[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Qemu-devel] [PATCH for-4.1 00/35] tcg: Move the softmmu tlb to CPUNegat
|
From: |
Richard Henderson |
|
Subject: |
[Qemu-devel] [PATCH for-4.1 00/35] tcg: Move the softmmu tlb to CPUNegativeOffsetState |
|
Date: |
Sat, 23 Mar 2019 12:08:50 -0700 |
This started merely as an attempt to reduce the size of each
softmmu lookup by using smaller offsets from env. But in the
end it also represents a significant cleanup in the boilerplate
that each target must define.
With respect to the initial goal, here are the relevant code
snips generated for loading the mask & table fields for a
qemu_ld from an aarch64 guest on the indicated host.
BEFORE:
x86_64:
0x7f9698c5f73b: 48 23 bd 98 32 00 00 andq 0x3298(%rbp), %rdi
0x7f9698c5f742: 48 03 bd d8 32 00 00 addq 0x32d8(%rbp), %rdi
aarch64:
0xffff9e001e28: 91400e61 add x1, x19, #3, lsl #12
0xffff9e001e2c: f9414c20 ldr x0, [x1, #0x298]
0xffff9e001e30: f9416c21 ldr x1, [x1, #0x2d8]
aarch32:
0xa2b7f0d4: e2862a03 add r2, r6, #0x3000
0xa2b7f0d8: e592c20c ldr ip, [r2, #0x20c]
0xa2b7f0dc: e592222c ldr r2, [r2, #0x22c]
AFTER:
x86_64:
0x7fa40a000154: 48 23 7d e0 andq -0x20(%rbp), %rdi
0x7fa40a000158: 48 03 7d e8 addq -0x18(%rbp), %rdi
aarch64:
0xffffa20001b4: a97e0660 ldp x0, x1, [x19, #-0x20]
aarch32:
0xa2c7f0d4: e14604d8 ldrd r0, r1, [r6, #-0x48]
The other tcg hosts do not see as significant difference. PPC and
mips have 16-bit signed offsets, and have no load-pair/multiple.
S390x has 20-bit signed offsets and, like x86, uses a read-operate
instruction form. Sparc and RISC-V have 13 and 12-bit signed offsets
respectively, and so do avoid an extra add insn in this case, but
do not have load-pair/multiple.
All that said, in the end I'm most happy with the diffstat result.
r~
Richard Henderson (35):
tcg: Fold CPUTLBWindow into CPUTLBDesc
tcg: Split out target/arch/cpu-param.h
tcg: Create struct CPUTLB
cpu: Define CPUArchState with typedef
cpu: Define ArchCPU
cpu: Replace ENV_GET_CPU with env_cpu
cpu: Introduce env_archcpu
target/alpha: Use env_cpu, env_archcpu
target/arm: Use env_cpu, env_archcpu
target/cris: Use env_cpu, env_archcpu
target/hppa: Use env_cpu, env_archcpu
target/i386: Use env_cpu, env_archcpu
target/lm32: Use env_cpu, env_archcpu
target/m68k: Use env_cpu, env_archcpu
target/microblaze: Use env_cpu, env_archcpu
target/mips: Use env_cpu, env_archcpu
target/moxie: Use env_cpu, env_archcpu
target/nios2: Use env_cpu, env_archcpu
target/openrisc: Use env_cpu, env_archcpu
target/ppc: Use env_cpu, env_archcpu
target/riscv: Use env_cpu, env_archcpu
target/s390x: Use env_cpu, env_archcpu
target/sh4: Use env_cpu, env_archcpu
target/sparc: Use env_cpu, env_archcpu
target/tilegx: Use env_cpu
target/tricore: Use env_cpu
target/unicore32: Use env_cpu, env_archcpu
target/xtensa: Use env_cpu, env_archcpu
cpu: Move ENV_OFFSET to exec/gen-icount.h
cpu: Introduce CPUNegativeOffsetState
cpu: Move icount_decr to CPUNegativeOffsetState
cpu: Move the softmmu tlb to CPUNegativeOffsetState
cpu: Remove CPU_COMMON
tcg/aarch64: Use LDP to load tlb mask+table
tcg/arm: Use LDRD to load tlb mask+table
Makefile.target | 1 +
accel/tcg/atomic_template.h | 8 +-
accel/tcg/softmmu_template.h | 24 +-
include/exec/cpu-all.h | 46 +++
include/exec/cpu-defs.h | 113 ++++--
include/exec/cpu_ldst.h | 6 +-
include/exec/cpu_ldst_template.h | 6 +-
include/exec/cpu_ldst_useronly_template.h | 6 +-
include/exec/gen-icount.h | 14 +-
include/exec/softmmu-semi.h | 16 +-
include/qom/cpu.h | 28 +-
linux-user/cpu_loop-common.h | 2 +-
linux-user/m68k/target_cpu.h | 2 +-
target/alpha/cpu-param.h | 19 +
target/alpha/cpu.h | 40 +-
target/arm/cpu-param.h | 22 ++
target/arm/cpu.h | 52 +--
target/cris/cpu-param.h | 5 +
target/cris/cpu.h | 25 +-
target/hppa/cpu-param.h | 22 ++
target/hppa/cpu.h | 38 +-
target/i386/cpu-param.h | 14 +
target/i386/cpu.h | 40 +-
target/lm32/cpu-param.h | 5 +
target/lm32/cpu.h | 25 +-
target/m68k/cpu-param.h | 9 +
target/m68k/cpu.h | 28 +-
target/microblaze/cpu-param.h | 6 +
target/microblaze/cpu.h | 63 ++--
target/mips/cpu-param.h | 18 +
target/mips/cpu.h | 21 +-
target/mips/mips-defs.h | 15 -
target/moxie/cpu-param.h | 5 +
target/moxie/cpu.h | 29 +-
target/nios2/cpu-param.h | 9 +
target/nios2/cpu.h | 33 +-
target/openrisc/cpu-param.h | 5 +
target/openrisc/cpu.h | 31 +-
target/ppc/cpu-param.h | 25 ++
target/ppc/cpu.h | 54 +--
target/ppc/helper_regs.h | 4 +-
target/riscv/cpu-param.h | 11 +
target/riscv/cpu.h | 35 +-
target/s390x/cpu-param.h | 5 +
target/s390x/cpu.h | 31 +-
target/sh4/cpu-param.h | 9 +
target/sh4/cpu.h | 30 +-
target/sparc/cpu-param.h | 17 +
target/sparc/cpu.h | 36 +-
target/tilegx/cpu-param.h | 5 +
target/tilegx/cpu.h | 23 +-
target/tricore/cpu-param.h | 5 +
target/tricore/cpu.h | 22 +-
target/tricore/tricore-defs.h | 5 -
target/unicore32/cpu-param.h | 5 +
target/unicore32/cpu.h | 24 +-
target/xtensa/cpu-param.h | 9 +
target/xtensa/cpu.h | 40 +-
accel/tcg/cpu-exec.c | 23 +-
accel/tcg/cputlb.c | 193 +++++-----
accel/tcg/tcg-all.c | 6 +-
accel/tcg/tcg-runtime.c | 4 +-
accel/tcg/translate-all.c | 10 +-
accel/tcg/user-exec.c | 2 +-
bsd-user/main.c | 5 +-
bsd-user/syscall.c | 6 +-
cpus.c | 9 +-
hw/i386/kvmvapic.c | 4 +-
hw/i386/pc.c | 2 +-
hw/intc/mips_gic.c | 2 +-
hw/mips/mips_int.c | 2 +-
hw/nios2/cpu_pic.c | 5 +-
hw/ppc/ppc.c | 18 +-
hw/ppc/ppc405_uc.c | 2 +-
hw/ppc/ppc_booke.c | 4 +-
hw/sparc/leon3.c | 4 +-
hw/sparc/sun4m.c | 4 +-
hw/sparc64/sparc64.c | 2 +-
hw/unicore32/puv3.c | 2 +-
hw/xtensa/pic_cpu.c | 2 +-
linux-user/aarch64/cpu_loop.c | 6 +-
linux-user/aarch64/signal.c | 4 +-
linux-user/alpha/cpu_loop.c | 2 +-
linux-user/arm/cpu_loop.c | 4 +-
linux-user/cris/cpu_loop.c | 4 +-
linux-user/elfload.c | 6 +-
linux-user/hppa/cpu_loop.c | 2 +-
linux-user/i386/cpu_loop.c | 2 +-
linux-user/i386/signal.c | 2 +-
linux-user/m68k-sim.c | 3 +-
linux-user/m68k/cpu_loop.c | 4 +-
linux-user/main.c | 2 +-
linux-user/microblaze/cpu_loop.c | 2 +-
linux-user/mips/cpu_loop.c | 4 +-
linux-user/nios2/cpu_loop.c | 2 +-
linux-user/openrisc/cpu_loop.c | 2 +-
linux-user/ppc/cpu_loop.c | 2 +-
linux-user/riscv/cpu_loop.c | 4 +-
linux-user/s390x/cpu_loop.c | 2 +-
linux-user/sh4/cpu_loop.c | 2 +-
linux-user/signal.c | 8 +-
linux-user/sparc/cpu_loop.c | 2 +-
linux-user/syscall.c | 26 +-
linux-user/tilegx/cpu_loop.c | 2 +-
linux-user/uname.c | 2 +-
linux-user/vm86.c | 18 +-
linux-user/xtensa/cpu_loop.c | 2 +-
qom/cpu-common.c | 425 ++++++++++++++++++++++
qom/cpu.c | 408 +--------------------
target/alpha/helper.c | 8 +-
target/alpha/sys_helper.c | 8 +-
target/arm/arm-semi.c | 4 +-
target/arm/cpu64.c | 2 +-
target/arm/helper-a64.c | 4 +-
target/arm/helper.c | 160 ++++----
target/arm/op_helper.c | 21 +-
target/arm/translate-a64.c | 4 +-
target/arm/translate.c | 2 +-
target/arm/vfp_helper.c | 2 +-
target/cris/mmu.c | 3 +-
target/cris/op_helper.c | 10 +-
target/cris/translate.c | 2 +-
target/hppa/helper.c | 3 +-
target/hppa/int_helper.c | 4 +-
target/hppa/mem_helper.c | 10 +-
target/hppa/op_helper.c | 10 +-
target/i386/bpt_helper.c | 4 +-
target/i386/cpu.c | 4 +-
target/i386/excp_helper.c | 2 +-
target/i386/fpu_helper.c | 2 +-
target/i386/hax-all.c | 6 +-
target/i386/helper.c | 16 +-
target/i386/hvf/x86_decode.c | 22 +-
target/i386/hvf/x86_emu.c | 48 +--
target/i386/mem_helper.c | 4 +-
target/i386/misc_helper.c | 24 +-
target/i386/seg_helper.c | 14 +-
target/i386/smm_helper.c | 4 +-
target/i386/svm_helper.c | 22 +-
target/lm32/helper.c | 19 +-
target/lm32/op_helper.c | 6 +-
target/lm32/translate.c | 2 +-
target/m68k/helper.c | 33 +-
target/m68k/m68k-semi.c | 4 +-
target/m68k/op_helper.c | 14 +-
target/m68k/translate.c | 4 +-
target/microblaze/mmu.c | 5 +-
target/microblaze/op_helper.c | 2 +-
target/microblaze/translate.c | 2 +-
target/mips/helper.c | 15 +-
target/mips/op_helper.c | 25 +-
target/mips/translate.c | 3 +-
target/mips/translate_init.inc.c | 4 +-
target/moxie/helper.c | 6 +-
target/moxie/translate.c | 2 +-
target/nios2/mmu.c | 14 +-
target/nios2/op_helper.c | 2 +-
target/openrisc/exception_helper.c | 5 +-
target/openrisc/sys_helper.c | 8 +-
target/ppc/excp_helper.c | 14 +-
target/ppc/fpu_helper.c | 14 +-
target/ppc/kvm.c | 5 +-
target/ppc/misc_helper.c | 22 +-
target/ppc/mmu-hash64.c | 14 +-
target/ppc/mmu_helper.c | 116 +++---
target/ppc/translate_init.inc.c | 85 +++--
target/riscv/cpu_helper.c | 4 +-
target/riscv/csr.c | 12 +-
target/riscv/op_helper.c | 8 +-
target/s390x/cc_helper.c | 5 +-
target/s390x/diag.c | 2 +-
target/s390x/excp_helper.c | 6 +-
target/s390x/fpu_helper.c | 4 +-
target/s390x/gdbstub.c | 24 +-
target/s390x/helper.c | 7 +-
target/s390x/int_helper.c | 3 +-
target/s390x/interrupt.c | 6 +-
target/s390x/mem_helper.c | 30 +-
target/s390x/misc_helper.c | 50 +--
target/s390x/mmu_helper.c | 8 +-
target/s390x/sigp.c | 4 +-
target/sh4/helper.c | 26 +-
target/sh4/op_helper.c | 11 +-
target/sparc/fop_helper.c | 2 +-
target/sparc/helper.c | 8 +-
target/sparc/ldst_helper.c | 33 +-
target/sparc/mmu_helper.c | 10 +-
target/tilegx/helper.c | 2 +-
target/tricore/op_helper.c | 4 +-
target/unicore32/helper.c | 8 +-
target/unicore32/op_helper.c | 2 +-
target/unicore32/softmmu.c | 11 +-
target/unicore32/translate.c | 26 +-
target/unicore32/ucf64_helper.c | 2 +-
target/xtensa/dbg_helper.c | 4 +-
target/xtensa/exc_helper.c | 9 +-
target/xtensa/helper.c | 2 +-
target/xtensa/mmu_helper.c | 11 +-
target/xtensa/xtensa-semi.c | 2 +-
tcg/aarch64/tcg-target.inc.c | 36 +-
tcg/arm/tcg-target.inc.c | 149 +++-----
tcg/i386/tcg-target.inc.c | 6 +-
tcg/mips/tcg-target.inc.c | 45 +--
tcg/ppc/tcg-target.inc.c | 32 +-
tcg/riscv/tcg-target.inc.c | 37 +-
tcg/s390/tcg-target.inc.c | 13 +-
tcg/sparc/tcg-target.inc.c | 40 +-
docs/devel/tracing.txt | 4 +-
qom/Makefile.objs | 2 +-
scripts/tracetool/format/tcg_helper_c.py | 2 +-
210 files changed, 1841 insertions(+), 2256 deletions(-)
create mode 100644 target/alpha/cpu-param.h
create mode 100644 target/arm/cpu-param.h
create mode 100644 target/cris/cpu-param.h
create mode 100644 target/hppa/cpu-param.h
create mode 100644 target/i386/cpu-param.h
create mode 100644 target/lm32/cpu-param.h
create mode 100644 target/m68k/cpu-param.h
create mode 100644 target/microblaze/cpu-param.h
create mode 100644 target/mips/cpu-param.h
create mode 100644 target/moxie/cpu-param.h
create mode 100644 target/nios2/cpu-param.h
create mode 100644 target/openrisc/cpu-param.h
create mode 100644 target/ppc/cpu-param.h
create mode 100644 target/riscv/cpu-param.h
create mode 100644 target/s390x/cpu-param.h
create mode 100644 target/sh4/cpu-param.h
create mode 100644 target/sparc/cpu-param.h
create mode 100644 target/tilegx/cpu-param.h
create mode 100644 target/tricore/cpu-param.h
create mode 100644 target/unicore32/cpu-param.h
create mode 100644 target/xtensa/cpu-param.h
create mode 100644 qom/cpu-common.c
--
2.17.1
- [Qemu-devel] [PATCH for-4.1 00/35] tcg: Move the softmmu tlb to CPUNegativeOffsetState,
Richard Henderson <=
- [Qemu-devel] [PATCH 16/35] target/mips: Use env_cpu, env_archcpu, Richard Henderson, 2019/03/23
- [Qemu-devel] [PATCH 30/35] cpu: Introduce CPUNegativeOffsetState, Richard Henderson, 2019/03/23
- [Qemu-devel] [PATCH 32/35] cpu: Move the softmmu tlb to CPUNegativeOffsetState, Richard Henderson, 2019/03/23
- [Qemu-devel] [PATCH 29/35] cpu: Move ENV_OFFSET to exec/gen-icount.h, Richard Henderson, 2019/03/23
- [Qemu-devel] [PATCH 25/35] target/tilegx: Use env_cpu, Richard Henderson, 2019/03/23
- [Qemu-devel] [PATCH 33/35] cpu: Remove CPU_COMMON, Richard Henderson, 2019/03/23
- [Qemu-devel] [PATCH 31/35] cpu: Move icount_decr to CPUNegativeOffsetState, Richard Henderson, 2019/03/23