[TIR][Arith] Add more strict checking in imm construction and folding. #12515
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
wrongtest-intellif
force-pushed
the
make_more_strict_imm_check
branch
from
44624e5 to
0375ec5
Compare
wrongtest-intellif
force-pushed
the
make_more_strict_imm_check
branch
from
0375ec5 to
4240d54
Compare
ganler
reviewed
Aug 20, 2022
ganler
reviewed
Aug 20, 2022
include/tvm/tir/op.h
Outdated
| @@ -911,7 +911,9 @@ inline PrimExpr MakeConstScalar(DataType t, ValueType value, Span span = Span()) | |||
| if (t.is_uint()) { | |||
| // Use IntImm if it is a small integer | |||
| uint64_t uval = static_cast<uint64_t>(value); | |||
| if (uval <= static_cast<uint64_t>(std::numeric_limits<int64_t>::max())) { | |||
| if (static_cast<int64_t>(value) < 0) { | |||
There was a problem hiding this comment.
Maybe here we should do value < static_cast<ValueType>(0). Because if value is of uint64_t, it will be regarded as a negative number if it is greater than numeric_limits<int64_t>::max().
wrongtest-intellif
force-pushed
the
make_more_strict_imm_check
branch
from
f1719af to
5e5d84d
Compare
wrongtest-intellif
force-pushed
the
make_more_strict_imm_check
branch
from
74aaa1c to
9f85312
Compare
wrongtest-intellif
force-pushed
the
make_more_strict_imm_check
branch
2 times, most recently
from
22db840 to
13aa737
Compare
wrongtest-intellif
force-pushed
the
make_more_strict_imm_check
branch
from
13aa737 to
f900f81
Compare
|
cc @ganler Hi, could you kindly take another look at the change~? |
ganler
approved these changes
Aug 29, 2022
junrushao
approved these changes
Aug 31, 2022
There was a problem hiding this comment.
Thanks for the tremendous amount of work improving IntImm semantics! This is indeed super helpful and preventative of potential flaws at all levels of the system :-)
On the other hand, now that we have stronger enforcement of more formal semantics, it's entirely possible that some models might stop working due to our fix to this UB.
Therefore, I would suggest that we leave this PR open for 3 days, and if there is no objection, please feel free to merge it in yourself @wrongtest-intellif!
junrushao
pushed a commit
that referenced
this pull request
Sep 15, 2022
This change tries to fix an issue due to #12515. Previously the logic for `-2147483648` is `parse(-literal)` = `-parse(literal)`, and all integer literals are converted to i32 (either the literal value actually overflow or not). Since after #12515, parse `2147483648` results in an i64 typed integer rather than i32, `-2147483648` then becomes an i64 integer too, which is not reasonable.
xinetzone
pushed a commit
to daobook/tvm
that referenced
this pull request
Nov 25, 2022
apache#12515) * Add more strict check in tir imm construction and folding. * fix bool-compare compile error * fix some illegal imm construction in testcases * do not test i64 overflow behaviour because it is not consistent on cython and ctypes * fix float32 testcase * auto-inferred dtype should be int64 when value exceeds int32 range * add floatimm range check for fp16 and fp32 * add more folding testcases and fix store fp32 folding result to double * fix i386 fp16 cases
xinetzone
pushed a commit
to daobook/tvm
that referenced
this pull request
Nov 25, 2022
This change tries to fix an issue due to apache#12515. Previously the logic for `-2147483648` is `parse(-literal)` = `-parse(literal)`, and all integer literals are converted to i32 (either the literal value actually overflow or not). Since after apache#12515, parse `2147483648` results in an i64 typed integer rather than i32, `-2147483648` then becomes an i64 integer too, which is not reasonable.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Refer to issues #12377 and #12378. The behavior of compile time constant folding and general path computation differs.
The changes are:
Add more strict value range checking when build IntImm object.
For constant folding on integers
For constant folding on floats
cc @junrushao1994