blob: 1ac1d8967db3902a751ca3f8f0cde123f6ab2de8 [file] [log] [blame]
// Copyright (c) 2012 The Chromium Authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
// This file defines the GLES2 command buffer commands.
#ifndef GPU_COMMAND_BUFFER_COMMON_GLES2_CMD_FORMAT_H_
#define GPU_COMMAND_BUFFER_COMMON_GLES2_CMD_FORMAT_H_
#include <KHR/khrplatform.h>
#include <stddef.h>
#include <stdint.h>
#include <string.h>
#include "base/atomicops.h"
#include "base/logging.h"
#include "base/macros.h"
#include "gpu/command_buffer/common/bitfield_helpers.h"
#include "gpu/command_buffer/common/cmd_buffer_common.h"
#include "gpu/command_buffer/common/constants.h"
#include "gpu/command_buffer/common/gles2_cmd_ids.h"
#include "gpu/command_buffer/common/gles2_cmd_utils.h"
// GL types are forward declared to avoid including the GL headers. The problem
// is determining which GL headers to include from code that is common to the
// client and service sides (GLES2 or one of several GL implementations).
typedef unsigned int GLenum;
typedef unsigned int GLbitfield;
typedef unsigned int GLuint;
typedef int GLint;
typedef int GLsizei;
typedef unsigned char GLboolean;
typedef signed char GLbyte;
typedef short GLshort;
typedef unsigned char GLubyte;
typedef unsigned short GLushort;
typedef unsigned long GLulong;
typedef float GLfloat;
typedef float GLclampf;
typedef double GLdouble;
typedef double GLclampd;
typedef void GLvoid;
typedef khronos_intptr_t GLintptr;
typedef khronos_ssize_t GLsizeiptr;
typedef struct __GLsync *GLsync;
typedef int64_t GLint64;
typedef uint64_t GLuint64;
namespace gpu {
namespace gles2 {
// Command buffer is GPU_COMMAND_BUFFER_ENTRY_ALIGNMENT byte aligned.
#pragma pack(push, 4)
static_assert(GPU_COMMAND_BUFFER_ENTRY_ALIGNMENT == 4,
"pragma pack alignment must be equal to "
"GPU_COMMAND_BUFFER_ENTRY_ALIGNMENT");
namespace id_namespaces {
// These are used when contexts share resources.
enum IdNamespaces {
kBuffers,
kFramebuffers,
kProgramsAndShaders,
kRenderbuffers,
kTextures,
kQueries,
kVertexArrays,
kSamplers,
kTransformFeedbacks,
kSyncs,
kNumIdNamespaces
};
enum RangeIdNamespaces { kPaths, kNumRangeIdNamespaces };
// These numbers must not change
static_assert(kBuffers == 0, "kBuffers should equal 0");
static_assert(kFramebuffers == 1, "kFramebuffers should equal 1");
static_assert(kProgramsAndShaders == 2, "kProgramsAndShaders should equal 2");
static_assert(kRenderbuffers == 3, "kRenderbuffers should equal 3");
static_assert(kTextures == 4, "kTextures should equal 4");
static_assert(kPaths == 0, "kPaths should equal 0");
} // namespace id_namespaces
// Used for some glGetXXX commands that return a result through a pointer. We
// need to know if the command succeeded or not and the size of the result. If
// the command failed its result size will 0.
template <typename T>
struct SizedResult {
typedef T Type;
T* GetData() {
return static_cast<T*>(static_cast<void*>(&data));
}
// Returns the total size in bytes of the SizedResult for a given number of
// results including the size field.
static size_t ComputeSize(size_t num_results) {
return sizeof(T) * num_results + sizeof(uint32_t); // NOLINT
}
// Returns the maximum number of results for a given buffer size.
static uint32_t ComputeMaxResults(size_t size_of_buffer) {
return (size_of_buffer >= sizeof(uint32_t)) ?
((size_of_buffer - sizeof(uint32_t)) / sizeof(T)) : 0; // NOLINT
}
// Set the size for a given number of results.
void SetNumResults(size_t num_results) {
size = sizeof(T) * num_results; // NOLINT
}
// Get the number of elements in the result
int32_t GetNumResults() const {
return size / sizeof(T); // NOLINT
}
// Copy the result.
void CopyResult(void* dst) const {
memcpy(dst, &data, size);
}
uint32_t size; // in bytes.
int32_t data; // this is just here to get an offset.
};
static_assert(sizeof(SizedResult<int8_t>) == 8,
"size of SizedResult<int8_t> should be 8");
static_assert(offsetof(SizedResult<int8_t>, size) == 0,
"offset of SizedResult<int8_t>.size should be 0");
static_assert(offsetof(SizedResult<int8_t>, data) == 4,
"offset of SizedResult<int8_t>.data should be 4");
// The data for one attrib or uniform from GetProgramInfoCHROMIUM.
struct ProgramInput {
uint32_t type; // The type (GL_VEC3, GL_MAT3, GL_SAMPLER_2D, etc.
int32_t size; // The size (how big the array is for uniforms)
uint32_t location_offset; // offset from ProgramInfoHeader to 'size'
// locations for uniforms, 1 for attribs.
uint32_t name_offset; // offset from ProgrmaInfoHeader to start of name.
uint32_t name_length; // length of the name.
};
// The format of the bucket filled out by GetProgramInfoCHROMIUM
struct ProgramInfoHeader {
uint32_t link_status;
uint32_t num_attribs;
uint32_t num_uniforms;
// ProgramInput inputs[num_attribs + num_uniforms];
};
// The data for one UniformBlock from GetProgramInfoCHROMIUM
struct UniformBlockInfo {
uint32_t binding; // UNIFORM_BLOCK_BINDING
uint32_t data_size; // UNIFORM_BLOCK_DATA_SIZE
uint32_t name_offset; // offset from UniformBlocksHeader to start of name.
uint32_t name_length; // UNIFORM_BLOCK_NAME_LENGTH
uint32_t active_uniforms; // UNIFORM_BLOCK_ACTIVE_UNIFORMS
// offset from UniformBlocksHeader to |active_uniforms| indices.
uint32_t active_uniform_offset;
// UNIFORM_BLOCK_REFERENDED_BY_VERTEX_SHADER
uint32_t referenced_by_vertex_shader;
// UNIFORM_BLOCK_REFERENDED_BY_FRAGMENT_SHADER
uint32_t referenced_by_fragment_shader;
};
// The format of the bucket filled out by GetUniformBlocksCHROMIUM
struct UniformBlocksHeader {
uint32_t num_uniform_blocks;
// UniformBlockInfo uniform_blocks[num_uniform_blocks];
};
// The data for one TransformFeedbackVarying from
// GetTransformFeedbackVaringCHROMIUM.
struct TransformFeedbackVaryingInfo {
uint32_t size;
uint32_t type;
uint32_t name_offset; // offset from Header to start of name.
uint32_t name_length; // including the null terminator.
};
// The format of the bucket filled out by GetTransformFeedbackVaryingsCHROMIUM
struct TransformFeedbackVaryingsHeader {
uint32_t transform_feedback_buffer_mode;
uint32_t num_transform_feedback_varyings;
// TransformFeedbackVaryingInfo varyings[num_transform_feedback_varyings];
};
// Parameters of a uniform that can be queried through glGetActiveUniformsiv,
// but not through glGetActiveUniform.
struct UniformES3Info {
int32_t block_index;
int32_t offset;
int32_t array_stride;
int32_t matrix_stride;
int32_t is_row_major;
};
// The format of the bucket filled out by GetUniformsivES3CHROMIUM
struct UniformsES3Header {
uint32_t num_uniforms;
// UniformES3Info uniforms[num_uniforms];
};
// The format of QuerySync used by EXT_occlusion_query_boolean
struct QuerySync {
void Reset() {
process_count = 0;
result = 0;
}
base::subtle::Atomic32 process_count;
uint64_t result;
};
struct AsyncUploadSync {
void Reset() {
base::subtle::Release_Store(&async_upload_token, 0);
}
void SetAsyncUploadToken(uint32_t token) {
DCHECK_NE(token, 0u);
base::subtle::Release_Store(&async_upload_token, token);
}
bool HasAsyncUploadTokenPassed(uint32_t token) {
DCHECK_NE(token, 0u);
uint32_t current_token = base::subtle::Acquire_Load(&async_upload_token);
return (current_token - token < 0x80000000);
}
base::subtle::Atomic32 async_upload_token;
};
struct DisjointValueSync {
void Reset() {
base::subtle::Release_Store(&disjoint_count, 0);
}
void SetDisjointCount(uint32_t token) {
DCHECK_NE(token, 0u);
base::subtle::Release_Store(&disjoint_count, token);
}
uint32_t GetDisjointCount() {
return base::subtle::Acquire_Load(&disjoint_count);
}
base::subtle::Atomic32 disjoint_count;
};
static_assert(sizeof(QuerySync) == 12, "size of QuerySync should be 12");
static_assert(offsetof(QuerySync, process_count) == 0,
"offset of QuerySync.process_count should be 0");
static_assert(offsetof(QuerySync, result) == 4,
"offset of QuerySync.result should be 4");
static_assert(sizeof(ProgramInput) == 20, "size of ProgramInput should be 20");
static_assert(offsetof(ProgramInput, type) == 0,
"offset of ProgramInput.type should be 0");
static_assert(offsetof(ProgramInput, size) == 4,
"offset of ProgramInput.size should be 4");
static_assert(offsetof(ProgramInput, location_offset) == 8,
"offset of ProgramInput.location_offset should be 8");
static_assert(offsetof(ProgramInput, name_offset) == 12,
"offset of ProgramInput.name_offset should be 12");
static_assert(offsetof(ProgramInput, name_length) == 16,
"offset of ProgramInput.name_length should be 16");
static_assert(sizeof(ProgramInfoHeader) == 12,
"size of ProgramInfoHeader should be 12");
static_assert(offsetof(ProgramInfoHeader, link_status) == 0,
"offset of ProgramInfoHeader.link_status should be 0");
static_assert(offsetof(ProgramInfoHeader, num_attribs) == 4,
"offset of ProgramInfoHeader.num_attribs should be 4");
static_assert(offsetof(ProgramInfoHeader, num_uniforms) == 8,
"offset of ProgramInfoHeader.num_uniforms should be 8");
static_assert(sizeof(UniformBlockInfo) == 32,
"size of UniformBlockInfo should be 32");
static_assert(offsetof(UniformBlockInfo, binding) == 0,
"offset of UniformBlockInfo.binding should be 0");
static_assert(offsetof(UniformBlockInfo, data_size) == 4,
"offset of UniformBlockInfo.data_size should be 4");
static_assert(offsetof(UniformBlockInfo, name_offset) == 8,
"offset of UniformBlockInfo.name_offset should be 8");
static_assert(offsetof(UniformBlockInfo, name_length) == 12,
"offset of UniformBlockInfo.name_length should be 12");
static_assert(offsetof(UniformBlockInfo, active_uniforms) == 16,
"offset of UniformBlockInfo.active_uniforms should be 16");
static_assert(offsetof(UniformBlockInfo, active_uniform_offset) == 20,
"offset of UniformBlockInfo.active_uniform_offset should be 20");
static_assert(offsetof(UniformBlockInfo, referenced_by_vertex_shader) == 24,
"offset of UniformBlockInfo.referenced_by_vertex_shader "
"should be 24");
static_assert(offsetof(UniformBlockInfo, referenced_by_fragment_shader) == 28,
"offset of UniformBlockInfo.referenced_by_fragment_shader "
"should be 28");
static_assert(sizeof(UniformBlocksHeader) == 4,
"size of UniformBlocksHeader should be 4");
static_assert(offsetof(UniformBlocksHeader, num_uniform_blocks) == 0,
"offset of UniformBlocksHeader.num_uniform_blocks should be 0");
namespace cmds {
#include "../common/gles2_cmd_format_autogen.h"
// These are hand written commands.
// TODO(gman): Attempt to make these auto-generated.
struct GenMailboxCHROMIUM {
typedef GenMailboxCHROMIUM ValueType;
static const CommandId kCmdId = kGenMailboxCHROMIUM;
static const cmd::ArgFlags kArgFlags = cmd::kFixed;
static const uint8_t cmd_flags = CMD_FLAG_SET_TRACE_LEVEL(3);
CommandHeader header;
};
struct CreateAndConsumeTextureCHROMIUMImmediate {
typedef CreateAndConsumeTextureCHROMIUMImmediate ValueType;
static const CommandId kCmdId = kCreateAndConsumeTextureCHROMIUMImmediate;
static const cmd::ArgFlags kArgFlags = cmd::kAtLeastN;
static const uint8_t cmd_flags = CMD_FLAG_SET_TRACE_LEVEL(1);
static uint32_t ComputeDataSize() {
return static_cast<uint32_t>(sizeof(GLbyte) * 64); // NOLINT
}
static uint32_t ComputeSize() {
return static_cast<uint32_t>(sizeof(ValueType) +
ComputeDataSize()); // NOLINT
}
void SetHeader(uint32_t size_in_bytes) {
header.SetCmdByTotalSize<ValueType>(size_in_bytes);
}
void Init(GLenum _target, uint32_t _client_id, const GLbyte* _mailbox) {
SetHeader(ComputeSize());
target = _target;
client_id = _client_id;
memcpy(ImmediateDataAddress(this), _mailbox, ComputeDataSize());
}
void* Set(void* cmd,
GLenum _target,
uint32_t _client_id,
const GLbyte* _mailbox) {
static_cast<ValueType*>(cmd)->Init(_target, _client_id, _mailbox);
const uint32_t size = ComputeSize();
return NextImmediateCmdAddressTotalSize<ValueType>(cmd, size);
}
gpu::CommandHeader header;
uint32_t target;
uint32_t client_id;
};
static_assert(sizeof(CreateAndConsumeTextureCHROMIUMImmediate) == 12,
"size of CreateAndConsumeTextureCHROMIUMImmediate should be 12");
static_assert(offsetof(CreateAndConsumeTextureCHROMIUMImmediate, header) == 0,
"offset of CreateAndConsumeTextureCHROMIUMImmediate.header "
"should be 0");
static_assert(offsetof(CreateAndConsumeTextureCHROMIUMImmediate, target) == 4,
"offset of CreateAndConsumeTextureCHROMIUMImmediate.target "
"should be 4");
static_assert(
offsetof(CreateAndConsumeTextureCHROMIUMImmediate, client_id) == 8,
"offset of CreateAndConsumeTextureCHROMIUMImmediate.client_id should be 8");
#pragma pack(pop)
} // namespace cmd
} // namespace gles2
} // namespace gpu
#endif // GPU_COMMAND_BUFFER_COMMON_GLES2_CMD_FORMAT_H_