[qcms] Add matrix transform flag and matrix extraction api

third_party/qcms/src/transform.c

 /* vim: set ts=8 sw=8 noexpandtab: */
 //  qcms
 //  Copyright (C) 2009 Mozilla Corporation
 //  Copyright (C) 1998-2007 Marti Maria
 //
 // Permission is hereby granted, free of charge, to any person obtaining 
 // a copy of this software and associated documentation files (the "Software"), 
 // to deal in the Software without restriction, including without limitation 
 // the rights to use, copy, modify, merge, publish, distribute, sublicense, 
 // and/or sell copies of the Software, and to permit persons to whom the Software 
@@ skipping 1347 matching lines @@
                 transform->matrix[0][0] = result.m[0][0];
                 transform->matrix[1][0] = result.m[0][1];
                 transform->matrix[2][0] = result.m[0][2];
                 transform->matrix[0][1] = result.m[1][0];
                 transform->matrix[1][1] = result.m[1][1];
                 transform->matrix[2][1] = result.m[1][2];
                 transform->matrix[0][2] = result.m[2][0];
                 transform->matrix[1][2] = result.m[2][1];
                 transform->matrix[2][2] = result.m[2][2];
 
+                /* Flag transform as matrix. */
+                transform->transform_flags |= TRANSFORM_FLAG_MATRIX;

[Stephen White, 2015/05/27 17:17:49]

Does transform_flags need to be initialized to zero somewhere?

In the non-HAVE_POSIX_MEMALIGN case transform_alloc() use calloc(), which means
the above is fine. It's the posix_memalign() case I'm concerned about.

[Noel Gordon, 2015/05/27 17:42:56]

calloc-ed as you noted, that's the normal case for us.  Dunno if we use
HAVE_POSIX_MEMALIGN, but other chromium users might .. so yes, let's zero it.

+
         } else if (in->color_space == GRAY_SIGNATURE) {
                 if (in_type != QCMS_DATA_GRAY_8 &&
                                 in_type != QCMS_DATA_GRAYA_8){
                         assert(0 && "input type");
                         transform_free(transform);
                         return NULL;
                 }
 
                 transform->input_gamma_table_gray = build_input_gamma_table(in->grayTRC);
                 if (!transform->input_gamma_table_gray) {
@@ skipping 46 matching lines @@
 
 void qcms_transform_data_type(qcms_transform *transform, void *src, void *dest, size_t length, qcms_output_type type)
 {
         static const struct _qcms_format_type output_rgbx = { 0, 2 };
         static const struct _qcms_format_type output_bgrx = { 2, 0 };
 
         transform->transform_fn(transform, src, dest, length, type == QCMS_OUTPUT_BGRX ? output_bgrx : output_rgbx);
 }
 
 qcms_bool qcms_supports_iccv4;
+
 void qcms_enable_iccv4()
 {
         qcms_supports_iccv4 = true;
 }
+
+static inline qcms_bool transform_is_matrix(qcms_transform *t)
+{
+        return (t->transform_flags & TRANSFORM_FLAG_MATRIX) ? true : false;
+}
+
+qcms_bool qcms_transform_is_matrix(qcms_transform *t)
+{
+        return transform_is_matrix(t);
+}
+
+float qcms_transform_get_matrix(qcms_transform *t, int i, int j)

[robert.bradford, 2015/05/27 16:03:27]

Perhaps call it _matrix_value - in case in the future there is a matrix type to
expose it as?

[robert.bradford, 2015/05/27 16:03:27]

unsigned int or size_t?

[Noel Gordon, 2015/05/27 17:21:55]

The i, j kinda does that for me too.  Let's leave it as is for now.

[Noel Gordon, 2015/05/27 17:21:55]

Yes, made it unsigned.

+{
+        assert(transform_is_matrix(t));
+
+        // Return transform matrix elements in row major order (permute i and j)
+
+        return t->matrix[j][i];

[robert.bradford, 2015/05/27 16:03:27]

Should we be doing bounds checks on array dereference?

[Noel Gordon, 2015/05/27 17:21:55]

Done.

+}