Skip to content

Commit

Permalink
Adopted explicit buffers for low/mid-level attrs
Browse files Browse the repository at this point in the history 
It's really frustrating that it's impossible to create an uninitialized
expression with the scope of a compound-literal...

(I'm going to ignore that this is technically possible with alloca.)

The lack of uninitialized compound-literals forces each of our attribute
lists to make a decision: 1. Use an implicit buffer and pay for
zero-initialization? or 2. Use an explicit buffer, adding code noising
and risking out-of-date buffer sizes.

As a compromise, this commit adopts explicit buffers in most of the
low/mid-level layers. Where the code is already pretty noisy, but also
heavily scrutinized and iterated over to reduce code/stack costs. This
leaves the high-level layers with the hopefully safer and more readable
implicit buffers.

You can see this zero initializing has a surprisingly high code cost,
for what is otherwise a noop:

           code          stack
  before: 33828           2632
  after:  33656 (-0.5%)   2632 (+0.0%)
  • Loading branch information
@geky
geky committed May 11, 2024
1 parent b36663f commit a9e3cad
Showing 1 changed file with 54 additions and 24 deletions.
78 changes: 54 additions & 24 deletions lfs.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -3500,8 +3500,9 @@ static int lfsr_rbyd_appendcksum(lfs_t *lfs, lfsr_rbyd_t *rbyd) {
return err;
}

uint8_t ecksum_buf[LFSR_ECKSUM_DSIZE];
err = lfsr_rbyd_appendattr_(lfs, rbyd, LFSR_ATTR(
LFSR_TAG_ECKSUM, 0, LFSR_DATA_ECKSUM(&ecksum)));
LFSR_TAG_ECKSUM, 0, LFSR_DATA_ECKSUM_(&ecksum, ecksum_buf)));
if (err) {
return err;
}
Expand Down Expand Up @@ -6042,11 +6043,12 @@ static int lfsr_mdir_commit__(lfs_t *lfs, lfsr_mdir_t *mdir,
}

// write our new shrub tag
uint8_t shrub_buf[LFSR_SHRUB_DSIZE];
err = lfsr_rbyd_appendattr(lfs, &rbyd_,
rid - lfs_smax32(start_rid, 0),
LFSR_ATTR(
LFSR_TAG_BSHRUB, 0,
LFSR_DATA_SHRUB(&shrub)));
LFSR_DATA_SHRUB_(&shrub, shrub_buf)));
if (err) {
return err;
}
Expand Down Expand Up @@ -6144,13 +6146,14 @@ static int lfsr_mdir_commit__(lfs_t *lfs, lfsr_mdir_t *mdir,
} else if (lfsr_tag_key(attrs[i].tag) == LFSR_TAG_SHRUBTRUNK) {
const lfsr_shrub_t *shrub = attrs[i].cat;

uint8_t shrub_buf[LFSR_SHRUB_DSIZE];
int err = lfsr_rbyd_appendattr(lfs, &rbyd_,
rid - lfs_smax32(start_rid, 0),
LFSR_ATTR(
lfsr_tag_mode(attrs[i].tag) | LFSR_TAG_BSHRUB,
attrs[i].delta,
// note we use the staged trunk here
LFSR_DATA_SHRUB(shrub)));
LFSR_DATA_SHRUB_(shrub, shrub_buf)));
if (err) {
return err;
}
Expand Down Expand Up @@ -6421,8 +6424,11 @@ static int lfsr_mdir_compact__(lfs_t *lfs, lfsr_mdir_t *mdir_,
}

// write the new shrub tag
uint8_t shrub_buf[LFSR_SHRUB_DSIZE];
err = lfsr_rbyd_appendcompactattr(lfs, &mdir_->rbyd,
LFSR_ATTR(tag, weight, LFSR_DATA_SHRUB(&shrub)));
LFSR_ATTR(
tag, weight,
LFSR_DATA_SHRUB_(&shrub, shrub_buf)));
if (err) {
LFS_ASSERT(err != LFS_ERR_RANGE);
return err;
Expand Down Expand Up @@ -6790,17 +6796,21 @@ static int lfsr_mdir_commit(lfs_t *lfs, lfsr_mdir_t *mdir,
return err;
}

uint8_t mdir_buf[LFSR_MPTR_DSIZE];
uint8_t msibling_buf[LFSR_MPTR_DSIZE];
err = lfsr_btree_commit(lfs, &mtree_.u.btree,
0, LFSR_ATTRS(
LFSR_ATTR(
LFSR_TAG_MDIR, +lfsr_mleafweight(lfs),
LFSR_DATA_MPTR(lfsr_mdir_mptr(&mdir_))),
LFSR_DATA_MPTR_(
lfsr_mdir_mptr(&mdir_), mdir_buf)),
LFSR_ATTR_CAT_(
LFSR_TAG_NAME, +lfsr_mleafweight(lfs),
&split_data, 1),
LFSR_ATTR(
LFSR_TAG_MDIR, 0,
LFSR_DATA_MPTR(lfsr_mdir_mptr(&msibling_)))));
LFSR_DATA_MPTR_(
lfsr_mdir_mptr(&msibling_), msibling_buf))));
if (err) {
return err;
}
Expand All @@ -6810,17 +6820,21 @@ static int lfsr_mdir_commit(lfs_t *lfs, lfsr_mdir_t *mdir,
// mark as unerased in case of failure
lfs->mtree.u.btree.eoff = -1;

uint8_t mdir_buf[LFSR_MPTR_DSIZE];
uint8_t msibling_buf[LFSR_MPTR_DSIZE];
err = lfsr_btree_commit(lfs, &mtree_.u.btree,
lfsr_mid_bid(lfs, mdir->mid), LFSR_ATTRS(
LFSR_ATTR(
LFSR_TAG_MDIR, 0,
LFSR_DATA_MPTR(lfsr_mdir_mptr(&mdir_))),
LFSR_DATA_MPTR_(
lfsr_mdir_mptr(&mdir_), mdir_buf)),
LFSR_ATTR_CAT_(
LFSR_TAG_NAME, +lfsr_mleafweight(lfs),
&split_data, 1),
LFSR_ATTR(
LFSR_TAG_MDIR, 0,
LFSR_DATA_MPTR(lfsr_mdir_mptr(&msibling_)))));
LFSR_DATA_MPTR_(
lfsr_mdir_mptr(&msibling_), msibling_buf))));
if (err) {
return err;
}
Expand Down Expand Up @@ -6880,11 +6894,13 @@ static int lfsr_mdir_commit(lfs_t *lfs, lfsr_mdir_t *mdir,
lfs->mtree.u.btree.eoff = -1;

// update our mtree
uint8_t mdir_buf[LFSR_MPTR_DSIZE];
err = lfsr_btree_commit(lfs, &mtree_.u.btree,
lfsr_mid_bid(lfs, mdir->mid), LFSR_ATTRS(
LFSR_ATTR(
LFSR_TAG_MDIR, 0,
LFSR_DATA_MPTR(lfsr_mdir_mptr(&mdir_)))));
LFSR_DATA_MPTR_(
lfsr_mdir_mptr(&mdir_), mdir_buf))));
if (err) {
return err;
}
Expand Down Expand Up @@ -6940,14 +6956,15 @@ static int lfsr_mdir_commit(lfs_t *lfs, lfsr_mdir_t *mdir,
//
// note end_rid=0 here will delete any files leftover from a split
// in our mroot
uint8_t mtree_buf[LFS_MAX(LFSR_MPTR_DSIZE, LFSR_BTREE_DSIZE)];
err = lfsr_mdir_commit_(lfs, &mroot_, -1, 0, NULL, -1, LFSR_ATTRS(
(lfsr_mtree_ismptr(&mtree_))
? LFSR_ATTR(
LFSR_TAG_SUB | LFSR_TAG_MDIR, 0,
LFSR_DATA_MPTR(&mtree_.u.mptr.mptr))
LFSR_DATA_MPTR_(&mtree_.u.mptr.mptr, mtree_buf))
: LFSR_ATTR(
LFSR_TAG_SUB | LFSR_TAG_MTREE, 0,
LFSR_DATA_BTREE(&mtree_.u.btree))));
LFSR_DATA_BTREE_(&mtree_.u.btree, mtree_buf))));
if (err) {
LFS_ASSERT(err != LFS_ERR_RANGE);
return err;
Expand Down Expand Up @@ -6978,11 +6995,14 @@ static int lfsr_mdir_commit(lfs_t *lfs, lfsr_mdir_t *mdir,
mrootchild = mrootparent_;

// commit mrootchild
uint8_t mrootchild_buf[LFSR_MPTR_DSIZE];
err = lfsr_mdir_commit_(lfs, &mrootparent_, -1, -1, NULL,
-1, LFSR_ATTRS(
LFSR_ATTR(
LFSR_TAG_MROOT, 0,
LFSR_DATA_MPTR(lfsr_mdir_mptr(&mrootchild_)))));
LFSR_DATA_MPTR_(
lfsr_mdir_mptr(&mrootchild_),
mrootchild_buf))));
if (err) {
LFS_ASSERT(err != LFS_ERR_RANGE);
LFS_ASSERT(err != LFS_ERR_NOENT);
Expand Down Expand Up @@ -7010,14 +7030,17 @@ static int lfsr_mdir_commit(lfs_t *lfs, lfsr_mdir_t *mdir,
return err;
}

uint8_t mrootchild_buf[LFSR_MPTR_DSIZE];
err = lfsr_mdir_commit__(lfs, &mrootanchor_, -1, -1,
-1, LFSR_ATTRS(
LFSR_ATTR(
LFSR_TAG_MAGIC, 0,
LFSR_DATA_BUF("littlefs", 8)),
LFSR_ATTR(
LFSR_TAG_MROOT, 0,
LFSR_DATA_MPTR(lfsr_mdir_mptr(&mrootchild_)))));
LFSR_DATA_MPTR_(
lfsr_mdir_mptr(&mrootchild_),
mrootchild_buf))));
if (err) {
LFS_ASSERT(err != LFS_ERR_RANGE);
LFS_ASSERT(err != LFS_ERR_NOENT);
Expand Down Expand Up @@ -10418,7 +10441,7 @@ static int lfsr_file_carve(lfs_t *lfs, lfsr_file_t *file,
bptr_.data = lfsr_data_slice(bptr_.data,
lfs->cfg->fragment_size,
-1);

err = lfsr_bshrub_commit(lfs, file, bid, LFSR_ATTRS(
LFSR_ATTR_CAT(
LFSR_TAG_GROW | LFSR_TAG_SUB | LFSR_TAG_DATA,
Expand All @@ -10427,7 +10450,7 @@ static int lfsr_file_carve(lfs_t *lfs, lfsr_file_t *file,
lfs->cfg->fragment_size)),
LFSR_ATTR(
LFSR_TAG_BLOCK, +(weight_ - lfs->cfg->fragment_size),
LFSR_DATA_BPTR(&bptr_))));
LFSR_DATA_BPTR_(&bptr_, left.buf))));
if (err) {
return err;
}
Expand All @@ -10451,7 +10474,7 @@ static int lfsr_file_carve(lfs_t *lfs, lfsr_file_t *file,
LFSR_ATTR(
LFSR_TAG_GROW | LFSR_TAG_SUB | LFSR_TAG_BLOCK,
-(weight_ - lfsr_data_size(bptr_.data)),
LFSR_DATA_BPTR(&bptr_)),
LFSR_DATA_BPTR_(&bptr_, right.buf)),
LFSR_ATTR_CAT(
LFSR_TAG_DATA,
+(weight_ - lfsr_data_size(bptr_.data)),
Expand Down Expand Up @@ -10925,9 +10948,12 @@ static int lfsr_file_flush_(lfs_t *lfs, lfsr_file_t *file,
lfs_off_t block_end = block_start + lfsr_data_size(bptr.data);

// and write it into our tree
uint8_t bptr_buf[LFSR_BPTR_DSIZE];
err = lfsr_file_carve(lfs, file,
block_start, block_end - block_start,
LFSR_ATTR(LFSR_TAG_BLOCK, 0, LFSR_DATA_BPTR(&bptr)));
LFSR_ATTR(
LFSR_TAG_BLOCK, 0,
LFSR_DATA_BPTR_(&bptr, bptr_buf)));
if (err) {
return err;
}
Expand Down Expand Up @@ -11036,7 +11062,9 @@ fragment:;
// our tree
int err = lfsr_file_carve(lfs, file,
fragment_start, fragment_end - fragment_start,
LFSR_ATTR_CAT_(LFSR_TAG_DATA, 0, datas, data_count));
LFSR_ATTR_CAT_(
LFSR_TAG_DATA, 0,
datas, data_count));
if (err && err != LFS_ERR_RANGE) {
return err;
}
Expand Down Expand Up @@ -11614,9 +11642,10 @@ int lfsr_file_truncate(lfs_t *lfs, lfsr_file_t *file, lfs_off_t size_) {
} else {
// truncate our btree
err = lfsr_file_carve(lfs, file,
lfs_min32(size, size_),
size - lfs_min32(size, size_),
LFSR_ATTR(LFSR_TAG_DATA, +size_ - size, LFSR_DATA_NULL()));
lfs_min32(size, size_), size - lfs_min32(size, size_),
LFSR_ATTR(
LFSR_TAG_DATA, +size_ - size,
LFSR_DATA_NULL()));
if (err) {
goto failed;
}
Expand Down Expand Up @@ -11731,9 +11760,10 @@ int lfsr_file_fruncate(lfs_t *lfs, lfsr_file_t *file, lfs_off_t size_) {
} else {
// fruncate our btree
err = lfsr_file_carve(lfs, file,
0,
lfs_smax32(size - size_, 0),
LFSR_ATTR(LFSR_TAG_DATA, +size_ - size, LFSR_DATA_NULL()));
0, lfs_smax32(size - size_, 0),
LFSR_ATTR(
LFSR_TAG_DATA, +size_ - size,
LFSR_DATA_NULL()));
if (err) {
goto failed;
}
Expand Down

0 comments on commit a9e3cad

Please sign in to comment.