[26/30] libcamera: software_isp: gpu: Do a synchronous BenchMark print after syncOutput
diff mbox series

Message ID 20260618122245.946138-27-bryan.odonoghue@linaro.org
State New
Headers show
Series
  • RFC/RFT: gpuisp: Multipass with speed optimisations on top
Related show

Commit Message

Bryan O'Donoghue June 18, 2026, 12:22 p.m. UTC 
Doing an optional synchronous print after syncOutput. The GPU has finished
at this point so we can interrogate how long each shader took to execute.

Only really interesting for a debug scenario.

Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>
---
 .../software_isp/software_isp_pipeline_gpu.cpp    | 15 +++++++++++++++
 .../software_isp/software_isp_pipeline_gpu.h      |  2 ++
 2 files changed, 17 insertions(+)

Patch
diff mbox series 

diff --git a/src/libcamera/software_isp/software_isp_pipeline_gpu.cpp b/src/libcamera/software_isp/software_isp_pipeline_gpu.cpp
index 14f6d2b8c..d7bed8837 100644
--- a/src/libcamera/software_isp/software_isp_pipeline_gpu.cpp
+++ b/src/libcamera/software_isp/software_isp_pipeline_gpu.cpp
@@ -122,6 +122,12 @@  int SoftwareIspPipelineGpu::initBayerShaders(PixelFormat inputFormat, PixelForma
 	return ret;
 }
 
+void SoftwareIspPipelineGpu::initShaderBenchMarks(void)
+{
+	gpuIspShaderPassBlcNormalise_.initShaderBenchMark();
+	gpuIspShaderPassDemosiac_.initShaderBenchMark();
+}
+
 int SoftwareIspPipelineGpu::configure(const StreamConfiguration &inputCfg,
 				      const std::vector<std::reference_wrapper<const StreamConfiguration>> &outputCfgs,
 				      [[maybe_unused]] bool ccmEnabled)
@@ -283,6 +289,12 @@  int SoftwareIspPipelineGpu::processGPU(FrameBuffer *input, FrameBuffer *output,
 	return pipelineResult;
 }
 
+void SoftwareIspPipelineGpu::printShaderBenchMarks(void)
+{
+	gpuIspShaderPassBlcNormalise_.printShaderBenchMark();
+	gpuIspShaderPassDemosiac_.printShaderBenchMark();
+}
+
 void SoftwareIspPipelineGpu::process(uint32_t frame, FrameBuffer *input, FrameBuffer *output, const DebayerParams &params)
 {
 	bench_.startFrame();
@@ -323,6 +335,7 @@  void SoftwareIspPipelineGpu::process(uint32_t frame, FrameBuffer *input, FrameBu
 
 	egl_.syncOutput();
 	bench_.finishFrame();
+	printShaderBenchMarks();
 
 	outputBufferReady.emit(output);
 	inputBufferReady.emit(input);
@@ -349,6 +362,8 @@  int SoftwareIspPipelineGpu::start()
 	if (initBayerShaders(inputPixelFormat_, outputPixelFormat_))
 		return -EINVAL;
 
+	initShaderBenchMarks();
+
 	/* Raw bayer input as texture */
 	eglImageBayerIn_ = std::make_unique<eGLImage>(gpuIspShaderPassBlcNormalise_.glFormat_, inputConfig_.stride / gpuIspShaderPassBlcNormalise_.getBytesPerPixel(), height_, inputConfig_.stride, GL_TEXTURE0, 0);
 
diff --git a/src/libcamera/software_isp/software_isp_pipeline_gpu.h b/src/libcamera/software_isp/software_isp_pipeline_gpu.h
index 8a922f658..b32d4cad3 100644
--- a/src/libcamera/software_isp/software_isp_pipeline_gpu.h
+++ b/src/libcamera/software_isp/software_isp_pipeline_gpu.h
@@ -63,6 +63,8 @@  private:
 	int getShaderVariableLocations();
 	void setShaderVariableValues(const DebayerParams &params);
 	int processGPU(FrameBuffer *input, FrameBuffer *output, const DebayerParams &params, std::optional<MappedFrameBuffer> *mappedInputBuffer, std::optional<DmaSyncer> *inputBufferDmaSyncer);
+	void initShaderBenchMarks(void);
+	void printShaderBenchMarks(void);
 
 	/* Pointer to object representing input texture */
 	std::unique_ptr<eGLImage> eglImageBayerIn_;