diff options
Diffstat (limited to 'src/run.cpp')
| -rw-r--r-- | src/run.cpp | 33 |
1 files changed, 25 insertions, 8 deletions
diff --git a/src/run.cpp b/src/run.cpp index bc9a533..b990a4c 100644 --- a/src/run.cpp +++ b/src/run.cpp @@ -40,13 +40,13 @@ static double min(double v1, double v2); typedef void (*benchmark)(const Chain**); typedef benchmark (*generator)(int64 chains_per_thread, int64 bytes_per_line, int64 bytes_per_chain, - int64 stride, int64 loop_length, bool prefetch); + int64 stride, int64 loop_length, int32 prefetch_hint); static benchmark chase_pointers(int64 chains_per_thread, int64 bytes_per_line, int64 bytes_per_chain, - int64 stride, int64 loop_length, bool prefetch); + int64 stride, int64 loop_length, int32 prefetch_hint); static benchmark follow_streams(int64 chains_per_thread, int64 bytes_per_line, int64 bytes_per_chain, - int64 stride, int64 loop_length, bool prefetch); + int64 stride, int64 loop_length, int32 prefetch_hint); Lock Run::global_mutex; int64 Run::_ops_per_chain = 0; @@ -121,7 +121,7 @@ int Run::run() { benchmark bench = gen(this->exp->chains_per_thread, this->exp->bytes_per_line, this->exp->bytes_per_chain, this->exp->stride, this->exp->loop_length, - this->exp->prefetch); + this->exp->prefetch_hint); volatile static double istart = 0; volatile static double istop = 0; @@ -171,7 +171,7 @@ int Run::run() { benchmark bench = gen(this->exp->chains_per_thread, this->exp->bytes_per_line, this->exp->bytes_per_chain, this->exp->stride, this->exp->loop_length, - this->exp->prefetch); + this->exp->prefetch_hint); for (int e = 0; e < this->exp->experiments; e++) { // barrier @@ -374,7 +374,7 @@ static benchmark chase_pointers(int64 chains_per_thread, // memory loading per t int64 bytes_per_chain, // ignored int64 stride, // ignored int64 loop_length, // length of the inner loop - bool prefetch // prefetch + int32 prefetch_hint // use of prefetching ) { // Create Compiler. AsmJit::Compiler c; @@ -417,8 +417,25 @@ static benchmark chase_pointers(int64 chains_per_thread, // memory loading per t c.mov(positions[i], ptr(positions[i], offsetof(Chain, next))); // Prefetch next - if (prefetch) + switch (prefetch_hint) + { + case Experiment::T0: c.prefetch(ptr(positions[i]), AsmJit::PREFETCH_T0); + break; + case Experiment::T1: + c.prefetch(ptr(positions[i]), AsmJit::PREFETCH_T1); + break; + case Experiment::T2: + c.prefetch(ptr(positions[i]), AsmJit::PREFETCH_T2); + break; + case Experiment::NTA: + c.prefetch(ptr(positions[i]), AsmJit::PREFETCH_NTA); + break; + case Experiment::NONE: + default: + break; + + } } // Wait @@ -485,7 +502,7 @@ static benchmark follow_streams(int64 chains_per_thread, // memory loading per t int64 bytes_per_chain, // ignored int64 stride, // ignored int64 loop_length, // ignored - bool prefetch // ignored + int32 prefetch_hint // ignored ) { return 0; /* |