[MLIR][IR] add -mlir-print-unique-ssa-ids to AsmPrinter #91241
Conversation
|
Thank you for submitting a Pull Request (PR) to the LLVM Project! This PR will be automatically labeled and the relevant teams will be If you wish to, you can add reviewers by using the "Reviewers" section on this page. If this is not working for you, it is probably because you do not have write If you have received no comments on your PR for a week, you can request a review If you have further questions, they may be answered by the LLVM GitHub User Guide. You can also ask questions in a comment on this PR, on the LLVM Discord or on the forums. |
|
@llvm/pr-subscribers-mlir @llvm/pr-subscribers-mlir-core Author: Scott Manley (rscottmanley) ChangesAdd an option to unique the numbers of values, block arguments and naming conflicts when requested and/or printing generic op form. This is helpful when debugging. For example, if you have: And you get a verifier error which says opB's "operand #0 does not dominate this use", it looks like %0 does dominate the use. This is not intuitive. If these were numbered uniquely, it would look like: And thus, much clearer as to why you are getting the error since %0 is out of scope. Since generic op form should aim to give you the most possible information, it seems like a good idea to use unique numbers in this situation. Adding an option also gives those an option to use it outside of generic op form. Full diff: https://github.com/llvm/llvm-project/pull/91241.diff 3 Files Affected:
diff --git a/mlir/include/mlir/IR/OperationSupport.h b/mlir/include/mlir/IR/OperationSupport.h
index e661bb87a27ed0..e0b88d3b1bb4f3 100644
--- a/mlir/include/mlir/IR/OperationSupport.h
+++ b/mlir/include/mlir/IR/OperationSupport.h
@@ -1219,6 +1219,9 @@ class OpPrintingFlags {
/// Return if the printer should print users of values.
bool shouldPrintValueUsers() const;
+ /// Return if printer should use unique IDs.
+ bool shouldPrintUniqueIDs() const;
+
private:
/// Elide large elements attributes if the number of elements is larger than
/// the upper limit.
@@ -1249,6 +1252,9 @@ class OpPrintingFlags {
/// Print users of values.
bool printValueUsersFlag : 1;
+
+ /// Print unique ids for values, block arguments and naming conflicts
+ bool printUniqueIDsFlag : 1;
};
//===----------------------------------------------------------------------===//
diff --git a/mlir/lib/IR/AsmPrinter.cpp b/mlir/lib/IR/AsmPrinter.cpp
index e915b97d9ff17b..72c84b57222092 100644
--- a/mlir/lib/IR/AsmPrinter.cpp
+++ b/mlir/lib/IR/AsmPrinter.cpp
@@ -189,6 +189,11 @@ struct AsmPrinterOptions {
"mlir-print-value-users", llvm::cl::init(false),
llvm::cl::desc(
"Print users of operation results and block arguments as a comment")};
+
+ llvm::cl::opt<bool> printUniqueIDs{
+ "mlir-print-unique-ids", llvm::cl::init(false),
+ llvm::cl::desc("Print unique id numbers for values, block arguments and "
+ "naming conflicts across all regions")};
};
} // namespace
@@ -206,7 +211,7 @@ OpPrintingFlags::OpPrintingFlags()
: printDebugInfoFlag(false), printDebugInfoPrettyFormFlag(false),
printGenericOpFormFlag(false), skipRegionsFlag(false),
assumeVerifiedFlag(false), printLocalScope(false),
- printValueUsersFlag(false) {
+ printValueUsersFlag(false), printUniqueIDsFlag(false) {
// Initialize based upon command line options, if they are available.
if (!clOptions.isConstructed())
return;
@@ -224,6 +229,7 @@ OpPrintingFlags::OpPrintingFlags()
printLocalScope = clOptions->printLocalScopeOpt;
skipRegionsFlag = clOptions->skipRegionsOpt;
printValueUsersFlag = clOptions->printValueUsers;
+ printUniqueIDsFlag = clOptions->printUniqueIDs;
}
/// Enable the elision of large elements attributes, by printing a '...'
@@ -350,6 +356,11 @@ bool OpPrintingFlags::shouldPrintValueUsers() const {
return printValueUsersFlag;
}
+/// Return if the printer should use unique IDs.
+bool OpPrintingFlags::shouldPrintUniqueIDs() const {
+ return printUniqueIDsFlag || shouldPrintGenericOpForm();
+}
+
//===----------------------------------------------------------------------===//
// NewLineCounter
//===----------------------------------------------------------------------===//
@@ -1369,8 +1380,14 @@ SSANameState::SSANameState(Operation *op, const OpPrintingFlags &printerFlags)
while (!nameContext.empty()) {
Region *region;
UsedNamesScopeTy *parentScope;
- std::tie(region, nextValueID, nextArgumentID, nextConflictID, parentScope) =
- nameContext.pop_back_val();
+
+ if (printerFlags.shouldPrintUniqueIDs())
+ // When printing unique IDs, ignore saved ID counts from parent regions
+ std::tie(region, std::ignore, std::ignore, std::ignore, parentScope) =
+ nameContext.pop_back_val();
+ else
+ std::tie(region, nextValueID, nextArgumentID, nextConflictID,
+ parentScope) = nameContext.pop_back_val();
// When we switch from one subtree to another, pop the scopes(needless)
// until the parent scope.
diff --git a/mlir/test/IR/print-unique-ids.mlir b/mlir/test/IR/print-unique-ids.mlir
new file mode 100644
index 00000000000000..8be57e0259de19
--- /dev/null
+++ b/mlir/test/IR/print-unique-ids.mlir
@@ -0,0 +1,40 @@
+// RUN: mlir-opt -mlir-print-unique-ids %s | FileCheck %s
+
+// CHECK: %arg5
+// CHECK: %15
+module {
+ func.func @uniqueConflicts(%arg0 : memref<i32>, %arg1 : memref<i32>) {
+ %c0 = arith.constant 0 : index
+ %c1 = arith.constant 1 : index
+ %c8 = arith.constant 8 : index
+ scf.for %arg2 = %c0 to %c8 step %c1 {
+ %a = memref.load %arg0[] : memref<i32>
+ %b = memref.load %arg1[] : memref<i32>
+ %0 = arith.addi %a, %b : i32
+ %1 = arith.subi %a, %b : i32
+ scf.for %arg3 = %c0 to %c8 step %c1 {
+ %a2 = memref.load %arg0[] : memref<i32>
+ %b2 = memref.load %arg1[] : memref<i32>
+ %2 = arith.addi %a2, %b2 : i32
+ %3 = arith.subi %a2, %b2 : i32
+ scf.yield
+ }
+ scf.for %arg3 = %c0 to %c8 step %c1 {
+ %a2 = memref.load %arg0[] : memref<i32>
+ %b2 = memref.load %arg1[] : memref<i32>
+ %2 = arith.addi %a2, %b2 : i32
+ %3 = arith.subi %a2, %b2 : i32
+ scf.yield
+ }
+ scf.yield
+ }
+ scf.for %arg2 = %c0 to %c8 step %c1 {
+ %a = memref.load %arg0[] : memref<i32>
+ %b = memref.load %arg1[] : memref<i32>
+ %0 = arith.addi %a, %b : i32
+ %1 = arith.subi %a, %b : i32
+ scf.yield
+ }
+ return
+ }
+}
|
mlir/lib/IR/AsmPrinter.cpp
Outdated
| @@ -189,6 +189,11 @@ struct AsmPrinterOptions { | |||
| "mlir-print-value-users", llvm::cl::init(false), | |||
| llvm::cl::desc( | |||
| "Print users of operation results and block arguments as a comment")}; | |||
|
|
|||
| llvm::cl::opt<bool> printUniqueIDs{ | |||
| "mlir-print-unique-ids", llvm::cl::init(false), | |||
There was a problem hiding this comment.
Should this be -mlir-print-unique-ssa-ids to be more explicit? "unique ids" seems a bit generic.
Or an option with a named value: -mlir-print-ssa-ids-scope={local, global}
There was a problem hiding this comment.
+1 to being explicit about ssa
There was a problem hiding this comment.
Unless you feel strongly it should be a named option, I like -mlir-print-unique-ssa-ids.
mlir/test/IR/print-unique-ids.mlir
Outdated
| @@ -0,0 +1,40 @@ | |||
| // RUN: mlir-opt -mlir-print-unique-ids %s | FileCheck %s | |||
There was a problem hiding this comment.
I would also add the negative test:
// RUN: mlir-opt %s | FileCheck %s --check-prefix=LOCAL_SCOPE
And:
// LOCAL_SCOPE-NOT: %arg5
// LOCAL_SCOPE-NOT: %15
mlir/lib/IR/AsmPrinter.cpp
Outdated
| @@ -189,6 +189,11 @@ struct AsmPrinterOptions { | |||
| "mlir-print-value-users", llvm::cl::init(false), | |||
| llvm::cl::desc( | |||
| "Print users of operation results and block arguments as a comment")}; | |||
|
|
|||
| llvm::cl::opt<bool> printUniqueIDs{ | |||
| "mlir-print-unique-ids", llvm::cl::init(false), | |||
There was a problem hiding this comment.
+1 to being explicit about ssa
mlir/test/IR/print-unique-ids.mlir
Outdated
| // CHECK: %arg5 | ||
| // CHECK: %15 | ||
| module { | ||
| func.func @uniqueConflicts(%arg0 : memref<i32>, %arg1 : memref<i32>) { |
There was a problem hiding this comment.
This test seems unnecessarily large. Don't you only need to test a function with two scf.for ops?
There was a problem hiding this comment.
I suppose I wanted to make sure it worked for nesting depths > 1, but I can reduce.
rscottmanley
force-pushed
the
rscottmanley/mlir-asm-printer
branch
from
04c08ab to
1cde917
Compare
rscottmanley
changed the title
rscottmanley
force-pushed
the
rscottmanley/mlir-asm-printer
branch
from
1cde917 to
ce517a2
Compare
Add an option to print unique SSA IDs of values, block arguments and naming
conflicts when requested and/or printing generic op form. This is helpful
when debugging. For example, if you have:
scf.for
%0 =
%1 = opA %0
scf.for
%0 =
%1 = opB %0
And you get a verifier error which says opB's "operand #0 does not dominate
this use", it looks like %0 does dominate the use. This is not intuitive. If
these were numbered uniquely, it would look like:
scf.for
%0 =
%1 = opA %0
scf.for
%2 =
%3 = opB %0
And thus, much clearer as to why you are getting the error since %0 is out
of scope. Since generic op form should aim to give you the most possible
information, it seems like a good idea to use unique numbers in this
situation. Adding an option also gives those an option to use it outside
of generic op form.
rscottmanley
force-pushed
the
rscottmanley/mlir-asm-printer
branch
from
ce517a2 to
11fe987
Compare
|
@rscottmanley Congratulations on having your first Pull Request (PR) merged into the LLVM Project! Your changes will be combined with recent changes from other authors, then tested Please check whether problems have been caused by your change specifically, as How to do this, and the rest of the post-merge process, is covered in detail here. If your change does cause a problem, it may be reverted, or you can revert it yourself. If you don't get any reports, no action is required from you. Your changes are working as expected, well done! |
Add an option to unique the numbers of values, block arguments and naming conflicts when requested and/or printing generic op form. This is helpful when debugging. For example, if you have:
And you get a verifier error which says opB's "operand #0 does not dominate this use", it looks like %0 does dominate the use. This is not intuitive. If these were numbered uniquely, it would look like:
And thus, much clearer as to why you are getting the error since %0 is out of scope. Since generic op form should aim to give you the most possible information, it seems like a good idea to use unique numbers in this situation. Adding an option also gives those an option to use it outside of generic op form.