[Mlir-commits] [mlir] [mlir][sparse] fold explicit value during sparsification (PR #90530)
Aart Bik
llvmlistbot at llvm.org
Mon Apr 29 15:48:48 PDT 2024
================
@@ -498,9 +498,17 @@ static Value genTensorLoad(CodegenEnv &env, OpBuilder &builder, ExprId exp) {
Value val = env.exp(exp).val;
if (val)
return val;
- // Load during insertion.
+ // Get tensor operand.
linalg::GenericOp op = env.op();
+ Location loc = op.getLoc();
OpOperand *t = &op->getOpOperand(env.exp(exp).tensor);
+ // Fold binary-valued tensor into explicit value.
+ const auto stt = getSparseTensorType(t->get());
+ if (stt.hasEncoding()) {
+ if (auto explVal = stt.getExplicitVal())
----------------
aartbik wrote:
We think alike, I noticed that too. Seems cleaner indeed.
https://github.com/llvm/llvm-project/pull/90530
More information about the Mlir-commits
mailing list