Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

IREE Bump to Oct 1, 2024 #788

Merged
merged 4 commits into from
Oct 1, 2024
Merged

IREE Bump to Oct 1, 2024 #788

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
155a5af
IREE Bump to 18th Sep, 2024
Abhishek-Varma Sep 19, 2024
72ec3cd
Test fix
Abhishek-Varma Sep 19, 2024
f254e21
Update xrt_device.cc
makslevental Sep 27, 2024
925d4b0
fix
makslevental Oct 1, 2024
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
2 changes: 2 additions & 0 deletions build_tools/ci/cpu_comparison/run.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -142,6 +142,8 @@ def generate_aie_vmfb(
f"--iree-amd-aie-vitis-install-dir={config.vitis_dir}",
f"--iree-hal-dump-executable-files-to={config.output_dir}",
"--iree-scheduling-optimize-bindings=false",
"--iree-hal-memoization=false",
"--iree-hal-indirect-command-buffers=false",
f"--mlir-disable-threading",
"--mlir-elide-resource-strings-if-larger=10",
]
Expand Down
4 changes: 4 additions & 0 deletions build_tools/ci/run_matmul_test.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -405,6 +405,8 @@ function run_matmul_test() {
--iree-amd-aie-enable-chess=${use_chess} \
--iree-amdaie-enable-packet-flow=${enable_packet_flow} \
--iree-hal-dump-executable-files-to=$PWD \
--iree-hal-memoization=false \
--iree-hal-indirect-command-buffers=false \
--mlir-elide-resource-strings-if-larger=10 \
--iree-amd-aie-show-invoked-commands"

Expand All @@ -416,6 +418,8 @@ function run_matmul_test() {
set +e

echo "**** Generating matmul .vmfb file for ${name} ****"
${IREE_COMPILE_EXE} "${matmul_ir}" \
${compilation_flags} --compile-to=vm -o "${matmul_vmfb}.vm"
${IREE_COMPILE_EXE} "${matmul_ir}" \
${compilation_flags} -o "${matmul_vmfb}"

Expand Down
7 changes: 3 additions & 4 deletions compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/AMDAIEConvertCoreForallToFor.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -27,8 +27,7 @@ namespace {

/// Converts `scf.forall` into nested `scf.for` and then coalesce the `scf.for`
/// loops.
LogicalResult coreForallToFor(RewriterBase &rewriter,
AMDAIE::CoreOp coreOp) {
LogicalResult coreForallToFor(RewriterBase &rewriter, AMDAIE::CoreOp coreOp) {
WalkResult res = coreOp->walk([&](scf::ForallOp forallOp) {
SmallVector<Operation *> forOpResults;
if (failed(scf::forallToForLoop(rewriter, forallOp, &forOpResults))) {
Expand All @@ -55,12 +54,12 @@ class AMDAIEConvertCoreForallToForPass
AMDAIEConvertCoreForallToForPass> {
public:
void getDependentDialects(DialectRegistry &registry) const override {
registry.insert<scf::SCFDialect>();
registry.insert<scf::SCFDialect, affine::AffineDialect>();
}

AMDAIEConvertCoreForallToForPass() = default;
AMDAIEConvertCoreForallToForPass(
const AMDAIEConvertCoreForallToForPass &pass){};
const AMDAIEConvertCoreForallToForPass &pass) {};
void runOnOperation() override;
};

Expand Down
11 changes: 7 additions & 4 deletions compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/AMDAIETileAndFuse.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -343,13 +343,15 @@ void AMDAIETileAndFusePass::runOnOperation() {
if (isTilingReductionDimension(consumerOp, tileSizesVal)) {
tileAndFuseOptions.setFusionControlFn(
[&](tensor::ExtractSliceOp sliceOp, OpResult originalProducer,
bool isDestinationOperand) -> std::tuple<bool, bool> {
return {false, false};
bool isDestinationOperand)
-> std::optional<scf::SCFTileAndFuseOptions::ControlFnResult> {
return std::nullopt;
});
} else {
tileAndFuseOptions.setFusionControlFn(
[&](tensor::ExtractSliceOp sliceOp, OpResult originalProducer,
bool isDestinationOperand) -> std::tuple<bool, bool> {
bool isDestinationOperand)
-> std::optional<scf::SCFTileAndFuseOptions::ControlFnResult> {
bool fusableOp =
TypeSwitch<Operation *, bool>(originalProducer.getOwner())
// List ops that shouldnt be fused.
Expand All @@ -360,7 +362,8 @@ void AMDAIETileAndFusePass::runOnOperation() {
return op->getDialect() ==
context->getLoadedDialect<linalg::LinalgDialect>();
});
return {fusableOp, false};
if (!fusableOp) return std::nullopt;
return scf::SCFTileAndFuseOptions::ControlFnResult{false};
});
}

Expand Down
51 changes: 23 additions & 28 deletions compiler/plugins/target/AMD-AIE/iree-amd-aie/Transforms/test/convert_core_forall_to_for.mlir
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,15 +1,14 @@
// RUN: iree-opt --pass-pipeline="builtin.module(iree-amdaie-convert-core-forall-to-for,canonicalize)" --split-input-file %s | FileCheck %s

// CHECK-LABEL: @test_single
// CHECK-DAG: %[[C0:.+]] = arith.constant 0 : index
// CHECK-DAG: %[[C1:.+]] = arith.constant 1 : index
// CHECK-DAG: %[[C2:.+]] = arith.constant 2 : index
// CHECK-DAG: %[[C4:.+]] = arith.constant 4 : index
// CHECK-DAG: amdaie.core
// CHECK-DAG: %[[C0:.+]] = arith.constant 0 : index
// CHECK-DAG: %[[C1:.+]] = arith.constant 1 : index
// CHECK-DAG: %[[C2:.+]] = arith.constant 2 : index
// CHECK-DAG: %[[C4:.+]] = arith.constant 4 : index
// CHECK: scf.for %[[ARG0:.+]] = %[[C0]] to %[[C4]] step %[[C1]] {
// CHECK-DAG: %[[REM:.+]] = arith.remsi %[[ARG0]], %[[C2]] : index
// CHECK-DAG: %[[DIV:.+]] = arith.divsi %[[ARG0]], %[[C2]] : index
// CHECK-DAG: func.call @callee(%[[DIV]], %[[REM]]) : (index, index) -> ()
// CHECK-DAG: %[[D:.+]]:2 = affine.delinearize_index %[[ARG0]] into (%[[C2]], %[[C2]]) : index, index
// CHECK-DAG: func.call @callee(%[[D]]#0, %[[D]]#1) : (index, index) -> ()
module @test_single {
func.func private @callee(%i: index, %j: index)
%c0 = arith.constant 0 : index
Expand All @@ -28,20 +27,18 @@ module @test_single {
// -----

// CHECK-LABEL: @test_multi
// CHECK-DAG: %[[C0:.+]] = arith.constant 0 : index
// CHECK-DAG: %[[C1:.+]] = arith.constant 1 : index
// CHECK-DAG: %[[C2:.+]] = arith.constant 2 : index
// CHECK-DAG: %[[C4:.+]] = arith.constant 4 : index
// CHECK-DAG: %[[C16:.+]] = arith.constant 16 : index
// CHECK-DAG: amdaie.core
// CHECK-DAG: %[[C0:.+]] = arith.constant 0 : index
// CHECK-DAG: %[[C1:.+]] = arith.constant 1 : index
// CHECK-DAG: %[[C2:.+]] = arith.constant 2 : index
// CHECK-DAG: %[[C4:.+]] = arith.constant 4 : index
// CHECK-DAG: %[[C16:.+]] = arith.constant 16 : index
// CHECK: scf.for %[[ARG0:.+]] = %[[C0]] to %[[C4]] step %[[C1]] {
// CHECK-DAG: %[[REM:.+]] = arith.remsi %[[ARG0]], %[[C2]] : index
// CHECK-DAG: %[[DIV:.+]] = arith.divsi %[[ARG0]], %[[C2]] : index
// CHECK-DAG: func.call @callee(%[[DIV]], %[[REM]]) : (index, index) -> ()
// CHECK-DAG: %[[D:.+]]:2 = affine.delinearize_index %[[ARG0]] into (%[[C2]], %[[C2]]) : index, index
// CHECK-DAG: func.call @callee(%[[D]]#0, %[[D]]#1) : (index, index) -> ()
// CHECK-DAG: scf.for %[[ARG1:.+]] = %[[C0]] to %[[C16]] step %[[C1]] {
// CHECK-DAG: %[[REM1:.+]] = arith.remsi %[[ARG1]], %[[C4]] : index
// CHECK-DAG: %[[DIV1:.+]] = arith.divsi %[[ARG1]], %[[C4]] : index
// CHECK-DAG: func.call @callee(%[[DIV1]], %[[REM1]]) : (index, index) -> ()
// CHECK-DAG: %[[D:.+]]:2 = affine.delinearize_index %[[ARG0]] into (%[[C4]], %[[C4]]) : index, index
// CHECK-DAG: func.call @callee(%[[D]]#0, %[[D]]#1) : (index, index) -> ()
module @test_multi {
func.func private @callee(%i: index, %j: index)
%c0 = arith.constant 0 : index
Expand All @@ -63,19 +60,17 @@ module @test_multi {
// -----

// CHECK-LABEL: @test_nested
// CHECK-DAG: %[[C0:.+]] = arith.constant 0 : index
// CHECK-DAG: %[[C1:.+]] = arith.constant 1 : index
// CHECK-DAG: %[[C2:.+]] = arith.constant 2 : index
// CHECK-DAG: %[[C4:.+]] = arith.constant 4 : index
// CHECK-DAG: %[[C16:.+]] = arith.constant 16 : index
// CHECK-DAG: amdaie.core
// CHECK-DAG: %[[C0:.+]] = arith.constant 0 : index
// CHECK-DAG: %[[C1:.+]] = arith.constant 1 : index
// CHECK-DAG: %[[C2:.+]] = arith.constant 2 : index
// CHECK-DAG: %[[C4:.+]] = arith.constant 4 : index
// CHECK-DAG: %[[C16:.+]] = arith.constant 16 : index
// CHECK: scf.for %[[ARG0:.+]] = %[[C0]] to %[[C16]] step %[[C1]] {
// CHECK-DAG: %[[REM0:.+]] = arith.remsi %[[ARG0]], %[[C4]] : index
// CHECK-DAG: %[[DIV0:.+]] = arith.divsi %[[ARG0]], %[[C4]] : index
// CHECK-DAG: %[[D1:.+]]:2 = affine.delinearize_index %[[ARG0]] into (%[[C4]], %[[C4]]) : index, index
// CHECK-DAG: scf.for %[[ARG1:.+]] = %[[C0]] to %[[C4]] step %[[C1]] {
// CHECK-DAG: %[[REM1:.+]] = arith.remsi %[[ARG1]], %[[C2]] : index
// CHECK-DAG: %[[DIV1:.+]] = arith.divsi %[[ARG1]], %[[C2]] : index
// CHECK-DAG: func.call @callee(%[[DIV0]], %[[REM0]], %[[DIV1]], %[[REM1]]) : (index, index, index, index) -> ()
// CHECK-DAG: %[[D2:.+]]:2 = affine.delinearize_index %[[ARG1]] into (%[[C2]], %[[C2]]) : index, index
// CHECK-DAG: func.call @callee(%[[D1]]#0, %[[D1]]#1, %[[D2]]#0, %[[D2]]#1) : (index, index, index, index) -> ()
module @test_nested {
func.func private @callee(%i: index, %j: index, %k: index, %l: index)
%c0 = arith.constant 0 : index
Expand Down
8 changes: 4 additions & 4 deletions .../plugins/target/AMD-AIE/iree-amd-aie/Transforms/test/fuse_consumer_into_loop_scf_for.mlir
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -62,12 +62,12 @@ module {
// CHECK: %[[SECOND_LOOP:.*]]:2 = scf.for %[[IV0:.*]] = %{{.*}} to %{{.*}} step %{{.*}} iter_args(%[[ITER_ARG_1:.*]] = %[[FIRST_LOOP]], %[[ITER_ARG_3:.*]] = %[[UNPACK_OUT]])
// CHECK: {
// CHECK: %[[MATMUL:.*]] = linalg.generic
// CHECK: affine.apply
// CHECK: affine.apply
// CHECK: %[[iv0:.*]] = affine.apply #[[UNPACK_RESULT_MAP0]](%[[IV0]])
// CHECK: %[[iv1:.*]] = affine.apply #[[UNPACK_RESULT_MAP1]](%[[IV0]])
// CHECK: %[[TILED_UNPACK_DEST:.*]] = tensor.extract_slice %[[ITER_ARG_3]][0, 0, %[[iv0]], %[[iv1]]] [1, 1, 32, 32] [1, 1, 1, 1]
// CHECK: %[[TILED_UNPACK:.*]] = tensor.unpack %[[MATMUL]] outer_dims_perm = [0, 1, 3, 2] inner_dims_pos = [2, 3] inner_tiles = [4, 8] into %[[TILED_UNPACK_DEST]]
// CHECK: %[[iv0:.*]] = affine.apply #[[UNPACK_RESULT_MAP0]](%[[IV0]])
// CHECK: %[[iv1:.*]] = affine.apply #[[UNPACK_RESULT_MAP1]](%[[IV0]])
// CHECK: %[[YIELD_MATMUL:.*]] = tensor.insert_slice %[[MATMUL]] into %[[ITER_ARG_1]]
// CHECK: %[[YIELD_UNPACK:.*]] = tensor.insert_slice %[[TILED_UNPACK]] into %[[ITER_ARG_3]]
// CHECK: scf.yield %[[YIELD_MATMUL]], %[[YIELD_UNPACK]]
Expand Down Expand Up @@ -159,12 +159,12 @@ module {
// CHECK: arith.addi
// CHECK: }
// CHECK: %[[YIELD_MATMUL:.*]] = tensor.insert_slice %[[MATMUL]] into %[[ITER_ARG_1]]
// CHECK: affine.apply
// CHECK: affine.apply
// CHECK: %[[iv0:.*]] = affine.apply #[[UNPACK_RESULT_MAP0]](%[[IV0]])
// CHECK: %[[iv1:.*]] = affine.apply #[[UNPACK_RESULT_MAP1]](%[[IV0]])
// CHECK: %[[TILED_UNPACK_DEST:.*]] = tensor.extract_slice %[[ITER_ARG_3]][0, 0, %[[iv0]], %[[iv1]]] [1, 1, 32, 32] [1, 1, 1, 1]
// CHECK: %[[TILED_UNPACK:.*]] = tensor.unpack %[[FUSED_CONSUMER]] outer_dims_perm = [0, 1, 3, 2] inner_dims_pos = [2, 3] inner_tiles = [4, 8] into %[[TILED_UNPACK_DEST]]
// CHECK: %[[iv0:.*]] = affine.apply #[[UNPACK_RESULT_MAP0]](%[[IV0]])
// CHECK: %[[iv1:.*]] = affine.apply #[[UNPACK_RESULT_MAP1]](%[[IV0]])
// CHECK: %[[YIELD_ELEM:.*]] = tensor.insert_slice %[[FUSED_CONSUMER]] into %[[ITER_ARG_2]]
// CHECK: %[[YIELD_UNPACK:.*]] = tensor.insert_slice %[[TILED_UNPACK]] into %[[ITER_ARG_3]]
// CHECK: scf.yield %[[YIELD_MATMUL]], %[[YIELD_ELEM]], %[[YIELD_UNPACK]]
Expand Down
14 changes: 7 additions & 7 deletions ...ugins/target/AMD-AIE/iree-amd-aie/Transforms/test/fuse_consumer_into_loop_scf_forall.mlir
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -59,15 +59,15 @@ module {
// CHECK: %[[SECOND_LOOP:.*]]:2 = scf.forall (%[[IV0:.*]], %[[IV1:.*]]) in (2, 2) shared_outs(%[[ITER_ARG_1:.*]] = %[[FIRST_LOOP]], %[[ITER_ARG_3:.*]] = %[[UNPACK_OUT]])
// CHECK: {
// CHECK: %[[MATMUL:.*]] = linalg.generic
// CHECK: affine.apply
// CHECK: affine.apply
// CHECK: %[[iv0:.*]] = affine.apply #[[UNPACK_RESULT_MAP0]](%[[IV0]])
// CHECK: %[[iv1:.*]] = affine.apply #[[UNPACK_RESULT_MAP1]](%[[IV1]])
// CHECK: %[[TILED_UNPACK_DEST:.*]] = tensor.extract_slice %[[ITER_ARG_3]][0, 0, %[[iv0]], %[[iv1]]] [1, 1, 32, 32] [1, 1, 1, 1]
// CHECK: %[[TILED_UNPACK:.*]] = tensor.unpack %[[MATMUL]] outer_dims_perm = [0, 1, 3, 2] inner_dims_pos = [2, 3] inner_tiles = [4, 8] into %[[TILED_UNPACK_DEST]]
// CHECK: %[[iv0:.*]] = affine.apply #[[UNPACK_RESULT_MAP0]](%[[IV0]])
// CHECK: %[[iv1:.*]] = affine.apply #[[UNPACK_RESULT_MAP1]](%[[IV1]])
// CHECK: scf.forall.in_parallel {
// CHECK: tensor.parallel_insert_slice %[[TILED_UNPACK]] into %[[ITER_ARG_3]][0, 0, %[[iv0]], %[[iv1]]] [1, 1, 32, 32] [1, 1, 1, 1]
// CHECK: tensor.parallel_insert_slice %[[MATMUL]] into %[[ITER_ARG_1]][0, 0, %[[IV1]], %[[IV0]], 0, 0] [1, 1, 4, 8, 4, 8] [1, 1, 1, 1, 1, 1]
// CHECK: tensor.parallel_insert_slice %[[TILED_UNPACK]] into %[[ITER_ARG_3]][0, 0, %[[iv0]], %[[iv1]]] [1, 1, 32, 32] [1, 1, 1, 1]
// CHECK: }
// CHECK: }
// CHECK: %[[SECOND_UNPACK:.*]] = tensor.unpack %[[SECOND_LOOP]]#1 inner_dims_pos = [0, 1] inner_tiles = [64, 64] into %[[SECOND_UNPACK_OUT]] :
Expand Down Expand Up @@ -153,16 +153,16 @@ module {
// CHECK: {
// CHECK: arith.addi
// CHECK: }
// CHECK: affine.apply
// CHECK: affine.apply
// CHECK: %[[iv0:.*]] = affine.apply #[[UNPACK_RESULT_MAP0]](%[[IV0]])
// CHECK: %[[iv1:.*]] = affine.apply #[[UNPACK_RESULT_MAP1]](%[[IV1]])
// CHECK: %[[TILED_UNPACK_DEST:.*]] = tensor.extract_slice %[[ITER_ARG_3]][0, 0, %[[iv0]], %[[iv1]]] [1, 1, 32, 32] [1, 1, 1, 1]
// CHECK: %[[TILED_UNPACK:.*]] = tensor.unpack %[[FUSED_CONSUMER]] outer_dims_perm = [0, 1, 3, 2] inner_dims_pos = [2, 3] inner_tiles = [4, 8] into %[[TILED_UNPACK_DEST]]
// CHECK: %[[iv0:.*]] = affine.apply #[[UNPACK_RESULT_MAP0]](%[[IV0]])
// CHECK: %[[iv1:.*]] = affine.apply #[[UNPACK_RESULT_MAP1]](%[[IV1]])
// CHECK: scf.forall.in_parallel {
// CHECK: tensor.parallel_insert_slice %[[TILED_UNPACK]] into %[[ITER_ARG_3]][0, 0, %[[iv0]], %[[iv1]]] [1, 1, 32, 32] [1, 1, 1, 1]
// CHECK: tensor.parallel_insert_slice %[[FUSED_CONSUMER]] into %[[ITER_ARG_2]][0, 0, %[[IV1]], %[[IV0]], 0, 0] [1, 1, 4, 8, 4, 8] [1, 1, 1, 1, 1, 1]
// CHECK: tensor.parallel_insert_slice %[[MATMUL]] into %[[ITER_ARG_1]][0, 0, %[[IV1]], %[[IV0]], 0, 0] [1, 1, 4, 8, 4, 8] [1, 1, 1, 1, 1, 1]
// CHECK: tensor.parallel_insert_slice %[[FUSED_CONSUMER]] into %[[ITER_ARG_2]][0, 0, %[[IV1]], %[[IV0]], 0, 0] [1, 1, 4, 8, 4, 8] [1, 1, 1, 1, 1, 1]
// CHECK: tensor.parallel_insert_slice %[[TILED_UNPACK]] into %[[ITER_ARG_3]][0, 0, %[[iv0]], %[[iv1]]] [1, 1, 32, 32] [1, 1, 1, 1]
// CHECK: }
// CHECK: }
// CHECK: %[[SECOND_UNPACK:.*]] = tensor.unpack %[[SECOND_LOOP]]#2 inner_dims_pos = [0, 1] inner_tiles = [64, 64] into %[[SECOND_UNPACK_OUT]] :
Expand Down
5 changes: 3 additions & 2 deletions runtime/src/iree-amd-aie/driver/xrt/direct_command_buffer.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -340,8 +340,9 @@ static iree_status_t iree_hal_xrt_direct_command_buffer_dispatch(
std::vector<xrt::bo> bos;
// TODO(max): do we need multiple descriptor sets ever for AIE?
uint32_t set = 0;
iree_hal_xrt_direct_command_buffer_push_descriptor_set(
base_command_buffer, set, bindings.count, bindings.values);
IREE_RETURN_AND_END_ZONE_IF_ERROR(
z0, iree_hal_xrt_direct_command_buffer_push_descriptor_set(
base_command_buffer, set, bindings.count, bindings.values));
for (iree_host_size_t j = 0; j < bindings.count; ++j) {
xrt::bo arg_buffer =
xrt::bo(*command_buffer->descriptor_sets[set].bindings[j],
Expand Down
12 changes: 10 additions & 2 deletions tests/conftest.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,7 +7,7 @@

from iree.compiler import ir
from iree.compiler._mlir_libs import get_dialect_registry
from iree.compiler.api import Session, Output, Source
from iree.compiler.api import Session, Output, Source, _initializeGlobalCL
from iree.compiler.extras import types as T
from iree.runtime import VmModule
from iree.runtime import get_driver, Config, SystemContext
Expand Down Expand Up @@ -49,8 +49,16 @@ def pytest_addoption(parser):
parser.addoption("--iree-aie-debug", action="store_true")


@pytest.fixture(scope="session")
def global_cl_args(request):
_initializeGlobalCL(
"--iree-hal-memoization=false",
"--iree-hal-indirect-command-buffers=false",
)


@pytest.fixture
def iree_session(request, pytestconfig) -> Session:
def iree_session(request, pytestconfig, global_cl_args) -> Session:
s = Session()
s.context.append_dialect_registry(get_dialect_registry())
s.context.load_all_available_dialects()
Expand Down
7 changes: 2 additions & 5 deletions tests/test_matmul.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -147,8 +147,5 @@ def test_matmul(
arg0 = np.random.randint(-1, 3, (M, K), dtype=lhs_rhs_type)
arg1 = np.random.randint(-1, 3, (K, N), dtype=lhs_rhs_type)
with invokable_module(session, module, device) as module:
for i in range(num_repeat_runs):
results = module[matmul_name](arg0, arg1).to_host()
assert np.array_equal(
results, (arg0.astype(acc_type) @ arg1.astype(acc_type))
)
results = module[matmul_name](arg0, arg1).to_host()
assert np.array_equal(results, (arg0.astype(acc_type) @ arg1.astype(acc_type)))
2 changes: 1 addition & 1 deletion third_party/iree
Open in desktop
Submodule iree updated 344 files
Loading