耗时问题

[0x1306a94]

• 我的场景是在Linux服务器上动态生成海报，统计相关阶段耗时发现在Linux平台耗时高
• 测试工程 pag_cpp_test
• macOS M1 环境下

==============Performance Report Begin==============
PAGFile Load: 8ms
TemplateData Parser: 0ms
PAGSurface MakeOffscreen: 60ms
Fill data: 1ms
Processor BeforeCreatingPlayer: 0ms
PAGPlayer setup: 0ms
Processor AfterCreatingPlayer: 26ms
GetFrameDataAt: 7ms
==============Performance Report End==============

• Linux 环境 Ubuntu 18.04.6 LTS Intel(R) Xeon(R) CPU E5-2620 v3 @ 2.40GHz 32G 开发环境

==============Performance Report Begin==============
PAGFile Load: 0ms
TemplateData Parser: 0ms
PAGSurface MakeOffscreen: 2ms
Fill data: 0ms
Processor BeforeCreatingPlayer: 0ms
PAGPlayer setup: 0ms
Processor AfterCreatingPlayer: 330ms
==============Performance Report End==============

• 发现 Linux平台上 AfterCreatingPlayer处理阶段在开发环境以及生产环境耗时都很高。
• 代码如下，其中 priceLayer 是模版中的图层，symbolLayer是后添加的。
• 因为需要实现前一个符号和后面的文本字体大小不一致，所以采用代码中动态添加符号图层，然后通过Matrix来调整位置。
• 这个耗时应该如何优化

void PosterProcessor::AfterCreatingPlayer(std::shared_ptr<pag::PAGComposition> composition, std::shared_ptr<pag::PAGPlayer> player) {
    if (composition == nullptr || player == nullptr) {
        return;
    }

    if (this->symbolLayer == nullptr || this->priceLayer == nullptr) {
        return;
    }

    this->symbolLayer->resetMatrix();
    this->priceLayer->resetMatrix();

    // 获取最终的准确的图层位置
    player->flush();

    auto canvasRect = pag::Rect::MakeWH(composition->width(), composition->height());
    auto symbolRect = player->getBounds(symbolLayer);
    auto priceRect = player->getBounds(priceLayer);

    auto priceTx = canvasRect.centerX() - priceRect.centerX() + symbolRect.width() * 0.5;
    auto priceTy = canvasRect.centerY() - priceRect.centerY();
    auto symbolTx = (canvasRect.width() - priceRect.width()) * 0.5 - symbolRect.width() * 0.5;
    auto symbolTy = (canvasRect.centerY() + priceRect.height() * 0.5) - symbolRect.height() * 0.8;

    do {
        auto totalMatrix = this->symbolLayer->getTotalMatrix();
        totalMatrix.setTranslate(symbolTx, symbolTy);
        this->symbolLayer->setMatrix(totalMatrix);
    } while (0);

    do {
        auto totalMatrix = this->priceLayer->getTotalMatrix();
        totalMatrix.setTranslate(priceTx, priceTy);
        this->priceLayer->setMatrix(totalMatrix);
    } while (0);
}

[0x1306a94]

我应该知道原因了，应该是Linux环境没有GPU设备，所以采用了CPU模拟导致耗时高的原因