kernel: introduce per process FPU struct (PFPU) for more flexible machine specific fpu handling

introducing the PFPU structue which allows the machine specific
code some flexibility on how to handle the FPU process state.

for example, in the pc and pc64 kernel, the FPsave structure is
arround 512 bytes. with avx512, it could grow up to 2K. instead
of embedding that into the Proc strucutre, it is more effective
to allocate it on first use of the fpu, as most processes do not
use simd or floating point in the first place. also, the FPsave
structure has special 16 byte alignment constraint, which further
favours dynamic allocation.

this gets rid of the memmoves in pc/pc64 kernels for the aligment.

there is also devproc, which is now checking if the fpsave area
is actually valid before reading it, avoiding debuggers to see
garbage data.

the Notsave structure is gone now, as it was not used on any
machine.

2 files changed, 11 insertions, 15 deletions

diff --git a/sys/src/9/kw/dat.h b/sys/src/9/kw/dat.h
index ed0f15bcc..84e33ac2d 100644
--- a/sys/src/9/kw/dat.h
+++ b/sys/src/9/kw/dat.h
@@ -1,13 +1,13 @@
 typedef struct Conf	Conf;
 typedef struct Confmem	Confmem;
 typedef struct FPsave	FPsave;
+typedef struct PFPU	PFPU;
 typedef struct ISAConf	ISAConf;
 typedef struct Label	Label;
 typedef struct Lock	Lock;
 typedef struct Memcache	Memcache;
 typedef struct MMMU	MMMU;
 typedef struct Mach	Mach;
-typedef struct Notsave	Notsave;
 typedef struct Page	Page;
 typedef struct Pcidev	Pcidev;
 typedef struct PhysUart	PhysUart;
@@ -45,13 +45,13 @@ struct Label
 	uintptr	pc;
 };
 
+/*
+ * emulated floating point
+ */
 enum{
 	Nfpctlregs = 16,
 };
 
-/*
- * emulated floating point
- */
 struct FPsave
 {
 	ulong	status;
@@ -61,9 +61,12 @@ struct FPsave
 	int	fpstate;
 };
 
-/*
- * FPsave.status
- */
+struct PFPU
+{
+	int	fpstate;
+	FPsave	fpsave[1];
+};
+
 enum
 {
 	FPinit,
@@ -99,13 +102,6 @@ struct Conf
 };
 
 /*
- *  things saved in the Proc structure during a notify
- */
-struct Notsave {
-	int	emptiness;
-};
-
-/*
  *  MMU stuff in Mach.
  */
 struct MMMU
diff --git a/sys/src/9/kw/fpiarm.c b/sys/src/9/kw/fpiarm.c
index 8735f87ba..80b7ceae0 100644
--- a/sys/src/9/kw/fpiarm.c
+++ b/sys/src/9/kw/fpiarm.c
@@ -726,7 +726,7 @@ fpiarm(Ureg *ur)
 
 	if(up == nil)
 		panic("fpiarm not in a process");
-	ufp = &up->fpsave;
+	ufp = up->fpsave;
 	/*
 	 * because all the emulated fp state is in the proc structure,
 	 * it need not be saved/restored