Compiler method for generating instructions for vector operations on a multi-endian processor

Patent metadata

Field	Value
Publication number	US-9588746-B2
Application number	US-201414576391-A
Country	US
Kind code	B2
Filing date	Dec 19, 2014
Priority date	Dec 19, 2014
Publication date	Mar 7, 2017
Grant date	Mar 7, 2017

Abstract

A compiler includes a vector instruction processing mechanism that generates instructions for vector instructions in a way that assures correct operation in a bi-endian environment, wherein the processor architecture contains instructions with an inherent endian bias. The compiler uses a code generation endian preference that is specified by the user, and that determines a natural element order. When the compiler processes a computer program, it generates instructions for vector operations by determining whether the vector instruction has an endian bias that matches the specified endian preference (i.e., the inherent element order of the vector instruction matches the natural element order). When the vector instruction has no endian bias, or when the endian bias of the vector instruction matches the specified endian preference, the compiler generates one or more instructions for the vector instruction as it normally does. When the endian bias of the vector instruction does not match the specified endian preference, the compiler generates instructions to fix the mismatch.

First claim

The invention claimed is:

1. A computer-implemented method executed by at least one processor for a compiler to process a plurality of instructions in a computer program, the method comprising:
the compiler reading the plurality of instructions;
the compiler determining an endian preference that defines a natural element order for vector instructions in the plurality of instructions; and
the compiler determining when a vector instruction in the plurality of instructions has an inherent element order that is a mismatch to the natural element order, and in response, generating at least one instruction to fix the mismatch.

2. The method of claim 1 further comprising determining when the vector instruction specifies a first element number, and in response, generating an instruction that references a second element number computed by subtracting the first element number from a number of elements in the vector minus one.

3. The method of claim 1 further comprising determining when the vector instruction specifies odd elements, and in response, generating an instruction that specifies even elements, and determining when the vector instruction specifies even elements, and in response, generating an instruction that specifies odd elements.

4. The method of claim 1 further comprising determining when the vector instruction is a vector load instruction, and in response, generating a vector element reverse instruction after the vector load instruction.

5. The method of claim 1 further comprising determining when the vector instruction is a vector store instruction, and in response, generating a vector element reverse instruction before the vector store instruction.

6. The method of claim 4 wherein each vector element reverse instruction reverses order of a plurality of elements of a vector register.

7. The method of claim 6 wherein the plurality of elements of the vector register comprises one of:
a plurality of bytes;
a plurality of halfwords;
a plurality of words;
a plurality of double-words;
a plurality of quadwords; and
a plurality of elements larger than quadwords.

8. The method of claim 1 further comprising determining when the vector instruction is an instruction that has input arguments treated as an extended vector, and in response, generating a vector instruction with an inverted order of the input arguments.

9. The method of claim 1 further comprising determining when the vector instruction refers to a high half of at least one vector register, and in response, generating an instruction that refers to a low half of the at least one vector register, and determining when the vector instruction refers to a low half of at least one vector register, and in response, generating an instruction that refers to the high half of the at least one vector register.

10. A computer-implemented method executed by at least one processor for a compiler to process a plurality of instructions in a computer program, the method comprising the steps of:
determining an endian preference that defines a natural element order for vector instructions in the plurality of instructions;
selecting a vector instruction in the plurality of instructions;
when the selected vector instruction has an inherent element order that does not match the natural element order:
determining when the vector instruction specifies a first element number, and in response, generating an instruction that references a second element number computed by subtracting the first element number from a number of elements in the vector minus one;
determining when the vector instruction specifies odd elements, and in response, generating an instruction that specifies even elements;
determining when the vector instruction specifies even elements, and in response, generating an instruction that specifies odd elements;
determining when the vector instruction is a vector load instruction, and in response, generating a vector element reverse instruction after the vector load instruction;
determining when the vector instruction is a vector store instruction, and in response, generating a vector element reverse instruction before the vector store instruction;
wherein each vector element reverse instruction reverses order of a plurality of elements of a vector register, wherein the plurality of elements of the vector register comprises one of:
a plurality of bytes;
a plurality of halfwords;
a plurality of words; and
a plurality of double-words;
determining when the vector instruction is an instruction that has input arguments treated as an extended vector, and in response, generating a vector instruction with an inverted order of the input arguments;
determining when the vector instruction refers to a high half of at least one vector register, and in response, generating an instruction that refers to a low half of the at least one vector register; and
determining when the vector instruction refers to the low half of at least one vector register, and in response, generating an instruction that refers to the high half of the at least one vector register.

Assignees

• IBM

Inventors

• Gschwind Michael Karl
• Ji Jin Song
• Schmidt William J

Classifications

• G06F8/445 Primary

  Exploiting fine grain parallelism, i.e. parallelism at instruction level (run-time instruction scheduling G06F9/3836) · CPC title

• G06F9/30145

  Instruction analysis, e.g. decoding, instruction word fields · CPC title

• G06F8/41 Primary

  Compilation · CPC title

• G06F9/30036

  Instructions to perform operations on packed data, e.g. vector, tile or matrix operations · CPC title