* Removed all Id tags.
[cacao.git] / src / vm / jit / x86_64 / asmpart.S
index 19537200e4befe31aec2acbfec165a651fc55f7f..58ab38124a894b21f02c4bb48d912dfc5a899660 100644 (file)
@@ -1,6 +1,6 @@
 /* src/vm/jit/x86_64/asmpart.S - Java-C interface functions for x86_64
 
-   Copyright (C) 1996-2005, 2006 R. Grafl, A. Krall, C. Kruegel,
+   Copyright (C) 1996-2005, 2006, 2007 R. Grafl, A. Krall, C. Kruegel,
    C. Oates, R. Obermaisser, M. Platter, M. Probst, S. Ring,
    E. Steiner, C. Thalinger, D. Thuernbeck, P. Tomsich, C. Ullrich,
    J. Wenninger, Institut f. Computersprachen - TU Wien
    Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
    02110-1301, USA.
 
-   Contact: cacao@cacaojvm.org
-
-   Authors: Andreas Krall
-            Reinhard Grafl
-            Christian Thalinger
-
-   Changes: Edwin Steiner
-
-   $Id: asmpart.S 5110 2006-07-11 22:23:55Z twisti $
-
 */
 
 
@@ -40,7 +30,6 @@
 #include "vm/jit/x86_64/arch.h"
 #include "vm/jit/x86_64/md-abi.h"
 #include "vm/jit/x86_64/md-asm.h"
-#include "vm/jit/x86_64/offsets.h"
 
 #include "vm/jit/abi-asm.h"
 #include "vm/jit/methodheader.h"
@@ -57,6 +46,7 @@
        .globl asm_vm_call_method_float
        .globl asm_vm_call_method_double
        .globl asm_vm_call_method_exception_handler
+       .globl asm_vm_call_method_end
 
        .globl asm_call_jit_compiler
 
 
        .globl asm_abstractmethoderror
 
-       .globl asm_wrapper_patcher
+       .globl asm_patcher_wrapper
 
+#if defined(ENABLE_REPLACEMENT)
        .globl asm_replacement_out
        .globl asm_replacement_in
+#endif
 
        .globl asm_builtin_f2i
        .globl asm_builtin_f2l
        .globl asm_builtin_d2i
        .globl asm_builtin_d2l
 
-       .globl asm_criticalsections
-       .globl asm_getclassvalues_atomic
+       .globl asm_compare_and_swap
+       .globl asm_memory_barrier
 
 
 /********************* function asm_calljavafunction ***************************
@@ -124,96 +116,59 @@ asm_vm_call_method_double:
        mov     s3,4*8(sp)
        mov     s4,5*8(sp)
 
-       mov     a0,itmp1                    /* move method pointer for compiler   */
-       xor     %rbp,%rbp                   /* set argument stack frame to zero   */
+       mov     a0,6*8(sp)                  /* store method PV                    */
 
-       test    a1,a1                       /* maybe we have no args...           */
-       jle     L_copy_done
+       mov     sp,s0                       /* save stack pointer                 */
 
-       mov     a1,itmp3                    /* arg count                          */
-       mov     a2,itmp2                    /* pointer to arg block               */
+       mov     a1,t0                       /* address of data structure          */
+       mov     a2,itmp1                    /* number of stack arguments          */
 
-       mov     itmp2,%r14                  /* save argument block pointer        */
-       mov     itmp3,%r15                  /* save argument count                */
+       mov     0*8(t0),a0
+       mov     1*8(t0),a1
+       mov     2*8(t0),a2
+       mov     3*8(t0),a3
+       mov     4*8(t0),a4
+       mov     5*8(t0),a5
 
-       sub     $sizevmarg,itmp2            /* initialize pointer (smaller code)  */
-       add     $1,itmp3                    /* initialize argument count          */
-       xor     %r12,%r12                   /* initialize integer argument counter*/
-       xor     %r13,%r13                   /* initialize float argument counter  */
+       movq    6*8(t0),fa0
+       movq    7*8(t0),fa1
+       movq    8*8(t0),fa2
+       movq    9*8(t0),fa3
+       movq    10*8(t0),fa4
+       movq    11*8(t0),fa5
+       movq    12*8(t0),fa6
+       movq    13*8(t0),fa7
 
-L_register_copy:
-       add     $sizevmarg,itmp2            /* goto next argument block           */
-       dec     itmp3                       /* argument count - 1                 */
-       jz      L_register_copy_done
-       andb    $0x02,offvmargtype(itmp2)   /* is this a float/double type?       */
-       jnz     L_register_handle_float     /* yes, handle it                     */
+       cmp     $0,itmp1l
+       je      L_asm_vm_call_method_stack_copy_done
 
-       cmp     $INT_ARG_CNT,%r12           /* are we out of integer argument     */
-       je      L_register_copy             /* register? yes, next loop           */
+       mov     itmp1,itmp2
+       add     $1,itmp2                    /* keep stack 16-byte aligned         */
+       and     $0xfffffffffffffffe,itmp2
+       shl     $3,itmp2                    /* calculate stack size               */
+       sub     itmp2,sp                    /* create stack frame                 */
+       mov     sp,itmp2                    /* temporary stack pointer            */
 
-       lea     jumptable_integer(%rip),%rbp
-       mov     0(%rbp,%r12,8),%rbx
-       inc     %r12                      /* integer argument counter + 1         */
-       jmp     *%rbx
+L_asm_vm_call_method_stack_copy_loop:
+       mov     14*8(t0),itmp3              /* load argument                      */
+       mov     itmp3,0(itmp2)              /* store argument on stack            */
 
-L_register_handle_float:
-       cmp     $FLT_ARG_CNT,%r13         /* are we out of float argument         */
-       je      L_register_copy           /* register? yes, next loop             */
+       sub     $1,itmp1l                   /* subtract 1 argument                */
+       add     $8,t0                       /* set address of next argument       */
+       add     $8,itmp2                    /* increase SP                        */
 
-       lea     jumptable_float(%rip),%rbp
-       mov     0(%rbp,%r13,8),%rbx
-       inc     %r13                      /* float argument counter + 1           */
-       jmp     *%rbx
-       
-L_register_copy_done:
-       mov     %r15,%rbp                 /* calculate remaining arguments        */
-       sub     %r12,%rbp                 /* - integer arguments in registers     */
-       sub     %r13,%rbp                 /* - float arguments in registers       */
-       jle     L_copy_done               /* are all assigned to registers?       */
-
-       and     $0xfffffffffffffffe,%rbp  /* keep stack 16-byte aligned           */
-       shl     $3,%rbp                   /* calculate stack size                 */
-       sub     %rbp,sp                   /* stack frame for arguments            */
-       mov     sp,%rbx                   /* use %rbx as temp sp                  */
-
-       sub     $sizevmarg,%r14           /* initialize pointer (smaller code)    */
-       add     $1,%r15                   /* initialize argument count            */
-               
-L_stack_copy_loop:
-       add     $sizevmarg,%r14           /* goto next argument block             */
-       dec     %r15                      /* are there any arguments left?        */
-       jz      L_copy_done               /* no test needed after dec             */
-
-       andb    $0x02,offvmargtype(%r14)    /* is this a float/double type?       */
-       jnz     L_stack_handle_float
-       dec     %r12                        /* arguments assigned to registers    */
-       jge     L_stack_copy_loop
-       jmp     L_stack_copy
-
-L_stack_handle_float:
-       dec     %r13                        /* arguments assigned to registers    */
-       jge     L_stack_copy_loop
-
-L_stack_copy:
-       mov     offvmargdata(%r14),itmp3    /* copy s8 argument onto stack        */
-       mov     itmp3,0(%rbx)
-       add     $8,%rbx                     /* increase sp to next argument       */
-       jmp     L_stack_copy_loop
-
-L_copy_done:
-                                           /* itmp1 still contains method pointer*/
-       lea     L_asm_call_jit_compiler(%rip),mptr
-       mov     sp,itmp3                    /* calculate the old stack pointer    */
-       add     bp,itmp3
-       mov     mptr,6*8(itmp3)
-       lea     (6*8-256)(itmp3),mptr       /* We subtract 256 to force the next  */
+       cmp     $0,itmp1l
+       jg      L_asm_vm_call_method_stack_copy_loop
+
+L_asm_vm_call_method_stack_copy_done:
+       lea     (6*8-256)(s0),mptr          /* We subtract 256 to force the next  */
                                            /* move instruction to have a 32-bit  */
                                            /* offset.                            */
 
-       mov     (0*8+256)(mptr),itmp3       /* method call as in Java             */
-       call    *itmp3                      /* call JIT compiler                  */
+       mov     (0*8+256)(mptr),itmp3       /* load PV                            */
+       call    *itmp3
 
-       add     bp,sp                       /* remove argument stack frame if any */
+       mov     s0,sp                       /* restore SP                         */
 
 L_asm_vm_call_method_return:
        mov     0*8(sp),%rbx                /* restore callee saved registers     */
@@ -230,69 +185,8 @@ asm_vm_call_method_exception_handler:
        call    builtin_throw_exception@PLT
        jmp     L_asm_vm_call_method_return
 
-
-jumptable_integer:
-       .quad   handle_a0
-       .quad   handle_a1
-       .quad   handle_a2
-       .quad   handle_a3
-       .quad   handle_a4
-       .quad   handle_a5
-
-handle_a0:
-       mov     offvmargdata(itmp2),a0
-       jmp     L_register_copy
-handle_a1:
-       mov     offvmargdata(itmp2),a1
-       jmp     L_register_copy
-handle_a2:
-       mov     offvmargdata(itmp2),a2
-       jmp     L_register_copy
-handle_a3:
-       mov     offvmargdata(itmp2),a3
-       jmp     L_register_copy
-handle_a4:
-       mov     offvmargdata(itmp2),a4
-       jmp     L_register_copy
-handle_a5:
-       mov     offvmargdata(itmp2),a5
-       jmp     L_register_copy
-
-
-jumptable_float:
-       .quad   handle_fa0
-       .quad   handle_fa1
-       .quad   handle_fa2
-       .quad   handle_fa3
-       .quad   handle_fa4
-       .quad   handle_fa5
-       .quad   handle_fa6
-       .quad   handle_fa7
-
-handle_fa0:
-       movq    offvmargdata(itmp2),fa0
-       jmp     L_register_copy
-handle_fa1:
-       movq    offvmargdata(itmp2),fa1
-       jmp     L_register_copy
-handle_fa2:
-       movq    offvmargdata(itmp2),fa2
-       jmp     L_register_copy
-handle_fa3:
-       movq    offvmargdata(itmp2),fa3
-       jmp     L_register_copy
-handle_fa4:
-       movq    offvmargdata(itmp2),fa4
-       jmp     L_register_copy
-handle_fa5:
-       movq    offvmargdata(itmp2),fa5
-       jmp     L_register_copy
-handle_fa6:
-       movq    offvmargdata(itmp2),fa6
-       jmp     L_register_copy
-handle_fa7:
-       movq    offvmargdata(itmp2),fa7
-       jmp     L_register_copy
+asm_vm_call_method_end:
+       nop
 
 
 /****************** function asm_call_jit_compiler *****************************
@@ -384,7 +278,7 @@ L_asm_handle_exception_stack_loop:
        mov     t0,4*8(sp)                  /* save maybe-leaf flag               */
 
        mov     xpc,a0                      /* exception pc                       */
-       call    codegen_findmethod@PLT
+       call    codegen_get_pv_from_pc@PLT
        mov     v0,2*8(sp)                  /* save data segment pointer          */
         
        mov     0*8(sp),a0                  /* pass exception pointer             */
@@ -511,7 +405,7 @@ asm_abstractmethoderror:
        jmp     L_asm_handle_exception
 
 
-/* asm_wrapper_patcher *********************************************************
+/* asm_patcher_wrapper *********************************************************
 
    XXX
 
@@ -526,10 +420,10 @@ asm_abstractmethoderror:
 
 *******************************************************************************/
 
-asm_wrapper_patcher:
+asm_patcher_wrapper:
        push    bp                          /* save base pointer                  */
        mov     sp,bp                       /* move actual sp to bp               */
-       sub     $((3+ARG_CNT+TMP_CNT)*8+sizestackframeinfo),sp
+       sub     $(3+ARG_CNT+TMP_CNT)*8,sp
        and     $0xfffffffffffffff0,sp      /* align sp to 16-byte (this is for   */
                                            /* leaf functions)                    */
 
@@ -539,25 +433,13 @@ asm_wrapper_patcher:
        mov     itmp1,0*8(sp)               /* save itmp1 and itmp2               */
        mov     itmp2,1*8(sp)               /* can be used by some instructions   */
 
-       mov     sp,a0                       /* create stackframe info             */
-       add     $((3+ARG_CNT+TMP_CNT)*8),a0
-       xor     a1,a1                       /* if pv is NULL, use findmethod      */
-       mov     bp,a2                       /* pass java sp                       */
-       add     $((6+1)*8),a2
-       mov     ((5+1)*8)(bp),a3            /* pass ra to java function           */
-       mov     a3,a4                       /* xpc is equal to ra                 */
-       call    stacktrace_create_extern_stackframeinfo@PLT
-
-       mov     bp,a0                       /* pass stack pointer                 */
-       add     $((1+1)*8),a0               /* skip function pointer              */
-       mov     1*8(bp),itmp3               /* get function pointer               */
-       call    *itmp3                      /* call the patcher function          */
+       mov     bp,a0                       /* pass SP of patcher stub            */
+       add     $(1*8),a0
+       mov     $0,a1                       /* pass PV (if NULL, use findmethod)  */
+       mov     $0,a2                       /* pass RA (it's on the stack)        */
+       call    patcher_wrapper@PLT
        mov     v0,2*8(sp)                  /* save return value                  */
 
-       mov     sp,a0                       /* remove stackframe info             */
-       add     $((3+ARG_CNT+TMP_CNT)*8),a0
-       call    stacktrace_remove_stackframeinfo@PLT
-
        RESTORE_ARGUMENT_REGISTERS(3)
        RESTORE_TEMPORARY_REGISTERS(3+ARG_CNT)
 
@@ -567,17 +449,18 @@ asm_wrapper_patcher:
 
        mov     bp,sp                       /* restore original sp                */
        pop     bp                          /* restore bp                         */
-       add     $(5*8),sp                   /* remove patcher stackframe, keep ra */
+       add     $(5*8),sp                   /* remove patcher stackframe, keep RA */
 
        test    itmp3,itmp3                 /* exception thrown?                  */
-       jz      L_asm_wrapper_patcher_exception
+       jne     L_asm_patcher_wrapper_exception
        ret                                 /* call new patched code              */
 
-L_asm_wrapper_patcher_exception:
-       call    exceptions_get_and_clear_exception@PLT
+L_asm_patcher_wrapper_exception:
+       mov     itmp3,xptr                  /* get exception                      */
        pop     xpc                         /* get and remove return address      */
        jmp     L_asm_handle_exception
 
+#if defined(ENABLE_REPLACEMENT)
 
 /* asm_replacement_out *********************************************************
 
@@ -659,57 +542,78 @@ asm_replacement_out:
    This function never returns!
 
    C prototype:
-      void asm_replacement_in(executionstate *es);
+      void asm_replacement_in(executionstate *es, replace_safestack_t *st);
 
 *******************************************************************************/
 
 asm_replacement_in:
-       mov     a0,%rbp                     /* executionstate *es                 */
+       /* get arguments */
+       mov     a1,s1                       /* replace_safestack_t *st            */
+       mov     a0,%rbp                     /* executionstate *es == safe stack   */
+
+       /* switch to the safe stack */
+       mov     %rbp,sp
+
+       /* call replace_build_execution_state(st) */
+       mov             s1,a0
+       call    replace_build_execution_state@PLT
 
        /* set new sp */
-       mov     (offes_sp)(%rbp),%rsp
-       
-       /* store address of new code */
-       push    (offes_pc)(%rbp)
-       
+       mov     (offes_sp)(%rbp),sp
+
+       /* push address of new code */
+       pushq   (offes_pc)(%rbp)
+
+       /* allocate an executionstate_t on the stack */
+       sub             $(sizeexecutionstate),sp
+
+       /* call replace_free_safestack(st,& of allocated executionstate_t) */
+       mov             sp,a1
+       mov             s1,a0
+       call    replace_free_safestack@PLT
+
        /* copy registers from execution state */
-       movq    (XMM0 *8+offes_fltregs)(%rbp),%xmm0
-       movq    (XMM1 *8+offes_fltregs)(%rbp),%xmm1
-       movq    (XMM2 *8+offes_fltregs)(%rbp),%xmm2
-       movq    (XMM3 *8+offes_fltregs)(%rbp),%xmm3
-       movq    (XMM4 *8+offes_fltregs)(%rbp),%xmm4
-       movq    (XMM5 *8+offes_fltregs)(%rbp),%xmm5
-       movq    (XMM6 *8+offes_fltregs)(%rbp),%xmm6
-       movq    (XMM7 *8+offes_fltregs)(%rbp),%xmm7
-       movq    (XMM8 *8+offes_fltregs)(%rbp),%xmm8
-       movq    (XMM9 *8+offes_fltregs)(%rbp),%xmm9
-       movq    (XMM10*8+offes_fltregs)(%rbp),%xmm10
-       movq    (XMM11*8+offes_fltregs)(%rbp),%xmm11
-       movq    (XMM12*8+offes_fltregs)(%rbp),%xmm12
-       movq    (XMM13*8+offes_fltregs)(%rbp),%xmm13
-       movq    (XMM14*8+offes_fltregs)(%rbp),%xmm14
-       movq    (XMM15*8+offes_fltregs)(%rbp),%xmm15
-
-       mov     (RAX*8+offes_intregs)(%rbp),%rax
-       mov     (RBX*8+offes_intregs)(%rbp),%rbx
-       mov     (RCX*8+offes_intregs)(%rbp),%rcx
-       mov     (RDX*8+offes_intregs)(%rbp),%rdx
-       mov     (RSI*8+offes_intregs)(%rbp),%rsi
-       mov     (RDI*8+offes_intregs)(%rbp),%rdi
-       mov     (R8 *8+offes_intregs)(%rbp),%r8
-       mov     (R9 *8+offes_intregs)(%rbp),%r9
-       mov     (R10*8+offes_intregs)(%rbp),%r10
-       mov     (R11*8+offes_intregs)(%rbp),%r11
-       mov     (R12*8+offes_intregs)(%rbp),%r12
-       mov     (R13*8+offes_intregs)(%rbp),%r13
-       mov     (R14*8+offes_intregs)(%rbp),%r14
-       mov     (R15*8+offes_intregs)(%rbp),%r15
-
-       mov     (RBP*8+offes_intregs)(%rbp),%rbp
+       movq    (XMM0 *8+offes_fltregs)(sp),%xmm0
+       movq    (XMM1 *8+offes_fltregs)(sp),%xmm1
+       movq    (XMM2 *8+offes_fltregs)(sp),%xmm2
+       movq    (XMM3 *8+offes_fltregs)(sp),%xmm3
+       movq    (XMM4 *8+offes_fltregs)(sp),%xmm4
+       movq    (XMM5 *8+offes_fltregs)(sp),%xmm5
+       movq    (XMM6 *8+offes_fltregs)(sp),%xmm6
+       movq    (XMM7 *8+offes_fltregs)(sp),%xmm7
+       movq    (XMM8 *8+offes_fltregs)(sp),%xmm8
+       movq    (XMM9 *8+offes_fltregs)(sp),%xmm9
+       movq    (XMM10*8+offes_fltregs)(sp),%xmm10
+       movq    (XMM11*8+offes_fltregs)(sp),%xmm11
+       movq    (XMM12*8+offes_fltregs)(sp),%xmm12
+       movq    (XMM13*8+offes_fltregs)(sp),%xmm13
+       movq    (XMM14*8+offes_fltregs)(sp),%xmm14
+       movq    (XMM15*8+offes_fltregs)(sp),%xmm15
+
+       mov     (RAX*8+offes_intregs)(sp),%rax
+       mov     (RBX*8+offes_intregs)(sp),%rbx
+       mov     (RCX*8+offes_intregs)(sp),%rcx
+       mov     (RDX*8+offes_intregs)(sp),%rdx
+       mov     (RSI*8+offes_intregs)(sp),%rsi
+       mov     (RDI*8+offes_intregs)(sp),%rdi
+       mov     (RBP*8+offes_intregs)(sp),%rbp
+       mov     (R8 *8+offes_intregs)(sp),%r8
+       mov     (R9 *8+offes_intregs)(sp),%r9
+       mov     (R10*8+offes_intregs)(sp),%r10
+       mov     (R11*8+offes_intregs)(sp),%r11
+       mov     (R12*8+offes_intregs)(sp),%r12
+       mov     (R13*8+offes_intregs)(sp),%r13
+       mov     (R14*8+offes_intregs)(sp),%r14
+       mov     (R15*8+offes_intregs)(sp),%r15
+
+       /* pop the execution state off the stack */
+       add             $(sizeexecutionstate),sp
 
        /* jump to new code */
        ret
 
+#endif /* defined(ENABLE_REPLACEMENT) */
+
 
 /* asm_builtin_x2x *************************************************************
 *                                                                              *
@@ -773,33 +677,34 @@ asm_builtin_d2l:
        ret
 
 
-asm_getclassvalues_atomic:
-_crit_restart:
-_crit_begin:
-       movl    offbaseval(a0),itmp1l
-       movl    offdiffval(a0),itmp2l
-       movl    offbaseval(a1),itmp3l
-_crit_end:
-       movl    itmp1l,offcast_super_baseval(a2)
-       movl    itmp2l,offcast_super_diffval(a2)
-       movl    itmp3l,offcast_sub_baseval(a2)
+/* asm_compare_and_swap ********************************************************
+
+   Does an atomic compare and swap.  Required for the lock
+   implementation.
+
+*******************************************************************************/
+
+asm_compare_and_swap:
+       mov     a1,v0                       /* v0 is %rax                         */
+       lock cmpxchg a2,(a0)
        ret
 
-       .data
-               
-asm_criticalsections:
-#if defined(ENABLE_THREADS)
-       .quad   _crit_begin
-       .quad   _crit_end
-       .quad   _crit_restart
-#endif
-       .quad 0
+
+/* asm_memory_barrier **********************************************************
+
+   A memory barrier for the Java Memory Model.
+
+*******************************************************************************/
+
+asm_memory_barrier:
+       mfence
+       ret
 
 
-/* Disable exec-stacks, required for Gentoo ***********************************/
+/* disable exec-stacks ********************************************************/
 
-#if defined(__GCC__) && defined(__ELF__)
-       .section .note.GNU-stack,"",@progbits
+#if defined(__linux__) && defined(__ELF__)
+       .section .note.GNU-stack,"",%progbits
 #endif