Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
drm/nouveau: Add DRM_IOCTL_NOUVEAU_GET_ZCULL_INFO
Add kernel-side support for using the zcull hardware in nvidia gpus.
zcull aims to improve memory bandwidth by using an early approximate
depth test, similar to hierarchical Z on an AMD card.
Add a new ioctl that exposes zcull information that has been read
from the hardware. Userspace uses each of these parameters either
in a heuristic for determining zcull region parameters or in the
calculation of a buffer size.
It appears the hardware hasn't changed its structure for these
values since FERMI_C (circa 2011), so the assumption is that it
won't change on us too quickly, and is therefore reasonable to
include in UAPI.
This bypasses the nvif layer and instead accesses nvkm_gr directly,
which mirrors existing usage of nvkm_gr_units(). There is no nvif
object for nvkm_gr yet, and adding one is not trivial.
Signed-off-by: Mel Henning <mhenning@darkrefraction.com>
Link: https://patch.msgid.link/20260219-zcull3-v3-2-dbe6a716f104@darkrefraction.com
Signed-off-by: Danilo Krummrich <dakr@kernel.org>
authored by
Mel Henning
and committed by
Danilo Krummrich
196b2b95
27e125e5
+97
+29
drivers/gpu/drm/nouveau/nouveau_abi16.c
+29
drivers/gpu/drm/nouveau/nouveau_abi16.c
···
334
334
}
335
335
336
336
int
337
+
nouveau_abi16_ioctl_get_zcull_info(ABI16_IOCTL_ARGS)
338
+
{
339
+
struct nouveau_drm *drm = nouveau_drm(dev);
340
+
struct nvkm_gr *gr = nvxx_gr(drm);
341
+
struct drm_nouveau_get_zcull_info *out = data;
342
+
343
+
if (gr->has_zcull_info) {
344
+
const struct nvkm_gr_zcull_info *i = &gr->zcull_info;
345
+
346
+
out->width_align_pixels = i->width_align_pixels;
347
+
out->height_align_pixels = i->height_align_pixels;
348
+
out->pixel_squares_by_aliquots = i->pixel_squares_by_aliquots;
349
+
out->aliquot_total = i->aliquot_total;
350
+
out->zcull_region_byte_multiplier = i->zcull_region_byte_multiplier;
351
+
out->zcull_region_header_size = i->zcull_region_header_size;
352
+
out->zcull_subregion_header_size = i->zcull_subregion_header_size;
353
+
out->subregion_count = i->subregion_count;
354
+
out->subregion_width_align_pixels = i->subregion_width_align_pixels;
355
+
out->subregion_height_align_pixels = i->subregion_height_align_pixels;
356
+
out->ctxsw_size = i->ctxsw_size;
357
+
out->ctxsw_align = i->ctxsw_align;
358
+
359
+
return 0;
360
+
} else {
361
+
return -ENOTTY;
362
+
}
363
+
}
364
+
365
+
int
337
366
nouveau_abi16_ioctl_channel_alloc(ABI16_IOCTL_ARGS)
338
367
{
339
368
struct drm_nouveau_channel_alloc *init = data;
+1
drivers/gpu/drm/nouveau/nouveau_abi16.h
+1
drivers/gpu/drm/nouveau/nouveau_abi16.h
···
6
6
struct drm_device *dev, void *data, struct drm_file *file_priv
7
7
8
8
int nouveau_abi16_ioctl_getparam(ABI16_IOCTL_ARGS);
9
+
int nouveau_abi16_ioctl_get_zcull_info(ABI16_IOCTL_ARGS);
9
10
int nouveau_abi16_ioctl_channel_alloc(ABI16_IOCTL_ARGS);
10
11
int nouveau_abi16_ioctl_channel_free(ABI16_IOCTL_ARGS);
11
12
int nouveau_abi16_ioctl_grobj_alloc(ABI16_IOCTL_ARGS);
+1
drivers/gpu/drm/nouveau/nouveau_drm.c
+1
drivers/gpu/drm/nouveau/nouveau_drm.c
···
1304
1304
DRM_IOCTL_DEF_DRV(NOUVEAU_GROBJ_ALLOC, nouveau_abi16_ioctl_grobj_alloc, DRM_RENDER_ALLOW),
1305
1305
DRM_IOCTL_DEF_DRV(NOUVEAU_NOTIFIEROBJ_ALLOC, nouveau_abi16_ioctl_notifierobj_alloc, DRM_RENDER_ALLOW),
1306
1306
DRM_IOCTL_DEF_DRV(NOUVEAU_GPUOBJ_FREE, nouveau_abi16_ioctl_gpuobj_free, DRM_RENDER_ALLOW),
1307
+
DRM_IOCTL_DEF_DRV(NOUVEAU_GET_ZCULL_INFO, nouveau_abi16_ioctl_get_zcull_info, DRM_RENDER_ALLOW),
1307
1308
DRM_IOCTL_DEF_DRV(NOUVEAU_SVM_INIT, nouveau_svmm_init, DRM_RENDER_ALLOW),
1308
1309
DRM_IOCTL_DEF_DRV(NOUVEAU_SVM_BIND, nouveau_svmm_bind, DRM_RENDER_ALLOW),
1309
1310
DRM_IOCTL_DEF_DRV(NOUVEAU_GEM_NEW, nouveau_gem_ioctl_new, DRM_RENDER_ALLOW),
+66
include/uapi/drm/nouveau_drm.h
+66
include/uapi/drm/nouveau_drm.h
···
432
432
__u64 push_ptr;
433
433
};
434
434
435
+
struct drm_nouveau_get_zcull_info {
436
+
/**
437
+
* @width_align_pixels: required alignment for region widths, in pixels
438
+
* (typically #TPC's * 16).
439
+
*/
440
+
__u32 width_align_pixels;
441
+
/**
442
+
* @height_align_pixels: required alignment for region heights, in
443
+
* pixels (typically 32).
444
+
*/
445
+
__u32 height_align_pixels;
446
+
/**
447
+
* @pixel_squares_by_aliquots: the pixel area covered by an aliquot
448
+
* (typically #Zcull_banks * 16 * 16).
449
+
*/
450
+
__u32 pixel_squares_by_aliquots;
451
+
/**
452
+
* @aliquot_total: the total aliquot pool available in hardware
453
+
*/
454
+
__u32 aliquot_total;
455
+
/**
456
+
* @zcull_region_byte_multiplier: the size of an aliquot in bytes, which
457
+
* is used for save/restore operations on a region
458
+
*/
459
+
__u32 zcull_region_byte_multiplier;
460
+
/**
461
+
* @zcull_region_header_size: the region header size in bytes, which is
462
+
* used for save/restore operations on a region
463
+
*/
464
+
__u32 zcull_region_header_size;
465
+
/**
466
+
* @zcull_subregion_header_size: the subregion header size in bytes,
467
+
* which is used for save/restore operations on a region
468
+
*/
469
+
__u32 zcull_subregion_header_size;
470
+
/**
471
+
* @subregion_count: the total number of subregions the hardware
472
+
* supports
473
+
*/
474
+
__u32 subregion_count;
475
+
/**
476
+
* @subregion_width_align_pixels: required alignment for subregion
477
+
* widths, in pixels (typically #TPC's * 16).
478
+
*/
479
+
__u32 subregion_width_align_pixels;
480
+
/**
481
+
* @subregion_height_align_pixels: required alignment for subregion
482
+
* heights, in pixels
483
+
*/
484
+
__u32 subregion_height_align_pixels;
485
+
486
+
/**
487
+
* @ctxsw_size: the size, in bytes, of a zcull context switching region.
488
+
* Will be zero if the kernel does not support zcull context switching.
489
+
*/
490
+
__u32 ctxsw_size;
491
+
/**
492
+
* @ctxsw_align: the alignment, in bytes, of a zcull context switching
493
+
* region
494
+
*/
495
+
__u32 ctxsw_align;
496
+
};
497
+
435
498
#define DRM_NOUVEAU_GETPARAM 0x00
436
499
#define DRM_NOUVEAU_SETPARAM 0x01 /* deprecated */
437
500
#define DRM_NOUVEAU_CHANNEL_ALLOC 0x02
···
508
445
#define DRM_NOUVEAU_VM_INIT 0x10
509
446
#define DRM_NOUVEAU_VM_BIND 0x11
510
447
#define DRM_NOUVEAU_EXEC 0x12
448
+
#define DRM_NOUVEAU_GET_ZCULL_INFO 0x13
511
449
#define DRM_NOUVEAU_GEM_NEW 0x40
512
450
#define DRM_NOUVEAU_GEM_PUSHBUF 0x41
513
451
#define DRM_NOUVEAU_GEM_CPU_PREP 0x42
···
577
513
#define DRM_IOCTL_NOUVEAU_VM_INIT DRM_IOWR(DRM_COMMAND_BASE + DRM_NOUVEAU_VM_INIT, struct drm_nouveau_vm_init)
578
514
#define DRM_IOCTL_NOUVEAU_VM_BIND DRM_IOWR(DRM_COMMAND_BASE + DRM_NOUVEAU_VM_BIND, struct drm_nouveau_vm_bind)
579
515
#define DRM_IOCTL_NOUVEAU_EXEC DRM_IOWR(DRM_COMMAND_BASE + DRM_NOUVEAU_EXEC, struct drm_nouveau_exec)
516
+
517
+
#define DRM_IOCTL_NOUVEAU_GET_ZCULL_INFO DRM_IOR (DRM_COMMAND_BASE + DRM_NOUVEAU_GET_ZCULL_INFO, struct drm_nouveau_get_zcull_info)
580
518
#if defined(__cplusplus)
581
519
}
582
520
#endif