[NVPTX] Limit a sparsity selector in sparse MMA intrinsics. #154984
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
sparsity selector in sparse MMA intrinsics.
|
@llvm/pr-subscribers-backend-nvptx Author: Kirill Vedernikov (kvederni) ChangesThis PR fixes NVPTX tests in LLVM testing by adding more limitations for a sparsity selector in sparse MMA intrinsics. Full diff: https://github.com/llvm/llvm-project/pull/154984.diff 2 Files Affected:
diff --git a/llvm/include/llvm/IR/IntrinsicsNVVM.td b/llvm/include/llvm/IR/IntrinsicsNVVM.td
index cd7a0bc9c4b48..130fa27e4f870 100644
--- a/llvm/include/llvm/IR/IntrinsicsNVVM.td
+++ b/llvm/include/llvm/IR/IntrinsicsNVVM.td
@@ -2161,6 +2161,7 @@ class NVVM_MMA_SP<WMMA_REGS A, WMMA_REGS B, WMMA_REGS C, WMMA_REGS D>
// The range [0;num_threads) is for the sparsity selector that indicates the threads
// which contribute metadata.
int num_threads = !if(!or(!and(!eq(A.geom, "m16n8k32"), !eq(A.ptx_elt_type, "bf16")),
+ !and(!eq(A.geom, "m16n8k32"), !eq(A.ptx_elt_type, "f16")),
!and(!eq(A.geom, "m16n8k16"), !eq(A.ptx_elt_type, "tf32")),
!and(!eq(A.geom, "m16n8k32"), !eq(A.ptx_elt_type, "u8")),
!and(!eq(A.geom, "m16n8k32"), !eq(A.ptx_elt_type, "s8")),
@@ -2175,7 +2176,11 @@ class NVVM_MMA_SP<WMMA_REGS A, WMMA_REGS B, WMMA_REGS C, WMMA_REGS D>
!eq(A.ptx_elt_type, "e3m2"),
!eq(A.ptx_elt_type, "e2m3"),
!eq(A.ptx_elt_type, "e2m1"))),
- 1, 4));
+ 1,
+ !if(!and(!eq(A.geom, "m16n8k128"),
+ !or(!eq(A.ptx_elt_type, "s4"),
+ !eq(A.ptx_elt_type, "u4"))),
+ 1, 4)));
let IntrProperties = [IntrNoMem, IntrNoCallback, ImmArg<ArgIndex<pos>>,
Range<ArgIndex<pos>, 0, num_threads>];
}
diff --git a/llvm/test/CodeGen/NVPTX/wmma.py b/llvm/test/CodeGen/NVPTX/wmma.py
index f4f166c4018d0..6d73bce46da7c 100644
--- a/llvm/test/CodeGen/NVPTX/wmma.py
+++ b/llvm/test/CodeGen/NVPTX/wmma.py
@@ -1135,6 +1135,7 @@ def sp_selector_gen(op):
# (geom, type) -> allowed selector range
range_01 = {
("m16n8k32", "bf16"),
+ ("m16n8k32", "f16"),
("m16n8k16", "tf32"),
("m16n8k32", "u8"),
("m16n8k32", "s8"),
@@ -1154,6 +1155,11 @@ def sp_selector_gen(op):
"e2m1",
]:
return range(1)
+ if op.a.geom == "m16n8k128" and op.a.mma_type.ptx_type in [
+ "u4",
+ "s4",
+ ]:
+ return range(1)
return range(4)
|
|
@llvm/pr-subscribers-llvm-ir Author: Kirill Vedernikov (kvederni) ChangesThis PR fixes NVPTX tests in LLVM testing by adding more limitations for a sparsity selector in sparse MMA intrinsics. Full diff: https://github.com/llvm/llvm-project/pull/154984.diff 2 Files Affected:
diff --git a/llvm/include/llvm/IR/IntrinsicsNVVM.td b/llvm/include/llvm/IR/IntrinsicsNVVM.td
index cd7a0bc9c4b48..130fa27e4f870 100644
--- a/llvm/include/llvm/IR/IntrinsicsNVVM.td
+++ b/llvm/include/llvm/IR/IntrinsicsNVVM.td
@@ -2161,6 +2161,7 @@ class NVVM_MMA_SP<WMMA_REGS A, WMMA_REGS B, WMMA_REGS C, WMMA_REGS D>
// The range [0;num_threads) is for the sparsity selector that indicates the threads
// which contribute metadata.
int num_threads = !if(!or(!and(!eq(A.geom, "m16n8k32"), !eq(A.ptx_elt_type, "bf16")),
+ !and(!eq(A.geom, "m16n8k32"), !eq(A.ptx_elt_type, "f16")),
!and(!eq(A.geom, "m16n8k16"), !eq(A.ptx_elt_type, "tf32")),
!and(!eq(A.geom, "m16n8k32"), !eq(A.ptx_elt_type, "u8")),
!and(!eq(A.geom, "m16n8k32"), !eq(A.ptx_elt_type, "s8")),
@@ -2175,7 +2176,11 @@ class NVVM_MMA_SP<WMMA_REGS A, WMMA_REGS B, WMMA_REGS C, WMMA_REGS D>
!eq(A.ptx_elt_type, "e3m2"),
!eq(A.ptx_elt_type, "e2m3"),
!eq(A.ptx_elt_type, "e2m1"))),
- 1, 4));
+ 1,
+ !if(!and(!eq(A.geom, "m16n8k128"),
+ !or(!eq(A.ptx_elt_type, "s4"),
+ !eq(A.ptx_elt_type, "u4"))),
+ 1, 4)));
let IntrProperties = [IntrNoMem, IntrNoCallback, ImmArg<ArgIndex<pos>>,
Range<ArgIndex<pos>, 0, num_threads>];
}
diff --git a/llvm/test/CodeGen/NVPTX/wmma.py b/llvm/test/CodeGen/NVPTX/wmma.py
index f4f166c4018d0..6d73bce46da7c 100644
--- a/llvm/test/CodeGen/NVPTX/wmma.py
+++ b/llvm/test/CodeGen/NVPTX/wmma.py
@@ -1135,6 +1135,7 @@ def sp_selector_gen(op):
# (geom, type) -> allowed selector range
range_01 = {
("m16n8k32", "bf16"),
+ ("m16n8k32", "f16"),
("m16n8k16", "tf32"),
("m16n8k32", "u8"),
("m16n8k32", "s8"),
@@ -1154,6 +1155,11 @@ def sp_selector_gen(op):
"e2m1",
]:
return range(1)
+ if op.a.geom == "m16n8k128" and op.a.mma_type.ptx_type in [
+ "u4",
+ "s4",
+ ]:
+ return range(1)
return range(4)
|
durga4github
approved these changes
Aug 22, 2025
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.
This PR fixes NVPTX tests in LLVM testing by adding more limitations for a sparsity selector in sparse MMA intrinsics.
The previous PR that is merged to llvm:main is PR150950. The merge to llvm:main is d9c6b7b.