Re: [PATCH v3 7/7] revision.c: refactor basic topo-order logic
On 10/25/2018 5:43 AM, Jeff King wrote: On Thu, Oct 11, 2018 at 12:21:44PM -0400, Derrick Stolee wrote: 2. INDEGREE: using the indegree_queue priority queue (ordered by maximizing the generation number), add one to the in- degree of each parent for each commit that is walked. Since we walk in order of decreasing generation number, we know that discovering an in-degree value of 0 means the value for that commit was not initialized, so should be initialized to two. (Recall that in-degree value "1" is what we use to say a commit is ready for output.) As we iterate the parents of a commit during this walk, ensure the EXPLORE walk has walked beyond their generation numbers. I wondered how this would work for INFINITY. We can't know the order of a bunch of INFINITY nodes at all, so we never know when their in-degree values are "done". But if I understand the EXPLORE walk, we'd basically walk all of INFINITY down to something with a real generation number. Is that right? But after that, I'm not totally clear on why we need this INDEGREE walk. The INDEGREE walk is an important element for Kahn's algorithm. The final output order is dictated by peeling commits of "indegree zero" to ensure all children are output before their parents. (Note: since we use literal 0 to mean "uninitialized", we peel commits when the indegree slab has value 1.) This walk replaces the indegree logic from sort_in_topological_order(). That method performs one walk that fills the indegree slab, then another walk that peels the commits with indegree 0 and inserts them into a list. I guess my big question here was: if we have generation numbers, do we need Kahn's algorithm? That is, in a fully populated set of generation numbers (i.e., no INFINITY), we could always just pick a commit with the highest generation number to show. So if we EXPLORE down to a real generation number in phase 1, why do we need to care about INDEGREE anymore? Or am I wrong that we always have a real generation number (i.e., not INFINITY) after EXPLORE? (And if so, why is exploring to a real generation number a bad idea; presumably it's due to a worst-case that goes deeper than we'd otherwise need to here). The issue is that we our final order (used by level 3) is unrelated to generation number. Yes, if we prioritized by generation number then we could output the commits in _some_ order that doesn't violate topological constraints. However, we are asking for a different priority, which is different than the generation number priority. In the case of "--topo-order", we want to output the commits reachable from the second parent of a merge before the commits reachable from the first parent. However, in most cases the generation number of the first parent is higher than the second parent (there are more things in the merge chain than in a small topic that got merged). The INDEGREE is what allows us to know when we can peel these commits while still respecting the priority we want at the end. Thanks, -Stolee
Re: [PATCH v3 7/7] revision.c: refactor basic topo-order logic
On Thu, Oct 11, 2018 at 12:21:44PM -0400, Derrick Stolee wrote: > > > 2. INDEGREE: using the indegree_queue priority queue (ordered > > > by maximizing the generation number), add one to the in- > > > degree of each parent for each commit that is walked. Since > > > we walk in order of decreasing generation number, we know > > > that discovering an in-degree value of 0 means the value for > > > that commit was not initialized, so should be initialized to > > > two. (Recall that in-degree value "1" is what we use to say a > > > commit is ready for output.) As we iterate the parents of a > > > commit during this walk, ensure the EXPLORE walk has walked > > > beyond their generation numbers. > > I wondered how this would work for INFINITY. We can't know the order of > > a bunch of INFINITY nodes at all, so we never know when their in-degree > > values are "done". But if I understand the EXPLORE walk, we'd basically > > walk all of INFINITY down to something with a real generation number. Is > > that right? > > > > But after that, I'm not totally clear on why we need this INDEGREE walk. > > The INDEGREE walk is an important element for Kahn's algorithm. The final > output order is dictated by peeling commits of "indegree zero" to ensure all > children are output before their parents. (Note: since we use literal 0 to > mean "uninitialized", we peel commits when the indegree slab has value 1.) > > This walk replaces the indegree logic from sort_in_topological_order(). That > method performs one walk that fills the indegree slab, then another walk > that peels the commits with indegree 0 and inserts them into a list. I guess my big question here was: if we have generation numbers, do we need Kahn's algorithm? That is, in a fully populated set of generation numbers (i.e., no INFINITY), we could always just pick a commit with the highest generation number to show. So if we EXPLORE down to a real generation number in phase 1, why do we need to care about INDEGREE anymore? Or am I wrong that we always have a real generation number (i.e., not INFINITY) after EXPLORE? (And if so, why is exploring to a real generation number a bad idea; presumably it's due to a worst-case that goes deeper than we'd otherwise need to here). > [...] Everything else you said here made perfect sense. -Peff
Re: [PATCH v3 7/7] revision.c: refactor basic topo-order logic
On Fri, Sep 21, 2018 at 10:39 AM Derrick Stolee via GitGitGadget wrote: > > From: Derrick Stolee [...] > For the test above, I specifically selected a path that is changed > frequently, including by merge commits. A less-frequently-changed > path (such as 'README') has similar end-to-end time since we need > to walk the same number of commits (before determining we do not > have 100 hits). However, get get the benefit that the output is "get get"
Re: [PATCH v3 7/7] revision.c: refactor basic topo-order logic
On 10/11/2018 11:35 AM, Jeff King wrote: On Fri, Sep 21, 2018 at 10:39:36AM -0700, Derrick Stolee via GitGitGadget wrote: From: Derrick Stolee When running a command like 'git rev-list --topo-order HEAD', Git performed the following steps: [...] In the new algorithm, these three steps correspond to three different commit walks. We run these walks simultaneously, A minor nit, but this commit message doesn't mention the most basic thing up front: that its main purpose is to introduce a new algorithm for topo-order. ;) It's obvious in the context of reviewing the series, but somebody reading "git log" later may want a bit more. Perhaps: revision.c: implement generation-based topo-order algorithm as a subject, and/or an introductory paragraph like: The current --topo-order algorithm requires walking all commits we are going to output up front, topo-sorting them, all before outputting the first value. This patch introduces a new algorithm which uses stored generation numbers to incrementally walk in topo-order, outputting commits as we go. Other than that, I find this to be a wonderfully explanatory commit message. :) Good idea. I'll make that change. The walks are as follows: 1. EXPLORE: using the explore_queue priority queue (ordered by maximizing the generation number), parse each reachable commit until all commits in the queue have generation number strictly lower than needed. During this walk, update the UNINTERESTING flags as necessary. OK, this makes sense. If we know that everybody else in our queue is at generation X, then it is safe to output a commit at generation greater than X. I think this by itself would allow us to implement "show no parents before all of its children are shown", right? But --topo-order promises a bit more: "avoid showing commits no multiple lines of history intermixed". I guess also INFINITY generation numbers need more. For a real generation number, we know that "gen(A) == gen(B)" implies that there is no ancestry relationship between the two. But not so for INFINITY. Yeah, to deal with INFINITY (and ZERO, but that won't happen if generation_numbers_enabled() returns true), we treat gen(A) == gen(B) as a "no information" state. So, to output a commit at generation X, we need to have our maximum generation number in the unexplored area to be at most X - 1. You'll see strict inequality when checking generations. 2. INDEGREE: using the indegree_queue priority queue (ordered by maximizing the generation number), add one to the in- degree of each parent for each commit that is walked. Since we walk in order of decreasing generation number, we know that discovering an in-degree value of 0 means the value for that commit was not initialized, so should be initialized to two. (Recall that in-degree value "1" is what we use to say a commit is ready for output.) As we iterate the parents of a commit during this walk, ensure the EXPLORE walk has walked beyond their generation numbers. I wondered how this would work for INFINITY. We can't know the order of a bunch of INFINITY nodes at all, so we never know when their in-degree values are "done". But if I understand the EXPLORE walk, we'd basically walk all of INFINITY down to something with a real generation number. Is that right? But after that, I'm not totally clear on why we need this INDEGREE walk. The INDEGREE walk is an important element for Kahn's algorithm. The final output order is dictated by peeling commits of "indegree zero" to ensure all children are output before their parents. (Note: since we use literal 0 to mean "uninitialized", we peel commits when the indegree slab has value 1.) This walk replaces the indegree logic from sort_in_topological_order(). That method performs one walk that fills the indegree slab, then another walk that peels the commits with indegree 0 and inserts them into a list. 3. TOPO: using the topo_queue priority queue (ordered based on the sort_order given, which could be commit-date, author- date, or typical topo-order which treats the queue as a LIFO stack), remove a commit from the queue and decrement the in-degree of each parent. If a parent has an in-degree of one, then we add it to the topo_queue. Before we decrement the in-degree, however, ensure the INDEGREE walk has walked beyond that generation number. OK, this makes sense to make --author-date-order, etc, work. Potentially those numbers might have no relationship at all with the graph structure, but we promise "no parent before its children are shown", so this is really just a tie-breaker after the topo-sort anyway. As long as steps 1 and 2 are correct and produce a complete set of commits for one "layer", this should be OK. I guess I'm not 100% convinced that we don't have a case where we haven't yet parsed or considered some commit that we know cannot have an ancestry relationship with commits
Re: [PATCH v3 7/7] revision.c: refactor basic topo-order logic
On Fri, Sep 21, 2018 at 10:39:36AM -0700, Derrick Stolee via GitGitGadget wrote: > From: Derrick Stolee > > When running a command like 'git rev-list --topo-order HEAD', > Git performed the following steps: > [...] > In the new algorithm, these three steps correspond to three > different commit walks. We run these walks simultaneously, A minor nit, but this commit message doesn't mention the most basic thing up front: that its main purpose is to introduce a new algorithm for topo-order. ;) It's obvious in the context of reviewing the series, but somebody reading "git log" later may want a bit more. Perhaps: revision.c: implement generation-based topo-order algorithm as a subject, and/or an introductory paragraph like: The current --topo-order algorithm requires walking all commits we are going to output up front, topo-sorting them, all before outputting the first value. This patch introduces a new algorithm which uses stored generation numbers to incrementally walk in topo-order, outputting commits as we go. Other than that, I find this to be a wonderfully explanatory commit message. :) > The walks are as follows: > > 1. EXPLORE: using the explore_queue priority queue (ordered by >maximizing the generation number), parse each reachable >commit until all commits in the queue have generation >number strictly lower than needed. During this walk, update >the UNINTERESTING flags as necessary. OK, this makes sense. If we know that everybody else in our queue is at generation X, then it is safe to output a commit at generation greater than X. I think this by itself would allow us to implement "show no parents before all of its children are shown", right? But --topo-order promises a bit more: "avoid showing commits no multiple lines of history intermixed". I guess also INFINITY generation numbers need more. For a real generation number, we know that "gen(A) == gen(B)" implies that there is no ancestry relationship between the two. But not so for INFINITY. > 2. INDEGREE: using the indegree_queue priority queue (ordered >by maximizing the generation number), add one to the in- >degree of each parent for each commit that is walked. Since >we walk in order of decreasing generation number, we know >that discovering an in-degree value of 0 means the value for >that commit was not initialized, so should be initialized to >two. (Recall that in-degree value "1" is what we use to say a >commit is ready for output.) As we iterate the parents of a >commit during this walk, ensure the EXPLORE walk has walked >beyond their generation numbers. I wondered how this would work for INFINITY. We can't know the order of a bunch of INFINITY nodes at all, so we never know when their in-degree values are "done". But if I understand the EXPLORE walk, we'd basically walk all of INFINITY down to something with a real generation number. Is that right? But after that, I'm not totally clear on why we need this INDEGREE walk. > 3. TOPO: using the topo_queue priority queue (ordered based on >the sort_order given, which could be commit-date, author- >date, or typical topo-order which treats the queue as a LIFO >stack), remove a commit from the queue and decrement the >in-degree of each parent. If a parent has an in-degree of >one, then we add it to the topo_queue. Before we decrement >the in-degree, however, ensure the INDEGREE walk has walked >beyond that generation number. OK, this makes sense to make --author-date-order, etc, work. Potentially those numbers might have no relationship at all with the graph structure, but we promise "no parent before its children are shown", so this is really just a tie-breaker after the topo-sort anyway. As long as steps 1 and 2 are correct and produce a complete set of commits for one "layer", this should be OK. I guess I'm not 100% convinced that we don't have a case where we haven't yet parsed or considered some commit that we know cannot have an ancestry relationship with commits we are outputting. But it may have an author-date-order relationship. (I'm not at all convinced that this _is_ a problem, and I suspect it isn't; I'm only suggesting I haven't fully grokked the proof). > --- > object.h | 4 +- > revision.c | 196 +++-- > revision.h | 2 + > 3 files changed, 194 insertions(+), 8 deletions(-) I'll pause here on evaluating the actual code. It looks sane from a cursory read, but there's no point in digging further until I'm sure I fully understand the algorithm. I think that needs a little more brain power from me, and hopefully discussion around my comments above will help trigger that. -Peff
Re: [PATCH v3 7/7] revision.c: refactor basic topo-order logic
Derrick Stolee writes: > On 9/21/2018 1:39 PM, Derrick Stolee via GitGitGadget wrote: > Hello, Git contributors. > > I understand that this commit message and patch are pretty > daunting. There is a lot to read and digest. I would like to see if > anyone is willing to put the work in to review this patch, as I quite > like what it does, and the performance numbers below. I'll try to find time to review v3 of this patch series this week. >> In my local testing, I used the following Git commands on the >> Linux repository in three modes: HEAD~1 with no commit-graph, >> HEAD~1 with a commit-graph, and HEAD with a commit-graph. This >> allows comparing the benefits we get from parsing commits from >> the commit-graph and then again the benefits we get by >> restricting the set of commits we walk. >> >> Test: git rev-list --topo-order -100 HEAD >> HEAD~1, no commit-graph: 6.80 s >> HEAD~1, w/ commit-graph: 0.77 s >>HEAD, w/ commit-graph: 0.02 s >> >> Test: git rev-list --topo-order -100 HEAD -- tools >> HEAD~1, no commit-graph: 9.63 s >> HEAD~1, w/ commit-graph: 6.06 s >>HEAD, w/ commit-graph: 0.06 s > > If there is something I can do to make this easier to review, then > please let me know. > > Thanks, > -Stolee
Re: [PATCH v3 7/7] revision.c: refactor basic topo-order logic
On 9/21/2018 1:39 PM, Derrick Stolee via GitGitGadget wrote: From: Derrick Stolee When running a command like 'git rev-list --topo-order HEAD', Git performed the following steps: 1. Run limit_list(), which parses all reachable commits, adds them to a linked list, and distributes UNINTERESTING flags. If all unprocessed commits are UNINTERESTING, then it may terminate without walking all reachable commits. This does not occur if we do not specify UNINTERESTING commits. 2. Run sort_in_topological_order(), which is an implementation of Kahn's algorithm. It first iterates through the entire set of important commits and computes the in-degree of each (plus one, as we use 'zero' as a special value here). Then, we walk the commits in priority order, adding them to the priority queue if and only if their in-degree is one. As we remove commits from this priority queue, we decrement the in-degree of their parents. 3. While we are peeling commits for output, get_revision_1() uses pop_commit on the full list of commits computed by sort_in_topological_order(). In the new algorithm, these three steps correspond to three different commit walks. We run these walks simultaneously, and advance each only as far as necessary to satisfy the requirements of the 'higher order' walk. We know when we can pause each walk by using generation numbers from the commit- graph feature. Hello, Git contributors. I understand that this commit message and patch are pretty daunting. There is a lot to read and digest. I would like to see if anyone is willing to put the work in to review this patch, as I quite like what it does, and the performance numbers below. In my local testing, I used the following Git commands on the Linux repository in three modes: HEAD~1 with no commit-graph, HEAD~1 with a commit-graph, and HEAD with a commit-graph. This allows comparing the benefits we get from parsing commits from the commit-graph and then again the benefits we get by restricting the set of commits we walk. Test: git rev-list --topo-order -100 HEAD HEAD~1, no commit-graph: 6.80 s HEAD~1, w/ commit-graph: 0.77 s HEAD, w/ commit-graph: 0.02 s Test: git rev-list --topo-order -100 HEAD -- tools HEAD~1, no commit-graph: 9.63 s HEAD~1, w/ commit-graph: 6.06 s HEAD, w/ commit-graph: 0.06 s If there is something I can do to make this easier to review, then please let me know. Thanks, -Stolee
[PATCH v3 7/7] revision.c: refactor basic topo-order logic
From: Derrick Stolee When running a command like 'git rev-list --topo-order HEAD', Git performed the following steps: 1. Run limit_list(), which parses all reachable commits, adds them to a linked list, and distributes UNINTERESTING flags. If all unprocessed commits are UNINTERESTING, then it may terminate without walking all reachable commits. This does not occur if we do not specify UNINTERESTING commits. 2. Run sort_in_topological_order(), which is an implementation of Kahn's algorithm. It first iterates through the entire set of important commits and computes the in-degree of each (plus one, as we use 'zero' as a special value here). Then, we walk the commits in priority order, adding them to the priority queue if and only if their in-degree is one. As we remove commits from this priority queue, we decrement the in-degree of their parents. 3. While we are peeling commits for output, get_revision_1() uses pop_commit on the full list of commits computed by sort_in_topological_order(). In the new algorithm, these three steps correspond to three different commit walks. We run these walks simultaneously, and advance each only as far as necessary to satisfy the requirements of the 'higher order' walk. We know when we can pause each walk by using generation numbers from the commit- graph feature. Recall that the generation number of a commit satisfies: * If the commit has at least one parent, then the generation number is one more than the maximum generation number among its parents. * If the commit has no parent, then the generation number is one. There are two special generation numbers: * GENERATION_NUMBER_INFINITY: this value is 0x and indicates that the commit is not stored in the commit-graph and the generation number was not previously calculated. * GENERATION_NUMBER_ZERO: this value (0) is a special indicator to say that the commit-graph was generated by a version of Git that does not compute generation numbers (such as v2.18.0). Since we use generation_numbers_enabled() before using the new algorithm, we do not need to worry about GENERATION_NUMBER_ZERO. However, the existence of GENERATION_NUMBER_INFINITY implies the following weaker statement than the usual we expect from generation numbers: If A and B are commits with generation numbers gen(A) and gen(B) and gen(A) < gen(B), then A cannot reach B. Thus, we will walk in each of our stages until the "maximum unexpanded generation number" is strictly lower than the generation number of a commit we are about to use. The walks are as follows: 1. EXPLORE: using the explore_queue priority queue (ordered by maximizing the generation number), parse each reachable commit until all commits in the queue have generation number strictly lower than needed. During this walk, update the UNINTERESTING flags as necessary. 2. INDEGREE: using the indegree_queue priority queue (ordered by maximizing the generation number), add one to the in- degree of each parent for each commit that is walked. Since we walk in order of decreasing generation number, we know that discovering an in-degree value of 0 means the value for that commit was not initialized, so should be initialized to two. (Recall that in-degree value "1" is what we use to say a commit is ready for output.) As we iterate the parents of a commit during this walk, ensure the EXPLORE walk has walked beyond their generation numbers. 3. TOPO: using the topo_queue priority queue (ordered based on the sort_order given, which could be commit-date, author- date, or typical topo-order which treats the queue as a LIFO stack), remove a commit from the queue and decrement the in-degree of each parent. If a parent has an in-degree of one, then we add it to the topo_queue. Before we decrement the in-degree, however, ensure the INDEGREE walk has walked beyond that generation number. The implementations of these walks are in the following methods: * explore_walk_step and explore_to_depth * indegree_walk_step and compute_indegrees_to_depth * next_topo_commit and expand_topo_walk These methods have some patterns that may seem strange at first, but they are probably carry-overs from their equivalents in limit_list and sort_in_topological_order. One thing that is missing from this implementation is a proper way to stop walking when the entire queue is UNINTERESTING, so this implementation is not enabled by comparisions, such as in 'git rev-list --topo-order A..B'. This can be updated in the future. In my local testing, I used the following Git commands on the Linux repository in three modes: HEAD~1 with no commit-graph, HEAD~1 with a commit-graph, and HEAD with a commit-graph. This allows comparing the benefits we get from parsing commits from the commit-graph and then again the benefits we get by restricting the set of commits we walk. Test: git rev-list --topo-order -100 HEAD
