsigned mode, +extern void nv40_draw_arrays(struct pipe_context *, unsigned mode, unsigned start, unsigned count); -extern boolean nv40_draw_elements(struct pipe_context *pipe, +extern void nv40_draw_elements(struct pipe_context *pipe, struct pipe_buffer *indexBuffer, unsigned indexSize, unsigned mode, unsigned start, diff --git a/src/gallium/drivers/nv40/nv40_draw.c b/src/gallium/drivers/nv40/nv40_draw.c index b2f19ecb69..38aba92a14 100644 --- a/src/gallium/drivers/nv40/nv40_draw.c +++ b/src/gallium/drivers/nv40/nv40_draw.c @@ -226,7 +226,7 @@ nv40_draw_render_stage(struct nv40_context *nv40) return &render->stage; } -boolean +void nv40_draw_elements_swtnl(struct pipe_context *pipe, struct pipe_buffer *idxbuf, unsigned idxbuf_size, unsigned mode, unsigned start, unsigned count) @@ -237,7 +237,7 @@ nv40_draw_elements_swtnl(struct pipe_context *pipe, void *map; if (!nv40_state_validate_swtnl(nv40)) - return FALSE; + return; nv40->state.dirty &= ~(1ULL << NV40_STATE_VTXBUF); nv40_state_emit(nv40); @@ -277,8 +277,6 @@ nv40_draw_elements_swtnl(struct pipe_context *pipe, draw_flush(nv40->draw); pipe->flush(pipe, 0, NULL); - - return TRUE; } static INLINE void diff --git a/src/gallium/drivers/nv40/nv40_vbo.c b/src/gallium/drivers/nv40/nv40_vbo.c index b2753b8e2e..eecdf01915 100644 --- a/src/gallium/drivers/nv40/nv40_vbo.c +++ b/src/gallium/drivers/nv40/nv40_vbo.c @@ -164,7 +164,7 @@ nv40_vbo_static_attrib(struct nv40_context *nv40, struct nouveau_stateobj *so, return TRUE; } -boolean +void nv40_draw_arrays(struct pipe_context *pipe, unsigned mode, unsigned start, unsigned count) { @@ -174,8 +174,9 @@ nv40_draw_arrays(struct pipe_context *pipe, nv40_vbo_set_idxbuf(nv40, NULL, 0); if (FORCE_SWTNL || !nv40_state_validate(nv40)) { - return nv40_draw_elements_swtnl(pipe, NULL, 0, - mode, start, count); + nv40_draw_elements_swtnl(pipe, NULL, 0, + mode, start, count); + return; } while (count) { @@ -221,7 +222,6 @@ nv40_draw_arrays(struct pipe_context *pipe, } pipe->flush(pipe, 0, NULL); - return TRUE; } static INLINE void @@ -362,7 +362,7 @@ nv40_draw_elements_u32(struct nv40_context *nv40, void *ib, } } -static boolean +static void nv40_draw_elements_inline(struct pipe_context *pipe, struct pipe_buffer *ib, unsigned ib_size, unsigned mode, unsigned start, unsigned count) @@ -393,10 +393,9 @@ nv40_draw_elements_inline(struct pipe_context *pipe, } pipe_buffer_unmap(pscreen, ib); - return TRUE; } -static boolean +static void nv40_draw_elements_vbo(struct pipe_context *pipe, unsigned mode, unsigned start, unsigned count) { @@ -445,11 +444,9 @@ nv40_draw_elements_vbo(struct pipe_context *pipe, count -= vc; start = restart; } - - return TRUE; } -boolean +void nv40_draw_elements(struct pipe_context *pipe, struct pipe_buffer *indexBuffer, unsigned indexSize, unsigned mode, unsigned start, unsigned count) @@ -459,8 +456,9 @@ nv40_draw_elements(struct pipe_context *pipe, idxbuf = nv40_vbo_set_idxbuf(nv40, indexBuffer, indexSize); if (FORCE_SWTNL || !nv40_state_validate(nv40)) { - return nv40_draw_elements_swtnl(pipe, NULL, 0, - mode, start, count); + nv40_draw_elements_swtnl(pipe, NULL, 0, + mode, start, count); + return; } if (idxbuf) { @@ -471,7 +469,6 @@ nv40_draw_elements(struct pipe_context *pipe, } pipe->flush(pipe, 0, NULL); - return TRUE; } static boolean diff --git a/src/gallium/drivers/nv50/nv50_context.h b/src/gallium/drivers/nv50/nv50_context.h index 5578a5838f..cbd4c3ff86 100644 --- a/src/gallium/drivers/nv50/nv50_context.h +++ b/src/gallium/drivers/nv50/nv50_context.h @@ -191,9 +191,9 @@ nv50_surface_do_copy(struct nv50_screen *screen, struct pipe_surface *dst, extern struct draw_stage *nv50_draw_render_stage(struct nv50_context *nv50); /* nv50_vbo.c */ -extern boolean nv50_draw_arrays(struct pipe_context *, unsigned mode, +extern void nv50_draw_arrays(struct pipe_context *, unsigned mode, unsigned start, unsigned count); -extern boolean nv50_draw_elements(struct pipe_context *pipe, +extern void nv50_draw_elements(struct pipe_context *pipe, struct pipe_buffer *indexBuffer, unsigned indexSize, unsigned mode, unsigned start, diff --git a/src/gallium/drivers/nv50/nv50_vbo.c b/src/gallium/drivers/nv50/nv50_vbo.c index f7fa0659e8..ca8608e987 100644 --- a/src/gallium/drivers/nv50/nv50_vbo.c +++ b/src/gallium/drivers/nv50/nv50_vbo.c @@ -152,7 +152,7 @@ nv50_vbo_vtxelt_to_hw(struct pipe_vertex_element *ve) return (hw_type | hw_size); } -boolean +void nv50_draw_arrays(struct pipe_context *pipe, unsigned mode, unsigned start, unsigned count) { @@ -182,7 +182,9 @@ nv50_draw_arrays(struct pipe_context *pipe, unsigned mode, unsigned start, BEGIN_RING(chan, tesla, NV50TCL_VERTEX_END, 1); OUT_RING (chan, 0); - return ret; + /* XXX: not sure what to do if ret != TRUE: flush and retry? + */ + assert(ret); } static INLINE boolean @@ -275,7 +277,7 @@ nv50_draw_elements_inline_u32(struct nv50_context *nv50, uint32_t *map, return TRUE; } -boolean +void nv50_draw_elements(struct pipe_context *pipe, struct pipe_buffer *indexBuffer, unsigned indexSize, unsigned mode, unsigned start, unsigned count) @@ -317,8 +319,10 @@ nv50_draw_elements(struct pipe_context *pipe, OUT_RING (chan, 0); pipe_buffer_unmap(pscreen, indexBuffer); - - return ret; + + /* XXX: what to do if ret != TRUE? Flush and retry? + */ + assert(ret); } static INLINE boolean diff --git a/src/gallium/drivers/r300/r300_render.c b/src/gallium/drivers/r300/r300_render.c index 2d70ec2ac9..87ad30ac30 100644 --- a/src/gallium/drivers/r300/r300_render.c +++ b/src/gallium/drivers/r300/r300_render.c @@ -213,7 +213,7 @@ validate: } /* This is the fast-path drawing & emission for HW TCL. */ -boolean r300_draw_range_elements(struct pipe_context* pipe, +void r300_draw_range_elements(struct pipe_context* pipe, struct pipe_buffer* indexBuffer, unsigned indexSize, unsigned minIndex, @@ -225,30 +225,33 @@ boolean r300_draw_range_elements(struct pipe_context* pipe, struct r300_context* r300 = r300_context(pipe); if (!u_trim_pipe_prim(mode, &count)) { - return FALSE; + return; } if (count > 65535) { - return FALSE; + /* XXX: use aux/indices functions to split this into smaller + * primitives. + */ + return; } if (r300_nothing_to_draw(r300)) { - return TRUE; + return; } r300_update_derived_state(r300); if (!r300_setup_vertex_buffers(r300)) { - return FALSE; + return; } if (!r300->winsys->add_buffer(r300->winsys, indexBuffer, RADEON_GEM_DOMAIN_GTT, 0)) { - return FALSE; + return; } if (!r300->winsys->validate(r300->winsys)) { - return FALSE; + return; } r300_emit_dirty_state(r300); @@ -257,41 +260,42 @@ boolean r300_draw_range_elements(struct pipe_context* pipe, r300_emit_draw_elements(r300, indexBuffer, indexSize, minIndex, maxIndex, mode, start, count); - - return TRUE; } /* Simple helpers for context setup. Should probably be moved to util. */ -boolean r300_draw_elements(struct pipe_context* pipe, - struct pipe_buffer* indexBuffer, - unsigned indexSize, unsigned mode, - unsigned start, unsigned count) +void r300_draw_elements(struct pipe_context* pipe, + struct pipe_buffer* indexBuffer, + unsigned indexSize, unsigned mode, + unsigned start, unsigned count) { - return pipe->draw_range_elements(pipe, indexBuffer, indexSize, 0, ~0, - mode, start, count); + pipe->draw_range_elements(pipe, indexBuffer, indexSize, 0, ~0, + mode, start, count); } -boolean r300_draw_arrays(struct pipe_context* pipe, unsigned mode, +void r300_draw_arrays(struct pipe_context* pipe, unsigned mode, unsigned start, unsigned count) { struct r300_context* r300 = r300_context(pipe); if (!u_trim_pipe_prim(mode, &count)) { - return FALSE; + return; } if (count > 65535) { - return FALSE; + /* XXX: driver needs to handle this -- use the functions in + * aux/indices to split this into several smaller primitives. + */ + return; } if (r300_nothing_to_draw(r300)) { - return TRUE; + return; } r300_update_derived_state(r300); if (!r300_setup_vertex_buffers(r300)) { - return FALSE; + return; } r300_emit_dirty_state(r300); @@ -299,8 +303,6 @@ boolean r300_draw_arrays(struct pipe_context* pipe, unsigned mode, r300_emit_aos(r300, start); r300_emit_draw_arrays(r300, mode, count); - - return TRUE; } /**************************************************************************** @@ -309,7 +311,7 @@ boolean r300_draw_arrays(struct pipe_context* pipe, unsigned mode, ***************************************************************************/ /* SW TCL arrays, using Draw. */ -boolean r300_swtcl_draw_arrays(struct pipe_context* pipe, +void r300_swtcl_draw_arrays(struct pipe_context* pipe, unsigned mode, unsigned start, unsigned count) @@ -318,11 +320,11 @@ boolean r300_swtcl_draw_arrays(struct pipe_context* pipe, int i; if (!u_trim_pipe_prim(mode, &count)) { - return FALSE; + return; } if (r300_nothing_to_draw(r300)) { - return TRUE; + return; } for (i = 0; i < r300->vertex_buffer_count; i++) { @@ -345,12 +347,10 @@ boolean r300_swtcl_draw_arrays(struct pipe_context* pipe, pipe_buffer_unmap(pipe->screen, r300->vertex_buffer[i].buffer); draw_set_mapped_vertex_buffer(r300->draw, i, NULL); } - - return TRUE; } /* SW TCL elements, using Draw. */ -boolean r300_swtcl_draw_range_elements(struct pipe_context* pipe, +void r300_swtcl_draw_range_elements(struct pipe_context* pipe, struct pipe_buffer* indexBuffer, unsigned indexSize, unsigned minIndex, @@ -363,11 +363,11 @@ boolean r300_swtcl_draw_range_elements(struct pipe_context* pipe, int i; if (!u_trim_pipe_prim(mode, &count)) { - return FALSE; + return; } if (r300_nothing_to_draw(r300)) { - return TRUE; + return; } for (i = 0; i < r300->vertex_buffer_count; i++) { @@ -397,8 +397,6 @@ boolean r300_swtcl_draw_range_elements(struct pipe_context* pipe, pipe_buffer_unmap(pipe->screen, indexBuffer); draw_set_mapped_element_buffer_range(r300->draw, 0, start, start + count - 1, NULL); - - return TRUE; } /* Object for rendering using Draw. */ diff --git a/src/gallium/drivers/r300/r300_render.h b/src/gallium/drivers/r300/r300_render.h index da83069083..27b5e6a963 100644 --- a/src/gallium/drivers/r300/r300_render.h +++ b/src/gallium/drivers/r300/r300_render.h @@ -25,35 +25,35 @@ uint32_t r300_translate_primitive(unsigned prim); -boolean r300_draw_range_elements(struct pipe_context* pipe, - struct pipe_buffer* indexBuffer, - unsigned indexSize, - unsigned minIndex, - unsigned maxIndex, - unsigned mode, - unsigned start, - unsigned count); - -boolean r300_draw_elements(struct pipe_context* pipe, - struct pipe_buffer* indexBuffer, - unsigned indexSize, unsigned mode, - unsigned start, unsigned count); - -boolean r300_draw_arrays(struct pipe_context* pipe, unsigned mode, - unsigned start, unsigned count); - -boolean r300_swtcl_draw_arrays(struct pipe_context* pipe, - unsigned mode, - unsigned start, - unsigned count); - -boolean r300_swtcl_draw_range_elements(struct pipe_context* pipe, - struct pipe_buffer* indexBuffer, - unsigned indexSize, - unsigned minIndex, - unsigned maxIndex, - unsigned mode, - unsigned start, - unsigned count); +void r300_draw_range_elements(struct pipe_context* pipe, + struct pipe_buffer* indexBuffer, + unsigned indexSize, + unsigned minIndex, + unsigned maxIndex, + unsigned mode, + unsigned start, + unsigned count); + +void r300_draw_elements(struct pipe_context* pipe, + struct pipe_buffer* indexBuffer, + unsigned indexSize, unsigned mode, + unsigned start, unsigned count); + +void r300_draw_arrays(struct pipe_context* pipe, unsigned mode, + unsigned start, unsigned count); + +void r300_swtcl_draw_arrays(struct pipe_context* pipe, + unsigned mode, + unsigned start, + unsigned count); + +void r300_swtcl_draw_range_elements(struct pipe_context* pipe, + struct pipe_buffer* indexBuffer, + unsigned indexSize, + unsigned minIndex, + unsigned maxIndex, + unsigned mode, + unsigned start, + unsigned count); #endif /* R300_RENDER_H */ diff --git a/src/gallium/drivers/softpipe/sp_draw_arrays.c b/src/gallium/drivers/softpipe/sp_draw_arrays.c index d4045816d0..70a92fd4c6 100644 --- a/src/gallium/drivers/softpipe/sp_draw_arrays.c +++ b/src/gallium/drivers/softpipe/sp_draw_arrays.c @@ -88,11 +88,11 @@ softpipe_unmap_constant_buffers(struct softpipe_context *sp) } -boolean +void softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, unsigned start, unsigned count) { - return softpipe_draw_elements(pipe, NULL, 0, mode, start, count); + softpipe_draw_elements(pipe, NULL, 0, mode, start, count); } @@ -101,7 +101,7 @@ softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, * Basically, map the vertex buffers (and drawing surfaces), then hand off * the drawing to the 'draw' module. */ -boolean +void softpipe_draw_range_elements(struct pipe_context *pipe, struct pipe_buffer *indexBuffer, unsigned indexSize, @@ -168,21 +168,19 @@ softpipe_draw_range_elements(struct pipe_context *pipe, softpipe_unmap_constant_buffers(sp); sp->dirty_render_cache = TRUE; - - return TRUE; } -boolean +void softpipe_draw_elements(struct pipe_context *pipe, struct pipe_buffer *indexBuffer, unsigned indexSize, unsigned mode, unsigned start, unsigned count) { - return softpipe_draw_range_elements( pipe, indexBuffer, - indexSize, - 0, 0xffffffff, - mode, start, count ); + softpipe_draw_range_elements( pipe, indexBuffer, + indexSize, + 0, 0xffffffff, + mode, start, count ); } diff --git a/src/gallium/drivers/softpipe/sp_state.h b/src/gallium/drivers/softpipe/sp_state.h index d488fb8710..7a61422387 100644 --- a/src/gallium/drivers/softpipe/sp_state.h +++ b/src/gallium/drivers/softpipe/sp_state.h @@ -174,14 +174,14 @@ void softpipe_set_vertex_buffers(struct pipe_context *, void softpipe_update_derived( struct softpipe_context *softpipe ); -boolean softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, - unsigned start, unsigned count); - -boolean softpipe_draw_elements(struct pipe_context *pipe, - struct pipe_buffer *indexBuffer, - unsigned indexSize, - unsigned mode, unsigned start, unsigned count); -boolean +void softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, + unsigned start, unsigned count); + +void softpipe_draw_elements(struct pipe_context *pipe, + struct pipe_buffer *indexBuffer, + unsigned indexSize, + unsigned mode, unsigned start, unsigned count); +void softpipe_draw_range_elements(struct pipe_context *pipe, struct pipe_buffer *indexBuffer, unsigned indexSize, -- cgit v1.2.3 From d29f55546dec74ca77dce3a3bf581c251be1d397 Mon Sep 17 00:00:00 2001 From: Christoph Bumiller Date: Thu, 24 Dec 2009 12:39:42 +0100 Subject: nv50: make edgeflags work It doesn't seem to be possible to set the egdeflag in the vertex shader, so we need to fallback to pushing vertices through the FIFO and use method 0x15e4 if they are used. This only works if VP does MOV OUT[X] IN[Y] where X is the edgeflag output, and Y is saved so we can tell the correct input later. The VP still writes the useless values to wasted outputs as punishment. --- src/gallium/drivers/nv50/nv50_program.c | 19 +++++++++++--- src/gallium/drivers/nv50/nv50_program.h | 1 + src/gallium/drivers/nv50/nv50_screen.c | 3 +++ src/gallium/drivers/nv50/nv50_vbo.c | 45 ++++++++++++++++++++++++++++++++- 4 files changed, 64 insertions(+), 4 deletions(-) (limited to 'src/gallium/drivers/nv50/nv50_vbo.c') diff --git a/src/gallium/drivers/nv50/nv50_program.c b/src/gallium/drivers/nv50/nv50_program.c index 679c28ce4b..ce3fa5fc88 100644 --- a/src/gallium/drivers/nv50/nv50_program.c +++ b/src/gallium/drivers/nv50/nv50_program.c @@ -159,6 +159,8 @@ struct nv50_pc { unsigned insn_nr; boolean allow32; + + uint8_t edgeflag_out; }; static INLINE struct nv50_reg * @@ -2554,10 +2556,16 @@ prep_inspect_insn(struct nv50_pc *pc, const struct tgsi_full_instruction *insn) mask = dst->WriteMask; if (dst->File == TGSI_FILE_TEMPORARY) - reg = pc->temp; + reg = pc->temp; else - if (dst->File == TGSI_FILE_OUTPUT) - reg = pc->result; + if (dst->File == TGSI_FILE_OUTPUT) { + reg = pc->result; + + if (insn->Instruction.Opcode == TGSI_OPCODE_MOV && + dst->Index == pc->edgeflag_out && + insn->Src[0].Register.File == TGSI_FILE_INPUT) + pc->p->cfg.edgeflag_in = insn->Src[0].Register.Index; + } if (reg) { for (c = 0; c < 4; c++) { @@ -2856,6 +2864,9 @@ nv50_program_tx_prep(struct nv50_pc *pc) if (p->cfg.io_nr > first) p->cfg.io_nr = first; break; + case TGSI_SEMANTIC_EDGEFLAG: + pc->edgeflag_out = first; + break; /* case TGSI_SEMANTIC_CLIP_DISTANCE: p->cfg.clpd = MIN2(p->cfg.clpd, first); @@ -3104,6 +3115,8 @@ ctor_nv50_pc(struct nv50_pc *pc, struct nv50_program *p) p->cfg.two_side[0].hw = 0x40; p->cfg.two_side[1].hw = 0x40; + p->cfg.edgeflag_in = pc->edgeflag_out = 0xff; + switch (p->type) { case PIPE_SHADER_VERTEX: p->cfg.psiz = 0x40; diff --git a/src/gallium/drivers/nv50/nv50_program.h b/src/gallium/drivers/nv50/nv50_program.h index 4a90c372ce..461fec1d89 100644 --- a/src/gallium/drivers/nv50/nv50_program.h +++ b/src/gallium/drivers/nv50/nv50_program.h @@ -58,6 +58,7 @@ struct nv50_program { /* VP only */ uint8_t clpd, clpd_nr; uint8_t psiz; + uint8_t edgeflag_in; } cfg; }; diff --git a/src/gallium/drivers/nv50/nv50_screen.c b/src/gallium/drivers/nv50/nv50_screen.c index d443ca3ad0..2435f65ed2 100644 --- a/src/gallium/drivers/nv50/nv50_screen.c +++ b/src/gallium/drivers/nv50/nv50_screen.c @@ -441,6 +441,9 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) so_method(so, screen->tesla, NV50TCL_SCISSOR_ENABLE, 1); so_data (so, 1); + so_method(so, screen->tesla, 0x15e4, 1); + so_data (so, 1); /* default edgeflag to TRUE */ + so_emit(chan, so); so_ref (so, &screen->static_init); so_ref (NULL, &so); diff --git a/src/gallium/drivers/nv50/nv50_vbo.c b/src/gallium/drivers/nv50/nv50_vbo.c index f7fa0659e8..39324e30f6 100644 --- a/src/gallium/drivers/nv50/nv50_vbo.c +++ b/src/gallium/drivers/nv50/nv50_vbo.c @@ -372,6 +372,10 @@ nv50_vbo_static_attrib(struct nv50_context *nv50, unsigned attrib, so_data (so, fui(v[1])); break; case 1: + if (attrib == nv50->vertprog->cfg.edgeflag_in) { + so_method(so, tesla, 0x15e4, 1); + so_data (so, v[0] ? 1 : 0); + } so_method(so, tesla, NV50TCL_VTX_ATTR_1F(attrib), 1); so_data (so, fui(v[0])); break; @@ -401,6 +405,9 @@ nv50_vbo_validate(struct nv50_context *nv50) !(nv50->vtxbuf[i].buffer->usage & PIPE_BUFFER_USAGE_VERTEX)) nv50->vbo_fifo = 0xffff; + if (nv50->vertprog->cfg.edgeflag_in < 16) + nv50->vbo_fifo = 0xffff; /* vertprog can't set edgeflag */ + n_ve = MAX2(nv50->vtxelt_nr, nv50->state.vtxelt_nr); vtxattr = NULL; @@ -479,6 +486,9 @@ struct nv50_vbo_emitctx unsigned nr_ve; unsigned vtx_dwords; unsigned vtx_max; + + float edgeflag; + unsigned ve_edgeflag; }; static INLINE void @@ -622,6 +632,9 @@ emit_prepare(struct nv50_context *nv50, struct nv50_vbo_emitctx *emit, if (nv50_map_vbufs(nv50) == FALSE) return FALSE; + emit->ve_edgeflag = nv50->vertprog->cfg.edgeflag_in; + + emit->edgeflag = 0.5f; emit->nr_ve = 0; emit->vtx_dwords = 0; @@ -644,7 +657,8 @@ emit_prepare(struct nv50_context *nv50, struct nv50_vbo_emitctx *emit, desc = util_format_description(ve->src_format); assert(desc); - size = util_format_get_component_bits(ve->src_format, UTIL_FORMAT_COLORSPACE_RGB, 0); + size = util_format_get_component_bits( + ve->src_format, UTIL_FORMAT_COLORSPACE_RGB, 0); assert(ve->nr_components > 0 && ve->nr_components <= 4); @@ -686,10 +700,31 @@ emit_prepare(struct nv50_context *nv50, struct nv50_vbo_emitctx *emit, } emit->vtx_max = 512 / emit->vtx_dwords; + if (emit->ve_edgeflag < 16) + emit->vtx_max = 1; return TRUE; } +static INLINE void +set_edgeflag(struct nouveau_channel *chan, + struct nouveau_grobj *tesla, + struct nv50_vbo_emitctx *emit, uint32_t index) +{ + unsigned i = emit->ve_edgeflag; + + if (i < 16) { + float f = *((float *)(emit->map[i] + index * emit->stride[i])); + + if (emit->edgeflag != f) { + emit->edgeflag = f; + + BEGIN_RING(chan, tesla, 0x15e4, 1); + OUT_RING (chan, f ? 1 : 0); + } + } +} + static boolean nv50_push_arrays(struct nv50_context *nv50, unsigned start, unsigned count) { @@ -704,6 +739,8 @@ nv50_push_arrays(struct nv50_context *nv50, unsigned start, unsigned count) unsigned i, dw, nr = MIN2(count, emit.vtx_max); dw = nr * emit.vtx_dwords; + set_edgeflag(chan, tesla, &emit, 0); /* nr will be 1 */ + BEGIN_RING(chan, tesla, NV50TCL_VERTEX_DATA | 0x40000000, dw); for (i = 0; i < nr; ++i) emit_vtx_next(chan, &emit); @@ -729,6 +766,8 @@ nv50_push_elements_u32(struct nv50_context *nv50, uint32_t *map, unsigned count) unsigned i, dw, nr = MIN2(count, emit.vtx_max); dw = nr * emit.vtx_dwords; + set_edgeflag(chan, tesla, &emit, *map); + BEGIN_RING(chan, tesla, NV50TCL_VERTEX_DATA | 0x40000000, dw); for (i = 0; i < nr; ++i) emit_vtx(chan, &emit, *map++); @@ -754,6 +793,8 @@ nv50_push_elements_u16(struct nv50_context *nv50, uint16_t *map, unsigned count) unsigned i, dw, nr = MIN2(count, emit.vtx_max); dw = nr * emit.vtx_dwords; + set_edgeflag(chan, tesla, &emit, *map); + BEGIN_RING(chan, tesla, NV50TCL_VERTEX_DATA | 0x40000000, dw); for (i = 0; i < nr; ++i) emit_vtx(chan, &emit, *map++); @@ -779,6 +820,8 @@ nv50_push_elements_u08(struct nv50_context *nv50, uint8_t *map, unsigned count) unsigned i, dw, nr = MIN2(count, emit.vtx_max); dw = nr * emit.vtx_dwords; + set_edgeflag(chan, tesla, &emit, *map); + BEGIN_RING(chan, tesla, NV50TCL_VERTEX_DATA | 0x40000000, dw); for (i = 0; i < nr; ++i) emit_vtx(chan, &emit, *map++); -- cgit v1.2.3 From c84cc09d41a83caa96eca84c73284024d8d63024 Mon Sep 17 00:00:00 2001 From: Marcin Kościelnicki Date: Mon, 28 Dec 2009 16:23:40 +0000 Subject: nv50: Dehexify and bring up to date with new method defines. Signed-off-by: Francisco Jerez --- src/gallium/drivers/nv50/nv50_program.c | 4 +-- src/gallium/drivers/nv50/nv50_query.c | 4 +-- src/gallium/drivers/nv50/nv50_screen.c | 44 ++++++++++++-------------- src/gallium/drivers/nv50/nv50_state.c | 13 ++++---- src/gallium/drivers/nv50/nv50_state_validate.c | 20 ++++++------ src/gallium/drivers/nv50/nv50_surface.c | 6 ++-- src/gallium/drivers/nv50/nv50_transfer.c | 12 +++---- src/gallium/drivers/nv50/nv50_vbo.c | 38 +++++++++++----------- 8 files changed, 69 insertions(+), 72 deletions(-) (limited to 'src/gallium/drivers/nv50/nv50_vbo.c') diff --git a/src/gallium/drivers/nv50/nv50_program.c b/src/gallium/drivers/nv50/nv50_program.c index a101ac095c..b9910b430a 100644 --- a/src/gallium/drivers/nv50/nv50_program.c +++ b/src/gallium/drivers/nv50/nv50_program.c @@ -3498,7 +3498,7 @@ nv50_fragprog_validate(struct nv50_context *nv50) so_data (so, p->cfg.high_temp); so_method(so, tesla, NV50TCL_FP_RESULT_COUNT, 1); so_data (so, p->cfg.high_result); - so_method(so, tesla, NV50TCL_FP_CTRL_UNK19A8, 1); + so_method(so, tesla, NV50TCL_FP_CONTROL, 1); so_data (so, p->cfg.regs[2]); so_method(so, tesla, NV50TCL_FP_CTRL_UNK196C, 1); so_data (so, p->cfg.regs[3]); @@ -3670,7 +3670,7 @@ nv50_linkage_validate(struct nv50_context *nv50) so_method(so, tesla, NV50TCL_FP_INTERPOLANT_CTRL, 1); so_data (so, reg[4]); - so_method(so, tesla, 0x1540, 4); + so_method(so, tesla, NV50TCL_NOPERSPECTIVE_BITMAP(0), 4); so_datap (so, lin, 4); if (nv50->rasterizer->pipe.point_sprite) { diff --git a/src/gallium/drivers/nv50/nv50_query.c b/src/gallium/drivers/nv50/nv50_query.c index 268c9823f7..5d9e18218a 100644 --- a/src/gallium/drivers/nv50/nv50_query.c +++ b/src/gallium/drivers/nv50/nv50_query.c @@ -77,9 +77,9 @@ nv50_query_begin(struct pipe_context *pipe, struct pipe_query *pq) struct nouveau_grobj *tesla = nv50->screen->tesla; struct nv50_query *q = nv50_query(pq); - BEGIN_RING(chan, tesla, 0x1530, 1); + BEGIN_RING(chan, tesla, NV50TCL_SAMPLECNT_RESET, 1); OUT_RING (chan, 1); - BEGIN_RING(chan, tesla, 0x1514, 1); + BEGIN_RING(chan, tesla, NV50TCL_SAMPLECNT_ENABLE, 1); OUT_RING (chan, 1); q->ready = FALSE; diff --git a/src/gallium/drivers/nv50/nv50_screen.c b/src/gallium/drivers/nv50/nv50_screen.c index 5a1efd3998..15e4b6e5ca 100644 --- a/src/gallium/drivers/nv50/nv50_screen.c +++ b/src/gallium/drivers/nv50/nv50_screen.c @@ -231,8 +231,7 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) break; case 0x80: case 0x90: - /* this stupid name should be corrected. */ - tesla_class = NV54TCL; + tesla_class = NV84TCL; break; case 0xa0: switch (chipset) { @@ -242,7 +241,7 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) tesla_class = NVA0TCL; break; default: - tesla_class = 0x8597; + tesla_class = NVA8TCL; break; } break; @@ -287,7 +286,7 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) so_data (so, chan->vram->handle); so_method(so, screen->eng2d, NV50_2D_OPERATION, 1); so_data (so, NV50_2D_OPERATION_SRCCOPY); - so_method(so, screen->eng2d, 0x0290, 1); + so_method(so, screen->eng2d, NV50_2D_CLIP_ENABLE, 1); so_data (so, 0); so_method(so, screen->eng2d, 0x0888, 1); so_data (so, 1); @@ -297,34 +296,33 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) /* Static tesla init */ so = so_new(256, 20); - so_method(so, screen->tesla, 0x1558, 1); - so_data (so, 1); + so_method(so, screen->tesla, NV50TCL_COND_MODE, 1); + so_data (so, NV50TCL_COND_MODE_ALWAYS); so_method(so, screen->tesla, NV50TCL_DMA_NOTIFY, 1); so_data (so, screen->sync->handle); - so_method(so, screen->tesla, NV50TCL_DMA_UNK0(0), - NV50TCL_DMA_UNK0__SIZE); - for (i = 0; i < NV50TCL_DMA_UNK0__SIZE; i++) + so_method(so, screen->tesla, NV50TCL_DMA_ZETA, 11); + for (i = 0; i < 11; i++) so_data(so, chan->vram->handle); - so_method(so, screen->tesla, NV50TCL_DMA_UNK1(0), - NV50TCL_DMA_UNK1__SIZE); - for (i = 0; i < NV50TCL_DMA_UNK1__SIZE; i++) + so_method(so, screen->tesla, NV50TCL_DMA_COLOR(0), + NV50TCL_DMA_COLOR__SIZE); + for (i = 0; i < NV50TCL_DMA_COLOR__SIZE; i++) so_data(so, chan->vram->handle); - so_method(so, screen->tesla, 0x121c, 1); + so_method(so, screen->tesla, NV50TCL_RT_CONTROL, 1); so_data (so, 1); /* activate all 32 lanes (threads) in a warp */ - so_method(so, screen->tesla, 0x19a0, 1); + so_method(so, screen->tesla, NV50TCL_WARP_HALVES, 1); so_data (so, 0x2); so_method(so, screen->tesla, 0x1400, 1); so_data (so, 0xf); /* max TIC (bits 4:8) & TSC (ignored) bindings, per program type */ - so_method(so, screen->tesla, 0x13b4, 1); + so_method(so, screen->tesla, NV50TCL_TEX_LIMITS(0), 1); so_data (so, 0x54); - so_method(so, screen->tesla, 0x13bc, 1); + so_method(so, screen->tesla, NV50TCL_TEX_LIMITS(2), 1); so_data (so, 0x54); /* origin is top left (set to 1 for bottom left) */ - so_method(so, screen->tesla, 0x13ac, 1); + so_method(so, screen->tesla, NV50TCL_Y_ORIGIN_BOTTOM, 1); so_data (so, 0); so_method(so, screen->tesla, NV50TCL_VP_REG_ALLOC_RESULT, 1); so_data (so, 8); @@ -360,7 +358,7 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) // B = buffer ID (maybe more than 1 byte) // N = CB index used in shader instruction // P = program type (0 = VP, 2 = GP, 3 = FP) - so_method(so, screen->tesla, 0x1694, 1); + so_method(so, screen->tesla, NV50TCL_SET_PROGRAM_CB, 1); so_data (so, 0x000BBNP1); */ @@ -424,24 +422,24 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) /* Vertex array limits - max them out */ for (i = 0; i < 16; i++) { - so_method(so, screen->tesla, NV50TCL_UNK1080_OFFSET_HIGH(i), 2); + so_method(so, screen->tesla, NV50TCL_VERTEX_ARRAY_LIMIT_HIGH(i), 2); so_data (so, 0x000000ff); so_data (so, 0xffffffff); } - so_method(so, screen->tesla, NV50TCL_DEPTH_RANGE_NEAR, 2); + so_method(so, screen->tesla, NV50TCL_DEPTH_RANGE_NEAR(0), 2); so_data (so, fui(0.0)); so_data (so, fui(1.0)); /* no dynamic combination of TIC & TSC entries => only BIND_TIC used */ - so_method(so, screen->tesla, 0x1234, 1); + so_method(so, screen->tesla, NV50TCL_LINKED_TSC, 1); so_data (so, 1); /* activate first scissor rectangle */ - so_method(so, screen->tesla, NV50TCL_SCISSOR_ENABLE, 1); + so_method(so, screen->tesla, NV50TCL_SCISSOR_ENABLE(0), 1); so_data (so, 1); - so_method(so, screen->tesla, 0x15e4, 1); + so_method(so, screen->tesla, NV50TCL_EDGEFLAG_ENABLE, 1); so_data (so, 1); /* default edgeflag to TRUE */ so_emit(chan, so); diff --git a/src/gallium/drivers/nv50/nv50_state.c b/src/gallium/drivers/nv50/nv50_state.c index 88aef52d08..18a2b819d8 100644 --- a/src/gallium/drivers/nv50/nv50_state.c +++ b/src/gallium/drivers/nv50/nv50_state.c @@ -295,7 +295,7 @@ nv50_rasterizer_state_create(struct pipe_context *pipe, so_method(so, tesla, NV50TCL_SHADE_MODEL, 1); so_data (so, cso->flatshade ? NV50TCL_SHADE_MODEL_FLAT : NV50TCL_SHADE_MODEL_SMOOTH); - so_method(so, tesla, 0x1684, 1); + so_method(so, tesla, NV50TCL_PROVOKING_VERTEX_LAST, 1); so_data (so, cso->flatshade_first ? 0 : 1); so_method(so, tesla, NV50TCL_VERTEX_TWO_SIDE_ENABLE, 1); @@ -439,9 +439,8 @@ nv50_depth_stencil_alpha_state_create(struct pipe_context *pipe, so_data (so, 0); } - /* XXX: keep hex values until header is updated (names reversed) */ if (cso->stencil[0].enabled) { - so_method(so, tesla, 0x1380, 8); + so_method(so, tesla, NV50TCL_STENCIL_FRONT_ENABLE, 8); so_data (so, 1); so_data (so, nvgl_stencil_op(cso->stencil[0].fail_op)); so_data (so, nvgl_stencil_op(cso->stencil[0].zfail_op)); @@ -451,23 +450,23 @@ nv50_depth_stencil_alpha_state_create(struct pipe_context *pipe, so_data (so, cso->stencil[0].writemask); so_data (so, cso->stencil[0].valuemask); } else { - so_method(so, tesla, 0x1380, 1); + so_method(so, tesla, NV50TCL_STENCIL_FRONT_ENABLE, 1); so_data (so, 0); } if (cso->stencil[1].enabled) { - so_method(so, tesla, 0x1594, 5); + so_method(so, tesla, NV50TCL_STENCIL_BACK_ENABLE, 5); so_data (so, 1); so_data (so, nvgl_stencil_op(cso->stencil[1].fail_op)); so_data (so, nvgl_stencil_op(cso->stencil[1].zfail_op)); so_data (so, nvgl_stencil_op(cso->stencil[1].zpass_op)); so_data (so, nvgl_comparison_op(cso->stencil[1].func)); - so_method(so, tesla, 0x0f54, 3); + so_method(so, tesla, NV50TCL_STENCIL_BACK_FUNC_REF, 3); so_data (so, cso->stencil[1].ref_value); so_data (so, cso->stencil[1].writemask); so_data (so, cso->stencil[1].valuemask); } else { - so_method(so, tesla, 0x1594, 1); + so_method(so, tesla, NV50TCL_STENCIL_BACK_ENABLE, 1); so_data (so, 0); } diff --git a/src/gallium/drivers/nv50/nv50_state_validate.c b/src/gallium/drivers/nv50/nv50_state_validate.c index 871e8097b6..c8bdf9dc27 100644 --- a/src/gallium/drivers/nv50/nv50_state_validate.c +++ b/src/gallium/drivers/nv50/nv50_state_validate.c @@ -41,7 +41,7 @@ nv50_state_validate_fb(struct nv50_context *nv50) * FP result 0 always goes to RT[0], bits 4 - 6 are ignored. * Ambiguous assignment results in no rendering (no DATA_ERROR). */ - so_method(so, tesla, 0x121c, 1); + so_method(so, tesla, NV50TCL_RT_CONTROL, 1); so_data (so, fb->nr_cbufs | (0 << 4) | (1 << 7) | (2 << 10) | (3 << 13) | (4 << 16) | (5 << 19) | (6 << 22) | (7 << 25)); @@ -87,7 +87,7 @@ nv50_state_validate_fb(struct nv50_context *nv50) level[fb->cbufs[i]->level].tile_mode << 4); so_data(so, 0x00000000); - so_method(so, tesla, 0x1224, 1); + so_method(so, tesla, NV50TCL_RT_ARRAY_MODE, 1); so_data (so, 1); } @@ -124,22 +124,22 @@ nv50_state_validate_fb(struct nv50_context *nv50) level[fb->zsbuf->level].tile_mode << 4); so_data(so, 0x00000000); - so_method(so, tesla, 0x1538, 1); + so_method(so, tesla, NV50TCL_ZETA_ENABLE, 1); so_data (so, 1); so_method(so, tesla, NV50TCL_ZETA_HORIZ, 3); so_data (so, fb->zsbuf->width); so_data (so, fb->zsbuf->height); so_data (so, 0x00010001); } else { - so_method(so, tesla, 0x1538, 1); + so_method(so, tesla, NV50TCL_ZETA_ENABLE, 1); so_data (so, 0); } - so_method(so, tesla, NV50TCL_VIEWPORT_HORIZ, 2); + so_method(so, tesla, NV50TCL_VIEWPORT_HORIZ(0), 2); so_data (so, w << 16); so_data (so, h << 16); /* set window lower left corner */ - so_method(so, tesla, NV50TCL_WINDOW_LEFT, 2); + so_method(so, tesla, NV50TCL_WINDOW_OFFSET_X, 2); so_data (so, 0); so_data (so, 0); /* set screen scissor rectangle */ @@ -325,7 +325,7 @@ nv50_state_validate(struct nv50_context *nv50) nv50->state.scissor_enabled = rast->scissor; so = so_new(3, 0); - so_method(so, tesla, NV50TCL_SCISSOR_HORIZ, 2); + so_method(so, tesla, NV50TCL_SCISSOR_HORIZ(0), 2); if (nv50->state.scissor_enabled) { so_data(so, (s->maxx << 16) | s->minx); so_data(so, (s->maxy << 16) | s->miny); @@ -355,11 +355,11 @@ scissor_uptodate: so = so_new(14, 0); if (!bypass) { - so_method(so, tesla, NV50TCL_VIEWPORT_TRANSLATE(0), 3); + so_method(so, tesla, NV50TCL_VIEWPORT_TRANSLATE_X(0), 3); so_data (so, fui(nv50->viewport.translate[0])); so_data (so, fui(nv50->viewport.translate[1])); so_data (so, fui(nv50->viewport.translate[2])); - so_method(so, tesla, NV50TCL_VIEWPORT_SCALE(0), 3); + so_method(so, tesla, NV50TCL_VIEWPORT_SCALE_X(0), 3); so_data (so, fui(nv50->viewport.scale[0])); so_data (so, fui(nv50->viewport.scale[1])); so_data (so, fui(nv50->viewport.scale[2])); @@ -440,7 +440,7 @@ void nv50_so_init_sifc(struct nv50_context *nv50, so_data (so, 1); so_reloc (so, bo, offset, reloc | NOUVEAU_BO_HIGH, 0, 0); so_reloc (so, bo, offset, reloc | NOUVEAU_BO_LOW, 0, 0); - so_method(so, eng2d, NV50_2D_SIFC_UNK0800, 2); + so_method(so, eng2d, NV50_2D_SIFC_BITMAP_ENABLE, 2); so_data (so, 0); so_data (so, NV50_2D_SIFC_FORMAT_R8_UNORM); so_method(so, eng2d, NV50_2D_SIFC_WIDTH, 10); diff --git a/src/gallium/drivers/nv50/nv50_surface.c b/src/gallium/drivers/nv50/nv50_surface.c index 79655fc08d..6378132979 100644 --- a/src/gallium/drivers/nv50/nv50_surface.c +++ b/src/gallium/drivers/nv50/nv50_surface.c @@ -176,11 +176,11 @@ nv50_surface_fill(struct pipe_context *pipe, struct pipe_surface *dest, if (ret) return; - BEGIN_RING(chan, eng2d, 0x0580, 3); - OUT_RING (chan, 4); + BEGIN_RING(chan, eng2d, NV50_2D_DRAW_SHAPE, 3); + OUT_RING (chan, NV50_2D_DRAW_SHAPE_RECTANGLES); OUT_RING (chan, format); OUT_RING (chan, value); - BEGIN_RING(chan, eng2d, NV50_2D_RECT_X1, 4); + BEGIN_RING(chan, eng2d, NV50_2D_DRAW_POINT32_X(0), 4); OUT_RING (chan, destx); OUT_RING (chan, desty); OUT_RING (chan, width); diff --git a/src/gallium/drivers/nv50/nv50_transfer.c b/src/gallium/drivers/nv50/nv50_transfer.c index 4d9afa6fed..a2f1db2914 100644 --- a/src/gallium/drivers/nv50/nv50_transfer.c +++ b/src/gallium/drivers/nv50/nv50_transfer.c @@ -47,7 +47,7 @@ nv50_transfer_rect_m2mf(struct pipe_screen *pscreen, NV50_MEMORY_TO_MEMORY_FORMAT_LINEAR_IN, 1); OUT_RING (chan, 1); BEGIN_RING(chan, m2mf, - NV50_MEMORY_TO_MEMORY_FORMAT_PITCH_IN, 1); + NV04_MEMORY_TO_MEMORY_FORMAT_PITCH_IN, 1); OUT_RING (chan, src_pitch); src_offset += (sy * src_pitch) + (sx * cpp); } else { @@ -66,7 +66,7 @@ nv50_transfer_rect_m2mf(struct pipe_screen *pscreen, NV50_MEMORY_TO_MEMORY_FORMAT_LINEAR_OUT, 1); OUT_RING (chan, 1); BEGIN_RING(chan, m2mf, - NV50_MEMORY_TO_MEMORY_FORMAT_PITCH_OUT, 1); + NV04_MEMORY_TO_MEMORY_FORMAT_PITCH_OUT, 1); OUT_RING (chan, dst_pitch); dst_offset += (dy * dst_pitch) + (dx * cpp); } else { @@ -89,7 +89,7 @@ nv50_transfer_rect_m2mf(struct pipe_screen *pscreen, OUT_RELOCh(chan, src_bo, src_offset, src_reloc); OUT_RELOCh(chan, dst_bo, dst_offset, dst_reloc); BEGIN_RING(chan, m2mf, - NV50_MEMORY_TO_MEMORY_FORMAT_OFFSET_IN, 2); + NV04_MEMORY_TO_MEMORY_FORMAT_OFFSET_IN, 2); OUT_RELOCl(chan, src_bo, src_offset, src_reloc); OUT_RELOCl(chan, dst_bo, dst_offset, dst_reloc); if (src_bo->tile_flags) { @@ -107,7 +107,7 @@ nv50_transfer_rect_m2mf(struct pipe_screen *pscreen, dst_offset += (line_count * dst_pitch); } BEGIN_RING(chan, m2mf, - NV50_MEMORY_TO_MEMORY_FORMAT_LINE_LENGTH_IN, 4); + NV04_MEMORY_TO_MEMORY_FORMAT_LINE_LENGTH_IN, 4); OUT_RING (chan, width * cpp); OUT_RING (chan, line_count); OUT_RING (chan, 0x00000101); @@ -291,7 +291,7 @@ nv50_upload_sifc(struct nv50_context *nv50, /* NV50_2D_OPERATION_SRCCOPY assumed already set */ - BEGIN_RING(chan, eng2d, NV50_2D_SIFC_UNK0800, 2); + BEGIN_RING(chan, eng2d, NV50_2D_SIFC_BITMAP_ENABLE, 2); OUT_RING (chan, 0); OUT_RING (chan, src_format); BEGIN_RING(chan, eng2d, NV50_2D_SIFC_WIDTH, 10); @@ -334,6 +334,6 @@ nv50_upload_sifc(struct nv50_context *nv50, src += src_pitch; } - BEGIN_RING(chan, tesla, 0x1440, 1); + BEGIN_RING(chan, tesla, NV50TCL_CODE_CB_FLUSH, 1); OUT_RING (chan, 0); } diff --git a/src/gallium/drivers/nv50/nv50_vbo.c b/src/gallium/drivers/nv50/nv50_vbo.c index 39324e30f6..602adfc50d 100644 --- a/src/gallium/drivers/nv50/nv50_vbo.c +++ b/src/gallium/drivers/nv50/nv50_vbo.c @@ -99,19 +99,19 @@ nv50_vbo_size_to_hw(unsigned size, unsigned nr_c) { static const uint32_t hw_values[] = { 0, 0, 0, 0, - NV50TCL_VERTEX_ARRAY_ATTRIB_SIZE_8, - NV50TCL_VERTEX_ARRAY_ATTRIB_SIZE_8_8, - NV50TCL_VERTEX_ARRAY_ATTRIB_SIZE_8_8_8, - NV50TCL_VERTEX_ARRAY_ATTRIB_SIZE_8_8_8_8, - NV50TCL_VERTEX_ARRAY_ATTRIB_SIZE_16, - NV50TCL_VERTEX_ARRAY_ATTRIB_SIZE_16_16, - NV50TCL_VERTEX_ARRAY_ATTRIB_SIZE_16_16_16, - NV50TCL_VERTEX_ARRAY_ATTRIB_SIZE_16_16_16_16, + NV50TCL_VERTEX_ARRAY_ATTRIB_FORMAT_8, + NV50TCL_VERTEX_ARRAY_ATTRIB_FORMAT_8_8, + NV50TCL_VERTEX_ARRAY_ATTRIB_FORMAT_8_8_8, + NV50TCL_VERTEX_ARRAY_ATTRIB_FORMAT_8_8_8_8, + NV50TCL_VERTEX_ARRAY_ATTRIB_FORMAT_16, + NV50TCL_VERTEX_ARRAY_ATTRIB_FORMAT_16_16, + NV50TCL_VERTEX_ARRAY_ATTRIB_FORMAT_16_16_16, + NV50TCL_VERTEX_ARRAY_ATTRIB_FORMAT_16_16_16_16, 0, 0, 0, 0, - NV50TCL_VERTEX_ARRAY_ATTRIB_SIZE_32, - NV50TCL_VERTEX_ARRAY_ATTRIB_SIZE_32_32, - NV50TCL_VERTEX_ARRAY_ATTRIB_SIZE_32_32_32, - NV50TCL_VERTEX_ARRAY_ATTRIB_SIZE_32_32_32_32 }; + NV50TCL_VERTEX_ARRAY_ATTRIB_FORMAT_32, + NV50TCL_VERTEX_ARRAY_ATTRIB_FORMAT_32_32, + NV50TCL_VERTEX_ARRAY_ATTRIB_FORMAT_32_32_32, + NV50TCL_VERTEX_ARRAY_ATTRIB_FORMAT_32_32_32_32 }; /* we'd also have R11G11B10 and R10G10B10A2 */ @@ -198,7 +198,7 @@ nv50_draw_elements_inline_u08(struct nv50_context *nv50, uint8_t *map, return nv50_push_elements_u08(nv50, map, count); if (count & 1) { - BEGIN_RING(chan, tesla, 0x15e8, 1); + BEGIN_RING(chan, tesla, NV50TCL_VB_ELEMENT_U32, 1); OUT_RING (chan, map[0]); map++; count--; @@ -208,7 +208,7 @@ nv50_draw_elements_inline_u08(struct nv50_context *nv50, uint8_t *map, unsigned nr = count > 2046 ? 2046 : count; int i; - BEGIN_RING(chan, tesla, 0x400015f0, nr >> 1); + BEGIN_RING(chan, tesla, NV50TCL_VB_ELEMENT_U16 | 0x40000000, nr >> 1); for (i = 0; i < nr; i += 2) OUT_RING (chan, (map[i + 1] << 16) | map[i]); @@ -231,7 +231,7 @@ nv50_draw_elements_inline_u16(struct nv50_context *nv50, uint16_t *map, return nv50_push_elements_u16(nv50, map, count); if (count & 1) { - BEGIN_RING(chan, tesla, 0x15e8, 1); + BEGIN_RING(chan, tesla, NV50TCL_VB_ELEMENT_U32, 1); OUT_RING (chan, map[0]); map++; count--; @@ -241,7 +241,7 @@ nv50_draw_elements_inline_u16(struct nv50_context *nv50, uint16_t *map, unsigned nr = count > 2046 ? 2046 : count; int i; - BEGIN_RING(chan, tesla, 0x400015f0, nr >> 1); + BEGIN_RING(chan, tesla, NV50TCL_VB_ELEMENT_U16 | 0x40000000, nr >> 1); for (i = 0; i < nr; i += 2) OUT_RING (chan, (map[i + 1] << 16) | map[i]); @@ -266,7 +266,7 @@ nv50_draw_elements_inline_u32(struct nv50_context *nv50, uint32_t *map, while (count) { unsigned nr = count > 2047 ? 2047 : count; - BEGIN_RING(chan, tesla, 0x400015e8, nr); + BEGIN_RING(chan, tesla, NV50TCL_VB_ELEMENT_U32 | 0x40000000, nr); OUT_RINGp (chan, map, nr); count -= nr; @@ -373,7 +373,7 @@ nv50_vbo_static_attrib(struct nv50_context *nv50, unsigned attrib, break; case 1: if (attrib == nv50->vertprog->cfg.edgeflag_in) { - so_method(so, tesla, 0x15e4, 1); + so_method(so, tesla, NV50TCL_EDGEFLAG_ENABLE, 1); so_data (so, v[0] ? 1 : 0); } so_method(so, tesla, NV50TCL_VTX_ATTR_1F(attrib), 1); @@ -452,7 +452,7 @@ nv50_vbo_validate(struct nv50_context *nv50) NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0); /* vertex array limits */ - so_method(vtxbuf, tesla, 0x1080 + (i * 8), 2); + so_method(vtxbuf, tesla, NV50TCL_VERTEX_ARRAY_LIMIT_HIGH(i), 2); so_reloc (vtxbuf, bo, vb->buffer->size - 1, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0); -- cgit v1.2.3 From c77ade8fed2be933af3f493932cedee7ca868b04 Mon Sep 17 00:00:00 2001 From: Maarten Maathuis Date: Tue, 29 Dec 2009 23:59:08 +0100 Subject: nouveau: rewrite nouveau_stateobj to use BEGIN_RING properly - The previous solution was hacky and didn't do subchannel autobinding. - The beheaviour should match what libdrm_nouveau does closely. - The solution remains statically sized, but when debugging is on it will check for abuse. Signed-off-by: Maarten Maathuis --- src/gallium/drivers/nouveau/nouveau_stateobj.h | 288 ++++++++++++++++++------- src/gallium/drivers/nv10/nv10_screen.c | 1 - src/gallium/drivers/nv20/nv20_screen.c | 1 - src/gallium/drivers/nv30/nv30_fragprog.c | 2 +- src/gallium/drivers/nv30/nv30_fragtex.c | 4 +- src/gallium/drivers/nv30/nv30_screen.c | 3 +- src/gallium/drivers/nv30/nv30_state.c | 6 +- src/gallium/drivers/nv30/nv30_state_blend.c | 2 +- src/gallium/drivers/nv30/nv30_state_fb.c | 2 +- src/gallium/drivers/nv30/nv30_state_scissor.c | 2 +- src/gallium/drivers/nv30/nv30_state_stipple.c | 4 +- src/gallium/drivers/nv30/nv30_state_viewport.c | 2 +- src/gallium/drivers/nv30/nv30_vbo.c | 6 +- src/gallium/drivers/nv30/nv30_vertprog.c | 2 +- src/gallium/drivers/nv40/nv40_fragprog.c | 2 +- src/gallium/drivers/nv40/nv40_fragtex.c | 4 +- src/gallium/drivers/nv40/nv40_screen.c | 3 +- src/gallium/drivers/nv40/nv40_state.c | 6 +- src/gallium/drivers/nv40/nv40_state_blend.c | 2 +- src/gallium/drivers/nv40/nv40_state_fb.c | 2 +- src/gallium/drivers/nv40/nv40_state_scissor.c | 2 +- src/gallium/drivers/nv40/nv40_state_stipple.c | 4 +- src/gallium/drivers/nv40/nv40_state_viewport.c | 2 +- src/gallium/drivers/nv40/nv40_vbo.c | 6 +- src/gallium/drivers/nv40/nv40_vertprog.c | 2 +- src/gallium/drivers/nv50/nv50_program.c | 6 +- src/gallium/drivers/nv50/nv50_screen.c | 9 +- src/gallium/drivers/nv50/nv50_state.c | 6 +- src/gallium/drivers/nv50/nv50_state_validate.c | 13 +- src/gallium/drivers/nv50/nv50_tex.c | 10 +- src/gallium/drivers/nv50/nv50_vbo.c | 6 +- 31 files changed, 265 insertions(+), 145 deletions(-) (limited to 'src/gallium/drivers/nv50/nv50_vbo.c') diff --git a/src/gallium/drivers/nouveau/nouveau_stateobj.h b/src/gallium/drivers/nouveau/nouveau_stateobj.h index 77ff7dcf20..e844f6abb3 100644 --- a/src/gallium/drivers/nouveau/nouveau_stateobj.h +++ b/src/gallium/drivers/nouveau/nouveau_stateobj.h @@ -3,41 +3,95 @@ #include "util/u_debug.h" +#ifdef DEBUG +#define DEBUG_NOUVEAU_STATEOBJ +#endif /* DEBUG */ + struct nouveau_stateobj_reloc { struct nouveau_bo *bo; - unsigned offset; - unsigned packet; + struct nouveau_grobj *gr; + uint32_t push_offset; + uint32_t mthd; - unsigned data; + uint32_t data; unsigned flags; unsigned vor; unsigned tor; }; +struct nouveau_stateobj_start { + struct nouveau_grobj *gr; + uint32_t mthd; + uint32_t size; + unsigned offset; +}; + struct nouveau_stateobj { struct pipe_reference reference; - unsigned *push; + struct nouveau_stateobj_start *start; struct nouveau_stateobj_reloc *reloc; - unsigned *cur; - unsigned cur_packet; + /* Common memory pool for data. */ + uint32_t *pool; + unsigned pool_cur; + +#ifdef DEBUG_NOUVEAU_STATEOBJ + unsigned start_alloc; + unsigned reloc_alloc; + unsigned pool_alloc; +#endif /* DEBUG_NOUVEAU_STATEOBJ */ + + unsigned total; /* includes begin_ring */ + unsigned cur; /* excludes begin_ring, offset from "cur_start" */ + unsigned cur_start; unsigned cur_reloc; }; +static INLINE void +so_dump(struct nouveau_stateobj *so) +{ + unsigned i, nr, total = 0; + + for (i = 0; i < so->cur_start; i++) { + if (so->start[i].gr->subc > -1) + debug_printf("+0x%04x: 0x%08x\n", total++, + (so->start[i].size << 18) | (so->start[i].gr->subc << 13) + | so->start[i].mthd); + else + debug_printf("+0x%04x: 0x%08x\n", total++, + (so->start[i].size << 18) | so->start[i].mthd); + for (nr = 0; nr < so->start[i].size; nr++, total++) + debug_printf("+0x%04x: 0x%08x\n", total, + so->pool[so->start[i].offset + nr]); + } +} + static INLINE struct nouveau_stateobj * -so_new(unsigned push, unsigned reloc) +so_new(unsigned start, unsigned push, unsigned reloc) { struct nouveau_stateobj *so; so = MALLOC(sizeof(struct nouveau_stateobj)); pipe_reference_init(&so->reference, 1); - so->push = MALLOC(sizeof(unsigned) * push); - so->reloc = MALLOC(sizeof(struct nouveau_stateobj_reloc) * reloc); - - so->cur = so->push; - so->cur_reloc = so->cur_packet = 0; + so->total = so->cur = so->cur_start = so->cur_reloc = 0; + +#ifdef DEBUG_NOUVEAU_STATEOBJ + so->start_alloc = start; + so->reloc_alloc = reloc; + so->pool_alloc = push; +#endif /* DEBUG_NOUVEAU_STATEOBJ */ + + so->start = MALLOC(start * sizeof(struct nouveau_stateobj_start)); + so->reloc = MALLOC(reloc * sizeof(struct nouveau_stateobj_reloc)); + so->pool = MALLOC(push * sizeof(uint32_t)); + so->pool_cur = 0; + + if (!so->start || !so->reloc || !so->pool) { + debug_printf("malloc failed\n"); + assert(0); + } return so; } @@ -48,54 +102,115 @@ so_ref(struct nouveau_stateobj *ref, struct nouveau_stateobj **pso) struct nouveau_stateobj *so = *pso; int i; - if (pipe_reference(&(*pso)->reference, &ref->reference)) { - free(so->push); + if (pipe_reference(&(*pso)->reference, &ref->reference)) { + FREE(so->start); for (i = 0; i < so->cur_reloc; i++) nouveau_bo_ref(NULL, &so->reloc[i].bo); - free(so->reloc); - free(so); + FREE(so->reloc); + FREE(so->pool); + FREE(so); } *pso = ref; } static INLINE void -so_data(struct nouveau_stateobj *so, unsigned data) +so_data(struct nouveau_stateobj *so, uint32_t data) { - (*so->cur++) = (data); - so->cur_packet += 4; +#ifdef DEBUG_NOUVEAU_STATEOBJ + if (so->cur >= so->start[so->cur_start - 1].size) { + debug_printf("exceeding specified size\n"); + assert(0); + } +#endif /* DEBUG_NOUVEAU_STATEOBJ */ + + so->pool[so->start[so->cur_start - 1].offset + so->cur++] = data; } static INLINE void -so_datap(struct nouveau_stateobj *so, unsigned *data, unsigned size) +so_datap(struct nouveau_stateobj *so, uint32_t *data, unsigned size) { - so->cur_packet += (4 * size); +#ifdef DEBUG_NOUVEAU_STATEOBJ + if ((so->cur + size) > so->start[so->cur_start - 1].size) { + debug_printf("exceeding specified size\n"); + assert(0); + } +#endif /* DEBUG_NOUVEAU_STATEOBJ */ + while (size--) - (*so->cur++) = (*data++); + so->pool[so->start[so->cur_start - 1].offset + so->cur++] = + *data++; } static INLINE void so_method(struct nouveau_stateobj *so, struct nouveau_grobj *gr, unsigned mthd, unsigned size) { - so->cur_packet = (gr->subc << 13) | (1 << 18) | (mthd - 4); - so_data(so, (gr->subc << 13) | (size << 18) | mthd); + struct nouveau_stateobj_start *start; + +#ifdef DEBUG_NOUVEAU_STATEOBJ + if (so->start_alloc <= so->cur_start) { + debug_printf("exceeding num_start size\n"); + assert(0); + } else +#endif /* DEBUG_NOUVEAU_STATEOBJ */ + start = so->start; + +#ifdef DEBUG_NOUVEAU_STATEOBJ + if (so->cur_start > 0 && start[so->cur_start - 1].size > so->cur) { + debug_printf("previous so_method was not filled\n"); + assert(0); + } +#endif /* DEBUG_NOUVEAU_STATEOBJ */ + + so->start = start; + start[so->cur_start].gr = gr; + start[so->cur_start].mthd = mthd; + start[so->cur_start].size = size; + +#ifdef DEBUG_NOUVEAU_STATEOBJ + if (so->pool_alloc < (size + so->pool_cur)) { + debug_printf("exceeding num_pool size\n"); + assert(0); + } +#endif /* DEBUG_NOUVEAU_STATEOBJ */ + + start[so->cur_start].offset = so->pool_cur; + so->pool_cur += size; + + so->cur_start++; + /* The 1 is for *this* begin_ring. */ + so->total += so->cur + 1; + so->cur = 0; } static INLINE void so_reloc(struct nouveau_stateobj *so, struct nouveau_bo *bo, unsigned data, unsigned flags, unsigned vor, unsigned tor) { - struct nouveau_stateobj_reloc *r = &so->reloc[so->cur_reloc++]; - - r->bo = NULL; - nouveau_bo_ref(bo, &r->bo); - r->offset = so->cur - so->push; - r->packet = so->cur_packet; - r->data = data; - r->flags = flags; - r->vor = vor; - r->tor = tor; + struct nouveau_stateobj_reloc *r; + +#ifdef DEBUG_NOUVEAU_STATEOBJ + if (so->reloc_alloc <= so->cur_reloc) { + debug_printf("exceeding num_reloc size\n"); + assert(0); + } else +#endif /* DEBUG_NOUVEAU_STATEOBJ */ + r = so->reloc; + + so->reloc = r; + r[so->cur_reloc].bo = NULL; + nouveau_bo_ref(bo, &(r[so->cur_reloc].bo)); + r[so->cur_reloc].gr = so->start[so->cur_start-1].gr; + r[so->cur_reloc].push_offset = so->total + so->cur; + r[so->cur_reloc].data = data; + r[so->cur_reloc].flags = flags; + r[so->cur_reloc].mthd = so->start[so->cur_start-1].mthd + + (so->cur << 2); + r[so->cur_reloc].vor = vor; + r[so->cur_reloc].tor = tor; + so_data(so, data); + so->cur_reloc++; } /* Determine if this buffer object is referenced by this state object. */ @@ -111,15 +226,6 @@ so_bo_is_reloc(struct nouveau_stateobj *so, struct nouveau_bo *bo) return false; } -static INLINE void -so_dump(struct nouveau_stateobj *so) -{ - unsigned i, nr = so->cur - so->push; - - for (i = 0; i < nr; i++) - debug_printf("+0x%04x: 0x%08x\n", i, so->push[i]); -} - static INLINE void so_emit(struct nouveau_channel *chan, struct nouveau_stateobj *so) { @@ -127,75 +233,93 @@ so_emit(struct nouveau_channel *chan, struct nouveau_stateobj *so) unsigned nr, i; int ret = 0; - nr = so->cur - so->push; +#ifdef DEBUG_NOUVEAU_STATEOBJ + if (so->start[so->cur_start - 1].size > so->cur) { + debug_printf("emit: previous so_method was not filled\n"); + assert(0); + } +#endif /* DEBUG_NOUVEAU_STATEOBJ */ + + /* We cannot update total in case we so_emit again. */ + nr = so->total + so->cur; + /* This will flush if we need space. * We don't actually need the marker. */ if ((ret = nouveau_pushbuf_marker_emit(chan, nr, so->cur_reloc))) { debug_printf("so_emit failed marker emit with error %d\n", ret); - return; + assert(0); + } + + /* Submit data. This will ensure proper binding of objects. */ + for (i = 0; i < so->cur_start; i++) { + BEGIN_RING(chan, so->start[i].gr, so->start[i].mthd, so->start[i].size); + OUT_RINGp(chan, &(so->pool[so->start[i].offset]), so->start[i].size); } - pb->remaining -= nr; - memcpy(pb->cur, so->push, nr * 4); for (i = 0; i < so->cur_reloc; i++) { struct nouveau_stateobj_reloc *r = &so->reloc[i]; - if ((ret = nouveau_pushbuf_emit_reloc(chan, pb->cur + r->offset, - r->bo, r->data, 0, r->flags, - r->vor, r->tor))) { + if ((ret = nouveau_pushbuf_emit_reloc(chan, pb->cur - nr + + r->push_offset, r->bo, r->data, + 0, r->flags, r->vor, r->tor))) { debug_printf("so_emit failed reloc with error %d\n", ret); - goto out; + assert(0); } } -out: - pb->cur += nr; } static INLINE void so_emit_reloc_markers(struct nouveau_channel *chan, struct nouveau_stateobj *so) { struct nouveau_pushbuf *pb = chan->pushbuf; + struct nouveau_grobj *gr = NULL; unsigned i; int ret = 0; if (!so) return; - i = so->cur_reloc << 1; - /* This will flush if we need space. - * We don't actually need the marker. - */ - if ((ret = nouveau_pushbuf_marker_emit(chan, i, i))) { - debug_printf("so_emit_reloc_markers failed marker emit with" \ - "error %d\n", ret); - return; - } - pb->remaining -= i; - + /* If we need to flush in flush notify, then we have a problem anyway. */ for (i = 0; i < so->cur_reloc; i++) { struct nouveau_stateobj_reloc *r = &so->reloc[i]; - if ((ret = nouveau_pushbuf_emit_reloc(chan, pb->cur++, r->bo, - r->packet, 0, - (r->flags & (NOUVEAU_BO_VRAM | - NOUVEAU_BO_GART | - NOUVEAU_BO_RDWR)) | - NOUVEAU_BO_DUMMY, 0, 0))) { - debug_printf("so_emit_reloc_markers failed reloc" \ - "with error %d\n", ret); - pb->remaining += ((so->cur_reloc - i) << 1); - return; +#ifdef DEBUG_NOUVEAU_STATEOBJ + if (r->mthd & 0x40000000) { + debug_printf("error: NI mthd 0x%08X\n", r->mthd); + continue; + } +#endif /* DEBUG_NOUVEAU_STATEOBJ */ + + /* The object needs to be bound and the system must know the + * subchannel is being used. Otherwise it will discard it. + */ + if (gr != r->gr) { + BEGIN_RING(chan, r->gr, 0x100, 1); + OUT_RING(chan, 0); + gr = r->gr; + } + + /* Some relocs really don't like to be hammered, + * NOUVEAU_BO_DUMMY makes sure it only + * happens when needed. + */ + ret = OUT_RELOC(chan, r->bo, (r->gr->subc << 13) | (1<< 18) | + r->mthd, (r->flags & (NOUVEAU_BO_VRAM | NOUVEAU_BO_GART + | NOUVEAU_BO_RDWR)) | NOUVEAU_BO_DUMMY, 0, 0); + if (ret) { + debug_printf("OUT_RELOC failed %d\n", ret); + assert(0); } - if ((ret = nouveau_pushbuf_emit_reloc(chan, pb->cur++, r->bo, - r->data, 0, - r->flags | NOUVEAU_BO_DUMMY, - r->vor, r->tor))) { - debug_printf("so_emit_reloc_markers failed reloc" \ - "with error %d\n", ret); - pb->remaining += ((so->cur_reloc - i) << 1) - 1; - return; + + ret = OUT_RELOC(chan, r->bo, r->data, r->flags | + NOUVEAU_BO_DUMMY, r->vor, r->tor); + if (ret) { + debug_printf("OUT_RELOC failed %d\n", ret); + assert(0); } + + pb->remaining -= 2; } } diff --git a/src/gallium/drivers/nv10/nv10_screen.c b/src/gallium/drivers/nv10/nv10_screen.c index 6a39ddeaac..69a6dab866 100644 --- a/src/gallium/drivers/nv10/nv10_screen.c +++ b/src/gallium/drivers/nv10/nv10_screen.c @@ -180,7 +180,6 @@ nv10_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) NOUVEAU_ERR("Error creating 3D object: %d\n", ret); return FALSE; } - BIND_RING(chan, screen->celsius, 7); /* 2D engine setup */ screen->eng2d = nv04_surface_2d_init(&screen->base); diff --git a/src/gallium/drivers/nv20/nv20_screen.c b/src/gallium/drivers/nv20/nv20_screen.c index a0973f1ebd..d091335063 100644 --- a/src/gallium/drivers/nv20/nv20_screen.c +++ b/src/gallium/drivers/nv20/nv20_screen.c @@ -176,7 +176,6 @@ nv20_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) NOUVEAU_ERR("Error creating 3D object: %d\n", ret); return FALSE; } - BIND_RING(chan, screen->kelvin, 7); /* 2D engine setup */ screen->eng2d = nv04_surface_2d_init(&screen->base); diff --git a/src/gallium/drivers/nv30/nv30_fragprog.c b/src/gallium/drivers/nv30/nv30_fragprog.c index d1ff18e2df..2d565cb631 100644 --- a/src/gallium/drivers/nv30/nv30_fragprog.c +++ b/src/gallium/drivers/nv30/nv30_fragprog.c @@ -837,7 +837,7 @@ nv30_fragprog_validate(struct nv30_context *nv30) fp->buffer = pscreen->buffer_create(pscreen, 0x100, 0, fp->insn_len * 4); nv30_fragprog_upload(nv30, fp); - so = so_new(8, 1); + so = so_new(4, 4, 1); so_method(so, nv30->screen->rankine, NV34TCL_FP_ACTIVE_PROGRAM, 1); so_reloc (so, nouveau_bo(fp->buffer), 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD | NOUVEAU_BO_LOW | diff --git a/src/gallium/drivers/nv30/nv30_fragtex.c b/src/gallium/drivers/nv30/nv30_fragtex.c index b3293ee700..9893567891 100644 --- a/src/gallium/drivers/nv30/nv30_fragtex.c +++ b/src/gallium/drivers/nv30/nv30_fragtex.c @@ -106,7 +106,7 @@ nv30_fragtex_build(struct nv30_context *nv30, int unit) txs = tf->swizzle; - so = so_new(16, 2); + so = so_new(1, 8, 2); so_method(so, nv30->screen->rankine, NV34TCL_TX_OFFSET(unit), 8); so_reloc (so, bo, 0, tex_flags | NOUVEAU_BO_LOW, 0, 0); so_reloc (so, bo, txf, tex_flags | NOUVEAU_BO_OR, @@ -135,7 +135,7 @@ nv30_fragtex_validate(struct nv30_context *nv30) unit = ffs(samplers) - 1; samplers &= ~(1 << unit); - so = so_new(2, 0); + so = so_new(1, 1, 0); so_method(so, nv30->screen->rankine, NV34TCL_TX_ENABLE(unit), 1); so_data (so, 0); so_ref(so, &nv30->state.hw[NV30_STATE_FRAGTEX0 + unit]); diff --git a/src/gallium/drivers/nv30/nv30_screen.c b/src/gallium/drivers/nv30/nv30_screen.c index 760467f736..9ed48178dc 100644 --- a/src/gallium/drivers/nv30/nv30_screen.c +++ b/src/gallium/drivers/nv30/nv30_screen.c @@ -233,7 +233,6 @@ nv30_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) NOUVEAU_ERR("Error creating 3D object: %d\n", ret); return FALSE; } - BIND_RING(chan, screen->rankine, 7); /* 2D engine setup */ screen->eng2d = nv04_surface_2d_init(&screen->base); @@ -270,7 +269,7 @@ nv30_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) } /* Static rankine initialisation */ - so = so_new(128, 0); + so = so_new(36, 60, 0); so_method(so, screen->rankine, NV34TCL_DMA_NOTIFY, 1); so_data (so, screen->sync->handle); so_method(so, screen->rankine, NV34TCL_DMA_TEXTURE0, 2); diff --git a/src/gallium/drivers/nv30/nv30_state.c b/src/gallium/drivers/nv30/nv30_state.c index e6321b480f..a80dfb0488 100644 --- a/src/gallium/drivers/nv30/nv30_state.c +++ b/src/gallium/drivers/nv30/nv30_state.c @@ -14,7 +14,7 @@ nv30_blend_state_create(struct pipe_context *pipe, struct nv30_context *nv30 = nv30_context(pipe); struct nouveau_grobj *rankine = nv30->screen->rankine; struct nv30_blend_state *bso = CALLOC(1, sizeof(*bso)); - struct nouveau_stateobj *so = so_new(16, 0); + struct nouveau_stateobj *so = so_new(5, 8, 0); if (cso->blend_enable) { so_method(so, rankine, NV34TCL_BLEND_FUNC_ENABLE, 3); @@ -300,7 +300,7 @@ nv30_rasterizer_state_create(struct pipe_context *pipe, { struct nv30_context *nv30 = nv30_context(pipe); struct nv30_rasterizer_state *rsso = CALLOC(1, sizeof(*rsso)); - struct nouveau_stateobj *so = so_new(32, 0); + struct nouveau_stateobj *so = so_new(9, 19, 0); struct nouveau_grobj *rankine = nv30->screen->rankine; /*XXX: ignored: @@ -435,7 +435,7 @@ nv30_depth_stencil_alpha_state_create(struct pipe_context *pipe, { struct nv30_context *nv30 = nv30_context(pipe); struct nv30_zsa_state *zsaso = CALLOC(1, sizeof(*zsaso)); - struct nouveau_stateobj *so = so_new(32, 0); + struct nouveau_stateobj *so = so_new(5, 21, 0); struct nouveau_grobj *rankine = nv30->screen->rankine; so_method(so, rankine, NV34TCL_DEPTH_FUNC, 3); diff --git a/src/gallium/drivers/nv30/nv30_state_blend.c b/src/gallium/drivers/nv30/nv30_state_blend.c index 64cf9ae93a..c36d58c040 100644 --- a/src/gallium/drivers/nv30/nv30_state_blend.c +++ b/src/gallium/drivers/nv30/nv30_state_blend.c @@ -18,7 +18,7 @@ struct nv30_state_entry nv30_state_blend = { static boolean nv30_state_blend_colour_validate(struct nv30_context *nv30) { - struct nouveau_stateobj *so = so_new(2, 0); + struct nouveau_stateobj *so = so_new(1, 1, 0); struct pipe_blend_color *bcol = &nv30->blend_colour; so_method(so, nv30->screen->rankine, NV34TCL_BLEND_COLOR, 1); diff --git a/src/gallium/drivers/nv30/nv30_state_fb.c b/src/gallium/drivers/nv30/nv30_state_fb.c index 6f6d1740d6..2ed2ea55e8 100644 --- a/src/gallium/drivers/nv30/nv30_state_fb.c +++ b/src/gallium/drivers/nv30/nv30_state_fb.c @@ -10,7 +10,7 @@ nv30_state_framebuffer_validate(struct nv30_context *nv30) struct nv04_surface *rt[2], *zeta = NULL; uint32_t rt_enable = 0, rt_format = 0; int i, colour_format = 0, zeta_format = 0, depth_only = 0; - struct nouveau_stateobj *so = so_new(64, 10); + struct nouveau_stateobj *so = so_new(12, 18, 10); unsigned rt_flags = NOUVEAU_BO_RDWR | NOUVEAU_BO_VRAM; unsigned w = fb->width; unsigned h = fb->height; diff --git a/src/gallium/drivers/nv30/nv30_state_scissor.c b/src/gallium/drivers/nv30/nv30_state_scissor.c index 3ac7a8471e..ba61a9e24a 100644 --- a/src/gallium/drivers/nv30/nv30_state_scissor.c +++ b/src/gallium/drivers/nv30/nv30_state_scissor.c @@ -12,7 +12,7 @@ nv30_state_scissor_validate(struct nv30_context *nv30) return FALSE; nv30->state.scissor_enabled = rast->scissor; - so = so_new(3, 0); + so = so_new(1, 2, 0); so_method(so, nv30->screen->rankine, NV34TCL_SCISSOR_HORIZ, 2); if (nv30->state.scissor_enabled) { so_data (so, ((s->maxx - s->minx) << 16) | s->minx); diff --git a/src/gallium/drivers/nv30/nv30_state_stipple.c b/src/gallium/drivers/nv30/nv30_state_stipple.c index d0c791ac08..ed520a4f43 100644 --- a/src/gallium/drivers/nv30/nv30_state_stipple.c +++ b/src/gallium/drivers/nv30/nv30_state_stipple.c @@ -14,14 +14,14 @@ nv30_state_stipple_validate(struct nv30_context *nv30) if (rast->poly_stipple_enable) { unsigned i; - so = so_new(35, 0); + so = so_new(2, 33, 0); so_method(so, rankine, NV34TCL_POLYGON_STIPPLE_ENABLE, 1); so_data (so, 1); so_method(so, rankine, NV34TCL_POLYGON_STIPPLE_PATTERN(0), 32); for (i = 0; i < 32; i++) so_data(so, nv30->stipple[i]); } else { - so = so_new(2, 0); + so = so_new(1, 1, 0); so_method(so, rankine, NV34TCL_POLYGON_STIPPLE_ENABLE, 1); so_data (so, 0); } diff --git a/src/gallium/drivers/nv30/nv30_state_viewport.c b/src/gallium/drivers/nv30/nv30_state_viewport.c index c3eb413dac..2d7781292b 100644 --- a/src/gallium/drivers/nv30/nv30_state_viewport.c +++ b/src/gallium/drivers/nv30/nv30_state_viewport.c @@ -19,7 +19,7 @@ nv30_state_viewport_validate(struct nv30_context *nv30) return FALSE; nv30->state.viewport_bypass = bypass; - so = so_new(11, 0); + so = so_new(3, 10, 0); if (!bypass) { so_method(so, nv30->screen->rankine, NV34TCL_VIEWPORT_TRANSLATE_X, 8); diff --git a/src/gallium/drivers/nv30/nv30_vbo.c b/src/gallium/drivers/nv30/nv30_vbo.c index bccc805324..1c5db03ea2 100644 --- a/src/gallium/drivers/nv30/nv30_vbo.c +++ b/src/gallium/drivers/nv30/nv30_vbo.c @@ -491,9 +491,9 @@ nv30_vbo_validate(struct nv30_context *nv30) unsigned vb_flags = NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD; int hw; - vtxbuf = so_new(20, 18); + vtxbuf = so_new(3, 17, 18); so_method(vtxbuf, rankine, NV34TCL_VTXBUF_ADDRESS(0), nv30->vtxelt_nr); - vtxfmt = so_new(17, 0); + vtxfmt = so_new(1, 16, 0); so_method(vtxfmt, rankine, NV34TCL_VTXFMT(0), nv30->vtxelt_nr); for (hw = 0; hw < nv30->vtxelt_nr; hw++) { @@ -506,7 +506,7 @@ nv30_vbo_validate(struct nv30_context *nv30) if (!vb->stride) { if (!sattr) - sattr = so_new(16 * 5, 0); + sattr = so_new(16, 16 * 4, 0); if (nv30_vbo_static_attrib(nv30, sattr, hw, ve, vb)) { so_data(vtxbuf, 0); diff --git a/src/gallium/drivers/nv30/nv30_vertprog.c b/src/gallium/drivers/nv30/nv30_vertprog.c index 4e6d3d01c7..e77a5be3f2 100644 --- a/src/gallium/drivers/nv30/nv30_vertprog.c +++ b/src/gallium/drivers/nv30/nv30_vertprog.c @@ -686,7 +686,7 @@ nv30_vertprog_validate(struct nv30_context *nv30) assert(0); } - so = so_new(2, 0); + so = so_new(1, 1, 0); so_method(so, rankine, NV34TCL_VP_START_FROM_ID, 1); so_data (so, vp->exec->start); so_ref(so, &vp->so); diff --git a/src/gallium/drivers/nv40/nv40_fragprog.c b/src/gallium/drivers/nv40/nv40_fragprog.c index bb9c85cc43..1237066c39 100644 --- a/src/gallium/drivers/nv40/nv40_fragprog.c +++ b/src/gallium/drivers/nv40/nv40_fragprog.c @@ -919,7 +919,7 @@ nv40_fragprog_validate(struct nv40_context *nv40) fp->buffer = pscreen->buffer_create(pscreen, 0x100, 0, fp->insn_len * 4); nv40_fragprog_upload(nv40, fp); - so = so_new(4, 1); + so = so_new(2, 2, 1); so_method(so, nv40->screen->curie, NV40TCL_FP_ADDRESS, 1); so_reloc (so, nouveau_bo(fp->buffer), 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD | NOUVEAU_BO_LOW | diff --git a/src/gallium/drivers/nv40/nv40_fragtex.c b/src/gallium/drivers/nv40/nv40_fragtex.c index 44abc84596..aad9198210 100644 --- a/src/gallium/drivers/nv40/nv40_fragtex.c +++ b/src/gallium/drivers/nv40/nv40_fragtex.c @@ -108,7 +108,7 @@ nv40_fragtex_build(struct nv40_context *nv40, int unit) txs = tf->swizzle; - so = so_new(16, 2); + so = so_new(2, 9, 2); so_method(so, nv40->screen->curie, NV40TCL_TEX_OFFSET(unit), 8); so_reloc (so, bo, 0, tex_flags | NOUVEAU_BO_LOW, 0, 0); so_reloc (so, bo, txf, tex_flags | NOUVEAU_BO_OR, @@ -139,7 +139,7 @@ nv40_fragtex_validate(struct nv40_context *nv40) unit = ffs(samplers) - 1; samplers &= ~(1 << unit); - so = so_new(2, 0); + so = so_new(1, 1, 0); so_method(so, nv40->screen->curie, NV40TCL_TEX_ENABLE(unit), 1); so_data (so, 0); so_ref(so, &nv40->state.hw[NV40_STATE_FRAGTEX0 + unit]); diff --git a/src/gallium/drivers/nv40/nv40_screen.c b/src/gallium/drivers/nv40/nv40_screen.c index d01e712805..9e55e5a089 100644 --- a/src/gallium/drivers/nv40/nv40_screen.c +++ b/src/gallium/drivers/nv40/nv40_screen.c @@ -215,7 +215,6 @@ nv40_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) NOUVEAU_ERR("Error creating 3D object: %d\n", ret); return FALSE; } - BIND_RING(chan, screen->curie, 7); /* 2D engine setup */ screen->eng2d = nv04_surface_2d_init(&screen->base); @@ -252,7 +251,7 @@ nv40_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) } /* Static curie initialisation */ - so = so_new(128, 0); + so = so_new(16, 25, 0); so_method(so, screen->curie, NV40TCL_DMA_NOTIFY, 1); so_data (so, screen->sync->handle); so_method(so, screen->curie, NV40TCL_DMA_TEXTURE0, 2); diff --git a/src/gallium/drivers/nv40/nv40_state.c b/src/gallium/drivers/nv40/nv40_state.c index ed55d29aff..ed0ca9e02c 100644 --- a/src/gallium/drivers/nv40/nv40_state.c +++ b/src/gallium/drivers/nv40/nv40_state.c @@ -16,7 +16,7 @@ nv40_blend_state_create(struct pipe_context *pipe, struct nv40_context *nv40 = nv40_context(pipe); struct nouveau_grobj *curie = nv40->screen->curie; struct nv40_blend_state *bso = CALLOC(1, sizeof(*bso)); - struct nouveau_stateobj *so = so_new(16, 0); + struct nouveau_stateobj *so = so_new(5, 8, 0); if (cso->blend_enable) { so_method(so, curie, NV40TCL_BLEND_ENABLE, 3); @@ -310,7 +310,7 @@ nv40_rasterizer_state_create(struct pipe_context *pipe, { struct nv40_context *nv40 = nv40_context(pipe); struct nv40_rasterizer_state *rsso = CALLOC(1, sizeof(*rsso)); - struct nouveau_stateobj *so = so_new(32, 0); + struct nouveau_stateobj *so = so_new(8, 18, 0); struct nouveau_grobj *curie = nv40->screen->curie; /*XXX: ignored: @@ -445,7 +445,7 @@ nv40_depth_stencil_alpha_state_create(struct pipe_context *pipe, { struct nv40_context *nv40 = nv40_context(pipe); struct nv40_zsa_state *zsaso = CALLOC(1, sizeof(*zsaso)); - struct nouveau_stateobj *so = so_new(32, 0); + struct nouveau_stateobj *so = so_new(4, 21, 0); struct nouveau_grobj *curie = nv40->screen->curie; so_method(so, curie, NV40TCL_DEPTH_FUNC, 3); diff --git a/src/gallium/drivers/nv40/nv40_state_blend.c b/src/gallium/drivers/nv40/nv40_state_blend.c index 8cd05ce66e..3ff00a37f6 100644 --- a/src/gallium/drivers/nv40/nv40_state_blend.c +++ b/src/gallium/drivers/nv40/nv40_state_blend.c @@ -18,7 +18,7 @@ struct nv40_state_entry nv40_state_blend = { static boolean nv40_state_blend_colour_validate(struct nv40_context *nv40) { - struct nouveau_stateobj *so = so_new(2, 0); + struct nouveau_stateobj *so = so_new(1, 1, 0); struct pipe_blend_color *bcol = &nv40->blend_colour; so_method(so, nv40->screen->curie, NV40TCL_BLEND_COLOR, 1); diff --git a/src/gallium/drivers/nv40/nv40_state_fb.c b/src/gallium/drivers/nv40/nv40_state_fb.c index 1c7a7cd64f..a58fe9ddb1 100644 --- a/src/gallium/drivers/nv40/nv40_state_fb.c +++ b/src/gallium/drivers/nv40/nv40_state_fb.c @@ -19,7 +19,7 @@ nv40_state_framebuffer_validate(struct nv40_context *nv40) struct nv04_surface *rt[4], *zeta; uint32_t rt_enable, rt_format; int i, colour_format = 0, zeta_format = 0; - struct nouveau_stateobj *so = so_new(64, 10); + struct nouveau_stateobj *so = so_new(18, 24, 10); unsigned rt_flags = NOUVEAU_BO_RDWR | NOUVEAU_BO_VRAM; unsigned w = fb->width; unsigned h = fb->height; diff --git a/src/gallium/drivers/nv40/nv40_state_scissor.c b/src/gallium/drivers/nv40/nv40_state_scissor.c index cf58d33906..753a505e93 100644 --- a/src/gallium/drivers/nv40/nv40_state_scissor.c +++ b/src/gallium/drivers/nv40/nv40_state_scissor.c @@ -12,7 +12,7 @@ nv40_state_scissor_validate(struct nv40_context *nv40) return FALSE; nv40->state.scissor_enabled = rast->scissor; - so = so_new(3, 0); + so = so_new(1, 2, 0); so_method(so, nv40->screen->curie, NV40TCL_SCISSOR_HORIZ, 2); if (nv40->state.scissor_enabled) { so_data (so, ((s->maxx - s->minx) << 16) | s->minx); diff --git a/src/gallium/drivers/nv40/nv40_state_stipple.c b/src/gallium/drivers/nv40/nv40_state_stipple.c index b51024ad9b..2b371ebfec 100644 --- a/src/gallium/drivers/nv40/nv40_state_stipple.c +++ b/src/gallium/drivers/nv40/nv40_state_stipple.c @@ -14,14 +14,14 @@ nv40_state_stipple_validate(struct nv40_context *nv40) if (rast->poly_stipple_enable) { unsigned i; - so = so_new(35, 0); + so = so_new(2, 33, 0); so_method(so, curie, NV40TCL_POLYGON_STIPPLE_ENABLE, 1); so_data (so, 1); so_method(so, curie, NV40TCL_POLYGON_STIPPLE_PATTERN(0), 32); for (i = 0; i < 32; i++) so_data(so, nv40->stipple[i]); } else { - so = so_new(2, 0); + so = so_new(1, 1, 0); so_method(so, curie, NV40TCL_POLYGON_STIPPLE_ENABLE, 1); so_data (so, 0); } diff --git a/src/gallium/drivers/nv40/nv40_state_viewport.c b/src/gallium/drivers/nv40/nv40_state_viewport.c index 665d2d5fca..9919ba1d0b 100644 --- a/src/gallium/drivers/nv40/nv40_state_viewport.c +++ b/src/gallium/drivers/nv40/nv40_state_viewport.c @@ -19,7 +19,7 @@ nv40_state_viewport_validate(struct nv40_context *nv40) return FALSE; nv40->state.viewport_bypass = bypass; - so = so_new(11, 0); + so = so_new(2, 9, 0); if (!bypass) { so_method(so, nv40->screen->curie, NV40TCL_VIEWPORT_TRANSLATE_X, 8); diff --git a/src/gallium/drivers/nv40/nv40_vbo.c b/src/gallium/drivers/nv40/nv40_vbo.c index 90087f0bee..a777898f68 100644 --- a/src/gallium/drivers/nv40/nv40_vbo.c +++ b/src/gallium/drivers/nv40/nv40_vbo.c @@ -491,9 +491,9 @@ nv40_vbo_validate(struct nv40_context *nv40) unsigned vb_flags = NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD; int hw; - vtxbuf = so_new(20, 18); + vtxbuf = so_new(3, 17, 18); so_method(vtxbuf, curie, NV40TCL_VTXBUF_ADDRESS(0), nv40->vtxelt_nr); - vtxfmt = so_new(17, 0); + vtxfmt = so_new(1, 16, 0); so_method(vtxfmt, curie, NV40TCL_VTXFMT(0), nv40->vtxelt_nr); for (hw = 0; hw < nv40->vtxelt_nr; hw++) { @@ -506,7 +506,7 @@ nv40_vbo_validate(struct nv40_context *nv40) if (!vb->stride) { if (!sattr) - sattr = so_new(16 * 5, 0); + sattr = so_new(16, 16 * 4, 0); if (nv40_vbo_static_attrib(nv40, sattr, hw, ve, vb)) { so_data(vtxbuf, 0); diff --git a/src/gallium/drivers/nv40/nv40_vertprog.c b/src/gallium/drivers/nv40/nv40_vertprog.c index afbb2cb47b..8d80fcad38 100644 --- a/src/gallium/drivers/nv40/nv40_vertprog.c +++ b/src/gallium/drivers/nv40/nv40_vertprog.c @@ -886,7 +886,7 @@ check_gpu_resources: assert(0); } - so = so_new(7, 0); + so = so_new(3, 4, 0); so_method(so, curie, NV40TCL_VP_START_FROM_ID, 1); so_data (so, vp->exec->start); so_method(so, curie, NV40TCL_VP_ATTRIB_EN, 2); diff --git a/src/gallium/drivers/nv50/nv50_program.c b/src/gallium/drivers/nv50/nv50_program.c index 2d0b1818ef..af0759e503 100644 --- a/src/gallium/drivers/nv50/nv50_program.c +++ b/src/gallium/drivers/nv50/nv50_program.c @@ -3579,7 +3579,7 @@ nv50_vertprog_validate(struct nv50_context *nv50) nv50_program_validate_data(nv50, p); nv50_program_validate_code(nv50, p); - so = so_new(13, 2); + so = so_new(5, 8, 2); so_method(so, tesla, NV50TCL_VP_ADDRESS_HIGH, 2); so_reloc (so, p->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0); @@ -3615,7 +3615,7 @@ nv50_fragprog_validate(struct nv50_context *nv50) nv50_program_validate_data(nv50, p); nv50_program_validate_code(nv50, p); - so = so_new(64, 2); + so = so_new(6, 7, 2); so_method(so, tesla, NV50TCL_FP_ADDRESS_HIGH, 2); so_reloc (so, p->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0); @@ -3783,7 +3783,7 @@ nv50_linkage_validate(struct nv50_context *nv50) } /* now fill the stateobj */ - so = so_new(64, 0); + so = so_new(6, 58, 0); n = (m + 3) / 4; so_method(so, tesla, NV50TCL_VP_RESULT_MAP_SIZE, 1); diff --git a/src/gallium/drivers/nv50/nv50_screen.c b/src/gallium/drivers/nv50/nv50_screen.c index 1778a74517..28e2b35dea 100644 --- a/src/gallium/drivers/nv50/nv50_screen.c +++ b/src/gallium/drivers/nv50/nv50_screen.c @@ -251,7 +251,6 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) nv50_screen_destroy(pscreen); return NULL; } - BIND_RING(chan, screen->m2mf, 1); /* 2D object */ ret = nouveau_grobj_alloc(chan, 0xbeef502d, NV50_2D, &screen->eng2d); @@ -260,7 +259,6 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) nv50_screen_destroy(pscreen); return NULL; } - BIND_RING(chan, screen->eng2d, 2); /* 3D object */ switch (chipset & 0xf0) { @@ -296,7 +294,6 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) nv50_screen_destroy(pscreen); return NULL; } - BIND_RING(chan, screen->tesla, 3); /* Sync notifier */ ret = nouveau_notifier_alloc(chan, 0xbeef0301, 1, &screen->sync); @@ -307,7 +304,7 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) } /* Static M2MF init */ - so = so_new(32, 0); + so = so_new(1, 3, 0); so_method(so, screen->m2mf, NV04_MEMORY_TO_MEMORY_FORMAT_DMA_NOTIFY, 3); so_data (so, screen->sync->handle); so_data (so, chan->vram->handle); @@ -316,7 +313,7 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) so_ref (NULL, &so); /* Static 2D init */ - so = so_new(64, 0); + so = so_new(4, 7, 0); so_method(so, screen->eng2d, NV50_2D_DMA_NOTIFY, 4); so_data (so, screen->sync->handle); so_data (so, chan->vram->handle); @@ -332,7 +329,7 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) so_ref(NULL, &so); /* Static tesla init */ - so = so_new(256, 20); + so = so_new(40, 84, 20); so_method(so, screen->tesla, NV50TCL_COND_MODE, 1); so_data (so, NV50TCL_COND_MODE_ALWAYS); diff --git a/src/gallium/drivers/nv50/nv50_state.c b/src/gallium/drivers/nv50/nv50_state.c index 30b2b0f91b..fd0a33d7f7 100644 --- a/src/gallium/drivers/nv50/nv50_state.c +++ b/src/gallium/drivers/nv50/nv50_state.c @@ -35,7 +35,7 @@ static void * nv50_blend_state_create(struct pipe_context *pipe, const struct pipe_blend_state *cso) { - struct nouveau_stateobj *so = so_new(64, 0); + struct nouveau_stateobj *so = so_new(5, 24, 0); struct nouveau_grobj *tesla = nv50_context(pipe)->screen->tesla; struct nv50_blend_stateobj *bso = CALLOC_STRUCT(nv50_blend_stateobj); unsigned cmask = 0, i; @@ -280,7 +280,7 @@ static void * nv50_rasterizer_state_create(struct pipe_context *pipe, const struct pipe_rasterizer_state *cso) { - struct nouveau_stateobj *so = so_new(64, 0); + struct nouveau_stateobj *so = so_new(15, 21, 0); struct nouveau_grobj *tesla = nv50_context(pipe)->screen->tesla; struct nv50_rasterizer_stateobj *rso = CALLOC_STRUCT(nv50_rasterizer_stateobj); @@ -425,7 +425,7 @@ nv50_depth_stencil_alpha_state_create(struct pipe_context *pipe, { struct nouveau_grobj *tesla = nv50_context(pipe)->screen->tesla; struct nv50_zsa_stateobj *zsa = CALLOC_STRUCT(nv50_zsa_stateobj); - struct nouveau_stateobj *so = so_new(64, 0); + struct nouveau_stateobj *so = so_new(8, 22, 0); so_method(so, tesla, NV50TCL_DEPTH_WRITE_ENABLE, 1); so_data (so, cso->depth.writemask ? 1 : 0); diff --git a/src/gallium/drivers/nv50/nv50_state_validate.c b/src/gallium/drivers/nv50/nv50_state_validate.c index 682786345e..f83232f43c 100644 --- a/src/gallium/drivers/nv50/nv50_state_validate.c +++ b/src/gallium/drivers/nv50/nv50_state_validate.c @@ -33,7 +33,7 @@ static void nv50_state_validate_fb(struct nv50_context *nv50) { struct nouveau_grobj *tesla = nv50->screen->tesla; - struct nouveau_stateobj *so = so_new(128, 18); + struct nouveau_stateobj *so = so_new(32, 79, 18); struct pipe_framebuffer_state *fb = &nv50->framebuffer; unsigned i, w, h, gw = 0; @@ -299,7 +299,7 @@ nv50_state_validate(struct nv50_context *nv50) so_ref(nv50->rasterizer->so, &nv50->state.rast); if (nv50->dirty & NV50_NEW_BLEND_COLOUR) { - so = so_new(5, 0); + so = so_new(1, 4, 0); so_method(so, tesla, NV50TCL_BLEND_COLOR(0), 4); so_data (so, fui(nv50->blend_colour.color[0])); so_data (so, fui(nv50->blend_colour.color[1])); @@ -310,7 +310,7 @@ nv50_state_validate(struct nv50_context *nv50) } if (nv50->dirty & NV50_NEW_STIPPLE) { - so = so_new(33, 0); + so = so_new(1, 32, 0); so_method(so, tesla, NV50TCL_POLYGON_STIPPLE_PATTERN(0), 32); for (i = 0; i < 32; i++) so_data(so, util_bswap32(nv50->stipple.stipple[i])); @@ -327,7 +327,7 @@ nv50_state_validate(struct nv50_context *nv50) goto scissor_uptodate; nv50->state.scissor_enabled = rast->scissor; - so = so_new(3, 0); + so = so_new(1, 2, 0); so_method(so, tesla, NV50TCL_SCISSOR_HORIZ(0), 2); if (nv50->state.scissor_enabled) { so_data(so, (s->maxx << 16) | s->minx); @@ -356,7 +356,7 @@ scissor_uptodate: goto viewport_uptodate; nv50->state.viewport_bypass = bypass; - so = so_new(14, 0); + so = so_new(5, 9, 0); if (!bypass) { so_method(so, tesla, NV50TCL_VIEWPORT_TRANSLATE_X(0), 3); so_data (so, fui(nv50->viewport.translate[0])); @@ -400,7 +400,8 @@ viewport_uptodate: for (i = 0; i < PIPE_SHADER_TYPES; ++i) nr += nv50->sampler_nr[i]; - so = so_new(nr * 8 + 24 * PIPE_SHADER_TYPES + 2, 4); + so = so_new(1+ 5 * PIPE_SHADER_TYPES, 1+ 19 * PIPE_SHADER_TYPES + + nr * 8, PIPE_SHADER_TYPES * 2); nv50_validate_samplers(nv50, so, PIPE_SHADER_VERTEX); nv50_validate_samplers(nv50, so, PIPE_SHADER_FRAGMENT); diff --git a/src/gallium/drivers/nv50/nv50_tex.c b/src/gallium/drivers/nv50/nv50_tex.c index c4ca096d6a..bef548b728 100644 --- a/src/gallium/drivers/nv50/nv50_tex.c +++ b/src/gallium/drivers/nv50/nv50_tex.c @@ -199,16 +199,18 @@ nv50_tex_validate(struct nv50_context *nv50) { struct nouveau_stateobj *so; struct nouveau_grobj *tesla = nv50->screen->tesla; - unsigned p, push, nrlc; + unsigned p, start, push, nrlc; - for (nrlc = 0, push = 0, p = 0; p < PIPE_SHADER_TYPES; ++p) { + for (nrlc = 0, start = 0, push = 0, p = 0; p < PIPE_SHADER_TYPES; ++p) { + start += MAX2(nv50->miptree_nr[p], nv50->state.miptree_nr[p]); push += MAX2(nv50->miptree_nr[p], nv50->state.miptree_nr[p]); nrlc += nv50->miptree_nr[p]; } - push = push * 11 + 23 * PIPE_SHADER_TYPES + 4; + start = start * 2 + 4 * PIPE_SHADER_TYPES + 2; + push = push * 9 + 19 * PIPE_SHADER_TYPES + 2; nrlc = nrlc * 2 + 2 * PIPE_SHADER_TYPES; - so = so_new(push, nrlc); + so = so_new(start, push, nrlc); if (nv50_validate_textures(nv50, so, PIPE_SHADER_VERTEX) == FALSE || nv50_validate_textures(nv50, so, PIPE_SHADER_FRAGMENT) == FALSE) { diff --git a/src/gallium/drivers/nv50/nv50_vbo.c b/src/gallium/drivers/nv50/nv50_vbo.c index df18c2dd20..f2e510fba6 100644 --- a/src/gallium/drivers/nv50/nv50_vbo.c +++ b/src/gallium/drivers/nv50/nv50_vbo.c @@ -354,7 +354,7 @@ nv50_vbo_static_attrib(struct nv50_context *nv50, unsigned attrib, so = *pso; if (!so) - *pso = so = so_new(nv50->vtxelt_nr * 5, 0); + *pso = so = so_new(nv50->vtxelt_nr, nv50->vtxelt_nr * 4, 0); switch (ve->nr_components) { case 4: @@ -415,8 +415,8 @@ nv50_vbo_validate(struct nv50_context *nv50) n_ve = MAX2(nv50->vtxelt_nr, nv50->state.vtxelt_nr); vtxattr = NULL; - vtxbuf = so_new(n_ve * 7, nv50->vtxelt_nr * 4); - vtxfmt = so_new(n_ve + 1, 0); + vtxbuf = so_new(n_ve * 2, n_ve * 5, nv50->vtxelt_nr * 4); + vtxfmt = so_new(1, n_ve, 0); so_method(vtxfmt, tesla, NV50TCL_VERTEX_ARRAY_ATTRIB(0), n_ve); for (i = 0; i < nv50->vtxelt_nr; i++) { -- cgit v1.2.3