[Intel-gfx] [PATCH] drm/i915: Add Wa_14015150844
kernel test robot
lkp at intel.com
Wed Aug 30 18:34:22 UTC 2023
Hi Shekhar,
kernel test robot noticed the following build errors:
[auto build test ERROR on drm-tip/drm-tip]
url: https://github.com/intel-lab-lkp/linux/commits/Shekhar-Chauhan/drm-i915-Add-Wa_14015150844/20230831-000233
base: git://anongit.freedesktop.org/drm/drm-tip drm-tip
patch link: https://lore.kernel.org/r/20230830160001.2395993-1-shekhar.chauhan%40intel.com
patch subject: [Intel-gfx] [PATCH] drm/i915: Add Wa_14015150844
config: x86_64-defconfig (https://download.01.org/0day-ci/archive/20230831/202308310238.2Dnlp2IN-lkp@intel.com/config)
compiler: gcc-11 (Debian 11.3.0-12) 11.3.0
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20230831/202308310238.2Dnlp2IN-lkp@intel.com/reproduce)
If you fix the issue in a separate patch/commit (i.e. not just a new version of
the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <lkp at intel.com>
| Closes: https://lore.kernel.org/oe-kbuild-all/202308310238.2Dnlp2IN-lkp@intel.com/
All errors (new ones prefixed by >>):
drivers/gpu/drm/i915/gt/intel_workarounds.c: In function 'rcs_engine_wa_init':
>> drivers/gpu/drm/i915/gt/intel_workarounds.c:2330:13: error: implicit declaration of function 'IS_DG2_GRAPHICS_STEP'; did you mean 'IS_GRAPHICS_STEP'? [-Werror=implicit-function-declaration]
2330 | if (IS_DG2_GRAPHICS_STEP(i915, G10, STEP_C0, STEP_FOREVER) ||
| ^~~~~~~~~~~~~~~~~~~~
| IS_GRAPHICS_STEP
>> drivers/gpu/drm/i915/gt/intel_workarounds.c:2330:40: error: 'G10' undeclared (first use in this function)
2330 | if (IS_DG2_GRAPHICS_STEP(i915, G10, STEP_C0, STEP_FOREVER) ||
| ^~~
drivers/gpu/drm/i915/gt/intel_workarounds.c:2330:40: note: each undeclared identifier is reported only once for each function it appears in
>> drivers/gpu/drm/i915/gt/intel_workarounds.c:2331:40: error: 'G11' undeclared (first use in this function)
2331 | IS_DG2_GRAPHICS_STEP(i915, G11, STEP_B0, STEP_FOREVER) ||
| ^~~
>> drivers/gpu/drm/i915/gt/intel_workarounds.c:2333:13: error: implicit declaration of function 'IS_MTL_GRAPHICS_STEP'; did you mean 'INTEL_GRAPHICS_STEP'? [-Werror=implicit-function-declaration]
2333 | IS_MTL_GRAPHICS_STEP(i915, M, STEP_A0, STEP_FOREVER) ||
| ^~~~~~~~~~~~~~~~~~~~
| INTEL_GRAPHICS_STEP
>> drivers/gpu/drm/i915/gt/intel_workarounds.c:2333:40: error: 'M' undeclared (first use in this function)
2333 | IS_MTL_GRAPHICS_STEP(i915, M, STEP_A0, STEP_FOREVER) ||
| ^
>> drivers/gpu/drm/i915/gt/intel_workarounds.c:2334:40: error: 'P' undeclared (first use in this function)
2334 | IS_MTL_GRAPHICS_STEP(i915, P, STEP_A0, STEP_FOREVER)) {
| ^
cc1: some warnings being treated as errors
vim +2330 drivers/gpu/drm/i915/gt/intel_workarounds.c
2292
2293 static void
2294 rcs_engine_wa_init(struct intel_engine_cs *engine, struct i915_wa_list *wal)
2295 {
2296 struct drm_i915_private *i915 = engine->i915;
2297 struct intel_gt *gt = engine->gt;
2298
2299 if (IS_GFX_GT_IP_STEP(gt, IP_VER(12, 70), STEP_A0, STEP_B0) ||
2300 IS_GFX_GT_IP_STEP(gt, IP_VER(12, 71), STEP_A0, STEP_B0)) {
2301 /* Wa_22014600077 */
2302 wa_mcr_masked_en(wal, GEN10_CACHE_MODE_SS,
2303 ENABLE_EU_COUNT_FOR_TDL_FLUSH);
2304 }
2305
2306 if (IS_GFX_GT_IP_STEP(gt, IP_VER(12, 70), STEP_A0, STEP_B0) ||
2307 IS_GFX_GT_IP_STEP(gt, IP_VER(12, 71), STEP_A0, STEP_B0) ||
2308 IS_DG2(i915)) {
2309 /* Wa_1509727124 */
2310 wa_mcr_masked_en(wal, GEN10_SAMPLER_MODE,
2311 SC_DISABLE_POWER_OPTIMIZATION_EBB);
2312 }
2313
2314 if (IS_GFX_GT_IP_STEP(gt, IP_VER(12, 70), STEP_A0, STEP_B0) ||
2315 IS_DG2(i915)) {
2316 /* Wa_22012856258 */
2317 wa_mcr_masked_en(wal, GEN8_ROW_CHICKEN2,
2318 GEN12_DISABLE_READ_SUPPRESSION);
2319 }
2320
2321 if (IS_DG2(i915)) {
2322 /*
2323 * Wa_22010960976:dg2
2324 * Wa_14013347512:dg2
2325 */
2326 wa_mcr_masked_dis(wal, XEHP_HDC_CHICKEN0,
2327 LSC_L1_FLUSH_CTL_3D_DATAPORT_FLUSH_EVENTS_MASK);
2328 }
2329
> 2330 if (IS_DG2_GRAPHICS_STEP(i915, G10, STEP_C0, STEP_FOREVER) ||
> 2331 IS_DG2_GRAPHICS_STEP(i915, G11, STEP_B0, STEP_FOREVER) ||
2332 IS_DG2_G12(i915) ||
> 2333 IS_MTL_GRAPHICS_STEP(i915, M, STEP_A0, STEP_FOREVER) ||
> 2334 IS_MTL_GRAPHICS_STEP(i915, P, STEP_A0, STEP_FOREVER)) {
2335 /* Wa_14015150844 */
2336 wa_mcr_masked_dis(wal, XEHP_HDC_CHICKEN0,
2337 ATOMIC_CHAINING_TYPED_WRITES);
2338 }
2339
2340 if (IS_DG2_G11(i915) || IS_DG2_G10(i915)) {
2341 /* Wa_22014600077:dg2 */
2342 wa_mcr_add(wal, GEN10_CACHE_MODE_SS, 0,
2343 _MASKED_BIT_ENABLE(ENABLE_EU_COUNT_FOR_TDL_FLUSH),
2344 0 /* Wa_14012342262 write-only reg, so skip verification */,
2345 true);
2346 }
2347
2348 if (IS_DG2(i915) || IS_ALDERLAKE_P(i915) || IS_ALDERLAKE_S(i915) ||
2349 IS_DG1(i915) || IS_ROCKETLAKE(i915) || IS_TIGERLAKE(i915)) {
2350 /*
2351 * Wa_1606700617:tgl,dg1,adl-p
2352 * Wa_22010271021:tgl,rkl,dg1,adl-s,adl-p
2353 * Wa_14010826681:tgl,dg1,rkl,adl-p
2354 * Wa_18019627453:dg2
2355 */
2356 wa_masked_en(wal,
2357 GEN9_CS_DEBUG_MODE1,
2358 FF_DOP_CLOCK_GATE_DISABLE);
2359 }
2360
2361 if (IS_ALDERLAKE_P(i915) || IS_ALDERLAKE_S(i915) || IS_DG1(i915) ||
2362 IS_ROCKETLAKE(i915) || IS_TIGERLAKE(i915)) {
2363 /* Wa_1606931601:tgl,rkl,dg1,adl-s,adl-p */
2364 wa_mcr_masked_en(wal, GEN8_ROW_CHICKEN2, GEN12_DISABLE_EARLY_READ);
2365
2366 /*
2367 * Wa_1407928979:tgl A*
2368 * Wa_18011464164:tgl[B0+],dg1[B0+]
2369 * Wa_22010931296:tgl[B0+],dg1[B0+]
2370 * Wa_14010919138:rkl,dg1,adl-s,adl-p
2371 */
2372 wa_write_or(wal, GEN7_FF_THREAD_MODE,
2373 GEN12_FF_TESSELATION_DOP_GATE_DISABLE);
2374
2375 /* Wa_1406941453:tgl,rkl,dg1,adl-s,adl-p */
2376 wa_mcr_masked_en(wal,
2377 GEN10_SAMPLER_MODE,
2378 ENABLE_SMALLPL);
2379 }
2380
2381 if (IS_ALDERLAKE_P(i915) || IS_ALDERLAKE_S(i915) ||
2382 IS_ROCKETLAKE(i915) || IS_TIGERLAKE(i915)) {
2383 /* Wa_1409804808 */
2384 wa_mcr_masked_en(wal, GEN8_ROW_CHICKEN2,
2385 GEN12_PUSH_CONST_DEREF_HOLD_DIS);
2386
2387 /* Wa_14010229206 */
2388 wa_mcr_masked_en(wal, GEN9_ROW_CHICKEN4, GEN12_DISABLE_TDL_PUSH);
2389 }
2390
2391 if (IS_ROCKETLAKE(i915) || IS_TIGERLAKE(i915) || IS_ALDERLAKE_P(i915)) {
2392 /*
2393 * Wa_1607297627
2394 *
2395 * On TGL and RKL there are multiple entries for this WA in the
2396 * BSpec; some indicate this is an A0-only WA, others indicate
2397 * it applies to all steppings so we trust the "all steppings."
2398 */
2399 wa_masked_en(wal,
2400 RING_PSMI_CTL(RENDER_RING_BASE),
2401 GEN12_WAIT_FOR_EVENT_POWER_DOWN_DISABLE |
2402 GEN8_RC_SEMA_IDLE_MSG_DISABLE);
2403 }
2404
2405 if (GRAPHICS_VER(i915) == 11) {
2406 /* This is not an Wa. Enable for better image quality */
2407 wa_masked_en(wal,
2408 _3D_CHICKEN3,
2409 _3D_CHICKEN3_AA_LINE_QUALITY_FIX_ENABLE);
2410
2411 /*
2412 * Wa_1405543622:icl
2413 * Formerly known as WaGAPZPriorityScheme
2414 */
2415 wa_write_or(wal,
2416 GEN8_GARBCNTL,
2417 GEN11_ARBITRATION_PRIO_ORDER_MASK);
2418
2419 /*
2420 * Wa_1604223664:icl
2421 * Formerly known as WaL3BankAddressHashing
2422 */
2423 wa_write_clr_set(wal,
2424 GEN8_GARBCNTL,
2425 GEN11_HASH_CTRL_EXCL_MASK,
2426 GEN11_HASH_CTRL_EXCL_BIT0);
2427 wa_write_clr_set(wal,
2428 GEN11_GLBLINVL,
2429 GEN11_BANK_HASH_ADDR_EXCL_MASK,
2430 GEN11_BANK_HASH_ADDR_EXCL_BIT0);
2431
2432 /*
2433 * Wa_1405733216:icl
2434 * Formerly known as WaDisableCleanEvicts
2435 */
2436 wa_mcr_write_or(wal,
2437 GEN8_L3SQCREG4,
2438 GEN11_LQSC_CLEAN_EVICT_DISABLE);
2439
2440 /* Wa_1606682166:icl */
2441 wa_write_or(wal,
2442 GEN7_SARCHKMD,
2443 GEN7_DISABLE_SAMPLER_PREFETCH);
2444
2445 /* Wa_1409178092:icl */
2446 wa_mcr_write_clr_set(wal,
2447 GEN11_SCRATCH2,
2448 GEN11_COHERENT_PARTIAL_WRITE_MERGE_ENABLE,
2449 0);
2450
2451 /* WaEnable32PlaneMode:icl */
2452 wa_masked_en(wal, GEN9_CSFE_CHICKEN1_RCS,
2453 GEN11_ENABLE_32_PLANE_MODE);
2454
2455 /*
2456 * Wa_1408767742:icl[a2..forever],ehl[all]
2457 * Wa_1605460711:icl[a0..c0]
2458 */
2459 wa_write_or(wal,
2460 GEN7_FF_THREAD_MODE,
2461 GEN12_FF_TESSELATION_DOP_GATE_DISABLE);
2462
2463 /* Wa_22010271021 */
2464 wa_masked_en(wal,
2465 GEN9_CS_DEBUG_MODE1,
2466 FF_DOP_CLOCK_GATE_DISABLE);
2467 }
2468
2469 /*
2470 * Intel platforms that support fine-grained preemption (i.e., gen9 and
2471 * beyond) allow the kernel-mode driver to choose between two different
2472 * options for controlling preemption granularity and behavior.
2473 *
2474 * Option 1 (hardware default):
2475 * Preemption settings are controlled in a global manner via
2476 * kernel-only register CS_DEBUG_MODE1 (0x20EC). Any granularity
2477 * and settings chosen by the kernel-mode driver will apply to all
2478 * userspace clients.
2479 *
2480 * Option 2:
2481 * Preemption settings are controlled on a per-context basis via
2482 * register CS_CHICKEN1 (0x2580). CS_CHICKEN1 is saved/restored on
2483 * context switch and is writable by userspace (e.g., via
2484 * MI_LOAD_REGISTER_IMMEDIATE instructions placed in a batch buffer)
2485 * which allows different userspace drivers/clients to select
2486 * different settings, or to change those settings on the fly in
2487 * response to runtime needs. This option was known by name
2488 * "FtrPerCtxtPreemptionGranularityControl" at one time, although
2489 * that name is somewhat misleading as other non-granularity
2490 * preemption settings are also impacted by this decision.
2491 *
2492 * On Linux, our policy has always been to let userspace drivers
2493 * control preemption granularity/settings (Option 2). This was
2494 * originally mandatory on gen9 to prevent ABI breakage (old gen9
2495 * userspace developed before object-level preemption was enabled would
2496 * not behave well if i915 were to go with Option 1 and enable that
2497 * preemption in a global manner). On gen9 each context would have
2498 * object-level preemption disabled by default (see
2499 * WaDisable3DMidCmdPreemption in gen9_ctx_workarounds_init), but
2500 * userspace drivers could opt-in to object-level preemption as they
2501 * saw fit. For post-gen9 platforms, we continue to utilize Option 2;
2502 * even though it is no longer necessary for ABI compatibility when
2503 * enabling a new platform, it does ensure that userspace will be able
2504 * to implement any workarounds that show up requiring temporary
2505 * adjustments to preemption behavior at runtime.
2506 *
2507 * Notes/Workarounds:
2508 * - Wa_14015141709: On DG2 and early steppings of MTL,
2509 * CS_CHICKEN1[0] does not disable object-level preemption as
2510 * it is supposed to (nor does CS_DEBUG_MODE1[0] if we had been
2511 * using Option 1). Effectively this means userspace is unable
2512 * to disable object-level preemption on these platforms/steppings
2513 * despite the setting here.
2514 *
2515 * - Wa_16013994831: May require that userspace program
2516 * CS_CHICKEN1[10] when certain runtime conditions are true.
2517 * Userspace requires Option 2 to be in effect for their update of
2518 * CS_CHICKEN1[10] to be effective.
2519 *
2520 * Other workarounds may appear in the future that will also require
2521 * Option 2 behavior to allow proper userspace implementation.
2522 */
2523 if (GRAPHICS_VER(i915) >= 9)
2524 wa_masked_en(wal,
2525 GEN7_FF_SLICE_CS_CHICKEN1,
2526 GEN9_FFSC_PERCTX_PREEMPT_CTRL);
2527
2528 if (IS_SKYLAKE(i915) ||
2529 IS_KABYLAKE(i915) ||
2530 IS_COFFEELAKE(i915) ||
2531 IS_COMETLAKE(i915)) {
2532 /* WaEnableGapsTsvCreditFix:skl,kbl,cfl */
2533 wa_write_or(wal,
2534 GEN8_GARBCNTL,
2535 GEN9_GAPS_TSV_CREDIT_DISABLE);
2536 }
2537
2538 if (IS_BROXTON(i915)) {
2539 /* WaDisablePooledEuLoadBalancingFix:bxt */
2540 wa_masked_en(wal,
2541 FF_SLICE_CS_CHICKEN2,
2542 GEN9_POOLED_EU_LOAD_BALANCING_FIX_DISABLE);
2543 }
2544
2545 if (GRAPHICS_VER(i915) == 9) {
2546 /* WaContextSwitchWithConcurrentTLBInvalidate:skl,bxt,kbl,glk,cfl */
2547 wa_masked_en(wal,
2548 GEN9_CSFE_CHICKEN1_RCS,
2549 GEN9_PREEMPT_GPGPU_SYNC_SWITCH_DISABLE);
2550
2551 /* WaEnableLbsSlaRetryTimerDecrement:skl,bxt,kbl,glk,cfl */
2552 wa_mcr_write_or(wal,
2553 BDW_SCRATCH1,
2554 GEN9_LBS_SLA_RETRY_TIMER_DECREMENT_ENABLE);
2555
2556 /* WaProgramL3SqcReg1DefaultForPerf:bxt,glk */
2557 if (IS_GEN9_LP(i915))
2558 wa_mcr_write_clr_set(wal,
2559 GEN8_L3SQCREG1,
2560 L3_PRIO_CREDITS_MASK,
2561 L3_GENERAL_PRIO_CREDITS(62) |
2562 L3_HIGH_PRIO_CREDITS(2));
2563
2564 /* WaOCLCoherentLineFlush:skl,bxt,kbl,cfl */
2565 wa_mcr_write_or(wal,
2566 GEN8_L3SQCREG4,
2567 GEN8_LQSC_FLUSH_COHERENT_LINES);
2568
2569 /* Disable atomics in L3 to prevent unrecoverable hangs */
2570 wa_write_clr_set(wal, GEN9_SCRATCH_LNCF1,
2571 GEN9_LNCF_NONIA_COHERENT_ATOMICS_ENABLE, 0);
2572 wa_mcr_write_clr_set(wal, GEN8_L3SQCREG4,
2573 GEN8_LQSQ_NONIA_COHERENT_ATOMICS_ENABLE, 0);
2574 wa_mcr_write_clr_set(wal, GEN9_SCRATCH1,
2575 EVICTION_PERF_FIX_ENABLE, 0);
2576 }
2577
2578 if (IS_HASWELL(i915)) {
2579 /* WaSampleCChickenBitEnable:hsw */
2580 wa_masked_en(wal,
2581 HSW_HALF_SLICE_CHICKEN3, HSW_SAMPLE_C_PERFORMANCE);
2582
2583 wa_masked_dis(wal,
2584 CACHE_MODE_0_GEN7,
2585 /* enable HiZ Raw Stall Optimization */
2586 HIZ_RAW_STALL_OPT_DISABLE);
2587 }
2588
2589 if (IS_VALLEYVIEW(i915)) {
2590 /* WaDisableEarlyCull:vlv */
2591 wa_masked_en(wal,
2592 _3D_CHICKEN3,
2593 _3D_CHICKEN_SF_DISABLE_OBJEND_CULL);
2594
2595 /*
2596 * WaVSThreadDispatchOverride:ivb,vlv
2597 *
2598 * This actually overrides the dispatch
2599 * mode for all thread types.
2600 */
2601 wa_write_clr_set(wal,
2602 GEN7_FF_THREAD_MODE,
2603 GEN7_FF_SCHED_MASK,
2604 GEN7_FF_TS_SCHED_HW |
2605 GEN7_FF_VS_SCHED_HW |
2606 GEN7_FF_DS_SCHED_HW);
2607
2608 /* WaPsdDispatchEnable:vlv */
2609 /* WaDisablePSDDualDispatchEnable:vlv */
2610 wa_masked_en(wal,
2611 GEN7_HALF_SLICE_CHICKEN1,
2612 GEN7_MAX_PS_THREAD_DEP |
2613 GEN7_PSD_SINGLE_PORT_DISPATCH_ENABLE);
2614 }
2615
2616 if (IS_IVYBRIDGE(i915)) {
2617 /* WaDisableEarlyCull:ivb */
2618 wa_masked_en(wal,
2619 _3D_CHICKEN3,
2620 _3D_CHICKEN_SF_DISABLE_OBJEND_CULL);
2621
2622 if (0) { /* causes HiZ corruption on ivb:gt1 */
2623 /* enable HiZ Raw Stall Optimization */
2624 wa_masked_dis(wal,
2625 CACHE_MODE_0_GEN7,
2626 HIZ_RAW_STALL_OPT_DISABLE);
2627 }
2628
2629 /*
2630 * WaVSThreadDispatchOverride:ivb,vlv
2631 *
2632 * This actually overrides the dispatch
2633 * mode for all thread types.
2634 */
2635 wa_write_clr_set(wal,
2636 GEN7_FF_THREAD_MODE,
2637 GEN7_FF_SCHED_MASK,
2638 GEN7_FF_TS_SCHED_HW |
2639 GEN7_FF_VS_SCHED_HW |
2640 GEN7_FF_DS_SCHED_HW);
2641
2642 /* WaDisablePSDDualDispatchEnable:ivb */
2643 if (IS_IVB_GT1(i915))
2644 wa_masked_en(wal,
2645 GEN7_HALF_SLICE_CHICKEN1,
2646 GEN7_PSD_SINGLE_PORT_DISPATCH_ENABLE);
2647 }
2648
2649 if (GRAPHICS_VER(i915) == 7) {
2650 /* WaBCSVCSTlbInvalidationMode:ivb,vlv,hsw */
2651 wa_masked_en(wal,
2652 RING_MODE_GEN7(RENDER_RING_BASE),
2653 GFX_TLB_INVALIDATE_EXPLICIT | GFX_REPLAY_MODE);
2654
2655 /* WaDisable_RenderCache_OperationalFlush:ivb,vlv,hsw */
2656 wa_masked_dis(wal, CACHE_MODE_0_GEN7, RC_OP_FLUSH_ENABLE);
2657
2658 /*
2659 * BSpec says this must be set, even though
2660 * WaDisable4x2SubspanOptimization:ivb,hsw
2661 * WaDisable4x2SubspanOptimization isn't listed for VLV.
2662 */
2663 wa_masked_en(wal,
2664 CACHE_MODE_1,
2665 PIXEL_SUBSPAN_COLLECT_OPT_DISABLE);
2666
2667 /*
2668 * BSpec recommends 8x4 when MSAA is used,
2669 * however in practice 16x4 seems fastest.
2670 *
2671 * Note that PS/WM thread counts depend on the WIZ hashing
2672 * disable bit, which we don't touch here, but it's good
2673 * to keep in mind (see 3DSTATE_PS and 3DSTATE_WM).
2674 */
2675 wa_masked_field_set(wal,
2676 GEN7_GT_MODE,
2677 GEN6_WIZ_HASHING_MASK,
2678 GEN6_WIZ_HASHING_16x4);
2679 }
2680
2681 if (IS_GRAPHICS_VER(i915, 6, 7))
2682 /*
2683 * We need to disable the AsyncFlip performance optimisations in
2684 * order to use MI_WAIT_FOR_EVENT within the CS. It should
2685 * already be programmed to '1' on all products.
2686 *
2687 * WaDisableAsyncFlipPerfMode:snb,ivb,hsw,vlv
2688 */
2689 wa_masked_en(wal,
2690 RING_MI_MODE(RENDER_RING_BASE),
2691 ASYNC_FLIP_PERF_DISABLE);
2692
2693 if (GRAPHICS_VER(i915) == 6) {
2694 /*
2695 * Required for the hardware to program scanline values for
2696 * waiting
2697 * WaEnableFlushTlbInvalidationMode:snb
2698 */
2699 wa_masked_en(wal,
2700 GFX_MODE,
2701 GFX_TLB_INVALIDATE_EXPLICIT);
2702
2703 /* WaDisableHiZPlanesWhenMSAAEnabled:snb */
2704 wa_masked_en(wal,
2705 _3D_CHICKEN,
2706 _3D_CHICKEN_HIZ_PLANE_DISABLE_MSAA_4X_SNB);
2707
2708 wa_masked_en(wal,
2709 _3D_CHICKEN3,
2710 /* WaStripsFansDisableFastClipPerformanceFix:snb */
2711 _3D_CHICKEN3_SF_DISABLE_FASTCLIP_CULL |
2712 /*
2713 * Bspec says:
2714 * "This bit must be set if 3DSTATE_CLIP clip mode is set
2715 * to normal and 3DSTATE_SF number of SF output attributes
2716 * is more than 16."
2717 */
2718 _3D_CHICKEN3_SF_DISABLE_PIPELINED_ATTR_FETCH);
2719
2720 /*
2721 * BSpec recommends 8x4 when MSAA is used,
2722 * however in practice 16x4 seems fastest.
2723 *
2724 * Note that PS/WM thread counts depend on the WIZ hashing
2725 * disable bit, which we don't touch here, but it's good
2726 * to keep in mind (see 3DSTATE_PS and 3DSTATE_WM).
2727 */
2728 wa_masked_field_set(wal,
2729 GEN6_GT_MODE,
2730 GEN6_WIZ_HASHING_MASK,
2731 GEN6_WIZ_HASHING_16x4);
2732
2733 /* WaDisable_RenderCache_OperationalFlush:snb */
2734 wa_masked_dis(wal, CACHE_MODE_0, RC_OP_FLUSH_ENABLE);
2735
2736 /*
2737 * From the Sandybridge PRM, volume 1 part 3, page 24:
2738 * "If this bit is set, STCunit will have LRA as replacement
2739 * policy. [...] This bit must be reset. LRA replacement
2740 * policy is not supported."
2741 */
2742 wa_masked_dis(wal,
2743 CACHE_MODE_0,
2744 CM0_STC_EVICT_DISABLE_LRA_SNB);
2745 }
2746
2747 if (IS_GRAPHICS_VER(i915, 4, 6))
2748 /* WaTimedSingleVertexDispatch:cl,bw,ctg,elk,ilk,snb */
2749 wa_add(wal, RING_MI_MODE(RENDER_RING_BASE),
2750 0, _MASKED_BIT_ENABLE(VS_TIMER_DISPATCH),
2751 /* XXX bit doesn't stick on Broadwater */
2752 IS_I965G(i915) ? 0 : VS_TIMER_DISPATCH, true);
2753
2754 if (GRAPHICS_VER(i915) == 4)
2755 /*
2756 * Disable CONSTANT_BUFFER before it is loaded from the context
2757 * image. For as it is loaded, it is executed and the stored
2758 * address may no longer be valid, leading to a GPU hang.
2759 *
2760 * This imposes the requirement that userspace reload their
2761 * CONSTANT_BUFFER on every batch, fortunately a requirement
2762 * they are already accustomed to from before contexts were
2763 * enabled.
2764 */
2765 wa_add(wal, ECOSKPD(RENDER_RING_BASE),
2766 0, _MASKED_BIT_ENABLE(ECO_CONSTANT_BUFFER_SR_DISABLE),
2767 0 /* XXX bit doesn't stick on Broadwater */,
2768 true);
2769 }
2770
--
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki
More information about the Intel-gfx
mailing list