From patchwork Mon Feb 24 15:20:46 2014 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 8bit X-Patchwork-Submitter: =?utf-8?b?TWFyZWsgT2zFocOhaw==?= X-Patchwork-Id: 3709761 Return-Path: X-Original-To: patchwork-dri-devel@patchwork.kernel.org Delivered-To: patchwork-parsemail@patchwork1.web.kernel.org Received: from mail.kernel.org (mail.kernel.org [198.145.19.201]) by patchwork1.web.kernel.org (Postfix) with ESMTP id 715889F3E3 for ; Mon, 24 Feb 2014 15:21:31 +0000 (UTC) Received: from mail.kernel.org (localhost [127.0.0.1]) by mail.kernel.org (Postfix) with ESMTP id 7828520160 for ; Mon, 24 Feb 2014 15:21:30 +0000 (UTC) Received: from gabe.freedesktop.org (gabe.freedesktop.org [131.252.210.177]) by mail.kernel.org (Postfix) with ESMTP id 933D02015A for ; Mon, 24 Feb 2014 15:21:29 +0000 (UTC) Received: from gabe.freedesktop.org (localhost [127.0.0.1]) by gabe.freedesktop.org (Postfix) with ESMTP id 88B32FAAEB; Mon, 24 Feb 2014 07:21:26 -0800 (PST) X-Original-To: dri-devel@lists.freedesktop.org Delivered-To: dri-devel@lists.freedesktop.org Received: from mail-ea0-f177.google.com (mail-ea0-f177.google.com [209.85.215.177]) by gabe.freedesktop.org (Postfix) with ESMTP id 8FA2DFAAE9 for ; Mon, 24 Feb 2014 07:21:11 -0800 (PST) Received: by mail-ea0-f177.google.com with SMTP id h10so1806392eak.36 for ; Mon, 24 Feb 2014 07:21:10 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20120113; h=from:to:subject:date:message-id:in-reply-to:references:mime-version :content-type:content-transfer-encoding; bh=BdFapXyaMTm1jIvKfj3F1T4riYnGlLqjsNmej01CThw=; b=cS9I5lAYMfxlBN+WxuHRjFEtXbGpiwtUidRCTy2n+4XCaMDNa/2bo1jBCI9pfMrpEC 5Y5ZgszzOKJZcZOBUgnJzwyP7mTPwCvyN/w4AJm8w9R9+6XgRTVycKledCXT1u5eF5Nb XrYtQ4Gmv06wys6C5niI9+VA64XjreGmhhr9Nor8n8cErX4JkzRpgx7+NsjsZDCT7/E2 jDh/jmKc9ZqWRpgm1CvYyrLaIbCKCIxEt7+0iTvBA00zOriWge2LHEZhVWXVTwixAZEt LuG2gVHlmxFJL7Qv+tpf9I5++EnmMM8/9sGkHE9OMekL86dETetBER6uhLGXPKQs9CJX M1/Q== X-Received: by 10.15.41.140 with SMTP id s12mr25403627eev.4.1393255270677; Mon, 24 Feb 2014 07:21:10 -0800 (PST) Received: from localhost.localdomain ([194.228.11.204]) by mx.google.com with ESMTPSA id o43sm64864402eef.12.2014.02.24.07.21.09 for (version=TLSv1.1 cipher=ECDHE-RSA-RC4-SHA bits=128/128); Mon, 24 Feb 2014 07:21:10 -0800 (PST) From: =?UTF-8?q?Marek=20Ol=C5=A1=C3=A1k?= To: dri-devel@lists.freedesktop.org Subject: [PATCH 6/6] drm/radeon: limit how much memory TTM can move per IB according to VRAM usage Date: Mon, 24 Feb 2014 16:20:46 +0100 Message-Id: <1393255246-8296-7-git-send-email-maraeo@gmail.com> X-Mailer: git-send-email 1.8.3.2 In-Reply-To: <1393255246-8296-1-git-send-email-maraeo@gmail.com> References: <1393255246-8296-1-git-send-email-maraeo@gmail.com> MIME-Version: 1.0 X-BeenThere: dri-devel@lists.freedesktop.org X-Mailman-Version: 2.1.13 Precedence: list List-Id: Direct Rendering Infrastructure - Development List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: dri-devel-bounces@lists.freedesktop.org Errors-To: dri-devel-bounces@lists.freedesktop.org X-Spam-Status: No, score=-2.7 required=5.0 tests=BAYES_00, DKIM_ADSP_CUSTOM_MED, DKIM_SIGNED,FREEMAIL_FROM,RCVD_IN_BL_SPAMCOP_NET,RCVD_IN_DNSWL_MED, RP_MATCHES_RCVD, T_DKIM_INVALID, UNPARSEABLE_RELAY autolearn=ham version=3.3.1 X-Spam-Checker-Version: SpamAssassin 3.3.1 (2010-03-16) on mail.kernel.org X-Virus-Scanned: ClamAV using ClamSMTP From: Marek Olšák Signed-off-by: Marek Olšák --- drivers/gpu/drm/radeon/radeon_cs.c | 2 +- drivers/gpu/drm/radeon/radeon_object.c | 87 +++++++++++++++++++++++++++++++--- drivers/gpu/drm/radeon/radeon_object.h | 3 +- 3 files changed, 84 insertions(+), 8 deletions(-) diff --git a/drivers/gpu/drm/radeon/radeon_cs.c b/drivers/gpu/drm/radeon/radeon_cs.c index 1ba1a48..2338eef 100644 --- a/drivers/gpu/drm/radeon/radeon_cs.c +++ b/drivers/gpu/drm/radeon/radeon_cs.c @@ -157,7 +157,7 @@ static int radeon_cs_parser_relocs(struct radeon_cs_parser *p) sort_relocs_for_validation(&p->validated); - return radeon_bo_list_validate(&p->ticket, &p->validated, p->ring); + return radeon_bo_list_validate(p->rdev, &p->ticket, &p->validated, p->ring); } static int radeon_cs_get_ring(struct radeon_cs_parser *p, u32 ring, s32 priority) diff --git a/drivers/gpu/drm/radeon/radeon_object.c b/drivers/gpu/drm/radeon/radeon_object.c index 19042ae..e1a5af7 100644 --- a/drivers/gpu/drm/radeon/radeon_object.c +++ b/drivers/gpu/drm/radeon/radeon_object.c @@ -368,29 +368,104 @@ void radeon_bo_fini(struct radeon_device *rdev) arch_phys_wc_del(rdev->mc.vram_mtrr); } -int radeon_bo_list_validate(struct ww_acquire_ctx *ticket, +/* Returns how many bytes TTM can move per IB. + */ +static u64 radeon_bo_get_threshold_for_moves(struct radeon_device *rdev) +{ + u64 real_vram_size = rdev->mc.real_vram_size; + u64 vram_usage = rdev->vram_usage; + + /* This function is based on the current VRAM usage. + * + * - If all of VRAM is free, allow relocating the number of bytes that + * is equal to 1/4 of the size of VRAM for this IB. + + * - If more than one half of VRAM is occupied, only allow relocating + * 1 MB of data for this IB. + * + * - From 0 to one half of used VRAM, the threshold decreases + * linearly. + * __________________ + * 1/4 of -|\ | + * VRAM | \ | + * | \ | + * | \ | + * | \ | + * | \ | + * | \ | + * | \________|1 MB + * |----------------| + * VRAM 0 % 100 % + * used used + * + * Note: It's a threshold, not a limit. The threshold must be crossed + * for buffer relocations to stop, so any buffer of an arbitrary size + * can be moved as long as the threshold isn't crossed before + * the relocation takes place. We don't want to disable buffer + * relocations completely. + * + * The idea is that buffers should be placed in VRAM at creation time + * and TTM should only do a minimum number of relocations during + * command submission. In practice, you need to submit at least + * a dozen IBs to move all buffers to VRAM if they are in GTT. + * + * Also, things can get pretty crazy under memory pressure and actual + * VRAM usage can change a lot, so playing safe even at 50% does + * consistently increase performance. + */ + + u64 half_vram = real_vram_size >> 1; + u64 half_free_vram = vram_usage >= half_vram ? 0 : half_vram - vram_usage; + u64 bytes_moved_threshold = half_free_vram >> 1; + return max(bytes_moved_threshold, 1024*1024ull); +} + +int radeon_bo_list_validate(struct radeon_device *rdev, + struct ww_acquire_ctx *ticket, struct list_head *head, int ring) { struct radeon_bo_list *lobj; struct radeon_bo *bo; - u32 domain; int r; + u64 bytes_moved = 0, initial_bytes_moved; + u64 bytes_moved_threshold = radeon_bo_get_threshold_for_moves(rdev); r = ttm_eu_reserve_buffers(ticket, head); if (unlikely(r != 0)) { return r; } + list_for_each_entry(lobj, head, tv.head) { bo = lobj->bo; if (!bo->pin_count) { - domain = lobj->domain; - + u32 domain = lobj->domain; + u32 current_domain = + radeon_mem_type_to_domain(bo->tbo.mem.mem_type); + + /* Check if this buffer will be moved and don't move it + * if we have moved too many buffers for this IB already. + * + * Note that this allows moving at least one buffer of + * any size, because it doesn't take the current "bo" + * into account. We don't want to disallow buffer moves + * completely. + */ + if (current_domain != RADEON_GEM_DOMAIN_CPU && + (domain & current_domain) == 0 && /* will be moved */ + bytes_moved > bytes_moved_threshold) { + /* don't move it */ + domain = current_domain; + } + retry: radeon_ttm_placement_from_domain(bo, domain); if (ring == R600_RING_TYPE_UVD_INDEX) radeon_uvd_force_into_uvd_segment(bo); - r = ttm_bo_validate(&bo->tbo, &bo->placement, - true, false); + + initial_bytes_moved = rdev->num_bytes_moved; + r = ttm_bo_validate(&bo->tbo, &bo->placement, true, false); + bytes_moved += rdev->num_bytes_moved - initial_bytes_moved; + if (unlikely(r)) { if (r != -ERESTARTSYS && domain != lobj->alt_domain) { domain = lobj->alt_domain; diff --git a/drivers/gpu/drm/radeon/radeon_object.h b/drivers/gpu/drm/radeon/radeon_object.h index 6c3ca9e..7dff64d 100644 --- a/drivers/gpu/drm/radeon/radeon_object.h +++ b/drivers/gpu/drm/radeon/radeon_object.h @@ -138,7 +138,8 @@ extern int radeon_bo_evict_vram(struct radeon_device *rdev); extern void radeon_bo_force_delete(struct radeon_device *rdev); extern int radeon_bo_init(struct radeon_device *rdev); extern void radeon_bo_fini(struct radeon_device *rdev); -extern int radeon_bo_list_validate(struct ww_acquire_ctx *ticket, +extern int radeon_bo_list_validate(struct radeon_device *rdev, + struct ww_acquire_ctx *ticket, struct list_head *head, int ring); extern int radeon_bo_fbdev_mmap(struct radeon_bo *bo, struct vm_area_struct *vma);