accel/tcg: add tracepoints for cpu_loop_exit_atomic

We try to avoid using cpu_loop_exit_atomic as it brings in an all-core
sync point. However on some cpu/kernel/benchmark combinations it is
starting to show up in the performance profile. To make it easier to
see whats going on add tracepoints for the slow path so we can see
what is triggering the wait.

It seems for a modern CPU it can be quite a bit, for example:

./qemu-system-aarch64 \
           -machine type=virt,virtualization=on,pflash0=rom,pflash1=efivars,gic-version=max \
           -smp 4 \
           -accel tcg \
           -device virtio-net-pci,netdev=unet \
           -device virtio-scsi-pci \
           -device scsi-hd,drive=hd \
           -netdev user,id=unet,hostfwd=tcp::2222-:22 \
           -blockdev driver=raw,node-name=hd,file.driver=host_device,file.filename=/dev/zen-ssd2/trixie-arm64,discard=unmap \
           -serial mon:stdio \
           -blockdev node-name=rom,driver=file,filename=(pwd)/pc-bios/edk2-aarch64-code.fd,read-only=true \
           -blockdev node-name=efivars,driver=file,filename=$HOME/images/qemu-arm64-efivars \
           -m 8192 \
           -object memory-backend-memfd,id=mem,size=8G,share=on \
           -kernel /home/alex/lsrc/linux.git/builds/arm64/arch/arm64/boot/Image -append "root=/dev/sda2 console=ttyAMA0 systemd.unit=benchmark-stress-ng.service" \
           -display none -d trace:load_atom\*_fallback,trace:store_atom\*_fallback

With:

  -cpu neoverse-v1,pauth-impdef=on => 2203343

With:

  -cpu cortex-a76 => 0

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20241023113406.1284676-9-alex.bennee@linaro.org>
This commit is contained in:
Alex Bennée 2024-10-23 12:33:56 +01:00
parent b6a48d2a4b
commit b24bad34bf
3 changed files with 22 additions and 1 deletions

View File

@ -168,6 +168,7 @@ static uint64_t load_atomic8_or_exit(CPUState *cpu, uintptr_t ra, void *pv)
#endif
/* Ultimate fallback: re-execute in serial context. */
trace_load_atom8_or_exit_fallback(ra);
cpu_loop_exit_atomic(cpu, ra);
}
@ -212,6 +213,7 @@ static Int128 load_atomic16_or_exit(CPUState *cpu, uintptr_t ra, void *pv)
}
/* Ultimate fallback: re-execute in serial context. */
trace_load_atom16_or_exit_fallback(ra);
cpu_loop_exit_atomic(cpu, ra);
}
@ -519,6 +521,7 @@ static uint64_t load_atom_8(CPUState *cpu, uintptr_t ra,
if (HAVE_al8) {
return load_atom_extract_al8x2(pv);
}
trace_load_atom8_fallback(memop, ra);
cpu_loop_exit_atomic(cpu, ra);
default:
g_assert_not_reached();
@ -563,6 +566,7 @@ static Int128 load_atom_16(CPUState *cpu, uintptr_t ra,
break;
case MO_64:
if (!HAVE_al8) {
trace_load_atom16_fallback(memop, ra);
cpu_loop_exit_atomic(cpu, ra);
}
a = load_atomic8(pv);
@ -570,6 +574,7 @@ static Int128 load_atom_16(CPUState *cpu, uintptr_t ra,
break;
case -MO_64:
if (!HAVE_al8) {
trace_load_atom16_fallback(memop, ra);
cpu_loop_exit_atomic(cpu, ra);
}
a = load_atom_extract_al8x2(pv);
@ -897,6 +902,7 @@ static void store_atom_2(CPUState *cpu, uintptr_t ra,
g_assert_not_reached();
}
trace_store_atom2_fallback(memop, ra);
cpu_loop_exit_atomic(cpu, ra);
}
@ -961,6 +967,7 @@ static void store_atom_4(CPUState *cpu, uintptr_t ra,
return;
}
}
trace_store_atom4_fallback(memop, ra);
cpu_loop_exit_atomic(cpu, ra);
default:
g_assert_not_reached();
@ -1029,6 +1036,7 @@ static void store_atom_8(CPUState *cpu, uintptr_t ra,
default:
g_assert_not_reached();
}
trace_store_atom8_fallback(memop, ra);
cpu_loop_exit_atomic(cpu, ra);
}
@ -1107,5 +1115,6 @@ static void store_atom_16(CPUState *cpu, uintptr_t ra,
default:
g_assert_not_reached();
}
trace_store_atom16_fallback(memop, ra);
cpu_loop_exit_atomic(cpu, ra);
}

View File

@ -12,3 +12,15 @@ memory_notdirty_set_dirty(uint64_t vaddr) "0x%" PRIx64
# translate-all.c
translate_block(void *tb, uintptr_t pc, const void *tb_code) "tb:%p, pc:0x%"PRIxPTR", tb_code:%p"
# ldst_atomicity
load_atom2_fallback(uint32_t memop, uintptr_t ra) "mop:0x%"PRIx32", ra:0x%"PRIxPTR""
load_atom4_fallback(uint32_t memop, uintptr_t ra) "mop:0x%"PRIx32", ra:0x%"PRIxPTR""
load_atom8_or_exit_fallback(uintptr_t ra) "ra:0x%"PRIxPTR""
load_atom8_fallback(uint32_t memop, uintptr_t ra) "mop:0x%"PRIx32", ra:0x%"PRIxPTR""
load_atom16_fallback(uint32_t memop, uintptr_t ra) "mop:0x%"PRIx32", ra:0x%"PRIxPTR""
load_atom16_or_exit_fallback(uintptr_t ra) "ra:0x%"PRIxPTR""
store_atom2_fallback(uint32_t memop, uintptr_t ra) "mop:0x%"PRIx32", ra:0x%"PRIxPTR""
store_atom4_fallback(uint32_t memop, uintptr_t ra) "mop:0x%"PRIx32", ra:0x%"PRIxPTR""
store_atom8_fallback(uint32_t memop, uintptr_t ra) "mop:0x%"PRIx32", ra:0x%"PRIxPTR""
store_atom16_fallback(uint32_t memop, uintptr_t ra) "mop:0x%"PRIx32", ra:0x%"PRIxPTR""

View File

@ -29,7 +29,7 @@
#include "exec/page-protection.h"
#include "exec/helper-proto.h"
#include "qemu/atomic128.h"
#include "trace/trace-root.h"
#include "trace.h"
#include "tcg/tcg-ldst.h"
#include "internal-common.h"
#include "internal-target.h"