The problem is not limited to arm64 and it is not limited to 64-bit values. Attached is a simple change that makes the 32-bit multiply on arm64 and arm use the same codegen strategy as the arm64 64-bit multiply (no constant folding + twice useRegisterAtStart + defineReuseInput). My reduced test case, converted to 32-bit in the obvious way, triggers the assert on both arm64 and arm. (On x86 and x64 multiply is weird because it uses fixed registers, so it's not straightforward to test there.)
Bug 1707774 Comment 21 Edit History
Note: The actual edited comment in the bug view page will always show the original commenter’s name and original timestamp.
The problem is not limited to arm64 and it is not limited to 64-bit values. Attached is a simple change that makes the 32-bit multiply on arm64 and arm use the same codegen strategy as the arm64 64-bit multiply (no constant folding + twice useRegisterAtStart + defineReuseInput). My reduced test case from comment 11, converted to 32-bit in the obvious way, triggers the assert on both arm64 and arm. (On x86 and x64 multiply is weird because it uses fixed registers, so it's not straightforward to test there.)