Fix field types in InlineeCallInfo. OS #15566165 #5490
Conversation
rajatd
force-pushed
the
inlineeCallInfo-type
branch
2 times, most recently
from
78b1840 to
8fa0e00
Compare
lib/Backend/InlineeFrameInfo.cpp
Outdated
| { | ||
| this->PopulateParent(inlinee); | ||
| #if TARGET_32 | ||
| const size_t offsetMask = (~(size_t)0) >> 4; |
There was a problem hiding this comment.
nit: Why not use uint32 here? Why hardcode 4 instead of (32 - ksizeofInlineeStartOffset) or (sizeof(uint) * CHAR_BIT - ksizeofInlineeStartOffset)?
There was a problem hiding this comment.
Yep, I should use uint32. Regarding the shift, I only want to make sure we are within the 28-bit limit on x86. currentOffset is uint32, so there's no point checking for 32 bit limit
In reply to: 204944353 [](ancestors = 204944353)
There was a problem hiding this comment.
Actually, regardless of which platform I want to do this check on, hardcoding 4 is probably not the best idea. I'll change that. Thanks.
lib/Backend/Encoder.cpp
Outdated
| uint32 Encoder::GetCurrentOffset() const | ||
| { | ||
| Assert(m_pc - m_encodeBuffer <= UINT_MAX); // encode buffer size is uint32 | ||
| AssertOrFailFast(m_pc - m_encodeBuffer <= UINT_MAX); // encode buffer size is uint32 |
There was a problem hiding this comment.
AssertOrFailFast [](start = 4, length = 16)
m_encodeBufferSize is a UINT32, so this failfast is redundant. Given that it can get called multiple times per instr, let's leave it as an assert.
rajatd
force-pushed
the
inlineeCallInfo-type
branch
from
8fa0e00 to
a1fc0ae
Compare
…o. OS #15566165 Merge pull request #5490 from rajatd:inlineeCallInfo-type
5 participants
No description provided.