FreeCalypso > hg > freecalypso-tools
comparison toolchain/binutils-patches/elf32-arm.patch @ 0:e7502631a0f9
initial import from freecalypso-sw rev 1033:5ab737ac3ad7
| author | Mychaela Falconia <falcon@freecalypso.org> |
|---|---|
| date | Sat, 11 Jun 2016 00:13:35 +0000 |
| parents | |
| children |
comparison
equal
deleted
inserted
replaced
| -1:000000000000 | 0:e7502631a0f9 |
|---|---|
| 1 # The present patch to the ARM ELF linking code in binutils (made against | |
| 2 # binutils-2.21.1a) is a hack that affects the generation of Thumb->ARM | |
| 3 # call veneers. The binutils linker can make these veneers in two forms: | |
| 4 # | |
| 5 # "Short" "Long" | |
| 6 # (16-bit) bx pc (16-bit) bx pc | |
| 7 # (16-bit) nop (16-bit) nop | |
| 8 # (32-bit) b dest (32-bit) ldr pc, [pc, #-4] | |
| 9 # (32-bit) .long target | |
| 10 # | |
| 11 # For code that's going to run on the Calypso ARM7 processor, the | |
| 12 # "short" form is the one we want: because of the way the address space | |
| 13 # is laid out, the range of an ARM branch instruction will always be | |
| 14 # sufficient for us, whereas the long version is more costly in terms | |
| 15 # of the number of cycles required, especially when executing from | |
| 16 # external memory. | |
| 17 # | |
| 18 # Unfortunately, the code which decides between these two forms | |
| 19 # (located in the arm_type_of_stub() function in bfd/elf32-arm.c) | |
| 20 # is broken. The original code (which you can see below in the | |
| 21 # "before" part of the patch hunk) exhibits two problems: | |
| 22 # | |
| 23 # 1. The distance that matters here is not from the original | |
| 24 # call point to the destination, but from the stub to | |
| 25 # the destination - and in this function we don't know | |
| 26 # where the stub will end up. | |
| 27 # | |
| 28 # 2. Because it's an ARM branch, the limits should be | |
| 29 # ARM_MAX_[BF]WD_BRANCH_OFFSET, not the THM_ ones. | |
| 30 # | |
| 31 # The 2nd problem would be trivial to fix, but the 1st one | |
| 32 # much less so: this function doesn't know where the stub | |
| 33 # will end up, and considering that for the present purpose | |
| 34 # of the FreeCalypso project always emitting the "short" | |
| 35 # version would be perfect, there is no justification for | |
| 36 # expending the time and effort to restructure the linker | |
| 37 # to do the proper relaxation for the case at hand. | |
| 38 # | |
| 39 # Therefore, the patch we apply is a hack: we simply force the | |
| 40 # use of the "short" form unconditionally. | |
| 41 | |
| 42 *** elf32-arm.c Wed May 11 07:29:12 2011 | |
| 43 --- elf32-arm-patched.c Mon Aug 19 03:24:38 2013 | |
| 44 *************** | |
| 45 *** 3203,3211 **** | |
| 46 : arm_stub_long_branch_v4t_thumb_arm); | |
| 47 | |
| 48 /* Handle v4t short branches. */ | |
| 49 ! if ((stub_type == arm_stub_long_branch_v4t_thumb_arm) | |
| 50 ! && (branch_offset <= THM_MAX_FWD_BRANCH_OFFSET) | |
| 51 ! && (branch_offset >= THM_MAX_BWD_BRANCH_OFFSET)) | |
| 52 stub_type = arm_stub_short_branch_v4t_thumb_arm; | |
| 53 } | |
| 54 } | |
| 55 --- 3203,3209 ---- | |
| 56 : arm_stub_long_branch_v4t_thumb_arm); | |
| 57 | |
| 58 /* Handle v4t short branches. */ | |
| 59 ! if (stub_type == arm_stub_long_branch_v4t_thumb_arm) | |
| 60 stub_type = arm_stub_short_branch_v4t_thumb_arm; | |
| 61 } | |
| 62 } |