Skip to content

Replace LZCNT with BSR instruction in hdr_histogram #230

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 3 commits into from
Oct 6, 2015
Merged

Replace LZCNT with BSR instruction in hdr_histogram #230

Changes from 1 commit
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
a5600f8
Replace LZCNT with BSR instruction in hdr_histogram
SeverinLeonhardt Oct 2, 2015
f4b75d1
Fix constness in hdr_histogram's hdr_bsr64
SeverinLeonhardt Oct 6, 2015
0d95364
Check for 0 before calling __builtin_clzll
SeverinLeonhardt Oct 6, 2015
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
31 changes: 23 additions & 8 deletions src/third_party/hdr_histogram/hdr_histogram.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,21 +23,36 @@

#include "hdr_histogram.hpp"

inline int32_t hdr_clz64(uint64_t x) {
/**
* Get the most significant bit set.
*
* The result is a one-based index, zero means no bits are set.
*/
inline int32_t hdr_bsr64(uint64_t x) {
#if defined(_MSC_VER)
unsigned long index;
# if defined(_M_AMD64)
return (int32_t)__lzcnt64(x);
char isNonzero = _BitScanReverse64(&index, x);
if (isNonzero)
return index + 1;
else
return 0;
# else
// On 32-bit this needs to be split into two operations
int32_t lz = (int32_t)__lzcnt((unsigned long)(x >> 32));
if (lz == 32) {
const char isNonzero = _BitScanReverse(&index, (unsigned long)(x >> 32));
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This shouldn't be const because it's reused for the next BSR.

if (isNonzero)
return index + 32 + 1;
else {
// Scan the last 32 bits by truncating the 64-bit value
return (int32_t)__lzcnt((unsigned long)x) + 32;
isNonzero = _BitScanReverse(&index, (unsigned long)x);
if (isNonzero)
return index + 1;
else
return 0;
}
return lz;
# endif
#else
return (int32_t)__builtin_clzll(x);
return 64 - (int32_t)__builtin_clzll(x);
#endif
}

Expand Down Expand Up @@ -132,7 +147,7 @@ static int64_t power(int64_t base, int64_t exp)

static int32_t get_bucket_index(struct hdr_histogram* h, int64_t value)
{
int32_t pow2ceiling = 64 - hdr_clz64(value | h->sub_bucket_mask); // smallest power of 2 containing value
int32_t pow2ceiling = hdr_bsr64(value | h->sub_bucket_mask); // smallest power of 2 containing value
return pow2ceiling - (int32_t)h->unit_magnitude - (h->sub_bucket_half_count_magnitude + 1);
}

Expand Down