Bug#611249: gcc-4.4: please update powerpcspe patches
On 17.02.2011 22:53, Sebastian Andrzej Siewior wrote: * Matthias Klose | 2011-02-16 23:25:53 [+0100]: - powerpc_remove_many.diff this helps to to detect opcodes which are not part of the current CPU because without -many gas won't touch them. This currently could break the kernel build as the 603 on steroids cpus use performance counter opcodes which are not available on the steroid less 603 core. no space at end of spec? I can't follow. It ended with -many and there was space at the end. I could end one unless you are talking about something else. could you just attach a debdiff against -12? same for the gcc-4.5 package please. Sure thing. I attached 4.4, 4.5 will follow tomorrow. ahh, and I forgot to ask about the gcc-4.6 package. -- To UNSUBSCRIBE, email to debian-gcc-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/4d5e5ac7.9010...@debian.org
Bug#611249: gcc-4.4: please update powerpcspe patches
* Matthias Klose | 2011-02-16 23:25:53 [+0100]:
- powerpc_remove_many.diff
this helps to to detect opcodes which are not part of the current
CPU because without -many gas won't touch them. This currently could
break the kernel build as the 603 on steroids cpus use performance
counter opcodes which are not available on the steroid less 603 core.
no space at end of spec?
I can't follow. It ended with -many and there was space at the end. I
could end one unless you are talking about something else.
could you just attach a debdiff against -12? same for the gcc-4.5 package
please.
Sure thing. I attached 4.4, 4.5 will follow tomorrow.
thanks, Matthias
Sebastian
diff -u gcc-4.4-4.4.5/debian/rules.patch gcc-4.4-4.4.5/debian/rules.patch
--- gcc-4.4-4.4.5/debian/rules.patch
+++ gcc-4.4-4.4.5/debian/rules.patch
@@ -82,6 +82,9 @@
linux-atomic-builtin-expect \
mips-fix-loongson2f-nop$(if $(with_linaro_branch),-linaro) \
gcc-system-root \
+ no_fpr_in_libgcc \
+ powerpc_remove_many \
+ pr44364 \
ifneq ($(with_linaro_branch),yes)
debian_patches += \
diff -u gcc-4.4-4.4.5/debian/changelog gcc-4.4-4.4.5/debian/changelog
--- gcc-4.4-4.4.5/debian/changelog
+++ gcc-4.4-4.4.5/debian/changelog
@@ -1,3 +1,13 @@
+gcc-4.4 (4.4.5-13) unstable; urgency=low
+
+ [ Sebastian Andrzej Siewior ]
+ * PR target/44364
+ * Remove -many on powerpcspe (__SPE__)
+ * Remove classic FPU opcodes from libgcc if target has no support for them
+(powerpcspe)
+
+ -- Sebastian Andrzej Siewior sebast...@breakpoint.cc Thu, 17 Feb 2011
22:27:30 +0100
+
gcc-4.4 (4.4.5-12) unstable; urgency=medium
* Fix build failure on ia64.
only in patch2:
unchanged:
--- gcc-4.4-4.4.5.orig/debian/patches/powerpc_remove_many.diff
+++ gcc-4.4-4.4.5/debian/patches/powerpc_remove_many.diff
@@ -0,0 +1,33 @@
+# DP: Subject: [PATCH] remove -many on __SPE__ target
+# DP: this helps to to detect opcodes which are not part of the current
+# DP: CPU because without -many gas won't touch them. This currently could
+# DP: break the kernel build as the 603 on steroids cpus use performance
+# DP: counter opcodes which are not available on the steroid less 603 core.
+
+Index: gcc-4.4.5/src/gcc/config/rs6000/rs6000.h
+===
+--- gcc-4.4.5.orig/src/gcc/config/rs6000/rs6000.h 2009-04-10
01:23:07.0 +0200
gcc-4.4.5/src/gcc/config/rs6000/rs6000.h 2011-02-13 17:57:17.0
+0100
+@@ -83,6 +83,12 @@
+ #define ASM_CPU_POWER7_SPEC -mpower4 -maltivec
+ #endif
+
++#ifndef __SPE__
++#define ASM_CPU_SPU_MANY_NOT_SPE -many
++#else
++#define ASM_CPU_SPU_MANY_NOT_SPE
++#endif
++
+ /* Common ASM definitions used by ASM_SPEC among the various targets
+for handling -mcpu=xxx switches. */
+ #define ASM_CPU_SPEC \
+@@ -147,7 +153,8 @@
+ %{mcpu=e300c3: -me300} \
+ %{mcpu=e500mc: -me500mc} \
+ %{maltivec: -maltivec} \
+--many
++ \
++ASM_CPU_SPU_MANY_NOT_SPE
+
+ #define CPP_DEFAULT_SPEC
+
only in patch2:
unchanged:
--- gcc-4.4-4.4.5.orig/debian/patches/no_fpr_in_libgcc.diff
+++ gcc-4.4-4.4.5/debian/patches/no_fpr_in_libgcc.diff
@@ -0,0 +1,55 @@
+# DP: It does not really harm by including them since nobody should use them
+# DP: but gas does not wont to assmebly hard float since they are not
+# DP: available on this cpu. Upstream did not respond.
+
+Index: gcc-4.4.5/src/gcc/config/rs6000/crtresfpr.asm
+===
+--- gcc-4.4.5.orig/src/gcc/config/rs6000/crtresfpr.asm 2011-02-13
17:25:36.0 +0100
gcc-4.4.5/src/gcc/config/rs6000/crtresfpr.asm 2011-02-13
17:26:14.0 +0100
+@@ -33,6 +33,7 @@
+
+ /* On PowerPC64 Linux, these functions are provided by the linker. */
+ #ifndef __powerpc64__
++#ifndef __NO_FPRS__
+
+ /* Routines for restoring floating point registers, called by the compiler.
*/
+ /* Called with r11 pointing to the stack header word of the caller of the */
+@@ -77,3 +78,4 @@
+ FUNC_END(_restfpr_14)
+
+ #endif
++#endif
+Index: gcc-4.4.5/src/gcc/config/rs6000/crtresxfpr.asm
+===
+--- gcc-4.4.5.orig/src/gcc/config/rs6000/crtresxfpr.asm2011-02-13
17:25:36.0 +0100
gcc-4.4.5/src/gcc/config/rs6000/crtresxfpr.asm 2011-02-13
17:26:29.0 +0100
+@@ -33,6 +33,7 @@
+
+ /* On PowerPC64 Linux, these functions are provided by the linker. */
+ #ifndef __powerpc64__
++#ifndef __NO_FPRS__
+
+ /* Routines for restoring floating point registers, called by the compiler.
*/
+ /* Called with r11 pointing to the stack header word of the caller of the */
+@@ -82,3 +83,4 @@
+ FUNC_END(_restfpr_14_x)
+
+ #endif
++#endif
+Index: gcc-4.4.5/src/gcc/config/rs6000/crtsavfpr.asm
+===
+--- gcc-4.4.5.orig/src/gcc/config/rs6000/crtsavfpr.asm 2011-02-13
17:25:36.0 +0100
Bug#611249: gcc-4.4: please update powerpcspe patches
Okay, I finally made it. Here is my list of gcc for powerpcspe:
- no_fpr_in_libgcc.diff
It does not really harm by including them since nobody should use them
but gas does not wont to assmebly hard float since they are not
available on this cpu. Upstream did respond.
- powerpc_remove_many.diff
this helps to to detect opcodes which are not part of the current
CPU because without -many gas won't touch them. This currently could
break the kernel build as the 603 on steroids cpus use performance
counter opcodes which are not available on the steroid less 603 core.
- pr44364.diff
This patch made it into trunk. I don't get it into the 4.4 branch
because I can't prove that it is a regression since 4.3. If you worry
about caller-save.c changes I could ifdef __SPE__ them.
I dropped pr44606 because I saw it in svn-updates.diff as of the -12
release. I also dropped the lwsync eieio patches because this
nightmare is allready addressed in gas.
Sebastian
Index: gcc-4.4.5/src/gcc/config/rs6000/crtresfpr.asm
===
--- gcc-4.4.5.orig/src/gcc/config/rs6000/crtresfpr.asm 2011-02-13
17:25:36.0 +0100
+++ gcc-4.4.5/src/gcc/config/rs6000/crtresfpr.asm 2011-02-13
17:26:14.0 +0100
@@ -33,6 +33,7 @@
/* On PowerPC64 Linux, these functions are provided by the linker. */
#ifndef __powerpc64__
+#ifndef __NO_FPRS__
/* Routines for restoring floating point registers, called by the compiler. */
/* Called with r11 pointing to the stack header word of the caller of the */
@@ -77,3 +78,4 @@
FUNC_END(_restfpr_14)
#endif
+#endif
Index: gcc-4.4.5/src/gcc/config/rs6000/crtresxfpr.asm
===
--- gcc-4.4.5.orig/src/gcc/config/rs6000/crtresxfpr.asm 2011-02-13
17:25:36.0 +0100
+++ gcc-4.4.5/src/gcc/config/rs6000/crtresxfpr.asm 2011-02-13
17:26:29.0 +0100
@@ -33,6 +33,7 @@
/* On PowerPC64 Linux, these functions are provided by the linker. */
#ifndef __powerpc64__
+#ifndef __NO_FPRS__
/* Routines for restoring floating point registers, called by the compiler. */
/* Called with r11 pointing to the stack header word of the caller of the */
@@ -82,3 +83,4 @@
FUNC_END(_restfpr_14_x)
#endif
+#endif
Index: gcc-4.4.5/src/gcc/config/rs6000/crtsavfpr.asm
===
--- gcc-4.4.5.orig/src/gcc/config/rs6000/crtsavfpr.asm 2011-02-13
17:25:36.0 +0100
+++ gcc-4.4.5/src/gcc/config/rs6000/crtsavfpr.asm 2011-02-13
17:26:42.0 +0100
@@ -33,6 +33,7 @@
/* On PowerPC64 Linux, these functions are provided by the linker. */
#ifndef __powerpc64__
+#ifndef __NO_FPRS__
/* Routines for saving floating point registers, called by the compiler. */
/* Called with r11 pointing to the stack header word of the caller of the */
@@ -77,3 +78,4 @@
FUNC_END(_savefpr_14)
#endif
+#endif
Subject: [PATCH] remove -many on __SPE__ target
Signed-off-by: Sebastian Andrzej Siewior sebast...@breakpoint.cc
---
Index: gcc-4.4.5/src/gcc/config/rs6000/rs6000.h
===
--- gcc-4.4.5.orig/src/gcc/config/rs6000/rs6000.h 2009-04-10
01:23:07.0 +0200
+++ gcc-4.4.5/src/gcc/config/rs6000/rs6000.h2011-02-13 17:57:17.0
+0100
@@ -83,6 +83,12 @@
#define ASM_CPU_POWER7_SPEC -mpower4 -maltivec
#endif
+#ifndef __SPE__
+#define ASM_CPU_SPU_MANY_NOT_SPE -many
+#else
+#define ASM_CPU_SPU_MANY_NOT_SPE
+#endif
+
/* Common ASM definitions used by ASM_SPEC among the various targets
for handling -mcpu=xxx switches. */
#define ASM_CPU_SPEC \
@@ -147,7 +153,8 @@
%{mcpu=e300c3: -me300} \
%{mcpu=e500mc: -me500mc} \
%{maltivec: -maltivec} \
--many
+ \
+ASM_CPU_SPU_MANY_NOT_SPE
#define CPP_DEFAULT_SPEC
Index: gcc/caller-save.c
===
--- a/src/gcc/caller-save.c (revision 160245)
+++ b/src/gcc/caller-save.c (working copy)
@@ -1080,7 +1080,7 @@ insert_restore (struct insn_chain *chain
/* Check that insn to restore REGNO in save_mode[regno] is
correct. */
reg_save_code (regno, save_mode[regno]) = 0)
-mem = adjust_address (mem, save_mode[regno], 0);
+mem = adjust_address_nv (mem, save_mode[regno], 0);
else
mem = copy_rtx (mem);
@@ -1161,7 +1161,7 @@ insert_save (struct insn_chain *chain, i
/* Check that insn to save REGNO in save_mode[regno] is
correct. */
reg_save_code (regno, save_mode[regno]) = 0)
-mem = adjust_address (mem, save_mode[regno], 0);
+mem = adjust_address_nv (mem, save_mode[regno], 0);
else
mem = copy_rtx (mem);
Index: gcc/config/rs6000/e500.h
===
--- a/src/gcc/config/rs6000/e500.h (revision 160245)
+++ b/src/gcc/config/rs6000/e500.h (working copy)
@@ -43,3 +43,10 @@
Bug#611249: gcc-4.4: please update powerpcspe patches
On 16.02.2011 22:44, Sebastian Andrzej Siewior wrote: Okay, I finally made it. Here is my list of gcc for powerpcspe: - no_fpr_in_libgcc.diff It does not really harm by including them since nobody should use them but gas does not wont to assmebly hard float since they are not available on this cpu. Upstream did respond. please provide # DP: lines in the patch (same for the others). - powerpc_remove_many.diff this helps to to detect opcodes which are not part of the current CPU because without -many gas won't touch them. This currently could break the kernel build as the 603 on steroids cpus use performance counter opcodes which are not available on the steroid less 603 core. no space at end of spec? - pr44364.diff This patch made it into trunk. I don't get it into the 4.4 branch because I can't prove that it is a regression since 4.3. If you worry about caller-save.c changes I could ifdef __SPE__ them. I dropped pr44606 because I saw it in svn-updates.diff as of the -12 release. I also dropped the lwsync eieio patches because this nightmare is allready addressed in gas. could you just attach a debdiff against -12? same for the gcc-4.5 package please. thanks, Matthias -- To UNSUBSCRIBE, email to debian-gcc-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/4d5c4ef1.7060...@debian.org
