mirror of
https://github.com/nod-ai/AMD-SHARK-Studio.git
synced 2026-04-03 03:00:17 -04:00
Enable --device_allocator=caching
This commit is contained in:
@@ -117,6 +117,7 @@ def compile_through_fx(
|
||||
def set_iree_runtime_flags():
|
||||
vulkan_runtime_flags = [
|
||||
f"--vulkan_large_heap_block_size={args.vulkan_large_heap_block_size}",
|
||||
f"--device_allocator=caching",
|
||||
f"--vulkan_validation_layers={'true' if args.vulkan_validation_layers else 'false'}",
|
||||
]
|
||||
if args.enable_rgp:
|
||||
|
||||
@@ -70,6 +70,7 @@ def get_iree_common_args():
|
||||
return [
|
||||
"--iree-stream-resource-index-bits=64",
|
||||
"--iree-vm-target-index-bits=64",
|
||||
"--iree-vm-bytecode-module-strip-source-map=true",
|
||||
"--iree-util-zero-fill-elided-attrs",
|
||||
]
|
||||
|
||||
|
||||
@@ -22,7 +22,7 @@ from shark.parser import shark_args
|
||||
# Get the default gpu args given the architecture.
|
||||
def get_iree_gpu_args():
|
||||
ireert.flags.FUNCTION_INPUT_VALIDATION = False
|
||||
ireert.flags.parse_flags("--cuda_allow_inline_execution")
|
||||
ireert.flags.parse_flags("--cuda_allow_inline_execution", "--device_allocator=caching")
|
||||
# TODO: Give the user_interface to pass the sm_arch.
|
||||
sm_arch = get_cuda_sm_cc()
|
||||
if (
|
||||
|
||||
@@ -139,9 +139,8 @@ def get_vulkan_triple_flag(device_name="", extra_args=[]):
|
||||
|
||||
|
||||
def get_iree_vulkan_args(extra_args=[]):
|
||||
# vulkan_flag = ["--iree-flow-demote-i64-to-i32"]
|
||||
res_vulkan_flag = ["--device_allocator=caching"]
|
||||
|
||||
res_vulkan_flag = []
|
||||
vulkan_triple_flag = None
|
||||
for arg in extra_args:
|
||||
if "-iree-vulkan-target-triple=" in arg:
|
||||
|
||||
Reference in New Issue
Block a user