This is the mail archive of the glibc-cvs@sourceware.org mailing list for the glibc project.
Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]
Other format:	[Raw text]
GNU C Library master sources branch hjl/pr18661 created. glibc-2.21-660-ge52b127

From: hjl at sourceware dot org
To: glibc-cvs at sourceware dot org
Date: 28 Jul 2015 13:01:32 -0000
Subject: GNU C Library master sources branch hjl/pr18661 created. glibc-2.21-660-ge52b127
This is an automated email from the git hooks/post-receive script. It was
generated because a ref change was pushed to the repository containing
the project "GNU C Library master sources".

The branch, hjl/pr18661 has been created
        at  e52b127fe28fce951cab0bcc55904b68dd39a6c3 (commit)

- Log -----------------------------------------------------------------
http://sourceware.org/git/gitweb.cgi?p=glibc.git;a=commitdiff;h=e52b127fe28fce951cab0bcc55904b68dd39a6c3

commit e52b127fe28fce951cab0bcc55904b68dd39a6c3
Author: H.J. Lu <hjl.tools@gmail.com>
Date:   Tue Jul 28 05:47:54 2015 -0700

    Check AVX/AVX512 in elf_machine_runtime_setup

diff --git a/sysdeps/x86_64/dl-machine.h b/sysdeps/x86_64/dl-machine.h
index cae6db3..7ec39dc 100644
--- a/sysdeps/x86_64/dl-machine.h
+++ b/sysdeps/x86_64/dl-machine.h
@@ -26,6 +26,7 @@
 #include <sysdep.h>
 #include <tls.h>
 #include <dl-tlsdesc.h>
+#include <cpuid.h>
 
 /* Return nonzero iff ELF header is compatible with the running host.  */
 static inline int __attribute__ ((unused))
@@ -70,6 +71,27 @@ elf_machine_runtime_setup (struct link_map *l, int lazy, int profile)
 
   if (l->l_info[DT_JMPREL] && lazy)
     {
+      unsigned int eax, ebx, ecx, edx;
+      bool has_avx, has_avx512f;
+
+      __cpuid (1, eax, ebx, ecx, edx);
+      if ((ecx & (bit_AVX | bit_OSXSAVE)) == (bit_AVX | bit_OSXSAVE))
+	{
+	  __cpuid_count (7, 0, eax, ebx, ecx, edx);
+	  has_avx512f = (ebx & bit_AVX512F) != 0;
+
+	  /* Verify that ZMM, YMM and XMM states are enabled.  */
+	  asm ("xgetbv" : "=a" (eax), "=d" (edx) : "c" (0));
+	  has_avx = (eax & 6) == 6;
+	  if (has_avx512f)
+	    has_avx512f = (eax & 0xe6) == 0xe6;
+	}
+      else
+	{
+	  has_avx = false;
+	  has_avx512f = false;
+	}
+
       /* The GOT entries for functions in the PLT have not yet been filled
 	 in.  Their initial contents will arrange when called to push an
 	 offset into the .rel.plt section, push _GLOBAL_OFFSET_TABLE_[1],
@@ -94,7 +116,12 @@ elf_machine_runtime_setup (struct link_map *l, int lazy, int profile)
 	 end in this function.  */
       if (__glibc_unlikely (profile))
 	{
-	  *(ElfW(Addr) *) (got + 2) = (ElfW(Addr)) &_dl_runtime_profile;
+	  if (has_avx)
+	    *(ElfW(Addr) *) (got + 2) = (ElfW(Addr)) &_dl_runtime_profile;
+	  else if (has_avx512f)
+	    *(ElfW(Addr) *) (got + 2) = (ElfW(Addr)) &_dl_runtime_profile;
+	  else
+	    *(ElfW(Addr) *) (got + 2) = (ElfW(Addr)) &_dl_runtime_profile;
 
 	  if (GLRO(dl_profile) != NULL
 	      && _dl_name_match_p (GLRO(dl_profile), l))
@@ -103,9 +130,17 @@ elf_machine_runtime_setup (struct link_map *l, int lazy, int profile)
 	    GL(dl_profile_map) = l;
 	}
       else
-	/* This function will get called to fix up the GOT entry indicated by
-	   the offset on the stack, and then jump to the resolved address.  */
-	*(ElfW(Addr) *) (got + 2) = (ElfW(Addr)) &_dl_runtime_resolve;
+	{
+	  /* This function will get called to fix up the GOT entry
+	     indicated by the offset on the stack, and then jump to
+	     the resolved address.  */
+	  if (has_avx)
+	    *(ElfW(Addr) *) (got + 2) = (ElfW(Addr)) &_dl_runtime_resolve;
+	  else if (has_avx512f)
+	    *(ElfW(Addr) *) (got + 2) = (ElfW(Addr)) &_dl_runtime_resolve;
+	  else
+	    *(ElfW(Addr) *) (got + 2) = (ElfW(Addr)) &_dl_runtime_resolve;
+	}
     }
 
   if (l->l_info[ADDRIDX (DT_TLSDESC_GOT)] && lazy)

http://sourceware.org/git/gitweb.cgi?p=glibc.git;a=commitdiff;h=856a4116f1f939e81197afce989fa91a1c1ea079

commit 856a4116f1f939e81197afce989fa91a1c1ea079
Author: H.J. Lu <hjl.tools@gmail.com>
Date:   Sat Jul 11 13:25:25 2015 -0700

    Save and restore xmm0-xmm7 in _dl_runtime_resolve

diff --git a/sysdeps/x86_64/dl-trampoline.S b/sysdeps/x86_64/dl-trampoline.S
index 678c57f..c527881 100644
--- a/sysdeps/x86_64/dl-trampoline.S
+++ b/sysdeps/x86_64/dl-trampoline.S
@@ -27,26 +27,36 @@
 /* Area on stack to save and restore registers used for parameter
    passing when calling _dl_fixup.  */
 #ifdef __ILP32__
-/* X32 saves RCX, RDX, RSI, RDI, R8 and R9 plus RAX.  */
-# define REGISTER_SAVE_AREA	(8 * 7)
-# define REGISTER_SAVE_RAX	0
+/* X32 saves RCX, RDX, RSI, RDI, R8 and R9 plus RAX as well as XMM0 to
+   XMM7.  */
+# define REGISTER_SAVE_AREA	(8 * 7 + 16 * 8)
+/* Align XMM register save area to 16 bytes.  */
+# define REGISTER_SAVE_XMM0	0
 # define PRESERVE_BND_REGS_PREFIX
 #else
-/* X86-64 saves RCX, RDX, RSI, RDI, R8 and R9 plus RAX as well as BND0,
-   BND1, BND2, BND3.  */
-# define REGISTER_SAVE_AREA	(8 * 7 + 16 * 4)
+/* X86-64 saves RCX, RDX, RSI, RDI, R8 and R9 plus RAX as well as
+   BND0, BND1, BND2, BND3 and XMM0 to XMM7. */
+# define REGISTER_SAVE_AREA	(8 * 7 + 16 * 4 + 16 * 8)
 /* Align bound register save area to 16 bytes.  */
 # define REGISTER_SAVE_BND0	0
 # define REGISTER_SAVE_BND1	(REGISTER_SAVE_BND0 + 16)
 # define REGISTER_SAVE_BND2	(REGISTER_SAVE_BND1 + 16)
 # define REGISTER_SAVE_BND3	(REGISTER_SAVE_BND2 + 16)
-# define REGISTER_SAVE_RAX	(REGISTER_SAVE_BND3 + 16)
+# define REGISTER_SAVE_XMM0	(REGISTER_SAVE_BND3 + 16)
 # ifdef HAVE_MPX_SUPPORT
 #  define PRESERVE_BND_REGS_PREFIX bnd
 # else
 #  define PRESERVE_BND_REGS_PREFIX .byte 0xf2
 # endif
 #endif
+#define REGISTER_SAVE_XMM1	(REGISTER_SAVE_XMM0 + 16)
+#define REGISTER_SAVE_XMM2	(REGISTER_SAVE_XMM1 + 16)
+#define REGISTER_SAVE_XMM3	(REGISTER_SAVE_XMM2 + 16)
+#define REGISTER_SAVE_XMM4	(REGISTER_SAVE_XMM3 + 16)
+#define REGISTER_SAVE_XMM5	(REGISTER_SAVE_XMM4 + 16)
+#define REGISTER_SAVE_XMM6	(REGISTER_SAVE_XMM5 + 16)
+#define REGISTER_SAVE_XMM7	(REGISTER_SAVE_XMM6 + 16)
+#define REGISTER_SAVE_RAX	(REGISTER_SAVE_XMM7 + 16)
 #define REGISTER_SAVE_RCX	(REGISTER_SAVE_RAX + 8)
 #define REGISTER_SAVE_RDX	(REGISTER_SAVE_RCX + 8)
 #define REGISTER_SAVE_RSI	(REGISTER_SAVE_RDX + 8)
@@ -54,6 +64,10 @@
 #define REGISTER_SAVE_R8	(REGISTER_SAVE_RDI + 8)
 #define REGISTER_SAVE_R9	(REGISTER_SAVE_R8 + 8)
 
+#if (REGISTER_SAVE_AREA % 16) != 8
+# error REGISTER_SAVE_AREA must be odd multples of 8
+#endif
+
 	.text
 	.globl _dl_runtime_resolve
 	.type _dl_runtime_resolve, @function
@@ -71,6 +85,15 @@ _dl_runtime_resolve:
 	movq %rdi, REGISTER_SAVE_RDI(%rsp)
 	movq %r8, REGISTER_SAVE_R8(%rsp)
 	movq %r9, REGISTER_SAVE_R9(%rsp)
+	# movaps is 1-byte shorter.
+	movaps %xmm0, REGISTER_SAVE_XMM0(%rsp)
+	movaps %xmm1, REGISTER_SAVE_XMM1(%rsp)
+	movaps %xmm2, REGISTER_SAVE_XMM2(%rsp)
+	movaps %xmm3, REGISTER_SAVE_XMM3(%rsp)
+	movaps %xmm4, REGISTER_SAVE_XMM4(%rsp)
+	movaps %xmm5, REGISTER_SAVE_XMM5(%rsp)
+	movaps %xmm6, REGISTER_SAVE_XMM6(%rsp)
+	movaps %xmm7, REGISTER_SAVE_XMM7(%rsp)
 #ifndef __ILP32__
 	# We also have to preserve bound registers.  These are nops if
 	# Intel MPX isn't available or disabled.
@@ -123,6 +146,15 @@ _dl_runtime_resolve:
 	movq REGISTER_SAVE_RDX(%rsp), %rdx
 	movq REGISTER_SAVE_RCX(%rsp), %rcx
 	movq REGISTER_SAVE_RAX(%rsp), %rax
+	# movaps is 1-byte shorter.
+	movaps REGISTER_SAVE_XMM0(%rsp), %xmm0
+	movaps REGISTER_SAVE_XMM1(%rsp), %xmm1
+	movaps REGISTER_SAVE_XMM2(%rsp), %xmm2
+	movaps REGISTER_SAVE_XMM3(%rsp), %xmm3
+	movaps REGISTER_SAVE_XMM4(%rsp), %xmm4
+	movaps REGISTER_SAVE_XMM5(%rsp), %xmm5
+	movaps REGISTER_SAVE_XMM6(%rsp), %xmm6
+	movaps REGISTER_SAVE_XMM7(%rsp), %xmm7
 	# Adjust stack(PLT did 2 pushes)
 	addq $(REGISTER_SAVE_AREA + 16), %rsp
 	cfi_adjust_cfa_offset(-(REGISTER_SAVE_AREA + 16))

http://sourceware.org/git/gitweb.cgi?p=glibc.git;a=commitdiff;h=568bd0a08f7383197b338f3f40a57b7b9cdc02ab

commit 568bd0a08f7383197b338f3f40a57b7b9cdc02ab
Author: H.J. Lu <hjl.tools@gmail.com>
Date:   Sun Jul 12 14:41:20 2015 -0700

    Align the stack before calling __errno_location

diff --git a/sysdeps/x86_64/fpu/s_cosf.S b/sysdeps/x86_64/fpu/s_cosf.S
index b7868ce..bea10ef 100644
--- a/sysdeps/x86_64/fpu/s_cosf.S
+++ b/sysdeps/x86_64/fpu/s_cosf.S
@@ -310,8 +310,14 @@ L(arg_inf_or_nan):
 	/* Here if |x| is Inf or NAN */
 	jne	L(skip_errno_setting)	/* in case of x is NaN */
 
+	/* Align stack to 16 bytes.  */
+	subq	$8, %rsp
+	cfi_adjust_cfa_offset (8)
 	/* Here if x is Inf. Set errno to EDOM.  */
 	call	JUMPTARGET(__errno_location)
+	addq	$8, %rsp
+	cfi_adjust_cfa_offset (-8)
+
 	movl	$EDOM, (%rax)
 
 	.p2align	4
diff --git a/sysdeps/x86_64/fpu/s_sincosf.S b/sysdeps/x86_64/fpu/s_sincosf.S
index 21db70a..a2f3133 100644
--- a/sysdeps/x86_64/fpu/s_sincosf.S
+++ b/sysdeps/x86_64/fpu/s_sincosf.S
@@ -354,8 +354,14 @@ L(arg_inf_or_nan):
 	/* Here if |x| is Inf or NAN */
 	jne	L(skip_errno_setting)	/* in case of x is NaN */
 
+	/* Align stack to 16 bytes.  */
+	subq	$8, %rsp
+	cfi_adjust_cfa_offset (8)
 	/* Here if x is Inf. Set errno to EDOM.  */
 	call	JUMPTARGET(__errno_location)
+	addq	$8, %rsp
+	cfi_adjust_cfa_offset (-8)
+
 	movl	$EDOM, (%rax)
 
 	.p2align	4
diff --git a/sysdeps/x86_64/fpu/s_sinf.S b/sysdeps/x86_64/fpu/s_sinf.S
index dc92164..90afbe8 100644
--- a/sysdeps/x86_64/fpu/s_sinf.S
+++ b/sysdeps/x86_64/fpu/s_sinf.S
@@ -336,8 +336,14 @@ L(arg_inf_or_nan):
 	/* Here if |x| is Inf or NAN */
 	jne	L(skip_errno_setting)	/* in case of x is NaN */
 
+	/* Align stack to 16 bytes.  */
+	subq	$8, %rsp
+	cfi_adjust_cfa_offset (8)
 	/* Here if x is Inf. Set errno to EDOM.  */
 	call	JUMPTARGET(__errno_location)
+	addq	$8, %rsp
+	cfi_adjust_cfa_offset (-8)
+
 	movl	$EDOM, (%rax)
 
 	.p2align	4

http://sourceware.org/git/gitweb.cgi?p=glibc.git;a=commitdiff;h=a1cb3d630d9a28458220464f3935aef1157f69a2

commit a1cb3d630d9a28458220464f3935aef1157f69a2
Author: H.J. Lu <hjl.tools@gmail.com>
Date:   Sun Jul 12 14:40:25 2015 -0700

    Align the stack before calling __gettimeofday

diff --git a/sysdeps/unix/sysv/linux/x86_64/lowlevellock.S b/sysdeps/unix/sysv/linux/x86_64/lowlevellock.S
index 0935db5..23f3def 100644
--- a/sysdeps/unix/sysv/linux/x86_64/lowlevellock.S
+++ b/sysdeps/unix/sysv/linux/x86_64/lowlevellock.S
@@ -394,8 +394,8 @@ __lll_timedwait_tid:
 	movq	%rdi, %r12
 	movq	%rsi, %r13
 
-	subq	$16, %rsp
-	cfi_adjust_cfa_offset(16)
+	subq	$24, %rsp
+	cfi_adjust_cfa_offset(24)
 
 	/* Get current time.  */
 2:	movq	%rsp, %rdi
@@ -441,8 +441,8 @@ __lll_timedwait_tid:
 	jne	1f
 4:	xorl	%eax, %eax
 
-8:	addq	$16, %rsp
-	cfi_adjust_cfa_offset(-16)
+8:	addq	$24, %rsp
+	cfi_adjust_cfa_offset(-24)
 	popq	%r13
 	cfi_adjust_cfa_offset(-8)
 	cfi_restore(%r13)

http://sourceware.org/git/gitweb.cgi?p=glibc.git;a=commitdiff;h=1fc73512dcc30036ffa9ef7363a9eb3442972e27

commit 1fc73512dcc30036ffa9ef7363a9eb3442972e27
Author: H.J. Lu <hjl.tools@gmail.com>
Date:   Sun Jul 12 14:38:58 2015 -0700

    Align stack before calling __setcontext

diff --git a/sysdeps/unix/sysv/linux/x86_64/__start_context.S b/sysdeps/unix/sysv/linux/x86_64/__start_context.S
index 52a5afa..57ee2b9 100644
--- a/sysdeps/unix/sysv/linux/x86_64/__start_context.S
+++ b/sysdeps/unix/sysv/linux/x86_64/__start_context.S
@@ -33,6 +33,11 @@ ENTRY(__start_context)
 
 	popq	%rdi			/* This is the next context.  */
 	cfi_adjust_cfa_offset(-8)
+
+	/* Align stack to 16 bytes.  */
+	subq	$8, %rsp
+	cfi_adjust_cfa_offset (8)
+
 	testq	%rdi, %rdi
 	je	2f			/* If it is zero exit.  */
 

-----------------------------------------------------------------------


hooks/post-receive
-- 
GNU C Library master sources
Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]