[tlively]

Move makeSignExt, makePackedFieldGet, getSignExtValue, getAlmostSignExt, getZeroExtValue, Provider::getMaxBits, and Provider::getSignExtKeptBits out of src/ir/bits.h into src/ir/bits.cpp, and add missing direct #includes to files that previously relied on transitive includes from bits.h.

[gemini-code-assist]

Code Review

This pull request introduces a new src/ir/bits.cpp file to centralize bit manipulation utilities and updates various headers and build configurations to support this module. The review highlighted a missing direct include for ir/properties.h and identified precision loss issues in the bit-tracking logic for RemSInt32 and RemSInt64 when dealing with negative divisors, providing actionable code suggestions to resolve these bugs.

[gemini-code-assist]

The header ir/boolean.h is included but not directly used in this file. Instead, Properties::emitsBoolean is used, which is defined in ir/properties.h. To adhere to the principle of direct includes and avoid relying on transitive includes, ir/properties.h should be included directly.

Suggested change

	#include "ir/boolean.h"
	#include "ir/properties.h"

[gemini-code-assist]

When the divisor c is negative, c->value.geti32() returns a negative value. Passing a negative value to ceilLog2 (which expects an unsigned value) causes it to be cast to a very large unsigned integer, resulting in ceilLog2 returning 32 and losing precision in the bit-tracking analysis. Using the absolute value of the divisor avoids this precision loss.

      case RemSInt32: {
        if (auto* c = binary->right->dynCast<Const>()) {
          auto maxBitsLeft = getMaxBits(binary->left);
          // if left may be negative, the result may be negative
          if (maxBitsLeft == 32) {
            return 32;
          }
          auto val = c->value.geti32();
          auto absVal = val < 0 ? -uint32_t(val) : uint32_t(val);
          auto bitsRight = Index(ceilLog2(absVal));
          return std::min(maxBitsLeft, bitsRight);
        }

[gemini-code-assist]

Similar to RemSInt32, when the divisor c is negative, c->value.geti64() returns a negative value. Passing this to ceilLog2 causes a cast to a very large unsigned integer, resulting in ceilLog2 returning 64 and losing precision. Using the absolute value of the divisor avoids this precision loss.

      case RemSInt64: {
        if (auto* c = binary->right->dynCast<Const>()) {
          auto maxBitsLeft = getMaxBits(binary->left);
          // if left may be negative, the result may be negative
          if (maxBitsLeft == 64) {
            return 64;
          }
          auto val = c->value.geti64();
          auto absVal = val < 0 ? -uint64_t(val) : uint64_t(val);
          auto bitsRight = Index(ceilLog2(absVal));
          return std::min(maxBitsLeft, bitsRight);
        }