zcommon: Refactor FPU state handling in fletcher4

Currently calls to kfpu_begin() and kfpu_end() are split between
the init() and fini() functions of the particular SIMD
implementation. This was done in #14247 as an optimization measure
for the ABD adapter. Unfortunately the split complicates FPU
handling on platforms that use a local FPU state buffer, like
Windows and macOS.

To ease porting, we introduce a boolean struct member in
fletcher_4_ops_t, indicating use of the FPU, and move the FPU state
handling from the SIMD implementations to the call sites.

Reviewed-by: Tino Reichardt <[email protected]>
Reviewed-by: Richard Yao <[email protected]>
Reviewed-by: Jorgen Lundman <[email protected]>
Signed-off-by: Attila Fülöp <[email protected]>
Closes #14600 

1 files changed, 1 insertions, 2 deletions

diff --git a/module/zcommon/zfs_fletcher_aarch64_neon.c b/module/zcommon/zfs_fletcher_aarch64_neon.c
index 8f0339728..cd5fe545a 100644
--- a/module/zcommon/zfs_fletcher_aarch64_neon.c
+++ b/module/zcommon/zfs_fletcher_aarch64_neon.c
@@ -52,7 +52,6 @@ ZFS_NO_SANITIZE_UNDEFINED
 static void
 fletcher_4_aarch64_neon_init(fletcher_4_ctx_t *ctx)
 {
-	kfpu_begin();
 	memset(ctx->aarch64_neon, 0, 4 * sizeof (zfs_fletcher_aarch64_neon_t));
 }
 
@@ -70,7 +69,6 @@ fletcher_4_aarch64_neon_fini(fletcher_4_ctx_t *ctx, zio_cksum_t *zcp)
 	    8 * ctx->aarch64_neon[3].v[1] - 8 * ctx->aarch64_neon[2].v[1] +
 	    ctx->aarch64_neon[1].v[1];
 	ZIO_SET_CHECKSUM(zcp, A, B, C, D);
-	kfpu_end();
 }
 
 #define	NEON_INIT_LOOP()			\
@@ -205,6 +203,7 @@ const fletcher_4_ops_t fletcher_4_aarch64_neon_ops = {
 	.compute_byteswap = fletcher_4_aarch64_neon_byteswap,
 	.fini_byteswap = fletcher_4_aarch64_neon_fini,
 	.valid = fletcher_4_aarch64_neon_valid,
+	.uses_fpu = B_TRUE,
 	.name = "aarch64_neon"
 };