Some issues are related with memory fragmentation, which may cause
the linear memory cannot be allocated. In WAMR, the memory managed
by the system is often trivial, but linear memory usually directly allocates
a large block and often remains unchanged for a long time. Their sensitivity
and contribution to fragmentation are different, which is suitable for
different allocation strategies. If we can control the linear memory's allocation,
do not make it from system heap, the overhead of heap management might
be avoided.
Add `mem_alloc_usage_t usage` as the first argument for user defined
malloc/realloc/free functions when `WAMR_BUILD_ALLOC_WITH_USAGE` cmake
variable is set as 1, and make passing `Alloc_For_LinearMemory` to the
argument when allocating the linear memory.
Enhance the GC subtyping checks:
- Fix issues in the type equivalence check
- Enable the recursive type subtyping check
- Add a equivalence type flag in defined types of aot file, if there is an
equivalence type before, just set it true and re-use the previous type
- Normalize the defined types for interpreter and AOT
- Enable spec test case type-equivalence.wast and type-subtyping.wast,
and enable some commented cases
- Enable set WAMR_BUILD_SANITIZER from cmake variable
When thread manager is enabled, the aux stack of exec_env may be allocated
by wasm_cluster_allocate_aux_stack or disabled by setting aux_stack_bottom
as UINTPTR_MAX directly. For the latter, no need to free it.
And fix an issue when paring `--gc-heap-size=n` argument for iwasm, and
fix a variable shadowed warning in fast-jit.
- Add new API wasm_runtime_load_ex() in wasm_export.h
and wasm_module_new_ex in wasm_c_api.h
- Put aot_create_perf_map() into a separated file aot_perf_map.c
- In perf.map, function names include user specified module name
- Enhance the script to help flamegraph generations
Fix the warnings and issues reported:
- in Windows platform
- by CodeQL static code analyzing
- by Coverity static code analyzing
And update CodeQL script to build exception handling and memory features.
Adding a new cmake flag (cache variable) `WAMR_BUILD_MEMORY64` to enable
the memory64 feature, it can only be enabled on the 64-bit platform/target and
can only use software boundary check. And when it is enabled, it can support both
i32 and i64 linear memory types. The main modifications are:
- wasm loader & mini-loader: loading and bytecode validating process
- wasm runtime: memory instantiating process
- classic-interpreter: wasm code executing process
- Support memory64 memory in related runtime APIs
- Modify main function type check when it's memory64 wasm file
- Modify `wasm_runtime_invoke_native` and `wasm_runtime_invoke_native_raw` to
handle registered native function pointer argument when memory64 is enabled
- memory64 classic-interpreter spec test in `test_wamr.sh` and in CI
Currently, it supports memory64 memory wasm file that uses core spec
(including bulk memory proposal) opcodes and threads opcodes.
ps.
https://github.com/bytecodealliance/wasm-micro-runtime/issues/3091https://github.com/bytecodealliance/wasm-micro-runtime/pull/3240https://github.com/bytecodealliance/wasm-micro-runtime/pull/3260
Some environment may call wasm_runtime_full_init/wasm_runtime_init multiple
times without knowing that runtime is initialized or not, it is better to add lock
and increase reference count during initialization.
ps. https://github.com/bytecodealliance/wasm-micro-runtime/discussions/3253.
This allows to know the beginning of the wasm address space. At the moment
to achieve that, we need to apply a `hack wasm_runtime_addr_app_to_native(X)-X`
to get the beginning of WASM memory in the nativ code, but I don't see a good
reason why not to allow zero address as a parameter value for this function.
This PR adds a max_memory_pages parameter to module instantiation APIs,
to allow overriding the max memory defined in the WASM module.
Sticking to the max memory defined in the module is quite limiting when
using shared memory in production. If targeted devices have different
memory constraints, many wasm files have to be generated with different
max memory values. And device constraints may not be known in advance.
Being able to set the max memory value during module instantiation allows
to reuse the same wasm module, e.g. by retrying instantiation with different
max memory value.
The wasm_interp_call_func_bytecode is called for the first time with the empty
module/exec_env to generate a global_handle_table. Before that happens though,
the function checks if the module instance has bounds check enabled. Because
the module instance is null, the program crashes. This PR added an extra check to
prevent the crashes.
Implement the GC (Garbage Collection) feature for interpreter mode,
AOT mode and LLVM-JIT mode, and support most features of the latest
spec proposal, and also enable the stringref feature.
Use `cmake -DWAMR_BUILD_GC=1/0` to enable/disable the feature,
and `wamrc --enable-gc` to generate the AOT file with GC supported.
And update the AOT file version from 2 to 3 since there are many AOT
ABI breaks, including the changes of AOT file format, the changes of
AOT module/memory instance layouts, the AOT runtime APIs for the
AOT code to invoke and so on.
With this approach we can omit using memset() for the newly allocated memory
therefore the physical pages are not being used unless touched by the program.
This also simplifies the implementation.
Check whether the arguments are NULL before calling bh_hash_map_find,
or lots of "HashMap find elem failed: map or key is NULL" warnings may
be dumped. Reported in #3053.
It seems that some users want to wrap rather large chunk of code
with wasm_runtime_begin_blocking_op/wasm_runtime_end_blocking_op.
If the wrapped code happens to have a call to
e.g. wasm_runtime_spawn_exec_env, WASM_SUSPEND_FLAG_BLOCKING is
inherited to the child exec_env and it may cause unexpected behaviors.
- Enable quick aot entry when hw bound check is disabled
- Remove unnecessary ret_type argument in the quick aot entries
- Declare detailed prototype of aot function to call in each quick aot entry
Enhance the statistic of wasm function execution time, or the performance
profiling feature:
- Add os_time_thread_cputime_us() to get the cputime of a thread,
and use it to calculate the execution time of a wasm function
- Support the statistic of the children execution time of a function,
and dump it in wasm_runtime_dump_perf_profiling
- Expose two APIs:
wasm_runtime_sum_wasm_exec_time
wasm_runtime_get_wasm_func_exec_time
And rename os_time_get_boot_microsecond to os_time_get_boot_us.
For shared memory, the max memory size must be defined in advanced. Re-allocation
for growing memory can't be used as it might change the base address, therefore when
OS_ENABLE_HW_BOUND_CHECK is enabled the memory is mmaped, and if the flag is
disabled, the memory is allocated. This change introduces a flag that allows users to use
mmap for reserving memory address space even if the OS_ENABLE_HW_BOUND_CHECK
is disabled.
When using the wasm-c-api and there's a trap, `wasm_func_call()` returns
a `wasm_trap_t *` object. No matter which thread crashes, the trap contains
the stack frames of the main thread.
With this PR, when there's an exception, the stack frames of the thread
where the exception occurs are stored into the thread cluster.
`wasm_func_call()` can then return those stack frames.
Allow to invoke the quick call entry wasm_runtime_quick_invoke_c_api_import to
call the wasm-c-api import functions to speedup the calling process, which reduces
the data copying.
Use `wamrc --invoke-c-api-import` to generate the optimized AOT code, and set
`jit_options->quick_invoke_c_api_import` true in wasm_engine_new when LLVM JIT
is enabled.
In some scenarios there may be lots of callings to AOT/JIT functions from the
host embedder, which expects good performance for the calling process, while
in the current implementation, runtime calls the wasm_runtime_invoke_native
to prepare the array of registers and stacks for the invokeNative assemble code,
and the latter then puts the elements in the array to physical registers and
native stacks and calls the AOT/JIT function, there may be many data copying
and handlings which impact the performance.
This PR registers some quick AOT/JIT entries for some simple wasm signatures,
and let runtime call the entry to directly invoke the AOT/JIT function instead of
calling wasm_runtime_invoke_native, which speedups the calling process.
We may extend the mechanism next to allow the developer to register his quick
AOT/JIT entries to speedup the calling process of invoking the AOT/JIT functions
for some specific signatures.
Add an API to set segue flags for wasm-c-api LLVM JIT mode:
```C
wasm_config_t *
wasm_config_set_segue_flags(wasm_config_t *config, uint32 segue_flags);
```
And refactor the original perf support
- use WAMR_BUILD_LINUX_PERF as the cmake compilation control
- use WASM_ENABLE_LINUX_PERF as the compiler macro
- use `wamrc --enable-linux-perf` to generate aot file which contains fp operations
- use `iwasm --enable-linux-perf` to create perf map for `perf record`
The host embedder may also want to terminate the wasm instance
for single-threading mode, and it should work by setting exception
to the wasm instance.
Support new a wasm_config_t, set allocation and linux_perf_support
options to it, and then pass it to wasm_engine_new_with_config to
new an engine with private configuration.
Add an extra argument `os_file_handle file` for `os_mmap` to support
mapping file from a file fd, and remove `os_get_invalid_handle` from
`posix_file.c` and `win_file.c`, instead, add it in the `platform_internal.h`
files to remove the dependency on libc-wasi.
Signed-off-by: Huang Qi <huangqi3@xiaomi.com>
To run it locally:
```bash
export TSAN_OPTIONS=suppressions=<path_to_tsan_suppressions.txt>
./test_wamr.sh <your flags> -T tsan
```
An example for wasi-threads would look like:
```bash
export TSAN_OPTIONS=suppressions=<path_to_tsan_suppressions.txt>
./test_wamr.sh -w -s wasi_certification -t fast-interp -T tsan
```
Split memory instance's field `uint32 ref_count` into `bool is_shared_memory`
and `uint16 ref_count`, and lock the memory only when `is_shared_memory`
flag is true, no need to acquire a lock for non-shared memory when shared
memory feature is enabled.
