[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[PULL 04/27] qemu/atomic128: Add x86_64 atomic128-ldst.h
|
From: |
Richard Henderson |
|
Subject: |
[PULL 04/27] qemu/atomic128: Add x86_64 atomic128-ldst.h |
|
Date: |
Tue, 30 May 2023 11:59:26 -0700 |
With CPUINFO_ATOMIC_VMOVDQA, we can perform proper atomic
load/store without cmpxchg16b.
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
host/include/x86_64/host/atomic128-ldst.h | 68 +++++++++++++++++++++++
1 file changed, 68 insertions(+)
create mode 100644 host/include/x86_64/host/atomic128-ldst.h
diff --git a/host/include/x86_64/host/atomic128-ldst.h
b/host/include/x86_64/host/atomic128-ldst.h
new file mode 100644
index 0000000000..adc9332f91
--- /dev/null
+++ b/host/include/x86_64/host/atomic128-ldst.h
@@ -0,0 +1,68 @@
+/*
+ * SPDX-License-Identifier: GPL-2.0-or-later
+ * Load/store for 128-bit atomic operations, x86_64 version.
+ *
+ * Copyright (C) 2023 Linaro, Ltd.
+ *
+ * See docs/devel/atomics.rst for discussion about the guarantees each
+ * atomic primitive is meant to provide.
+ */
+
+#ifndef AARCH64_ATOMIC128_LDST_H
+#define AARCH64_ATOMIC128_LDST_H
+
+#ifdef CONFIG_INT128_TYPE
+#include "host/cpuinfo.h"
+#include "tcg/debug-assert.h"
+
+/*
+ * Through clang 16, with -mcx16, __atomic_load_n is incorrectly
+ * expanded to a read-write operation: lock cmpxchg16b.
+ */
+
+#define HAVE_ATOMIC128_RO likely(cpuinfo & CPUINFO_ATOMIC_VMOVDQA)
+#define HAVE_ATOMIC128_RW 1
+
+static inline Int128 atomic16_read_ro(const Int128 *ptr)
+{
+ Int128Alias r;
+
+ tcg_debug_assert(HAVE_ATOMIC128_RO);
+ asm("vmovdqa %1, %0" : "=x" (r.i) : "m" (*ptr));
+
+ return r.s;
+}
+
+static inline Int128 atomic16_read_rw(Int128 *ptr)
+{
+ __int128_t *ptr_align = __builtin_assume_aligned(ptr, 16);
+ Int128Alias r;
+
+ if (HAVE_ATOMIC128_RO) {
+ asm("vmovdqa %1, %0" : "=x" (r.i) : "m" (*ptr_align));
+ } else {
+ r.i = __sync_val_compare_and_swap_16(ptr_align, 0, 0);
+ }
+ return r.s;
+}
+
+static inline void atomic16_set(Int128 *ptr, Int128 val)
+{
+ __int128_t *ptr_align = __builtin_assume_aligned(ptr, 16);
+ Int128Alias new = { .s = val };
+
+ if (HAVE_ATOMIC128_RO) {
+ asm("vmovdqa %1, %0" : "=m"(*ptr_align) : "x" (new.i));
+ } else {
+ __int128_t old;
+ do {
+ old = *ptr_align;
+ } while (!__sync_bool_compare_and_swap_16(ptr_align, old, new.i));
+ }
+}
+#else
+/* Provide QEMU_ERROR stubs. */
+#include "host/include/generic/host/atomic128-ldst.h"
+#endif
+
+#endif /* AARCH64_ATOMIC128_LDST_H */
--
2.34.1
- [PULL 00/27] tcg patch queue, Richard Henderson, 2023/05/30
- [PULL 01/27] tcg: Fix register move type in tcg_out_ld_helper_ret, Richard Henderson, 2023/05/30
- [PULL 06/27] tcg/aarch64: Rename temporaries, Richard Henderson, 2023/05/30
- [PULL 03/27] meson: Split test for __int128_t type from __int128_t arithmetic, Richard Henderson, 2023/05/30
- [PULL 02/27] accel/tcg: Fix check for page writeability in load_atomic16_or_exit, Richard Henderson, 2023/05/30
- [PULL 04/27] qemu/atomic128: Add x86_64 atomic128-ldst.h,
Richard Henderson <=
- [PULL 14/27] accel/tcg: Add x86_64 load_atom_extract_al16_or_al8, Richard Henderson, 2023/05/30
- [PULL 17/27] tcg: Remove TCG_TARGET_TLB_DISPLACEMENT_BITS, Richard Henderson, 2023/05/30
- [PULL 21/27] decodetree: Do not remove output_file from /dev, Richard Henderson, 2023/05/30
- [PULL 22/27] tests/decode: Convert tests to meson, Richard Henderson, 2023/05/30
- [PULL 24/27] scripts/decodetree: Pass lvalue-formatter function to str_extract(), Richard Henderson, 2023/05/30
- [PULL 13/27] accel/tcg: Extract store_atom_insert_al16 to host header, Richard Henderson, 2023/05/30
- [PULL 18/27] decodetree: Add --test-for-error, Richard Henderson, 2023/05/30
- [PULL 15/27] accel/tcg: Add aarch64 lse2 load_atom_extract_al16_or_al8, Richard Henderson, 2023/05/30
- [PULL 19/27] decodetree: Fix recursion in prop_format and build_tree, Richard Henderson, 2023/05/30
- [PULL 27/27] tests/decode: Add tests for various named-field cases, Richard Henderson, 2023/05/30