[mlir][arith] Add printing integration tests

[pingshiyu]

Printing operations regression tests, from the original larger PR that has been broken down: #92272

[llvmbot]

@llvm/pr-subscribers-mlir

@llvm/pr-subscribers-mlir-arith

Author: Jacob Yu (pingshiyu)

Changes

Printing operations regression tests, from the original larger PR that has been broken down: #92272

---

Full diff: https://github.com/llvm/llvm-project/pull/98184.diff

1 Files Affected:

• (added) mlir/test/Integration/Dialect/Arith/CPU/print.mlir (+47)

diff --git a/mlir/test/Integration/Dialect/Arith/CPU/print.mlir b/mlir/test/Integration/Dialect/Arith/CPU/print.mlir
new file mode 100644
index 0000000000000..d322c23952bab
--- /dev/null
+++ b/mlir/test/Integration/Dialect/Arith/CPU/print.mlir
@@ -0,0 +1,47 @@
+// RUN: mlir-opt %s --convert-scf-to-cf --convert-cf-to-llvm --convert-vector-to-llvm \
+// RUN:             --convert-func-to-llvm --convert-arith-to-llvm | \
+// RUN:   mlir-cpu-runner -e entry -entry-point-result=void \
+// RUN:                   --shared-libs=%mlir_c_runner_utils | \
+// RUN:   FileCheck %s --match-full-lines
+
+func.func @i1() {
+    // printing i1 values
+    // print(0 : i1) = '0'; print(1 : i1) = '1'; print(-1 : i1) = '1'
+    // CHECK:      0
+    // CHECK-NEXT: 1
+    // CHECK-NEXT: 1
+    %false = arith.constant false
+    %true = arith.constant 1 : i1
+    %true_as_n1 = arith.constant -1 : i1
+    vector.print %false : i1
+    vector.print %true : i1
+    vector.print %true_as_n1 : i1
+    return
+}
+
+func.func @index() {
+    // printing index values
+    // print(0 : index) = '0'; print(1 : index) = '1'; print(-1 : index) = '2^w - 1'
+    // index constants are printed as unsigned
+    // vector.print(arith.constant(x)) ~= toUnsignedRepr x
+    // CHECK-NEXT: 0
+    // CHECK-NEXT: 1
+    // CHECK-NEXT: 18446744073709551615
+    // CHECK-NEXT: 63
+    %c0 = arith.constant 0 : index
+    %c1 = arith.constant 1 : index
+    %cn1 = arith.constant -1 : index
+    %c63 = arith.constant 63 : index
+
+    vector.print %c0 : index
+    vector.print %c1 : index
+    vector.print %cn1 : index
+    vector.print %c63 : index
+    return
+}
+
+func.func @entry() {
+    func.call @i1() : () -> ()
+    func.call @index() : () -> ()
+    return
+}

[kuhar]

Could we use either 0 / 1 or false / true for consistency here?

[kuhar]

Could we specify the index type in the mlir-opt pipeline above? I'm worried this can be printed as u64 or u32 depending on the host

[joker-eph]

Thanks for adding tests, we surely have holes here and there and not enough people closing them... :)

Printing operations regression tests, from the original larger PR that has been broken down: #92272

Please provide a standalone description, we don't need to refer another PR here and instead I'd rather see here a clear "why" these tests are added.

Thanks!

[joker-eph]

From what I can tell, this is not testing anything about arith is it? Seems like this is all about testing "vector.print" in itself, or even just the testing runtime we have.
Can you try to make the file location more relevant?

Can we also consolidate all the testing for the runtime in a single execution? (each new lit test isn't free: it'll fork/exec a compiler process, and then execute the result here).

[banach-space]

Can we also consolidate all the testing for the runtime in a single execution? (each new lit test isn't free: it'll fork/exec a compiler process, and then execute the result here).

I advocated for keeping them in separate files - to make sure the tests are minimal and to make it easier to identify what's being tested. That has made reviewing much easier, but I didn't really think about the runtime cost. Consolidation makes sense, but l'd wait until all PR's from #100121 are complete. Is that OK?

[banach-space]

From what I can tell, this is duplicating https://github.com/llvm/llvm-project/blob/main/mlir/test/Integration/Dialect/Vector/CPU/print-int.mlir, no?