Efficient data compression in processing systems
Certain aspects of the present disclosure provide techniques and apparatus for efficiently performing operations using data compression. An example method generally includes identifying, for a block of data samples, a number of leading bits to remove from each data sample in the block of data samples. A block of compressed data samples is generated based on the identified number of leading bits and truncation of a number of least significant bits from each data sample in the block of data samples. A bitstream including the block of compressed data samples and an indication of a type of compression applied to the block of data samples is generated and output for further processing.
1 . A method for data compression, comprising:
identifying, for a block of data samples, a number of leading bits to remove from each data sample in the block of data samples;
generating a block of compressed data samples based on the identified number of leading bits and truncation of a number of least significant bits from each data sample in the block of data samples;
generating a bitstream including the block of compressed data samples and an indication of a type of compression applied to the block of data samples; and
outputting the generated bitstream.
2 . The method of claim 1 , wherein the indication of the type of compression applied to the block of data samples comprises a first header indicating whether the bitstream includes compressed data and a second header indicating the number of leading bits removed from each data sample in the block of compressed data samples, the second header being located subsequent to the first header in the generated bitstream.
3 . The method of claim 2 , wherein the second header comprises a single field indicating the number of leading bits removed from each data sample in the block of compressed data samples.
4 . The method of claim 3 , further comprising:
receiving the generated bitstream and,
for each respective compressed data sample in the block of compressed data samples in the received bitstream:
adding the number of leading bits as most significant bits to the respective compressed data sample; and
adding the number of truncated least significant bits to the respective compressed data sample.
5 . The method of claim 1 , wherein the indication of the type of compression applied to the block of data samples comprises a first header indicating whether the bitstream includes compressed data and a plurality of second headers, and wherein each second header of the plurality of second headers identifies the number of leading bits removed from each data sample in a corresponding block of compressed data samples.
6 . The method of claim 5 , further comprising:
receiving the generated bitstream; and
for each respective compressed data sample in the block of compressed data samples:
adding the number of leading bits identified in a corresponding second header as most significant bits to the respective compressed data sample;
reconstructing the block of data samples based on the block of compressed data samples and the indicated type of compression applied to the plurality of data samples.
7 . The method of claim 1 , wherein generating the block of compressed data samples comprises, for each respective data sample in the block of data samples:
applying a right shift to the respective data sample according to the number of least significant bits identified for truncation, wherein a most significant bit of the least significant bits identified for truncation is designated as a carry bit; and
adding the carry bit to the right-shifted respective data sample.
8 . A method for data decompression, comprising:
receiving a bitstream including a block of compressed data samples and an indication of a type of compression applied to the block of compressed data samples;
generating a reconstructed data block by reconstructing each respective compressed data sample in the block of compressed data samples based on the indicated type of compression applied to the block of compressed data samples, wherein reconstructing each respective compressed data sample comprises adding a number of leading bits and a number of trailing bits to each respective compressed data sample based on the indicated type of compression applied to the block of compressed data samples; and
outputting the reconstructed data block.
9 . The method of claim 8 , wherein the indication of the type of compression applied to the block of data samples comprises a first header indicating whether the bitstream includes compressed data and a second header indicating the number of leading bits removed from each data sample in the block of compressed data samples, the second header being located subsequent to the first header in the received bitstream.
10 . The method of claim 9 , wherein the second header comprises a single field indicating the number of leading bits to add to each compressed data sample in the block of compressed data samples.
11 . The method of claim 10 , wherein generating the reconstructed data block comprises, for each respective compressed data sample:
adding the number of leading bits as most significant bits to the respective compressed data sample; and
adding the number of trailing bits as least significant bits to the respective compressed data sample.
12 . The method of claim 8 , wherein the indication of the type of compression applied to the block of data samples comprises a first header indicating whether the bitstream includes compressed data and a plurality of second headers, and wherein each second header of the plurality of second headers identifies the number of leading bits removed from each data sample in a corresponding block of compressed data samples.
13 . The method of claim 12 , wherein generating the reconstructed data block comprises, for each respective compressed data sample:
adding the number of leading bits identified in a corresponding second header as most significant bits to the respective compressed data sample; and
reconstructing the block of data samples based on the block of compressed data samples and the indicated type of compression applied to the plurality of data samples.
14 . The method of claim 8 , wherein reconstructing each respective compressed data sample in the block of compressed data samples comprises applying a left shift to the respective data sample according to the number of trailing bits to add to each respective compressed data sample.
15 . An arithmetic shifter for compressing data, comprising:
an arithmetic right shifter configured to shift a value in an input register by a number of bits;
a multiplexer configured to output a carry bit from the number of bits based on a selector signal identifying that a compression operation is being executed; and
an adder configured to add the shifted value generated by the arithmetic right shifter to an output of the multiplexer, wherein a data compression operation executed using the arithmetic shifter comprises an operation executable via invoking a single instruction on a processing system including the arithmetic shifter.