Messages in this thread Patch in this message | | | Date | Mon, 24 Oct 2005 13:56:06 -0400 (EDT) | From | Nicolas Pitre <> | Subject | [PATCH 1/5] crypto/sha1.c: avoid useless memcpy() |
| |
The current code unconditionally copy the first block for every call to sha1_update(). This can be avoided if there is no pending partial block. This is always the case on the first call to sha1_update() (if the length is >= 64 of course0.
In the case where sha1_update() is called successively with len=64, a 8.5% performance increase can be observed on i386, and 8.2% on ARm.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Index: linux-2.6/crypto/sha1.c =================================================================== --- linux-2.6.orig/crypto/sha1.c +++ linux-2.6/crypto/sha1.c @@ -48,23 +48,26 @@ static void sha1_update(void *ctx, const u8 *data, unsigned int len) { struct sha1_ctx *sctx = ctx; - unsigned int i, j; + unsigned int partial, done; u32 temp[SHA_WORKSPACE_WORDS]; - j = (sctx->count >> 3) & 0x3f; + partial = (sctx->count >> 3) & 0x3f; sctx->count += len << 3; + done = 0; - if ((j + len) > 63) { - memcpy(&sctx->buffer[j], data, (i = 64-j)); - sha_transform(sctx->state, sctx->buffer, temp); - for ( ; i + 63 < len; i += 64) { - sha_transform(sctx->state, &data[i], temp); + if ((partial + len) > 63) { + if (partial) { + done = 64 - partial; + memcpy(sctx->buffer + partial, data, done); + sha_transform(sctx->state, sctx->buffer, temp); + partial = 0; } - j = 0; + for ( ; done + 63 < len; done += 64) + sha_transform(sctx->state, data + done, temp); } - else i = 0; + if (len - done) + memcpy(sctx->buffer + partial, data + done, len - done); memset(temp, 0, sizeof(temp)); - memcpy(&sctx->buffer[j], &data[i], len - i); } - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |