1.0.22.15: 16 byte align assembly routines on x86 and x86-64

 * Modern processors like it that way, and x86-64 was mostly there
   already.

 * Also fix a stupid typo in bug 419.

diff --git a/BUGS b/BUGS
index f58784a..ddd560b 100644 (file)
--- a/BUGS
+++ b/BUGS
@@ -1818,7 +1818,7 @@ WORKAROUND:
 
   (ASSERT (EQUAL (BUG419 42) '(1 2 3 4 5 6))) => failure
 
-  Note: as of SBCL 1.0.26.29 this bug no longer affects user code, as
+  Note: as of SBCL 1.0.16.29 this bug no longer affects user code, as
   SB-INT:TRULY-DYNAMIC-EXTENT needs to be used instead of
   DYNAMIC-EXTENT for this to happen. Proper fix for this bug requires
   (Nikodemus thinks) storing the relevant LAMBDA-VARs in a

diff --git a/src/runtime/x86-64-assem.S b/src/runtime/x86-64-assem.S
index 5f1ff8b..85d5238 100644 (file)
--- a/src/runtime/x86-64-assem.S
+++ b/src/runtime/x86-64-assem.S
@@ -246,7 +246,7 @@ LsingleValue:
        .text
        .globl  GNAME(fpu_save)
        TYPE(GNAME(fpu_save))
-       .align  2,0x90
+       .align  align_16byte,0x90
 GNAME(fpu_save):
        fnsave  (%rdi)          # Save the NPX state. (resets NPX)
        ret
@@ -254,7 +254,7 @@ GNAME(fpu_save):
 
        .globl  GNAME(fpu_restore)
        TYPE(GNAME(fpu_restore))
-       .align  2,0x90
+       .align  align_16byte,0x90
 GNAME(fpu_restore):
        frstor  (%rdi)          # Restore the NPX state.
        ret
@@ -367,6 +367,7 @@ GNAME(fun_end_breakpoint_guts):
 multiple_value_return:
        
        .globl  GNAME(fun_end_breakpoint_trap)
+       .align  align_16byte,0x90
 GNAME(fun_end_breakpoint_trap):
        TRAP
        .byte   trap_FunEndBreakpoint

diff --git a/src/runtime/x86-assem.S b/src/runtime/x86-assem.S
index 4a5090c..315643b 100644 (file)
--- a/src/runtime/x86-assem.S
+++ b/src/runtime/x86-assem.S
@@ -335,7 +335,7 @@ GNAME(fpu_restore):
  * the undefined-function trampoline
  */
        .text
-       .align  align_4byte,0x90
+       .align  align_16byte,0x90
        .globl GNAME(undefined_tramp)
        TYPE(GNAME(undefined_tramp))
         .byte   0, 0, 0, SIMPLE_FUN_HEADER_WIDETAG
@@ -352,7 +352,7 @@ GNAME(undefined_tramp):
  * the closure trampoline
  */
        .text
-       .align  align_4byte,0x90
+       .align  align_16byte,0x90
        .globl GNAME(closure_tramp)
        TYPE(GNAME(closure_tramp))
         .byte   0, 0, 0, SIMPLE_FUN_HEADER_WIDETAG
@@ -368,7 +368,7 @@ GNAME(closure_tramp):
        SIZE(GNAME(closure_tramp))
 
        .text
-       .align  align_4byte,0x90
+       .align  align_16byte,0x90
        .globl GNAME(funcallable_instance_tramp)
        TYPE(GNAME(funcallable_instance_tramp))
 GNAME(funcallable_instance_tramp):
@@ -383,7 +383,7 @@ GNAME(funcallable_instance_tramp):
  */
        .text
        .globl  GNAME(fun_end_breakpoint_guts)
-       .align  align_4byte
+       .align  align_16byte
 GNAME(fun_end_breakpoint_guts):
        /* Multiple Value return */
        jc      multiple_value_return
@@ -410,7 +410,7 @@ GNAME(fun_end_breakpoint_end):
 \f
        .globl  GNAME(do_pending_interrupt)
        TYPE(GNAME(do_pending_interrupt))
-       .align  align_4byte,0x90
+       .align  align_16byte,0x90
 GNAME(do_pending_interrupt):
        TRAP
        .byte   trap_PendingInterrupt
@@ -451,7 +451,7 @@ GNAME(do_pending_interrupt):
 #define DEFINE_ALLOC_TO_EAX(name,size)                          \
         .globl  GNAME(name);                                    \
         TYPE(GNAME(name));                                      \
-        .align  align_4byte,0x90;                               \
+        .align  align_16byte,0x90;                              \
 GNAME(name):                                                    \
         pushl   %ecx;              /* Save ECX and EDX       */ \
         pushl   %edx;                                           \
@@ -464,7 +464,7 @@ GNAME(name):                                                    \
 #define DEFINE_ALLOC_TO_ECX(name,size)                          \
         .globl  GNAME(name);                                    \
         TYPE(GNAME(name));                                      \
-        .align  align_4byte,0x90;                               \
+        .align  align_16byte,0x90;                              \
 GNAME(name):                                                    \
         pushl   %eax;              /* Save EAX and EDX       */ \
         pushl   %edx;                                           \
@@ -478,7 +478,7 @@ GNAME(name):                                                    \
 #define DEFINE_ALLOC_TO_EDX(name,size)                          \
         .globl  GNAME(name);                                    \
         TYPE(GNAME(name));                                      \
-        .align  align_4byte,0x90;                               \
+        .align  align_16byte,0x90;                              \
 GNAME(name):                                                    \
         pushl   %eax;               /* Save EAX and ECX      */ \
         pushl   %ecx;                                           \
@@ -492,7 +492,7 @@ GNAME(name):                                                    \
 #define DEFINE_ALLOC_TO_REG(name,reg,size)                      \
         .globl  GNAME(name);                                    \
         TYPE(GNAME(name));                                      \
-        .align  align_4byte,0x90;                               \
+        .align  align_16byte,0x90;                              \
 GNAME(name):                                                    \
         pushl   %eax;              /* Save EAX, ECX, and EDX */ \
         pushl   %ecx;                                           \
@@ -547,7 +547,7 @@ DEFINE_ALLOC_TO_REG(alloc_16_to_edi,%edi,$16)
 
 /* This routine handles an overflow with eax=crfp+size. So the
    size=eax-crfp. */
-        .align  align_4byte
+        .align  align_16byte
         .globl  GNAME(alloc_overflow_eax)
        TYPE(GNAME(alloc_overflow_eax))
 GNAME(alloc_overflow_eax):
@@ -559,7 +559,7 @@ GNAME(alloc_overflow_eax):
         ret
         SIZE(GNAME(alloc_overflow_eax))
 
-        .align  align_4byte
+        .align  align_16byte
         .globl  GNAME(alloc_overflow_ecx)
        TYPE(GNAME(alloc_overflow_ecx))
 GNAME(alloc_overflow_ecx):
@@ -572,7 +572,7 @@ GNAME(alloc_overflow_ecx):
         ret
         SIZE(GNAME(alloc_overflow_ecx))
 
-        .align  align_4byte
+        .align  align_16byte
         .globl  GNAME(alloc_overflow_edx)
         TYPE(GNAME(alloc_overflow_edx))
 GNAME(alloc_overflow_edx):
@@ -587,7 +587,7 @@ GNAME(alloc_overflow_edx):
 
 /* This routine handles an overflow with ebx=crfp+size. So the
    size=ebx-crfp. */
-        .align  align_4byte
+        .align  align_16byte
         .globl  GNAME(alloc_overflow_ebx)
         TYPE(GNAME(alloc_overflow_ebx))
 GNAME(alloc_overflow_ebx):
@@ -604,7 +604,7 @@ GNAME(alloc_overflow_ebx):
 
 /* This routine handles an overflow with esi=crfp+size. So the
    size=esi-crfp. */
-        .align  align_4byte
+        .align  align_16byte
         .globl  GNAME(alloc_overflow_esi)
         TYPE(GNAME(alloc_overflow_esi))
 GNAME(alloc_overflow_esi):
@@ -619,7 +619,7 @@ GNAME(alloc_overflow_esi):
         ret
         SIZE(GNAME(alloc_overflow_esi))
 
-        .align  align_4byte
+        .align  align_16byte
         .globl  GNAME(alloc_overflow_edi)
         TYPE(GNAME(alloc_overflow_edi))
 GNAME(alloc_overflow_edi):
@@ -642,7 +642,7 @@ GNAME(alloc_overflow_edi):
         * and EIP from the exception context and use it to fake
         * up a stack frame which will skip over the system SEH
         * code. */
-       .align  align_4byte
+       .align  align_16byte
        .globl  GNAME(exception_handler_wrapper)
        TYPE(GNAME(exception_handler_wrapper))
 GNAME(exception_handler_wrapper):
@@ -673,7 +673,7 @@ GNAME(exception_handler_wrapper):
 #endif
 
 #ifdef LISP_FEATURE_DARWIN
-        .align align_4byte
+        .align align_16byte
         .globl GNAME(call_into_lisp_tramp)
         TYPE(GNAME(call_into_lisp_tramp))
 GNAME(call_into_lisp_tramp):
@@ -722,7 +722,7 @@ GNAME(call_into_lisp_tramp):
         SIZE(call_into_lisp_tramp)
 #endif
         
-       .align  align_4byte,0x90
+       .align  align_16byte,0x90
        .globl  GNAME(post_signal_tramp)
        TYPE(GNAME(post_signal_tramp))
 GNAME(post_signal_tramp):
@@ -747,7 +747,7 @@ GNAME(post_signal_tramp):
 \f
        .globl GNAME(fast_bzero_pointer)
        .data
-        .align  align_4byte
+        .align  align_16byte
 GNAME(fast_bzero_pointer):
         /* Variable containing a pointer to the bzero function to use.
          * Initially points to a basic function.  Change this variable
@@ -755,7 +755,7 @@ GNAME(fast_bzero_pointer):
         .long GNAME(fast_bzero_base)
 \f
        .text
-       .align  align_8byte,0x90
+       .align  align_16byte,0x90
        .globl GNAME(fast_bzero)
        TYPE(GNAME(fast_bzero))
 GNAME(fast_bzero):        
@@ -765,7 +765,7 @@ GNAME(fast_bzero):
         
 \f      
        .text
-       .align  align_8byte,0x90
+       .align  align_16byte,0x90
        .globl GNAME(fast_bzero_detect)
        TYPE(GNAME(fast_bzero_detect))
 GNAME(fast_bzero_detect):
@@ -803,7 +803,7 @@ Lrestore:
         
 \f
        .text
-       .align  align_8byte,0x90
+       .align  align_16byte,0x90
        .globl GNAME(fast_bzero_sse)
        TYPE(GNAME(fast_bzero_sse))
         
@@ -849,7 +849,7 @@ Lend_sse:
                 
 \f
        .text
-       .align  align_8byte,0x90
+       .align  align_16byte,0x90
        .globl GNAME(fast_bzero_base)
        TYPE(GNAME(fast_bzero_base))
         

diff --git a/version.lisp-expr b/version.lisp-expr
index 2155293..c97b3e4 100644 (file)
--- a/version.lisp-expr
+++ b/version.lisp-expr
@@ -17,4 +17,4 @@
 ;;; checkins which aren't released. (And occasionally for internal
 ;;; versions, especially for internal versions off the main CVS
 ;;; branch, it gets hairier, e.g. "0.pre7.14.flaky4.13".)
-"1.0.22.14"
+"1.0.22.15"