Merge pull request #253 from fastfloat/update_ci_to_ubuntu24

update CI to ubuntu 24 + safe a shift value to a variable (for elegance)

Author: lemire

.github/workflows/ubuntu22-gcc13.yml renamed to .github/workflows/ubuntu24.yml

@@ -1,23 +1,23 @@
-name: Ubuntu 22.04 CI (GCC 13)
+name: Ubuntu 24.04 CI (GCC 13)
 
 on: [push, pull_request]
 
 jobs:
   ubuntu-build:
-    runs-on: ubuntu-22.04
+    runs-on: ubuntu-24.04
     steps:
       - uses: actions/checkout@v3
       - name: Use cmake
         run: |
           mkdir build &&
           cd build &&
-          CXX=g++-13 CXXFLAGS=-Werror cmake -DFASTFLOAT_TEST=ON  ..  &&
+          CXXFLAGS=-Werror cmake -DFASTFLOAT_TEST=ON  ..  &&
           cmake --build .   &&
           ctest --output-on-failure
       - name: Use cmake CXX23
         run: |
           mkdir build20 &&
           cd build20 &&
-          CXX=g++-13 CXXFLAGS=-Werror cmake -DFASTFLOAT_CONSTEXPR_TESTS=ON -DFASTFLOAT_FIXEDWIDTH_TESTS=ON -DFASTFLOAT_CXX_STANDARD=23 -DFASTFLOAT_TEST=ON  ..  &&
+          CXXFLAGS=-Werror cmake -DFASTFLOAT_CONSTEXPR_TESTS=ON -DFASTFLOAT_FIXEDWIDTH_TESTS=ON -DFASTFLOAT_CXX_STANDARD=23 -DFASTFLOAT_TEST=ON  ..  &&
           cmake --build .   &&
           ctest --output-on-failure

include/fast_float/decimal_to_binary.h

@@ -127,8 +127,9 @@ adjusted_mantissa compute_float(int64_t q, uint64_t w)  noexcept  {
   // but in practice, we can win big with the compute_product_approximation if its additional branch
   // is easily predicted. Which is best is data specific.
   int upperbit = int(product.high >> 63);
+  int shift = upperbit + 64 - binary::mantissa_explicit_bits() - 3;
 
-  answer.mantissa = product.high >> (upperbit + 64 - binary::mantissa_explicit_bits() - 3);
+  answer.mantissa = product.high >> shift;
 
   answer.power2 = int32_t(detail::power(int32_t(q)) + upperbit - lz - binary::minimum_exponent());
   if (answer.power2 <= 0) { // we have a subnormal?
@@ -164,7 +165,7 @@ adjusted_mantissa compute_float(int64_t q, uint64_t w)  noexcept  {
     // To be in-between two floats we need that in doing
     //   answer.mantissa = product.high >> (upperbit + 64 - binary::mantissa_explicit_bits() - 3);
     // ... we dropped out only zeroes. But if this happened, then we can go back!!!
-    if((answer.mantissa  << (upperbit + 64 - binary::mantissa_explicit_bits() - 3)) ==  product.high) {
+    if((answer.mantissa  << shift) ==  product.high) {
       answer.mantissa &= ~uint64_t(1);          // flip it so that we do not round up
     }
   }