gl_rasterizer_cache: Partially implement several non-2D texture types.

This commit is contained in:
bunnei 2018-09-06 19:08:45 -04:00
parent 0731383124
commit f165a85398

View File

@ -245,7 +245,8 @@ static bool IsFormatBCn(PixelFormat format) {
} }
template <bool morton_to_gl, PixelFormat format> template <bool morton_to_gl, PixelFormat format>
void MortonCopy(u32 stride, u32 block_height, u32 height, std::vector<u8>& gl_buffer, VAddr addr) { void MortonCopy(u32 stride, u32 block_height, u32 height, u8* gl_buffer, size_t gl_buffer_size,
VAddr addr) {
constexpr u32 bytes_per_pixel = SurfaceParams::GetFormatBpp(format) / CHAR_BIT; constexpr u32 bytes_per_pixel = SurfaceParams::GetFormatBpp(format) / CHAR_BIT;
constexpr u32 gl_bytes_per_pixel = CachedSurface::GetGLBytesPerPixel(format); constexpr u32 gl_bytes_per_pixel = CachedSurface::GetGLBytesPerPixel(format);
@ -255,18 +256,18 @@ void MortonCopy(u32 stride, u32 block_height, u32 height, std::vector<u8>& gl_bu
const u32 tile_size{IsFormatBCn(format) ? 4U : 1U}; const u32 tile_size{IsFormatBCn(format) ? 4U : 1U};
const std::vector<u8> data = Tegra::Texture::UnswizzleTexture( const std::vector<u8> data = Tegra::Texture::UnswizzleTexture(
addr, tile_size, bytes_per_pixel, stride, height, block_height); addr, tile_size, bytes_per_pixel, stride, height, block_height);
const size_t size_to_copy{std::min(gl_buffer.size(), data.size())}; const size_t size_to_copy{std::min(gl_buffer_size, data.size())};
gl_buffer.assign(data.begin(), data.begin() + size_to_copy); memcpy(gl_buffer, data.data(), size_to_copy);
} else { } else {
// TODO(bunnei): Assumes the default rendering GOB size of 16 (128 lines). We should // TODO(bunnei): Assumes the default rendering GOB size of 16 (128 lines). We should
// check the configuration for this and perform more generic un/swizzle // check the configuration for this and perform more generic un/swizzle
LOG_WARNING(Render_OpenGL, "need to use correct swizzle/GOB parameters!"); LOG_WARNING(Render_OpenGL, "need to use correct swizzle/GOB parameters!");
VideoCore::MortonCopyPixels128(stride, height, bytes_per_pixel, gl_bytes_per_pixel, VideoCore::MortonCopyPixels128(stride, height, bytes_per_pixel, gl_bytes_per_pixel,
Memory::GetPointer(addr), gl_buffer.data(), morton_to_gl); Memory::GetPointer(addr), gl_buffer, morton_to_gl);
} }
} }
static constexpr std::array<void (*)(u32, u32, u32, std::vector<u8>&, VAddr), static constexpr std::array<void (*)(u32, u32, u32, u8*, size_t, VAddr),
SurfaceParams::MaxPixelFormat> SurfaceParams::MaxPixelFormat>
morton_to_gl_fns = { morton_to_gl_fns = {
// clang-format off // clang-format off
@ -323,7 +324,7 @@ static constexpr std::array<void (*)(u32, u32, u32, std::vector<u8>&, VAddr),
// clang-format on // clang-format on
}; };
static constexpr std::array<void (*)(u32, u32, u32, std::vector<u8>&, VAddr), static constexpr std::array<void (*)(u32, u32, u32, u8*, size_t, VAddr),
SurfaceParams::MaxPixelFormat> SurfaceParams::MaxPixelFormat>
gl_to_morton_fns = { gl_to_morton_fns = {
// clang-format off // clang-format off
@ -441,29 +442,51 @@ CachedSurface::CachedSurface(const SurfaceParams& params)
texture.Create(); texture.Create();
const auto& rect{params.GetRect()}; const auto& rect{params.GetRect()};
OpenGLState cur_state = OpenGLState::GetCurState();
// Keep track of previous texture bindings // Keep track of previous texture bindings
GLuint old_tex = cur_state.texture_units[0].texture; OpenGLState cur_state = OpenGLState::GetCurState();
const auto& old_tex = cur_state.texture_units[0];
SCOPE_EXIT({
cur_state.texture_units[0] = old_tex;
cur_state.Apply();
});
cur_state.texture_units[0].texture = texture.handle; cur_state.texture_units[0].texture = texture.handle;
cur_state.texture_units[0].target = SurfaceTargetToGL(params.target);
cur_state.Apply(); cur_state.Apply();
glActiveTexture(GL_TEXTURE0); glActiveTexture(GL_TEXTURE0);
const auto& format_tuple = GetFormatTuple(params.pixel_format, params.component_type); const auto& format_tuple = GetFormatTuple(params.pixel_format, params.component_type);
if (!format_tuple.compressed) { if (!format_tuple.compressed) {
// Only pre-create the texture for non-compressed textures. // Only pre-create the texture for non-compressed textures.
switch (params.target) {
case SurfaceParams::SurfaceTarget::Texture1D:
glTexImage1D(SurfaceTargetToGL(params.target), 0, format_tuple.internal_format,
rect.GetWidth(), 0, format_tuple.format, format_tuple.type, nullptr);
break;
case SurfaceParams::SurfaceTarget::Texture2D:
glTexImage2D(SurfaceTargetToGL(params.target), 0, format_tuple.internal_format,
rect.GetWidth(), rect.GetHeight(), 0, format_tuple.format,
format_tuple.type, nullptr);
break;
case SurfaceParams::SurfaceTarget::Texture3D:
case SurfaceParams::SurfaceTarget::Texture2DArray:
glTexImage3D(SurfaceTargetToGL(params.target), 0, format_tuple.internal_format,
rect.GetWidth(), rect.GetHeight(), params.depth, 0, format_tuple.format,
format_tuple.type, nullptr);
break;
default:
LOG_CRITICAL(Render_OpenGL, "Unimplemented surface target={}",
static_cast<u32>(params.target));
UNREACHABLE();
glTexImage2D(GL_TEXTURE_2D, 0, format_tuple.internal_format, rect.GetWidth(), glTexImage2D(GL_TEXTURE_2D, 0, format_tuple.internal_format, rect.GetWidth(),
rect.GetHeight(), 0, format_tuple.format, format_tuple.type, nullptr); rect.GetHeight(), 0, format_tuple.format, format_tuple.type, nullptr);
} }
}
glTexParameteri(SurfaceTargetToGL(params.target), GL_TEXTURE_MAX_LEVEL, 0); glTexParameteri(SurfaceTargetToGL(params.target), GL_TEXTURE_MAX_LEVEL, 0);
glTexParameteri(SurfaceTargetToGL(params.target), GL_TEXTURE_MIN_FILTER, GL_LINEAR); glTexParameteri(SurfaceTargetToGL(params.target), GL_TEXTURE_MIN_FILTER, GL_LINEAR);
glTexParameteri(SurfaceTargetToGL(params.target), GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE); glTexParameteri(SurfaceTargetToGL(params.target), GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
glTexParameteri(SurfaceTargetToGL(params.target), GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE); glTexParameteri(SurfaceTargetToGL(params.target), GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
// Restore previous texture bindings
cur_state.texture_units[0].texture = old_tex;
cur_state.Apply();
} }
static void ConvertS8Z24ToZ24S8(std::vector<u8>& data, u32 width, u32 height) { static void ConvertS8Z24ToZ24S8(std::vector<u8>& data, u32 width, u32 height) {
@ -548,13 +571,24 @@ void CachedSurface::LoadGLBuffer() {
MICROPROFILE_SCOPE(OpenGL_SurfaceLoad); MICROPROFILE_SCOPE(OpenGL_SurfaceLoad);
if (params.is_tiled) { if (params.is_tiled) {
gl_buffer.resize(copy_size); // TODO(bunnei): This only unswizzles and copies a 2D texture - we do not yet know how to do
// this for 3D textures, etc.
switch (params.target) {
case SurfaceParams::SurfaceTarget::Texture2D:
// Pass impl. to the fallback code below
break;
default:
LOG_CRITICAL(HW_GPU, "Unimplemented tiled load for target={}",
static_cast<u32>(params.target));
UNREACHABLE();
}
gl_buffer.resize(params.depth * copy_size);
morton_to_gl_fns[static_cast<size_t>(params.pixel_format)]( morton_to_gl_fns[static_cast<size_t>(params.pixel_format)](
params.width, params.block_height, params.height, gl_buffer, params.addr); params.width, params.block_height, params.height, gl_buffer.data(), copy_size,
params.addr);
} else { } else {
const u8* const texture_src_data_end = texture_src_data + copy_size; const u8* const texture_src_data_end{texture_src_data + (params.depth * copy_size)};
gl_buffer.assign(texture_src_data, texture_src_data_end); gl_buffer.assign(texture_src_data, texture_src_data_end);
} }
@ -574,7 +608,7 @@ void CachedSurface::UploadGLTexture(GLuint read_fb_handle, GLuint draw_fb_handle
MICROPROFILE_SCOPE(OpenGL_TextureUL); MICROPROFILE_SCOPE(OpenGL_TextureUL);
ASSERT(gl_buffer.size() == ASSERT(gl_buffer.size() ==
params.width * params.height * GetGLBytesPerPixel(params.pixel_format)); params.width * params.height * GetGLBytesPerPixel(params.pixel_format) * params.depth);
const auto& rect{params.GetRect()}; const auto& rect{params.GetRect()};
@ -587,8 +621,13 @@ void CachedSurface::UploadGLTexture(GLuint read_fb_handle, GLuint draw_fb_handle
GLuint target_tex = texture.handle; GLuint target_tex = texture.handle;
OpenGLState cur_state = OpenGLState::GetCurState(); OpenGLState cur_state = OpenGLState::GetCurState();
GLuint old_tex = cur_state.texture_units[0].texture; const auto& old_tex = cur_state.texture_units[0];
SCOPE_EXIT({
cur_state.texture_units[0] = old_tex;
cur_state.Apply();
});
cur_state.texture_units[0].texture = target_tex; cur_state.texture_units[0].texture = target_tex;
cur_state.texture_units[0].target = SurfaceTargetToGL(params.target);
cur_state.Apply(); cur_state.Apply();
// Ensure no bad interactions with GL_UNPACK_ALIGNMENT // Ensure no bad interactions with GL_UNPACK_ALIGNMENT
@ -597,20 +636,62 @@ void CachedSurface::UploadGLTexture(GLuint read_fb_handle, GLuint draw_fb_handle
glActiveTexture(GL_TEXTURE0); glActiveTexture(GL_TEXTURE0);
if (tuple.compressed) { if (tuple.compressed) {
switch (params.target) {
case SurfaceParams::SurfaceTarget::Texture2D:
glCompressedTexImage2D(
SurfaceTargetToGL(params.target), 0, tuple.internal_format,
static_cast<GLsizei>(params.width), static_cast<GLsizei>(params.height), 0,
static_cast<GLsizei>(params.size_in_bytes), &gl_buffer[buffer_offset]);
break;
case SurfaceParams::SurfaceTarget::Texture3D:
case SurfaceParams::SurfaceTarget::Texture2DArray:
glCompressedTexImage3D(
SurfaceTargetToGL(params.target), 0, tuple.internal_format,
static_cast<GLsizei>(params.width), static_cast<GLsizei>(params.height),
static_cast<GLsizei>(params.depth), 0, static_cast<GLsizei>(params.size_in_bytes),
&gl_buffer[buffer_offset]);
break;
default:
LOG_CRITICAL(Render_OpenGL, "Unimplemented surface target={}",
static_cast<u32>(params.target));
UNREACHABLE();
glCompressedTexImage2D( glCompressedTexImage2D(
GL_TEXTURE_2D, 0, tuple.internal_format, static_cast<GLsizei>(params.width), GL_TEXTURE_2D, 0, tuple.internal_format, static_cast<GLsizei>(params.width),
static_cast<GLsizei>(params.height), 0, static_cast<GLsizei>(params.size_in_bytes), static_cast<GLsizei>(params.height), 0, static_cast<GLsizei>(params.size_in_bytes),
&gl_buffer[buffer_offset]); &gl_buffer[buffer_offset]);
}
} else { } else {
switch (params.target) {
case SurfaceParams::SurfaceTarget::Texture1D:
glTexSubImage1D(SurfaceTargetToGL(params.target), 0, x0,
static_cast<GLsizei>(rect.GetWidth()), tuple.format, tuple.type,
&gl_buffer[buffer_offset]);
break;
case SurfaceParams::SurfaceTarget::Texture2D:
glTexSubImage2D(SurfaceTargetToGL(params.target), 0, x0, y0,
static_cast<GLsizei>(rect.GetWidth()),
static_cast<GLsizei>(rect.GetHeight()), tuple.format, tuple.type,
&gl_buffer[buffer_offset]);
break;
case SurfaceParams::SurfaceTarget::Texture3D:
case SurfaceParams::SurfaceTarget::Texture2DArray:
glTexSubImage3D(SurfaceTargetToGL(params.target), 0, x0, y0, 0,
static_cast<GLsizei>(rect.GetWidth()),
static_cast<GLsizei>(rect.GetHeight()), params.depth, tuple.format,
tuple.type, &gl_buffer[buffer_offset]);
break;
default:
LOG_CRITICAL(Render_OpenGL, "Unimplemented surface target={}",
static_cast<u32>(params.target));
UNREACHABLE();
glTexSubImage2D(GL_TEXTURE_2D, 0, x0, y0, static_cast<GLsizei>(rect.GetWidth()), glTexSubImage2D(GL_TEXTURE_2D, 0, x0, y0, static_cast<GLsizei>(rect.GetWidth()),
static_cast<GLsizei>(rect.GetHeight()), tuple.format, tuple.type, static_cast<GLsizei>(rect.GetHeight()), tuple.format, tuple.type,
&gl_buffer[buffer_offset]); &gl_buffer[buffer_offset]);
} }
}
glPixelStorei(GL_UNPACK_ROW_LENGTH, 0); glPixelStorei(GL_UNPACK_ROW_LENGTH, 0);
cur_state.texture_units[0].texture = old_tex;
cur_state.Apply();
} }
RasterizerCacheOpenGL::RasterizerCacheOpenGL() { RasterizerCacheOpenGL::RasterizerCacheOpenGL() {