Use of prefetching in bufferiszero.c is quite questionable:
- prefetches are issued just a few CPU cycles before the corresponding
line would be hit by demand loads;
- they are done for simple access patterns, i.e. where hardware
prefetchers can perform better;
- they compete for load ports in loops that should be limited by load
port throughput rather than ALU throughput.
Signed-off-by: Alexander Monakov <amona...@ispras.ru>
Signed-off-by: Mikhail Romanov <mmroma...@ispras.ru>
---
util/bufferiszero.c | 3 ---
1 file changed, 3 deletions(-)
diff --git a/util/bufferiszero.c b/util/bufferiszero.c
index c037d11d04..cb3eb2543f 100644
--- a/util/bufferiszero.c
+++ b/util/bufferiszero.c
@@ -49,7 +49,6 @@ buffer_is_zero_len_4_plus(const void *buf, size_t len)
const uint64_t *e = (uint64_t *)(((uintptr_t)buf + len) & -8);
for (; p + 8 <= e; p += 8) {
- __builtin_prefetch(p + 8);
if (t) {
return false;
}
@@ -79,7 +78,6 @@ buffer_zero_sse2(const void *buf, size_t len)
/* Loop over 16-byte aligned blocks of 64. */
while (likely(p <= e)) {
- __builtin_prefetch(p);
t = _mm_cmpeq_epi8(t, zero);
if (unlikely(_mm_movemask_epi8(t) != 0xFFFF)) {
return false;
@@ -110,7 +108,6 @@ buffer_zero_avx2(const void *buf, size_t len)
/* Loop over 32-byte aligned blocks of 128. */
while (p <= e) {
- __builtin_prefetch(p);
if (unlikely(!_mm256_testz_si256(t, t))) {
return false;
}
--
2.32.0
