On Mon, 8 Sep 2025, H.J. Lu wrote:
> Don't check tls_model attribute when optimizing TLS access since
>
> 1. -ftls-model=initial-exec also can specify tls_model without tls_model
> attribute.
There's a difference: -ftls-model is usable only for upgrading the model;
if you pass -ftls-model=global-dynamic without -fpic, you still get local-exec
or initial-exec. Just a correction in case you missed it.
> 2. Linker can optimize TLS access at link-time.
> 3. LTO should perform the similar optimization.
I generally agree that GCC should perform this optimization. That said, with
this patch we arrive at a situation with both
a) ld.bfd does not obey --no-relax to switch off TLS relaxation, and
b) gcc unconditionally performs TLS access optimization
which makes investigation of TLS-related toolchain bugs difficult. Would you
consider adding a GCC flag to control this optimization? And using the flag
in the testcases?
> Since C, C++, and Fortran front-ends now set the TLS access model after
> a variable has been fully processed, not in the middle of processing it,
> partially revert
>
> commit 82e629c26647313be406c41a01e6868cfad0f289
> Author: Alexander Monakov <amona...@ispras.ru>
> Date: Wed Oct 26 16:37:34 2022 +0300
>
> ipa-visibility: remove assert in TLS optimization [PR107353]
>
> to restore assert if not compiling for shared library. When compiling
> for shared library, the recomputed model may be weaker.
I don't understand why (for shared libraries), can you explain?
> Linker may
> issue an error if the front-end assigned model isn't supported in shared
> library. But compiler doesn't know if the generated code will be linked
> into executable or shared library.
>
> gcc/
>
> PR other/107353
> PR middle-end/121352
> * ipa-visibility.cc (function_and_variable_visibility): Don't
> check tls_model attribute. Restore assert if not compiling for
> shared library.
> * doc/extend.texi: Update the tls_model attribute.
>
> gcc/testsuite/
>
> PR middle-end/121352
> * gcc.dg/tls/vis-attr-hidden-gd.c: Scan for tls-local-dynamic.
> * gcc.dg/tls/vis-flag-hidden-gd.c: Likewise.
> * gcc.dg/tls/vis-pragma-hidden-gd.c: Likewise.
>
> Signed-off-by: H.J. Lu <hjl.to...@gmail.com>
> ---
> gcc/doc/extend.texi | 5 ++++-
> gcc/ipa-visibility.cc | 19 ++++++++++++-------
> gcc/testsuite/gcc.dg/tls/vis-attr-hidden-gd.c | 4 ++--
> gcc/testsuite/gcc.dg/tls/vis-flag-hidden-gd.c | 4 ++--
> .../gcc.dg/tls/vis-pragma-hidden-gd.c | 4 ++--
> 5 files changed, 22 insertions(+), 14 deletions(-)
>
> diff --git a/gcc/doc/extend.texi b/gcc/doc/extend.texi
> index 2922d9e9839..879a759714b 100644
> --- a/gcc/doc/extend.texi
> +++ b/gcc/doc/extend.texi
> @@ -7563,7 +7563,10 @@ The @code{tls_model} attribute sets thread-local
> storage model
> overriding @option{-ftls-model=} command-line switch on a per-variable
> basis.
> The @var{tls_model} argument should be one of @code{global-dynamic},
> -@code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
> +@code{local-dynamic}, @code{initial-exec} or @code{local-exec}. The
> +@code{tls_model} attribute specifies the weakest @acronym{TLS} access
> +model. GCC may optimize @acronym{TLS} access to a stronger @acronym{TLS}
> +access model.
I think with this patch the attribute does not fully override the -ftls-model
option. If so, the 'overriding' claim in the previous paragraph should be
changed. If you add a command-line flag to control the optimization, this
section will have to be reworded.
>
> Not all targets support this attribute.
>
> diff --git a/gcc/ipa-visibility.cc b/gcc/ipa-visibility.cc
> index 8097a03e240..d2283507a31 100644
> --- a/gcc/ipa-visibility.cc
> +++ b/gcc/ipa-visibility.cc
> @@ -883,17 +883,22 @@ function_and_variable_visibility (bool whole_program)
> tree decl = vnode->decl;
>
> /* Upgrade TLS access model based on optimized visibility status,
> - unless it was specified explicitly or no references remain. */
> + unless no references remain. */
> if (DECL_THREAD_LOCAL_P (decl)
> - && !lookup_attribute ("tls_model", DECL_ATTRIBUTES (decl))
> && vnode->ref_list.referring.length ())
> {
> enum tls_model new_model = decl_default_tls_model (decl);
> - STATIC_ASSERT (TLS_MODEL_GLOBAL_DYNAMIC <
> TLS_MODEL_LOCAL_DYNAMIC);
> - STATIC_ASSERT (TLS_MODEL_INITIAL_EXEC < TLS_MODEL_LOCAL_EXEC);
> - /* We'd prefer to assert that recomputed model is not weaker than
> - what the front-end assigned, but cannot: see PR 107353. */
> - if (new_model >= decl_tls_model (decl))
> + /* Assert that the recomputed model is not weaker than
> + what the front-end assigned if not comping for shared
> + library. When compiling for shared library, the
> + recomputed model may be weaker. Linker may issue an
> + error if the front-end assigned model isn't supported
> + in shared library. But compiler doesn't know if the
> + generated code will be linked into executable or shared
> + library. */
> + tls_model model = decl_tls_model (decl);
> + gcc_checking_assert (flag_shlib || new_model >= model);
I think this can trip if a variable is redeclared such that initial declaration
implies a stronger model:
// compile with gcc -O2 -std=c11 -fno-common
__thread int x;
__attribute__((common))
__thread int x;
int *f()
{
return &x;
}
Can you check?
Thanks.
Alexander
> + if (new_model > model)
> set_decl_tls_model (decl, new_model);
> }
> }
> diff --git a/gcc/testsuite/gcc.dg/tls/vis-attr-hidden-gd.c
> b/gcc/testsuite/gcc.dg/tls/vis-attr-hidden-gd.c
> index e32565588c8..4512bff04ad 100644
> --- a/gcc/testsuite/gcc.dg/tls/vis-attr-hidden-gd.c
> +++ b/gcc/testsuite/gcc.dg/tls/vis-attr-hidden-gd.c
> @@ -3,11 +3,11 @@
> /* { dg-require-effective-target tls } */
> /* { dg-options "-O2 -fPIC -fdump-ipa-whole-program" } */
>
> -// tls_model should be global-dynamic due to explicitly specified attribute
> __attribute__((visibility("hidden")))
> __attribute__((tls_model("global-dynamic")))
> __thread int x;
>
> void reference() { x++; }
>
> -/* { dg-final { scan-ipa-dump "Varpool flags: tls-global-dynamic"
> "whole-program" } } */
> +// tls_model should be local-dynamic due to hidden visibility.
> +/* { dg-final { scan-ipa-dump "Varpool flags: tls-local-dynamic"
> "whole-program" } } */
> diff --git a/gcc/testsuite/gcc.dg/tls/vis-flag-hidden-gd.c
> b/gcc/testsuite/gcc.dg/tls/vis-flag-hidden-gd.c
> index cad41e0c8e6..fba79cfd75a 100644
> --- a/gcc/testsuite/gcc.dg/tls/vis-flag-hidden-gd.c
> +++ b/gcc/testsuite/gcc.dg/tls/vis-flag-hidden-gd.c
> @@ -4,10 +4,10 @@
> /* { dg-options "-O2 -fPIC -fdump-ipa-whole-program -fvisibility=hidden" } */
>
>
> -// tls_model should be global-dynamic due to explicitly specified attribute
> __attribute__((tls_model("global-dynamic")))
> __thread int x;
>
> void reference() { x++; }
>
> -/* { dg-final { scan-ipa-dump "Varpool flags: tls-global-dynamic"
> "whole-program" } } */
> +// tls_model should be local-dynamic due to hidden visibility.
> +/* { dg-final { scan-ipa-dump "Varpool flags: tls-local-dynamic"
> "whole-program" } } */
> diff --git a/gcc/testsuite/gcc.dg/tls/vis-pragma-hidden-gd.c
> b/gcc/testsuite/gcc.dg/tls/vis-pragma-hidden-gd.c
> index 3b3598134fe..3649b6f7272 100644
> --- a/gcc/testsuite/gcc.dg/tls/vis-pragma-hidden-gd.c
> +++ b/gcc/testsuite/gcc.dg/tls/vis-pragma-hidden-gd.c
> @@ -6,7 +6,6 @@
>
> #pragma GCC visibility push(hidden)
>
> -// tls_model should be global-dynamic due to explicitly specified attribute
> __attribute__((tls_model("global-dynamic")))
> __thread int x;
>
> @@ -14,4 +13,5 @@ __thread int x;
>
> void reference() { x++; }
>
> -/* { dg-final { scan-ipa-dump "Varpool flags: tls-global-dynamic"
> "whole-program" } } */
> +// tls_model should be local-dynamic due to hidden visibility.
> +/* { dg-final { scan-ipa-dump "Varpool flags: tls-local-dynamic"
> "whole-program" } } */
>
