Refine Montgomery GPU helpers to avoid kernel branching
This commit is contained in:
1 parent
f12919e220
commit
24eaec7a21
5 files changed
+46
-22
No files matched your search
@@ -17,7 +17,7 @@ internal static class DivisorByDivisorKernels
|
||||
}
|
||||
|
||||
ulong exponent = exponents[index];
|
||||
hits[index] = ULongExtensions.Pow2MontgomeryModWindowedGpu(divisor, exponent, keepMontgomery: false) == 1UL ? (byte)1 : (byte)0;
|
||||
hits[index] = ULongExtensions.Pow2MontgomeryModWindowedGpuConvertToStandard(divisor, exponent) == 1UL ? (byte)1 : (byte)0;
|
||||
}
|
||||
|
||||
public static void ComputeRemainderDeltasKernel(Index1D index, ArrayView<ulong> gaps, ArrayView<byte> deltas, byte modulus)
|
||||
@@ -149,6 +149,6 @@ internal static class DivisorByDivisorKernels
|
||||
|
||||
ulong exponent = exponents[index];
|
||||
|
||||
results[index] = ULongExtensions.Pow2MontgomeryModWindowedGpu(divisor, exponent, keepMontgomery: true);
|
||||
results[index] = ULongExtensions.Pow2MontgomeryModWindowedGpuKeepMontgomery(divisor, exponent);
|
||||
}
|
||||
}
|
||||
Reference in new issue
Block a user