Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
13 changes: 7 additions & 6 deletions tests/kernel/wave_gemm_mxfp_test.py
Original file line number Diff line number Diff line change
Expand Up @@ -397,9 +397,9 @@ def testScaledBatchedGemmMXFP4Codegen(use_water_backend: bool, tmp_path: Path):
# We encode the exact registers and wait counts as we want to know if
# they suddenly change due to backend or upstream MLIR changes.
if use_water_backend:
vgpr_count = 160
vgpr_count = 150
vgpr_spill_count = 0
sgpr_count = 59
sgpr_count = 58
sgpr_spill_count = 0
waitcounts = [
"s_waitcnt lgkmcnt(0)",
Expand All @@ -413,10 +413,11 @@ def testScaledBatchedGemmMXFP4Codegen(use_water_backend: bool, tmp_path: Path):
"s_waitcnt lgkmcnt(1)",
"s_waitcnt vmcnt(0) lgkmcnt(0)",
"s_waitcnt vmcnt(0)",
"s_waitcnt lgkmcnt(7)",
"s_waitcnt lgkmcnt(6)",
"s_waitcnt lgkmcnt(5)",
"s_waitcnt lgkmcnt(4)",
"s_waitcnt lgkmcnt(10)",
"s_waitcnt lgkmcnt(11)",
"s_waitcnt lgkmcnt(2)",
"s_waitcnt lgkmcnt(1)",
"s_waitcnt lgkmcnt(2)",
"s_waitcnt lgkmcnt(3)",
"s_waitcnt lgkmcnt(1)",
"s_waitcnt lgkmcnt(0)",
Expand Down
1 change: 1 addition & 0 deletions water/tools/water-opt/water-opt.cpp
Original file line number Diff line number Diff line change
Expand Up @@ -48,6 +48,7 @@ void registerWaterTestDialect(DialectRegistry &registry);
} // namespace mlir::water::test

int main(int argc, char **argv) {
mlir::arith::registerArithIntRangeNarrowingPass();
mlir::arith::registerArithIntRangeOptsPass();
mlir::registerCSEPass();
mlir::registerCanonicalizerPass();
Expand Down
1 change: 1 addition & 0 deletions wave_lang/kernel/wave/water.py
Original file line number Diff line number Diff line change
Expand Up @@ -460,6 +460,7 @@ def add_transform(transform: str, entry_point: str) -> tuple[str, dict[str, Any]
*add_opt(canonicalize_cse),
*add_opt("loop-invariant-code-motion"),
*add_opt("int-range-optimizations"),
*add_opt(("arith-int-range-narrowing", {"int-bitwidths-supported": "8,16,32"})),
"convert-scf-to-cf",
("convert-amdgpu-to-rocdl", {"chipset": target_chip}),
("water-alloc-to-alloca", {}, "gpu.module"),
Expand Down
Loading