[BOLT] link_fdata support for hardcoded offsets in no-LBR #113157
Closed
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
For example, the below fails as the script tries to match it with the LBR mode: ``` // RUN: link_fdata --no-lbr ... // FDATA: 1 foo 0 10 ```
Allow using hardcoded offsets to blocks, eg: ``` // RUN: link_fdata --no-lbr ... // FDATA: 1 foo 0 10 ``` For LBR mode this was already supported. See for example: - tail-duplication-complex.s
paschalis-mpeis
changed the title
paschalis-mpeis
requested review from
aaupov,
ayermolo,
dcci,
maksfb and
rafaelauler
as code owners
Member
|
@llvm/pr-subscribers-bolt Author: Paschalis Mpeis (paschalis-mpeis) ChangesAllow using hardcoded offsets to blocks, eg: For LBR mode this was already supported. See for example:
Full diff: https://github.com/llvm/llvm-project/pull/113157.diff 2 Files Affected:
diff --git a/bolt/test/AArch64/fdata-nolbr.s b/bolt/test/AArch64/fdata-nolbr.s
new file mode 100644
index 00000000000000..5d7c0a4014986e
--- /dev/null
+++ b/bolt/test/AArch64/fdata-nolbr.s
@@ -0,0 +1,19 @@
+# Check that using link_fdata tool in non-lbr mode allows using hardcoded
+# addresses for basic block offsets.
+
+# REQUIRES: system-linux
+
+# RUN: %clang %cflags -o %t %s
+# RUN: %clang %s %cflags -Wl,-q -o %t
+# RUN: link_fdata --no-lbr %s %t %t.fdata
+# RUN: cat %t.fdata | FileCheck %s
+
+ .text
+ .globl foo
+ .type foo, %function
+foo:
+# FDATA: 1 foo 0 10
+ ret
+
+# CHECK: no_lbr
+# CHECK-NEXT: 1 foo 0 10
diff --git a/bolt/test/link_fdata.py b/bolt/test/link_fdata.py
index 3837e394ccc87b..613d58248bb052 100755
--- a/bolt/test/link_fdata.py
+++ b/bolt/test/link_fdata.py
@@ -58,7 +58,7 @@
fdata_match = fdata_pat.match(profile_line)
preagg_match = preagg_pat.match(profile_line)
nolbr_match = nolbr_pat.match(profile_line)
- if fdata_match:
+ if fdata_match and not (nolbr_match and args.no_lbr):
src_dst, execnt, mispred = fdata_match.groups()
# Split by whitespaces not preceded by a backslash (negative lookbehind)
chunks = re.split(r"(?<!\\) +", src_dst)
|
Member
Author
|
Just adding that people could (and are) using LBR mode with link_fdata to set hardcoded offsets to blocks, regardless of architecture. So there is a way without this minor patch. Based on that I could be closing this PR unless I hear otherwise in the next couple of days. |
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.
Allow using hardcoded offsets to blocks, eg:
For LBR mode this was already supported. See for example: