[Perf] (MG_Backend/DirectGLES): dedup per-draw indexed UBO/SSBO binds with a shadow cache; BindCurrentProgramWithResources 5.1% -> 3.1%

This commit is contained in:
2026-07-12 06:32:18 -04:00
parent 25395a9f9a
commit ad9ee99521
3 changed files with 63 additions and 13 deletions
+15 -13
View File
@@ -208,7 +208,7 @@ namespace MobileGL::MG_Backend::DirectGLES {
auto& point = MG_State::pGLContext->GetBufferBindingPoint(target, i);
auto& obj = point.GetBoundObject();
if (!obj) {
g_GLESFuncs.glBindBufferBase(glTarget, static_cast<GLuint>(i), 0);
BindBufferBaseCached(glTarget, static_cast<GLuint>(i), 0);
continue;
}
@@ -222,12 +222,12 @@ namespace MobileGL::MG_Backend::DirectGLES {
const auto& range = point.GetRange();
auto backendBufferId = backendResource->id;
if (range.start == 0 && range.end >= obj->GetSize()) {
g_GLESFuncs.glBindBufferBase(glTarget, static_cast<GLuint>(i), backendBufferId);
BindBufferBaseCached(glTarget, static_cast<GLuint>(i), backendBufferId);
} else {
const auto start = std::min(range.start, obj->GetSize());
const auto end = std::min(range.end, obj->GetSize());
g_GLESFuncs.glBindBufferRange(glTarget, static_cast<GLuint>(i), backendBufferId,
static_cast<GLintptr>(start), static_cast<GLsizeiptr>(end - start));
BindBufferRangeCached(glTarget, static_cast<GLuint>(i), backendBufferId,
static_cast<GLintptr>(start), static_cast<GLsizeiptr>(end - start));
}
}
}
@@ -1083,7 +1083,7 @@ namespace MobileGL::MG_Backend::DirectGLES {
g_GLESFuncs.glBindBuffer(GL_UNIFORM_BUFFER, 0);
backendProgram.SetLastUploadedGlobalUboVersion(uboContentVersion);
}
g_GLESFuncs.glBindBufferBase(GL_UNIFORM_BUFFER, 0, backendProgram.GetBackendGlobalUBOId());
BufferImpl::BindBufferBaseCached(GL_UNIFORM_BUFFER, 0, backendProgram.GetBackendGlobalUBOId());
}
{
@@ -1111,12 +1111,13 @@ namespace MobileGL::MG_Backend::DirectGLES {
if (bufferObj) {
auto* backendResource = BufferImpl::EnsureBufferResource(bufferObj);
if (backendResource && backendResource->id != 0) {
BufferImpl::BindBufferId(GL_UNIFORM_BUFFER, backendResource->id);
// glBindBufferBase/Range set the generic GL_UNIFORM_BUFFER binding
// as a side effect, so no separate BindBufferId is needed here.
if (range.end == 0) {
g_GLESFuncs.glBindBufferBase(GL_UNIFORM_BUFFER, lastUBOBinding,
backendResource->id);
BufferImpl::BindBufferBaseCached(GL_UNIFORM_BUFFER, lastUBOBinding,
backendResource->id);
} else {
g_GLESFuncs.glBindBufferRange(
BufferImpl::BindBufferRangeCached(
GL_UNIFORM_BUFFER, lastUBOBinding, backendResource->id,
(GLintptr)range.start, (GLintptr)(range.end - range.start));
}
@@ -1235,8 +1236,8 @@ namespace MobileGL::MG_Backend::DirectGLES {
if (paramsBinding >= 0) {
auto* resource = BufferImpl::EnsureBufferResource(drawIndirectBuffer);
if (resource && resource->id != 0) {
g_GLESFuncs.glBindBufferBase(GL_SHADER_STORAGE_BUFFER, static_cast<GLuint>(paramsBinding),
resource->id);
BufferImpl::BindBufferBaseCached(GL_SHADER_STORAGE_BUFFER, static_cast<GLuint>(paramsBinding),
resource->id);
}
}
for (GLsizei i = 0; i < drawcount; ++i) {
@@ -1282,8 +1283,8 @@ namespace MobileGL::MG_Backend::DirectGLES {
if (paramsBinding >= 0) {
auto* resource = BufferImpl::EnsureBufferResource(drawIndirectBuffer);
if (resource && resource->id != 0) {
g_GLESFuncs.glBindBufferBase(GL_SHADER_STORAGE_BUFFER, static_cast<GLuint>(paramsBinding),
resource->id);
BufferImpl::BindBufferBaseCached(GL_SHADER_STORAGE_BUFFER, static_cast<GLuint>(paramsBinding),
resource->id);
}
}
for (GLsizei i = 0; i < drawcount; ++i) {
@@ -3871,6 +3872,7 @@ namespace MobileGL::MG_Backend::DirectGLES {
// Conservatively drop the redundant-glUseProgram guard: re-issuing one bind
// after a MakeCurrent is cheaper than trusting a possibly-reset context.
PrgramImpl::g_lastUsedBackendProgramId = 0;
BufferImpl::InvalidateIndexedBufferBindingCache();
// eglSwapInterval requires a current context; a request made while none was
// current (and dropped by the driver) is retried here.
ApplyRequestedSwapInterval();
@@ -749,6 +749,47 @@ namespace MobileGL::MG_Backend::DirectGLES {
g_boundArrayBufferKnown = false;
}
namespace {
// Shadow of the GL indexed buffer bindings so redundant glBindBufferBase/Range
// (same index + id + range) are skipped. isBase distinguishes a whole-buffer
// base bind from a sub-range bind. Fresh/reset context: every point is base(0)
// == unbound, which matches the GL default.
struct IndexedBufferBinding {
Uint id = 0;
GLintptr offset = 0;
GLsizeiptr size = 0;
Bool isBase = true;
};
constexpr SizeT kMaxIndexedBufferBindings = 64;
IndexedBufferBinding g_indexedUBOBindings[kMaxIndexedBufferBindings];
IndexedBufferBinding g_indexedSSBOBindings[kMaxIndexedBufferBindings];
IndexedBufferBinding* IndexedBindingShadow(GLenum glTarget, Uint index) {
if (index >= kMaxIndexedBufferBindings) return nullptr; // out of range: never cache
if (glTarget == GL_UNIFORM_BUFFER) return &g_indexedUBOBindings[index];
if (glTarget == GL_SHADER_STORAGE_BUFFER) return &g_indexedSSBOBindings[index];
return nullptr;
}
} // namespace
void BindBufferBaseCached(GLenum glTarget, Uint index, Uint id) {
auto* s = IndexedBindingShadow(glTarget, index);
if (s && s->isBase && s->id == id) return;
g_GLESFuncs.glBindBufferBase(glTarget, index, id);
if (s) *s = {id, 0, 0, true};
}
void BindBufferRangeCached(GLenum glTarget, Uint index, Uint id, GLintptr offset, GLsizeiptr size) {
auto* s = IndexedBindingShadow(glTarget, index);
if (s && !s->isBase && s->id == id && s->offset == offset && s->size == size) return;
g_GLESFuncs.glBindBufferRange(glTarget, index, id, offset, size);
if (s) *s = {id, offset, size, false};
}
void InvalidateIndexedBufferBindingCache() {
for (auto& b : g_indexedUBOBindings) b = {};
for (auto& b : g_indexedSSBOBindings) b = {};
}
void TrimBufferPool() {
const std::lock_guard<std::mutex> lock(g_poolMutex);
if (g_pooledBytes <= kMaxPoolBytes) return;
@@ -177,6 +177,13 @@ namespace MobileGL::MG_Backend::DirectGLES {
// glBindBuffer with a redundant-bind cache for GL_ARRAY_BUFFER.
void BindBufferId(GLenum target, Uint id);
void InvalidateArrayBufferBindingCache();
// Redundant-bind cache for INDEXED buffer bindings (glBindBufferBase/Range on
// GL_UNIFORM_BUFFER / GL_SHADER_STORAGE_BUFFER): skips the GL call when the
// (id, range) already at that index matches, like the array-buffer/texture/
// sampler caches already do. Invalidated on MakeCurrent (context may reset).
void BindBufferBaseCached(GLenum glTarget, Uint index, Uint id);
void BindBufferRangeCached(GLenum glTarget, Uint index, Uint id, GLintptr offset, GLsizeiptr size);
void InvalidateIndexedBufferBindingCache();
// Buffer-storage pool maintenance. TrimBufferPool evicts over-budget entries
// (called once per frame from Present); ClearBufferPool drops all pooled ids
// without glDeleteBuffers (called when the ES context is going away).