From c33402e25563e8c80a650e1062527db4c5865bbc Mon Sep 17 00:00:00 2001 From: Franz Sirl Date: Wed, 29 Jan 2003 23:53:54 +0000 Subject: ppc_closure.S: Recode to fit shared libs. 2003-01-29 Franz Sirl * src/powerpc/ppc_closure.S: Recode to fit shared libs. From-SVN: r62112 --- libffi/src/powerpc/ppc_closure.S | 194 ++++++++++++++++++++++++--------------- 1 file changed, 121 insertions(+), 73 deletions(-) (limited to 'libffi/src') diff --git a/libffi/src/powerpc/ppc_closure.S b/libffi/src/powerpc/ppc_closure.S index a9ea9c7..e402fb5 100644 --- a/libffi/src/powerpc/ppc_closure.S +++ b/libffi/src/powerpc/ppc_closure.S @@ -1,14 +1,13 @@ #define LIBFFI_ASM #include -.globl ffi_closure_helper_SYSV + .file "ppc_closure.S" ENTRY(ffi_closure_SYSV) .LFB1: stwu %r1,-144(%r1) .LCFI0: mflr %r0 - stw %r31,140(%r1) .LCFI1: stw %r0,148(%r1) @@ -63,87 +62,136 @@ ENTRY(ffi_closure_SYSV) # look up the proper starting point in table # by using return type as offset addi %r5,%r1,112 # get pointer to results area - addis %r4,0,.L60@ha # get address of jump table - addi %r4,%r4,.L60@l - slwi %r3,%r3,2 # now multiply return type by 4 - lwzx %r3,%r4,%r3 # get the contents of that table value - add %r3,%r3,%r4 # add contents of table to table address + bl .Lget_ret_type0_addr # get pointer to .Lret_type0 into LR + mflr %r4 # move to r4 + slwi %r3,%r3,4 # now multiply return type by 16 + add %r3,%r3,%r4 # add contents of table to table address mtctr %r3 bctr # jump to it .LFE1: - .align 2 -.L60: - .long .L44-.L60 # FFI_TYPE_VOID - .long .L50-.L60 # FFI_TYPE_INT - .long .L47-.L60 # FFI_TYPE_FLOAT - .long .L46-.L60 # FFI_TYPE_DOUBLE - .long .L46-.L60 # FFI_TYPE_LONGDOUBLE - .long .L56-.L60 # FFI_TYPE_UINT8 - .long .L55-.L60 # FFI_TYPE_SINT8 - .long .L58-.L60 # FFI_TYPE_UINT16 - .long .L57-.L60 # FFI_TYPE_SINT16 - .long .L50-.L60 # FFI_TYPE_UINT32 - .long .L50-.L60 # FFI_TYPE_SINT32 - .long .L48-.L60 # FFI_TYPE_UINT64 - .long .L48-.L60 # FFI_TYPE_SINT64 - .long .L44-.L60 # FFI_TYPE_STRUCT - .long .L50-.L60 # FFI_TYPE_POINTER - - -# case double -.L46: - lfd %f1,0(%r5) - b .L44 -# case float -.L47: +# Each of the ret_typeX code fragments has to be exactly 16 bytes long +# (4 instructions). For cache effectiveness we align to a 16 byte boundary +# first. + .align 4 + + nop + nop + nop +.Lget_ret_type0_addr: + blrl + +# case FFI_TYPE_VOID +.Lret_type0: + b .Lfinish + nop + nop + nop + +# case FFI_TYPE_INT +.Lret_type1: + lwz %r3,0(%r5) + b .Lfinish + nop + nop + +# case FFI_TYPE_FLOAT +.Lret_type2: lfs %f1,0(%r5) - b .L44 - -# case long long -.L48: + b .Lfinish + nop + nop + +# case FFI_TYPE_DOUBLE +.Lret_type3: + lfd %f1,0(%r5) + b .Lfinish + nop + nop + +# case FFI_TYPE_LONGDOUBLE +.Lret_type4: + lfd %f1,0(%r5) + b .Lfinish + nop + nop + +# case FFI_TYPE_UINT8 +.Lret_type5: + lbz %r3,3(%r5) + b .Lfinish + nop + nop + +# case FFI_TYPE_SINT8 +.Lret_type6: + lbz %r3,3(%r5) + extsb %r3,%r3 + b .Lfinish + nop + +# case FFI_TYPE_UINT16 +.Lret_type7: + lhz %r3,2(%r5) + b .Lfinish + nop + nop + +# case FFI_TYPE_SINT16 +.Lret_type8: + lha %r3,2(%r5) + b .Lfinish + nop + nop + +# case FFI_TYPE_UINT32 +.Lret_type9: + lwz %r3,0(%r5) + b .Lfinish + nop + nop + +# case FFI_TYPE_SINT32 +.Lret_type10: + lwz %r3,0(%r5) + b .Lfinish + nop + nop + +# case FFI_TYPE_UINT64 +.Lret_type11: lwz %r3,0(%r5) lwz %r4,4(%r5) - b .L44 - -# case default / int32 / pointer -.L50: + b .Lfinish + nop + +# case FFI_TYPE_SINT64 +.Lret_type12: lwz %r3,0(%r5) - b .L44 - -# case signed int8 -.L55: - addi %r5,%r5,3 - lbz %r3,0(%r5) - extsb %r3,%r3 - b .L44 - -# case unsigned int8 -.L56: - addi %r5,%r5,3 - lbz %r3,0(%r5) - b .L44 - -# case signed int16 -.L57: - addi %r5,%r5,2 - lhz %r3,0(%r5) - extsh %r3,%r3 - b .L44 - -#case unsigned int16 -.L58: - addi %r5,%r5,2 - lhz %r3,0(%r5) - -# case void / done -.L44: + lwz %r4,4(%r5) + b .Lfinish + nop + +# case FFI_TYPE_STRUCT +.Lret_type13: + b .Lfinish + nop + nop + nop + +# case FFI_TYPE_POINTER +.Lret_type14: + lwz %r3,0(%r5) + b .Lfinish + nop + nop + +# case done +.Lfinish: - lwz %r11,0(%r1) - lwz %r0,4(%r11) + lwz %r0,148(%r1) mtlr %r0 - lwz %r31,-4(%r11) - mr %r1,%r11 + addi %r1,%r1,144 blr END(ffi_closure_SYSV) -- cgit v1.1