Changes from v1:
* Frame pointer backtrace linkage retained. Not that gdb seems to
use this at all; the tcg_register_jit patch is still required in
order to get a proper backtrace.
* Several patches re-ordered in order to reduce churn. Especially
the qemu_ld/st related patches.
* Avoid using ADRP for 32-bit values. It would be exceedingly rare
to save an instruction for this case.
* Don't merge tcg_out_movr{,_sp} with their callers.
I believe I've incorporated all of Claudio's feedback from v1.
r~
Richard Henderson (26):
tcg-aarch64: Properly detect SIGSEGV writes
tcg-aarch64: Use intptr_t apropriately
tcg-aarch64: Use TCGType and TCGMemOp constants
tcg-aarch64: Use MOVN in tcg_out_movi
tcg-aarch64: Use ORRI in tcg_out_movi
tcg-aarch64: Special case small constants in tcg_out_movi
tcg-aarch64: Use adrp in tcg_out_movi
tcg-aarch64: Use symbolic names for branches
tcg-aarch64: Create tcg_out_brcond
tcg-aarch64: Use CBZ and CBNZ
tcg-aarch64: Reuse LR in translated code
tcg-aarch64: Introduce tcg_out_insn_3314
tcg-aarch64: Implement tcg_register_jit
tcg-aarch64: Avoid add with zero in tlb load
tcg-aarch64: Use tcg_out_call for qemu_ld/st
tcg-aarch64: Use ADR to pass the return address to the ld/st helpers
tcg-aarch64: Use TCGMemOp in qemu_ld/st
tcg-aarch64: Pass qemu_ld/st arguments directly
tcg-aarch64: Implement TCG_TARGET_HAS_new_ldst
tcg-aarch64: Support stores of zero
tcg-aarch64: Introduce tcg_out_insn_3507
tcg-aarch64: Merge aarch64_ldst_get_data/type into tcg_out_op
tcg-aarch64: Replace aarch64_ldst_op_data with TCGMemOp
tcg-aarch64: Replace aarch64_ldst_op_data with AArch64LdstType
tcg-aarch64: Prefer unsigned offsets before signed offsets for ldst
tcg-aarch64: Use tcg_out_mov in preference to tcg_out_movr
tcg/aarch64/tcg-target.c | 1076 ++++++++++++++++++++++++----------------------
tcg/aarch64/tcg-target.h | 34 +-
user-exec.c | 29 +-
3 files changed, 593 insertions(+), 546 deletions(-)
--
1.9.0
