X-Git-Url: https://git.libre-soc.org/?a=blobdiff_plain;f=src%2Fmesa%2Fstate_tracker%2Fst_draw.c;h=208ea735a78b06184f161607c23ed3e279684fb9;hb=86d63b53a20a747e3de769549572d895c7fbb736;hp=cbf4e59fb6d7db4ff59ddd735d710a1bdc3fd754;hpb=01bf5569c44389c1127bbb9e873c8a234ac92ff7;p=mesa.git diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index cbf4e59fb6d..208ea735a78 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -1,8 +1,8 @@ /************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * + * Copyright 2007 VMware, Inc. * All Rights Reserved. - * + * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the * "Software"), to deal in the Software without restriction, including @@ -10,19 +10,19 @@ * distribute, sub license, and/or sell copies of the Software, and to * permit persons to whom the Software is furnished to do so, subject to * the following conditions: - * + * * The above copyright notice and this permission notice (including the * next paragraph) shall be included in all copies or substantial portions * of the Software. - * + * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * + * **************************************************************************/ /* @@ -31,27 +31,28 @@ * this function whether the user called glBegin/End, glDrawArrays, * glDrawElements, glEvalMesh, or glCalList, etc. * - * We basically convert the VBO's vertex attribute/array information into - * Gallium vertex state, bind the vertex buffer objects and call - * pipe->draw_vbo(). - * * Authors: - * Keith Whitwell + * Keith Whitwell */ +#include "main/errors.h" #include "main/imports.h" #include "main/image.h" #include "main/bufferobj.h" #include "main/macros.h" -#include "main/mfeatures.h" +#include "main/varray.h" + +#include "compiler/glsl/ir_uniform.h" #include "vbo/vbo.h" #include "st_context.h" #include "st_atom.h" +#include "st_cb_bitmap.h" #include "st_cb_bufferobjects.h" #include "st_cb_xformfb.h" +#include "st_debug.h" #include "st_draw.h" #include "st_program.h" @@ -60,749 +61,32 @@ #include "util/u_inlines.h" #include "util/u_format.h" #include "util/u_prim.h" -#include "util/u_draw_quad.h" +#include "util/u_draw.h" #include "util/u_upload_mgr.h" #include "draw/draw_context.h" #include "cso_cache/cso_context.h" -#include "../glsl/ir_uniform.h" - - -static GLuint double_types[4] = { - PIPE_FORMAT_R64_FLOAT, - PIPE_FORMAT_R64G64_FLOAT, - PIPE_FORMAT_R64G64B64_FLOAT, - PIPE_FORMAT_R64G64B64A64_FLOAT -}; - -static GLuint float_types[4] = { - PIPE_FORMAT_R32_FLOAT, - PIPE_FORMAT_R32G32_FLOAT, - PIPE_FORMAT_R32G32B32_FLOAT, - PIPE_FORMAT_R32G32B32A32_FLOAT -}; - -static GLuint half_float_types[4] = { - PIPE_FORMAT_R16_FLOAT, - PIPE_FORMAT_R16G16_FLOAT, - PIPE_FORMAT_R16G16B16_FLOAT, - PIPE_FORMAT_R16G16B16A16_FLOAT -}; - -static GLuint uint_types_norm[4] = { - PIPE_FORMAT_R32_UNORM, - PIPE_FORMAT_R32G32_UNORM, - PIPE_FORMAT_R32G32B32_UNORM, - PIPE_FORMAT_R32G32B32A32_UNORM -}; - -static GLuint uint_types_scale[4] = { - PIPE_FORMAT_R32_USCALED, - PIPE_FORMAT_R32G32_USCALED, - PIPE_FORMAT_R32G32B32_USCALED, - PIPE_FORMAT_R32G32B32A32_USCALED -}; - -static GLuint uint_types_int[4] = { - PIPE_FORMAT_R32_UINT, - PIPE_FORMAT_R32G32_UINT, - PIPE_FORMAT_R32G32B32_UINT, - PIPE_FORMAT_R32G32B32A32_UINT -}; - -static GLuint int_types_norm[4] = { - PIPE_FORMAT_R32_SNORM, - PIPE_FORMAT_R32G32_SNORM, - PIPE_FORMAT_R32G32B32_SNORM, - PIPE_FORMAT_R32G32B32A32_SNORM -}; - -static GLuint int_types_scale[4] = { - PIPE_FORMAT_R32_SSCALED, - PIPE_FORMAT_R32G32_SSCALED, - PIPE_FORMAT_R32G32B32_SSCALED, - PIPE_FORMAT_R32G32B32A32_SSCALED -}; - -static GLuint int_types_int[4] = { - PIPE_FORMAT_R32_SINT, - PIPE_FORMAT_R32G32_SINT, - PIPE_FORMAT_R32G32B32_SINT, - PIPE_FORMAT_R32G32B32A32_SINT -}; - -static GLuint ushort_types_norm[4] = { - PIPE_FORMAT_R16_UNORM, - PIPE_FORMAT_R16G16_UNORM, - PIPE_FORMAT_R16G16B16_UNORM, - PIPE_FORMAT_R16G16B16A16_UNORM -}; - -static GLuint ushort_types_scale[4] = { - PIPE_FORMAT_R16_USCALED, - PIPE_FORMAT_R16G16_USCALED, - PIPE_FORMAT_R16G16B16_USCALED, - PIPE_FORMAT_R16G16B16A16_USCALED -}; - -static GLuint ushort_types_int[4] = { - PIPE_FORMAT_R16_UINT, - PIPE_FORMAT_R16G16_UINT, - PIPE_FORMAT_R16G16B16_UINT, - PIPE_FORMAT_R16G16B16A16_UINT -}; - -static GLuint short_types_norm[4] = { - PIPE_FORMAT_R16_SNORM, - PIPE_FORMAT_R16G16_SNORM, - PIPE_FORMAT_R16G16B16_SNORM, - PIPE_FORMAT_R16G16B16A16_SNORM -}; - -static GLuint short_types_scale[4] = { - PIPE_FORMAT_R16_SSCALED, - PIPE_FORMAT_R16G16_SSCALED, - PIPE_FORMAT_R16G16B16_SSCALED, - PIPE_FORMAT_R16G16B16A16_SSCALED -}; - -static GLuint short_types_int[4] = { - PIPE_FORMAT_R16_SINT, - PIPE_FORMAT_R16G16_SINT, - PIPE_FORMAT_R16G16B16_SINT, - PIPE_FORMAT_R16G16B16A16_SINT -}; - -static GLuint ubyte_types_norm[4] = { - PIPE_FORMAT_R8_UNORM, - PIPE_FORMAT_R8G8_UNORM, - PIPE_FORMAT_R8G8B8_UNORM, - PIPE_FORMAT_R8G8B8A8_UNORM -}; - -static GLuint ubyte_types_scale[4] = { - PIPE_FORMAT_R8_USCALED, - PIPE_FORMAT_R8G8_USCALED, - PIPE_FORMAT_R8G8B8_USCALED, - PIPE_FORMAT_R8G8B8A8_USCALED -}; - -static GLuint ubyte_types_int[4] = { - PIPE_FORMAT_R8_UINT, - PIPE_FORMAT_R8G8_UINT, - PIPE_FORMAT_R8G8B8_UINT, - PIPE_FORMAT_R8G8B8A8_UINT -}; - -static GLuint byte_types_norm[4] = { - PIPE_FORMAT_R8_SNORM, - PIPE_FORMAT_R8G8_SNORM, - PIPE_FORMAT_R8G8B8_SNORM, - PIPE_FORMAT_R8G8B8A8_SNORM -}; - -static GLuint byte_types_scale[4] = { - PIPE_FORMAT_R8_SSCALED, - PIPE_FORMAT_R8G8_SSCALED, - PIPE_FORMAT_R8G8B8_SSCALED, - PIPE_FORMAT_R8G8B8A8_SSCALED -}; - -static GLuint byte_types_int[4] = { - PIPE_FORMAT_R8_SINT, - PIPE_FORMAT_R8G8_SINT, - PIPE_FORMAT_R8G8B8_SINT, - PIPE_FORMAT_R8G8B8A8_SINT -}; - -static GLuint fixed_types[4] = { - PIPE_FORMAT_R32_FIXED, - PIPE_FORMAT_R32G32_FIXED, - PIPE_FORMAT_R32G32B32_FIXED, - PIPE_FORMAT_R32G32B32A32_FIXED -}; - - - -/** - * Return a PIPE_FORMAT_x for the given GL datatype and size. - */ -enum pipe_format -st_pipe_vertex_format(GLenum type, GLuint size, GLenum format, - GLboolean normalized, GLboolean integer) -{ - assert((type >= GL_BYTE && type <= GL_DOUBLE) || - type == GL_FIXED || type == GL_HALF_FLOAT || - type == GL_INT_2_10_10_10_REV || - type == GL_UNSIGNED_INT_2_10_10_10_REV); - assert(size >= 1); - assert(size <= 4); - assert(format == GL_RGBA || format == GL_BGRA); - - if (type == GL_INT_2_10_10_10_REV || - type == GL_UNSIGNED_INT_2_10_10_10_REV) { - assert(size == 4); - assert(!integer); - - if (format == GL_BGRA) { - if (type == GL_INT_2_10_10_10_REV) { - if (normalized) - return PIPE_FORMAT_B10G10R10A2_SNORM; - else - return PIPE_FORMAT_B10G10R10A2_SSCALED; - } else { - if (normalized) - return PIPE_FORMAT_B10G10R10A2_UNORM; - else - return PIPE_FORMAT_B10G10R10A2_USCALED; - } - } else { - if (type == GL_INT_2_10_10_10_REV) { - if (normalized) - return PIPE_FORMAT_R10G10B10A2_SNORM; - else - return PIPE_FORMAT_R10G10B10A2_SSCALED; - } else { - if (normalized) - return PIPE_FORMAT_R10G10B10A2_UNORM; - else - return PIPE_FORMAT_R10G10B10A2_USCALED; - } - } - } - - if (format == GL_BGRA) { - /* this is an odd-ball case */ - assert(type == GL_UNSIGNED_BYTE); - assert(normalized); - return PIPE_FORMAT_B8G8R8A8_UNORM; - } - - if (integer) { - switch (type) { - case GL_INT: return int_types_int[size-1]; - case GL_SHORT: return short_types_int[size-1]; - case GL_BYTE: return byte_types_int[size-1]; - case GL_UNSIGNED_INT: return uint_types_int[size-1]; - case GL_UNSIGNED_SHORT: return ushort_types_int[size-1]; - case GL_UNSIGNED_BYTE: return ubyte_types_int[size-1]; - default: assert(0); return 0; - } - } - else if (normalized) { - switch (type) { - case GL_DOUBLE: return double_types[size-1]; - case GL_FLOAT: return float_types[size-1]; - case GL_HALF_FLOAT: return half_float_types[size-1]; - case GL_INT: return int_types_norm[size-1]; - case GL_SHORT: return short_types_norm[size-1]; - case GL_BYTE: return byte_types_norm[size-1]; - case GL_UNSIGNED_INT: return uint_types_norm[size-1]; - case GL_UNSIGNED_SHORT: return ushort_types_norm[size-1]; - case GL_UNSIGNED_BYTE: return ubyte_types_norm[size-1]; - case GL_FIXED: return fixed_types[size-1]; - default: assert(0); return 0; - } - } - else { - switch (type) { - case GL_DOUBLE: return double_types[size-1]; - case GL_FLOAT: return float_types[size-1]; - case GL_HALF_FLOAT: return half_float_types[size-1]; - case GL_INT: return int_types_scale[size-1]; - case GL_SHORT: return short_types_scale[size-1]; - case GL_BYTE: return byte_types_scale[size-1]; - case GL_UNSIGNED_INT: return uint_types_scale[size-1]; - case GL_UNSIGNED_SHORT: return ushort_types_scale[size-1]; - case GL_UNSIGNED_BYTE: return ubyte_types_scale[size-1]; - case GL_FIXED: return fixed_types[size-1]; - default: assert(0); return 0; - } - } - return PIPE_FORMAT_NONE; /* silence compiler warning */ -} - - -/** - * This is very similar to vbo_all_varyings_in_vbos() but we are - * only interested in per-vertex data. See bug 38626. - */ -static GLboolean -all_varyings_in_vbos(const struct gl_client_array *arrays[]) -{ - GLuint i; - - for (i = 0; i < VERT_ATTRIB_MAX; i++) - if (arrays[i]->StrideB && - !arrays[i]->InstanceDivisor && - !_mesa_is_bufferobj(arrays[i]->BufferObj)) - return GL_FALSE; - - return GL_TRUE; -} - - -/** - * Examine the active arrays to determine if we have interleaved - * vertex arrays all living in one VBO, or all living in user space. - */ -static GLboolean -is_interleaved_arrays(const struct st_vertex_program *vp, - const struct st_vp_variant *vpv, - const struct gl_client_array **arrays) -{ - GLuint attr; - const struct gl_buffer_object *firstBufObj = NULL; - GLint firstStride = -1; - const GLubyte *firstPtr = NULL; - GLboolean userSpaceBuffer = GL_FALSE; - - for (attr = 0; attr < vpv->num_inputs; attr++) { - const GLuint mesaAttr = vp->index_to_input[attr]; - const struct gl_client_array *array = arrays[mesaAttr]; - const struct gl_buffer_object *bufObj = array->BufferObj; - const GLsizei stride = array->StrideB; /* in bytes */ - - if (attr == 0) { - /* save info about the first array */ - firstStride = stride; - firstPtr = array->Ptr; - firstBufObj = bufObj; - userSpaceBuffer = !bufObj || !bufObj->Name; - } - else { - /* check if other arrays interleave with the first, in same buffer */ - if (stride != firstStride) - return GL_FALSE; /* strides don't match */ - - if (bufObj != firstBufObj) - return GL_FALSE; /* arrays in different VBOs */ - - if (abs(array->Ptr - firstPtr) > firstStride) - return GL_FALSE; /* arrays start too far apart */ - - if ((!_mesa_is_bufferobj(bufObj)) != userSpaceBuffer) - return GL_FALSE; /* mix of VBO and user-space arrays */ - } - } - - return GL_TRUE; -} - - -/** - * Set up for drawing interleaved arrays that all live in one VBO - * or all live in user space. - * \param vbuffer returns vertex buffer info - * \param velements returns vertex element info - * \return GL_TRUE for success, GL_FALSE otherwise (probably out of memory) - */ -static GLboolean -setup_interleaved_attribs(struct gl_context *ctx, - const struct st_vertex_program *vp, - const struct st_vp_variant *vpv, - const struct gl_client_array **arrays, - struct pipe_vertex_buffer *vbuffer, - struct pipe_vertex_element velements[]) -{ - GLuint attr; - const GLubyte *low_addr = NULL; - GLboolean usingVBO; /* all arrays in a VBO? */ - struct gl_buffer_object *bufobj; - GLsizei stride; - - /* Find the lowest address of the arrays we're drawing, - * Init bufobj and stride. - */ - if (vpv->num_inputs) { - const GLuint mesaAttr0 = vp->index_to_input[0]; - const struct gl_client_array *array = arrays[mesaAttr0]; - - /* Since we're doing interleaved arrays, we know there'll be at most - * one buffer object and the stride will be the same for all arrays. - * Grab them now. - */ - bufobj = array->BufferObj; - stride = array->StrideB; - - low_addr = arrays[vp->index_to_input[0]]->Ptr; - - for (attr = 1; attr < vpv->num_inputs; attr++) { - const GLubyte *start = arrays[vp->index_to_input[attr]]->Ptr; - low_addr = MIN2(low_addr, start); - } - } - else { - /* not sure we'll ever have zero inputs, but play it safe */ - bufobj = NULL; - stride = 0; - low_addr = 0; - } - - /* are the arrays in user space? */ - usingVBO = _mesa_is_bufferobj(bufobj); - - for (attr = 0; attr < vpv->num_inputs; attr++) { - const GLuint mesaAttr = vp->index_to_input[attr]; - const struct gl_client_array *array = arrays[mesaAttr]; - unsigned src_offset = (unsigned) (array->Ptr - low_addr); - GLuint element_size = array->_ElementSize; - - assert(element_size == array->Size * _mesa_sizeof_type(array->Type)); - - velements[attr].src_offset = src_offset; - velements[attr].instance_divisor = array->InstanceDivisor; - velements[attr].vertex_buffer_index = 0; - velements[attr].src_format = st_pipe_vertex_format(array->Type, - array->Size, - array->Format, - array->Normalized, - array->Integer); - assert(velements[attr].src_format); - } - - /* - * Return the vbuffer info and setup user-space attrib info, if needed. - */ - if (vpv->num_inputs == 0) { - /* just defensive coding here */ - vbuffer->buffer = NULL; - vbuffer->user_buffer = NULL; - vbuffer->buffer_offset = 0; - vbuffer->stride = 0; - } - else if (usingVBO) { - /* all interleaved arrays in a VBO */ - struct st_buffer_object *stobj = st_buffer_object(bufobj); - - if (!stobj || !stobj->buffer) { - /* probably out of memory (or zero-sized buffer) */ - return GL_FALSE; - } - - vbuffer->buffer = stobj->buffer; - vbuffer->user_buffer = NULL; - vbuffer->buffer_offset = pointer_to_offset(low_addr); - vbuffer->stride = stride; - } - else { - /* all interleaved arrays in user memory */ - vbuffer->buffer = NULL; - vbuffer->user_buffer = low_addr; - vbuffer->buffer_offset = 0; - vbuffer->stride = stride; - } - - return GL_TRUE; -} - - -/** - * Set up a separate pipe_vertex_buffer and pipe_vertex_element for each - * vertex attribute. - * \param vbuffer returns vertex buffer info - * \param velements returns vertex element info - * \return GL_TRUE for success, GL_FALSE otherwise (probably out of memory) - */ -static GLboolean -setup_non_interleaved_attribs(struct gl_context *ctx, - const struct st_vertex_program *vp, - const struct st_vp_variant *vpv, - const struct gl_client_array **arrays, - struct pipe_vertex_buffer vbuffer[], - struct pipe_vertex_element velements[]) -{ - GLuint attr; - - for (attr = 0; attr < vpv->num_inputs; attr++) { - const GLuint mesaAttr = vp->index_to_input[attr]; - const struct gl_client_array *array = arrays[mesaAttr]; - struct gl_buffer_object *bufobj = array->BufferObj; - GLsizei stride = array->StrideB; - - assert(array->_ElementSize == array->Size * _mesa_sizeof_type(array->Type)); - - if (_mesa_is_bufferobj(bufobj)) { - /* Attribute data is in a VBO. - * Recall that for VBOs, the gl_client_array->Ptr field is - * really an offset from the start of the VBO, not a pointer. - */ - struct st_buffer_object *stobj = st_buffer_object(bufobj); - - if (!stobj || !stobj->buffer) { - /* probably out of memory (or zero-sized buffer) */ - return GL_FALSE; - } - - vbuffer[attr].buffer = stobj->buffer; - vbuffer[attr].user_buffer = NULL; - vbuffer[attr].buffer_offset = pointer_to_offset(array->Ptr); - } - else { - /* wrap user data */ - void *ptr; - - if (array->Ptr) { - ptr = (void *) array->Ptr; - } - else { - /* no array, use ctx->Current.Attrib[] value */ - ptr = (void *) ctx->Current.Attrib[mesaAttr]; - stride = 0; - } - - assert(ptr); - - vbuffer[attr].buffer = NULL; - vbuffer[attr].user_buffer = ptr; - vbuffer[attr].buffer_offset = 0; - } - - /* common-case setup */ - vbuffer[attr].stride = stride; /* in bytes */ - - velements[attr].src_offset = 0; - velements[attr].instance_divisor = array->InstanceDivisor; - velements[attr].vertex_buffer_index = attr; - velements[attr].src_format = st_pipe_vertex_format(array->Type, - array->Size, - array->Format, - array->Normalized, - array->Integer); - assert(velements[attr].src_format); - } - - return GL_TRUE; -} - - -static void -setup_index_buffer(struct st_context *st, - const struct _mesa_index_buffer *ib, - struct pipe_index_buffer *ibuffer) -{ - struct gl_buffer_object *bufobj = ib->obj; - - ibuffer->index_size = vbo_sizeof_ib_type(ib->type); - - /* get/create the index buffer object */ - if (_mesa_is_bufferobj(bufobj)) { - /* indices are in a real VBO */ - ibuffer->buffer = st_buffer_object(bufobj)->buffer; - ibuffer->offset = pointer_to_offset(ib->ptr); - } - else if (st->indexbuf_uploader) { - u_upload_data(st->indexbuf_uploader, 0, ib->count * ibuffer->index_size, - ib->ptr, &ibuffer->offset, &ibuffer->buffer); - } - else { - /* indices are in user space memory */ - ibuffer->user_buffer = ib->ptr; - } - - cso_set_index_buffer(st->cso_context, ibuffer); -} - - -/** - * Prior to drawing, check that any uniforms referenced by the - * current shader have been set. If a uniform has not been set, - * issue a warning. - */ -static void -check_uniforms(struct gl_context *ctx) -{ - struct gl_shader_program *shProg[3] = { - ctx->Shader.CurrentVertexProgram, - ctx->Shader.CurrentGeometryProgram, - ctx->Shader.CurrentFragmentProgram, - }; - unsigned j; - - for (j = 0; j < 3; j++) { - unsigned i; - - if (shProg[j] == NULL || !shProg[j]->LinkStatus) - continue; - - for (i = 0; i < shProg[j]->NumUserUniformStorage; i++) { - const struct gl_uniform_storage *u = &shProg[j]->UniformStorage[i]; - if (!u->initialized) { - _mesa_warning(ctx, - "Using shader with uninitialized uniform: %s", - u->name); - } - } - } -} - - -/* - * Notes on primitive restart: - * The code below is used when the gallium driver does not support primitive - * restart itself. We map the index buffer, find the restart indexes, unmap - * the index buffer then draw the sub-primitives delineated by the restarts. - * A couple possible optimizations: - * 1. Save the list of sub-primitive (start, count) values in a list attached - * to the index buffer for re-use in subsequent draws. The list would be - * invalidated when the contents of the buffer changed. - * 2. If drawing triangle strips or quad strips, create a new index buffer - * that uses duplicated vertices to render the disjoint strips as one - * long strip. We'd have to be careful to avoid using too much memory - * for this. - * Finally, some apps might perform better if they don't use primitive restart - * at all rather than this fallback path. Set MESA_EXTENSION_OVERRIDE to - * "-GL_NV_primitive_restart" to test that. - */ - - -struct sub_primitive -{ - unsigned start, count; -}; - - -/** - * Scan the elements array to find restart indexes. Return a list - * of primitive (start,count) pairs to indicate how to draw the sub- - * primitives delineated by the restart index. - */ -static struct sub_primitive * -find_sub_primitives(const void *elements, unsigned element_size, - unsigned start, unsigned end, unsigned restart_index, - unsigned *num_sub_prims) -{ - const unsigned max_prims = end - start; - struct sub_primitive *sub_prims; - unsigned i, cur_start, cur_count, num; - - sub_prims = (struct sub_primitive *) - malloc(max_prims * sizeof(struct sub_primitive)); - - if (!sub_prims) { - *num_sub_prims = 0; - return NULL; - } - - cur_start = start; - cur_count = 0; - num = 0; - -#define SCAN_ELEMENTS(TYPE) \ - for (i = start; i < end; i++) { \ - if (((const TYPE *) elements)[i] == restart_index) { \ - if (cur_count > 0) { \ - assert(num < max_prims); \ - sub_prims[num].start = cur_start; \ - sub_prims[num].count = cur_count; \ - num++; \ - } \ - cur_start = i + 1; \ - cur_count = 0; \ - } \ - else { \ - cur_count++; \ - } \ - } \ - if (cur_count > 0) { \ - assert(num < max_prims); \ - sub_prims[num].start = cur_start; \ - sub_prims[num].count = cur_count; \ - num++; \ - } - - switch (element_size) { - case 1: - SCAN_ELEMENTS(ubyte); - break; - case 2: - SCAN_ELEMENTS(ushort); - break; - case 4: - SCAN_ELEMENTS(uint); - break; - default: - assert(0 && "bad index_size in find_sub_primitives()"); - } - -#undef SCAN_ELEMENTS - - *num_sub_prims = num; - - return sub_prims; -} - /** - * For gallium drivers that don't support the primitive restart - * feature, handle it here by breaking up the indexed primitive into - * sub-primitives. + * Set the restart index. */ static void -handle_fallback_primitive_restart(struct cso_context *cso, - struct pipe_context *pipe, - const struct _mesa_index_buffer *ib, - struct pipe_index_buffer *ibuffer, - struct pipe_draw_info *orig_info) +setup_primitive_restart(struct gl_context *ctx, struct pipe_draw_info *info) { - const unsigned start = orig_info->start; - const unsigned count = orig_info->count; - struct pipe_draw_info info = *orig_info; - struct pipe_transfer *transfer = NULL; - unsigned instance, i; - const void *ptr = NULL; - struct sub_primitive *sub_prims; - unsigned num_sub_prims; - - assert(info.indexed); - assert(ibuffer->buffer || ibuffer->user_buffer); - assert(ib); - - if (!ibuffer->buffer || !ibuffer->user_buffer || !ib) - return; + if (ctx->Array._PrimitiveRestart) { + unsigned index_size = info->index_size; - info.primitive_restart = FALSE; - info.instance_count = 1; + info->restart_index = + _mesa_primitive_restart_index(ctx, index_size); - if (_mesa_is_bufferobj(ib->obj)) { - ptr = pipe_buffer_map_range(pipe, ibuffer->buffer, - start * ibuffer->index_size, /* start */ - count * ibuffer->index_size, /* length */ - PIPE_TRANSFER_READ, &transfer); - if (!ptr) - return; - - ptr = (uint8_t*)ptr + (ibuffer->offset - start * ibuffer->index_size); - } - else { - ptr = ib->ptr; - if (!ptr) - return; - } - - sub_prims = find_sub_primitives(ptr, ibuffer->index_size, - 0, count, orig_info->restart_index, - &num_sub_prims); - - if (transfer) - pipe_buffer_unmap(pipe, transfer); - - /* Now draw the sub primitives. - * Need to loop over instances as well to preserve draw order. - */ - for (instance = 0; instance < orig_info->instance_count; instance++) { - info.start_instance = instance + orig_info->start_instance; - for (i = 0; i < num_sub_prims; i++) { - info.start = sub_prims[i].start; - info.count = sub_prims[i].count; - if (u_trim_pipe_prim(info.mode, &info.count)) { - cso_draw_vbo(cso, &info); - } - } + /* Enable primitive restart only when the restart index can have an + * effect. This is required for correctness in radeonsi VI support. + * Other hardware may also benefit from taking a faster, non-restart path + * when possible. + */ + if (index_size == 4 || info->restart_index < (1 << (index_size * 8))) + info->primitive_restart = true; } - - if (sub_prims) - free(sub_prims); } @@ -814,208 +98,247 @@ static unsigned translate_prim(const struct gl_context *ctx, unsigned prim) { /* GL prims should match Gallium prims, spot-check a few */ - assert(GL_POINTS == PIPE_PRIM_POINTS); - assert(GL_QUADS == PIPE_PRIM_QUADS); - assert(GL_TRIANGLE_STRIP_ADJACENCY == PIPE_PRIM_TRIANGLE_STRIP_ADJACENCY); - - /* Avoid quadstrips if it's easy to do so: - * Note: it's important to do the correct trimming if we change the - * prim type! We do that wherever this function is called. - */ - if (prim == GL_QUAD_STRIP && - ctx->Light.ShadeModel != GL_FLAT && - ctx->Polygon.FrontMode == GL_FILL && - ctx->Polygon.BackMode == GL_FILL) - prim = GL_TRIANGLE_STRIP; + STATIC_ASSERT(GL_POINTS == PIPE_PRIM_POINTS); + STATIC_ASSERT(GL_QUADS == PIPE_PRIM_QUADS); + STATIC_ASSERT(GL_TRIANGLE_STRIP_ADJACENCY == PIPE_PRIM_TRIANGLE_STRIP_ADJACENCY); + STATIC_ASSERT(GL_PATCHES == PIPE_PRIM_PATCHES); return prim; } - -/** - * Setup vertex arrays and buffers prior to drawing. - * \return GL_TRUE for success, GL_FALSE otherwise (probably out of memory) - */ -static GLboolean -st_validate_varrays(struct gl_context *ctx, - const struct gl_client_array **arrays) +static inline void +prepare_draw(struct st_context *st, struct gl_context *ctx) { - struct st_context *st = st_context(ctx); - const struct st_vertex_program *vp; - const struct st_vp_variant *vpv; - struct pipe_vertex_buffer vbuffer[PIPE_MAX_SHADER_INPUTS]; - struct pipe_vertex_element velements[PIPE_MAX_ATTRIBS]; - unsigned num_vbuffers, num_velements; - - /* must get these after state validation! */ - vp = st->vp; - vpv = st->vp_variant; - - memset(velements, 0, sizeof(struct pipe_vertex_element) * vpv->num_inputs); + /* Mesa core state should have been validated already */ + assert(ctx->NewState == 0x0); - /* - * Setup the vbuffer[] and velements[] arrays. - */ - if (is_interleaved_arrays(vp, vpv, arrays)) { - if (!setup_interleaved_attribs(ctx, vp, vpv, arrays, vbuffer, - velements)) { - return GL_FALSE; - } + if (unlikely(!st->bitmap.cache.empty)) + st_flush_bitmap_cache(st); - num_vbuffers = 1; - num_velements = vpv->num_inputs; - if (num_velements == 0) - num_vbuffers = 0; - } - else { - if (!setup_non_interleaved_attribs(ctx, vp, vpv, arrays, - vbuffer, velements)) { - return GL_FALSE; - } + st_invalidate_readpix_cache(st); - num_vbuffers = vpv->num_inputs; - num_velements = vpv->num_inputs; + /* Validate state. */ + if ((st->dirty | ctx->NewDriverState) & ST_PIPELINE_RENDER_STATE_MASK || + st->gfx_shaders_may_be_dirty) { + st_validate_state(st, ST_PIPELINE_RENDER); } - - cso_set_vertex_buffers(st->cso_context, num_vbuffers, vbuffer); - cso_set_vertex_elements(st->cso_context, num_velements, velements); - - return GL_TRUE; } - /** * This function gets plugged into the VBO module and is called when * we have something to render. * Basically, translate the information into the format expected by gallium. */ -void +static void st_draw_vbo(struct gl_context *ctx, - const struct gl_client_array **arrays, const struct _mesa_prim *prims, GLuint nr_prims, const struct _mesa_index_buffer *ib, GLboolean index_bounds_valid, GLuint min_index, GLuint max_index, - struct gl_transform_feedback_object *tfb_vertcount) + struct gl_transform_feedback_object *tfb_vertcount, + unsigned stream, + struct gl_buffer_object *indirect) { struct st_context *st = st_context(ctx); - struct pipe_context *pipe = st->pipe; - struct pipe_index_buffer ibuffer = {0}; struct pipe_draw_info info; unsigned i; - GLboolean new_array = - st->dirty.st && - (st->dirty.mesa & (_NEW_ARRAY | _NEW_PROGRAM | _NEW_BUFFER_OBJECT)) != 0; + unsigned start = 0; - /* Mesa core state should have been validated already */ - assert(ctx->NewState == 0x0); + prepare_draw(st, ctx); - /* Validate state. */ - if (st->dirty.st) { - GLboolean vertDataEdgeFlags; - - vertDataEdgeFlags = arrays[VERT_ATTRIB_EDGEFLAG]->BufferObj && - arrays[VERT_ATTRIB_EDGEFLAG]->BufferObj->Name; - if (vertDataEdgeFlags != st->vertdata_edgeflags) { - st->vertdata_edgeflags = vertDataEdgeFlags; - st->dirty.st |= ST_NEW_EDGEFLAGS_DATA; - } + if (st->vertex_array_out_of_memory) + return; - st_validate_state(st); + /* Initialize pipe_draw_info. */ + info.primitive_restart = false; + info.vertices_per_patch = ctx->TessCtrlProgram.patch_vertices; + info.indirect = NULL; + info.count_from_stream_output = NULL; - if (new_array) { - if (!st_validate_varrays(ctx, arrays)) { - /* probably out of memory, no-op the draw call */ - return; - } - } + if (ib) { + struct gl_buffer_object *bufobj = ib->obj; -#if 0 - if (MESA_VERBOSE & VERBOSE_GLSL) { - check_uniforms(ctx); + /* Get index bounds for user buffers. */ + if (!index_bounds_valid && st->draw_needs_minmax_index) { + vbo_get_minmax_indices(ctx, prims, ib, &min_index, &max_index, + nr_prims); } -#else - (void) check_uniforms; -#endif - } - util_draw_init_info(&info); - if (ib) { - /* Get index bounds for user buffers. */ - if (!index_bounds_valid) - if (!all_varyings_in_vbos(arrays)) - vbo_get_minmax_indices(ctx, prims, ib, &min_index, &max_index, - nr_prims); + info.index_size = ib->index_size; + info.min_index = min_index; + info.max_index = max_index; - setup_index_buffer(st, ib, &ibuffer); + if (_mesa_is_bufferobj(bufobj)) { + /* indices are in a real VBO */ + info.has_user_indices = false; + info.index.resource = st_buffer_object(bufobj)->buffer; - info.indexed = TRUE; - if (min_index != ~0 && max_index != ~0) { - info.min_index = min_index; - info.max_index = max_index; + /* Return if the bound element array buffer doesn't have any backing + * storage. (nothing to do) + */ + if (!info.index.resource) + return; + + start = pointer_to_offset(ib->ptr) / info.index_size; + } else { + /* indices are in user space memory */ + info.has_user_indices = true; + info.index.user = ib->ptr; } - /* The VBO module handles restart for the non-indexed GLDrawArrays - * so we only set these fields for indexed drawing: - */ - info.primitive_restart = ctx->Array.PrimitiveRestart; - info.restart_index = ctx->Array.RestartIndex; + setup_primitive_restart(ctx, &info); } else { + info.index_size = 0; + info.has_user_indices = false; + /* Transform feedback drawing is always non-indexed. */ /* Set info.count_from_stream_output. */ if (tfb_vertcount) { - st_transform_feedback_draw_init(tfb_vertcount, &info); + if (!st_transform_feedback_draw_init(tfb_vertcount, stream, &info)) + return; } } + assert(!indirect); + /* do actual drawing */ for (i = 0; i < nr_prims; i++) { - info.mode = translate_prim( ctx, prims[i].mode ); - info.start = prims[i].start; info.count = prims[i].count; + + /* Skip no-op draw calls. */ + if (!info.count && !tfb_vertcount) + continue; + + info.mode = translate_prim(ctx, prims[i].mode); + info.start = start + prims[i].start; + info.start_instance = prims[i].base_instance; info.instance_count = prims[i].num_instances; info.index_bias = prims[i].basevertex; + info.drawid = prims[i].draw_id; if (!ib) { info.min_index = info.start; info.max_index = info.start + info.count - 1; } - if (info.count_from_stream_output) { - cso_draw_vbo(st->cso_context, &info); - } - else if (info.primitive_restart) { - if (st->sw_primitive_restart) { - /* Handle primitive restart for drivers that doesn't support it */ - handle_fallback_primitive_restart(st->cso_context, pipe, ib, - &ibuffer, &info); - } - else { - /* don't trim, restarts might be inside index list */ - cso_draw_vbo(st->cso_context, &info); - } + if (ST_DEBUG & DEBUG_DRAW) { + debug_printf("st/draw: mode %s start %u count %u index_size %d\n", + u_prim_name(info.mode), + info.start, + info.count, + info.index_size); } - else if (u_trim_pipe_prim(info.mode, &info.count)) - cso_draw_vbo(st->cso_context, &info); + + /* Don't call u_trim_pipe_prim. Drivers should do it if they need it. */ + cso_draw_vbo(st->cso_context, &info); } +} - if (ib && st->indexbuf_uploader && !_mesa_is_bufferobj(ib->obj)) { - pipe_resource_reference(&ibuffer.buffer, NULL); +static void +st_indirect_draw_vbo(struct gl_context *ctx, + GLuint mode, + struct gl_buffer_object *indirect_data, + GLsizeiptr indirect_offset, + unsigned draw_count, + unsigned stride, + struct gl_buffer_object *indirect_draw_count, + GLsizeiptr indirect_draw_count_offset, + const struct _mesa_index_buffer *ib) +{ + struct st_context *st = st_context(ctx); + struct pipe_draw_info info; + struct pipe_draw_indirect_info indirect; + + assert(stride); + prepare_draw(st, ctx); + + if (st->vertex_array_out_of_memory) + return; + + memset(&indirect, 0, sizeof(indirect)); + util_draw_init_info(&info); + info.start = 0; /* index offset / index size */ + + if (ib) { + struct gl_buffer_object *bufobj = ib->obj; + + /* indices are always in a real VBO */ + assert(_mesa_is_bufferobj(bufobj)); + + info.index_size = ib->index_size; + info.index.resource = st_buffer_object(bufobj)->buffer; + info.start = pointer_to_offset(ib->ptr) / info.index_size; + + /* Primitive restart is not handled by the VBO module in this case. */ + setup_primitive_restart(ctx, &info); + } + + info.mode = translate_prim(ctx, mode); + info.vertices_per_patch = ctx->TessCtrlProgram.patch_vertices; + info.indirect = &indirect; + indirect.buffer = st_buffer_object(indirect_data)->buffer; + indirect.offset = indirect_offset; + + if (ST_DEBUG & DEBUG_DRAW) { + debug_printf("st/draw indirect: mode %s drawcount %d index_size %d\n", + u_prim_name(info.mode), + draw_count, + info.index_size); + } + + if (!st->has_multi_draw_indirect) { + int i; + + assert(!indirect_draw_count); + indirect.draw_count = 1; + for (i = 0; i < draw_count; i++) { + info.drawid = i; + cso_draw_vbo(st->cso_context, &info); + indirect.offset += stride; + } + } else { + indirect.draw_count = draw_count; + indirect.stride = stride; + if (indirect_draw_count) { + indirect.indirect_draw_count = + st_buffer_object(indirect_draw_count)->buffer; + indirect.indirect_draw_count_offset = indirect_draw_count_offset; + } + cso_draw_vbo(st->cso_context, &info); } } void -st_init_draw(struct st_context *st) +st_init_draw_functions(struct dd_function_table *functions) { - struct gl_context *ctx = st->ctx; + functions->Draw = st_draw_vbo; + functions->DrawIndirect = st_indirect_draw_vbo; +} - vbo_set_draw_func(ctx, st_draw_vbo); -#if FEATURE_feedback || FEATURE_rastpos - st->draw = draw_create(st->pipe); /* for selection/feedback */ +void +st_destroy_draw(struct st_context *st) +{ + draw_destroy(st->draw); +} + +/** + * Getter for the draw_context, so that initialization of it can happen only + * when needed (the TGSI exec machines take up quite a bit of memory). + */ +struct draw_context * +st_get_draw_context(struct st_context *st) +{ + if (!st->draw) { + st->draw = draw_create(st->pipe); + if (!st->draw) { + _mesa_error(st->ctx, GL_OUT_OF_MEMORY, "feedback fallback allocation"); + return NULL; + } + } /* Disable draw options that might convert points/lines to tris, etc. * as that would foul-up feedback/selection mode. @@ -1024,14 +347,88 @@ st_init_draw(struct st_context *st) draw_wide_point_threshold(st->draw, 1000.0f); draw_enable_line_stipple(st->draw, FALSE); draw_enable_point_sprites(st->draw, FALSE); -#endif -} + return st->draw; +} -void -st_destroy_draw(struct st_context *st) +/** + * Draw a quad with given position, texcoords and color. + */ +bool +st_draw_quad(struct st_context *st, + float x0, float y0, float x1, float y1, float z, + float s0, float t0, float s1, float t1, + const float *color, + unsigned num_instances) { -#if FEATURE_feedback || FEATURE_rastpos - draw_destroy(st->draw); -#endif + struct pipe_vertex_buffer vb = {0}; + struct st_util_vertex *verts; + + vb.stride = sizeof(struct st_util_vertex); + + u_upload_alloc(st->pipe->stream_uploader, 0, + 4 * sizeof(struct st_util_vertex), 4, + &vb.buffer_offset, &vb.buffer.resource, (void **) &verts); + if (!vb.buffer.resource) { + return false; + } + + /* lower-left */ + verts[0].x = x0; + verts[0].y = y1; + verts[0].z = z; + verts[0].r = color[0]; + verts[0].g = color[1]; + verts[0].b = color[2]; + verts[0].a = color[3]; + verts[0].s = s0; + verts[0].t = t0; + + /* lower-right */ + verts[1].x = x1; + verts[1].y = y1; + verts[1].z = z; + verts[1].r = color[0]; + verts[1].g = color[1]; + verts[1].b = color[2]; + verts[1].a = color[3]; + verts[1].s = s1; + verts[1].t = t0; + + /* upper-right */ + verts[2].x = x1; + verts[2].y = y0; + verts[2].z = z; + verts[2].r = color[0]; + verts[2].g = color[1]; + verts[2].b = color[2]; + verts[2].a = color[3]; + verts[2].s = s1; + verts[2].t = t1; + + /* upper-left */ + verts[3].x = x0; + verts[3].y = y0; + verts[3].z = z; + verts[3].r = color[0]; + verts[3].g = color[1]; + verts[3].b = color[2]; + verts[3].a = color[3]; + verts[3].s = s0; + verts[3].t = t1; + + u_upload_unmap(st->pipe->stream_uploader); + + cso_set_vertex_buffers(st->cso_context, 0, 1, &vb); + + if (num_instances > 1) { + cso_draw_arrays_instanced(st->cso_context, PIPE_PRIM_TRIANGLE_FAN, 0, 4, + 0, num_instances); + } else { + cso_draw_arrays(st->cso_context, PIPE_PRIM_TRIANGLE_FAN, 0, 4); + } + + pipe_resource_reference(&vb.buffer.resource, NULL); + + return true; }