From patchwork Mon Feb 4 15:27:24 2013 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: "Lespiau, Damien" X-Patchwork-Id: 2092921 Return-Path: X-Original-To: patchwork-intel-gfx@patchwork.kernel.org Delivered-To: patchwork-process-083081@patchwork2.kernel.org Received: from gabe.freedesktop.org (gabe.freedesktop.org [131.252.210.177]) by patchwork2.kernel.org (Postfix) with ESMTP id 50774DFE82 for ; Mon, 4 Feb 2013 15:44:41 +0000 (UTC) Received: from gabe.freedesktop.org (localhost [127.0.0.1]) by gabe.freedesktop.org (Postfix) with ESMTP id 27D42E614C for ; Mon, 4 Feb 2013 07:44:41 -0800 (PST) X-Original-To: intel-gfx@lists.freedesktop.org Delivered-To: intel-gfx@lists.freedesktop.org Received: from mga02.intel.com (mga02.intel.com [134.134.136.20]) by gabe.freedesktop.org (Postfix) with ESMTP id 2C1A5E5C5E for ; Mon, 4 Feb 2013 07:29:27 -0800 (PST) Received: from orsmga001.jf.intel.com ([10.7.209.18]) by orsmga101.jf.intel.com with ESMTP; 04 Feb 2013 07:29:26 -0800 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="4.84,600,1355126400"; d="scan'208";a="257441790" Received: from unknown (HELO dyon.amr.corp.intel.com) ([10.255.12.132]) by orsmga001.jf.intel.com with ESMTP; 04 Feb 2013 07:29:20 -0800 From: Damien Lespiau To: intel-gfx@lists.freedesktop.org Date: Mon, 4 Feb 2013 15:27:24 +0000 Message-Id: <1359991705-5254-30-git-send-email-damien.lespiau@intel.com> X-Mailer: git-send-email 1.7.7.5 In-Reply-To: <1359991705-5254-1-git-send-email-damien.lespiau@intel.com> References: <1359991705-5254-1-git-send-email-damien.lespiau@intel.com> Subject: [Intel-gfx] [PATCH 29/90] assembler: Import brw_eu.c X-BeenThere: intel-gfx@lists.freedesktop.org X-Mailman-Version: 2.1.13 Precedence: list List-Id: Intel graphics driver community testing & development List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , MIME-Version: 1.0 Sender: intel-gfx-bounces+patchwork-intel-gfx=patchwork.kernel.org@lists.freedesktop.org Errors-To: intel-gfx-bounces+patchwork-intel-gfx=patchwork.kernel.org@lists.freedesktop.org Another step the road of importing Mesa's emission code. Signed-off-by: Damien Lespiau --- assembler/Makefile.am | 1 + assembler/brw_eu.c | 269 +++++++++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 270 insertions(+), 0 deletions(-) create mode 100644 assembler/brw_eu.c diff --git a/assembler/Makefile.am b/assembler/Makefile.am index 48e38d0..e4468a4 100644 --- a/assembler/Makefile.am +++ b/assembler/Makefile.am @@ -11,6 +11,7 @@ libbrw_la_SOURCES = \ brw_disasm.c \ brw_defines.h \ brw_eu.h \ + brw_eu.c \ brw_eu_compact.c \ brw_reg.h \ brw_structs.h \ diff --git a/assembler/brw_eu.c b/assembler/brw_eu.c new file mode 100644 index 0000000..1641c95 --- /dev/null +++ b/assembler/brw_eu.c @@ -0,0 +1,269 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include + +#include "gen4asm.h" +#include "brw_context.h" +#include "brw_defines.h" +#include "brw_eu.h" + +#include "ralloc.h" + +/* Returns the corresponding conditional mod for swapping src0 and + * src1 in e.g. CMP. + */ +uint32_t +brw_swap_cmod(uint32_t cmod) +{ + switch (cmod) { + case BRW_CONDITIONAL_Z: + case BRW_CONDITIONAL_NZ: + return cmod; + case BRW_CONDITIONAL_G: + return BRW_CONDITIONAL_L; + case BRW_CONDITIONAL_GE: + return BRW_CONDITIONAL_LE; + case BRW_CONDITIONAL_L: + return BRW_CONDITIONAL_G; + case BRW_CONDITIONAL_LE: + return BRW_CONDITIONAL_GE; + default: + return ~0; + } +} + + +/* How does predicate control work when execution_size != 8? Do I + * need to test/set for 0xffff when execution_size is 16? + */ +void brw_set_predicate_control_flag_value( struct brw_compile *p, GLuint value ) +{ + p->current->header.predicate_control = BRW_PREDICATE_NONE; + + if (value != 0xff) { + if (value != p->flag_value) { + brw_push_insn_state(p); + brw_MOV(p, brw_flag_reg(0, 0), brw_imm_uw(value)); + p->flag_value = value; + brw_pop_insn_state(p); + } + + p->current->header.predicate_control = BRW_PREDICATE_NORMAL; + } +} + +void brw_set_predicate_control( struct brw_compile *p, GLuint pc ) +{ + p->current->header.predicate_control = pc; +} + +void brw_set_predicate_inverse(struct brw_compile *p, bool predicate_inverse) +{ + p->current->header.predicate_inverse = predicate_inverse; +} + +void brw_set_conditionalmod( struct brw_compile *p, GLuint conditional ) +{ + p->current->header.destreg__conditionalmod = conditional; +} + +void brw_set_flag_reg(struct brw_compile *p, int reg, int subreg) +{ + p->current->bits2.da1.flag_reg_nr = reg; + p->current->bits2.da1.flag_subreg_nr = subreg; +} + +void brw_set_access_mode( struct brw_compile *p, GLuint access_mode ) +{ + p->current->header.access_mode = access_mode; +} + +void +brw_set_compression_control(struct brw_compile *p, + enum brw_compression compression_control) +{ + p->compressed = (compression_control == BRW_COMPRESSION_COMPRESSED); + + if (p->brw->intel.gen >= 6) { + /* Since we don't use the 32-wide support in gen6, we translate + * the pre-gen6 compression control here. + */ + switch (compression_control) { + case BRW_COMPRESSION_NONE: + /* This is the "use the first set of bits of dmask/vmask/arf + * according to execsize" option. + */ + p->current->header.compression_control = GEN6_COMPRESSION_1Q; + break; + case BRW_COMPRESSION_2NDHALF: + /* For 8-wide, this is "use the second set of 8 bits." */ + p->current->header.compression_control = GEN6_COMPRESSION_2Q; + break; + case BRW_COMPRESSION_COMPRESSED: + /* For 16-wide instruction compression, use the first set of 16 bits + * since we don't do 32-wide dispatch. + */ + p->current->header.compression_control = GEN6_COMPRESSION_1H; + break; + default: + assert(!"not reached"); + p->current->header.compression_control = GEN6_COMPRESSION_1H; + break; + } + } else { + p->current->header.compression_control = compression_control; + } +} + +void brw_set_mask_control( struct brw_compile *p, GLuint value ) +{ + p->current->header.mask_control = value; +} + +void brw_set_saturate( struct brw_compile *p, bool enable ) +{ + p->current->header.saturate = enable; +} + +void brw_set_acc_write_control(struct brw_compile *p, GLuint value) +{ + if (p->brw->intel.gen >= 6) + p->current->header.acc_wr_control = value; +} + +void brw_push_insn_state( struct brw_compile *p ) +{ + assert(p->current != &p->stack[BRW_EU_MAX_INSN_STACK-1]); + memcpy(p->current+1, p->current, sizeof(struct brw_instruction)); + p->compressed_stack[p->current - p->stack] = p->compressed; + p->current++; +} + +void brw_pop_insn_state( struct brw_compile *p ) +{ + assert(p->current != p->stack); + p->current--; + p->compressed = p->compressed_stack[p->current - p->stack]; +} + + +/*********************************************************************** + */ +void +brw_init_compile(struct brw_context *brw, struct brw_compile *p, void *mem_ctx) +{ + memset(p, 0, sizeof(*p)); + + p->brw = brw; + /* + * Set the initial instruction store array size to 1024, if found that + * isn't enough, then it will double the store size at brw_next_insn() + * until out of memory. + */ + p->store_size = 1024; + p->store = rzalloc_array(mem_ctx, struct brw_instruction, p->store_size); + p->nr_insn = 0; + p->current = p->stack; + p->compressed = false; + memset(p->current, 0, sizeof(p->current[0])); + + p->mem_ctx = mem_ctx; + + /* Some defaults? + */ + brw_set_mask_control(p, BRW_MASK_ENABLE); /* what does this do? */ + brw_set_saturate(p, 0); + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + brw_set_predicate_control_flag_value(p, 0xff); + + /* Set up control flow stack */ + p->if_stack_depth = 0; + p->if_stack_array_size = 16; + p->if_stack = rzalloc_array(mem_ctx, int, p->if_stack_array_size); + + p->loop_stack_depth = 0; + p->loop_stack_array_size = 16; + p->loop_stack = rzalloc_array(mem_ctx, int, p->loop_stack_array_size); + p->if_depth_in_loop = rzalloc_array(mem_ctx, int, p->loop_stack_array_size); + + brw_init_compaction_tables(&brw->intel); +} + + +const GLuint *brw_get_program( struct brw_compile *p, + GLuint *sz ) +{ + brw_compact_instructions(p); + + *sz = p->next_insn_offset; + return (const GLuint *)p->store; +} + +void +brw_dump_compile(struct brw_compile *p, FILE *out, int start, int end) +{ + struct brw_context *brw = p->brw; + struct intel_context *intel = &brw->intel; + void *store = p->store; + bool dump_hex = false; + + for (int offset = start; offset < end;) { + struct brw_instruction *insn = store + offset; + struct brw_instruction uncompacted; + printf("0x%08x: ", offset); + + if (insn->header.cmpt_control) { + struct brw_compact_instruction *compacted = (void *)insn; + if (dump_hex) { + printf("0x%08x 0x%08x ", + ((uint32_t *)insn)[1], + ((uint32_t *)insn)[0]); + } + + brw_uncompact_instruction(intel, &uncompacted, compacted); + insn = &uncompacted; + offset += 8; + } else { + if (dump_hex) { + printf("0x%08x 0x%08x 0x%08x 0x%08x ", + ((uint32_t *)insn)[3], + ((uint32_t *)insn)[2], + ((uint32_t *)insn)[1], + ((uint32_t *)insn)[0]); + } + offset += 16; + } + + brw_disasm(stdout, insn, p->brw->intel.gen); + } +}