#include <map>
#include <algorithm>
#include <memory>
+#include <functional>
#include <sstream>
#include <backends/cxxrtl/cxxrtl_capi.h>
+#ifndef __has_attribute
+# define __has_attribute(x) 0
+#endif
+
// CXXRTL essentially uses the C++ compiler as a hygienic macro engine that feeds an instruction selector.
// It generates a lot of specialized template functions with relatively large bodies that, when inlined
// into the caller and (for those with loops) unrolled, often expose many new optimization opportunities.
// Because of this, most of the CXXRTL runtime must be always inlined for best performance.
-#ifndef __has_attribute
-# define __has_attribute(x) 0
-#endif
#if __has_attribute(always_inline)
#define CXXRTL_ALWAYS_INLINE inline __attribute__((__always_inline__))
#else
#define CXXRTL_ALWAYS_INLINE inline
#endif
+// Conversely, some functions in the generated code are extremely large yet very cold, with both of these
+// properties being extreme enough to confuse C++ compilers into spending pathological amounts of time
+// on a futile (the code becomes worse) attempt to optimize the least important parts of code.
+#if __has_attribute(optnone)
+#define CXXRTL_EXTREMELY_COLD __attribute__((__optnone__))
+#elif __has_attribute(optimize)
+#define CXXRTL_EXTREMELY_COLD __attribute__((__optimize__(0)))
+#else
+#define CXXRTL_EXTREMELY_COLD
+#endif
+
+// CXXRTL uses assert() to check for C++ contract violations (which may result in e.g. undefined behavior
+// of the simulation code itself), and CXXRTL_ASSERT to check for RTL contract violations (which may at
+// most result in undefined simulation results).
+//
+// Though by default, CXXRTL_ASSERT() expands to assert(), it may be overridden e.g. when integrating
+// the simulation into another process that should survive violating RTL contracts.
+#ifndef CXXRTL_ASSERT
+#ifndef CXXRTL_NDEBUG
+#define CXXRTL_ASSERT(x) assert(x)
+#else
+#define CXXRTL_ASSERT(x)
+#endif
+#endif
namespace cxxrtl {
return sext_cast<NewBits>()(*this);
}
+ // Bit replication is far more efficient than the equivalent concatenation.
+ template<size_t Count>
+ CXXRTL_ALWAYS_INLINE
+ value<Bits * Count> repeat() const {
+ static_assert(Bits == 1, "repeat() is implemented only for 1-bit values");
+ return *this ? value<Bits * Count>().bit_not() : value<Bits * Count>();
+ }
+
// Operations with run-time parameters (offsets, amounts, etc).
//
// These operations are used for computations.
value<Bits> next;
wire() = default;
- constexpr wire(const value<Bits> &init) : curr(init), next(init) {}
+ explicit constexpr wire(const value<Bits> &init) : curr(init), next(init) {}
template<typename... Init>
explicit constexpr wire(Init ...init) : curr{init...}, next{init...} {}
// Tag class to disambiguate values/wires and their aliases.
struct debug_alias {};
+// Tag declaration to disambiguate values and debug outlines.
+using debug_outline = ::_cxxrtl_outline;
+
// This structure is intended for consumption via foreign function interfaces, like Python's ctypes.
// Because of this it uses a C-style layout that is easy to parse rather than more idiomatic C++.
//
struct debug_item : ::cxxrtl_object {
// Object types.
enum : uint32_t {
- VALUE = CXXRTL_VALUE,
- WIRE = CXXRTL_WIRE,
- MEMORY = CXXRTL_MEMORY,
- ALIAS = CXXRTL_ALIAS,
+ VALUE = CXXRTL_VALUE,
+ WIRE = CXXRTL_WIRE,
+ MEMORY = CXXRTL_MEMORY,
+ ALIAS = CXXRTL_ALIAS,
+ OUTLINE = CXXRTL_OUTLINE,
};
// Object flags.
zero_at = 0;
curr = item.data;
next = item.data;
+ outline = nullptr;
}
template<size_t Bits>
zero_at = 0;
curr = const_cast<chunk_t*>(item.data);
next = nullptr;
+ outline = nullptr;
}
template<size_t Bits>
zero_at = 0;
curr = item.curr.data;
next = item.next.data;
+ outline = nullptr;
}
template<size_t Width>
zero_at = zero_offset;
curr = item.data.empty() ? nullptr : item.data[0].data;
next = nullptr;
+ outline = nullptr;
}
template<size_t Bits>
zero_at = 0;
curr = const_cast<chunk_t*>(item.data);
next = nullptr;
+ outline = nullptr;
}
template<size_t Bits>
zero_at = 0;
curr = const_cast<chunk_t*>(item.curr.data);
next = nullptr;
+ outline = nullptr;
+ }
+
+ template<size_t Bits>
+ debug_item(debug_outline &group, const value<Bits> &item, size_t lsb_offset = 0) {
+ static_assert(sizeof(item) == value<Bits>::chunks * sizeof(chunk_t),
+ "value<Bits> is not compatible with C layout");
+ type = OUTLINE;
+ flags = DRIVEN_COMB;
+ width = Bits;
+ lsb_at = lsb_offset;
+ depth = 1;
+ zero_at = 0;
+ curr = const_cast<chunk_t*>(item.data);
+ next = nullptr;
+ outline = &group;
}
};
static_assert(std::is_standard_layout<debug_item>::value, "debug_item is not compatible with C layout");
} // namespace cxxrtl
-// Internal structure used to communicate with the implementation of the C interface.
+// Internal structures used to communicate with the implementation of the C interface.
+
typedef struct _cxxrtl_toplevel {
std::unique_ptr<cxxrtl::module> module;
} *cxxrtl_toplevel;
+typedef struct _cxxrtl_outline {
+ std::function<void()> eval;
+} *cxxrtl_outline;
+
// Definitions of internal Yosys cells. Other than the functions in this namespace, CXXRTL is fully generic
// and indepenent of Yosys implementation details.
//
template<size_t BitsY, size_t BitsA, size_t BitsB>
CXXRTL_ALWAYS_INLINE
value<BitsY> shl_uu(const value<BitsA> &a, const value<BitsB> &b) {
- return a.template zcast<BitsY>().template shl(b);
+ return a.template zcast<BitsY>().shl(b);
}
template<size_t BitsY, size_t BitsA, size_t BitsB>
CXXRTL_ALWAYS_INLINE
value<BitsY> shl_su(const value<BitsA> &a, const value<BitsB> &b) {
- return a.template scast<BitsY>().template shl(b);
+ return a.template scast<BitsY>().shl(b);
}
template<size_t BitsY, size_t BitsA, size_t BitsB>
CXXRTL_ALWAYS_INLINE
value<BitsY> sshl_uu(const value<BitsA> &a, const value<BitsB> &b) {
- return a.template zcast<BitsY>().template shl(b);
+ return a.template zcast<BitsY>().shl(b);
}
template<size_t BitsY, size_t BitsA, size_t BitsB>
CXXRTL_ALWAYS_INLINE
value<BitsY> sshl_su(const value<BitsA> &a, const value<BitsB> &b) {
- return a.template scast<BitsY>().template shl(b);
+ return a.template scast<BitsY>().shl(b);
}
template<size_t BitsY, size_t BitsA, size_t BitsB>
CXXRTL_ALWAYS_INLINE
value<BitsY> shr_uu(const value<BitsA> &a, const value<BitsB> &b) {
- return a.template shr(b).template zcast<BitsY>();
+ return a.shr(b).template zcast<BitsY>();
}
template<size_t BitsY, size_t BitsA, size_t BitsB>
CXXRTL_ALWAYS_INLINE
value<BitsY> shr_su(const value<BitsA> &a, const value<BitsB> &b) {
- return a.template shr(b).template scast<BitsY>();
+ return a.shr(b).template scast<BitsY>();
}
template<size_t BitsY, size_t BitsA, size_t BitsB>
CXXRTL_ALWAYS_INLINE
value<BitsY> sshr_uu(const value<BitsA> &a, const value<BitsB> &b) {
- return a.template shr(b).template zcast<BitsY>();
+ return a.shr(b).template zcast<BitsY>();
}
template<size_t BitsY, size_t BitsA, size_t BitsB>
CXXRTL_ALWAYS_INLINE
value<BitsY> sshr_su(const value<BitsA> &a, const value<BitsB> &b) {
- return a.template sshr(b).template scast<BitsY>();
+ return a.sshr(b).template scast<BitsY>();
}
template<size_t BitsY, size_t BitsA, size_t BitsB>
projects / yosys.git / blobdiff