From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:36346) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1beRaP-0001Ky-LK for qemu-devel@nongnu.org; Mon, 29 Aug 2016 14:47:10 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1beRaN-0001pl-PH for qemu-devel@nongnu.org; Mon, 29 Aug 2016 14:47:08 -0400 Received: from mail-qk0-x243.google.com ([2607:f8b0:400d:c09::243]:33840) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1beRaN-0001pX-G7 for qemu-devel@nongnu.org; Mon, 29 Aug 2016 14:47:07 -0400 Received: by mail-qk0-x243.google.com with SMTP id r128so10894777qkc.1 for ; Mon, 29 Aug 2016 11:47:07 -0700 (PDT) Sender: Richard Henderson From: Richard Henderson Date: Mon, 29 Aug 2016 11:46:20 -0700 Message-Id: <1472496380-19706-10-git-send-email-rth@twiddle.net> In-Reply-To: <1472496380-19706-1-git-send-email-rth@twiddle.net> References: <1472496380-19706-1-git-send-email-rth@twiddle.net> Subject: [Qemu-devel] [PATCH v3 9/9] cutils: Remove ppc buffer zero checking List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: qemu-devel@nongnu.org Cc: pbonzini@redhat.com, vijay.kilari@gmail.com For ppc64le, gcc6 does extremely poorly with the Altivec code. Moreover, on POWER7 and POWER8, a hand-optimized Altivec version turns out to be no faster than the revised integer version, and therefore not worth the effort. Signed-off-by: Richard Henderson --- util/bufferiszero.c | 55 +---------------------------------------------------- 1 file changed, 1 insertion(+), 54 deletions(-) diff --git a/util/bufferiszero.c b/util/bufferiszero.c index 6d13d7f..3b39f82 100644 --- a/util/bufferiszero.c +++ b/util/bufferiszero.c @@ -29,35 +29,6 @@ /* vector definitions */ -extern void link_error(void); - -#define ACCEL_BUFFER_ZERO(NAME, SIZE, VECTYPE, NONZERO) \ -static bool NAME(const void *buf, size_t len) \ -{ \ - const void *end = buf + len; \ - do { \ - const VECTYPE *p = buf; \ - VECTYPE t; \ - __builtin_prefetch(buf + SIZE); \ - barrier(); \ - if (SIZE == sizeof(VECTYPE) * 4) { \ - t = (p[0] | p[1]) | (p[2] | p[3]); \ - } else if (SIZE == sizeof(VECTYPE) * 8) { \ - t = p[0] | p[1]; \ - t |= p[2] | p[3]; \ - t |= p[4] | p[5]; \ - t |= p[6] | p[7]; \ - } else { \ - link_error(); \ - } \ - if (unlikely(NONZERO(t))) { \ - return false; \ - } \ - buf += SIZE; \ - } while (buf < end); \ - return true; \ -} - typedef bool (*accel_zero_fn)(const void *, size_t); static bool @@ -98,31 +69,7 @@ buffer_zero_int(const void *buf, size_t len) } } -#ifdef __ALTIVEC__ -#include -/* The altivec.h header says we're allowed to undef these for - * C++ compatibility. Here we don't care about C++, but we - * undef them anyway to avoid namespace pollution. - * altivec.h may redefine the bool macro as vector type. - * Reset it to POSIX semantics. - */ -#undef vector -#undef pixel -#undef bool -#define bool _Bool -#define DO_NONZERO(X) vec_any_ne(X, (__vector unsigned char){ 0 }) -ACCEL_BUFFER_ZERO(buffer_zero_ppc, 128, __vector unsigned char, DO_NONZERO) - -static bool select_accel_fn(const void *buf, size_t len) -{ - uintptr_t ibuf = (uintptr_t)buf; - if (len % 128 == 0 && ibuf % sizeof(__vector unsigned char) == 0) { - return buffer_zero_ppc(buf, len); - } - return buffer_zero_int(buf, len); -} - -#elif defined(CONFIG_AVX2_OPT) || defined(__SSE2__) +#if defined(CONFIG_AVX2_OPT) || defined(__SSE2__) #include #include -- 2.7.4