[igt-dev] [PATCH i-g-t v3 2/8] lib/xe: Introduce Xe library
Zbigniew Kempczyński
zbigniew.kempczynski at intel.com
Tue Feb 28 10:04:55 UTC 2023
On Fri, Feb 24, 2023 at 07:46:35PM +0100, Michal Wajdeczko wrote:
>
>
> On 24.02.2023 11:44, Zbigniew Kempczyński wrote:
> > Xe, is a new driver for Intel GPUs that supports both integrated
> > and discrete platforms starting with Tiger Lake (first Intel Xe
> > Architecture).
> >
> > Series was split to allow easier review. Library, drm uapi, tests,
> > tools and other were squashed according to code subject.
> >
> > This patch introduces library used for Xe tests. As there's not
> > too trivial to calculate credits for squashed subjects full series
> > credits are:
> >
> > Co-developed-by: Matthew Brost
> > [commits: 90 / lines changed: 12574]
> > Co-developed-by: Mauro Carvalho Chehab
> > [commits: 28 / lines changed: 1873]
> > Co-developed-by: Rodrigo Vivi
> > [commits: 15 / lines changed: 1317]
> > Co-developed-by: Jason Ekstrand
> > [commits: 14 / lines changed: 1418]
> > Co-developed-by: Francois Dugast
> > [commits: 8 / lines changed: 1082]
> > Co-developed-by: Philippe Lecluse
> > [commits: 6 / lines changed: 560]
> > Co-developed-by: Zbigniew Kempczyński
> > [commits: 4 / lines changed: 1091]
> > Co-developed-by: Matthew Auld
> > [commits: 3 / lines changed: 35]
> > Co-developed-by: Niranjana Vishwanathapura
> > [commits: 2 / lines changed: 66]
> > Co-developed-by: Maarten Lankhorst
> > [commits: 2 / lines changed: 774]
> > Co-developed-by: Ryszard Knop
> > [commits: 1 / lines changed: 12]
> > Co-developed-by: Thomas Hellström
> > [commits: 1 / lines changed: 12]
> > Signed-off-by: Rodrigo Vivi <rodrigo.vivi at intel.com>
> > Signed-off-by: Zbigniew Kempczyński <zbigniew.kempczynski at intel.com>
> > ---
> > lib/drmtest.c | 3 +
> > lib/drmtest.h | 1 +
> > lib/meson.build | 4 +
> > lib/xe/xe_compute.c | 402 ++++++++++++++++++++++++++++++++++++++
> > lib/xe/xe_compute.h | 29 +++
> > lib/xe/xe_ioctl.c | 434 +++++++++++++++++++++++++++++++++++++++++
> > lib/xe/xe_ioctl.h | 89 +++++++++
> > lib/xe/xe_query.c | 465 ++++++++++++++++++++++++++++++++++++++++++++
> > lib/xe/xe_query.h | 64 ++++++
> > lib/xe/xe_spin.c | 139 +++++++++++++
> > lib/xe/xe_spin.h | 48 +++++
> > meson.build | 8 +
> > 12 files changed, 1686 insertions(+)
> > create mode 100644 lib/xe/xe_compute.c
> > create mode 100644 lib/xe/xe_compute.h
> > create mode 100644 lib/xe/xe_ioctl.c
> > create mode 100644 lib/xe/xe_ioctl.h
> > create mode 100644 lib/xe/xe_query.c
> > create mode 100644 lib/xe/xe_query.h
> > create mode 100644 lib/xe/xe_spin.c
> > create mode 100644 lib/xe/xe_spin.h
> >
> > diff --git a/lib/drmtest.c b/lib/drmtest.c
> > index 16e80bdf..859f66ff 100644
> > --- a/lib/drmtest.c
> > +++ b/lib/drmtest.c
> > @@ -189,6 +189,7 @@ static const struct module {
> > { DRIVER_V3D, "v3d" },
> > { DRIVER_VC4, "vc4" },
> > { DRIVER_VGEM, "vgem" },
> > + { DRIVER_XE, "xe" },
> > {}
> > };
> >
> > @@ -547,6 +548,8 @@ static const char *chipset_to_str(int chipset)
> > return "panfrost";
> > case DRIVER_MSM:
> > return "msm";
> > + case DRIVER_XE:
> > + return "xe";
> > case DRIVER_ANY:
> > return "any";
> > default:
> > diff --git a/lib/drmtest.h b/lib/drmtest.h
> > index b5debd44..448ac03b 100644
> > --- a/lib/drmtest.h
> > +++ b/lib/drmtest.h
> > @@ -51,6 +51,7 @@
> > #define DRIVER_V3D (1 << 4)
> > #define DRIVER_PANFROST (1 << 5)
> > #define DRIVER_MSM (1 << 6)
> > +#define DRIVER_XE (1 << 7)
> >
> > /*
> > * Exclude DRVER_VGEM from DRIVER_ANY since if you run on a system
> > diff --git a/lib/meson.build b/lib/meson.build
> > index c5131d9a..768ce90b 100644
> > --- a/lib/meson.build
> > +++ b/lib/meson.build
> > @@ -98,6 +98,10 @@ lib_sources = [
> > 'veboxcopy_gen12.c',
> > 'igt_msm.c',
> > 'igt_dsc.c',
> > + 'xe/xe_compute.c',
> > + 'xe/xe_ioctl.c',
> > + 'xe/xe_query.c',
> > + 'xe/xe_spin.c'
> > ]
> >
> > lib_deps = [
> > diff --git a/lib/xe/xe_compute.c b/lib/xe/xe_compute.c
> > new file mode 100644
> > index 00000000..477c39bb
> > --- /dev/null
> > +++ b/lib/xe/xe_compute.c
> > @@ -0,0 +1,402 @@
> > +/* SPDX-License-Identifier: MIT */
> > +/*
> > + * Copyright © 2023 Intel Corporation
> > + *
> > + * Authors:
> > + * Francois Dugast <francois.dugast at intel.com>
> > + */
> > +
> > +#include "xe_compute.h"
> > +
> > +#define PIPE_CONTROL 0x7a000004
> > +#define MI_LOAD_REGISTER_IMM 0x11000001
> > +#define PIPELINE_SELECT 0x69040302
> > +#define MEDIA_VFE_STATE 0x70000007
> > +#define STATE_BASE_ADDRESS 0x61010014
> > +#define MEDIA_STATE_FLUSH 0x0
> > +#define MEDIA_INTERFACE_DESCRIPTOR_LOAD 0x70020002
> > +#define GPGPU_WALKER 0x7105000d
> > +#define MI_BATCH_BUFFER_END (0xA << 23)
>
> what criteria was used to select only above cmds to have their friendly
> mnemonics ?
>
+Francois
Could you provide some comment + kernel.cl?
> > +
> > +// generated with: ocloc -file kernel.cl -device tgllp && xxd -i kernel_Gen12LPlp.gen
>
> is this kernel.cl available somewhere ?
>
Agree, I would also like to include it here.
> > +unsigned char tgllp_kernel_square_bin[] = {
>
> static const ?
It cannot be static, it is copied to kernel bo in test (tests/xe/xe_compute.c)
>
> > + 0x61, 0x00, 0x03, 0x80, 0x20, 0x02, 0x05, 0x03, 0x04, 0x00, 0x10, 0x00,
> > + 0x00, 0x00, 0x00, 0x00, 0x66, 0x01, 0x00, 0x80, 0x20, 0x82, 0x01, 0x80,
> > + 0x00, 0x80, 0x00, 0x01, 0xc0, 0x04, 0xc0, 0x04, 0x41, 0x01, 0x20, 0x22,
> > + 0x16, 0x09, 0x11, 0x03, 0x49, 0x00, 0x04, 0xa2, 0x12, 0x09, 0x11, 0x03,
> > + 0x40, 0x01, 0x04, 0x00, 0x60, 0x06, 0x05, 0x05, 0x04, 0x04, 0x00, 0x01,
> > + 0x05, 0x01, 0x58, 0x00, 0x40, 0x00, 0x24, 0x00, 0x60, 0x06, 0x05, 0x0a,
> > + 0x04, 0x04, 0x00, 0x01, 0x05, 0x02, 0x58, 0x00, 0x40, 0x02, 0x0c, 0xa0,
> > + 0x02, 0x05, 0x10, 0x07, 0x40, 0x02, 0x0e, 0xa6, 0x02, 0x0a, 0x10, 0x07,
> > + 0x70, 0x02, 0x04, 0x00, 0x60, 0x02, 0x01, 0x00, 0x05, 0x0c, 0x46, 0x52,
> > + 0x84, 0x08, 0x00, 0x00, 0x70, 0x02, 0x24, 0x00, 0x60, 0x02, 0x01, 0x00,
> > + 0x05, 0x0e, 0x46, 0x52, 0x84, 0x08, 0x00, 0x00, 0x72, 0x00, 0x02, 0x80,
> > + 0x50, 0x0d, 0x04, 0x00, 0x05, 0x00, 0x05, 0x1d, 0x05, 0x00, 0x05, 0x00,
> > + 0x22, 0x00, 0x05, 0x01, 0x00, 0xc0, 0x00, 0x00, 0x90, 0x00, 0x00, 0x00,
> > + 0x90, 0x00, 0x00, 0x00, 0x69, 0x00, 0x10, 0x60, 0x02, 0x0c, 0x20, 0x00,
> > + 0x69, 0x00, 0x12, 0x66, 0x02, 0x0e, 0x20, 0x00, 0x40, 0x02, 0x14, 0xa0,
> > + 0x32, 0x10, 0x10, 0x08, 0x40, 0x02, 0x16, 0xa6, 0x32, 0x12, 0x10, 0x08,
> > + 0x31, 0xa0, 0x04, 0x00, 0x00, 0x00, 0x14, 0x18, 0x14, 0x14, 0x00, 0xcc,
> > + 0x00, 0x00, 0x16, 0x00, 0x31, 0x91, 0x24, 0x00, 0x00, 0x00, 0x14, 0x1a,
> > + 0x14, 0x16, 0x00, 0xcc, 0x00, 0x00, 0x16, 0x00, 0x40, 0x00, 0x10, 0xa0,
> > + 0x4a, 0x10, 0x10, 0x08, 0x40, 0x00, 0x12, 0xa6, 0x4a, 0x12, 0x10, 0x08,
> > + 0x41, 0x20, 0x18, 0x20, 0x00, 0x18, 0x00, 0x18, 0x41, 0x21, 0x1a, 0x26,
> > + 0x00, 0x1a, 0x00, 0x1a, 0x31, 0xa2, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00,
> > + 0x14, 0x10, 0x02, 0xcc, 0x14, 0x18, 0x96, 0x00, 0x31, 0x93, 0x24, 0x00,
> > + 0x00, 0x00, 0x00, 0x00, 0x14, 0x12, 0x02, 0xcc, 0x14, 0x1a, 0x96, 0x00,
> > + 0x25, 0x00, 0x05, 0x00, 0x00, 0x40, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
> > + 0x10, 0x00, 0x00, 0x00, 0x61, 0x00, 0x7f, 0x64, 0x00, 0x03, 0x10, 0x00,
> > + 0x31, 0x44, 0x03, 0x80, 0x00, 0x00, 0x0c, 0x1c, 0x0c, 0x03, 0x00, 0xa0,
> > + 0x00, 0x00, 0x78, 0x02, 0x61, 0x24, 0x03, 0x80, 0x20, 0x02, 0x01, 0x00,
> > + 0x05, 0x1c, 0x46, 0x00, 0x00, 0x00, 0x00, 0x00, 0x61, 0x00, 0x04, 0x80,
> > + 0xa0, 0x4a, 0x01, 0x20, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
> > + 0x31, 0x01, 0x03, 0x80, 0x04, 0x00, 0x00, 0x00, 0x0c, 0x7f, 0x20, 0x70,
> > + 0x00, 0x00, 0x00, 0x00, 0x60, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
> > + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
> > + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
> > + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
> > + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
> > + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
> > + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
> > + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
> > + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
> > + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
> > + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
> > + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
> > + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
> > + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
> > + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
> > + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
> > + 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00
> > +};
> > +unsigned int tgllp_kernel_square_length = sizeof(tgllp_kernel_square_bin);
>
> static const ?
Same as above.
>
> > +
> > +/**
> > + * tgllp_create_indirect_data:
> > + * @addr_bo_buffer_batch: pointer to batch buffer
> > + * @addr_input: input buffer gpu offset
> > + * @addr_output: output buffer gpu offset
> > + *
> > + * Prepares indirect data for compute pipeline.
> > + */
> > +void tgllp_create_indirect_data(uint32_t *addr_bo_buffer_batch,
> > + uint64_t addr_input, uint64_t addr_output)
> > +{
> > + int b = 0;
>
> nit: pls add separation line for clarity
>
Agree, I will send in v4 (for other functions too).
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000200;
> > + addr_bo_buffer_batch[b++] = 0x00000001;
> > + addr_bo_buffer_batch[b++] = 0x00000001;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = addr_input & 0xffffffff;
> > + addr_bo_buffer_batch[b++] = addr_input >> 32;
> > + addr_bo_buffer_batch[b++] = addr_output & 0xffffffff;
> > + addr_bo_buffer_batch[b++] = addr_output >> 32;
> > + addr_bo_buffer_batch[b++] = 0x00000400;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000200;
> > + addr_bo_buffer_batch[b++] = 0x00000001;
> > + addr_bo_buffer_batch[b++] = 0x00000001;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00010000;
> > + addr_bo_buffer_batch[b++] = 0x00030002;
> > + addr_bo_buffer_batch[b++] = 0x00050004;
> > + addr_bo_buffer_batch[b++] = 0x00070006;
> > + addr_bo_buffer_batch[b++] = 0x00090008;
> > + addr_bo_buffer_batch[b++] = 0x000B000A;
> > + addr_bo_buffer_batch[b++] = 0x000D000C;
> > + addr_bo_buffer_batch[b++] = 0x000F000E;
> > + addr_bo_buffer_batch[b++] = 0x00110010;
> > + addr_bo_buffer_batch[b++] = 0x00130012;
> > + addr_bo_buffer_batch[b++] = 0x00150014;
> > + addr_bo_buffer_batch[b++] = 0x00170016;
> > + addr_bo_buffer_batch[b++] = 0x00190018;
> > + addr_bo_buffer_batch[b++] = 0x001B001A;
> > + addr_bo_buffer_batch[b++] = 0x001D001C;
> > + addr_bo_buffer_batch[b++] = 0x001F001E;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00210020;
> > + addr_bo_buffer_batch[b++] = 0x00230022;
> > + addr_bo_buffer_batch[b++] = 0x00250024;
> > + addr_bo_buffer_batch[b++] = 0x00270026;
> > + addr_bo_buffer_batch[b++] = 0x00290028;
> > + addr_bo_buffer_batch[b++] = 0x002B002A;
> > + addr_bo_buffer_batch[b++] = 0x002D002C;
> > + addr_bo_buffer_batch[b++] = 0x002F002E;
> > + addr_bo_buffer_batch[b++] = 0x00310030;
> > + addr_bo_buffer_batch[b++] = 0x00330032;
> > + addr_bo_buffer_batch[b++] = 0x00350034;
> > + addr_bo_buffer_batch[b++] = 0x00370036;
> > + addr_bo_buffer_batch[b++] = 0x00390038;
> > + addr_bo_buffer_batch[b++] = 0x003B003A;
> > + addr_bo_buffer_batch[b++] = 0x003D003C;
> > + addr_bo_buffer_batch[b++] = 0x003F003E;
> > +}
> > +
> > +/**
> > + * tgllp_create_surface_state:
> > + * @addr_bo_buffer_batch: pointer to batch buffer
> > + * @addr_input: input buffer gpu offset
> > + * @addr_output: output buffer gpu offset
> > + *
> > + * Prepares surface state for compute pipeline.
> > + */
> > +void tgllp_create_surface_state(uint32_t *addr_bo_buffer_batch,
> > + uint64_t addr_input, uint64_t addr_output)
> > +{
> > + int b = 0;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x87FD4000;
> > + addr_bo_buffer_batch[b++] = 0x04000000;
> > + addr_bo_buffer_batch[b++] = 0x001F007F;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00004000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = addr_input & 0xffffffff;
> > + addr_bo_buffer_batch[b++] = addr_input >> 32;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x87FD4000;
> > + addr_bo_buffer_batch[b++] = 0x04000000;
> > + addr_bo_buffer_batch[b++] = 0x001F007F;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00004000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = addr_output & 0xffffffff;
> > + addr_bo_buffer_batch[b++] = addr_output >> 32;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000040;
> > + addr_bo_buffer_batch[b++] = 0x00000080;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > +}
> > +
> > +/**
> > + * tgllp_create_dynamic_state:
> > + * @addr_bo_buffer_batch: pointer to batch buffer
> > + * @offset_kernel: gpu offset of the shader
> > + *
> > + * Prepares dynamic state for compute pipeline.
> > + */
> > +void tgllp_create_dynamic_state(uint32_t *addr_bo_buffer_batch,
> > + uint64_t offset_kernel)
> > +{
> > + int b = 0;
> > + addr_bo_buffer_batch[b++] = offset_kernel;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00180000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x000000C0;
> > + addr_bo_buffer_batch[b++] = 0x00060000;
> > + addr_bo_buffer_batch[b++] = 0x00000010;
> > + addr_bo_buffer_batch[b++] = 0x00000003;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > +}
> > +
> > +/**
> > + * tgllp_create_batch_compute:
> > + * @addr_bo_buffer_batch: pointer to batch buffer
> > + * @addr_surface_state_base: gpu offset of surface state data
> > + * @addr_dynamic_state_base: gpu offset of dynamic state data
> > + * @addr_indirect_object_base: gpu offset of indirect object data
> > + * @offset_indirect_data_start: gpu offset of indirect data start
> > + *
> > + * Prepares compute pipeline.
> > + */
> > +void tgllp_create_batch_compute(uint32_t *addr_bo_buffer_batch,
> > + uint64_t addr_surface_state_base,
> > + uint64_t addr_dynamic_state_base,
> > + uint64_t addr_indirect_object_base,
> > + uint64_t offset_indirect_data_start)
> > +{
> > + int b = 0;
> > + addr_bo_buffer_batch[b++] = MI_LOAD_REGISTER_IMM;
> > + addr_bo_buffer_batch[b++] = 0x00002580;
> > + addr_bo_buffer_batch[b++] = 0x00060002;
> > + addr_bo_buffer_batch[b++] = PIPELINE_SELECT;
> > + addr_bo_buffer_batch[b++] = MI_LOAD_REGISTER_IMM;
> > + addr_bo_buffer_batch[b++] = 0x00007034;
> > + addr_bo_buffer_batch[b++] = 0x60000321;
> > + addr_bo_buffer_batch[b++] = PIPE_CONTROL;
> > + addr_bo_buffer_batch[b++] = 0x00100000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = MI_LOAD_REGISTER_IMM;
> > + addr_bo_buffer_batch[b++] = 0x0000E404;
> > + addr_bo_buffer_batch[b++] = 0x00000100;
> > + addr_bo_buffer_batch[b++] = PIPE_CONTROL;
> > + addr_bo_buffer_batch[b++] = 0x00101021;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = MEDIA_VFE_STATE;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00A70100;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x07820000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = PIPE_CONTROL;
> > + addr_bo_buffer_batch[b++] = 0x00100420;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = STATE_BASE_ADDRESS;
> > + addr_bo_buffer_batch[b++] = 0x00000001;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00040000;
> > + addr_bo_buffer_batch[b++] = (addr_surface_state_base & 0xffffffff) | 0x1;
> > + addr_bo_buffer_batch[b++] = addr_surface_state_base >> 32;
> > + addr_bo_buffer_batch[b++] = (addr_dynamic_state_base & 0xffffffff) | 0x1;
> > + addr_bo_buffer_batch[b++] = addr_dynamic_state_base >> 32;
> > + addr_bo_buffer_batch[b++] = (addr_indirect_object_base & 0xffffffff) | 0x1;
> > + addr_bo_buffer_batch[b++] = (addr_indirect_object_base >> 32) | 0xffff0000;
> > + addr_bo_buffer_batch[b++] = (addr_indirect_object_base & 0xffffffff) | 0x41;
> > + addr_bo_buffer_batch[b++] = addr_indirect_object_base >> 32;
> > + addr_bo_buffer_batch[b++] = 0xFFFFF001;
> > + addr_bo_buffer_batch[b++] = 0x00010001;
> > + addr_bo_buffer_batch[b++] = 0xFFFFF001;
> > + addr_bo_buffer_batch[b++] = 0xFFFFF001;
> > + addr_bo_buffer_batch[b++] = (addr_surface_state_base & 0xffffffff) | 0x1;
> > + addr_bo_buffer_batch[b++] = addr_surface_state_base >> 32;
> > + addr_bo_buffer_batch[b++] = 0x003BF000;
> > + addr_bo_buffer_batch[b++] = 0x00000041;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = PIPE_CONTROL;
> > + addr_bo_buffer_batch[b++] = 0x00100000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = PIPE_CONTROL;
> > + addr_bo_buffer_batch[b++] = 0x00100000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = MEDIA_STATE_FLUSH;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = MEDIA_INTERFACE_DESCRIPTOR_LOAD;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000020;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = GPGPU_WALKER;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000c80;
> > + addr_bo_buffer_batch[b++] = offset_indirect_data_start;
> > + addr_bo_buffer_batch[b++] = 0x8000000f;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000002;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000001;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000001;
> > + addr_bo_buffer_batch[b++] = 0xffffffff;
> > + addr_bo_buffer_batch[b++] = 0xffffffff;
> > + addr_bo_buffer_batch[b++] = MEDIA_STATE_FLUSH;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = PIPE_CONTROL;
> > + addr_bo_buffer_batch[b++] = 0x00100000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = PIPE_CONTROL;
> > + addr_bo_buffer_batch[b++] = 0x00100120;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = 0x00000000;
> > + addr_bo_buffer_batch[b++] = MI_BATCH_BUFFER_END;
> > +}
> > diff --git a/lib/xe/xe_compute.h b/lib/xe/xe_compute.h
> > new file mode 100644
> > index 00000000..de763101
> > --- /dev/null
> > +++ b/lib/xe/xe_compute.h
> > @@ -0,0 +1,29 @@
> > +/* SPDX-License-Identifier: MIT */
> > +/*
> > + * Copyright © 2023 Intel Corporation
> > + *
> > + * Authors:
> > + * Francois Dugast <francois.dugast at intel.com>
> > + */
> > +
> > +#ifndef XE_COMPUTE_H
> > +#define XE_COMPUTE_H
> > +
> > +#include <stdint.h>
> > +
> > +void tgllp_create_indirect_data(uint32_t *addr_bo_buffer_batch,
> > + uint64_t addr_input, uint64_t addr_output);
> > +void tgllp_create_surface_state(uint32_t *addr_bo_buffer_batch,
> > + uint64_t addr_input, uint64_t addr_output);
> > +void tgllp_create_dynamic_state(uint32_t *addr_bo_buffer_batch,
> > + uint64_t offset_kernel);
> > +void tgllp_create_batch_compute(uint32_t *addr_bo_buffer_batch,
> > + uint64_t addr_surface_state_base,
> > + uint64_t addr_dynamic_state_base,
> > + uint64_t addr_indirect_object_base,
> > + uint64_t offset_indirect_data_start);
> > +
> > +extern unsigned char tgllp_kernel_square_bin[];
> > +extern unsigned int tgllp_kernel_square_length;
>
> do we need to expose this as raw data ? maybe as function:
>
> const char* tgllp_kernel_square(uint32_t *out_length);
>
May we do this refactor later? I think we will add compute pipelines
for other platforms so this would be good time to clean this.
> > +
> > +#endif /* XE_COMPUTE_H */
> > diff --git a/lib/xe/xe_ioctl.c b/lib/xe/xe_ioctl.c
> > new file mode 100644
> > index 00000000..d34af2dd
> > --- /dev/null
> > +++ b/lib/xe/xe_ioctl.c
> > @@ -0,0 +1,434 @@
> > +// SPDX-License-Identifier: MIT
> > +/*
> > + * Copyright © 2023 Intel Corporation
> > + *
> > + * Authors:
> > + * Jason Ekstrand <jason at jlekstrand.net>
> > + * Maarten Lankhorst <maarten.lankhorst at linux.intel.com>
> > + * Matthew Brost <matthew.brost at intel.com>
> > + */
> > +
> > +#ifdef HAVE_LIBGEN_H
> > +#include <libgen.h>
> > +#endif
> > +
> > +#include <errno.h>
> > +#include <fcntl.h>
> > +#include <getopt.h>
> > +#include <pciaccess.h>
> > +#include <signal.h>
> > +#include <stdio.h>
> > +#include <stdlib.h>
> > +#include <string.h>
> > +#include <termios.h>
> > +#include <unistd.h>
> > +#include <sys/ioctl.h>
> > +#include <sys/mman.h>
> > +#include <sys/stat.h>
> > +#include <sys/syscall.h>
> > +#include <sys/types.h>
> > +#include <sys/utsname.h>
> > +#include <sys/wait.h>
> > +
> > +#include "drmtest.h"
> > +#include "config.h"
>
> nit: wrong include order
>
You mean sorting local headers?
> > +
> > +#ifdef HAVE_VALGRIND
> > +#include <valgrind/valgrind.h>
> > +#include <valgrind/memcheck.h>
> > +
> > +#define VG(x) x
> > +#else
> > +#define VG(x) do {} while (0)
> > +#endif
>
> VG seems unused, drop it ?
>
Agree, looks like copy-paste code from i915.
> > +
> > +#include "xe_ioctl.h"
> > +#include "xe_query.h"
> > +#include "igt_syncobj.h"
> > +#include "ioctl_wrappers.h"
> > +
> > +uint32_t xe_cs_prefetch_size(int fd)
> > +{
> > + return 512;
> > +}
> > +
> > +uint32_t xe_vm_create(int fd, uint32_t flags, uint64_t ext)
> > +{
> > + struct drm_xe_vm_create create = {
> > + .extensions = ext,
> > + .flags = flags,
> > + };
> > +
> > + igt_assert_eq(igt_ioctl(fd, DRM_IOCTL_XE_VM_CREATE, &create), 0);
> > +
> > + return create.vm_id;
> > +}
> > +
> > +void xe_vm_unbind_all_async(int fd, uint32_t vm, uint32_t engine,
> > + uint32_t bo, struct drm_xe_sync *sync,
> > + uint32_t num_syncs)
> > +{
> > + __xe_vm_bind_assert(fd, vm, engine, bo, 0, 0, 0,
> > + XE_VM_BIND_OP_UNMAP_ALL | XE_VM_BIND_FLAG_ASYNC,
> > + sync, num_syncs, 0, 0);
> > +}
> > +
> > +void xe_vm_bind_array(int fd, uint32_t vm, uint32_t engine,
> > + struct drm_xe_vm_bind_op *bind_ops,
> > + uint32_t num_bind, struct drm_xe_sync *sync,
> > + uint32_t num_syncs)
> > +{
> > + struct drm_xe_vm_bind bind = {
> > + .vm_id = vm,
> > + .num_binds = num_bind,
> > + .vector_of_binds = (uintptr_t)bind_ops,
> > + .num_syncs = num_syncs,
> > + .syncs = (uintptr_t)sync,
> > + .engine_id = engine,
> > + };
> > +
> > + igt_assert(num_bind > 1);
> > + igt_assert_eq(igt_ioctl(fd, DRM_IOCTL_XE_VM_BIND, &bind), 0);
> > +}
> > +
> > +int __xe_vm_bind(int fd, uint32_t vm, uint32_t engine, uint32_t bo,
> > + uint64_t offset, uint64_t addr, uint64_t size, uint32_t op,
> > + struct drm_xe_sync *sync, uint32_t num_syncs, uint32_t region,
> > + uint64_t ext)
> > +{
> > + struct drm_xe_vm_bind bind = {
> > + .extensions = ext,
> > + .vm_id = vm,
> > + .num_binds = 1,
> > + .bind.obj = bo,
> > + .bind.obj_offset = offset,
> > + .bind.range = size,
> > + .bind.addr = addr,
> > + .bind.op = op,
> > + .bind.region = region,
> > + .num_syncs = num_syncs,
> > + .syncs = (uintptr_t)sync,
> > + .engine_id = engine,
> > + };
> > +
> > + if (igt_ioctl(fd, DRM_IOCTL_XE_VM_BIND, &bind))
> > + return -errno;
> > +
> > + return 0;
> > +}
> > +
> > +void __xe_vm_bind_assert(int fd, uint32_t vm, uint32_t engine, uint32_t bo,
> > + uint64_t offset, uint64_t addr, uint64_t size,
> > + uint32_t op, struct drm_xe_sync *sync,
> > + uint32_t num_syncs, uint32_t region, uint64_t ext)
> > +{
> > + igt_assert_eq(__xe_vm_bind(fd, vm, engine, bo, offset, addr, size,
> > + op, sync, num_syncs, region, ext), 0);
> > +}
> > +
> > +void xe_vm_bind(int fd, uint32_t vm, uint32_t bo, uint64_t offset,
> > + uint64_t addr, uint64_t size,
> > + struct drm_xe_sync *sync, uint32_t num_syncs)
> > +{
> > + __xe_vm_bind_assert(fd, vm, 0, bo, offset, addr, size,
> > + XE_VM_BIND_OP_MAP, sync, num_syncs, 0, 0);
> > +}
> > +
> > +void xe_vm_unbind(int fd, uint32_t vm, uint64_t offset,
> > + uint64_t addr, uint64_t size,
> > + struct drm_xe_sync *sync, uint32_t num_syncs)
> > +{
> > + __xe_vm_bind_assert(fd, vm, 0, 0, offset, addr, size,
> > + XE_VM_BIND_OP_UNMAP, sync, num_syncs, 0, 0);
> > +}
> > +
> > +void xe_vm_prefetch_async(int fd, uint32_t vm, uint32_t engine, uint64_t offset,
> > + uint64_t addr, uint64_t size,
> > + struct drm_xe_sync *sync, uint32_t num_syncs,
> > + uint32_t region)
> > +{
> > + __xe_vm_bind_assert(fd, vm, engine, 0, offset, addr, size,
> > + XE_VM_BIND_OP_PREFETCH | XE_VM_BIND_FLAG_ASYNC,
> > + sync, num_syncs, region, 0);
> > +}
> > +
> > +void xe_vm_bind_async(int fd, uint32_t vm, uint32_t engine, uint32_t bo,
> > + uint64_t offset, uint64_t addr, uint64_t size,
> > + struct drm_xe_sync *sync, uint32_t num_syncs)
> > +{
> > + __xe_vm_bind_assert(fd, vm, engine, bo, offset, addr, size,
> > + XE_VM_BIND_OP_MAP | XE_VM_BIND_FLAG_ASYNC, sync,
> > + num_syncs, 0, 0);
> > +}
> > +
> > +void xe_vm_bind_async_flags(int fd, uint32_t vm, uint32_t engine, uint32_t bo,
> > + uint64_t offset, uint64_t addr, uint64_t size,
> > + struct drm_xe_sync *sync, uint32_t num_syncs,
> > + uint32_t flags)
> > +{
> > + __xe_vm_bind_assert(fd, vm, engine, bo, offset, addr, size,
> > + XE_VM_BIND_OP_MAP | XE_VM_BIND_FLAG_ASYNC | flags,
> > + sync, num_syncs, 0, 0);
> > +}
> > +
> > +void xe_vm_bind_userptr_async(int fd, uint32_t vm, uint32_t engine,
> > + uint64_t userptr, uint64_t addr, uint64_t size,
> > + struct drm_xe_sync *sync, uint32_t num_syncs)
> > +{
> > + __xe_vm_bind_assert(fd, vm, engine, 0, userptr, addr, size,
> > + XE_VM_BIND_OP_MAP_USERPTR | XE_VM_BIND_FLAG_ASYNC,
> > + sync, num_syncs, 0, 0);
> > +}
> > +
> > +void xe_vm_bind_userptr_async_flags(int fd, uint32_t vm, uint32_t engine,
> > + uint64_t userptr, uint64_t addr,
> > + uint64_t size, struct drm_xe_sync *sync,
> > + uint32_t num_syncs, uint32_t flags)
> > +{
> > + __xe_vm_bind_assert(fd, vm, engine, 0, userptr, addr, size,
> > + XE_VM_BIND_OP_MAP_USERPTR | XE_VM_BIND_FLAG_ASYNC |
> > + flags, sync, num_syncs, 0, 0);
> > +}
> > +
> > +void xe_vm_unbind_async(int fd, uint32_t vm, uint32_t engine,
> > + uint64_t offset, uint64_t addr, uint64_t size,
> > + struct drm_xe_sync *sync, uint32_t num_syncs)
> > +{
> > + __xe_vm_bind_assert(fd, vm, engine, 0, offset, addr, size,
> > + XE_VM_BIND_OP_UNMAP | XE_VM_BIND_FLAG_ASYNC, sync,
> > + num_syncs, 0, 0);
> > +}
> > +
> > +static void __xe_vm_bind_sync(int fd, uint32_t vm, uint32_t bo, uint64_t offset,
> > + uint64_t addr, uint64_t size, uint32_t op)
> > +{
> > + struct drm_xe_sync sync = {
> > + .flags = DRM_XE_SYNC_SYNCOBJ | DRM_XE_SYNC_SIGNAL,
> > + .handle = syncobj_create(fd, 0),
> > + };
> > +
> > + __xe_vm_bind_assert(fd, vm, 0, bo, offset, addr, size, op, &sync, 1, 0,
> > + 0);
> > +
> > + igt_assert(syncobj_wait(fd, &sync.handle, 1, INT64_MAX, 0, NULL));
> > + syncobj_destroy(fd, sync.handle);
> > +}
> > +
> > +void xe_vm_bind_sync(int fd, uint32_t vm, uint32_t bo, uint64_t offset,
> > + uint64_t addr, uint64_t size)
> > +{
> > + __xe_vm_bind_sync(fd, vm, bo, offset, addr, size, XE_VM_BIND_OP_MAP);
> > +}
> > +
> > +void xe_vm_unbind_sync(int fd, uint32_t vm, uint64_t offset,
> > + uint64_t addr, uint64_t size)
> > +{
> > + __xe_vm_bind_sync(fd, vm, 0, offset, addr, size, XE_VM_BIND_OP_UNMAP);
> > +}
> > +
> > +void xe_vm_destroy(int fd, uint32_t vm)
> > +{
> > + struct drm_xe_vm_destroy destroy = {
> > + .vm_id = vm,
> > + };
> > +
> > + igt_assert_eq(igt_ioctl(fd, DRM_IOCTL_XE_VM_DESTROY, &destroy), 0);
> > +}
> > +
> > +uint32_t xe_bo_create_flags(int fd, uint32_t vm, uint64_t size, uint32_t flags)
> > +{
> > + struct drm_xe_gem_create create = {
> > + .vm_id = vm,
> > + .size = size,
> > + .flags = flags,
> > + };
> > +
> > + igt_assert_eq(igt_ioctl(fd, DRM_IOCTL_XE_GEM_CREATE, &create), 0);
> > +
> > + return create.handle;
> > +}
> > +
> > +uint32_t xe_bo_create(int fd, int gt, uint32_t vm, uint64_t size)
> > +{
> > + struct drm_xe_gem_create create = {
> > + .vm_id = vm,
> > + .size = size,
> > + .flags = vram_if_possible(fd, gt),
> > + };
> > + int err;
> > +
> > + err = igt_ioctl(fd, DRM_IOCTL_XE_GEM_CREATE, &create);
> > + igt_assert_eq(err, 0);
> > +
> > + return create.handle;
> > +}
> > +
> > +uint32_t xe_bind_engine_create(int fd, uint32_t vm, uint64_t ext)
> > +{
> > + struct drm_xe_engine_class_instance instance = {
> > + .engine_class = DRM_XE_ENGINE_CLASS_VM_BIND,
> > + };
> > + struct drm_xe_engine_create create = {
> > + .extensions = ext,
> > + .vm_id = vm,
> > + .width = 1,
> > + .num_placements = 1,
> > + .instances = to_user_pointer(&instance),
> > + };
> > +
> > + igt_assert_eq(igt_ioctl(fd, DRM_IOCTL_XE_ENGINE_CREATE, &create), 0);
> > +
> > + return create.engine_id;
> > +}
> > +
> > +uint32_t xe_engine_create(int fd, uint32_t vm,
> > + struct drm_xe_engine_class_instance *instance,
> > + uint64_t ext)
> > +{
> > + struct drm_xe_engine_create create = {
> > + .extensions = ext,
> > + .vm_id = vm,
> > + .width = 1,
> > + .num_placements = 1,
> > + .instances = to_user_pointer(instance),
> > + };
> > +
> > + igt_assert_eq(igt_ioctl(fd, DRM_IOCTL_XE_ENGINE_CREATE, &create), 0);
> > +
> > + return create.engine_id;
> > +}
> > +
> > +uint32_t xe_engine_create_class(int fd, uint32_t vm, uint16_t class)
> > +{
> > + struct drm_xe_engine_class_instance instance = {
> > + .engine_class = class,
> > + .engine_instance = 0,
> > + .gt_id = 0,
> > + };
> > + struct drm_xe_engine_create create = {
> > + .vm_id = vm,
> > + .width = 1,
> > + .num_placements = 1,
> > + .instances = to_user_pointer(&instance),
> > + };
> > +
> > + igt_assert_eq(igt_ioctl(fd, DRM_IOCTL_XE_ENGINE_CREATE, &create), 0);
> > +
> > + return create.engine_id;
> > +}
> > +
> > +void xe_engine_destroy(int fd, uint32_t engine)
> > +{
> > + struct drm_xe_engine_destroy destroy = {
> > + .engine_id = engine,
> > + };
> > +
> > + igt_assert_eq(igt_ioctl(fd, DRM_IOCTL_XE_ENGINE_DESTROY, &destroy), 0);
> > +}
> > +
> > +uint64_t xe_bo_mmap_offset(int fd, uint32_t bo)
> > +{
> > + struct drm_xe_gem_mmap_offset mmo = {
> > + .handle = bo,
> > + };
> > +
> > + igt_assert_eq(igt_ioctl(fd, DRM_IOCTL_XE_GEM_MMAP_OFFSET, &mmo), 0);
> > +
> > + return mmo.offset;
> > +}
> > +
> > +void *xe_bo_map(int fd, uint32_t bo, size_t size)
> > +{
> > + uint64_t mmo;
> > + void *map;
> > +
> > + mmo = xe_bo_mmap_offset(fd, bo);
> > + map = mmap(NULL, size, PROT_WRITE, MAP_SHARED, fd, mmo);
> > + igt_assert(map != MAP_FAILED);
> > +
> > + return map;
> > +}
> > +
> > +static int __xe_exec(int fd, struct drm_xe_exec *exec)
> > +{
> > + int err = 0;
> > +
> > + if (igt_ioctl(fd, DRM_IOCTL_XE_EXEC, exec)) {
> > + err = -errno;
> > + igt_assume(err != 0);
> > + }
> > + errno = 0;
> > + return err;
> > +}
> > +
> > +void xe_exec(int fd, struct drm_xe_exec *exec)
> > +{
> > + igt_assert_eq(__xe_exec(fd, exec), 0);
> > +}
> > +
> > +void xe_exec_sync(int fd, uint32_t engine, uint64_t addr,
> > + struct drm_xe_sync *sync, uint32_t num_syncs)
> > +{
> > + struct drm_xe_exec exec = {
> > + .engine_id = engine,
> > + .syncs = (uintptr_t)sync,
> > + .num_syncs = num_syncs,
> > + .address = addr,
> > + .num_batch_buffer = 1,
> > + };
> > +
> > + igt_assert_eq(__xe_exec(fd, &exec), 0);
> > +}
> > +
> > +void xe_exec_wait(int fd, uint32_t engine, uint64_t addr)
> > +{
> > + struct drm_xe_sync sync = {
> > + .flags = DRM_XE_SYNC_SYNCOBJ | DRM_XE_SYNC_SIGNAL,
> > + .handle = syncobj_create(fd, 0),
> > + };
> > +
> > + xe_exec_sync(fd, engine, addr, &sync, 1);
> > +
> > + igt_assert(syncobj_wait(fd, &sync.handle, 1, INT64_MAX, 0, NULL));
> > + syncobj_destroy(fd, sync.handle);
> > +}
> > +
> > +void xe_wait_ufence(int fd, uint64_t *addr, uint64_t value,
> > + struct drm_xe_engine_class_instance *eci,
> > + int64_t timeout)
> > +{
> > + struct drm_xe_wait_user_fence wait = {
> > + .addr = to_user_pointer(addr),
> > + .op = DRM_XE_UFENCE_WAIT_EQ,
> > + .flags = !eci ? DRM_XE_UFENCE_WAIT_SOFT_OP : 0,
> > + .value = value,
> > + .mask = DRM_XE_UFENCE_WAIT_U64,
> > + .timeout = timeout,
> > + .num_engines = eci ? 1 :0,
> > + .instances = eci ? to_user_pointer(eci) : 0,
> > + };
> > +
> > + igt_assert_eq(igt_ioctl(fd, DRM_IOCTL_XE_WAIT_USER_FENCE, &wait), 0);
> > +}
> > +
> > +void xe_force_gt_reset(int fd, int gt)
> > +{
> > + char reset_string[128];
> > +
> > + sprintf(reset_string, "cat /sys/kernel/debug/dri/0/gt%d/force_reset", gt);
>
> this seems wrong: you can't assume Xe will be always dri 0
>
Agree, dri/n should be get from fd open path. I will fix it.
> (another topic is why reset is triggered by reading debugfs, but that's
> not your fault)
>
> > + system(reset_string);
> > +}
> > +
> > +void xe_vm_madvise(int fd, uint32_t vm, uint64_t addr, uint64_t size,
> > + uint32_t property, uint32_t value)
> > +{
> > + struct drm_xe_vm_madvise madvise = {
> > + .vm_id = vm,
> > + .range = size,
> > + .addr = addr,
> > + .property = property,
> > + .value = value,
> > + };
> > +
> > + igt_assert_eq(igt_ioctl(fd, DRM_IOCTL_XE_VM_MADVISE, &madvise), 0);
> > +}
> > diff --git a/lib/xe/xe_ioctl.h b/lib/xe/xe_ioctl.h
> > new file mode 100644
> > index 00000000..70704364
> > --- /dev/null
> > +++ b/lib/xe/xe_ioctl.h
> > @@ -0,0 +1,89 @@
> > +/* SPDX-License-Identifier: MIT */
> > +/*
> > + * Copyright © 2023 Intel Corporation
> > + *
> > + * Authors:
> > + * Jason Ekstrand <jason at jlekstrand.net>
> > + * Maarten Lankhorst <maarten.lankhorst at linux.intel.com>
> > + * Matthew Brost <matthew.brost at intel.com>
> > + */
> > +
> > +#ifndef XE_IOCTL_H
> > +#define XE_IOCTL_H
> > +
> > +#include <stdint.h>
> > +#include <stdbool.h>
> > +#include <sys/mman.h>
> > +#include <xe_drm.h>
>
> do you really need all these headers in this file ?
>
Ok, stdint.h and xe_drm.h are enough.
> > +
> > +uint32_t xe_cs_prefetch_size(int fd);
> > +uint32_t xe_vm_create(int fd, uint32_t flags, uint64_t ext);
> > +int __xe_vm_bind(int fd, uint32_t vm, uint32_t engine, uint32_t bo,
> > + uint64_t offset, uint64_t addr, uint64_t size, uint32_t op,
> > + struct drm_xe_sync *sync, uint32_t num_syncs, uint32_t region,
> > + uint64_t ext);
> > +void __xe_vm_bind_assert(int fd, uint32_t vm, uint32_t engine, uint32_t bo,
> > + uint64_t offset, uint64_t addr, uint64_t size,
> > + uint32_t op, struct drm_xe_sync *sync,
> > + uint32_t num_syncs, uint32_t region, uint64_t ext);
> > +void xe_vm_bind(int fd, uint32_t vm, uint32_t bo, uint64_t offset,
> > + uint64_t addr, uint64_t size,
> > + struct drm_xe_sync *sync, uint32_t num_syncs);
> > +void xe_vm_unbind(int fd, uint32_t vm, uint64_t offset,
> > + uint64_t addr, uint64_t size,
> > + struct drm_xe_sync *sync, uint32_t num_syncs);
> > +void xe_vm_prefetch_async(int fd, uint32_t vm, uint32_t engine,
> > + uint64_t offset, uint64_t addr, uint64_t size,
> > + struct drm_xe_sync *sync, uint32_t num_syncs,
> > + uint32_t region);
> > +void xe_vm_bind_async(int fd, uint32_t vm, uint32_t engine, uint32_t bo,
> > + uint64_t offset, uint64_t addr, uint64_t size,
> > + struct drm_xe_sync *sync, uint32_t num_syncs);
> > +void xe_vm_bind_userptr_async(int fd, uint32_t vm, uint32_t engine,
> > + uint64_t userptr, uint64_t addr, uint64_t size,
> > + struct drm_xe_sync *sync, uint32_t num_syncs);
> > +void xe_vm_bind_async_flags(int fd, uint32_t vm, uint32_t engine, uint32_t bo,
> > + uint64_t offset, uint64_t addr, uint64_t size,
> > + struct drm_xe_sync *sync, uint32_t num_syncs,
> > + uint32_t flags);
> > +void xe_vm_bind_userptr_async_flags(int fd, uint32_t vm, uint32_t engine,
> > + uint64_t userptr, uint64_t addr,
> > + uint64_t size, struct drm_xe_sync *sync,
> > + uint32_t num_syncs, uint32_t flags);
> > +void xe_vm_unbind_async(int fd, uint32_t vm, uint32_t engine,
> > + uint64_t offset, uint64_t addr, uint64_t size,
> > + struct drm_xe_sync *sync, uint32_t num_syncs);
> > +void xe_vm_bind_sync(int fd, uint32_t vm, uint32_t bo, uint64_t offset,
> > + uint64_t addr, uint64_t size);
> > +void xe_vm_unbind_sync(int fd, uint32_t vm, uint64_t offset,
> > + uint64_t addr, uint64_t size);
> > +void xe_vm_bind_array(int fd, uint32_t vm, uint32_t engine,
> > + struct drm_xe_vm_bind_op *bind_ops,
> > + uint32_t num_bind, struct drm_xe_sync *sync,
> > + uint32_t num_syncs);
> > +void xe_vm_unbind_all_async(int fd, uint32_t vm, uint32_t engine,
> > + uint32_t bo, struct drm_xe_sync *sync,
> > + uint32_t num_syncs);
> > +void xe_vm_destroy(int fd, uint32_t vm);
> > +uint32_t xe_bo_create_flags(int fd, uint32_t vm, uint64_t size, uint32_t flags);
> > +uint32_t xe_bo_create(int fd, int gt, uint32_t vm, uint64_t size);
> > +uint32_t xe_engine_create(int fd, uint32_t vm,
> > + struct drm_xe_engine_class_instance *instance,
> > + uint64_t ext);
> > +uint32_t xe_bind_engine_create(int fd, uint32_t vm, uint64_t ext);
> > +uint32_t xe_engine_create_class(int fd, uint32_t vm, uint16_t class);
> > +void xe_engine_destroy(int fd, uint32_t engine);
> > +uint64_t xe_bo_mmap_offset(int fd, uint32_t bo);
> > +void *xe_bo_map(int fd, uint32_t bo, size_t size);
> > +void xe_exec(int fd, struct drm_xe_exec *exec);
> > +void xe_exec_sync(int fd, uint32_t engine, uint64_t addr,
> > + struct drm_xe_sync *sync, uint32_t num_syncs);
> > +void xe_exec_wait(int fd, uint32_t engine, uint64_t addr);
> > +void xe_wait_ufence(int fd, uint64_t *addr, uint64_t value,
> > + struct drm_xe_engine_class_instance *eci,
> > + int64_t timeout);
> > +void xe_force_gt_reset(int fd, int gt);
> > +void xe_vm_madvise(int fd, uint32_t vm, uint64_t addr, uint64_t size,
> > + uint32_t property, uint32_t value);
> > +
> > +#endif /* XE_IOCTL_H */
> > diff --git a/lib/xe/xe_query.c b/lib/xe/xe_query.c
> > new file mode 100644
> > index 00000000..c31cad46
> > --- /dev/null
> > +++ b/lib/xe/xe_query.c
> > @@ -0,0 +1,465 @@
> > +// SPDX-License-Identifier: MIT
> > +/*
> > + * Copyright © 2023 Intel Corporation
> > + *
> > + * Authors:
> > + * Matthew Brost <matthew.brost at intel.com>
> > + */
> > +
> > +#include <stdlib.h>
> > +#include <pthread.h>
> > +
> > +#include "drmtest.h"
> > +#include "ioctl_wrappers.h"
> > +#include "igt_map.h"
> > +
> > +#include "xe_query.h"
> > +#include "xe_ioctl.h"
> > +
> > +static struct drm_xe_query_config *xe_query_config_new(int fd)
> > +{
> > + struct drm_xe_query_config *config;
> > + struct drm_xe_device_query query = {
> > + .extensions = 0,
> > + .query = DRM_XE_DEVICE_QUERY_CONFIG,
> > + .size = 0,
> > + .data = 0,
> > + };
> > +
> > + igt_assert_eq(igt_ioctl(fd, DRM_IOCTL_XE_DEVICE_QUERY, &query), 0);
> > +
> > + config = malloc(query.size);
> > + igt_assert(config);
> > +
> > + query.data = to_user_pointer(config);
> > + igt_assert_eq(igt_ioctl(fd, DRM_IOCTL_XE_DEVICE_QUERY, &query), 0);
> > +
> > + igt_assert(config->num_params > 0);
> > +
> > + return config;
> > +}
> > +
> > +static struct drm_xe_query_gts *xe_query_gts_new(int fd)
> > +{
> > + struct drm_xe_query_gts *gts;
> > + struct drm_xe_device_query query = {
> > + .extensions = 0,
> > + .query = DRM_XE_DEVICE_QUERY_GTS,
> > + .size = 0,
> > + .data = 0,
> > + };
> > +
> > + igt_assert_eq(igt_ioctl(fd, DRM_IOCTL_XE_DEVICE_QUERY, &query), 0);
> > +
> > + gts = malloc(query.size);
> > + igt_assert(gts);
> > +
> > + query.data = to_user_pointer(gts);
> > + igt_assert_eq(igt_ioctl(fd, DRM_IOCTL_XE_DEVICE_QUERY, &query), 0);
> > +
> > + return gts;
> > +}
> > +
> > +static uint64_t __memory_regions(const struct drm_xe_query_gts *gts)
> > +{
> > + uint64_t regions = 0;
> > + int i;
> > +
> > + for (i = 0; i < gts->num_gt; i++)
> > + regions |= gts->gts[i].native_mem_regions |
> > + gts->gts[i].slow_mem_regions;
> > +
> > + return regions;
> > +}
> > +
> > +static struct drm_xe_engine_class_instance *
> > +xe_query_engines_new(int fd, int *num_engines)
> > +{
> > + struct drm_xe_engine_class_instance *hw_engines;
> > + struct drm_xe_device_query query = {
> > + .extensions = 0,
> > + .query = DRM_XE_DEVICE_QUERY_ENGINES,
> > + .size = 0,
> > + .data = 0,
> > + };
> > +
> > + igt_assert(num_engines);
> > + igt_assert_eq(igt_ioctl(fd, DRM_IOCTL_XE_DEVICE_QUERY, &query), 0);
> > +
> > + hw_engines = malloc(query.size);
> > + igt_assert(hw_engines);
> > +
> > + query.data = to_user_pointer(hw_engines);
> > + igt_assert_eq(igt_ioctl(fd, DRM_IOCTL_XE_DEVICE_QUERY, &query), 0);
> > +
> > + *num_engines = query.size / sizeof(*hw_engines);
> > +
> > + return hw_engines;
> > +}
> > +
> > +static struct drm_xe_query_mem_usage *xe_query_mem_usage_new(int fd)
> > +{
> > + struct drm_xe_query_mem_usage *mem_usage;
> > + struct drm_xe_device_query query = {
> > + .extensions = 0,
> > + .query = DRM_XE_DEVICE_QUERY_MEM_USAGE,
> > + .size = 0,
> > + .data = 0,
> > + };
> > +
> > + igt_assert_eq(igt_ioctl(fd, DRM_IOCTL_XE_DEVICE_QUERY, &query), 0);
> > +
> > + mem_usage = malloc(query.size);
> > + igt_assert(mem_usage);
> > +
> > + query.data = to_user_pointer(mem_usage);
> > + igt_assert_eq(igt_ioctl(fd, DRM_IOCTL_XE_DEVICE_QUERY, &query), 0);
> > +
> > + return mem_usage;
> > +}
> > +
> > +/* FIXME: Make generic / multi-GT aware */
>
> why can't it be fixed right now ?
>
Ok, I'm going to fix it so it will be in v4.
> > +static uint64_t __mem_vram_size(struct drm_xe_query_mem_usage *mem_usage)
> > +{
> > + for (int i = 0; i < mem_usage->num_regions; i++)
> > + if (mem_usage->regions[i].mem_class == XE_MEM_REGION_CLASS_VRAM)
> > + return mem_usage->regions[i].total_size;
> > +
> > + return 0;
> > +}
> > +
> > +static bool __mem_has_vram(struct drm_xe_query_mem_usage *mem_usage)
> > +{
> > + for (int i = 0; i < mem_usage->num_regions; i++)
> > + if (mem_usage->regions[i].mem_class == XE_MEM_REGION_CLASS_VRAM)
> > + return true;
> > +
> > + return false;
> > +}
> > +
> > +static uint32_t __mem_default_alignment(struct drm_xe_query_mem_usage *mem_usage)
> > +{
> > + uint32_t alignment = XE_DEFAULT_ALIGNMENT;
> > +
> > + for (int i = 0; i < mem_usage->num_regions; i++)
> > + if (alignment < mem_usage->regions[i].min_page_size)
> > + alignment = mem_usage->regions[i].min_page_size;
> > +
> > + return alignment;
> > +}
> > +
> > +static bool xe_check_supports_faults(int fd)
> > +{
> > + bool supports_faults;
> > +
> > + struct drm_xe_vm_create create = {
> > + .flags = DRM_XE_VM_CREATE_ASYNC_BIND_OPS |
> > + DRM_XE_VM_CREATE_FAULT_MODE,
> > + };
> > +
> > + supports_faults = !igt_ioctl(fd, DRM_IOCTL_XE_VM_CREATE, &create);
> > +
> > + if (supports_faults)
> > + xe_vm_destroy(fd, create.vm_id);
> > +
> > + return supports_faults;
> > +}
> > +
> > +/**
> > + * xe_engine_class_string:
> > + * @engine_class: engine class
> > + *
> > + * Returns engine class name or 'unknown class engine' otherwise.
> > + */
> > +const char *xe_engine_class_string(uint32_t engine_class)
> > +{
> > + switch (engine_class) {
> > + case DRM_XE_ENGINE_CLASS_RENDER:
> > + return "DRM_XE_ENGINE_CLASS_RENDER";
> > + case DRM_XE_ENGINE_CLASS_COPY:
> > + return "DRM_XE_ENGINE_CLASS_COPY";
> > + case DRM_XE_ENGINE_CLASS_VIDEO_DECODE:
> > + return "DRM_XE_ENGINE_CLASS_VIDEO_DECODE";
> > + case DRM_XE_ENGINE_CLASS_VIDEO_ENHANCE:
> > + return "DRM_XE_ENGINE_CLASS_VIDEO_ENHANCE";
> > + case DRM_XE_ENGINE_CLASS_COMPUTE:
> > + return "DRM_XE_ENGINE_CLASS_COMPUTE";
> > + default:
> > + igt_warn("Engine class %x unknown\n", engine_class);
>
> %x without 0x prefix might be misleading
>
Ok, fixed.
> > + return "unknown engine class";
> > + }
> > +}
> > +
> > +static struct xe_device_cache {
> > + pthread_mutex_t cache_mutex;
> > + struct igt_map *map;
> > +} cache;
> > +
> > +static struct xe_device *find_in_cache_unlocked(int fd)
> > +{
> > + return igt_map_search(cache.map, &fd);
> > +}
> > +
> > +static struct xe_device *find_in_cache(int fd)
> > +{
> > + struct xe_device *xe_dev;
> > +
> > + pthread_mutex_lock(&cache.cache_mutex);
> > + xe_dev = find_in_cache_unlocked(fd);
> > + pthread_mutex_unlock(&cache.cache_mutex);
> > +
> > + return xe_dev;
> > +}
> > +
> > +/**
> > + * xe_device_get:
> > + * @fd: xe device fd
> > + *
> > + * Function creates and caches xe_device struct which contains configuration
> > + * data returned in few queries. Subsequent calls returns previously
> > + * created xe_device. To remove this from cache xe_device_put() must be
> > + * called.
> > + */
> > +struct xe_device *xe_device_get(int fd)
> > +{
> > + struct xe_device *xe_dev;
> > +
> > + xe_dev = find_in_cache(fd);
> > + if (xe_dev)
> > + return xe_dev;
> > +
> > + xe_dev = calloc(1, sizeof(*xe_dev));
> > + igt_assert(xe_dev);
> > +
> > + xe_dev->fd = fd;
> > + xe_dev->config = xe_query_config_new(fd);
> > + xe_dev->number_gt = xe_dev->config->info[XE_QUERY_CONFIG_GT_COUNT];
> > + xe_dev->va_bits = xe_dev->config->info[XE_QUERY_CONFIG_VA_BITS];
> > + xe_dev->gts = xe_query_gts_new(fd);
> > + xe_dev->memory_regions = __memory_regions(xe_dev->gts);
> > + xe_dev->hw_engines = xe_query_engines_new(fd, &xe_dev->number_hw_engines);
> > + xe_dev->mem_usage = xe_query_mem_usage_new(fd);
> > + xe_dev->vram_size = __mem_vram_size(xe_dev->mem_usage);
> > + xe_dev->default_alignment = __mem_default_alignment(xe_dev->mem_usage);
> > + xe_dev->has_vram = __mem_has_vram(xe_dev->mem_usage);
> > + xe_dev->supports_faults = xe_check_supports_faults(fd);
> > +
> > + igt_map_insert(cache.map, &xe_dev->fd, xe_dev);
> > +
> > + return xe_dev;
> > +}
> > +
> > +static void xe_device_free(struct xe_device *xe_dev)
> > +{
> > + free(xe_dev->config);
> > + free(xe_dev->gts);
> > + free(xe_dev->hw_engines);
> > + free(xe_dev->mem_usage);
> > + free(xe_dev);
> > +}
> > +
> > +static void delete_in_cache(struct igt_map_entry *entry)
> > +{
> > + xe_device_free((struct xe_device *)entry->data);
> > +}
> > +
> > +/**
> > + * xe_device_put:
> > + * @fd: xe device fd
> > + *
> > + * Remove previously allocated and cached xe_device (if any).
> > + */
> > +void xe_device_put(int fd)
> > +{
> > + pthread_mutex_lock(&cache.cache_mutex);
> > + if (find_in_cache_unlocked(fd))
> > + igt_map_remove(cache.map, &fd, delete_in_cache);
> > + pthread_mutex_unlock(&cache.cache_mutex);
> > +}
> > +
> > +static void xe_device_destroy_cache(void)
> > +{
> > + pthread_mutex_lock(&cache.cache_mutex);
> > + igt_map_destroy(cache.map, delete_in_cache);
> > + pthread_mutex_unlock(&cache.cache_mutex);
> > +}
> > +
> > +static void xe_device_cache_init(void)
> > +{
> > + pthread_mutex_init(&cache.cache_mutex, NULL);
> > + xe_device_destroy_cache();
> > + cache.map = igt_map_create(igt_map_hash_32, igt_map_equal_32);
> > +}
> > +
> > +#define RETV(__v) \
> > + struct xe_device *xe_dev;\
> > + xe_dev = find_in_cache(fd);\
> > + igt_assert(xe_dev);\
> > + return xe_dev->__v
>
> return from within macro is rather not welcomed
>
> can't we just define full function template instead?
> (if we really need to have a macro?)
>
> #define xe_dev_XXX(_X, _T) \
> _T xe_dev_##_X(int fd) \
> { \
> struct xe_device *xe_dev; \
> \
> xe_dev = find_in_cache(fd); \
> igt_assert(xe_dev); \
> return xe_dev->_X; \
> }
>
> xe_dev_XXX(number_gt, int)
> xe_dev_XXX(memory_regions, uint64_t)
> xe_dev_XXX(hw_engines, struct drm_xe_engine_class_instance*)
> ...
>
>
Macro was interoduced as I didn't wanted to touch most xe-igt
tests which uses such getters. But define template looks better
than my RETV() macro so I'll use it.
> > +
> > +/**
> > + * xe_number_gt:
> > + * @fd: xe device fd
> > + *
> > + * Return number of gts for xe device fd.
> > + */
> > +int xe_number_gt(int fd)
> > +{
> > + RETV(number_gt);
> > +}
> > +
> > +/**
> > + * all_memory_regions:
> > + * @fd: xe device fd
> > + *
> > + * Returns memory regions bitmask for xe device @fd.
> > + */
> > +uint64_t all_memory_regions(int fd)
> > +{
> > + RETV(memory_regions);
> > +}
> > +
> > +/**
> > + * system_memory:
> > + * @fd: xe device fd
> > + *
> > + * Returns system memory bitmask for xe device @fd.
> > + */
> > +uint64_t system_memory(int fd)
> > +{
> > + uint64_t regions = all_memory_regions(fd);
> > +
> > + return regions & 0x1;
> > +}
> > +
> > +/**
> > + * vram_memory:
> > + * @fd: xe device fd
> > + * @gt: gt id
> > + *
> > + * Returns vram memory bitmask for xe device @fd and @gt id.
> > + */
> > +uint64_t vram_memory(int fd, int gt)
> > +{
> > + uint64_t regions = all_memory_regions(fd);
> > +
> > + return regions & (0x2 << gt);
> > +}
> > +
> > +/**
> > + * vram_if_possible:
> > + * @fd: xe device fd
> > + * @gt: gt id
> > + *
> > + * Returns vram memory bitmask for xe device @fd and @gt id or system memory
> > + * if there's no vram memory available for @gt.
> > + */
> > +uint64_t vram_if_possible(int fd, int gt)
> > +{
> > + uint64_t regions = all_memory_regions(fd);
> > + uint64_t system_memory = regions & 0x1;
> > + uint64_t vram = regions & (0x2 << gt);
> > +
> > + return vram ? vram : system_memory;
> > +}
> > +
> > +/**
> > + * xe_hw_engines:
> > + * @fd: xe device fd
> > + *
> > + * Returns engines array of xe device @fd.
> > + */
> > +struct drm_xe_engine_class_instance *xe_hw_engines(int fd)
> > +{
> > + RETV(hw_engines);
> > +}
> > +
> > +/**
> > + * xe_hw_engine:
> > + * @fd: xe device fd
> > + * @idx: engine index
> > + *
> > + * Returns engine instance of xe device @fd and @idx.
> > + */
> > +struct drm_xe_engine_class_instance *xe_hw_engine(int fd, int idx)
> > +{
> > + struct xe_device *xe_dev;
> > +
> > + xe_dev = find_in_cache(fd);
> > + igt_assert(xe_dev);
> > + igt_assert(idx >= 0 && idx < xe_dev->number_hw_engines);
> > +
> > + return &xe_dev->hw_engines[idx];
> > +}
> > +
> > +/**
> > + * xe_number_hw_engine:
> > + * @fd: xe device fd
> > + *
> > + * Returns number of hw engines of xe device @fd.
> > + */
> > +int xe_number_hw_engines(int fd)
> > +{
> > + RETV(number_hw_engines);
> > +}
> > +
> > +/**
> > + * xe_has_vram:
> > + * @fd: xe device fd
> > + *
> > + * Returns true if xe device @fd has vram otherwise false.
> > + */
> > +bool xe_has_vram(int fd)
> > +{
> > + RETV(has_vram);
> > +}
> > +
> > +/**
> > + * xe_vram_size:
> > + * @fd: xe device fd
> > + *
> > + * Returns size of vram of xe device @fd.
> > + */
> > +uint64_t xe_vram_size(int fd)
> > +{
> > + RETV(vram_size);
> > +}
> > +
> > +/**
> > + * xe_get_default_alignment:
> > + * @fd: xe device fd
> > + *
> > + * Returns default alignment of objects for xe device @fd.
> > + */
> > +uint32_t xe_get_default_alignment(int fd)
> > +{
> > + RETV(default_alignment);
> > +}
> > +
> > +/**
> > + * xe_supports_faults:
> > + * @fd: xe device fd
> > + *
> > + * Returns true if xe device @fd allows creating vm in fault mode otherwise
> > + * false.
> > + */
> > +bool xe_supports_faults(int fd)
> > +{
> > + RETV(supports_faults);
> > +}
> > +
> > +/**
> > + * xe_va_bits:
> > + * @fd: xe device fd
> > + *
> > + * Returns number of virtual address bits used in xe device @fd.
> > + */
> > +uint32_t xe_va_bits(int fd)
> > +{
> > + RETV(va_bits);
> > +}
> > +
> > +igt_constructor
> > +{
> > + xe_device_cache_init();
> > +}
> > diff --git a/lib/xe/xe_query.h b/lib/xe/xe_query.h
> > new file mode 100644
> > index 00000000..f8398a41
> > --- /dev/null
> > +++ b/lib/xe/xe_query.h
> > @@ -0,0 +1,64 @@
> > +/* SPDX-License-Identifier: MIT */
> > +/*
> > + * Copyright © 2023 Intel Corporation
> > + *
> > + * Authors:
> > + * Matthew Brost <matthew.brost at intel.com>
> > + */
> > +
> > +#ifndef XE_QUERY_H
> > +#define XE_QUERY_H
> > +
> > +#include <stdint.h>
> > +#include <xe_drm.h>
> > +#include "igt_list.h"
> > +
> > +#define XE_DEFAULT_ALIGNMENT 0x1000
> > +#define XE_DEFAULT_ALIGNMENT_64K 0x10000
>
> nit: maybe time to define SZ_xxx
>
Agree.
> > +
> > +struct xe_device {
>
> shouldn't we document all struct members ?
>
You're right, this is widely used in tests so definitely
it should be documented.
> > + int fd;
> > +
> > + struct drm_xe_query_config *config;
> > + struct drm_xe_query_gts *gts;
> > + uint64_t memory_regions;
> > + struct drm_xe_engine_class_instance *hw_engines;
> > + int number_hw_engines;
>
> unsigned int ?
>
Ack.
> > + struct drm_xe_query_mem_usage *mem_usage;
> > + uint64_t vram_size;
> > + uint32_t default_alignment;
> > + bool has_vram;
> > + bool supports_faults;
> > +
> > + int number_gt;
>
> unsigned int ?
>
Ack.
> > + uint32_t va_bits;
> > +};
> > +
> > +#define for_each_hw_engine(__fd, __hwe) \
> > + for (int __i = 0; __i < xe_number_hw_engines(__fd) && \
> > + (__hwe = xe_hw_engine(__fd, __i)); ++__i)
> > +#define for_each_hw_engine_class(__class) \
> > + for (__class = 0; __class < DRM_XE_ENGINE_CLASS_COMPUTE + 1; \
> > + ++__class)
> > +#define for_each_gt(__fd, __gt) \
> > + for (__gt = 0; __gt < xe_number_gt(__fd); ++__gt)
> > +
> > +int xe_number_gt(int fd);
> > +uint64_t all_memory_regions(int fd);
> > +uint64_t system_memory(int fd);
> > +uint64_t vram_memory(int fd, int gt);
> > +uint64_t vram_if_possible(int fd, int gt);
>
> any reason why above function don't have any common prefix like "xe" ?
>
I didn't want to touch all tests at this moment (I tried to minimize
number of changes in xe-igt where all that variables were globals).
> > +struct drm_xe_engine_class_instance *xe_hw_engines(int fd);
> > +struct drm_xe_engine_class_instance *xe_hw_engine(int fd, int idx);
> > +int xe_number_hw_engines(int fd);
> > +bool xe_has_vram(int fd);
> > +uint64_t xe_vram_size(int fd);
> > +uint32_t xe_get_default_alignment(int fd);
> > +uint32_t xe_va_bits(int fd);
> > +bool xe_supports_faults(int fd);
> > +const char* xe_engine_class_string(uint32_t engine_class);
> > +
> > +struct xe_device *xe_device_get(int fd);
> > +void xe_device_put(int fd);
> > +
> > +#endif /* XE_QUERY_H */
> > diff --git a/lib/xe/xe_spin.c b/lib/xe/xe_spin.c
> > new file mode 100644
> > index 00000000..d7ce9aac
> > --- /dev/null
> > +++ b/lib/xe/xe_spin.c
> > @@ -0,0 +1,139 @@
> > +// SPDX-License-Identifier: MIT
> > +/*
> > + * Copyright © 2023 Intel Corporation
> > + *
> > + * Authors:
> > + * Matthew Brost <matthew.brost at intel.com>
> > + */
> > +
> > +#include <string.h>
> > +
> > +#include "drmtest.h"
> > +#include "igt.h"
> > +#include "igt_core.h"
> > +#include "igt_syncobj.h"
> > +#include "intel_reg.h"
> > +#include "xe_ioctl.h"
> > +#include "xe_spin.h"
> > +
>
> in other file even trivial getter functions were documented,
>
> why spin functions here are all undocumented ?
>
I think because code was developed in the xe-igt repo and noone
previously asked. And I didn't want to make too much changes
to avoid same work again if I would need to squash this series
one more time.
Thanks for the review.
All things which I think might be addressed in v4 will be sent
soon.
--
Zbigniew
> > +void xe_spin_init(struct xe_spin *spin, uint64_t addr, bool preempt)
> > +{
> > + uint64_t batch_offset = (char *)&spin->batch - (char *)spin;
> > + uint64_t batch_addr = addr + batch_offset;
> > + uint64_t start_offset = (char *)&spin->start - (char *)spin;
> > + uint64_t start_addr = addr + start_offset;
> > + uint64_t end_offset = (char *)&spin->end - (char *)spin;
> > + uint64_t end_addr = addr + end_offset;
> > + int b = 0;
> > +
> > + spin->start = 0;
> > + spin->end = 0xffffffff;
> > +
> > + spin->batch[b++] = MI_STORE_DWORD_IMM;
> > + spin->batch[b++] = start_addr;
> > + spin->batch[b++] = start_addr >> 32;
> > + spin->batch[b++] = 0xc0ffee;
> > +
> > + if (preempt)
> > + spin->batch[b++] = (0x5 << 23);
> > +
> > + spin->batch[b++] = MI_COND_BATCH_BUFFER_END | MI_DO_COMPARE | 2;
> > + spin->batch[b++] = 0;
> > + spin->batch[b++] = end_addr;
> > + spin->batch[b++] = end_addr >> 32;
> > +
> > + spin->batch[b++] = MI_BATCH_BUFFER_START | 1 << 8 | 1;
> > + spin->batch[b++] = batch_addr;
> > + spin->batch[b++] = batch_addr >> 32;
> > +
> > + igt_assert(b <= ARRAY_SIZE(spin->batch));
> > +}
> > +
> > +bool xe_spin_started(struct xe_spin *spin)
> > +{
> > + return spin->start != 0;
> > +}
> > +
> > +void xe_spin_wait_started(struct xe_spin *spin)
> > +{
> > + while(!xe_spin_started(spin));
> > +}
> > +
> > +void xe_spin_end(struct xe_spin *spin)
> > +{
> > + spin->end = 0;
> > +}
> > +
> > +void xe_cork_init(int fd, struct drm_xe_engine_class_instance *hwe,
> > + struct xe_cork *cork)
> > +{
> > + uint64_t addr = xe_get_default_alignment(fd);
> > + size_t bo_size = xe_get_default_alignment(fd);
> > + uint32_t vm, bo, engine, syncobj;
> > + struct xe_spin *spin;
> > + struct drm_xe_sync sync = {
> > + .flags = DRM_XE_SYNC_SYNCOBJ | DRM_XE_SYNC_SIGNAL,
> > + };
> > + struct drm_xe_exec exec = {
> > + .num_batch_buffer = 1,
> > + .num_syncs = 1,
> > + .syncs = to_user_pointer(&sync),
> > + };
> > +
> > + vm = xe_vm_create(fd, 0, 0);
> > +
> > + bo = xe_bo_create(fd, hwe->gt_id, vm, bo_size);
> > + spin = xe_bo_map(fd, bo, 0x1000);
> > +
> > + xe_vm_bind_sync(fd, vm, bo, 0, addr, bo_size);
> > +
> > + engine = xe_engine_create(fd, vm, hwe, 0);
> > + syncobj = syncobj_create(fd, 0);
> > +
> > + xe_spin_init(spin, addr, true);
> > + exec.engine_id = engine;
> > + exec.address = addr;
> > + sync.handle = syncobj;
> > + igt_assert_eq(igt_ioctl(fd, DRM_IOCTL_XE_EXEC, &exec), 0);
> > +
> > + cork->spin = spin;
> > + cork->fd = fd;
> > + cork->vm = vm;
> > + cork->bo = bo;
> > + cork->engine = engine;
> > + cork->syncobj = syncobj;
> > +}
> > +
> > +bool xe_cork_started(struct xe_cork *cork)
> > +{
> > + return xe_spin_started(cork->spin);
> > +}
> > +
> > +void xe_cork_wait_started(struct xe_cork *cork)
> > +{
> > + xe_spin_wait_started(cork->spin);
> > +}
> > +
> > +void xe_cork_end(struct xe_cork *cork)
> > +{
> > + xe_spin_end(cork->spin);
> > +}
> > +
> > +void xe_cork_wait_done(struct xe_cork *cork)
> > +{
> > + igt_assert(syncobj_wait(cork->fd, &cork->syncobj, 1, INT64_MAX, 0,
> > + NULL));
> > +}
> > +
> > +void xe_cork_fini(struct xe_cork *cork)
> > +{
> > + syncobj_destroy(cork->fd, cork->syncobj);
> > + xe_engine_destroy(cork->fd, cork->engine);
> > + xe_vm_destroy(cork->fd, cork->vm);
> > + gem_close(cork->fd, cork->bo);
> > +}
> > +
> > +uint32_t xe_cork_sync_handle(struct xe_cork *cork)
> > +{
> > + return cork->syncobj;
> > +}
> > diff --git a/lib/xe/xe_spin.h b/lib/xe/xe_spin.h
> > new file mode 100644
> > index 00000000..73f9a026
> > --- /dev/null
> > +++ b/lib/xe/xe_spin.h
> > @@ -0,0 +1,48 @@
> > +/* SPDX-License-Identifier: MIT */
> > +/*
> > + * Copyright © 2023 Intel Corporation
> > + *
> > + * Authors:
> > + * Matthew Brost <matthew.brost at intel.com>
> > + */
> > +
> > +#ifndef XE_SPIN_H
> > +#define XE_SPIN_H
> > +
> > +#include <stdint.h>
> > +#include <stdbool.h>
> > +
> > +#include "xe_query.h"
> > +
> > +/* Mapped GPU object */
> > +struct xe_spin {
> > + uint32_t batch[16];
> > + uint64_t pad;
> > + uint32_t start;
> > + uint32_t end;
> > +};
> > +
> > +void xe_spin_init(struct xe_spin *spin, uint64_t addr, bool preempt);
> > +bool xe_spin_started(struct xe_spin *spin);
> > +void xe_spin_wait_started(struct xe_spin *spin);
> > +void xe_spin_end(struct xe_spin *spin);
> > +
> > +struct xe_cork {
> > + struct xe_spin *spin;
> > + int fd;
> > + uint32_t vm;
> > + uint32_t bo;
> > + uint32_t engine;
> > + uint32_t syncobj;
> > +};
> > +
> > +void xe_cork_init(int fd, struct drm_xe_engine_class_instance *hwe,
> > + struct xe_cork *cork);
> > +bool xe_cork_started(struct xe_cork *cork);
> > +void xe_cork_wait_started(struct xe_cork *cork);
> > +void xe_cork_end(struct xe_cork *cork);
> > +void xe_cork_wait_done(struct xe_cork *cork);
> > +void xe_cork_fini(struct xe_cork *cork);
> > +uint32_t xe_cork_sync_handle(struct xe_cork *cork);
> > +
> > +#endif /* XE_SPIN_H */
> > diff --git a/meson.build b/meson.build
> > index e7a68503..2a7f6078 100644
> > --- a/meson.build
> > +++ b/meson.build
> > @@ -261,6 +261,7 @@ libexecdir = join_paths(get_option('libexecdir'), 'igt-gpu-tools')
> > amdgpudir = join_paths(libexecdir, 'amdgpu')
> > v3ddir = join_paths(libexecdir, 'v3d')
> > vc4dir = join_paths(libexecdir, 'vc4')
> > +xedir = join_paths(libexecdir, 'xe')
> > mandir = get_option('mandir')
> > pkgconfigdir = join_paths(libdir, 'pkgconfig')
> > python3 = find_program('python3', required : true)
> > @@ -308,12 +309,19 @@ if get_option('use_rpath')
> > vc4_rpathdir = join_paths(vc4_rpathdir, '..')
> > endforeach
> > vc4_rpathdir = join_paths(vc4_rpathdir, libdir)
> > +
> > + xedir_rpathdir = '$ORIGIN'
> > + foreach p : xedir.split('/')
> > + xedir_rpathdir = join_paths(xedir_rpathdir, '..')
> > + endforeach
> > + xedir_rpathdir = join_paths(xedir_rpathdir, libdir)
> > else
> > bindir_rpathdir = ''
> > libexecdir_rpathdir = ''
> > amdgpudir_rpathdir = ''
> > v3d_rpathdir = ''
> > vc4_rpathdir = ''
> > + xedir_rpathdir = ''
> > endif
> >
> > subdir('lib')
More information about the igt-dev
mailing list