No, you don't want a sine lookup table as if you had one you'd have to do two reverse lookups to get the angle then a bunch of comparisons to get U,V and W.
How fast do you want? Throw a large enough ROM lookup table at the problem and you can generate U,V & W signals as fast as you can read two channels via the ADC. Just concatenate the two ADC results with some of the low bits sliced off and use them as the address in the lookup table. The table stores precomputed U,V W and an error bit in a nibble. The error bit is set for addresses that aren't close enough to the expected sine/cosine circle from the encoder, and indicate an encoder fault. To save memory, pack the table two nibbles to a byte and slice off one bit of the combined address to select which nibble to return.
Throw hardware at the problem e.g. two Flash ADCs, a fast ROM and a sample clock generator, and you could update U, V & W at well over 10MHz!