From 93ee812d2dba27937bac796a4e7bc1d6e7adf1df Mon Sep 17 00:00:00 2001
From: Dave Rowenhorst <45666721+drowenhorst-nrl@users.noreply.github.com>
Date: Fri, 7 Jun 2024 18:29:25 -0400
Subject: [PATCH] NLPAR Chunk Patch ... again (#62)

Fixed
-----
- This time I think that edge case for NLPAR chunking of scans is really fixed.
- Wrote on a chalkboard 100 times, "I will run ALL the unit tests before release."

Signed-off by: David Rowenhorst <david.rowenhorst@nrl.navy.mil>
---
 CHANGELOG.rst            |  9 +++++++++
 pyebsdindex/__init__.py  |  2 +-
 pyebsdindex/nlpar_cpu.py | 20 ++++++++++++++++++--
 3 files changed, 28 insertions(+), 3 deletions(-)

diff --git a/CHANGELOG.rst b/CHANGELOG.rst
index 09c0462..a5a0d15 100644
--- a/CHANGELOG.rst
+++ b/CHANGELOG.rst
@@ -5,6 +5,15 @@ Changelog
 All notable changes to PyEBSDIndex will be documented in this file. The format is based
 on `Keep a Changelog <https://keepachangelog.com/en/1.1.0>`_.
 
+
+0.3.4 (2024-06-07)
+==================
+
+Fixed
+-----
+- This time I think that edge case for NLPAR chunking of scans is really fixed.
+- Wrote on a chalkboard 100 times, "I will run ALL the unit tests before release."
+
 0.3.3 (2024-06-07)
 ==================
 
diff --git a/pyebsdindex/__init__.py b/pyebsdindex/__init__.py
index 3644248..53da16a 100644
--- a/pyebsdindex/__init__.py
+++ b/pyebsdindex/__init__.py
@@ -7,7 +7,7 @@
 ]
 __description__ = "Python based tool for Radon based EBSD indexing"
 __name__ = "pyebsdindex"
-__version__ = "0.3.3"
+__version__ = "0.3.4"
 
 
 # Try to import only once - also will perform check that at least one GPU is found.
diff --git a/pyebsdindex/nlpar_cpu.py b/pyebsdindex/nlpar_cpu.py
index 9bec11c..2b00fc3 100644
--- a/pyebsdindex/nlpar_cpu.py
+++ b/pyebsdindex/nlpar_cpu.py
@@ -390,6 +390,7 @@ def calcnlpar(self, chunksize=0, searchradius=None, lam = None, dthresh = None,
 
     nthreadpos = numba.get_num_threads()
     #numba.set_num_threads(18)
+    #numba.set_num_threads(18)
     colstartcount = np.asarray([0,ncols],dtype=np.int64)
     if verbose >= 1:
       print("lambda:", self.lam, "search radius:", self.searchradius, "dthresh:", self.dthresh)
@@ -776,7 +777,7 @@ def _calcchunks(self, patdim, ncol, nrow, target_bytes=2e9, col_overlap=0, row_o
     colchunks[-1, 1] = ncol
 
     if ncolchunks > 1:
-      colchunks[-1, 0] = max(0, colchunks[-2, 1] - col_overlap)
+      colchunks[-1, 0] = max(0, colchunks[-2, 1] -  2*col_overlap-1)
 
     colchunks += col_offset
 
@@ -784,6 +785,21 @@ def _calcchunks(self, patdim, ncol, nrow, target_bytes=2e9, col_overlap=0, row_o
     #   if colchunks[i + 1, 0] >= ncol:
     #     colchunks = colchunks[0:i + 1, :]
 
+    rowchunks = []
+    row_overlap = int(row_overlap)
+    for r in range(nrowchunks):
+      rchunk = [int(r * rowstep) - row_overlap, int(r * rowstep + rowstepov) - row_overlap]
+      rowchunks.append(rchunk)
+      if rchunk[1] > nrow:
+        break
+
+    nrowchunks = len(rowchunks)
+    rowchunks = np.array(rowchunks, dtype=int)
+    rowchunks[0, 0] = 0
+    # for i in range(ncolchunks - 1):
+    #   if colchunks[i + 1, 0] >= ncol:
+    #     colchunks = colchunks[0:i + 1, :]
+
     rowchunks = []
     row_overlap = int(row_overlap)
     for r in range(nrowchunks):
@@ -798,7 +814,7 @@ def _calcchunks(self, patdim, ncol, nrow, target_bytes=2e9, col_overlap=0, row_o
     rowchunks[-1, 1] = nrow
 
     if nrowchunks > 1:
-      rowchunks[-1, 0] = max(0, rowchunks[-2, 1] - row_overlap)
+      rowchunks[-1, 0] = max(0, rowchunks[-2, 1] -  2*row_overlap-1)
 
     rowchunks += row_offset