Skip to content

Fix reverse mode for memref.alloca #2826

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
xys-syx wants to merge 1 commit into
base: main
Choose a base branch
from
+123 −1
Open

Fix reverse mode for memref.alloca #2826

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
14 changes: 13 additions & 1 deletion enzyme/Enzyme/MLIR/Interfaces/GradientUtils.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -225,8 +225,20 @@ void mlir::enzyme::MGradientUtils::setDiffe(mlir::Value val, mlir::Value toset,
}
assert(!isConstantValue(val));

// Reverse mode also uses invertedPointers for mutable values. For allocation
// results, forceAugmentedReturns() installs a placeholder shadow; the
// allocation derivative later calls setDiffe with the real shadow, so replace
// it here.
bool isMutable = cast<AutoDiffTypeInterface>(val.getType()).isMutable();
if (mode == DerivativeMode::ForwardMode ||
mode == DerivativeMode::ForwardModeSplit) {
mode == DerivativeMode::ForwardModeSplit || isMutable) {
Comment thread
vimarsh6739 marked this conversation as resolved.
if (isMutable) {
// Mutable arguments may already be mapped to a real user-provided shadow
// rather than a placeholder. Only replace placeholders here.
auto existing = invertedPointers.lookupOrNull(val);
Comment thread
vimarsh6739 marked this conversation as resolved.
if (existing && !existing.getDefiningOp<enzyme::PlaceholderOp>())
return;
}
setInvertedPointer(val, toset);
}
/*
Expand Down
110 changes: 110 additions & 0 deletions enzyme/test/MLIR/ReverseMode/alloca.mlir
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,110 @@
// RUN: %eopt --enzyme %s | FileCheck --implicit-check-not=enzyme.placeholder %s

func.func @foo_flat(%x : f64) -> f64 {
%buf = memref.alloca() : memref<f64>
memref.store %x, %buf[] : memref<f64>
%y = memref.load %buf[] : memref<f64>
return %y : f64
}

func.func @dfoo_flat(%x: f64, %dout : f64) -> f64 {
%dx = enzyme.autodiff @foo_flat(%x, %dout) {
activity = [#enzyme<activity enzyme_active>],
ret_activity = [#enzyme<activity enzyme_activenoneed>]
} : (f64, f64) -> (f64)
return %dx : f64
}

// CHECK-LABEL: func.func private @diffefoo_flat(
// CHECK-SAME: %[[X:[^,]+]]: f64,
// CHECK-SAME: %[[DOUT:[^)]+]]: f64) -> f64 {

// CHECK: %[[GX:.+]] = "enzyme.init"() : () -> !enzyme.Gradient<f64>
// CHECK: "enzyme.set"(%[[GX]], %{{.*}}) : (!enzyme.Gradient<f64>, f64) -> ()
// CHECK: %[[CS:.+]] = "enzyme.init"() : () -> !enzyme.Cache<memref<f64>>
// CHECK: %[[CL:.+]] = "enzyme.init"() : () -> !enzyme.Cache<memref<f64>>
// CHECK: %[[GY:.+]] = "enzyme.init"() : () -> !enzyme.Gradient<f64>
// CHECK: "enzyme.set"(%[[GY]], %{{.*}}) : (!enzyme.Gradient<f64>, f64) -> ()

// CHECK: %[[DBUF:.+]] = memref.alloca() : memref<f64>
// CHECK: %[[ZERO_INIT:.+]] = arith.constant 0.000000e+00 : f64
// CHECK: linalg.fill ins(%[[ZERO_INIT]] : f64) outs(%[[DBUF]] : memref<f64>)

// CHECK: %[[BUF:.+]] = memref.alloca() : memref<f64>
// CHECK: "enzyme.push"(%[[CS]], %[[DBUF]]) : (!enzyme.Cache<memref<f64>>, memref<f64>) -> ()
// CHECK: memref.store %[[X]], %[[BUF]][] : memref<f64>
// CHECK: "enzyme.push"(%[[CL]], %[[DBUF]]) : (!enzyme.Cache<memref<f64>>, memref<f64>) -> ()
// CHECK: memref.load %[[BUF]][] : memref<f64>
// CHECK: cf.br ^bb1
// CHECK: ^bb1:

// CHECK: %{{.+}} = "enzyme.get"(%[[GY]]) : (!enzyme.Gradient<f64>) -> f64
// CHECK: arith.addf %{{.+}}, %[[DOUT]] : f64
// CHECK: "enzyme.set"(%[[GY]], %{{.*}}) : (!enzyme.Gradient<f64>, f64) -> ()

// CHECK: %{{.+}} = "enzyme.get"(%[[GY]]) : (!enzyme.Gradient<f64>) -> f64
// CHECK: %[[POPL:.+]] = "enzyme.pop"(%[[CL]]) : (!enzyme.Cache<memref<f64>>) -> memref<f64>
// CHECK: memref.load %[[POPL]][] : memref<f64>
// CHECK: arith.addf
// CHECK: memref.store %{{.+}}, %[[POPL]][] : memref<f64>

// CHECK: %[[POPS:.+]] = "enzyme.pop"(%[[CS]]) : (!enzyme.Cache<memref<f64>>) -> memref<f64>
// CHECK: memref.load %[[POPS]][] : memref<f64>
// CHECK: %{{.+}} = "enzyme.get"(%[[GX]]) : (!enzyme.Gradient<f64>) -> f64
// CHECK: arith.addf
// CHECK: "enzyme.set"(%[[GX]], %{{.*}}) : (!enzyme.Gradient<f64>, f64) -> ()
// CHECK: %[[ZERO_CLEAR:.+]] = arith.constant 0.000000e+00 : f64
// CHECK: memref.store %[[ZERO_CLEAR]], %[[POPS]][] : memref<f64>

// CHECK: %{{.+}} = "enzyme.get"(%[[GX]]) : (!enzyme.Gradient<f64>) -> f64
// CHECK: return %{{.+}} : f64


// -----

func.func @load_arg(%mem: memref<f64>) -> f64 {
%buf = memref.alloca() : memref<f64>
%x = memref.load %mem[] : memref<f64>
memref.store %x, %buf[] : memref<f64>
%y = memref.load %buf[] : memref<f64>
return %y : f64
}

func.func @dload_arg(%mem: memref<f64>, %dmem: memref<f64>, %dout : f64) {
enzyme.autodiff @load_arg(%mem, %dmem, %dout) {
activity = [#enzyme<activity enzyme_dup>],
ret_activity = [#enzyme<activity enzyme_activenoneed>]
} : (memref<f64>, memref<f64>, f64) -> ()
return
}

// CHECK-LABEL: func.func private @diffeload_arg(
// CHECK-SAME: %[[MEM:[^,]+]]: memref<f64>,
// CHECK-SAME: %[[DMEM:[^,]+]]: memref<f64>,
// CHECK-SAME: %[[DOUT:[^)]+]]: f64) {
// CHECK: %[[DBUF:.+]] = memref.alloca() : memref<f64>
// CHECK: linalg.fill ins(%{{.+}} : f64) outs(%[[DBUF]] : memref<f64>)
// CHECK: %[[BUF:.+]] = memref.alloca() : memref<f64>
// CHECK: "enzyme.push"(%{{.+}}, %[[DMEM]]) : (!enzyme.Cache<memref<f64>>, memref<f64>) -> ()
// CHECK: memref.load %[[MEM]][] : memref<f64>
// CHECK: "enzyme.push"(%{{.+}}, %[[DBUF]]) : (!enzyme.Cache<memref<f64>>, memref<f64>) -> ()
// CHECK: memref.store %{{.+}}, %[[BUF]][] : memref<f64>
// CHECK: "enzyme.push"(%{{.+}}, %[[DBUF]]) : (!enzyme.Cache<memref<f64>>, memref<f64>) -> ()
// CHECK: memref.load %[[BUF]][] : memref<f64>
// CHECK: arith.addf %{{.+}}, %[[DOUT]] : f64
// CHECK: %[[POPL:.+]] = "enzyme.pop"(%{{.+}}) : (!enzyme.Cache<memref<f64>>) -> memref<f64>
// CHECK: memref.load %[[POPL]][] : memref<f64>
// CHECK: arith.addf
// CHECK: memref.store %{{.+}}, %[[POPL]][] : memref<f64>
// CHECK: %[[POPS:.+]] = "enzyme.pop"(%{{.+}}) : (!enzyme.Cache<memref<f64>>) -> memref<f64>
// CHECK: memref.load %[[POPS]][] : memref<f64>
// CHECK: "enzyme.get"
// CHECK: arith.addf
// CHECK: "enzyme.set"
// CHECK: memref.store %{{.+}}, %[[POPS]][] : memref<f64>
// CHECK: "enzyme.get"
// CHECK: %[[POPD:.+]] = "enzyme.pop"(%{{.+}}) : (!enzyme.Cache<memref<f64>>) -> memref<f64>
// CHECK: memref.load %[[POPD]][] : memref<f64>
// CHECK: arith.addf
// CHECK: memref.store %{{.+}}, %[[POPD]][] : memref<f64>
// CHECK: return
Loading