Move userland bits of i386 npx.h and amd64 fpu.h to x86 fpu.h.

Remove FPU types from compat/ia32/ia32_reg.h that are no longer needed.
Create machine/npx.h on amd64 to allow compiling i386 code that uses
this header.

The original npx.h and fpu.h define struct envxmm differently. Both
definitions have been included in the new x86 header as struct __envxmm32
and struct __envxmm64. During compilation either __envxmm32 or __envxmm64
is defined as envxmm depending on machine architecture. On amd64 the i386
struct is also available as struct envxmm32.

Reviewed by:	kib

1 files changed, 1 insertions, 121 deletions

diff --git a/sys/i386/include/npx.h b/sys/i386/include/npx.h
index bfecd56..86ac477 100644
--- a/sys/i386/include/npx.h
+++ b/sys/i386/include/npx.h
@@ -41,127 +41,7 @@
 #ifndef _MACHINE_NPX_H_
 #define	_MACHINE_NPX_H_
 
-/* Environment information of floating point unit */
-struct env87 {
-	int32_t		en_cw;		/* control word (16bits) */
-	int32_t		en_sw;		/* status word (16bits) */
-	int32_t		en_tw;		/* tag word (16bits) */
-	int32_t		en_fip;		/* fp instruction pointer */
-	uint16_t	en_fcs;		/* fp code segment selector */
-	uint16_t	en_opcode;	/* opcode last executed (11 bits ) */
-	int32_t		en_foo;		/* fp operand offset */
-	int32_t		en_fos;		/* fp operand segment selector */
-};
-
-/* Contents of each floating point accumulator */
-struct fpacc87 {
-#ifdef dontdef /* too unportable */
-	u_long	fp_mantlo;	/* mantissa low (31:0) */
-	u_long	fp_manthi;	/* mantissa high (63:32) */
-	int	fp_exp:15;	/* exponent */
-	int	fp_sgn:1;	/* mantissa sign */
-#else
-	u_char	fp_bytes[10];
-#endif
-};
-
-/* Floating point context */
-struct save87 {
-	struct	env87 sv_env;	/* floating point control/status */
-	struct	fpacc87	sv_ac[8];	/* accumulator contents, 0-7 */
-	u_char	sv_pad0[4];	/* padding for (now unused) saved status word */
-	/*
-	 * Bogus padding for emulators.  Emulators should use their own
-	 * struct and arrange to store into this struct (ending here)
-	 * before it is inspected for ptracing or for core dumps.  Some
-	 * emulators overwrite the whole struct.  We have no good way of
-	 * knowing how much padding to leave.  Leave just enough for the
-	 * GPL emulator's i387_union (176 bytes total).
-	 */
-	u_char	sv_pad[64];	/* padding; used by emulators */
-};
-
-struct  envxmm {
-	u_int16_t	en_cw;		/* control word (16bits) */
-	u_int16_t	en_sw;		/* status word (16bits) */
-	u_int16_t	en_tw;		/* tag word (16bits) */
-	u_int16_t	en_opcode;	/* opcode last executed (11 bits ) */
-	u_int32_t	en_fip;		/* floating point instruction pointer */
-	u_int16_t	en_fcs;		/* floating code segment selector */
-	u_int16_t	en_pad0;	/* padding */
-	u_int32_t	en_foo;		/* floating operand offset */
-	u_int16_t	en_fos;		/* floating operand segment selector */
-	u_int16_t	en_pad1;	/* padding */
-	u_int32_t	en_mxcsr;	/* SSE control/status register */
-	u_int32_t	en_mxcsr_mask;	/* valid bits in mxcsr */
-};
-
-/* Contents of each SSE extended accumulator */
-struct  xmmacc {
-	u_char	xmm_bytes[16];
-};
-
-/* Contents of the upper 16 bytes of each AVX extended accumulator */
-struct  ymmacc {
-	uint8_t  ymm_bytes[16];
-};
-
-struct  savexmm {
-	struct	envxmm	sv_env;
-	struct {
-		struct fpacc87	fp_acc;
-		u_char		fp_pad[6];      /* padding */
-	} sv_fp[8];
-	struct xmmacc	sv_xmm[8];
-	u_char sv_pad[224];
-} __aligned(16);
-
-union	savefpu {
-	struct	save87	sv_87;
-	struct	savexmm	sv_xmm;
-};
-
-struct xstate_hdr {
-	uint64_t xstate_bv;
-	uint8_t xstate_rsrv0[16];
-	uint8_t	xstate_rsrv[40];
-};
-
-struct savexmm_xstate {
-	struct xstate_hdr sx_hd;
-	struct ymmacc	sx_ymm[16];
-};
-
-struct savexmm_ymm {
-	struct	envxmm	sv_env;
-	struct {
-		struct fpacc87	fp_acc;
-		int8_t		fp_pad[6];      /* padding */
-	} sv_fp[8];
-	struct xmmacc	sv_xmm[16];
-	uint8_t sv_pad[96];
-	struct savexmm_xstate sv_xstate;
-} __aligned(64);
-
-/*
- * The hardware default control word for i387's and later coprocessors is
- * 0x37F, giving:
- *
- *	round to nearest
- *	64-bit precision
- *	all exceptions masked.
- *
- * We modify the affine mode bit and precision bits in this to give:
- *
- *	affine mode for 287's (if they work at all) (1 in bitfield 1<<12)
- *	53-bit precision (2 in bitfield 3<<8)
- *
- * 64-bit precision often gives bad results with high level languages
- * because it makes the results of calculations depend on whether
- * intermediate values are stored in memory or in FPU registers.
- */
-#define	__INITIAL_NPXCW__	0x127F
-#define	__INITIAL_MXCSR__	0x1F80
+#include <x86/fpu.h>
 
 #ifdef _KERNEL