summary refs log tree commit diff stats
path: root/target/arm/tcg
diff options
context:
space:
mode:
Diffstat (limited to 'target/arm/tcg')
-rw-r--r--target/arm/tcg/arith_helper.c296
-rw-r--r--target/arm/tcg/meson.build1
-rw-r--r--target/arm/tcg/op_addsub.c.inc103
3 files changed, 400 insertions, 0 deletions
diff --git a/target/arm/tcg/arith_helper.c b/target/arm/tcg/arith_helper.c
new file mode 100644
index 0000000000..9a555c7966
--- /dev/null
+++ b/target/arm/tcg/arith_helper.c
@@ -0,0 +1,296 @@
+/*
+ * ARM generic helpers for various arithmetical operations.
+ *
+ * This code is licensed under the GNU GPL v2 or later.
+ *
+ * SPDX-License-Identifier: GPL-2.0-or-later
+ */
+#include "qemu/osdep.h"
+#include "cpu.h"
+#include "exec/helper-proto.h"
+#include "qemu/crc32c.h"
+#include <zlib.h> /* for crc32 */
+
+/*
+ * Note that signed overflow is undefined in C.  The following routines are
+ * careful to use unsigned types where modulo arithmetic is required.
+ * Failure to do so _will_ break on newer gcc.
+ */
+
+/* Signed saturating arithmetic.  */
+
+/* Perform 16-bit signed saturating addition.  */
+static inline uint16_t add16_sat(uint16_t a, uint16_t b)
+{
+    uint16_t res;
+
+    res = a + b;
+    if (((res ^ a) & 0x8000) && !((a ^ b) & 0x8000)) {
+        if (a & 0x8000) {
+            res = 0x8000;
+        } else {
+            res = 0x7fff;
+        }
+    }
+    return res;
+}
+
+/* Perform 8-bit signed saturating addition.  */
+static inline uint8_t add8_sat(uint8_t a, uint8_t b)
+{
+    uint8_t res;
+
+    res = a + b;
+    if (((res ^ a) & 0x80) && !((a ^ b) & 0x80)) {
+        if (a & 0x80) {
+            res = 0x80;
+        } else {
+            res = 0x7f;
+        }
+    }
+    return res;
+}
+
+/* Perform 16-bit signed saturating subtraction.  */
+static inline uint16_t sub16_sat(uint16_t a, uint16_t b)
+{
+    uint16_t res;
+
+    res = a - b;
+    if (((res ^ a) & 0x8000) && ((a ^ b) & 0x8000)) {
+        if (a & 0x8000) {
+            res = 0x8000;
+        } else {
+            res = 0x7fff;
+        }
+    }
+    return res;
+}
+
+/* Perform 8-bit signed saturating subtraction.  */
+static inline uint8_t sub8_sat(uint8_t a, uint8_t b)
+{
+    uint8_t res;
+
+    res = a - b;
+    if (((res ^ a) & 0x80) && ((a ^ b) & 0x80)) {
+        if (a & 0x80) {
+            res = 0x80;
+        } else {
+            res = 0x7f;
+        }
+    }
+    return res;
+}
+
+#define ADD16(a, b, n) RESULT(add16_sat(a, b), n, 16);
+#define SUB16(a, b, n) RESULT(sub16_sat(a, b), n, 16);
+#define ADD8(a, b, n)  RESULT(add8_sat(a, b), n, 8);
+#define SUB8(a, b, n)  RESULT(sub8_sat(a, b), n, 8);
+#define PFX q
+
+#include "op_addsub.c.inc"
+
+/* Unsigned saturating arithmetic.  */
+static inline uint16_t add16_usat(uint16_t a, uint16_t b)
+{
+    uint16_t res;
+    res = a + b;
+    if (res < a) {
+        res = 0xffff;
+    }
+    return res;
+}
+
+static inline uint16_t sub16_usat(uint16_t a, uint16_t b)
+{
+    if (a > b) {
+        return a - b;
+    } else {
+        return 0;
+    }
+}
+
+static inline uint8_t add8_usat(uint8_t a, uint8_t b)
+{
+    uint8_t res;
+    res = a + b;
+    if (res < a) {
+        res = 0xff;
+    }
+    return res;
+}
+
+static inline uint8_t sub8_usat(uint8_t a, uint8_t b)
+{
+    if (a > b) {
+        return a - b;
+    } else {
+        return 0;
+    }
+}
+
+#define ADD16(a, b, n) RESULT(add16_usat(a, b), n, 16);
+#define SUB16(a, b, n) RESULT(sub16_usat(a, b), n, 16);
+#define ADD8(a, b, n)  RESULT(add8_usat(a, b), n, 8);
+#define SUB8(a, b, n)  RESULT(sub8_usat(a, b), n, 8);
+#define PFX uq
+
+#include "op_addsub.c.inc"
+
+/* Signed modulo arithmetic.  */
+#define SARITH16(a, b, n, op) do { \
+    int32_t sum; \
+    sum = (int32_t)(int16_t)(a) op (int32_t)(int16_t)(b); \
+    RESULT(sum, n, 16); \
+    if (sum >= 0) \
+        ge |= 3 << (n * 2); \
+    } while (0)
+
+#define SARITH8(a, b, n, op) do { \
+    int32_t sum; \
+    sum = (int32_t)(int8_t)(a) op (int32_t)(int8_t)(b); \
+    RESULT(sum, n, 8); \
+    if (sum >= 0) \
+        ge |= 1 << n; \
+    } while (0)
+
+
+#define ADD16(a, b, n) SARITH16(a, b, n, +)
+#define SUB16(a, b, n) SARITH16(a, b, n, -)
+#define ADD8(a, b, n)  SARITH8(a, b, n, +)
+#define SUB8(a, b, n)  SARITH8(a, b, n, -)
+#define PFX s
+#define ARITH_GE
+
+#include "op_addsub.c.inc"
+
+/* Unsigned modulo arithmetic.  */
+#define ADD16(a, b, n) do { \
+    uint32_t sum; \
+    sum = (uint32_t)(uint16_t)(a) + (uint32_t)(uint16_t)(b); \
+    RESULT(sum, n, 16); \
+    if ((sum >> 16) == 1) \
+        ge |= 3 << (n * 2); \
+    } while (0)
+
+#define ADD8(a, b, n) do { \
+    uint32_t sum; \
+    sum = (uint32_t)(uint8_t)(a) + (uint32_t)(uint8_t)(b); \
+    RESULT(sum, n, 8); \
+    if ((sum >> 8) == 1) \
+        ge |= 1 << n; \
+    } while (0)
+
+#define SUB16(a, b, n) do { \
+    uint32_t sum; \
+    sum = (uint32_t)(uint16_t)(a) - (uint32_t)(uint16_t)(b); \
+    RESULT(sum, n, 16); \
+    if ((sum >> 16) == 0) \
+        ge |= 3 << (n * 2); \
+    } while (0)
+
+#define SUB8(a, b, n) do { \
+    uint32_t sum; \
+    sum = (uint32_t)(uint8_t)(a) - (uint32_t)(uint8_t)(b); \
+    RESULT(sum, n, 8); \
+    if ((sum >> 8) == 0) \
+        ge |= 1 << n; \
+    } while (0)
+
+#define PFX u
+#define ARITH_GE
+
+#include "op_addsub.c.inc"
+
+/* Halved signed arithmetic.  */
+#define ADD16(a, b, n) \
+  RESULT(((int32_t)(int16_t)(a) + (int32_t)(int16_t)(b)) >> 1, n, 16)
+#define SUB16(a, b, n) \
+  RESULT(((int32_t)(int16_t)(a) - (int32_t)(int16_t)(b)) >> 1, n, 16)
+#define ADD8(a, b, n) \
+  RESULT(((int32_t)(int8_t)(a) + (int32_t)(int8_t)(b)) >> 1, n, 8)
+#define SUB8(a, b, n) \
+  RESULT(((int32_t)(int8_t)(a) - (int32_t)(int8_t)(b)) >> 1, n, 8)
+#define PFX sh
+
+#include "op_addsub.c.inc"
+
+/* Halved unsigned arithmetic.  */
+#define ADD16(a, b, n) \
+  RESULT(((uint32_t)(uint16_t)(a) + (uint32_t)(uint16_t)(b)) >> 1, n, 16)
+#define SUB16(a, b, n) \
+  RESULT(((uint32_t)(uint16_t)(a) - (uint32_t)(uint16_t)(b)) >> 1, n, 16)
+#define ADD8(a, b, n) \
+  RESULT(((uint32_t)(uint8_t)(a) + (uint32_t)(uint8_t)(b)) >> 1, n, 8)
+#define SUB8(a, b, n) \
+  RESULT(((uint32_t)(uint8_t)(a) - (uint32_t)(uint8_t)(b)) >> 1, n, 8)
+#define PFX uh
+
+#include "op_addsub.c.inc"
+
+static inline uint8_t do_usad(uint8_t a, uint8_t b)
+{
+    if (a > b) {
+        return a - b;
+    } else {
+        return b - a;
+    }
+}
+
+/* Unsigned sum of absolute byte differences.  */
+uint32_t HELPER(usad8)(uint32_t a, uint32_t b)
+{
+    uint32_t sum;
+    sum = do_usad(a, b);
+    sum += do_usad(a >> 8, b >> 8);
+    sum += do_usad(a >> 16, b >> 16);
+    sum += do_usad(a >> 24, b >> 24);
+    return sum;
+}
+
+/* For ARMv6 SEL instruction.  */
+uint32_t HELPER(sel_flags)(uint32_t flags, uint32_t a, uint32_t b)
+{
+    uint32_t mask;
+
+    mask = 0;
+    if (flags & 1) {
+        mask |= 0xff;
+    }
+    if (flags & 2) {
+        mask |= 0xff00;
+    }
+    if (flags & 4) {
+        mask |= 0xff0000;
+    }
+    if (flags & 8) {
+        mask |= 0xff000000;
+    }
+    return (a & mask) | (b & ~mask);
+}
+
+/*
+ * CRC helpers.
+ * The upper bytes of val (above the number specified by 'bytes') must have
+ * been zeroed out by the caller.
+ */
+uint32_t HELPER(crc32)(uint32_t acc, uint32_t val, uint32_t bytes)
+{
+    uint8_t buf[4];
+
+    stl_le_p(buf, val);
+
+    /* zlib crc32 converts the accumulator and output to one's complement.  */
+    return crc32(acc ^ 0xffffffff, buf, bytes) ^ 0xffffffff;
+}
+
+uint32_t HELPER(crc32c)(uint32_t acc, uint32_t val, uint32_t bytes)
+{
+    uint8_t buf[4];
+
+    stl_le_p(buf, val);
+
+    /* Linux crc32c converts the output to one's complement.  */
+    return crc32c(acc, buf, bytes) ^ 0xffffffff;
+}
diff --git a/target/arm/tcg/meson.build b/target/arm/tcg/meson.build
index 09238989c5..1f9077c372 100644
--- a/target/arm/tcg/meson.build
+++ b/target/arm/tcg/meson.build
@@ -40,6 +40,7 @@ arm_ss.add(files(
   'tlb_helper.c',
   'vec_helper.c',
   'tlb-insns.c',
+  'arith_helper.c',
 ))
 
 arm_ss.add(when: 'TARGET_AARCH64', if_true: files(
diff --git a/target/arm/tcg/op_addsub.c.inc b/target/arm/tcg/op_addsub.c.inc
new file mode 100644
index 0000000000..ca4a1893c3
--- /dev/null
+++ b/target/arm/tcg/op_addsub.c.inc
@@ -0,0 +1,103 @@
+/*
+ * ARMv6 integer SIMD operations.
+ *
+ * Copyright (c) 2007 CodeSourcery.
+ * Written by Paul Brook
+ *
+ * This code is licensed under the GPL.
+ */
+
+#ifdef ARITH_GE
+#define GE_ARG , void *gep
+#define DECLARE_GE uint32_t ge = 0
+#define SET_GE *(uint32_t *)gep = ge
+#else
+#define GE_ARG
+#define DECLARE_GE do{}while(0)
+#define SET_GE do{}while(0)
+#endif
+
+#define RESULT(val, n, width) \
+    res |= ((uint32_t)(glue(glue(uint,width),_t))(val)) << (n * width)
+
+uint32_t HELPER(glue(PFX,add16))(uint32_t a, uint32_t b GE_ARG)
+{
+    uint32_t res = 0;
+    DECLARE_GE;
+
+    ADD16(a, b, 0);
+    ADD16(a >> 16, b >> 16, 1);
+    SET_GE;
+    return res;
+}
+
+uint32_t HELPER(glue(PFX,add8))(uint32_t a, uint32_t b GE_ARG)
+{
+    uint32_t res = 0;
+    DECLARE_GE;
+
+    ADD8(a, b, 0);
+    ADD8(a >> 8, b >> 8, 1);
+    ADD8(a >> 16, b >> 16, 2);
+    ADD8(a >> 24, b >> 24, 3);
+    SET_GE;
+    return res;
+}
+
+uint32_t HELPER(glue(PFX,sub16))(uint32_t a, uint32_t b GE_ARG)
+{
+    uint32_t res = 0;
+    DECLARE_GE;
+
+    SUB16(a, b, 0);
+    SUB16(a >> 16, b >> 16, 1);
+    SET_GE;
+    return res;
+}
+
+uint32_t HELPER(glue(PFX,sub8))(uint32_t a, uint32_t b GE_ARG)
+{
+    uint32_t res = 0;
+    DECLARE_GE;
+
+    SUB8(a, b, 0);
+    SUB8(a >> 8, b >> 8, 1);
+    SUB8(a >> 16, b >> 16, 2);
+    SUB8(a >> 24, b >> 24, 3);
+    SET_GE;
+    return res;
+}
+
+uint32_t HELPER(glue(PFX,subaddx))(uint32_t a, uint32_t b GE_ARG)
+{
+    uint32_t res = 0;
+    DECLARE_GE;
+
+    ADD16(a, b >> 16, 0);
+    SUB16(a >> 16, b, 1);
+    SET_GE;
+    return res;
+}
+
+uint32_t HELPER(glue(PFX,addsubx))(uint32_t a, uint32_t b GE_ARG)
+{
+    uint32_t res = 0;
+    DECLARE_GE;
+
+    SUB16(a, b >> 16, 0);
+    ADD16(a >> 16, b, 1);
+    SET_GE;
+    return res;
+}
+
+#undef GE_ARG
+#undef DECLARE_GE
+#undef SET_GE
+#undef RESULT
+
+#undef ARITH_GE
+#undef PFX
+#undef ADD16
+#undef SUB16
+#undef ADD8
+#undef SUB8