From patchwork Thu Jun 8 17:23:22 2017
Content-Type: text/plain; charset="utf-8"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
X-Patchwork-Submitter: Philipp Zabel
X-Patchwork-Id: 9776127
Return-Path:
Received: from mail.wl.linuxfoundation.org (pdx-wl-mail.web.codeaurora.org
[172.30.200.125])
by pdx-korg-patchwork.web.codeaurora.org (Postfix) with ESMTP id
F1BC56034B for ;
Thu, 8 Jun 2017 17:23:32 +0000 (UTC)
Received: from mail.wl.linuxfoundation.org (localhost [127.0.0.1])
by mail.wl.linuxfoundation.org (Postfix) with ESMTP id D480C205F7
for ;
Thu, 8 Jun 2017 17:23:32 +0000 (UTC)
Received: by mail.wl.linuxfoundation.org (Postfix, from userid 486)
id C7BCE269A3; Thu, 8 Jun 2017 17:23:32 +0000 (UTC)
X-Spam-Checker-Version: SpamAssassin 3.3.1 (2010-03-16) on
pdx-wl-mail.web.codeaurora.org
X-Spam-Level:
X-Spam-Status: No, score=-4.2 required=2.0 tests=BAYES_00, RCVD_IN_DNSWL_MED
autolearn=ham version=3.3.1
Received: from gabe.freedesktop.org (gabe.freedesktop.org [131.252.210.177])
(using TLSv1.2 with cipher DHE-RSA-AES256-GCM-SHA384 (256/256
bits)) (No client certificate requested)
by mail.wl.linuxfoundation.org (Postfix) with ESMTPS id 51663205F7
for ;
Thu, 8 Jun 2017 17:23:32 +0000 (UTC)
Received: from gabe.freedesktop.org (localhost [127.0.0.1])
by gabe.freedesktop.org (Postfix) with ESMTP id 2BBBB6E410;
Thu, 8 Jun 2017 17:23:29 +0000 (UTC)
X-Original-To: dri-devel@lists.freedesktop.org
Delivered-To: dri-devel@lists.freedesktop.org
Received: from metis.ext.pengutronix.de (metis.ext.pengutronix.de
[IPv6:2001:67c:670:201:290:27ff:fe1d:cc33])
by gabe.freedesktop.org (Postfix) with ESMTPS id 862476E410
for ;
Thu, 8 Jun 2017 17:23:27 +0000 (UTC)
Received: from dude.hi.pengutronix.de ([2001:67c:670:100:1d::7]
helo=dude.pengutronix.de.)
by metis.ext.pengutronix.de with esmtp (Exim 4.84_2)
(envelope-from )
id 1dJ19Z-0005a8-Up; Thu, 08 Jun 2017 19:23:25 +0200
From: Philipp Zabel
To: dri-devel@lists.freedesktop.org
Subject: [PATCH] drm/imx: lock scanout transfers for consecutive bursts
Date: Thu, 8 Jun 2017 19:23:22 +0200
Message-Id: <20170608172322.17105-1-p.zabel@pengutronix.de>
X-Mailer: git-send-email 2.11.0
X-SA-Exim-Connect-IP: 2001:67c:670:100:1d::7
X-SA-Exim-Mail-From: p.zabel@pengutronix.de
X-SA-Exim-Scanned: No (on metis.ext.pengutronix.de);
SAEximRunCond expanded to false
X-PTX-Original-Recipient: dri-devel@lists.freedesktop.org
Cc: kernel@pengutronix.de
X-BeenThere: dri-devel@lists.freedesktop.org
X-Mailman-Version: 2.1.18
Precedence: list
List-Id: Direct Rendering Infrastructure - Development
List-Unsubscribe: ,
List-Archive:
List-Post:
List-Help:
List-Subscribe: ,
MIME-Version: 1.0
Errors-To: dri-devel-bounces@lists.freedesktop.org
Sender: "dri-devel"
X-Virus-Scanned: ClamAV using ClamSMTP
Because of its shallow queues and limited reordering ability, the i.MX6Q
memory controller likes AXI bursts of consecutive addresses a lot.
To optimize memory access performance, lock the IPU scanout channels for
a number of burst accesses each, before switching to the next channel.
The burst size and length of a locked burst chain are chosen not to
overshoot the stride.
Enabling the 8-burst channel lock on a single 1920x1080@60Hz RGBx
scanout (474 MiB/s of 64-byte IPU memory read accesses) reduces the
reported memory controller busy cycles from 46% to below 28% on an
otherwise idle i.MX6Q.
Tested-by: Lucas Stach
Signed-off-by: Philipp Zabel
---
drivers/gpu/drm/imx/ipuv3-plane.c | 38 +++++++++++++++++++++++++++++++++++---
1 file changed, 35 insertions(+), 3 deletions(-)
diff --git a/drivers/gpu/drm/imx/ipuv3-plane.c b/drivers/gpu/drm/imx/ipuv3-plane.c
index a9292e3d3c437..46e7e44b5ae12 100644
--- a/drivers/gpu/drm/imx/ipuv3-plane.c
+++ b/drivers/gpu/drm/imx/ipuv3-plane.c
@@ -496,6 +496,27 @@ static int ipu_chan_assign_axi_id(int ipu_chan)
}
}
+static void ipu_calculate_bursts(u32 width, u32 cpp, u32 stride,
+ u8 *burstsize, u8 *num_bursts)
+{
+ const unsigned int width_bytes = width * cpp;
+ unsigned int npb, bursts;
+
+ /* Maximum number of pixels per burst without overshooting stride */
+ for (npb = 64 / cpp; npb > 0; --npb) {
+ if (round_up(width_bytes, npb * cpp) <= stride)
+ break;
+ }
+ *burstsize = npb;
+
+ /* Maximum number of consecutive bursts without overshooting stride */
+ for (bursts = 8; bursts > 1; bursts /= 2) {
+ if (round_up(width_bytes, npb * cpp * bursts) <= stride)
+ break;
+ }
+ *num_bursts = bursts;
+}
+
static void ipu_plane_atomic_update(struct drm_plane *plane,
struct drm_plane_state *old_state)
{
@@ -509,6 +530,9 @@ static void ipu_plane_atomic_update(struct drm_plane *plane,
unsigned long alpha_eba = 0;
enum ipu_color_space ics;
unsigned int axi_id = 0;
+ const struct drm_format_info *info;
+ u8 burstsize, num_bursts;
+ u32 width, height;
int active;
if (ipu_plane->dp_flow == IPU_DP_FLOW_SYNC_FG)
@@ -583,15 +607,21 @@ static void ipu_plane_atomic_update(struct drm_plane *plane,
ipu_dmfc_config_wait4eot(ipu_plane->dmfc, drm_rect_width(dst));
+ width = drm_rect_width(&state->src) >> 16;
+ height = drm_rect_height(&state->src) >> 16;
+ info = drm_format_info(fb->format->format);
+ ipu_calculate_bursts(width, info->cpp[0], fb->pitches[0],
+ &burstsize, &num_bursts);
+
ipu_cpmem_zero(ipu_plane->ipu_ch);
- ipu_cpmem_set_resolution(ipu_plane->ipu_ch,
- drm_rect_width(&state->src) >> 16,
- drm_rect_height(&state->src) >> 16);
+ ipu_cpmem_set_resolution(ipu_plane->ipu_ch, width, height);
ipu_cpmem_set_fmt(ipu_plane->ipu_ch, fb->format->format);
+ ipu_cpmem_set_burstsize(ipu_plane->ipu_ch, burstsize);
ipu_cpmem_set_high_priority(ipu_plane->ipu_ch);
ipu_idmac_set_double_buffer(ipu_plane->ipu_ch, 1);
ipu_cpmem_set_stride(ipu_plane->ipu_ch, fb->pitches[0]);
ipu_cpmem_set_axi_id(ipu_plane->ipu_ch, axi_id);
+
switch (fb->format->format) {
case DRM_FORMAT_YUV420:
case DRM_FORMAT_YVU420:
@@ -631,6 +661,7 @@ static void ipu_plane_atomic_update(struct drm_plane *plane,
case DRM_FORMAT_RGBX8888_A8:
case DRM_FORMAT_BGRX8888_A8:
alpha_eba = drm_plane_state_to_eba(state, 1);
+ num_bursts = 0;
dev_dbg(ipu_plane->base.dev->dev, "phys = %lu %lu, x = %d, y = %d",
eba, alpha_eba, state->src.x1 >> 16, state->src.y1 >> 16);
@@ -656,6 +687,7 @@ static void ipu_plane_atomic_update(struct drm_plane *plane,
}
ipu_cpmem_set_buffer(ipu_plane->ipu_ch, 0, eba);
ipu_cpmem_set_buffer(ipu_plane->ipu_ch, 1, eba);
+ ipu_idmac_lock_enable(ipu_plane->ipu_ch, num_bursts);
ipu_plane_enable(ipu_plane);
}