Small fixes for slice edge cases (#476)

Author: dan-garvey

e2e_testing/torchscript/slice_like.py

@@ -32,7 +32,6 @@ def SliceModule_basic(module, tu: TestUtils):
 
 # ==============================================================================
 
-# This Test currently xfails due to https://github.com/llvm/torch-mlir/issues/448
 class SliceOutOfUpperBoundIndexModule(torch.nn.Module):
     def __init__(self):
         super().__init__()
@@ -43,8 +42,11 @@ def __init__(self):
         ([-1, -1, -1], torch.float32, True),
     ])
     def forward(self, x):
-        return x[:8, :5, 8:]
-
+        # TODO: remove hacky cat tensor once refbackend supports 0 size dim
+        result =  x[:8, :5, 8:]
+        cat_tensor = torch.ones((6,4,1), dtype=torch.float32)
+        return torch.cat((result,cat_tensor), dim=2)
+        
 
 @register_test_case(module_factory=lambda: SliceOutOfUpperBoundIndexModule())
 def SliceOutOfUpperBoundIndexModule_basic(module, tu: TestUtils):
@@ -90,7 +92,7 @@ def SliceOutOfLowerBoundStartIndexModule_basic(module, tu: TestUtils):
 
 # ==============================================================================
 
-# This Test currently xfails due to https://github.com/llvm/torch-mlir/issues/448
+
 class SliceEndSleStartModule(torch.nn.Module):
     def __init__(self):
         super().__init__()
@@ -101,7 +103,10 @@ def __init__(self):
         ([-1, -1, -1], torch.float32, True),
     ])
     def forward(self, x):
-        return x[:0, 4:3, :-7]
+        # TODO: remove hacky cat tensor once refbackend supports 0 size dim
+        result = x[:, 4:3, :]
+        cat_tensor = torch.ones((6,1,7), dtype=torch.float32)
+        return torch.cat((result, cat_tensor), dim=1)
 
 
 @register_test_case(module_factory=lambda: SliceEndSleStartModule())
@@ -110,7 +115,7 @@ def SliceEndSleStartModule_basic(module, tu: TestUtils):
 
 # ==============================================================================
 
-# This Test currently xfails due to https://github.com/llvm/torch-mlir/issues/448
+
 class SliceStartEqEndModule(torch.nn.Module):
     def __init__(self):
         super().__init__()
@@ -121,7 +126,10 @@ def __init__(self):
         ([-1, -1, -1], torch.float32, True),
     ])
     def forward(self, x):
-        return x[5:5, 3:3, -1:]
+        # TODO: remove hacky cat tensor once refbackend supports 0 size dim
+        result = x[5:5, :, :]
+        cat_tensor = torch.ones((1,4,7), dtype=torch.float32)
+        return torch.cat((result, cat_tensor), dim=0)
 
 
 @register_test_case(module_factory=lambda: SliceStartEqEndModule())

e2e_testing/torchscript/xfail_sets.py

@@ -17,13 +17,7 @@
     "QuantizedMLP_basic",
     "IouOfModule_basic",
 }
-# Fails due to https://github.com/llvm/torch-mlir/issues/448
-SIZE_ZERO_TENSOR_XFAILS = {
-    "SliceEndSleStartModule_basic",
-    "SliceStartEqEndModule_basic",
-    "SliceOutOfUpperBoundIndexModule_basic",
-}
-REFBACKEND_XFAIL_SET = set.union(COMMON_TORCH_MLIR_LOWERING_XFAILS, SIZE_ZERO_TENSOR_XFAILS)
+REFBACKEND_XFAIL_SET = COMMON_TORCH_MLIR_LOWERING_XFAILS
 
 # Write the TOSA set as a "passing" set as it is very early in development
 # and very few tests work yet.

lib/Conversion/TorchToLinalg/TorchToLinalg.cpp

@@ -3047,9 +3047,17 @@ class ConvertAtenSliceTensorOp : public OpConversionPattern<AtenSliceTensorOp> {
       return castIntToIndex(rewriter, loc, startOrEndBoundedByDimSize);
     };
 
+    if (op.start().getType().isa<OptionalType>() ||
+        op.end().getType().isa<OptionalType>())
+      return rewriter.notifyMatchFailure(op, "unimplemented optional type arg");
     Value start = adjustStartOrEnd(op.start(), adaptor.start(), zero);
     Value end = adjustStartOrEnd(op.end(), adaptor.end(), dimSize);
 
+    // end >= start ? end : start
+    Value endSgeStart = rewriter.create<arith::CmpIOp>(
+        loc, arith::CmpIPredicate::sge, end, start);
+    end = rewriter.create<SelectOp>(loc, endSgeStart, end, start);
+
     int64_t step;
     if (!matchPattern(op.step(), m_TorchConstantInt(&step))) {
       if (!op.step().getType().isa<Torch::NoneType>())