Draft
Conversation
…re Apple M-series
…dup on 100M cell raster
github-actions
bot
added
raster
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
| ncols = Rast_window_cols(); | ||
| if ((region.ew_res / region.ns_res >= 1.01) || | ||
| (region.ns_res / region.ew_res >= 1.01)) { | ||
| G_warning(_("E-W and N-S grid resolutions are different. Taking average.")); |
Contributor
There was a problem hiding this comment.
[pre-commit] reported by reviewdog 🐶
Suggested change
| G_warning(_("E-W and N-S grid resolutions are different. Taking average.")); | |
| G_warning( | |
| _("E-W and N-S grid resolutions are different. Taking average.")); |
| */ | ||
| } | ||
| else | ||
| G_ludcmp(normal_ptr, 6, index_ptr,&temp); |
Contributor
There was a problem hiding this comment.
[pre-commit] reported by reviewdog 🐶
Suggested change
| G_ludcmp(normal_ptr, 6, index_ptr,&temp); | |
| G_ludcmp(normal_ptr, 6, index_ptr, &temp); |
Comment on lines
80
to
+84
| /*-----------------------------------------------------------------------*/ | ||
| /* PROCESS INPUT RASTER AND WRITE OUT RASTER LINE BY LINE */ | ||
|
|
||
| /* Parallel row loop. Each thread gets its own local computation */ | ||
| /* buffers (t_window, t_obs) to avoid write contention. */ | ||
| /*-----------------------------------------------------------------------*/ | ||
|
|
||
| if (mparam != FEATURE) | ||
| for (wind_row = 0; wind_row < EDGE; wind_row++) | ||
| Rast_put_row(fd_out, row_out, | ||
| DCELL_TYPE); /* Write out the edge cells as NULL. */ | ||
| else | ||
| for (wind_row = 0; wind_row < EDGE; wind_row++) | ||
| Rast_put_row(fd_out, featrow_out, | ||
| CELL_TYPE); /* Write out the edge cells as NULL. */ | ||
|
|
||
| for (wind_row = 0; wind_row < wsize - 1; wind_row++) | ||
| Rast_get_row(fd_in, row_in + (wind_row * ncols), wind_row, DCELL_TYPE); | ||
| /* Read in enough of the first rows to */ | ||
| /* allow window to be examined. */ | ||
|
|
||
| #pragma omp parallel for schedule(dynamic) private(row, col, wind_row) |
Contributor
There was a problem hiding this comment.
[pre-commit] reported by reviewdog 🐶
Suggested change
| /*-----------------------------------------------------------------------*/ | |
| /* PROCESS INPUT RASTER AND WRITE OUT RASTER LINE BY LINE */ | |
| /* Parallel row loop. Each thread gets its own local computation */ | |
| /* buffers (t_window, t_obs) to avoid write contention. */ | |
| /*-----------------------------------------------------------------------*/ | |
| if (mparam != FEATURE) | |
| for (wind_row = 0; wind_row < EDGE; wind_row++) | |
| Rast_put_row(fd_out, row_out, | |
| DCELL_TYPE); /* Write out the edge cells as NULL. */ | |
| else | |
| for (wind_row = 0; wind_row < EDGE; wind_row++) | |
| Rast_put_row(fd_out, featrow_out, | |
| CELL_TYPE); /* Write out the edge cells as NULL. */ | |
| for (wind_row = 0; wind_row < wsize - 1; wind_row++) | |
| Rast_get_row(fd_in, row_in + (wind_row * ncols), wind_row, DCELL_TYPE); | |
| /* Read in enough of the first rows to */ | |
| /* allow window to be examined. */ | |
| #pragma omp parallel for schedule(dynamic) private(row, col, wind_row) | |
| /*-----------------------------------------------------------------------*/ | |
| /* Parallel row loop. Each thread gets its own local computation */ | |
| /* buffers (t_window, t_obs) to avoid write contention. */ | |
| /*-----------------------------------------------------------------------*/ | |
| #pragma omp parallel for schedule(dynamic) private(row, col, wind_row) |
Comment on lines
+111
to
+112
| (size_t)(row + wind_row - EDGE) * ncols + | ||
| col + wind_col - EDGE; |
Contributor
There was a problem hiding this comment.
[pre-commit] reported by reviewdog 🐶
Suggested change
| (size_t)(row + wind_row - EDGE) * ncols + | |
| col + wind_col - EDGE; | |
| (size_t)(row + wind_row - EDGE) * ncols + | |
| col + wind_col - EDGE; |
| Rast_set_d_null_value(row_out + col, 1); | ||
| } | ||
| found_null = TRUE; | ||
| Rast_set_c_null_value((CELL *)row_buffers[row] + col, 1); |
Contributor
There was a problem hiding this comment.
[pre-commit] reported by reviewdog 🐶
Suggested change
| Rast_set_c_null_value((CELL *)row_buffers[row] + col, 1); | |
| Rast_set_c_null_value( | |
| (CELL *)row_buffers[row] + col, 1); |
|
|
||
| /* row index in input matrix */ | ||
| double row_idx = (incellhd.north - ycoord1) / incellhd.ns_res; | ||
| if (GPJ_transform(&oproj, &iproj, &tproj, PJ_FWD, &x1, &y1, NULL) < 0) { |
Contributor
There was a problem hiding this comment.
[pre-commit] reported by reviewdog 🐶
Suggested change
| if (GPJ_transform(&oproj, &iproj, &tproj, PJ_FWD, &x1, &y1, NULL) < 0) { | |
| if (GPJ_transform(&oproj, &iproj, &tproj, PJ_FWD, &x1, &y1, NULL) < | |
| 0) { |
| double row_idx = (incellhd.north - ycoord1) / incellhd.ns_res; | ||
| if (GPJ_transform(&oproj, &iproj, &tproj, PJ_FWD, &x1, &y1, NULL) < 0) { | ||
| Rast_set_null_value(obufptr, 1, cell_type); | ||
| } else { |
Contributor
There was a problem hiding this comment.
[pre-commit] reported by reviewdog 🐶
Suggested change
| } else { | |
| } | |
| else { |
| interpolate(ibuffer, obufptr, cell_type, col_idx, row_idx, | ||
| &incellhd); | ||
| /* CALL OUR LOCK-FREE RAM INTERPOLATOR */ | ||
| interpolate_ram(full_map_array, obufptr, cell_type, c_idx, r_idx, &incellhd); |
Contributor
There was a problem hiding this comment.
[pre-commit] reported by reviewdog 🐶
Suggested change
| interpolate_ram(full_map_array, obufptr, cell_type, c_idx, r_idx, &incellhd); | |
| interpolate_ram(full_map_array, obufptr, cell_type, c_idx, | |
| r_idx, &incellhd); |
|
|
||
| xcoord2 = outcellhd.west + (outcellhd.ew_res / 2); | ||
| ycoord2 -= outcellhd.ns_res; | ||
| #pragma omp critical |
Contributor
There was a problem hiding this comment.
[pre-commit] reported by reviewdog 🐶
Suggested change
| #pragma omp critical | |
| #pragma omp critical |
|
|
||
| return buf; | ||
| } | ||
|
|
Contributor
There was a problem hiding this comment.
[pre-commit] reported by reviewdog 🐶
Suggested change
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This draft PR demonstrates a proof of concept parallelization of
r.param.scaleusing OpenMP, submitted as part of my GSoC 2026 proposal for "Parallelization of existing tools."Benchmark Results
100M cell raster, Apple M4, 8 cores — 3 runs each:
1.7x wall-time speedup on a 100M cell raster.
Technical Approach
The core blocker for parallelizing
r.param.scaleis the sequential sliding buffer inprocess(). The original implementation shuffles rows down after each row. Each row depends on the state left by the previous row, making direct row-level parallelism impossible without reading each row's neighborhood from disk multiple times, which would be slower than serial. The solution is the same RAM preload pattern used in the r.proj parallelization: load the entire input raster into a flat 2D array before the parallel region. With the full map in RAM, each output row's neighborhood window can be accessed independently. The outer row loop is parallelized with#pragma omp parallel for schedule(dynamic), with per-threadt_windowandt_obsbuffers to avoid write contention.Known Limitations
-Xclang -fopenmp). Full implementation will wireHAVE_OPENMPinto the configure system for Linux/Windows portability.memoryparameter consistent with other GRASS modules.This PR is not intended for merge. It demonstrates that the surface parameter computation is parallelizable and the performance gain is real.