github/concrete
1
1
Fork 0
mirror of https://github.com/zama-ai/concrete.git synced 2026-04-17 03:00:54 -04:00
Code Issues Packages Projects Releases Wiki Activity

feat(compiler): Add support for full unrolling of loops with SDFG-convertible ops

Browse Source 
This adds a new option `--unroll-loops-with-sdfg-convertible-ops`,
which causes loops containing SDFG-convertible operations to be fully
unrolled upon the extraction of SDFG-operations using the
`--emit-sdfg-ops` switch. This avoids constant roundtrips between an
SDFG-capable accelerator and the host during execution of a loop.

The option is limited to `scf.for` loops with static bounds and a
static step size. Since full unrolling of loops with large bounds
results in a large number of operations, the option is disabled by
default.
This commit is contained in:
Andi Drebes
2022-12-07 10:26:29 +01:00
committed by Andi Drebes
parent 2dce654406
commit e2e6df322e
9 changed files with 160 additions and 17 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
compiler/src/main.cpp
View File

@@ -183,6 +183,12 @@ llvm::cl::opt<bool> emitSDFGOps(
" graphs and emit them."),
llvm::cl::init(false));
llvm::cl::opt<bool> unrollLoopsWithSDFGConvertibleOps(
"unroll-loops-with-sdfg-convertible-ops",
llvm::cl::desc("Causes loops containing SDFG-convertible operations to be "
"fully unrolled."),
llvm::cl::init(false));
llvm::cl::opt<bool> dataflowParallelize(
"parallelize-dataflow",
llvm::cl::desc(
@@ -316,6 +322,8 @@ cmdlineCompilationOptions() {
options.dataflowParallelize = cmdline::dataflowParallelize;
options.batchConcreteOps = cmdline::batchConcreteOps;
options.emitSDFGOps = cmdline::emitSDFGOps;
options.unrollLoopsWithSDFGConvertibleOps =
cmdline::unrollLoopsWithSDFGConvertibleOps;
options.optimizeConcrete = cmdline::optimizeConcrete;
options.emitGPUOps = cmdline::emitGPUOps;