Skip to content
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

Avoid using YMM registers in gcm_ghash_vpclmulqdq_avx2_1. #2470

Merged
merged 1 commit into from
Mar 11, 2025
Merged

Avoid using YMM registers in gcm_ghash_vpclmulqdq_avx2_1. #2470

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
5 changes: 2 additions & 3 deletions crypto/fipsmodule/aes/asm/aes-gcm-avx10-x86_64.pl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -786,9 +786,6 @@ sub _ghash_update {
jae .Laad_loop_1x$local_label_suffix
.Laad_large_done$local_label_suffix:
# Issue the vzeroupper that is needed after using ymm or zmm registers.
# Do it here instead of at the end, to minimize overhead for small AADLEN.
vzeroupper
# GHASH the remaining data 16 bytes at a time, using xmm registers only.
.Laad_blockbyblock$local_label_suffix:
Expand All @@ -809,6 +806,8 @@ sub _ghash_update {
# Store the updated GHASH accumulator back to memory.
vpshufb $BSWAP_MASK_XMM, $GHASH_ACC_XMM, $GHASH_ACC_XMM
vmovdqu $GHASH_ACC_XMM, ($GHASH_ACC_PTR)
vzeroupper # This is needed after using ymm or zmm registers.
___
return $code;
}
Expand Down
12 changes: 10 additions & 2 deletions crypto/fipsmodule/aes/asm/aes-gcm-avx2-x86_64.pl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -461,10 +461,16 @@ sub _ghash_4x {
@{[ _save_xmmregs (6 .. 9) ]}
.seh_endprologue

vbroadcasti128 .Lbswap_mask(%rip), $BSWAP_MASK
# Load the bswap_mask and gfpoly constants. Since AADLEN is usually small,
# usually only 128-bit vectors will be used. So as an optimization, don't
# broadcast these constants to both 128-bit lanes quite yet.
vmovdqu .Lbswap_mask(%rip), $BSWAP_MASK_XMM
vmovdqu .Lgfpoly(%rip), $GFPOLY_XMM

# Load the GHASH accumulator.
vmovdqu ($GHASH_ACC_PTR), $GHASH_ACC_XMM
vpshufb $BSWAP_MASK_XMM, $GHASH_ACC_XMM, $GHASH_ACC_XMM
vbroadcasti128 .Lgfpoly(%rip), $GFPOLY


# Update GHASH with the remaining 16-byte block if any.
.Lghash_lastblock:
Expand All @@ -479,6 +485,8 @@ sub _ghash_4x {
# Store the updated GHASH accumulator back to memory.
vpshufb $BSWAP_MASK_XMM, $GHASH_ACC_XMM, $GHASH_ACC_XMM
vmovdqu $GHASH_ACC_XMM, ($GHASH_ACC_PTR)

vzeroupper
___
}
$code .= _end_func;
Expand Down