diff options
| author | Chris Liddell <chris.liddell@artifex.com> | 2013-10-03 15:15:53 +0100 |
|---|---|---|
| committer | Chris Liddell <chris.liddell@artifex.com> | 2013-10-10 11:29:56 +0100 |
| commit | 68fa228ade5d2f7496b96676aaa1f82be4584362 (patch) | |
| tree | d2067a8f4c0c6b173c8e8249bedaebd6cf87c10b /xps/xpsfapi.c | |
| parent | 83565d8264e96312df8c26055aaa0e23e57c4c8d (diff) | |
| download | ghostpdl-68fa228ade5d2f7496b96676aaa1f82be4584362.tar.gz | |
Bug 694519: improve PCL performance with FAPI
Correctly maintain "design_grid" parameter in the font/matrix pair
structure, for FAPI handled TTFs.
Change a "return" to a "return_error".
Retrieving glyph metrics was sub-optimal as my original implementation used
a gs_text_process() call, which uses the entire "show machinery".
This implementation calls gs_fapi_do_char() directly. This has several benefits:
we only need a "skeleton" text enumerator, and a "skeleton" graphics state,
cutting down the local overhead setup. We also cutout a whole lot of setup
done as part of gs_text_begin() and during gs_text_process() itself. Finally,
we avoid a lot of "stuff" related to showing entire strings, when we're only
ever handling a single glyph at a time.
A final optimisation is in the FAPI/FT code: if we're just retrieving metrics,
we no longer take the action of retrieving the glyph from FT - which is not
a very cheap operation.
This commit causes ~100 cluster tests to have pixel shift differences.
Diffstat (limited to 'xps/xpsfapi.c')
| -rw-r--r-- | xps/xpsfapi.c | 1 |
1 files changed, 1 insertions, 0 deletions
diff --git a/xps/xpsfapi.c b/xps/xpsfapi.c index d81abb75f..968ad8747 100644 --- a/xps/xpsfapi.c +++ b/xps/xpsfapi.c @@ -62,6 +62,7 @@ static const gs_fapi_font pl_ff_stub = { false, /* is_outline_font */ false, /* is_mtx_skipped */ false, /* is_vertical */ + false, /* metrics_only */ { {3,10}, {3,1},{3,5},{3,4},\ {3,3},{3,2},{3,0},{1,0}, |