[mlir][sparse][gpu] migration to sparse_tensor.print #83510
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Continuous efforts llvm#83357 for our sparse CUDA tests
aartbik
requested review from
PeimingLiu,
yinying-lisa-li and
matthias-springer
as code owners
|
@llvm/pr-subscribers-mlir @llvm/pr-subscribers-mlir-sparse Author: Aart Bik (aartbik) ChangesContinuous efforts #83357 for our sparse CUDA tests Full diff: https://github.com/llvm/llvm-project/pull/83510.diff 4 Files Affected:
diff --git a/mlir/test/Integration/Dialect/SparseTensor/GPU/CUDA/sparse-gemm-lib.mlir b/mlir/test/Integration/Dialect/SparseTensor/GPU/CUDA/sparse-gemm-lib.mlir
index 73faf7d9b7b9a3..984b3614eded56 100644
--- a/mlir/test/Integration/Dialect/SparseTensor/GPU/CUDA/sparse-gemm-lib.mlir
+++ b/mlir/test/Integration/Dialect/SparseTensor/GPU/CUDA/sparse-gemm-lib.mlir
@@ -62,23 +62,15 @@ module {
tensor<8x8xf32, #CSR>) -> tensor<8x8xf32, #CSR>
//
- // Verify computed result (expected output, with only 20 nonzeros).
+ // Verify computed result.
//
- // CHECK: ( ( 1, 39, 52, 0, 0, 0, 45, 51 ),
- // CHECK-SAME: ( 0, 0, 0, 0, 0, 0, 0, 0 ),
- // CHECK-SAME: ( 0, 0, 16, 0, 0, 0, 0, 0 ),
- // CHECK-SAME: ( 0, 0, 0, 25, 0, 0, 0, 0 ),
- // CHECK-SAME: ( 0, 0, 0, 0, 36, 0, 0, 0 ),
- // CHECK-SAME: ( 0, 117, 158, 0, 0, 0, 135, 144 ),
- // CHECK-SAME: ( 0, 156, 318, 0, 0, 0, 301, 324 ),
- // CHECK-SAME: ( 0, 208, 430, 0, 0, 0, 405, 436 ) )
- // CHECK-NEXT: 20
- %d = sparse_tensor.convert %Ccsr : tensor<8x8xf32, #CSR> to tensor<8x8xf32>
- %v = vector.transfer_read %d[%c0, %c0], %f0: tensor<8x8xf32>, vector<8x8xf32>
- vector.print %v : vector<8x8xf32>
- %nnz = sparse_tensor.number_of_entries %Ccsr : tensor<8x8xf32, #CSR>
- %x = sparse_tensor.number_of_entries %Ccsr : tensor<8x8xf32, #CSR>
- vector.print %nnz : index
+ // CHECK: ---- Sparse Tensor ----
+ // CHECK-NEXT: nse = 20
+ // CHECK-NEXT: pos[1] : ( 0, 5, 5, 6, 7, 8, 12, 16, 20,
+ // CHECK-NEXT: crd[1] : ( 0, 1, 2, 6, 7, 2, 3, 4, 1, 2, 6, 7, 1, 2, 6, 7, 1, 2, 6, 7,
+ // CHECK-NEXT: values : ( 1, 39, 52, 45, 51, 16, 25, 36, 117, 158, 135, 144, 156, 318, 301, 324, 208, 430, 405, 436,
+ // CHECK-NEXT: ----
+ sparse_tensor.print %Ccsr : tensor<8x8xf32, #CSR>
llvm.call @mgpuDestroySparseEnv(): () -> ()
return
diff --git a/mlir/test/Integration/Dialect/SparseTensor/GPU/CUDA/sparse-matmul-lib.mlir b/mlir/test/Integration/Dialect/SparseTensor/GPU/CUDA/sparse-matmul-lib.mlir
index 67b6e60ab5c869..0e7c75bc41aae2 100644
--- a/mlir/test/Integration/Dialect/SparseTensor/GPU/CUDA/sparse-matmul-lib.mlir
+++ b/mlir/test/Integration/Dialect/SparseTensor/GPU/CUDA/sparse-matmul-lib.mlir
@@ -66,32 +66,16 @@ module {
return %D: tensor<8x8xf32>
}
+ // Helper to dump dense tensor as series of vectors.
func.func @dump(%mat: tensor<8x8xf32>) {
%f0 = arith.constant 0.0 : f32
%c0 = arith.constant 0 : index
%c1 = arith.constant 1 : index
- %c2 = arith.constant 2 : index
- %c3 = arith.constant 3 : index
- %c4 = arith.constant 4 : index
- %c5 = arith.constant 5 : index
- %c6 = arith.constant 6 : index
- %c7 = arith.constant 7 : index
- %r0 = vector.transfer_read %mat[%c0,%c0], %f0 : tensor<8x8xf32>, vector<8xf32>
- vector.print %r0 : vector<8xf32>
- %r1 = vector.transfer_read %mat[%c1,%c0], %f0 : tensor<8x8xf32>, vector<8xf32>
- vector.print %r1 : vector<8xf32>
- %r2 = vector.transfer_read %mat[%c2,%c0], %f0 : tensor<8x8xf32>, vector<8xf32>
- vector.print %r2 : vector<8xf32>
- %r3 = vector.transfer_read %mat[%c3,%c0], %f0 : tensor<8x8xf32>, vector<8xf32>
- vector.print %r3 : vector<8xf32>
- %r4 = vector.transfer_read %mat[%c4,%c0], %f0 : tensor<8x8xf32>, vector<8xf32>
- vector.print %r4 : vector<8xf32>
- %r5 = vector.transfer_read %mat[%c5,%c0], %f0 : tensor<8x8xf32>, vector<8xf32>
- vector.print %r5 : vector<8xf32>
- %r6 = vector.transfer_read %mat[%c6,%c0], %f0 : tensor<8x8xf32>, vector<8xf32>
- vector.print %r6 : vector<8xf32>
- %r7 = vector.transfer_read %mat[%c7,%c0], %f0 : tensor<8x8xf32>, vector<8xf32>
- vector.print %r7 : vector<8xf32>
+ %c8 = arith.constant 8 : index
+ scf.for %i = %c0 to %c8 step %c1 {
+ %v = vector.transfer_read %mat[%i,%c0], %f0 : tensor<8x8xf32>, vector<8xf32>
+ vector.print %v : vector<8xf32>
+ }
return
}
diff --git a/mlir/test/Integration/Dialect/SparseTensor/GPU/CUDA/sparse-sampled-matmul-lib.mlir b/mlir/test/Integration/Dialect/SparseTensor/GPU/CUDA/sparse-sampled-matmul-lib.mlir
index 9b33f081512390..aad26556b00e1e 100644
--- a/mlir/test/Integration/Dialect/SparseTensor/GPU/CUDA/sparse-sampled-matmul-lib.mlir
+++ b/mlir/test/Integration/Dialect/SparseTensor/GPU/CUDA/sparse-sampled-matmul-lib.mlir
@@ -113,11 +113,13 @@ module {
//
// Print the result for verification.
//
- // CHECK: ( 11, 41.4, 42, 102.5, 93, 44.1, 164, 105.2, 255 )
- //
- %vm = sparse_tensor.values %0 : tensor<?x?xf32, #CSR> to memref<?xf32>
- %vv = vector.transfer_read %vm[%c0], %d0 : memref<?xf32>, vector<9xf32>
- vector.print %vv : vector<9xf32>
+ // CHECK: ---- Sparse Tensor ----
+ // CHECK-NEXT: nse = 9
+ // CHECK-NEXT: pos[1] : ( 0, 2, 4, 5, 7, 9,
+ // CHECK-NEXT: crd[1] : ( 0, 3, 1, 4, 2, 0, 3, 1, 4,
+ // CHECK-NEXT: values : ( 11, 41.4, 42, 102.5, 93, 44.1, 164, 105.2, 255,
+ // CHECK-NEXT: ----
+ sparse_tensor.print %0 : tensor<?x?xf32, #CSR>
// Create a much sparser sampling matrix.
%t = arith.constant sparse<[[0,0], [0,1], [1,0], [3,4], [7,7]],
@@ -137,11 +139,14 @@ module {
//
// Print the result for verification.
//
- // CHECK: ( ( 17, 18, 0, 0, 0, 0, 0, 0 ), ( 19, 0, 0, 0, 0, 0, 0, 0 ), ( 0, 0, 0, 0, 0, 0, 0, 0 ), ( 0, 0, 0, 0, 20, 0, 0, 0 ), ( 0, 0, 0, 0, 0, 0, 0, 0 ), ( 0, 0, 0, 0, 0, 0, 0, 0 ), ( 0, 0, 0, 0, 0, 0, 0, 0 ), ( 0, 0, 0, 0, 0, 0, 0, 21 ) )
+ // CHECK: ---- Sparse Tensor ----
+ // CHECK-NEXT: nse = 5
+ // CHECK-NEXT: pos[1] : ( 0, 2, 3, 3, 4, 4, 4, 4, 5,
+ // CHECK-NEXT: crd[1] : ( 0, 1, 0, 4, 7,
+ // CHECK-NEXT: values : ( 17, 18, 19, 20, 21,
+ // CHECK-NEXT: ----
//
- %d = sparse_tensor.convert %1 : tensor<?x?xf32, #CSR> to tensor<?x?xf32>
- %mm = vector.transfer_read %d[%c0, %c0], %d0 : tensor<?x?xf32>, vector<8x8xf32>
- vector.print %mm : vector<8x8xf32>
+ sparse_tensor.print %1 : tensor<?x?xf32, #CSR>
// Release the resources.
bufferization.dealloc_tensor %0 : tensor<?x?xf32, #CSR>
diff --git a/mlir/test/Integration/Dialect/SparseTensor/GPU/CUDA/sparse-sddmm-lib.mlir b/mlir/test/Integration/Dialect/SparseTensor/GPU/CUDA/sparse-sddmm-lib.mlir
index db5c154e5e438d..bd2c72271c613a 100644
--- a/mlir/test/Integration/Dialect/SparseTensor/GPU/CUDA/sparse-sddmm-lib.mlir
+++ b/mlir/test/Integration/Dialect/SparseTensor/GPU/CUDA/sparse-sddmm-lib.mlir
@@ -166,15 +166,22 @@ module {
// the result (which is why the block sparse version has actual results
// in the original zero positions).
//
- // CHECK: ( 5, 10, 24, 19, 53, 42, 55, 56 )
- // CHECK-NEXT: ( 5, 10, 8, 19, 24, 24, 40, 53, 42, 55, 56, 64 )
+ // CHECK: ---- Sparse Tensor ----
+ // CHECK-NEXT: nse = 8
+ // CHECK-NEXT: pos[1] : ( 0, 3, 5, 7, 8,
+ // CHECK-NEXT: crd[1] : ( 0, 1, 4, 1, 5, 2, 3, 2,
+ // CHECK-NEXT: values : ( 5, 10, 24, 19, 53, 42, 55, 56,
+ // CHECK-NEXT: ----
//
- %v0 = sparse_tensor.values %0 : tensor<?x?xf32, #CSR> to memref<?xf32>
- %vv0 = vector.transfer_read %v0[%c0], %d0 : memref<?xf32>, vector<8xf32>
- vector.print %vv0 : vector<8xf32>
- %v1 = sparse_tensor.values %1 : tensor<?x?xf32, #BSR> to memref<?xf32>
- %vv1 = vector.transfer_read %v1[%c0], %d0 : memref<?xf32>, vector<12xf32>
- vector.print %vv1 : vector<12xf32>
+ // CHECK: ---- Sparse Tensor ----
+ // CHECK-NEXT: nse = 12
+ // CHECK-NEXT: pos[1] : ( 0, 2, 3,
+ // CHECK-NEXT: crd[1] : ( 0, 2, 1,
+ // CHECK-NEXT: values : ( 5, 10, 8, 19, 24, 24, 40, 53, 42, 55, 56, 64,
+ // CHECK-NEXT: ----
+ //
+ sparse_tensor.print %0 : tensor<?x?xf32, #CSR>
+ sparse_tensor.print %1 : tensor<?x?xf32, #BSR>
// Release the resources.
bufferization.dealloc_tensor %0 : tensor<?x?xf32, #CSR>
|
PeimingLiu
approved these changes
Mar 1, 2024
yinying-lisa-li
approved these changes
Mar 1, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Continuous efforts #83357 for our sparse CUDA tests