From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from gabe.freedesktop.org (gabe.freedesktop.org [131.252.210.177]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.lore.kernel.org (Postfix) with ESMTPS id 84CBAC27C77 for ; Tue, 11 Jun 2024 10:41:09 +0000 (UTC) Received: from gabe.freedesktop.org (localhost [127.0.0.1]) by gabe.freedesktop.org (Postfix) with ESMTP id C70F410E598; Tue, 11 Jun 2024 10:41:08 +0000 (UTC) Authentication-Results: gabe.freedesktop.org; dkim=pass (2048-bit key; unprotected) header.d=intel.com header.i=@intel.com header.b="hfPWRz0Q"; dkim-atps=neutral Received: from mgamail.intel.com (mgamail.intel.com [198.175.65.12]) by gabe.freedesktop.org (Postfix) with ESMTPS id C698610E59B for ; Tue, 11 Jun 2024 10:41:06 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=intel.com; i=@intel.com; q=dns/txt; s=Intel; t=1718102466; x=1749638466; h=from:date:subject:mime-version:content-transfer-encoding: message-id:references:in-reply-to:to:cc; bh=HHDlIp41TXmGjua2BhL5dvBfFYTJo1GD4Pv2Dnni8AM=; b=hfPWRz0QgeRgeVR8JoPRRu12f7tfRTPzdux7vgdCT54/WLX7kBh3ISG/ BsU59b3PgZ4df3EqGLUt+kuG9GvO3VAnUKKI85wcyfkVq1CXIS89Fme0U t2hTmrA7jPX4BdFX7L+d8u4utoPeNUI1hK8Dy7N3BA9Ag3UjSedlck+Bh r+p4dX+U2vxva2OFTXOBGjMyQtWBSelgICEDGNXtg8x7pMJFxrEjyf1h8 KPkT+sMNMQ6vuELFBnf8dTXnHctygkSd5vBi2yQN0uCF1LmIcoMTJec3O yXW7Q2nQnaGFR7Ojcw/OycG08HO9R1siUEaPfBCZYNdZvFkR9XHcwbf+5 g==; X-CSE-ConnectionGUID: TUp7sjomTZ++5q15nniMSg== X-CSE-MsgGUID: qZE300qPS+6RWGZVJXztMw== X-IronPort-AV: E=McAfee;i="6600,9927,11099"; a="26217719" X-IronPort-AV: E=Sophos;i="6.08,229,1712646000"; d="scan'208";a="26217719" Received: from fmviesa002.fm.intel.com ([10.60.135.142]) by orvoesa104.jf.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 11 Jun 2024 03:41:06 -0700 X-CSE-ConnectionGUID: EkHI0fjISmWnng5KweTqPQ== X-CSE-MsgGUID: vb4Qj51IQsmWbJqeU8gjHQ== X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="6.08,229,1712646000"; d="scan'208";a="62563553" Received: from lab-ah.igk.intel.com ([10.102.138.202]) by fmviesa002-auth.fm.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 11 Jun 2024 03:41:04 -0700 From: Andrzej Hajda Date: Tue, 11 Jun 2024 12:40:51 +0200 Subject: [PATCH v6 2/5] lib/gpgpu_shader: tooling for preparing and running gpgpu shaders MIME-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 8bit Message-Id: <20240611-iga64_inline_ups-v6-2-b634ec43a610@intel.com> References: <20240611-iga64_inline_ups-v6-0-b634ec43a610@intel.com> In-Reply-To: <20240611-iga64_inline_ups-v6-0-b634ec43a610@intel.com> To: igt-dev@lists.freedesktop.org Cc: Kamil Konieczny , Dominik Grzegorzek , Christoph Manszewski , =?utf-8?q?Zbigniew_Kempczy=C5=84ski?= , Gwan-gyeong Mun , Andrzej Hajda X-Mailer: b4 0.13.0 X-Developer-Signature: v=1; a=openpgp-sha256; l=8051; i=andrzej.hajda@intel.com; h=from:subject:message-id; bh=HHDlIp41TXmGjua2BhL5dvBfFYTJo1GD4Pv2Dnni8AM=; b=owEB7QES/pANAwAKASNispPeEP3XAcsmYgBmaCm5d28dKR9NCVGRM0fBK5SWdzJ5CHlsihlAsdGT i3jiNz+JAbMEAAEKAB0WIQT8qEQxNN2/XeF/A00jYrKT3hD91wUCZmgpuQAKCRAjYrKT3hD9167jC/ 9zmIGCbY1cdpvfupYvkh4Aen2l82kWRkFGMpbE/YOSdZaMHVDtCYep5p+If3Hziex7V8p75gjFQobJ 8HlXvpp1rX3J2zmSGLlpYIc4xplKUOOKF9OyXe9G0FFnpzSVD4ji7bVLHcDn/afMkMsKL+1jmXHDYZ sDFwMmWBOj8ztS5iFcNoojqKGnt42LJ08O9VcKdr4320kVFvxLsrse/65n7cpgUYjrUPLPUNw3zX3t pLytDV/DR3dyx2r+9yMXSIRKUPItos83AKNEv4ycGMmq+punGFCH5JklDqLJBEkBrwyGVs+7Fbvw9g xyjExLhhJyHegyAYTz1C5SNsO4p59bVKNvV3ybBJaYQgEUmBegM1Pm3vfzmVpi9Zxevx9SDYenucib 3/12fPKpQB3I1dBBrrOfwSudUdT+EX/t5YbrFrXY5c6wNXlmS5zLTfa28INk0/yCq6Vt4RSdaaUew9 38jamA7aEV0kSROgdoN7wW4oeI4IVnl1EqFmpL6TZVyDI= X-Developer-Key: i=andrzej.hajda@intel.com; a=openpgp; fpr=FCA8443134DDBF5DE17F034D2362B293DE10FDD7 X-BeenThere: igt-dev@lists.freedesktop.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: Development mailing list for IGT GPU Tools List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: igt-dev-bounces@lists.freedesktop.org Sender: "igt-dev" Implement tooling for building shaders for specific generations. The library allows you to build and run shader from precompiled blocks and provides an abstraction layer over gpgpu pipeline. Signed-off-by: Dominik Grzegorzek Signed-off-by: Christoph Manszewski Signed-off-by: Andrzej Hajda --- lib/gpgpu_shader.c | 211 +++++++++++++++++++++++++++++++++++++++++++++++++++++ lib/gpgpu_shader.h | 38 ++++++++++ lib/meson.build | 1 + 3 files changed, 250 insertions(+) diff --git a/lib/gpgpu_shader.c b/lib/gpgpu_shader.c new file mode 100644 index 000000000000..d14301789421 --- /dev/null +++ b/lib/gpgpu_shader.c @@ -0,0 +1,211 @@ +// SPDX-License-Identifier: MIT +/* + * Copyright © 2024 Intel Corporation + * + * Author: Dominik Grzegorzek + */ + +#include + +#include "ioctl_wrappers.h" +#include "gpgpu_shader.h" +#include "gpu_cmds.h" + +#define SUPPORTED_GEN_VER 1200 /* Support TGL and up */ + +#define PAGE_SIZE 4096 +#define BATCH_STATE_SPLIT 2048 +/* VFE STATE params */ +#define THREADS (1 << 16) /* max value */ +#define GEN8_GPGPU_URB_ENTRIES 1 +#define GPGPU_URB_SIZE 0 +#define GPGPU_CURBE_SIZE 0 +#define GEN7_VFE_STATE_GPGPU_MODE 1 + +static uint32_t fill_sip(struct intel_bb *ibb, + const uint32_t sip[][4], + const size_t size) +{ + uint32_t *sip_dst; + uint32_t offset; + + intel_bb_ptr_align(ibb, 16); + sip_dst = intel_bb_ptr(ibb); + offset = intel_bb_offset(ibb); + + memcpy(sip_dst, sip, size); + + intel_bb_ptr_add(ibb, size); + + return offset; +} + +static void emit_sip(struct intel_bb *ibb, const uint64_t offset) +{ + intel_bb_out(ibb, GEN4_STATE_SIP | (3 - 2)); + intel_bb_out(ibb, lower_32_bits(offset)); + intel_bb_out(ibb, upper_32_bits(offset)); +} + +static void +__xelp_gpgpu_execfunc(struct intel_bb *ibb, + struct intel_buf *target, + unsigned int x_dim, unsigned int y_dim, + struct gpgpu_shader *shdr, + struct gpgpu_shader *sip, + uint64_t ring, bool explicit_engine) +{ + uint32_t interface_descriptor, sip_offset; + uint64_t engine; + + intel_bb_add_intel_buf(ibb, target, true); + + intel_bb_ptr_set(ibb, BATCH_STATE_SPLIT); + + interface_descriptor = gen8_fill_interface_descriptor(ibb, target, + shdr->instr, + 4 * shdr->size); + + if (sip && sip->size) + sip_offset = fill_sip(ibb, sip->instr, 4 * sip->size); + else + sip_offset = 0; + + intel_bb_ptr_set(ibb, 0); + + /* GPGPU pipeline */ + intel_bb_out(ibb, GEN7_PIPELINE_SELECT | GEN9_PIPELINE_SELECTION_MASK | + PIPELINE_SELECT_GPGPU); + + gen9_emit_state_base_address(ibb); + + xelp_emit_vfe_state(ibb, THREADS, GEN8_GPGPU_URB_ENTRIES, + GPGPU_URB_SIZE, GPGPU_CURBE_SIZE, true); + + gen7_emit_interface_descriptor_load(ibb, interface_descriptor); + + if (sip_offset) + emit_sip(ibb, sip_offset); + + gen8_emit_gpgpu_walk(ibb, 0, 0, x_dim * 16, y_dim); + + intel_bb_out(ibb, MI_BATCH_BUFFER_END); + intel_bb_ptr_align(ibb, 32); + + engine = explicit_engine ? ring : I915_EXEC_DEFAULT; + intel_bb_exec(ibb, intel_bb_offset(ibb), + engine | I915_EXEC_NO_RELOC, false); +} + +static void +__xehp_gpgpu_execfunc(struct intel_bb *ibb, + struct intel_buf *target, + unsigned int x_dim, unsigned int y_dim, + struct gpgpu_shader *shdr, + struct gpgpu_shader *sip, + uint64_t ring, bool explicit_engine) +{ + struct xehp_interface_descriptor_data idd; + uint32_t sip_offset; + uint64_t engine; + + intel_bb_add_intel_buf(ibb, target, true); + + intel_bb_ptr_set(ibb, BATCH_STATE_SPLIT); + + xehp_fill_interface_descriptor(ibb, target, shdr->instr, + 4 * shdr->size, &idd); + + if (sip && sip->size) + sip_offset = fill_sip(ibb, sip->instr, 4 * sip->size); + else + sip_offset = 0; + + intel_bb_ptr_set(ibb, 0); + + /* GPGPU pipeline */ + intel_bb_out(ibb, GEN7_PIPELINE_SELECT | GEN9_PIPELINE_SELECTION_MASK | + PIPELINE_SELECT_GPGPU); + xehp_emit_state_base_address(ibb); + xehp_emit_state_compute_mode(ibb); + xehp_emit_state_binding_table_pool_alloc(ibb); + xehp_emit_cfe_state(ibb, THREADS); + + if (sip_offset) + emit_sip(ibb, sip_offset); + + xehp_emit_compute_walk(ibb, 0, 0, x_dim * 16, y_dim, &idd, 0x0); + + intel_bb_out(ibb, MI_BATCH_BUFFER_END); + intel_bb_ptr_align(ibb, 32); + + engine = explicit_engine ? ring : I915_EXEC_DEFAULT; + intel_bb_exec(ibb, intel_bb_offset(ibb), + engine | I915_EXEC_NO_RELOC, false); + +} + +/** + * gpgpu_shader_exec: + * @ibb: pointer to initialized intel_bb + * @target: pointer to initialized intel_buf to be written by shader/sip + * @x_dim: gpgpu/compute walker thread group width + * @y_dim: gpgpu/compute walker thread group height + * @shdr: shader to be executed + * @sip: sip to be executed, can be NULL + * @ring: engine index + * @explicit_engine: whether to use provided engine index + * + * Execute provided shader in asynchronous fashion. To wait for completion, + * caller has to use the provided ibb handle. + */ +void gpgpu_shader_exec(struct intel_bb *ibb, + struct intel_buf *target, + unsigned int x_dim, unsigned int y_dim, + struct gpgpu_shader *shdr, + struct gpgpu_shader *sip, + uint64_t ring, bool explicit_engine) +{ + igt_require(shdr->gen_ver >= SUPPORTED_GEN_VER); + igt_assert(ibb->size >= PAGE_SIZE); + igt_assert(ibb->ptr == ibb->batch); + + if (shdr->gen_ver >= 1250) + __xehp_gpgpu_execfunc(ibb, target, x_dim, y_dim, shdr, sip, + ring, explicit_engine); + else + __xelp_gpgpu_execfunc(ibb, target, x_dim, y_dim, shdr, sip, + ring, explicit_engine); +} + +/** + * gpgpu_shader_create: + * @fd: drm fd - i915 or xe + * + * Creates empty shader. + * + * Returns: pointer to empty shader struct. + */ +struct gpgpu_shader *gpgpu_shader_create(int fd) +{ + struct gpgpu_shader *shdr = calloc(1, sizeof(struct gpgpu_shader)); + const struct intel_device_info *info; + + info = intel_get_device_info(intel_get_drm_devid(fd)); + shdr->gen_ver = 100 * info->graphics_ver + info->graphics_rel; + shdr->max_size = 16 * 4; + shdr->code = malloc(4 * shdr->max_size); + return shdr; +} + +/** + * gpgpu_shader_destroy: + * @shdr: pointer to shader struct created with 'gpgpu_shader_create' + * + * Frees resources of gpgpu_shader struct. + */ +void gpgpu_shader_destroy(struct gpgpu_shader *shdr) +{ + free(shdr->code); + free(shdr); +} diff --git a/lib/gpgpu_shader.h b/lib/gpgpu_shader.h new file mode 100644 index 000000000000..02f6f1aad1e3 --- /dev/null +++ b/lib/gpgpu_shader.h @@ -0,0 +1,38 @@ +/* SPDX-License-Identifier: MIT */ +/* + * Copyright © 2024 Intel Corporation + */ + +#ifndef GPGPU_SHADER_H +#define GPGPU_SHADER_H + +#include +#include +#include + +struct intel_bb; +struct intel_buf; + +struct gpgpu_shader { + uint32_t gen_ver; + uint32_t size; + uint32_t max_size; + union { + uint32_t *code; + uint32_t (*instr)[4]; + }; +}; + +struct gpgpu_shader *gpgpu_shader_create(int fd); +void gpgpu_shader_destroy(struct gpgpu_shader *shdr); + +void gpgpu_shader_dump(struct gpgpu_shader *shdr); + +void gpgpu_shader_exec(struct intel_bb *ibb, + struct intel_buf *target, + unsigned int x_dim, unsigned int y_dim, + struct gpgpu_shader *shdr, + struct gpgpu_shader *sip, + uint64_t ring, bool explicit_engine); + +#endif /* GPGPU_SHADER_H */ diff --git a/lib/meson.build b/lib/meson.build index e2f740c116f8..0a3084f8aea2 100644 --- a/lib/meson.build +++ b/lib/meson.build @@ -72,6 +72,7 @@ lib_sources = [ 'media_spin.c', 'media_fill.c', 'gpgpu_fill.c', + 'gpgpu_shader.c', 'gpu_cmds.c', 'rendercopy_i915.c', 'rendercopy_i830.c', -- 2.34.1