LeafLabs Garden » Topic: libmaple/ring

LeafLabs Garden » Topic: libmaple/ring_buffer.h http://forums.leaflabs.com/topic.php?id=689 A place to share, learn, and grow... en-US Fri, 22 Jan 2016 00:20:01 +0000 http://bbpress.org/?v=1.0.2 <![CDATA[Search]]> q http://forums.leaflabs.com/search.php mbolivar on "libmaple/ring_buffer.h" http://forums.leaflabs.com/topic.php?id=689#post-4131 Mon, 04 Apr 2011 12:45:46 +0000 mbolivar 4131@http://forums.leaflabs.com/ x893, Actually, the size of the buffer is 63, not 64. One byte is always left free to distinguish empty from full. This is what makes the logic correct. I was unable to reproduce your report, although it's possible that I misunderstood you. Can you please submit some sample code along with your description of the problem? For example, here is a commit I have just pushed into master which tests basic ring_buffer functionality: <a href="https://github.com/leaflabs/libmaple/commit/987fa5f433e0cd41b1621dad8b4e331cf82d49f3" rel="nofollow">https://github.com/leaflabs/libmaple/commit/987fa5f433e0cd41b1621dad8b4e331cf82d49f3</a> The relevant new file is examples/test-ring-buffer-insertion.cpp: <a href="https://github.com/leaflabs/libmaple/blob/987fa5f433e0cd41b1621dad8b4e331cf82d49f3/examples/test-ring-buffer-insertion.cpp" rel="nofollow">https://github.com/leaflabs/libmaple/blob/987fa5f433e0cd41b1621dad8b4e331cf82d49f3/examples/test-ring-buffer-insertion.cpp</a> Here is the output it gives on my Maple Rev 5; everything looks fine to me: <blockquote> Beginning test. resetting ring buffer. ring buffer contents: rb_push_insert()-ing 63 bytes. rb_full_count(rb) = 63 ring buffer contents: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63 ------------------------------ resetting ring buffer. ring buffer contents: rb_push_insert()-ing 64 bytes. rb_full_count(rb) = 63 ring buffer contents: 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64 ------------------------------ resetting ring buffer. ring buffer contents: rb_safe_insert()-ing 63 bytes. rb_full_count(rb) = 63 ring buffer contents: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63 ------------------------------ resetting ring buffer. ring buffer contents: rb_safe_insert()-ing 64 bytes. rb_full_count(rb) = 63 ring buffer contents: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63 ------------------------------ Test finished. </blockquote> In the future, when discussing problems, can you please report the version of libmaple or the Maple IDE that you are using? If you are using 0.0.9, the ring_buffer implementation had many bugs which have been fixed in subsequent snapshot releases. x893 on "libmaple/ring_buffer.h" http://forums.leaflabs.com/topic.php?id=689#post-4082 Tue, 29 Mar 2011 07:44:28 +0000 x893 4082@http://forums.leaflabs.com/ Logic for count build only on head/tail calculations not correct. Simple check any logic with next situation i send 64 bytes (64 = size of buffer) and rb_full_count(..) return zero ! For correct calculation need add counter to calculate insert/remove elements. Correctly name ring_buffer - queue with maximum capacity and count of queued elements. I implement this logic (counter/tail/head) in my CAN library and no lost elements in any case. mbolivar on "libmaple/ring_buffer.h" http://forums.leaflabs.com/topic.php?id=689#post-3933 Thu, 17 Mar 2011 08:11:44 +0000 mbolivar 3933@http://forums.leaflabs.com/ yes. sorry for the hasty implementation. <a href="https://github.com/leaflabs/libmaple/commit/12aca5b72d62081c50fafc41e448841a4328a500" rel="nofollow">https://github.com/leaflabs/libmaple/commit/12aca5b72d62081c50fafc41e448841a4328a500</a> geoffreymbrown on "libmaple/ring_buffer.h" http://forums.leaflabs.com/topic.php?id=689#post-3930 Thu, 17 Mar 2011 07:43:23 +0000 geoffreymbrown 3930@http://forums.leaflabs.com/ Isn't it easier to check "not empty" head != tail rather than full count != 0 ? mbolivar on "libmaple/ring_buffer.h" http://forums.leaflabs.com/topic.php?id=689#post-3927 Thu, 17 Mar 2011 07:22:40 +0000 mbolivar 3927@http://forums.leaflabs.com/ I'd rather not force everyone to pay for checking if the buffer is empty in rb_remove(), so I've added an rb_safe_remove() (in the refactor branch) instead: <a href="https://github.com/leaflabs/libmaple/commit/8097736f3be70ed3a5550334b31e8cfeddc0bfcf" rel="nofollow">https://github.com/leaflabs/libmaple/commit/8097736f3be70ed3a5550334b31e8cfeddc0bfcf</a> geoffreymbrown on "libmaple/ring_buffer.h" http://forums.leaflabs.com/topic.php?id=689#post-3925 Thu, 17 Mar 2011 07:01:39 +0000 geoffreymbrown 3925@http://forums.leaflabs.com/ Ok, I missed that. You might still want to make remove do the right thing when the buffer is empty. mbolivar on "libmaple/ring_buffer.h" http://forums.leaflabs.com/topic.php?id=689#post-3920 Thu, 17 Mar 2011 06:38:18 +0000 mbolivar 3920@http://forums.leaflabs.com/ Thanks for the report. This has been fixed in master for over a month: <a href="https://github.com/leaflabs/libmaple/blob/master/libmaple/ring_buffer.h" rel="nofollow">https://github.com/leaflabs/libmaple/blob/master/libmaple/ring_buffer.h</a> <a href="https://github.com/leaflabs/libmaple/blob/master/libmaple/usart.c#L83" rel="nofollow">https://github.com/leaflabs/libmaple/blob/master/libmaple/usart.c#L83</a> A snapshot was released with the new code (choose the 2011-02-13 snapshot release for your platform): <a href="http://static.leaflabs.com/pub/leaflabs/maple-ide/snapshots/" rel="nofollow">http://static.leaflabs.com/pub/leaflabs/maple-ide/snapshots/</a> geoffreymbrown on "libmaple/ring_buffer.h" http://forums.leaflabs.com/topic.php?id=689#post-3918 Wed, 16 Mar 2011 20:45:34 +0000 geoffreymbrown 3918@http://forums.leaflabs.com/ This looks pretty broken: static inline uint32 rb_full_count(ring_buffer *rb) { return rb->tail - rb->head; } Suppose with have a 4 place ring buffer where tail = 0, head = 1 (i.e. there are three items in the buffer). Type conversion isn't going to magically return 3. I think what you want is (rb->size + rb->tail - rb->head) & (rb->size - 1) Also, your insert and remove routines are definitely unsafe. Notice that in libmaple/uart.c, you don't check to see if the buffer is full. The right thing to do here is probably throw the character in the trash since you don't have flow control. Overflowing your buffer seems pretty crass. I noticed that you call rb_remove in your usart_getc routine but also don't check if the buffer is empty. At the least this should fail in benign way rather than manufacturing a buffer full of trash.