move unaligned memory access functions to bswap.h

This is just code movement, and moving the fpu/ include path from
target-dependent to target-independent Make variables.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

1 files changed, 474 insertions, 0 deletions

diff --git a/bswap.h b/bswap.h
index 82a79517db..f41bebed83 100644
--- a/bswap.h
+++ b/bswap.h
@@ -11,6 +11,8 @@
 #include <machine/bswap.h>
 #else
 
+#include "softfloat.h"
+
 #ifdef CONFIG_BYTESWAP_H
 #include <byteswap.h>
 #else
@@ -237,4 +239,476 @@ static inline uint32_t qemu_bswap_len(uint32_t value, int len)
     return bswap32(value) >> (32 - 8 * len);
 }
 
+typedef union {
+    float32 f;
+    uint32_t l;
+} CPU_FloatU;
+
+typedef union {
+    float64 d;
+#if defined(HOST_WORDS_BIGENDIAN)
+    struct {
+        uint32_t upper;
+        uint32_t lower;
+    } l;
+#else
+    struct {
+        uint32_t lower;
+        uint32_t upper;
+    } l;
+#endif
+    uint64_t ll;
+} CPU_DoubleU;
+
+typedef union {
+     floatx80 d;
+     struct {
+         uint64_t lower;
+         uint16_t upper;
+     } l;
+} CPU_LDoubleU;
+
+typedef union {
+    float128 q;
+#if defined(HOST_WORDS_BIGENDIAN)
+    struct {
+        uint32_t upmost;
+        uint32_t upper;
+        uint32_t lower;
+        uint32_t lowest;
+    } l;
+    struct {
+        uint64_t upper;
+        uint64_t lower;
+    } ll;
+#else
+    struct {
+        uint32_t lowest;
+        uint32_t lower;
+        uint32_t upper;
+        uint32_t upmost;
+    } l;
+    struct {
+        uint64_t lower;
+        uint64_t upper;
+    } ll;
+#endif
+} CPU_QuadU;
+
+/* unaligned/endian-independent pointer access */
+
+/*
+ * the generic syntax is:
+ *
+ * load: ld{type}{sign}{size}{endian}_p(ptr)
+ *
+ * store: st{type}{size}{endian}_p(ptr, val)
+ *
+ * Note there are small differences with the softmmu access API!
+ *
+ * type is:
+ * (empty): integer access
+ *   f    : float access
+ *
+ * sign is:
+ * (empty): for floats or 32 bit size
+ *   u    : unsigned
+ *   s    : signed
+ *
+ * size is:
+ *   b: 8 bits
+ *   w: 16 bits
+ *   l: 32 bits
+ *   q: 64 bits
+ *
+ * endian is:
+ * (empty): 8 bit access
+ *   be   : big endian
+ *   le   : little endian
+ */
+static inline int ldub_p(const void *ptr)
+{
+    return *(uint8_t *)ptr;
+}
+
+static inline int ldsb_p(const void *ptr)
+{
+    return *(int8_t *)ptr;
+}
+
+static inline void stb_p(void *ptr, int v)
+{
+    *(uint8_t *)ptr = v;
+}
+
+/* NOTE: on arm, putting 2 in /proc/sys/debug/alignment so that the
+   kernel handles unaligned load/stores may give better results, but
+   it is a system wide setting : bad */
+#if defined(HOST_WORDS_BIGENDIAN) || defined(WORDS_ALIGNED)
+
+/* conservative code for little endian unaligned accesses */
+static inline int lduw_le_p(const void *ptr)
+{
+#ifdef _ARCH_PPC
+    int val;
+    __asm__ __volatile__ ("lhbrx %0,0,%1" : "=r" (val) : "r" (ptr));
+    return val;
+#else
+    const uint8_t *p = ptr;
+    return p[0] | (p[1] << 8);
+#endif
+}
+
+static inline int ldsw_le_p(const void *ptr)
+{
+#ifdef _ARCH_PPC
+    int val;
+    __asm__ __volatile__ ("lhbrx %0,0,%1" : "=r" (val) : "r" (ptr));
+    return (int16_t)val;
+#else
+    const uint8_t *p = ptr;
+    return (int16_t)(p[0] | (p[1] << 8));
+#endif
+}
+
+static inline int ldl_le_p(const void *ptr)
+{
+#ifdef _ARCH_PPC
+    int val;
+    __asm__ __volatile__ ("lwbrx %0,0,%1" : "=r" (val) : "r" (ptr));
+    return val;
+#else
+    const uint8_t *p = ptr;
+    return p[0] | (p[1] << 8) | (p[2] << 16) | (p[3] << 24);
+#endif
+}
+
+static inline uint64_t ldq_le_p(const void *ptr)
+{
+    const uint8_t *p = ptr;
+    uint32_t v1, v2;
+    v1 = ldl_le_p(p);
+    v2 = ldl_le_p(p + 4);
+    return v1 | ((uint64_t)v2 << 32);
+}
+
+static inline void stw_le_p(void *ptr, int v)
+{
+#ifdef _ARCH_PPC
+    __asm__ __volatile__ ("sthbrx %1,0,%2" : "=m" (*(uint16_t *)ptr) : "r" (v), "r" (ptr));
+#else
+    uint8_t *p = ptr;
+    p[0] = v;
+    p[1] = v >> 8;
+#endif
+}
+
+static inline void stl_le_p(void *ptr, int v)
+{
+#ifdef _ARCH_PPC
+    __asm__ __volatile__ ("stwbrx %1,0,%2" : "=m" (*(uint32_t *)ptr) : "r" (v), "r" (ptr));
+#else
+    uint8_t *p = ptr;
+    p[0] = v;
+    p[1] = v >> 8;
+    p[2] = v >> 16;
+    p[3] = v >> 24;
+#endif
+}
+
+static inline void stq_le_p(void *ptr, uint64_t v)
+{
+    uint8_t *p = ptr;
+    stl_le_p(p, (uint32_t)v);
+    stl_le_p(p + 4, v >> 32);
+}
+
+/* float access */
+
+static inline float32 ldfl_le_p(const void *ptr)
+{
+    union {
+        float32 f;
+        uint32_t i;
+    } u;
+    u.i = ldl_le_p(ptr);
+    return u.f;
+}
+
+static inline void stfl_le_p(void *ptr, float32 v)
+{
+    union {
+        float32 f;
+        uint32_t i;
+    } u;
+    u.f = v;
+    stl_le_p(ptr, u.i);
+}
+
+static inline float64 ldfq_le_p(const void *ptr)
+{
+    CPU_DoubleU u;
+    u.l.lower = ldl_le_p(ptr);
+    u.l.upper = ldl_le_p(ptr + 4);
+    return u.d;
+}
+
+static inline void stfq_le_p(void *ptr, float64 v)
+{
+    CPU_DoubleU u;
+    u.d = v;
+    stl_le_p(ptr, u.l.lower);
+    stl_le_p(ptr + 4, u.l.upper);
+}
+
+#else
+
+static inline int lduw_le_p(const void *ptr)
+{
+    return *(uint16_t *)ptr;
+}
+
+static inline int ldsw_le_p(const void *ptr)
+{
+    return *(int16_t *)ptr;
+}
+
+static inline int ldl_le_p(const void *ptr)
+{
+    return *(uint32_t *)ptr;
+}
+
+static inline uint64_t ldq_le_p(const void *ptr)
+{
+    return *(uint64_t *)ptr;
+}
+
+static inline void stw_le_p(void *ptr, int v)
+{
+    *(uint16_t *)ptr = v;
+}
+
+static inline void stl_le_p(void *ptr, int v)
+{
+    *(uint32_t *)ptr = v;
+}
+
+static inline void stq_le_p(void *ptr, uint64_t v)
+{
+    *(uint64_t *)ptr = v;
+}
+
+/* float access */
+
+static inline float32 ldfl_le_p(const void *ptr)
+{
+    return *(float32 *)ptr;
+}
+
+static inline float64 ldfq_le_p(const void *ptr)
+{
+    return *(float64 *)ptr;
+}
+
+static inline void stfl_le_p(void *ptr, float32 v)
+{
+    *(float32 *)ptr = v;
+}
+
+static inline void stfq_le_p(void *ptr, float64 v)
+{
+    *(float64 *)ptr = v;
+}
+#endif
+
+#if !defined(HOST_WORDS_BIGENDIAN) || defined(WORDS_ALIGNED)
+
+static inline int lduw_be_p(const void *ptr)
+{
+#if defined(__i386__)
+    int val;
+    asm volatile ("movzwl %1, %0\n"
+                  "xchgb %b0, %h0\n"
+                  : "=q" (val)
+                  : "m" (*(uint16_t *)ptr));
+    return val;
+#else
+    const uint8_t *b = ptr;
+    return ((b[0] << 8) | b[1]);
+#endif
+}
+
+static inline int ldsw_be_p(const void *ptr)
+{
+#if defined(__i386__)
+    int val;
+    asm volatile ("movzwl %1, %0\n"
+                  "xchgb %b0, %h0\n"
+                  : "=q" (val)
+                  : "m" (*(uint16_t *)ptr));
+    return (int16_t)val;
+#else
+    const uint8_t *b = ptr;
+    return (int16_t)((b[0] << 8) | b[1]);
+#endif
+}
+
+static inline int ldl_be_p(const void *ptr)
+{
+#if defined(__i386__) || defined(__x86_64__)
+    int val;
+    asm volatile ("movl %1, %0\n"
+                  "bswap %0\n"
+                  : "=r" (val)
+                  : "m" (*(uint32_t *)ptr));
+    return val;
+#else
+    const uint8_t *b = ptr;
+    return (b[0] << 24) | (b[1] << 16) | (b[2] << 8) | b[3];
+#endif
+}
+
+static inline uint64_t ldq_be_p(const void *ptr)
+{
+    uint32_t a,b;
+    a = ldl_be_p(ptr);
+    b = ldl_be_p((uint8_t *)ptr + 4);
+    return (((uint64_t)a<<32)|b);
+}
+
+static inline void stw_be_p(void *ptr, int v)
+{
+#if defined(__i386__)
+    asm volatile ("xchgb %b0, %h0\n"
+                  "movw %w0, %1\n"
+                  : "=q" (v)
+                  : "m" (*(uint16_t *)ptr), "0" (v));
+#else
+    uint8_t *d = (uint8_t *) ptr;
+    d[0] = v >> 8;
+    d[1] = v;
+#endif
+}
+
+static inline void stl_be_p(void *ptr, int v)
+{
+#if defined(__i386__) || defined(__x86_64__)
+    asm volatile ("bswap %0\n"
+                  "movl %0, %1\n"
+                  : "=r" (v)
+                  : "m" (*(uint32_t *)ptr), "0" (v));
+#else
+    uint8_t *d = (uint8_t *) ptr;
+    d[0] = v >> 24;
+    d[1] = v >> 16;
+    d[2] = v >> 8;
+    d[3] = v;
+#endif
+}
+
+static inline void stq_be_p(void *ptr, uint64_t v)
+{
+    stl_be_p(ptr, v >> 32);
+    stl_be_p((uint8_t *)ptr + 4, v);
+}
+
+/* float access */
+
+static inline float32 ldfl_be_p(const void *ptr)
+{
+    union {
+        float32 f;
+        uint32_t i;
+    } u;
+    u.i = ldl_be_p(ptr);
+    return u.f;
+}
+
+static inline void stfl_be_p(void *ptr, float32 v)
+{
+    union {
+        float32 f;
+        uint32_t i;
+    } u;
+    u.f = v;
+    stl_be_p(ptr, u.i);
+}
+
+static inline float64 ldfq_be_p(const void *ptr)
+{
+    CPU_DoubleU u;
+    u.l.upper = ldl_be_p(ptr);
+    u.l.lower = ldl_be_p((uint8_t *)ptr + 4);
+    return u.d;
+}
+
+static inline void stfq_be_p(void *ptr, float64 v)
+{
+    CPU_DoubleU u;
+    u.d = v;
+    stl_be_p(ptr, u.l.upper);
+    stl_be_p((uint8_t *)ptr + 4, u.l.lower);
+}
+
+#else
+
+static inline int lduw_be_p(const void *ptr)
+{
+    return *(uint16_t *)ptr;
+}
+
+static inline int ldsw_be_p(const void *ptr)
+{
+    return *(int16_t *)ptr;
+}
+
+static inline int ldl_be_p(const void *ptr)
+{
+    return *(uint32_t *)ptr;
+}
+
+static inline uint64_t ldq_be_p(const void *ptr)
+{
+    return *(uint64_t *)ptr;
+}
+
+static inline void stw_be_p(void *ptr, int v)
+{
+    *(uint16_t *)ptr = v;
+}
+
+static inline void stl_be_p(void *ptr, int v)
+{
+    *(uint32_t *)ptr = v;
+}
+
+static inline void stq_be_p(void *ptr, uint64_t v)
+{
+    *(uint64_t *)ptr = v;
+}
+
+/* float access */
+
+static inline float32 ldfl_be_p(const void *ptr)
+{
+    return *(float32 *)ptr;
+}
+
+static inline float64 ldfq_be_p(const void *ptr)
+{
+    return *(float64 *)ptr;
+}
+
+static inline void stfl_be_p(void *ptr, float32 v)
+{
+    *(float32 *)ptr = v;
+}
+
+static inline void stfq_be_p(void *ptr, float64 v)
+{
+    *(float64 *)ptr = v;
+}
+
+#endif
+
 #endif /* BSWAP_H */