[clang][Modules] Raise empty.modulemap expected size to <70KB to fix RISC-V failure #123959
Conversation
…RISC-V failure I'm not sure why the test is larger for RISC-V than other targets, but we saw this before with llvm#111360. The file is just over the current 60KB limit: ``` 62772 /home/asb/llvm-project/build/stage2/tools/clang/test/Modules/Output/empty.modulemap.tmp/base.pcm ```
llvmbot
added
clang
|
@llvm/pr-subscribers-clang-modules Author: Alex Bradbury (asb) ChangesI'm not sure why the test is larger for RISC-V than other targets, but we saw this before with #111360. The file is just over the current 60KB limit: Full diff: https://github.com/llvm/llvm-project/pull/123959.diff 1 Files Affected:
diff --git a/clang/test/Modules/empty.modulemap b/clang/test/Modules/empty.modulemap
index f2d37c19d77bcc..8cad8b67b91155 100644
--- a/clang/test/Modules/empty.modulemap
+++ b/clang/test/Modules/empty.modulemap
@@ -13,8 +13,8 @@
// The module file should be identical each time we produce it.
// RUN: diff %t/base.pcm %t/check.pcm
//
-// We expect an empty module to be less than 60KB (and at least 10K, for now).
+// We expect an empty module to be less than 70KB (and at least 10K, for now).
// RUN: wc -c %t/base.pcm | FileCheck --check-prefix=CHECK-SIZE %s
-// CHECK-SIZE: {{(^|[^0-9])[1-5][0-9][0-9][0-9][0-9]($|[^0-9])}}
+// CHECK-SIZE: {{(^|[^0-9])[1-6][0-9][0-9][0-9][0-9]($|[^0-9])}}
module empty { header "Inputs/empty.h" export * }
|
|
@llvm/pr-subscribers-clang Author: Alex Bradbury (asb) ChangesI'm not sure why the test is larger for RISC-V than other targets, but we saw this before with #111360. The file is just over the current 60KB limit: Full diff: https://github.com/llvm/llvm-project/pull/123959.diff 1 Files Affected:
diff --git a/clang/test/Modules/empty.modulemap b/clang/test/Modules/empty.modulemap
index f2d37c19d77bcc..8cad8b67b91155 100644
--- a/clang/test/Modules/empty.modulemap
+++ b/clang/test/Modules/empty.modulemap
@@ -13,8 +13,8 @@
// The module file should be identical each time we produce it.
// RUN: diff %t/base.pcm %t/check.pcm
//
-// We expect an empty module to be less than 60KB (and at least 10K, for now).
+// We expect an empty module to be less than 70KB (and at least 10K, for now).
// RUN: wc -c %t/base.pcm | FileCheck --check-prefix=CHECK-SIZE %s
-// CHECK-SIZE: {{(^|[^0-9])[1-5][0-9][0-9][0-9][0-9]($|[^0-9])}}
+// CHECK-SIZE: {{(^|[^0-9])[1-6][0-9][0-9][0-9][0-9]($|[^0-9])}}
module empty { header "Inputs/empty.h" export * }
|
ChuanqiXu9
left a comment
ChuanqiXu9
left a comment
There was a problem hiding this comment.
CC @Bigcheese
But I think the patch itself is good to go. Let's avoid breaking the CI.
|
Thanks, I've gone ahead an merged. Any insight into why the module files may be larger on RISC-V vs other targets greatly appreciated - as I said in #111360 I'm not sure how to best inspect the module files. |
|
It could be due to builtins, I believe we always serialize them, and that would be different between targets. Does RISC-V have a lot of them compared to other targets? |
That's almost certainly it then, thanks! We have a lot of vector builtins in particular. |
4 participants
I'm not sure why the test is larger for RISC-V than other targets, but we saw this before with #111360.
The file is just over the current 60KB limit: