Intel® oneAPI DPC++ FPGA Optimization Guide

Developer Guide

Contents

FPGA Loop Directives

The following table summarizes loop directives:
FPGA Loop Directives
Pragma or Attribute
Description
Example
disable_loop_pipelining
Directs the
Intel® oneAPI
DPC++/C++
 Compiler
 to disable pipelining of a loop. 
[[intel::disable_loop_pipelining]]
for (int i = 1; i < N; i++) {
  int j = a[i-1];
  // Memory dependency induces a high-latency loop feedback path
  a[i] = foo(j)
}
ii
Forces a loop to have a loop initialization interval (II) of a specified value. 
// ii set to 5
[[intel::ii(5)]]
  for (int i = 0; i < N; ++i){
}
ivdep
Ignores memory dependencies between iterations of this loop 
// ivdep loop
[[intel::ivdep]] for (…) {}
//ivdep safelen 
[[intel::ivdep(safelen)]] for (;;) {} 
// ivdep accessor
[[intel::ivdep(accessorA)]] for (;;) {}
//ivdep array safelen
[[intel::ivdep(accessorA, safelen)]]
for (;;){}
loop_coalesce
Coalesces nested loops into a single loop without affecting the loop functionality. 
[[intel::loop_coalesce(2)]] 
for (int i = 0; i < N; i++)
  for (int j = 0; j < M; j++)
    sum[i][j] += i+j;
max_concurrency
Limits the number of iterations of a loop that can simultaneously execute at any time. 
//max concurrency set to 1
[[intel::max_concurrency(1)]] 
  for (int i = 0; i < c; ++i){
}
max_interleaving
Maximizes the throughput and hardware resource occupancy of pipelined inner loops in a loop nest. 
// Loop j is pipelined with ii=1
for (int j = 0; j < M; j++) {
  int a[N];
  // Loop i is pipelined with ii=2 
  [[intel::max_interleaving(1)]]
  for (int i = 1; i < N; i++) {
    a[i] = foo(i)
  }
  …
}
speculated_iterations
Improves the performance of pipelined loops. 
[[intel::speculated_iterations(1)]]
  while (m*m*m < N) {
    m += 1;
  }
  dst[0] = m;
unroll
Unrolls a loop in the kernel code. 
// unroll factor N set to 2
#pragma unroll 2
for(size_t k = 0; k < 4; k++){
  mac += data_in[(gid * 4) + k] * coeff[k];
}

Product and Performance Information

1

Performance varies by use, configuration and other factors. Learn more at www.Intel.com/PerformanceIndex.