diff mbox series

powerpc, libcpp: Fix gcc build with clang on power8 [PR97163]

Message ID 20200925090939.GC2131@tucnak
State New
Headers show
Series powerpc, libcpp: Fix gcc build with clang on power8 [PR97163] | expand

Commit Message

Jakub Jelinek Sept. 25, 2020, 9:09 a.m. UTC 
Hi!

libcpp has two specialized altivec implementations of search_line_fast,
one for power8+ and the other one otherwise.
Both use __attribute__((altivec(vector))) and the GCC builtins rather than
altivec.h and the APIs from there, which is fine, but should be restricted
to when libcpp is built with GCC, so that it can be relied on.
The second elif is
#elif (GCC_VERSION >= 4005) && defined(__ALTIVEC__) && defined (__BIG_ENDIAN__)
and thus e.g. when built with clang it isn't picked, but the first one was
just guarded with
#elif defined(_ARCH_PWR8) && defined(__ALTIVEC__)
and so according to the bugreporter clang fails miserably on that.

The following patch fixes that by adding the same GCC_VERSION requirement
as the second version.  I don't know where the 4.5 in there comes from and
the exact version doesn't matter that much, as long as it is above 4.2 that
clang pretends to be and smaller or equal to 4.8 as the oldest gcc we
support as bootstrap compiler ATM.
Furthermore, the patch fixes the comment, the version it is talking about is
not pre-GCC 5, but actually the GCC 5+ one.

Bootstrapped/regtested on powerpc64le-linux, ok for trunk?

2020-09-25  Jakub Jelinek  <jakub@redhat.com>

	PR bootstrap/97163
	* lex.c (search_line_fast): Only use _ARCH_PWR8 Altivec version
	for GCC >= 4.5.


	Jakub

Comments

Segher Boessenkool Sept. 25, 2020, 9:40 p.m. UTC | #1 
Hi!

On Fri, Sep 25, 2020 at 11:09:39AM +0200, Jakub Jelinek wrote:
> libcpp has two specialized altivec implementations of search_line_fast,
> one for power8+ and the other one otherwise.
> Both use __attribute__((altivec(vector))) and the GCC builtins rather than
> altivec.h and the APIs from there, which is fine, but should be restricted
> to when libcpp is built with GCC, so that it can be relied on.
> The second elif is
> #elif (GCC_VERSION >= 4005) && defined(__ALTIVEC__) && defined (__BIG_ENDIAN__)
> and thus e.g. when built with clang it isn't picked, but the first one was
> just guarded with
> #elif defined(_ARCH_PWR8) && defined(__ALTIVEC__)
> and so according to the bugreporter clang fails miserably on that.

Yeah.  This could be rewritten to use the more portable, more modern
intrinsics, but that would need a lot of testing etc.  Since the only
thing this does is speed up the compiler a few percent, and nothing
changes for a bootstrapped compiler (whatever the build compiler is),
it is just fine.

> 	PR bootstrap/97163
> 	* lex.c (search_line_fast): Only use _ARCH_PWR8 Altivec version
> 	for GCC >= 4.5.

Okay for trunk (and whatever backports you want of course, if any).
Thanks!


Segher
diff mbox series

Patch

--- libcpp/lex.c.jj	2020-07-28 15:39:56.434118065 +0200
+++ libcpp/lex.c	2020-09-24 18:09:06.358207369 +0200
@@ -531,11 +531,11 @@  init_vectorized_lexer (void)
   search_line_fast = impl;
 }
 
-#elif defined(_ARCH_PWR8) && defined(__ALTIVEC__)
+#elif (GCC_VERSION >= 4005) && defined(_ARCH_PWR8) && defined(__ALTIVEC__)
 
 /* A vection of the fast scanner using AltiVec vectorized byte compares
    and VSX unaligned loads (when VSX is available).  This is otherwise
-   the same as the pre-GCC 5 version.  */
+   the same as the AltiVec version.  */
 
 ATTRIBUTE_NO_SANITIZE_UNDEFINED
 static const uchar *