[compiler-rt][ubsan] Add support for f16 #129624
Merged
Conversation
This file contains hidden or 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
llvmbot
added
compiler-rt
compiler-rt:ubsan
|
@llvm/pr-subscribers-compiler-rt-sanitizer Author: Alexander Shaposhnikov (alexander-shaposhnikov) ChangesLLVM supports long double <-> f16 conversions so we can remove the old FIXME. Full diff: https://github.com/llvm/llvm-project/pull/129624.diff 2 Files Affected:
diff --git a/compiler-rt/lib/ubsan/ubsan_value.cpp b/compiler-rt/lib/ubsan/ubsan_value.cpp
index 6e88ebaf34d4b..609253ae16622 100644
--- a/compiler-rt/lib/ubsan/ubsan_value.cpp
+++ b/compiler-rt/lib/ubsan/ubsan_value.cpp
@@ -120,15 +120,11 @@ FloatMax Value::getFloatValue() const {
CHECK(getType().isFloatTy());
if (isInlineFloat()) {
switch (getType().getFloatBitWidth()) {
-#if 0
- // FIXME: OpenCL / NEON 'half' type. LLVM can't lower the conversion
- // from '__fp16' to 'long double'.
case 16: {
__fp16 Value;
- internal_memcpy(&Value, &Val, 4);
+ internal_memcpy(&Value, &Val, sizeof(Value));
return Value;
}
-#endif
case 32: {
float Value;
#if defined(__BYTE_ORDER__) && __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__
diff --git a/compiler-rt/test/ubsan/TestCases/Float/cast-overflow.cpp b/compiler-rt/test/ubsan/TestCases/Float/cast-overflow.cpp
index 859bc6a2fc9e7..05827f0f28e19 100644
--- a/compiler-rt/test/ubsan/TestCases/Float/cast-overflow.cpp
+++ b/compiler-rt/test/ubsan/TestCases/Float/cast-overflow.cpp
@@ -131,14 +131,19 @@ int main(int argc, char **argv) {
return 0;
}
case '5': {
- // CHECK-5: {{.*}}cast-overflow.cpp:[[@LINE+1]]:27: runtime error: {{.*}} is outside the range of representable values of type 'int'
+ // CHECK-5: {{.*}}cast-overflow.cpp:[[@LINE+1]]:28: runtime error: {{.*}} is outside the range of representable values of type 'int'
+ static int test_int = (__fp16)Inf;
+ return 0;
+ }
+ case '6': {
+ // CHECK-6: {{.*}}cast-overflow.cpp:[[@LINE+1]]:27: runtime error: {{.*}} is outside the range of representable values of type 'int'
static int test_int = NaN;
return 0;
}
// Integer -> floating point overflow.
- case '6': {
- // CHECK-6: cast-overflow.cpp:[[@LINE+2]]:{{27: runtime error: 3.40282e\+38 is outside the range of representable values of type 'int'| __int128 not supported}}
+ case '7': {
+ // CHECK-7: cast-overflow.cpp:[[@LINE+2]]:{{27: runtime error: 3.40282e\+38 is outside the range of representable values of type 'int'| __int128 not supported}}
#if defined(__SIZEOF_INT128__) && !defined(_WIN32)
static int test_int = (float)(FloatMaxAsUInt128 + 1);
return 0;
@@ -149,9 +154,9 @@ int main(int argc, char **argv) {
return 0;
#endif
}
- case '7': {
+ case '8': {
volatile long double ld = 300.0;
- // CHECK-7: {{.*}}cast-overflow.cpp:[[@LINE+1]]:14: runtime error: 300 is outside the range of representable values of type 'char'
+ // CHECK-8: {{.*}}cast-overflow.cpp:[[@LINE+1]]:14: runtime error: 300 is outside the range of representable values of type 'char'
char c = ld;
// `c` is allowed to contain UNDEF, thus we should not use
// its value as an exit code.
|
|
✅ With the latest revision this PR passed the C/C++ code formatter. |
alexander-shaposhnikov
force-pushed
the
f16_ubsan
branch
from
97eb971 to
13f6fec
Compare
vitalybuka
reviewed
Mar 4, 2025
vitalybuka
reviewed
Mar 4, 2025
|
also CI needs fixes |
|
regarding the CI - I'd have to reformat the whole file, otherwise things look a bit awkward |
Please reformat as NFC preparation patch |
alexander-shaposhnikov
force-pushed
the
f16_ubsan
branch
2 times, most recently
from
9707cb7 to
90f35d4
Compare
vitalybuka
approved these changes
Mar 4, 2025
alexander-shaposhnikov
force-pushed
the
f16_ubsan
branch
from
90f35d4 to
2a45423
Compare
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/204/builds/2290 Here is the relevant piece of the build log for the reference |
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/205/builds/2271 Here is the relevant piece of the build log for the reference |
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/30/builds/16860 Here is the relevant piece of the build log for the reference |
|
Hi, can this be fixed-forward or should we revert to investigate? |
|
I'll revert |
|
Thank you. If you need assistance reproducing the issue let me know. |
alexander-shaposhnikov
added a commit
that referenced
this pull request
Mar 4, 2025
This reverts commit 23a30e6. The commit has broken some build bots.
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/153/builds/24583 Here is the relevant piece of the build log for the reference |
|
@jplehr - yeah, if you could say a few words about how to reproduce this issue, this would be helpful. |
|
upd. I have a local repro now. |
jph-13
pushed a commit
to jph-13/llvm-project
that referenced
this pull request
Mar 21, 2025
LLVM supports long double <-> f16 conversions so we can remove the old FIXME.
jph-13
pushed a commit
to jph-13/llvm-project
that referenced
this pull request
Mar 21, 2025
This reverts commit 23a30e6. The commit has broken some build bots.
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.
LLVM supports long double <-> f16 conversions so we can remove the old FIXME.