exp tdim caching

.github/workflows/beta-and-nightly.yml

@@ -52,7 +52,7 @@ jobs:
   warnings:
     strategy:
       matrix:
-        rust: [ 1.75.0, beta ]
+        rust: [ 1.77.0, beta ]
         os: [ ubuntu-latest, macOS-latest ]
       fail-fast: false
     runs-on: ${{ matrix.os }}

.github/workflows/unix.yml

@@ -14,7 +14,7 @@ jobs:
     strategy:
       matrix:
         os: [ ubuntu-latest, macOS-latest ]
-        rust: [ 1.75.0 ]
+        rust: [ 1.77.0 ]
       fail-fast: false
 
     runs-on: ${{matrix.os}}

.travis/cross.sh

@@ -3,7 +3,7 @@
 set -ex
 
 export DEBIAN_FRONTEND=noninteractive
-export RUSTUP_TOOLCHAIN=1.75.0
+export RUSTUP_TOOLCHAIN=1.77.0
 
 if [ `whoami` != "root" ]
 then

Cargo.toml

@@ -172,6 +172,7 @@ prost-types = "0.11.0"
 py_literal = "0.4.0"
 rand = { version = "0.8.4", features = ["small_rng"] }
 rand_distr = "0.4"
+rapidhash = "1.2"
 rayon = "1.10"
 readings-probe = "0.1.3"
 regex = "1.5.4"

api/Cargo.toml

@@ -9,7 +9,7 @@ keywords = [ "NeuralNetworks" ]
 categories = [ "science" ]
 autobenches = false
 edition = "2021"
-rust-version = "1.75"
+rust-version = "1.77"
 include = [ "Cargo.toml", "src/**/*.rs", "LICENSE*", "tract.h" ]
 
 [dependencies]

api/proxy/Cargo.toml

@@ -9,7 +9,7 @@ keywords = [ "NeuralNetworks" ]
 categories = [ "science" ]
 autobenches = false
 edition = "2021"
-rust-version = "1.75"
+rust-version = "1.77"
 include = [ "Cargo.toml", "src/**/*.rs", "LICENSE*" ]
 
 [dependencies]

api/proxy/sys/Cargo.toml

@@ -9,6 +9,6 @@ keywords = [ "NeuralNetworks" ]
 categories = [ "science" ]
 autobenches = false
 edition = "2021"
-rust-version = "1.75"
+rust-version = "1.77"
 include = [ "Cargo.toml", "src/**/*.rs", "LICENSE*", "build.rs", "tract.h", "bindings.rs" ]
 

api/rs/Cargo.toml

@@ -9,7 +9,7 @@ keywords = [ "NeuralNetworks" ]
 categories = [ "science" ]
 autobenches = false
 edition = "2021"
-rust-version = "1.75"
+rust-version = "1.77"
 include = [ "Cargo.toml", "src/**/*.rs", "LICENSE*", "tests" ]
 
 [dependencies]

core/Cargo.toml

@@ -9,7 +9,7 @@ keywords = [ "TensorFlow", "NeuralNetworks" ]
 categories = [ "science" ]
 autobenches = false
 edition = "2021"
-rust-version = "1.75"
+rust-version = "1.77"
 
 [badges]
 maintenance = { status = "actively-developed" }

data/Cargo.toml

@@ -9,7 +9,7 @@ keywords = [ "TensorFlow", "NeuralNetworks" ]
 categories = [ "science" ]
 autobenches = false
 edition = "2021"
-rust-version = "1.75"
+rust-version = "1.77"
 
 [badges]
 maintenance = { status = "actively-developed" }
@@ -27,6 +27,7 @@ nom.workspace = true
 num-complex = { workspace = true, optional = true }
 num-integer.workspace = true
 num-traits.workspace = true
+rapidhash.workspace = true
 smallvec.workspace = true
 lazy_static.workspace = true
 scan_fmt.workspace = true

data/src/dim/sym.rs

@@ -28,6 +28,7 @@ pub struct SymbolScopeData {
     table: DefaultStringInterner,
     assertions: Vec<Assertion>,
     scenarios: Vec<(String, Vec<Assertion>)>,
+    cache_proven_positive_or_zero: RefCell<rapidhash::RapidHashMap<TDim, bool>>,
 }
 
 impl SymbolScope {
@@ -147,7 +148,7 @@ impl SymbolScope {
         let locked = self.0.lock();
         let locked = locked.borrow();
         if locked.scenarios.len() == 0 {
-            return Ok(None)
+            return Ok(None);
         }
         let mut maybe = None;
         for (ix, (_name, assertions)) in locked.scenarios.iter().enumerate() {
@@ -158,7 +159,7 @@ impl SymbolScope {
             } else if maybe.is_none() {
                 maybe = Some(ix);
             } else {
-                return Ok(None)
+                return Ok(None);
             }
         }
         if maybe.is_some() {
@@ -201,38 +202,41 @@ impl SymbolScopeData {
         if let TDim::Val(v) = t {
             return *v >= 0;
         }
-        let positives = self.assertions.iter().filter_map(|i| i.as_known_positive()).collect_vec();
-        let mut visited = vec![];
-        let mut todo = vec![t.clone()];
-        while let Some(t) = todo.pop() {
-            if t.to_i64().is_ok_and(|i| i >= 0) {
-                return true;
-            }
-            if t.inclusive_bound(self, false).is_some_and(|l| l >= 0) {
-                return true;
-            }
-            let syms = t.symbols();
-            for s in syms {
-                let me = t.guess_slope(&s);
-                for pos in &positives {
-                    if pos.symbols().contains(&s) {
-                        let other = pos.guess_slope(&s);
-                        if me.0.signum() == other.0.signum() {
-                            let new = t.clone() * me.1 * other.0.abs()
-                                - pos.clone() * me.0.abs() * other.1;
-                            if !visited.contains(&new) {
-                                todo.push(new);
+        *self.cache_proven_positive_or_zero.borrow_mut().entry(t.clone()).or_insert_with(|| {
+            let positives =
+                self.assertions.iter().filter_map(|i| i.as_known_positive()).collect_vec();
+            let mut visited = vec![];
+            let mut todo = vec![t.clone()];
+            while let Some(t) = todo.pop() {
+                if t.to_i64().is_ok_and(|i| i >= 0) {
+                    return true;
+                }
+                if t.inclusive_bound(self, false).is_some_and(|l| l >= 0) {
+                    return true;
+                }
+                let syms = t.symbols();
+                for s in syms {
+                    let me = t.guess_slope(&s);
+                    for pos in &positives {
+                        if pos.symbols().contains(&s) {
+                            let other = pos.guess_slope(&s);
+                            if me.0.signum() == other.0.signum() {
+                                let new = t.clone() * me.1 * other.0.abs()
+                                    - pos.clone() * me.0.abs() * other.1;
+                                if !visited.contains(&new) {
+                                    todo.push(new);
+                                }
                             }
                         }
                     }
                 }
+                visited.push(t);
+                if visited.len() > 10 {
+                    break;
+                }
             }
-            visited.push(t);
-            if visited.len() > 10 {
-                break;
-            }
-        }
-        false
+            false
+        })
     }
 }
 

extra/Cargo.toml

@@ -9,7 +9,7 @@ keywords = [ "TensorFlow", "NeuralNetworks" ]
 categories = [ "science" ]
 autobenches = false
 edition = "2021"
-rust-version = "1.75"
+rust-version = "1.77"
 
 [badges]
 maintenance = { status = "actively-developed" }

metal/Cargo.toml

@@ -12,7 +12,7 @@ keywords = [ "TensorFlow", "NeuralNetworks", "Metal" ]
 categories = [ "science" ]
 autobenches = false
 edition = "2021"
-rust-version = "1.75"
+rust-version = "1.77"
 
 [badges]
 maintenance = { status = "actively-developed" }

nnef/Cargo.toml

@@ -9,7 +9,7 @@ keywords = [ "NeuralNetworks", "NNEF" ]
 categories = [ "science" ]
 autobenches = false
 edition = "2021"
-rust-version = "1.75"
+rust-version = "1.77"
 
 [badges]
 maintenance = { status = "actively-developed" }

nnef/cli/Cargo.toml

@@ -12,7 +12,7 @@ keywords = [ "NeuralNetworks", "NNEF" ]
 categories = [ "science" ]
 autobenches = false
 edition = "2021"
-rust-version = "1.75"
+rust-version = "1.77"
 
 [dependencies]
 anyhow.workspace = true

nnef/nnef-resources/Cargo.toml

@@ -12,7 +12,7 @@ keywords = [ "NeuralNetworks", "NNEF" ]
 categories = [ "science" ]
 autobenches = false
 edition = "2021"
-rust-version = "1.75"
+rust-version = "1.77"
 
 [badges]
 maintenance = { status = "actively-developed" }

onnx/Cargo.toml

@@ -11,7 +11,7 @@ autobenches = false
 edition = "2021"
 exclude = [ "test_cases" ]
 # build = "build-proto-rs"
-rust-version = "1.75"
+rust-version = "1.77"
 
 
 [badges]

tensorflow/Cargo.toml

@@ -9,7 +9,7 @@ keywords = [ "TensorFlow", "NeuralNetworks", "ONNX" ]
 categories = [ "science" ]
 autobenches = false
 edition = "2021"
-rust-version = "1.75"
+rust-version = "1.77"
 # build = "build-proto.rs"
 
 [badges]