mu/043space.cc

295 lines
9.9 KiB
C++
Raw Normal View History

2015-05-19 17:38:23 +00:00
//: Spaces help isolate recipes from each other. You can create them at will,
2015-03-22 00:53:20 +00:00
//: and all addresses in arguments are implicitly based on the 'default-space'
//: (unless they have the /raw property)
//:
//: Spaces are often called 'scopes' in other languages. Stack frames are a
//: limited form of space that can't outlive callers.
4089 Clean up how we reclaim local scopes. It used to work like this (commit 3216): 1. Update refcounts of products after every instruction, EXCEPT: a) when instruction is a non-primitive and the callee starts with 'local-scope' (because it's already not decremented in 'return') OR: b) when instruction is primitive 'next-ingredient' or 'next-ingredient-without-typechecking', and its result is saved to a variable in the default space (because it's already incremented at the time of the call) 2. If a function starts with 'local-scope', force it to be reclaimed before each return. However, since locals may be returned, *very carefully* don't reclaim those. (See the logic in the old `escaping` and `should_update_refcount` functions.) However, this approach had issues. We needed two separate commands for 'local-scope' (reclaim locals on exit) and 'new-default-space' (programmer takes charge of reclaiming locals). The hard-coded reclamation duplicated refcounting logic. In addition to adding complexity, this implementation failed to work if a function overwrites default-space after setting up a local-scope (the old default-space is leaked). It also fails in the presence of continuations. Calling a continuation more than once was guaranteed to corrupt memory (commit 3986). After this commit, reclaiming local scopes now works like this: Update refcounts of products for every PRIMITIVE instruction. For non-primitive instructions, all the work happens in the `return` instruction: increment refcount of ingredients to `return` (unless -- one last bit of ugliness -- they aren't saved in the caller) decrement the refcount of the default-space use existing infrastructure for reclaiming as necessary if reclaiming default-space, first decrement refcount of each local again, use existing infrastructure for reclaiming as necessary This commit (finally!) completes the bulk[1] of step 2 of the plan in commit 3991. It was very hard until I gave up trying to tweak the existing implementation and just test-drove layer 43 from scratch. [1] There's still potential for memory corruption if we abuse `default-space`. I should probably try to add warnings about that at some point (todo in layer 45).
2017-10-23 06:14:19 +00:00
//:
2017-11-04 01:01:59 +00:00
//: Warning: messing with 'default-space' can corrupt memory. Don't share
//: default-space between recipes. Later we'll see how to chain spaces safely.
//:
//: Tests in this layer can write to a location as part of one type, and read
//: it as part of another. This is unsafe and insecure, and we'll stop doing
//: this once we switch to variable names.
2015-03-22 00:53:20 +00:00
//: Under the hood, a space is an array of locations in memory.
2016-09-17 21:53:00 +00:00
:(before "End Mu Types Initialization")
put(Type_abbreviations, "space", new_type_tree("address:array:location"));
2015-04-24 17:19:03 +00:00
:(scenario set_default_space)
def main [
# prepare default-space address
10:num/alloc-id, 11:num <- copy 0, 1000
# prepare default-space payload
1000:num <- copy 0 # alloc id of payload
1001:num <- copy 5 # length
# actual start of this recipe
default-space:space <- copy 10:&:@:location
# if default-space is 1000, then:
# 1000: alloc id
# 1001: array size
# 1002: location 0 (space for the chaining slot; described later; often unused)
# 1003: location 1 (space for the chaining slot; described later; often unused)
# 1004: local 2 (assuming it is a scalar)
2:num <- copy 93
2015-03-22 00:53:20 +00:00
]
+mem: storing 93 in location 1004
2015-03-22 00:53:20 +00:00
:(scenario lookup_sidesteps_default_space)
def main [
# prepare default-space address
10:num/alloc-id, 11:num <- copy 0, 1000
# prepare default-space payload
1000:num <- copy 0 # alloc id of payload
1001:num <- copy 5 # length
# prepare payload outside the local scope
2000:num/alloc-id, 2001:num <- copy 0, 34
2015-05-19 17:38:23 +00:00
# actual start of this recipe
default-space:space <- copy 10:&:@:location
# a local address
2:num, 3:num <- copy 0, 2000
20:num/raw <- copy *2:&:num
2015-04-03 18:42:50 +00:00
]
+mem: storing 2000 in location 1005
+mem: storing 34 in location 20
2015-04-03 18:42:50 +00:00
//: precondition: disable name conversion for 'default-space'
:(scenarios transform)
2015-05-23 19:30:58 +00:00
:(scenario convert_names_passes_default_space)
2015-10-07 07:22:49 +00:00
% Hide_errors = true;
def main [
default-space:num <- copy 0
x:num <- copy 1
2015-05-23 19:30:58 +00:00
]
+name: assign x 2
2015-05-23 19:30:58 +00:00
-name: assign default-space 1
-name: assign default-space 2
:(scenarios run)
2015-05-23 19:30:58 +00:00
:(before "End is_disqualified Special-cases")
2015-05-23 19:30:58 +00:00
if (x.name == "default-space")
x.initialized = true;
:(before "End is_special_name Special-cases")
2015-05-23 19:30:58 +00:00
if (s == "default-space") return true;
//: core implementation
2015-04-13 03:56:45 +00:00
:(before "End call Fields")
int default_space;
2015-05-24 00:58:10 +00:00
:(before "End call Constructor")
default_space = 0;
2015-03-22 00:53:20 +00:00
:(before "Begin canonize(x) Lookups")
2016-04-20 16:54:10 +00:00
absolutize(x);
2015-03-22 00:53:20 +00:00
:(code)
void absolutize(reagent& x) {
if (is_raw(x) || is_dummy(x)) return;
if (x.name == "default-space") return;
2016-08-17 19:18:27 +00:00
if (!x.initialized)
2017-05-26 23:43:18 +00:00
raise << to_original_string(current_instruction()) << ": reagent not initialized: '" << x.original_string << "'\n" << end();
x.set_value(address(x.value, space_base(x)));
x.properties.push_back(pair<string, string_tree*>("raw", NULL));
assert(is_raw(x));
2015-03-22 00:53:20 +00:00
}
//: hook replaced in a later layer
int space_base(const reagent& x) {
return current_call().default_space ? (current_call().default_space + /*skip alloc id*/1) : 0;
2016-01-20 21:41:19 +00:00
}
int address(int offset, int base) {
assert(offset >= 0);
2016-01-20 21:41:19 +00:00
if (base == 0) return offset; // raw
int size = get_or_insert(Memory, base);
2016-01-20 21:41:19 +00:00
if (offset >= size) {
// todo: test
raise << current_recipe_name() << ": location " << offset << " is out of bounds " << size << " at " << base << '\n' << end();
DUMP("");
exit(1);
2016-01-20 21:41:19 +00:00
return 0;
}
return base + /*skip length*/1 + offset;
}
//: reads and writes to the 'default-space' variable have special behavior
2016-01-20 21:41:19 +00:00
:(after "Begin Preprocess write_memory(x, data)")
2016-05-06 07:46:39 +00:00
if (x.name == "default-space") {
if (!is_mu_space(x))
2016-09-17 21:43:13 +00:00
raise << maybe(current_recipe_name()) << "'default-space' should be of type address:array:location, but is " << to_string(x.type) << '\n' << end();
if (SIZE(data) != 2)
raise << maybe(current_recipe_name()) << "'default-space' getting data from non-address\n" << end();
current_call().default_space = data.at(/*skip alloc id*/1);
2016-05-06 07:46:39 +00:00
return;
}
3309 Rip out everything to fix one failing unit test (commit 3290; type abbreviations). This commit does several things at once that I couldn't come up with a clean way to unpack: A. It moves to a new representation for type trees without changing the actual definition of the `type_tree` struct. B. It adds unit tests for our type metadata precomputation, so that errors there show up early and in a simpler setting rather than dying when we try to load Mu code. C. It fixes a bug, guarding against infinite loops when precomputing metadata for recursive shape-shifting containers. To do this it uses a dumb way of comparing type_trees, comparing their string representations instead. That is likely incredibly inefficient. Perhaps due to C, this commit has made Mu incredibly slow. Running all tests for the core and the edit/ app now takes 6.5 minutes rather than 3.5 minutes. == more notes and details I've been struggling for the past week now to back out of a bad design decision, a premature optimization from the early days: storing atoms directly in the 'value' slot of a cons cell rather than creating a special 'atom' cons cell and storing it on the 'left' slot. In other words, if a cons cell looks like this: o / | \ left val right ..then the type_tree (a b c) used to look like this (before this commit): o | \ a o | \ b o | \ c null ..rather than like this 'classic' approach to s-expressions which never mixes val and right (which is what we now have): o / \ o o | / \ a o o | / \ b o null | c The old approach made several operations more complicated, most recently the act of replacing a (possibly atom/leaf) sub-tree with another. That was the final straw that got me to realize the contortions I was going through to save a few type_tree nodes (cons cells). Switching to the new approach was hard partly because I've been using the old approach for so long and type_tree manipulations had pervaded everything. Another issue I ran into was the realization that my layers were not cleanly separated. Key parts of early layers (precomputing type metadata) existed purely for far later ones (shape-shifting types). Layers I got repeatedly stuck at: 1. the transform for precomputing type sizes (layer 30) 2. type-checks on merge instructions (layer 31) 3. the transform for precomputing address offsets in types (layer 36) 4. replace operations in supporting shape-shifting recipes (layer 55) After much thrashing I finally noticed that it wasn't the entirety of these layers that was giving me trouble, but just the type metadata precomputation, which had bugs that weren't manifesting until 30 layers later. Or, worse, when loading .mu files before any tests had had a chance to run. A common failure mode was running into types at run time that I hadn't precomputed metadata for at transform time. Digging into these bugs got me to realize that what I had before wasn't really very good, but a half-assed heuristic approach that did a whole lot of extra work precomputing metadata for utterly meaningless types like `((address number) 3)` which just happened to be part of a larger type like `(array (address number) 3)`. So, I redid it all. I switched the representation of types (because the old representation made unit tests difficult to retrofit) and added unit tests to the metadata precomputation. I also made layer 30 only do the minimal metadata precomputation it needs for the concepts introduced until then. In the process, I also made the precomputation more correct than before, and added hooks in the right place so that I could augment the logic when I introduced shape-shifting containers. == lessons learned There's several levels of hygiene when it comes to layers: 1. Every layer introduces precisely what it needs and in the simplest way possible. If I was building an app until just that layer, nothing would seem over-engineered. 2. Some layers are fore-shadowing features in future layers. Sometimes this is ok. For example, layer 10 foreshadows containers and arrays and so on without actually supporting them. That is a net win because it lets me lay out the core of Mu's data structures out in one place. But if the fore-shadowing gets too complex things get nasty. Not least because it can be hard to write unit tests for features before you provide the plumbing to visualize and manipulate them. 3. A layer is introducing features that are tested only in later layers. 4. A layer is introducing features with tests that are invalidated in later layers. (This I knew from early on to be an obviously horrendous idea.) Summary: avoid Level 2 (foreshadowing layers) as much as possible. Tolerate it indefinitely for small things where the code stays simple over time, but become strict again when things start to get more complex. Level 3 is mostly a net lose, but sometimes it can be expedient (a real case of the usually grossly over-applied term "technical debt"), and it's better than the conventional baseline of no layers and no scenarios. Just clean it up as soon as possible. Definitely avoid layer 4 at any time. == minor lessons Avoid unit tests for trivial things, write scenarios in context as much as possible. But within those margins unit tests are fine. Just introduce them before any scenarios (commit 3297). Reorganizing layers can be easy. Just merge layers for starters! Punt on resplitting them in some new way until you've gotten them to work. This is the wisdom of Refactoring: small steps. What made it hard was not wanting to merge *everything* between layer 30 and 55. The eventual insight was realizing I just need to move those two full-strength transforms and nothing else.
2016-09-10 01:32:52 +00:00
:(code)
bool is_mu_space(reagent/*copy*/ x) {
canonize_type(x);
if (!is_compound_type_starting_with(x.type, "address")) return false;
drop_from_type(x, "address");
if (!is_compound_type_starting_with(x.type, "array")) return false;
drop_from_type(x, "array");
return x.type && x.type->atom && x.type->name == "location";
3309 Rip out everything to fix one failing unit test (commit 3290; type abbreviations). This commit does several things at once that I couldn't come up with a clean way to unpack: A. It moves to a new representation for type trees without changing the actual definition of the `type_tree` struct. B. It adds unit tests for our type metadata precomputation, so that errors there show up early and in a simpler setting rather than dying when we try to load Mu code. C. It fixes a bug, guarding against infinite loops when precomputing metadata for recursive shape-shifting containers. To do this it uses a dumb way of comparing type_trees, comparing their string representations instead. That is likely incredibly inefficient. Perhaps due to C, this commit has made Mu incredibly slow. Running all tests for the core and the edit/ app now takes 6.5 minutes rather than 3.5 minutes. == more notes and details I've been struggling for the past week now to back out of a bad design decision, a premature optimization from the early days: storing atoms directly in the 'value' slot of a cons cell rather than creating a special 'atom' cons cell and storing it on the 'left' slot. In other words, if a cons cell looks like this: o / | \ left val right ..then the type_tree (a b c) used to look like this (before this commit): o | \ a o | \ b o | \ c null ..rather than like this 'classic' approach to s-expressions which never mixes val and right (which is what we now have): o / \ o o | / \ a o o | / \ b o null | c The old approach made several operations more complicated, most recently the act of replacing a (possibly atom/leaf) sub-tree with another. That was the final straw that got me to realize the contortions I was going through to save a few type_tree nodes (cons cells). Switching to the new approach was hard partly because I've been using the old approach for so long and type_tree manipulations had pervaded everything. Another issue I ran into was the realization that my layers were not cleanly separated. Key parts of early layers (precomputing type metadata) existed purely for far later ones (shape-shifting types). Layers I got repeatedly stuck at: 1. the transform for precomputing type sizes (layer 30) 2. type-checks on merge instructions (layer 31) 3. the transform for precomputing address offsets in types (layer 36) 4. replace operations in supporting shape-shifting recipes (layer 55) After much thrashing I finally noticed that it wasn't the entirety of these layers that was giving me trouble, but just the type metadata precomputation, which had bugs that weren't manifesting until 30 layers later. Or, worse, when loading .mu files before any tests had had a chance to run. A common failure mode was running into types at run time that I hadn't precomputed metadata for at transform time. Digging into these bugs got me to realize that what I had before wasn't really very good, but a half-assed heuristic approach that did a whole lot of extra work precomputing metadata for utterly meaningless types like `((address number) 3)` which just happened to be part of a larger type like `(array (address number) 3)`. So, I redid it all. I switched the representation of types (because the old representation made unit tests difficult to retrofit) and added unit tests to the metadata precomputation. I also made layer 30 only do the minimal metadata precomputation it needs for the concepts introduced until then. In the process, I also made the precomputation more correct than before, and added hooks in the right place so that I could augment the logic when I introduced shape-shifting containers. == lessons learned There's several levels of hygiene when it comes to layers: 1. Every layer introduces precisely what it needs and in the simplest way possible. If I was building an app until just that layer, nothing would seem over-engineered. 2. Some layers are fore-shadowing features in future layers. Sometimes this is ok. For example, layer 10 foreshadows containers and arrays and so on without actually supporting them. That is a net win because it lets me lay out the core of Mu's data structures out in one place. But if the fore-shadowing gets too complex things get nasty. Not least because it can be hard to write unit tests for features before you provide the plumbing to visualize and manipulate them. 3. A layer is introducing features that are tested only in later layers. 4. A layer is introducing features with tests that are invalidated in later layers. (This I knew from early on to be an obviously horrendous idea.) Summary: avoid Level 2 (foreshadowing layers) as much as possible. Tolerate it indefinitely for small things where the code stays simple over time, but become strict again when things start to get more complex. Level 3 is mostly a net lose, but sometimes it can be expedient (a real case of the usually grossly over-applied term "technical debt"), and it's better than the conventional baseline of no layers and no scenarios. Just clean it up as soon as possible. Definitely avoid layer 4 at any time. == minor lessons Avoid unit tests for trivial things, write scenarios in context as much as possible. But within those margins unit tests are fine. Just introduce them before any scenarios (commit 3297). Reorganizing layers can be easy. Just merge layers for starters! Punt on resplitting them in some new way until you've gotten them to work. This is the wisdom of Refactoring: small steps. What made it hard was not wanting to merge *everything* between layer 30 and 55. The eventual insight was realizing I just need to move those two full-strength transforms and nothing else.
2016-09-10 01:32:52 +00:00
}
2016-01-20 21:41:19 +00:00
:(scenario get_default_space)
def main [
# prepare default-space address
10:num/alloc-id, 11:num <- copy 0, 1000
# prepare default-space payload
1000:num <- copy 0 # alloc id of payload
1001:num <- copy 5 # length
# actual start of this recipe
default-space:space <- copy 10:space
2:space/raw <- copy default-space:space
2016-01-20 21:41:19 +00:00
]
+mem: storing 1000 in location 3
2016-01-20 21:41:19 +00:00
:(after "Begin Preprocess read_memory(x)")
2016-05-06 07:46:39 +00:00
if (x.name == "default-space") {
vector<double> result;
result.push_back(/*alloc id*/0);
2016-05-06 07:46:39 +00:00
result.push_back(current_call().default_space);
return result;
}
2016-01-20 21:41:19 +00:00
//:: fix 'get'
:(scenario lookup_sidesteps_default_space_in_get)
def main [
# prepare default-space address
10:num/alloc-id, 11:num <- copy 0, 1000
# prepare default-space payload
1000:num <- copy 0 # alloc id of payload
1001:num <- copy 5 # length
# prepare payload outside the local scope
2000:num/alloc-id, 2001:num/x, 2002:num/y <- copy 0, 34, 35
2015-05-19 17:38:23 +00:00
# actual start of this recipe
default-space:space <- copy 10:space
# a local address
2:num, 3:num <- copy 0, 2000
3000:num/raw <- get *2:&:point, 1:offset
2015-04-18 04:51:13 +00:00
]
+mem: storing 35 in location 3000
2015-04-18 04:51:13 +00:00
2016-04-20 16:58:07 +00:00
:(before "Read element" following "case GET:")
element.properties.push_back(pair<string, string_tree*>("raw", NULL));
2015-04-18 04:51:13 +00:00
//:: fix 'index'
:(scenario lookup_sidesteps_default_space_in_index)
def main [
# prepare default-space address
10:num/alloc-id, 11:num <- copy 0, 1000
# prepare default-space payload
1000:num <- copy 0 # alloc id of payload
1001:num <- copy 5 # length
# prepare an array address
20:num/alloc-id, 21:num <- copy 0, 2000
# prepare an array payload
2000:num/alloc-id, 2001:num/length, 2002:num/index:0, 2003:num/index:1 <- copy 0, 2, 34, 35
2015-05-19 17:38:23 +00:00
# actual start of this recipe
default-space:space <- copy 10:&:@:location
1:&:@:num <- copy 20:&:@:num/raw
3000:num/raw <- index *1:&:@:num, 1
2015-04-18 04:51:13 +00:00
]
+mem: storing 35 in location 3000
2015-04-18 04:51:13 +00:00
2016-04-20 16:58:07 +00:00
:(before "Read element" following "case INDEX:")
element.properties.push_back(pair<string, string_tree*>("raw", NULL));
2015-04-18 04:51:13 +00:00
4089 Clean up how we reclaim local scopes. It used to work like this (commit 3216): 1. Update refcounts of products after every instruction, EXCEPT: a) when instruction is a non-primitive and the callee starts with 'local-scope' (because it's already not decremented in 'return') OR: b) when instruction is primitive 'next-ingredient' or 'next-ingredient-without-typechecking', and its result is saved to a variable in the default space (because it's already incremented at the time of the call) 2. If a function starts with 'local-scope', force it to be reclaimed before each return. However, since locals may be returned, *very carefully* don't reclaim those. (See the logic in the old `escaping` and `should_update_refcount` functions.) However, this approach had issues. We needed two separate commands for 'local-scope' (reclaim locals on exit) and 'new-default-space' (programmer takes charge of reclaiming locals). The hard-coded reclamation duplicated refcounting logic. In addition to adding complexity, this implementation failed to work if a function overwrites default-space after setting up a local-scope (the old default-space is leaked). It also fails in the presence of continuations. Calling a continuation more than once was guaranteed to corrupt memory (commit 3986). After this commit, reclaiming local scopes now works like this: Update refcounts of products for every PRIMITIVE instruction. For non-primitive instructions, all the work happens in the `return` instruction: increment refcount of ingredients to `return` (unless -- one last bit of ugliness -- they aren't saved in the caller) decrement the refcount of the default-space use existing infrastructure for reclaiming as necessary if reclaiming default-space, first decrement refcount of each local again, use existing infrastructure for reclaiming as necessary This commit (finally!) completes the bulk[1] of step 2 of the plan in commit 3991. It was very hard until I gave up trying to tweak the existing implementation and just test-drove layer 43 from scratch. [1] There's still potential for memory corruption if we abuse `default-space`. I should probably try to add warnings about that at some point (todo in layer 45).
2017-10-23 06:14:19 +00:00
//:: 'local-scope' is a convenience operation to automatically deduce
2017-09-01 08:50:10 +00:00
//:: the amount of space to allocate in a default space with names
4089 Clean up how we reclaim local scopes. It used to work like this (commit 3216): 1. Update refcounts of products after every instruction, EXCEPT: a) when instruction is a non-primitive and the callee starts with 'local-scope' (because it's already not decremented in 'return') OR: b) when instruction is primitive 'next-ingredient' or 'next-ingredient-without-typechecking', and its result is saved to a variable in the default space (because it's already incremented at the time of the call) 2. If a function starts with 'local-scope', force it to be reclaimed before each return. However, since locals may be returned, *very carefully* don't reclaim those. (See the logic in the old `escaping` and `should_update_refcount` functions.) However, this approach had issues. We needed two separate commands for 'local-scope' (reclaim locals on exit) and 'new-default-space' (programmer takes charge of reclaiming locals). The hard-coded reclamation duplicated refcounting logic. In addition to adding complexity, this implementation failed to work if a function overwrites default-space after setting up a local-scope (the old default-space is leaked). It also fails in the presence of continuations. Calling a continuation more than once was guaranteed to corrupt memory (commit 3986). After this commit, reclaiming local scopes now works like this: Update refcounts of products for every PRIMITIVE instruction. For non-primitive instructions, all the work happens in the `return` instruction: increment refcount of ingredients to `return` (unless -- one last bit of ugliness -- they aren't saved in the caller) decrement the refcount of the default-space use existing infrastructure for reclaiming as necessary if reclaiming default-space, first decrement refcount of each local again, use existing infrastructure for reclaiming as necessary This commit (finally!) completes the bulk[1] of step 2 of the plan in commit 3991. It was very hard until I gave up trying to tweak the existing implementation and just test-drove layer 43 from scratch. [1] There's still potential for memory corruption if we abuse `default-space`. I should probably try to add warnings about that at some point (todo in layer 45).
2017-10-23 06:14:19 +00:00
:(scenario local_scope)
def main [
4089 Clean up how we reclaim local scopes. It used to work like this (commit 3216): 1. Update refcounts of products after every instruction, EXCEPT: a) when instruction is a non-primitive and the callee starts with 'local-scope' (because it's already not decremented in 'return') OR: b) when instruction is primitive 'next-ingredient' or 'next-ingredient-without-typechecking', and its result is saved to a variable in the default space (because it's already incremented at the time of the call) 2. If a function starts with 'local-scope', force it to be reclaimed before each return. However, since locals may be returned, *very carefully* don't reclaim those. (See the logic in the old `escaping` and `should_update_refcount` functions.) However, this approach had issues. We needed two separate commands for 'local-scope' (reclaim locals on exit) and 'new-default-space' (programmer takes charge of reclaiming locals). The hard-coded reclamation duplicated refcounting logic. In addition to adding complexity, this implementation failed to work if a function overwrites default-space after setting up a local-scope (the old default-space is leaked). It also fails in the presence of continuations. Calling a continuation more than once was guaranteed to corrupt memory (commit 3986). After this commit, reclaiming local scopes now works like this: Update refcounts of products for every PRIMITIVE instruction. For non-primitive instructions, all the work happens in the `return` instruction: increment refcount of ingredients to `return` (unless -- one last bit of ugliness -- they aren't saved in the caller) decrement the refcount of the default-space use existing infrastructure for reclaiming as necessary if reclaiming default-space, first decrement refcount of each local again, use existing infrastructure for reclaiming as necessary This commit (finally!) completes the bulk[1] of step 2 of the plan in commit 3991. It was very hard until I gave up trying to tweak the existing implementation and just test-drove layer 43 from scratch. [1] There's still potential for memory corruption if we abuse `default-space`. I should probably try to add warnings about that at some point (todo in layer 45).
2017-10-23 06:14:19 +00:00
local-scope
x:num <- copy 0
y:num <- copy 3
]
# allocate space for x and y, as well as the chaining slot at indices 0 and 1
+mem: array length is 4
:(before "End is_disqualified Special-cases")
if (x.name == "number-of-locals")
x.initialized = true;
:(before "End is_special_name Special-cases")
if (s == "number-of-locals") return true;
:(before "End Rewrite Instruction(curr, recipe result)")
4089 Clean up how we reclaim local scopes. It used to work like this (commit 3216): 1. Update refcounts of products after every instruction, EXCEPT: a) when instruction is a non-primitive and the callee starts with 'local-scope' (because it's already not decremented in 'return') OR: b) when instruction is primitive 'next-ingredient' or 'next-ingredient-without-typechecking', and its result is saved to a variable in the default space (because it's already incremented at the time of the call) 2. If a function starts with 'local-scope', force it to be reclaimed before each return. However, since locals may be returned, *very carefully* don't reclaim those. (See the logic in the old `escaping` and `should_update_refcount` functions.) However, this approach had issues. We needed two separate commands for 'local-scope' (reclaim locals on exit) and 'new-default-space' (programmer takes charge of reclaiming locals). The hard-coded reclamation duplicated refcounting logic. In addition to adding complexity, this implementation failed to work if a function overwrites default-space after setting up a local-scope (the old default-space is leaked). It also fails in the presence of continuations. Calling a continuation more than once was guaranteed to corrupt memory (commit 3986). After this commit, reclaiming local scopes now works like this: Update refcounts of products for every PRIMITIVE instruction. For non-primitive instructions, all the work happens in the `return` instruction: increment refcount of ingredients to `return` (unless -- one last bit of ugliness -- they aren't saved in the caller) decrement the refcount of the default-space use existing infrastructure for reclaiming as necessary if reclaiming default-space, first decrement refcount of each local again, use existing infrastructure for reclaiming as necessary This commit (finally!) completes the bulk[1] of step 2 of the plan in commit 3991. It was very hard until I gave up trying to tweak the existing implementation and just test-drove layer 43 from scratch. [1] There's still potential for memory corruption if we abuse `default-space`. I should probably try to add warnings about that at some point (todo in layer 45).
2017-10-23 06:14:19 +00:00
// rewrite 'local-scope' to
2017-09-01 08:50:10 +00:00
// ```
// default-space:space <- new location:type, number-of-locals:literal
// ```
// where number-of-locals is Name[recipe][""]
4089 Clean up how we reclaim local scopes. It used to work like this (commit 3216): 1. Update refcounts of products after every instruction, EXCEPT: a) when instruction is a non-primitive and the callee starts with 'local-scope' (because it's already not decremented in 'return') OR: b) when instruction is primitive 'next-ingredient' or 'next-ingredient-without-typechecking', and its result is saved to a variable in the default space (because it's already incremented at the time of the call) 2. If a function starts with 'local-scope', force it to be reclaimed before each return. However, since locals may be returned, *very carefully* don't reclaim those. (See the logic in the old `escaping` and `should_update_refcount` functions.) However, this approach had issues. We needed two separate commands for 'local-scope' (reclaim locals on exit) and 'new-default-space' (programmer takes charge of reclaiming locals). The hard-coded reclamation duplicated refcounting logic. In addition to adding complexity, this implementation failed to work if a function overwrites default-space after setting up a local-scope (the old default-space is leaked). It also fails in the presence of continuations. Calling a continuation more than once was guaranteed to corrupt memory (commit 3986). After this commit, reclaiming local scopes now works like this: Update refcounts of products for every PRIMITIVE instruction. For non-primitive instructions, all the work happens in the `return` instruction: increment refcount of ingredients to `return` (unless -- one last bit of ugliness -- they aren't saved in the caller) decrement the refcount of the default-space use existing infrastructure for reclaiming as necessary if reclaiming default-space, first decrement refcount of each local again, use existing infrastructure for reclaiming as necessary This commit (finally!) completes the bulk[1] of step 2 of the plan in commit 3991. It was very hard until I gave up trying to tweak the existing implementation and just test-drove layer 43 from scratch. [1] There's still potential for memory corruption if we abuse `default-space`. I should probably try to add warnings about that at some point (todo in layer 45).
2017-10-23 06:14:19 +00:00
if (curr.name == "local-scope") {
rewrite_default_space_instruction(curr);
}
:(code)
void rewrite_default_space_instruction(instruction& curr) {
if (!curr.ingredients.empty())
raise << "'" << to_original_string(curr) << "' can't take any ingredients\n" << end();
curr.name = "new";
curr.ingredients.push_back(reagent("location:type"));
curr.ingredients.push_back(reagent("number-of-locals:literal"));
if (!curr.products.empty())
2017-12-24 20:44:01 +00:00
raise << "local-scope can't take any results\n" << end();
curr.products.push_back(reagent("default-space:space"));
}
:(after "Begin Preprocess read_memory(x)")
2016-05-06 07:46:39 +00:00
if (x.name == "number-of-locals") {
vector<double> result;
result.push_back(Name[get(Recipe_ordinal, current_recipe_name())][""]);
if (result.back() == 0)
raise << "no space allocated for default-space in recipe " << current_recipe_name() << "; are you using names?\n" << end();
return result;
}
:(after "Begin Preprocess write_memory(x, data)")
2016-05-06 07:46:39 +00:00
if (x.name == "number-of-locals") {
raise << maybe(current_recipe_name()) << "can't write to special name 'number-of-locals'\n" << end();
return;
}
//:: all recipes must set default-space one way or another
:(before "End Globals")
2016-02-25 16:24:14 +00:00
bool Hide_missing_default_space_errors = true;
:(before "End Checks")
Transform.push_back(check_default_space); // idempotent
:(code)
void check_default_space(const recipe_ordinal r) {
2016-10-22 23:56:07 +00:00
if (Hide_missing_default_space_errors) return; // skip previous core tests; this is only for Mu code
const recipe& caller = get(Recipe, r);
// End check_default_space Special-cases
// assume recipes with only numeric addresses know what they're doing (usually tests)
if (!contains_non_special_name(r)) return;
trace(9991, "transform") << "--- check that recipe " << caller.name << " sets default-space" << end();
if (caller.steps.empty()) return;
4089 Clean up how we reclaim local scopes. It used to work like this (commit 3216): 1. Update refcounts of products after every instruction, EXCEPT: a) when instruction is a non-primitive and the callee starts with 'local-scope' (because it's already not decremented in 'return') OR: b) when instruction is primitive 'next-ingredient' or 'next-ingredient-without-typechecking', and its result is saved to a variable in the default space (because it's already incremented at the time of the call) 2. If a function starts with 'local-scope', force it to be reclaimed before each return. However, since locals may be returned, *very carefully* don't reclaim those. (See the logic in the old `escaping` and `should_update_refcount` functions.) However, this approach had issues. We needed two separate commands for 'local-scope' (reclaim locals on exit) and 'new-default-space' (programmer takes charge of reclaiming locals). The hard-coded reclamation duplicated refcounting logic. In addition to adding complexity, this implementation failed to work if a function overwrites default-space after setting up a local-scope (the old default-space is leaked). It also fails in the presence of continuations. Calling a continuation more than once was guaranteed to corrupt memory (commit 3986). After this commit, reclaiming local scopes now works like this: Update refcounts of products for every PRIMITIVE instruction. For non-primitive instructions, all the work happens in the `return` instruction: increment refcount of ingredients to `return` (unless -- one last bit of ugliness -- they aren't saved in the caller) decrement the refcount of the default-space use existing infrastructure for reclaiming as necessary if reclaiming default-space, first decrement refcount of each local again, use existing infrastructure for reclaiming as necessary This commit (finally!) completes the bulk[1] of step 2 of the plan in commit 3991. It was very hard until I gave up trying to tweak the existing implementation and just test-drove layer 43 from scratch. [1] There's still potential for memory corruption if we abuse `default-space`. I should probably try to add warnings about that at some point (todo in layer 45).
2017-10-23 06:14:19 +00:00
if (!starts_by_setting_default_space(caller))
raise << caller.name << " does not seem to start with 'local-scope' or 'default-space'\n" << end();
}
bool starts_by_setting_default_space(const recipe& r) {
return !r.steps.empty()
&& !r.steps.at(0).products.empty()
&& r.steps.at(0).products.at(0).name == "default-space";
}
2016-10-22 23:56:07 +00:00
:(after "Load Mu Prelude")
2016-02-25 16:24:14 +00:00
Hide_missing_default_space_errors = false;
:(after "Test Runs")
2016-02-25 16:24:14 +00:00
Hide_missing_default_space_errors = true;
:(after "Running Main")
2016-02-25 16:24:14 +00:00
Hide_missing_default_space_errors = false;
:(code)
bool contains_non_special_name(const recipe_ordinal r) {
2016-10-20 05:10:35 +00:00
for (map<string, int>::iterator p = Name[r].begin(); p != Name[r].end(); ++p) {
if (p->first.empty()) continue;
2016-02-10 23:46:50 +00:00
if (p->first.find("stash_") == 0) continue; // generated by rewrite_stashes_to_text (cross-layer)
2016-02-26 04:47:42 +00:00
if (!is_special_name(p->first))
return true;
}
return false;
}