[Mlir-commits] [mlir] [mlir][arith] Add comparison regression tests (PR #96974)

Jakub Kuderski llvmlistbot at llvm.org
Mon Jul 8 08:42:11 PDT 2024


================
@@ -0,0 +1,123 @@
+// 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 @cmpi_slt_i1(%v1 : i1, %v2 : i1) {
+  vector.print str "@cmpi_slt_i1\n"
+  %res = arith.cmpi slt, %v1, %v2 : i1
+  vector.print %res : i1
+  return
+}
+
+func.func @cmpi_sle_i1(%v1 : i1, %v2 : i1) {
+  vector.print str "@cmpi_sle_i1\n"
+  %res = arith.cmpi sle, %v1, %v2 : i1
+  vector.print %res : i1
+  return
+}
+
+func.func @cmpi_sgt_i1(%v1 : i1, %v2 : i1) {
+  vector.print str "@cmpi_sgt_i1\n"
+  %res = arith.cmpi sgt, %v1, %v2 : i1
+  vector.print %res : i1
+  return
+}
+
+func.func @cmpi_sge_i1(%v1 : i1, %v2 : i1) {
+  vector.print str "@cmpi_sge_i1\n"
+  %res = arith.cmpi sge, %v1, %v2 : i1
+  vector.print %res : i1
+  return
+}
+
+func.func @cmpi_signed() {
+  // ------------------------------------------------
+  // Test i1
+  // ------------------------------------------------
+  %false_i1 = arith.constant 0 : i1
+  %true_i1 = arith.constant 1 : i1
+  %true_i1_n1 = arith.constant -1 : i1
+
+  // sge 0 -1, sge 0 1, should be true
+  // since the bitvector `1` is interpreted as the int value -1 in signed comparisons
+  // sge 0 -1 == sge 0 1 == true (1)
+
+  // CHECK-LABEL: @cmpi_sge_i1
+  // CHECK-NEXT:  1
+  func.call @cmpi_sge_i1(%false_i1, %true_i1_n1) : (i1, i1) -> ()
+
+  // CHECK-LABEL: @cmpi_sge_i1
+  // CHECK-NEXT:  1
+  func.call @cmpi_sge_i1(%false_i1, %true_i1) : (i1, i1) -> ()
+
+  %false = arith.constant false
+  %true = arith.constant true
+
+  // signed comparisons on i1s
----------------
kuhar wrote:

+1, this is slightly misleading

https://github.com/llvm/llvm-project/pull/96974


More information about the Mlir-commits mailing list