mtl/src/gba/fixed.cpp
2024-08-03 16:18:20 -06:00

61 lines
2.0 KiB
C++

#include "mtl/target.hpp"
#include "mtl/fixed.hpp"
TARGET_ARM_MODE
namespace mtl {
fixed fixed::operator/(fixed rhs) const noexcept {
int32_t raw_result;
asm(
// This division implementation has two methods it can use.
// The fastest uses a left shift followed by a single division. The value is shifted
// first to preserve the decimal part. Unfortunately, this means large numerators
// will cause the operation to overflow. In this case, a compatible method will be
// used. This method uses two divisions, one to calculate the integral quotient,
// and one to calculate the decimal part. Both these methods work for negative numbers as well.
".arm;"
"movs r1, %[d];" // Load numerator and denominator, and check if negative or zero
"beq 4f;"
"movs r0, %[n];"
"blt 1f;"
"tst r0, #0x7e000000;" // Check if the numerator is large enough to overflow
"bne 3f;"
"b 2f;"
"1:" // check_negative
"mvn r2, r0;" // Check if the numerator is large enough to overflow.
"tst r2, #0x7e000000;"
"bne 3f;"
"2:" // fast_div // Fast method
"lsl r0, #6;" // Shift first to avoid truncation
"swi #0x60000;" // GBA Div syscall
"mov %[res], r0;"
"b 5f;"
"3:" // compat_div // Compatible method
"swi #0x60000;" // Compute quotient and shift
"lsl r2, r0, #6;"
"mov r0, r1;" // Div syscall puts the modulus in r1, use it as the numerator
"lsr r1, %[d], #6;" // Load the denominator again, shifted right to calculate decimal part
"swi #0x60000;"
"mov %[res], r2;" // Calculate the final result
"add %[res], r0;"
"b 5f;"
"4:" // zero_div
"teq %[n], %[d];" // Set result to largest possible negative/positive value.
"movmi %[res], #0x80000000;"
"movpl %[res], #0x7FFFFFFF;"
"5:"
: [res] "=r" (raw_result)
: [n] "r" (x),
[d] "r" (rhs.x)
: "r0", "r1", "r2", "r3"
);
return from_raw(raw_result);
}
} // namespace mtl
TARGET_END_MODE