Performance optimisation with pdf_cache_object/pdf_get_xref_entry

The recent change to holding pdf xrefs in a sparse format has resulted
in a significant decrease in speed (x10). Malc points out that some of
this (2x) can be recovered simply by making pdf_cache_object return the
entry which it found the object in.

This saves us having to immediately call pdf_get_xref_entry again
afterwards.

I am still thinking about ways to try and get the remaining time back.

3 files changed, 13 insertions, 19 deletions

diff --git a/include/mupdf/pdf/xref.h b/include/mupdf/pdf/xref.h
index 70826314..ac8525ac 100644
--- a/include/mupdf/pdf/xref.h
+++ b/include/mupdf/pdf/xref.h
@@ -66,7 +66,7 @@ struct pdf_xref_s
 	pdf_obj *pre_repair_trailer;
 };
 
-void pdf_cache_object(pdf_document *doc, int num, int gen);
+pdf_xref_entry *pdf_cache_object(pdf_document *doc, int num, int gen);
 
 int pdf_count_objects(pdf_document *doc);
 pdf_obj *pdf_resolve_indirect(pdf_obj *ref);
diff --git a/source/pdf/pdf-stream.c b/source/pdf/pdf-stream.c
index eb6b616e..f859719b 100644
--- a/source/pdf/pdf-stream.c
+++ b/source/pdf/pdf-stream.c
@@ -11,9 +11,8 @@ pdf_is_stream(pdf_document *doc, int num, int gen)
 	if (num <= 0 || num >= pdf_xref_len(doc))
 		return 0;
 
-	pdf_cache_object(doc, num, gen);
+	entry = pdf_cache_object(doc, num, gen);
 
-	entry = pdf_get_xref_entry(doc, num);
 	return entry->stm_ofs != 0 || entry->stm_buf;
 }
 
@@ -408,9 +407,7 @@ pdf_open_raw_renumbered_stream(pdf_document *doc, int num, int gen, int orig_num
 	if (num <= 0 || num >= pdf_xref_len(doc))
 		fz_throw(doc->ctx, FZ_ERROR_GENERIC, "object id out of range (%d %d R)", num, gen);
 
-	pdf_cache_object(doc, num, gen);
-
-	x = pdf_get_xref_entry(doc, num);
+	x = pdf_cache_object(doc, num, gen);
 	if (x->stm_ofs == 0)
 		fz_throw(doc->ctx, FZ_ERROR_GENERIC, "object is not a stream");
 
@@ -425,9 +422,7 @@ pdf_open_image_stream(pdf_document *doc, int num, int gen, int orig_num, int ori
 	if (num <= 0 || num >= pdf_xref_len(doc))
 		fz_throw(doc->ctx, FZ_ERROR_GENERIC, "object id out of range (%d %d R)", num, gen);
 
-	pdf_cache_object(doc, num, gen);
-
-	x = pdf_get_xref_entry(doc, num);
+	x = pdf_cache_object(doc, num, gen);
 	if (x->stm_ofs == 0 && x->stm_buf == NULL)
 		fz_throw(doc->ctx, FZ_ERROR_GENERIC, "object is not a stream");
 
diff --git a/source/pdf/pdf-xref.c b/source/pdf/pdf-xref.c
index d72ae55c..b7ac1e20 100644
--- a/source/pdf/pdf-xref.c
+++ b/source/pdf/pdf-xref.c
@@ -1833,7 +1833,7 @@ read_hinted_object(pdf_document *doc, int num)
 	return 1;
 }
 
-void
+pdf_xref_entry *
 pdf_cache_object(pdf_document *doc, int num, int gen)
 {
 	pdf_xref_entry *x;
@@ -1851,8 +1851,8 @@ object_updated:
 
 	x = pdf_get_xref_entry(doc, num);
 
-	if (x->obj)
-		return;
+	if (x->obj != NULL)
+		return x;
 
 	if (x->type == 'f')
 	{
@@ -1930,6 +1930,7 @@ object_updated:
 	}
 
 	pdf_set_obj_parent(x->obj, num);
+	return x;
 }
 
 pdf_obj *
@@ -1940,16 +1941,14 @@ pdf_load_object(pdf_document *doc, int num, int gen)
 
 	fz_try(ctx)
 	{
-		pdf_cache_object(doc, num, gen);
+		entry = pdf_cache_object(doc, num, gen);
 	}
 	fz_catch(ctx)
 	{
 		fz_rethrow_message(ctx, "cannot load object (%d %d R) into cache", num, gen);
 	}
 
-	entry = pdf_get_xref_entry(doc, num);
-
-	assert(entry->obj);
+	assert(entry->obj != NULL);
 
 	return pdf_keep_obj(entry->obj);
 }
@@ -1986,7 +1985,7 @@ pdf_resolve_indirect(pdf_obj *ref)
 
 		fz_try(ctx)
 		{
-			pdf_cache_object(doc, num, gen);
+			entry = pdf_cache_object(doc, num, gen);
 		}
 		fz_catch(ctx)
 		{
@@ -1994,8 +1993,8 @@ pdf_resolve_indirect(pdf_obj *ref)
 			fz_warn(ctx, "cannot load object (%d %d R) into cache", num, gen);
 			return NULL;
 		}
-		entry = pdf_get_xref_entry(doc, num);
-		if (!entry->obj)
+
+		if (entry->obj == NULL)
 			return NULL;
 		ref = entry->obj;
 	}