AMD amd64 architecture User Manual
102
General-Purpose Programming
AMD64 Technology
24592—Rev. 3.15—November 2009
For data that will be used only once in a procedure, consider using non-temporal accesses. Such
accesses are not burdened by the overhead of cache protocols.
accesses are not burdened by the overhead of cache protocols.
3.10.6 Keep Common Operands in Registers
Keep frequently used values in registers rather than in memory. This avoids the comparatively long
latencies for accessing memory.
latencies for accessing memory.
3.10.7 Avoid True Dependencies
Spread out true dependencies (write-read or flow dependencies) to increase the opportunities for
parallel execution. This spreading out is not necessary for anti-dependencies and output dependencies.
parallel execution. This spreading out is not necessary for anti-dependencies and output dependencies.
3.10.8 Avoid Store-to-Load Dependencies
Store-to-load dependencies occur when data is stored to memory, only to be read back shortly
thereafter. Hardware implementations of the architecture may contain means of accelerating such
store-to-load dependencies, allowing the load to obtain the store data before it has been written to
memory. However, this acceleration might be available only when the addresses and operand sizes of
the store and the dependent load are matched, and when both memory accesses are aligned.
Performance is typically optimized by avoiding such dependencies altogether and keeping the data,
including temporary variables, in registers.
thereafter. Hardware implementations of the architecture may contain means of accelerating such
store-to-load dependencies, allowing the load to obtain the store data before it has been written to
memory. However, this acceleration might be available only when the addresses and operand sizes of
the store and the dependent load are matched, and when both memory accesses are aligned.
Performance is typically optimized by avoiding such dependencies altogether and keeping the data,
including temporary variables, in registers.
3.10.9 Optimize Stack Allocation
When allocating space on the stack for local variables and/or outgoing parameters within a procedure,
adjust the stack pointer and use moves rather than pushes. This method of allocation allows random
access to the outgoing parameters, so that they can be set up when they are calculated instead of being
held in a register or memory until the procedure call. This method also reduces stack-pointer
dependencies.
adjust the stack pointer and use moves rather than pushes. This method of allocation allows random
access to the outgoing parameters, so that they can be set up when they are calculated instead of being
held in a register or memory until the procedure call. This method also reduces stack-pointer
dependencies.
3.10.10 Consider Repeat-Prefix Setup Time
The repeat instruction prefixes have a setup overhead. If the repeated count is variable, the overhead
can sometimes be avoided by substituting a simple loop to move or store the data. Repeated string
instructions can be expanded into equivalent sequences of inline loads and stores. For details, see
“Repeat Prefixes” in Volume 3.
can sometimes be avoided by substituting a simple loop to move or store the data. Repeated string
instructions can be expanded into equivalent sequences of inline loads and stores. For details, see
“Repeat Prefixes” in Volume 3.
3.10.11 Replace GPR with Media Instructions
Some integer-based programs can be made to run faster by using 128-bit media or 64-bit media
instructions. These instructions have their own register sets. Because of this, they relieve register
pressure on the GPR registers. For loads, stores, adds, shifts, etc., media instructions may be good
substitutes for general-purpose integer instructions. GPR registers are freed up, and the media
instructions increase opportunities for parallel operations.
instructions. These instructions have their own register sets. Because of this, they relieve register
pressure on the GPR registers. For loads, stores, adds, shifts, etc., media instructions may be good
substitutes for general-purpose integer instructions. GPR registers are freed up, and the media
instructions increase opportunities for parallel operations.