Lines Matching refs:time

156  * slope of the workload size vs frame time data is estimated. This slope
157 * tells us the estimated increase in frame time caused by a workload
168 * At this point, the result tells us the increase in frame time caused
171 * fragments in a draw call, we get the time cost of one operation.
174 * between workload size and frame time. Instead, there tends to be some
176 * workload sizes below some positive integer C yield the same frame time,
177 * and only workload sizes beyond C increase the frame time in a supposedly
266 STATE_FIND_HIGH_WORKLOAD, //!< Find an appropriate lower bound for the highest workload size we intend to use (one with high-enough frame time compared to workload size 1) for each program.
281 void addFrameTime (float time) { frameTimes.push_back(time); }
293 void prepareWorkload (int progNdx, int workload); //!< Calls setWorkloadSizeUniform and draws, in case the implementation does some draw-time compilation.
297 void adjustAndLogGridAndViewport (void); //!< Log grid and viewport sizes, after possibly reducing them to reduce draw time.
716 << " doesn't give high enough frame time for program " << m_measureProgramNdx
825 1000.0f/30.0f /* frame time (ms) */, 1000.0f/60.0f /* frame time cap (ms) */, 1000.0f /* target measure duration (ms) */));
865 // If call count is just 1, and the target frame time still wasn't reached, reduce grid or viewport size.
875 // Reduce grid or viewport size assuming draw call time scales proportionally.
964 << TestLog::ValueInfo("FrameTime", "Frame time", "us", QP_SAMPLE_VALUE_TAG_RESPONSE)
977 << TestLog::ValueInfo("MedianFrameTime", "Median frame time", "us", QP_SAMPLE_VALUE_TAG_RESPONSE)
1006 // \note For each estimator, .right.coefficient is the increase in draw time (in microseconds) when
1010 // The measurements of any single program can't tell us the final result (time of single operation),
1369 int modifyParamNdx, //!< Add a compile-time constant (2.0) to the parameter at this index. This is ignored if negative.
1406 // compile-time constant (2.0) is added. This is a quick and dirty way to deal with
1417 // is too heavy and dominates time-wise.
2055 // \note In order to reduce case count and thus total execution time, we don't test all input type combinations for every function.