Merge pull request #1855 from CEED/zach/fix-flops

jeremylt · web-flow · commit 07d8a9fd2c3c · 2025-07-14T05:51:17.000-06:00
Fix grad basis flop counts
diff --git a/interface/ceed-basis.c b/interface/ceed-basis.c
@@ -941,7 +941,7 @@ int CeedBasisGetFlopsEstimate(CeedBasis basis, CeedTransposeMode t_mode, CeedEva
             CeedInt inner_flops =
                 dim * (2 * Q_1d * Q_1d + (t_mode == CEED_TRANSPOSE ? 2 : 3) * Q_1d) + (dim - 1) * (2 * chebyshev_flops + d_chebyshev_flops);
 
-            *flops += num_points * Q_1d * (chebyshev_flops + d_chebyshev_flops) * num_comp * (inner_flops + (t_mode == CEED_TRANSPOSE ? 1 : 0));
+            *flops += num_points * Q_1d * (chebyshev_flops + d_chebyshev_flops + num_comp * (inner_flops + (t_mode == CEED_TRANSPOSE ? 1 : 0)));
           } else {
             *flops += num_points * (is_gpu ? num_comp : 1) * dim * (d_chebyshev_flops + (dim - 1) * chebyshev_flops);
           }

Original file line number	Diff line number	Diff line change
`@@ -941,7 +941,7 @@ int CeedBasisGetFlopsEstimate(CeedBasis basis, CeedTransposeMode t_mode, CeedEva`
`941`	`941`	`CeedInt inner_flops =`
`942`	`942`	`dim * (2 * Q_1d * Q_1d + (t_mode == CEED_TRANSPOSE ? 2 : 3) * Q_1d) + (dim - 1) * (2 * chebyshev_flops + d_chebyshev_flops);`
`943`	`943`
`944`		`- flops += num_points Q_1d * (chebyshev_flops + d_chebyshev_flops) * num_comp * (inner_flops + (t_mode == CEED_TRANSPOSE ? 1 : 0));`
	`944`	`+ flops += num_points Q_1d * (chebyshev_flops + d_chebyshev_flops + num_comp * (inner_flops + (t_mode == CEED_TRANSPOSE ? 1 : 0)));`
`945`	`945`	`} else {`
`946`	`946`	`flops += num_points (is_gpu ? num_comp : 1) * dim * (d_chebyshev_flops + (dim - 1) * chebyshev_flops);`
`947`	`947`	`}`