Merge commit 'refs/top-bases/t/no-hp-timing' into t/no-hp-timing

10 files changed, 80 insertions, 68 deletions

diff --git a/sysdeps/powerpc/powerpc32/fpu/__longjmp-common.S b/sysdeps/powerpc/powerpc32/fpu/__longjmp-common.S
index 9d34cd9165..13611541c2 100644
--- a/sysdeps/powerpc/powerpc32/fpu/__longjmp-common.S
+++ b/sysdeps/powerpc/powerpc32/fpu/__longjmp-common.S
@@ -17,6 +17,7 @@
    <http://www.gnu.org/licenses/>.  */
 
 #include <sysdep.h>
+#include <stap-probe.h>
 #define _ASM
 #ifdef __NO_VMX__
 # include <novmxsetjmp.h>
@@ -25,7 +26,7 @@
 #endif
 
 	.machine	"altivec"
-ENTRY (__longjmp)
+ENTRY (__longjmp_symbol)
 #ifndef __NO_VMX__
 # ifdef PIC
 	mflr    r6
@@ -43,16 +44,16 @@ ENTRY (__longjmp)
 #   endif
 	mtlr    r6
 	cfi_same_value (lr)
-	lwz     r5,RTLD_GLOBAL_RO_DL_HWCAP_OFFSET+4(r5)
+	lwz     r5,RTLD_GLOBAL_RO_DL_HWCAP_OFFSET+LOWORD(r5)
 #  else
 	lwz     r5,_dl_hwcap@got(r5)
 	mtlr    r6
 	cfi_same_value (lr)
-	lwz     r5,4(r5)
+	lwz     r5,LOWORD(r5)
 #  endif
 # else
-	lis	r5,(_dl_hwcap+4)@ha
-	lwz     r5,(_dl_hwcap+4)@l(r5)
+	lis	r5,(_dl_hwcap+LOWORD)@ha
+	lwz     r5,(_dl_hwcap+LOWORD)@l(r5)
 # endif
 	andis.	r5,r5,(PPC_FEATURE_HAS_ALTIVEC >> 16)
 	beq	L(no_vmx)
@@ -143,19 +144,22 @@ L(no_vmx):
 # endif
 	PTR_DEMANGLE2 (r0, r25)
 #endif
+	/* longjmp/longjmp_target probe expects longjmp first argument (4@3),
+	   second argument (-4@4), and target address (4@0), respectively.  */
+	LIBC_PROBE (longjmp, 3, 4@3, -4@4, 4@0)
 	mtlr r0
 	lwz r21,((JB_GPRS+7)*4)(r3)
 	lfd fp21,((JB_FPRS+7*2)*4)(r3)
 	lwz r22,((JB_GPRS+8)*4)(r3)
 	lfd fp22,((JB_FPRS+8*2)*4)(r3)
-	lwz r0,(JB_CR*4)(r3)
+	lwz r5,(JB_CR*4)(r3)
 	lwz r23,((JB_GPRS+9)*4)(r3)
 	lfd fp23,((JB_FPRS+9*2)*4)(r3)
 	lwz r24,((JB_GPRS+10)*4)(r3)
 	lfd fp24,((JB_FPRS+10*2)*4)(r3)
 	lwz r25,((JB_GPRS+11)*4)(r3)
 	lfd fp25,((JB_FPRS+11*2)*4)(r3)
-	mtcrf 0xFF,r0
+	mtcrf 0xFF,r5
 	lwz r26,((JB_GPRS+12)*4)(r3)
 	lfd fp26,((JB_FPRS+12*2)*4)(r3)
 	lwz r27,((JB_GPRS+13)*4)(r3)
@@ -168,6 +172,7 @@ L(no_vmx):
 	lfd fp30,((JB_FPRS+16*2)*4)(r3)
 	lwz r31,((JB_GPRS+17)*4)(r3)
 	lfd fp31,((JB_FPRS+17*2)*4)(r3)
+	LIBC_PROBE (longjmp_target, 3, 4@3, -4@4, 4@0)
 	mr r3,r4
 	blr
-END (__longjmp)
+END (__longjmp_symbol)
diff --git a/sysdeps/powerpc/powerpc32/fpu/__longjmp.S b/sysdeps/powerpc/powerpc32/fpu/__longjmp.S
index 96e50de378..cbd42be5cb 100644
--- a/sysdeps/powerpc/powerpc32/fpu/__longjmp.S
+++ b/sysdeps/powerpc/powerpc32/fpu/__longjmp.S
@@ -22,20 +22,21 @@
 
 #if defined NOT_IN_libc
 /* Build a non-versioned object for rtld-*.  */
+# define __longjmp_symbol __longjmp
 # include "__longjmp-common.S"
 
 #else /* !NOT_IN_libc */
 /* Build a versioned object for libc.  */
-default_symbol_version (__vmx__longjmp,__longjmp,GLIBC_2.3.4);
-# define __longjmp  __vmx__longjmp
+versioned_symbol (libc, __vmx__longjmp, __longjmp, GLIBC_2_3_4);
+# define __longjmp_symbol  __vmx__longjmp
 # include "__longjmp-common.S"
 
 # if defined SHARED && SHLIB_COMPAT (libc, GLIBC_2_0, GLIBC_2_3_4)
 #  define __NO_VMX__
 #  undef JB_SIZE
-symbol_version (__novmx__longjmp,__longjmp,GLIBC_2.0);
-#  undef __longjmp
-#  define __longjmp  __novmx__longjmp
+compat_symbol (libc, __novmx__longjmp, __longjmp, GLIBC_2_0);
+#  undef __longjmp_symbol
+#  define __longjmp_symbol  __novmx__longjmp
 #  include "__longjmp-common.S"
 # endif
 #endif /* !NOT_IN_libc */
diff --git a/sysdeps/powerpc/powerpc32/fpu/s_copysign.S b/sysdeps/powerpc/powerpc32/fpu/s_copysign.S
index 840891f1c3..1da24f492e 100644
--- a/sysdeps/powerpc/powerpc32/fpu/s_copysign.S
+++ b/sysdeps/powerpc/powerpc32/fpu/s_copysign.S
@@ -29,7 +29,7 @@ ENTRY(__copysign)
 	stwu	r1,-16(r1)
 	cfi_adjust_cfa_offset (16)
 	stfd	fp2,8(r1)
-	lwz	r3,8(r1)
+	lwz	r3,8+HIWORD(r1)
 	cmpwi   r3,0
 	addi    r1,r1,16
 	cfi_adjust_cfa_offset (-16)
diff --git a/sysdeps/powerpc/powerpc32/fpu/s_copysignl.S b/sysdeps/powerpc/powerpc32/fpu/s_copysignl.S
index 4ec8389b5d..2ad6de273d 100644
--- a/sysdeps/powerpc/powerpc32/fpu/s_copysignl.S
+++ b/sysdeps/powerpc/powerpc32/fpu/s_copysignl.S
@@ -30,7 +30,7 @@ ENTRY(__copysignl)
 	fmr	fp0,fp1
 	fabs	fp1,fp1
 	fcmpu	cr7,fp0,fp1
-	lwz	r3,8(r1)
+	lwz	r3,8+HIWORD(r1)
 	cmpwi	cr6,r3,0
 	addi	r1,r1,16
 	cfi_adjust_cfa_offset (-16)
diff --git a/sysdeps/powerpc/powerpc32/fpu/s_isnan.S b/sysdeps/powerpc/powerpc32/fpu/s_isnan.S
index 98d10daf68..024252a5ca 100644
--- a/sysdeps/powerpc/powerpc32/fpu/s_isnan.S
+++ b/sysdeps/powerpc/powerpc32/fpu/s_isnan.S
@@ -37,9 +37,11 @@ weak_alias (__isnan, isnan)
 
 /* It turns out that the 'double' version will also always work for
    single-precision.  */
+#ifndef __isnan
 strong_alias (__isnan, __isnanf)
 hidden_def (__isnanf)
 weak_alias (__isnanf, isnanf)
+#endif
 
 #ifdef NO_LONG_DOUBLE
 strong_alias (__isnan, __isnanl)
diff --git a/sysdeps/powerpc/powerpc32/fpu/s_lrint.S b/sysdeps/powerpc/powerpc32/fpu/s_lrint.S
index 27881f8cc7..249fda501f 100644
--- a/sysdeps/powerpc/powerpc32/fpu/s_lrint.S
+++ b/sysdeps/powerpc/powerpc32/fpu/s_lrint.S
@@ -24,10 +24,10 @@ ENTRY (__lrint)
 	stwu	r1,-16(r1)
 	fctiw	fp13,fp1
 	stfd	fp13,8(r1)
-	nop	/* Insure the following load is in a different dispatch group */
+	nop	/* Ensure the following load is in a different dispatch group */
 	nop	/* to avoid pipe stall on POWER4&5.  */
 	nop
-	lwz	r3,12(r1)
+	lwz	r3,8+LOWORD(r1)
 	addi	r1,r1,16
 	blr
 	END (__lrint)
diff --git a/sysdeps/powerpc/powerpc32/fpu/s_lround.S b/sysdeps/powerpc/powerpc32/fpu/s_lround.S
index 92dc3787d6..6309f864b7 100644
--- a/sysdeps/powerpc/powerpc32/fpu/s_lround.S
+++ b/sysdeps/powerpc/powerpc32/fpu/s_lround.S
@@ -67,7 +67,7 @@ ENTRY (__lround)
 	nop	/* Ensure the following load is in a different dispatch  */
 	nop	/* group to avoid pipe stall on POWER4&5.  */
 	nop
-	lwz	r3,12(r1)	/* Load return as integer.  */
+	lwz	r3,8+LOWORD(r1)	/* Load return as integer.  */
 .Lout:
 	addi	r1,r1,16
 	blr
diff --git a/sysdeps/powerpc/powerpc32/fpu/s_roundf.S b/sysdeps/powerpc/powerpc32/fpu/s_roundf.S
index 2ed9ca7b40..8cff1563a0 100644
--- a/sysdeps/powerpc/powerpc32/fpu/s_roundf.S
+++ b/sysdeps/powerpc/powerpc32/fpu/s_roundf.S
@@ -19,7 +19,7 @@
 #include <sysdep.h>
 
 	.section	.rodata.cst8,"aM",@progbits,8
-	.align	2
+	.align	3
 .LC0:	/* 2**23 */
 	.long 0x4b000000
 .LC1:	/* 0.5 */
@@ -60,7 +60,6 @@ ENTRY (__roundf )
 #ifdef SHARED
 	lfs	fp10,.LC1-.LC0(r9)
 #else
-	lis	r9,.LC1@ha
 	lfs	fp10,.LC1@l(r9)
 #endif
 	ble-	cr6,.L4
diff --git a/sysdeps/powerpc/powerpc32/fpu/setjmp-common.S b/sysdeps/powerpc/powerpc32/fpu/setjmp-common.S
index 46ea2b00f9..08efd64920 100644
--- a/sysdeps/powerpc/powerpc32/fpu/setjmp-common.S
+++ b/sysdeps/powerpc/powerpc32/fpu/setjmp-common.S
@@ -17,6 +17,7 @@
    <http://www.gnu.org/licenses/>.  */
 
 #include <sysdep.h>
+#include <stap-probe.h>
 #define _ASM
 #ifdef __NO_VMX__
 # include <novmxsetjmp.h>
@@ -25,7 +26,7 @@
 #endif
 
 	.machine	"altivec"
-ENTRY (__sigsetjmp)
+ENTRY (__sigsetjmp_symbol)
 
 #ifdef PTR_MANGLE
 	mr   r5,r1
@@ -35,6 +36,9 @@ ENTRY (__sigsetjmp)
 	stw  r1,(JB_GPR1*4)(3)
 #endif
 	mflr r0
+	/* setjmp probe expects longjmp first argument (4@3), second argument
+	   (-4@4), and target address (4@0), respectively.  */
+	LIBC_PROBE (setjmp, 3, 4@3, -4@4, 4@0)
 	stw  r14,((JB_GPRS+0)*4)(3)
 	stfd fp14,((JB_FPRS+0*2)*4)(3)
 #ifdef PTR_MANGLE
@@ -94,14 +98,14 @@ ENTRY (__sigsetjmp)
 #   else
 	lwz     r5,_rtld_global_ro@got(r5)
 #   endif
-	lwz     r5,RTLD_GLOBAL_RO_DL_HWCAP_OFFSET+4(r5)
+	lwz     r5,RTLD_GLOBAL_RO_DL_HWCAP_OFFSET+LOWORD(r5)
 #  else
 	lwz     r5,_dl_hwcap@got(r5)
-	lwz     r5,4(r5)
+	lwz     r5,LOWORD(r5)
 #  endif
 # else
-	lis	r6,(_dl_hwcap+4)@ha
-	lwz     r5,(_dl_hwcap+4)@l(r6)
+	lis	r6,(_dl_hwcap+LOWORD)@ha
+	lwz     r5,(_dl_hwcap+LOWORD)@l(r6)
 # endif
 	andis.	r5,r5,(PPC_FEATURE_HAS_ALTIVEC >> 16)
 	beq	L(no_vmx)
@@ -111,44 +115,43 @@ ENTRY (__sigsetjmp)
 	stw	r0,((JB_VRSAVE)*4)(3)
 	addi	r6,r5,16
 	beq+	L(aligned_save_vmx)
-	lvsr	v0,0,r5
-	vspltisb v1,-1         /* set v1 to all 1's */
-	vspltisb v2,0          /* set v2 to all 0's */
-	vperm   v3,v2,v1,v0   /* v3 contains shift mask with num all 1 bytes on left = misalignment  */
 
+	lvsr	v0,0,r5
+	lvsl	v1,0,r5
+	addi	r6,r5,-16
 
-	/* Special case for v20 we need to preserve what is in save area below v20 before obliterating it */
-	lvx     v5,0,r5
-	vperm   v20,v20,v20,v0
-	vsel    v5,v5,v20,v3
-	vsel    v20,v20,v2,v3
-	stvx    v5,0,r5
+# define save_misaligned_vmx(savevr,prevvr,shiftvr,tmpvr,savegpr,addgpr) \
+	addi	addgpr,addgpr,32;					 \
+	vperm	tmpvr,prevvr,savevr,shiftvr;				 \
+	stvx	tmpvr,0,savegpr
 
-#define save_2vmx_partial(savevr,prev_savevr,hivr,shiftvr,maskvr,savegpr,addgpr) \
-	addi    addgpr,addgpr,32; \
-	vperm   savevr,savevr,savevr,shiftvr; \
-	vsel    hivr,prev_savevr,savevr,maskvr; \
-	stvx    hivr,0,savegpr;
+	/*
+	 * We have to be careful not to corrupt the data below v20 and
+	 * above v31. To keep things simple we just rotate both ends in
+	 * the opposite direction to our main permute so we can use
+	 * the common macro.
+	 */
 
-	save_2vmx_partial(v21,v20,v5,v0,v3,r6,r5)
-	save_2vmx_partial(v22,v21,v5,v0,v3,r5,r6)
-	save_2vmx_partial(v23,v22,v5,v0,v3,r6,r5)
-	save_2vmx_partial(v24,v23,v5,v0,v3,r5,r6)
-	save_2vmx_partial(v25,v24,v5,v0,v3,r6,r5)
-	save_2vmx_partial(v26,v25,v5,v0,v3,r5,r6)
-	save_2vmx_partial(v27,v26,v5,v0,v3,r6,r5)
-	save_2vmx_partial(v28,v27,v5,v0,v3,r5,r6)
-	save_2vmx_partial(v29,v28,v5,v0,v3,r6,r5)
-	save_2vmx_partial(v30,v29,v5,v0,v3,r5,r6)
+	/* load and rotate data below v20 */
+	lvx	v2,0,r5
+	vperm	v2,v2,v2,v1
+	save_misaligned_vmx(v20,v2,v0,v3,r5,r6)
+	save_misaligned_vmx(v21,v20,v0,v3,r6,r5)
+	save_misaligned_vmx(v22,v21,v0,v3,r5,r6)
+	save_misaligned_vmx(v23,v22,v0,v3,r6,r5)
+	save_misaligned_vmx(v24,v23,v0,v3,r5,r6)
+	save_misaligned_vmx(v25,v24,v0,v3,r6,r5)
+	save_misaligned_vmx(v26,v25,v0,v3,r5,r6)
+	save_misaligned_vmx(v27,v26,v0,v3,r6,r5)
+	save_misaligned_vmx(v28,v27,v0,v3,r5,r6)
+	save_misaligned_vmx(v29,v28,v0,v3,r6,r5)
+	save_misaligned_vmx(v30,v29,v0,v3,r5,r6)
+	save_misaligned_vmx(v31,v30,v0,v3,r6,r5)
+	/* load and rotate data above v31 */
+	lvx	v2,0,r6
+	vperm	v2,v2,v2,v1
+	save_misaligned_vmx(v2,v31,v0,v3,r5,r6)
 
-	/* Special case for r31 we need to preserve what is in save area above v31 before obliterating it */
-	addi    r5,r5,32
-	vperm   v31,v31,v31,v0
-	lvx     v4,0,r5
-	vsel    v5,v30,v31,v3
-	stvx    v5,0,r6
-	vsel    v4,v31,v4,v3
-	stvx    v4,0,r5
 	b	L(no_vmx)
 
 L(aligned_save_vmx):
@@ -176,5 +179,5 @@ L(aligned_save_vmx):
 	stvx	31,0,r6
 L(no_vmx):
 #endif
-	b __sigjmp_save@local
-END (__sigsetjmp)
+	b __sigjmp_save_symbol@local
+END (__sigsetjmp_symbol)
diff --git a/sysdeps/powerpc/powerpc32/fpu/setjmp.S b/sysdeps/powerpc/powerpc32/fpu/setjmp.S
index 60cd350522..566aa34d0f 100644
--- a/sysdeps/powerpc/powerpc32/fpu/setjmp.S
+++ b/sysdeps/powerpc/powerpc32/fpu/setjmp.S
@@ -22,23 +22,25 @@
 
 #if defined NOT_IN_libc
 /* Build a non-versioned object for rtld-*.  */
+# define __sigsetjmp_symbol __sigsetjmp
+# define __sigjmp_save_symbol __sigjmp_save
 # include "setjmp-common.S"
 
 #else /* !NOT_IN_libc */
 /* Build a versioned object for libc.  */
-default_symbol_version (__vmx__sigsetjmp,__sigsetjmp,GLIBC_2.3.4)
-# define __sigsetjmp __vmx__sigsetjmp
-# define __sigjmp_save __vmx__sigjmp_save
+versioned_symbol (libc, __vmx__sigsetjmp, __sigsetjmp, GLIBC_2_3_4)
+# define __sigsetjmp_symbol __vmx__sigsetjmp
+# define __sigjmp_save_symbol __vmx__sigjmp_save
 # include "setjmp-common.S"
 
 # if defined SHARED && SHLIB_COMPAT (libc, GLIBC_2_0, GLIBC_2_3_4)
 #  define __NO_VMX__
-#  undef __sigsetjmp
-#  undef __sigjmp_save
+#  undef __sigsetjmp_symbol
+#  undef __sigjmp_save_symbol
 #  undef JB_SIZE
-symbol_version (__novmx__sigsetjmp,__sigsetjmp,GLIBC_2.0)
-#  define __sigsetjmp __novmx__sigsetjmp
-#  define __sigjmp_save __novmx__sigjmp_save
+compat_symbol (libc, __novmx__sigsetjmp, __sigsetjmp, GLIBC_2_0)
+#  define __sigsetjmp_symbol __novmx__sigsetjmp
+#  define __sigjmp_save_symbol __novmx__sigjmp_save
 #  include "setjmp-common.S"
 # endif
 #endif /* !NOT_IN_libc */