From patchwork Tue Aug 25 21:41:33 2015 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: "H.J. Lu" X-Patchwork-Id: 8440 Received: (qmail 12197 invoked by alias); 25 Aug 2015 21:41:37 -0000 Mailing-List: contact libc-alpha-help@sourceware.org; run by ezmlm Precedence: bulk List-Id: List-Unsubscribe: List-Subscribe: List-Archive: List-Post: List-Help: , Sender: libc-alpha-owner@sourceware.org Delivered-To: mailing list libc-alpha@sourceware.org Received: (qmail 12185 invoked by uid 89); 25 Aug 2015 21:41:36 -0000 Authentication-Results: sourceware.org; auth=none X-Virus-Found: No X-Spam-SWARE-Status: No, score=-1.3 required=5.0 tests=AWL, BAYES_00, FREEMAIL_FROM, RCVD_IN_DNSWL_LOW, SPF_PASS autolearn=ham version=3.3.2 X-HELO: mail-ob0-f181.google.com MIME-Version: 1.0 X-Received: by 10.182.148.65 with SMTP id tq1mr30543074obb.12.1440538893106; Tue, 25 Aug 2015 14:41:33 -0700 (PDT) In-Reply-To: <20150825213151.029E82C3A94@topped-with-meat.com> References: <20150825122223.GA1588@gmail.com> <20150825183521.2898C2C3A73@topped-with-meat.com> <20150825210854.AF8382C3A94@topped-with-meat.com> <20150825213151.029E82C3A94@topped-with-meat.com> Date: Tue, 25 Aug 2015 14:41:33 -0700 Message-ID: Subject: Re: [PATCH 2/7] Don't disable SSE in x86-64 ld.so From: "H.J. Lu" To: Roland McGrath Cc: GNU C Library On Tue, Aug 25, 2015 at 2:31 PM, Roland McGrath wrote: > I meant a comment for the CFLAGS-.os setting, but close enough. > > However, the x86_64/Makefile bit needs a comment for its CFLAGS-.os > setting that explains why -mno-mmx is necessary in ld.so code. We use -mno-mmx on i386 since mm registers are passed to pass __m64 arguments. On x86-64, I can't think of a good reason to use MMX in ld.so. Here is the updated patch. From f65ffc892e43fb90ed3235a30e3cecb0faa3f0d6 Mon Sep 17 00:00:00 2001 From: "H.J. Lu" Date: Tue, 28 Jul 2015 18:56:18 -0700 Subject: [PATCH] Don't disable SSE in x86-64 ld.so Since x86-64 ld.so preserves vector registers now, we can use SSE in x86-64 ld.so. We should run tst-ld-sse-use.sh only on i386. * sysdeps/x86/Makefile [$(subdir) == elf] (CFLAGS-.os, tests-special, $(objpfx)tst-ld-sse-use.out): Moved to ... * sysdeps/i386/Makefile [$(subdir) == elf] (CFLAGS-.os, tests-special, $(objpfx)tst-ld-sse-use.out): Here. Update comments. * sysdeps/x86_64/Makefile [$(subdir) == elf] (CFLAGS-.os): Add -mno-mmx for $(all-rtld-routines). * sysdeps/x86/tst-ld-sse-use.sh: Moved to ... * sysdeps/i386/tst-ld-sse-use.sh: Here. Replace x86-64 with i386. --- sysdeps/i386/Makefile | 14 ++++++++++++++ sysdeps/{x86 => i386}/tst-ld-sse-use.sh | 2 +- sysdeps/x86/Makefile | 11 ----------- sysdeps/x86_64/Makefile | 4 ++++ 4 files changed, 19 insertions(+), 12 deletions(-) rename sysdeps/{x86 => i386}/tst-ld-sse-use.sh (97%) diff --git a/sysdeps/i386/Makefile b/sysdeps/i386/Makefile index 717d8e7..168512f 100644 --- a/sysdeps/i386/Makefile +++ b/sysdeps/i386/Makefile @@ -83,3 +83,17 @@ endif ifeq ($(subdir),csu) gen-as-const-headers += tlsdesc.sym endif + +ifeq ($(subdir),elf) +# Make sure no code in ld.so uses mm/xmm/ymm/zmm registers on i386 since +# the first 3 mm/xmm/ymm/zmm registers are used to pass vector parameters +# which must be preserved. +CFLAGS-.os += $(if $(filter $(@F),$(patsubst %,%.os,$(all-rtld-routines))),\ + -mno-sse -mno-mmx) + +tests-special += $(objpfx)tst-ld-sse-use.out +$(objpfx)tst-ld-sse-use.out: ../sysdeps/i386/tst-ld-sse-use.sh $(objpfx)ld.so + @echo "Checking ld.so for SSE register use. This will take a few seconds..." + $(BASH) $< $(objpfx) '$(NM)' '$(OBJDUMP)' '$(READELF)' > $@; \ + $(evaluate-test) +endif diff --git a/sysdeps/x86/tst-ld-sse-use.sh b/sysdeps/i386/tst-ld-sse-use.sh similarity index 97% rename from sysdeps/x86/tst-ld-sse-use.sh rename to sysdeps/i386/tst-ld-sse-use.sh index 839de18..85a0651 100755 --- a/sysdeps/x86/tst-ld-sse-use.sh +++ b/sysdeps/i386/tst-ld-sse-use.sh @@ -1,5 +1,5 @@ #! /bin/bash -# Make sure no code in ld.so uses xmm/ymm/zmm registers on x86-64. +# Make sure no code in ld.so uses xmm/ymm/zmm registers on i386. # Copyright (C) 2009-2015 Free Software Foundation, Inc. # This file is part of the GNU C Library. diff --git a/sysdeps/x86/Makefile b/sysdeps/x86/Makefile index c262fdf..0de4f42 100644 --- a/sysdeps/x86/Makefile +++ b/sysdeps/x86/Makefile @@ -1,14 +1,3 @@ -ifeq ($(subdir),elf) -CFLAGS-.os += $(if $(filter $(@F),$(patsubst %,%.os,$(all-rtld-routines))),\ - -mno-sse -mno-mmx) - -tests-special += $(objpfx)tst-ld-sse-use.out -$(objpfx)tst-ld-sse-use.out: ../sysdeps/x86/tst-ld-sse-use.sh $(objpfx)ld.so - @echo "Checking ld.so for SSE register use. This will take a few seconds..." - $(BASH) $< $(objpfx) '$(NM)' '$(OBJDUMP)' '$(READELF)' > $@; \ - $(evaluate-test) -endif - ifeq ($(subdir),csu) gen-as-const-headers += cpu-features-offsets.sym rtld-global-offsets.sym endif diff --git a/sysdeps/x86_64/Makefile b/sysdeps/x86_64/Makefile index e50bcad..6c28318 100644 --- a/sysdeps/x86_64/Makefile +++ b/sysdeps/x86_64/Makefile @@ -19,6 +19,10 @@ gen-as-const-headers += locale-defines.sym endif ifeq ($(subdir),elf) +# There is no good reason to use MMX in x86-64 ld.so with GCC. +CFLAGS-.os += $(if $(filter $(@F),$(patsubst %,%.os,$(all-rtld-routines))),\ + -mno-mmx) + sysdep-dl-routines += tlsdesc dl-tlsdesc tests += ifuncmain8 -- 2.4.3