* IN THE SOFTWARE.
*/
+#include <assert.h>
#include <stdio.h>
#include <stdlib.h>
#include "gen_device_info.h"
#include "compiler/shader_enums.h"
+#include "util/macros.h"
static const struct gen_device_info gen_device_info_i965 = {
.gen = 4,
.has_negative_rhw_bug = true,
.num_slices = 1,
+ .num_subslices = { 1, },
+ .num_thread_per_eu = 4,
.max_vs_threads = 16,
.max_gs_threads = 2,
.max_wm_threads = 8 * 4,
.urb = {
.size = 256,
},
- .timebase_scale = 80,
+ .timestamp_frequency = 12500000,
};
static const struct gen_device_info gen_device_info_g4x = {
.has_surface_tile_offset = true,
.is_g4x = true,
.num_slices = 1,
+ .num_subslices = { 1, },
+ .num_thread_per_eu = 5,
.max_vs_threads = 32,
.max_gs_threads = 2,
.max_wm_threads = 10 * 5,
.urb = {
.size = 384,
},
- .timebase_scale = 80,
+ .timestamp_frequency = 12500000,
};
static const struct gen_device_info gen_device_info_ilk = {
.has_compr4 = true,
.has_surface_tile_offset = true,
.num_slices = 1,
+ .num_subslices = { 1, },
+ .num_thread_per_eu = 6,
.max_vs_threads = 72,
.max_gs_threads = 32,
.max_wm_threads = 12 * 6,
.urb = {
.size = 1024,
},
- .timebase_scale = 80,
+ .timestamp_frequency = 12500000,
};
static const struct gen_device_info gen_device_info_snb_gt1 = {
.has_surface_tile_offset = true,
.needs_unlit_centroid_workaround = true,
.num_slices = 1,
+ .num_subslices = { 1, },
+ .num_thread_per_eu = 6, /* Not confirmed */
.max_vs_threads = 24,
.max_gs_threads = 21, /* conservative; 24 if rendering disabled. */
.max_wm_threads = 40,
[MESA_SHADER_GEOMETRY] = 256,
},
},
- .timebase_scale = 80,
+ .timestamp_frequency = 12500000,
};
static const struct gen_device_info gen_device_info_snb_gt2 = {
.has_surface_tile_offset = true,
.needs_unlit_centroid_workaround = true,
.num_slices = 1,
+ .num_subslices = { 1, },
+ .num_thread_per_eu = 6, /* Not confirmed */
.max_vs_threads = 60,
.max_gs_threads = 60,
.max_wm_threads = 80,
[MESA_SHADER_GEOMETRY] = 256,
},
},
- .timebase_scale = 80,
+ .timestamp_frequency = 12500000,
};
#define GEN7_FEATURES \
.has_llc = true, \
.has_pln = true, \
.has_surface_tile_offset = true, \
- .timebase_scale = 80
+ .timestamp_frequency = 12500000
static const struct gen_device_info gen_device_info_ivb_gt1 = {
GEN7_FEATURES, .is_ivybridge = true, .gt = 1,
.num_slices = 1,
+ .num_subslices = { 1, },
+ .num_thread_per_eu = 6,
+ .l3_banks = 2,
.max_vs_threads = 36,
.max_tcs_threads = 36,
.max_tes_threads = 36,
static const struct gen_device_info gen_device_info_ivb_gt2 = {
GEN7_FEATURES, .is_ivybridge = true, .gt = 2,
.num_slices = 1,
+ .num_subslices = { 1, },
+ .num_thread_per_eu = 8, /* Not sure why this isn't a multiple of
+ * @max_wm_threads ... */
+ .l3_banks = 4,
.max_vs_threads = 128,
.max_tcs_threads = 128,
.max_tes_threads = 128,
static const struct gen_device_info gen_device_info_byt = {
GEN7_FEATURES, .is_baytrail = true, .gt = 1,
.num_slices = 1,
+ .num_subslices = { 1, },
+ .num_thread_per_eu = 8,
+ .l3_banks = 1,
.has_llc = false,
.max_vs_threads = 36,
.max_tcs_threads = 36,
static const struct gen_device_info gen_device_info_hsw_gt1 = {
HSW_FEATURES, .gt = 1,
.num_slices = 1,
+ .num_subslices = { 1, },
+ .num_thread_per_eu = 7,
+ .l3_banks = 2,
.max_vs_threads = 70,
.max_tcs_threads = 70,
.max_tes_threads = 70,
static const struct gen_device_info gen_device_info_hsw_gt2 = {
HSW_FEATURES, .gt = 2,
.num_slices = 1,
+ .num_subslices = { 2, },
+ .num_thread_per_eu = 7,
+ .l3_banks = 4,
.max_vs_threads = 280,
.max_tcs_threads = 256,
.max_tes_threads = 280,
static const struct gen_device_info gen_device_info_hsw_gt3 = {
HSW_FEATURES, .gt = 3,
.num_slices = 2,
+ .num_subslices = { 2, },
+ .num_thread_per_eu = 7,
+ .l3_banks = 8,
.max_vs_threads = 280,
.max_tcs_threads = 256,
.max_tes_threads = 280,
.max_tes_threads = 504, \
.max_gs_threads = 504, \
.max_wm_threads = 384, \
- .timebase_scale = 80
+ .timestamp_frequency = 12500000
static const struct gen_device_info gen_device_info_bdw_gt1 = {
GEN8_FEATURES, .gt = 1,
+ .is_broadwell = true,
.num_slices = 1,
+ .num_subslices = { 2, },
+ .num_thread_per_eu = 7,
+ .l3_banks = 2,
.max_cs_threads = 42,
.urb = {
.size = 192,
static const struct gen_device_info gen_device_info_bdw_gt2 = {
GEN8_FEATURES, .gt = 2,
+ .is_broadwell = true,
.num_slices = 1,
+ .num_subslices = { 3, },
+ .num_thread_per_eu = 7,
+ .l3_banks = 4,
.max_cs_threads = 56,
.urb = {
.size = 384,
static const struct gen_device_info gen_device_info_bdw_gt3 = {
GEN8_FEATURES, .gt = 3,
+ .is_broadwell = true,
.num_slices = 2,
+ .num_subslices = { 3, 3, },
+ .num_thread_per_eu = 7,
+ .l3_banks = 8,
.max_cs_threads = 56,
.urb = {
.size = 384,
GEN8_FEATURES, .is_cherryview = 1, .gt = 1,
.has_llc = false,
.num_slices = 1,
+ .num_subslices = { 2, },
+ .num_thread_per_eu = 7,
+ .l3_banks = 2,
.max_vs_threads = 80,
.max_tcs_threads = 80,
.max_tes_threads = 80,
}
};
-#define GEN9_FEATURES \
+#define GEN9_HW_INFO \
.gen = 9, \
- .has_hiz_and_separate_stencil = true, \
- .has_resource_streamer = true, \
- .must_use_separate_stencil = true, \
- .has_llc = true, \
- .has_pln = true, \
- .supports_simd16_3src = true, \
- .has_surface_tile_offset = true, \
.max_vs_threads = 336, \
.max_gs_threads = 336, \
.max_tcs_threads = 336, \
.max_tes_threads = 336, \
.max_cs_threads = 56, \
- .timebase_scale = 1000000000.0 / 12000000.0, \
+ .timestamp_frequency = 12000000, \
.urb = { \
.size = 384, \
.min_entries = { \
}
#define GEN9_LP_FEATURES \
- GEN9_FEATURES, \
- .is_broxton = 1, \
+ GEN8_FEATURES, \
+ GEN9_HW_INFO, \
.gt = 1, \
.has_llc = false, \
.num_slices = 1, \
+ .num_thread_per_eu = 6, \
.max_vs_threads = 112, \
.max_tcs_threads = 112, \
.max_tes_threads = 112, \
.max_gs_threads = 112, \
.max_cs_threads = 6 * 6, \
- .timebase_scale = 1000000000.0 / 19200123.0, \
+ .timestamp_frequency = 19200000, \
.urb = { \
.size = 192, \
.min_entries = { \
}, \
}
+#define GEN9_LP_FEATURES_3X6 \
+ GEN9_LP_FEATURES, \
+ .num_subslices = { 3, }
+
#define GEN9_LP_FEATURES_2X6 \
GEN9_LP_FEATURES, \
+ .num_subslices = { 2, }, \
.max_vs_threads = 56, \
.max_tcs_threads = 56, \
.max_tes_threads = 56, \
}, \
}
+#define GEN9_FEATURES \
+ GEN8_FEATURES, \
+ GEN9_HW_INFO, \
+ .num_thread_per_eu = 7
+
static const struct gen_device_info gen_device_info_skl_gt1 = {
GEN9_FEATURES, .gt = 1,
+ .is_skylake = true,
.num_slices = 1,
+ .num_subslices = { 2, },
+ .l3_banks = 2,
.urb.size = 192,
};
static const struct gen_device_info gen_device_info_skl_gt2 = {
GEN9_FEATURES, .gt = 2,
+ .is_skylake = true,
.num_slices = 1,
+ .num_subslices = { 3, },
+ .l3_banks = 4,
};
static const struct gen_device_info gen_device_info_skl_gt3 = {
GEN9_FEATURES, .gt = 3,
+ .is_skylake = true,
.num_slices = 2,
+ .num_subslices = { 3, 3, },
+ .l3_banks = 8,
};
static const struct gen_device_info gen_device_info_skl_gt4 = {
GEN9_FEATURES, .gt = 4,
+ .is_skylake = true,
.num_slices = 3,
+ .num_subslices = { 3, 3, 3, },
+ .l3_banks = 12,
/* From the "L3 Allocation and Programming" documentation:
*
* "URB is limited to 1008KB due to programming restrictions. This is not a
};
static const struct gen_device_info gen_device_info_bxt = {
- GEN9_LP_FEATURES
+ GEN9_LP_FEATURES_3X6,
+ .is_broxton = true,
+ .l3_banks = 2,
};
static const struct gen_device_info gen_device_info_bxt_2x6 = {
- GEN9_LP_FEATURES_2X6
+ GEN9_LP_FEATURES_2X6,
+ .is_broxton = true,
+ .l3_banks = 1,
};
/*
* Note: for all KBL SKUs, the PRM says SKL for GS entries, not SKL+.
.max_cs_threads = 7 * 6,
.urb.size = 192,
.num_slices = 1,
+ .num_subslices = { 2, },
+ .l3_banks = 2,
};
static const struct gen_device_info gen_device_info_kbl_gt1_5 = {
.max_cs_threads = 7 * 6,
.num_slices = 1,
+ .num_subslices = { 3, },
+ .l3_banks = 4,
};
static const struct gen_device_info gen_device_info_kbl_gt2 = {
.gt = 2,
.num_slices = 1,
+ .num_subslices = { 3, },
+ .l3_banks = 4,
};
static const struct gen_device_info gen_device_info_kbl_gt3 = {
.gt = 3,
.num_slices = 2,
+ .num_subslices = { 3, 3, },
+ .l3_banks = 8,
};
static const struct gen_device_info gen_device_info_kbl_gt4 = {
*/
.urb.size = 1008 / 3,
.num_slices = 3,
+ .num_subslices = { 3, 3, 3, },
+ .l3_banks = 12,
};
static const struct gen_device_info gen_device_info_glk = {
- GEN9_LP_FEATURES
+ GEN9_LP_FEATURES_3X6,
+ .is_geminilake = true,
+ .l3_banks = 2,
};
+/*TODO: Initialize l3_banks when we know the number. */
static const struct gen_device_info gen_device_info_glk_2x6 = {
- GEN9_LP_FEATURES_2X6
+ GEN9_LP_FEATURES_2X6,
+ .is_geminilake = true,
+};
+
+static const struct gen_device_info gen_device_info_cfl_gt1 = {
+ GEN9_FEATURES,
+ .is_coffeelake = true,
+ .gt = 1,
+
+ .num_slices = 1,
+ .num_subslices = { 2, },
+ .l3_banks = 2,
+};
+static const struct gen_device_info gen_device_info_cfl_gt2 = {
+ GEN9_FEATURES,
+ .is_coffeelake = true,
+ .gt = 2,
+
+ .num_slices = 1,
+ .num_subslices = { 3, },
+ .l3_banks = 4,
+};
+
+static const struct gen_device_info gen_device_info_cfl_gt3 = {
+ GEN9_FEATURES,
+ .is_coffeelake = true,
+ .gt = 3,
+
+ .num_slices = 2,
+ .num_subslices = { 3, 3, },
+ .l3_banks = 8,
+};
+
+#define GEN10_HW_INFO \
+ .gen = 10, \
+ .num_thread_per_eu = 7, \
+ .max_vs_threads = 728, \
+ .max_gs_threads = 432, \
+ .max_tcs_threads = 432, \
+ .max_tes_threads = 624, \
+ .max_cs_threads = 56, \
+ .timestamp_frequency = 19200000, \
+ .urb = { \
+ .size = 256, \
+ .min_entries = { \
+ [MESA_SHADER_VERTEX] = 64, \
+ [MESA_SHADER_TESS_EVAL] = 34, \
+ }, \
+ .max_entries = { \
+ [MESA_SHADER_VERTEX] = 3936, \
+ [MESA_SHADER_TESS_CTRL] = 896, \
+ [MESA_SHADER_TESS_EVAL] = 2064, \
+ [MESA_SHADER_GEOMETRY] = 832, \
+ }, \
+ }
+
+#define subslices(args...) { args, }
+
+#define GEN10_FEATURES(_gt, _slices, _subslices, _l3) \
+ GEN8_FEATURES, \
+ GEN10_HW_INFO, \
+ .gt = _gt, \
+ .num_slices = _slices, \
+ .num_subslices = _subslices, \
+ .l3_banks = _l3
+
+static const struct gen_device_info gen_device_info_cnl_2x8 = {
+ /* GT0.5 */
+ GEN10_FEATURES(1, 1, subslices(2), 2),
+ .is_cannonlake = true,
+};
+
+static const struct gen_device_info gen_device_info_cnl_3x8 = {
+ /* GT1 */
+ GEN10_FEATURES(1, 1, subslices(3), 3),
+ .is_cannonlake = true,
+};
+
+static const struct gen_device_info gen_device_info_cnl_4x8 = {
+ /* GT 1.5 */
+ GEN10_FEATURES(1, 2, subslices(2, 2), 6),
+ .is_cannonlake = true,
+};
+
+static const struct gen_device_info gen_device_info_cnl_5x8 = {
+ /* GT2 */
+ GEN10_FEATURES(2, 2, subslices(3, 2), 6),
+ .is_cannonlake = true,
};
bool
* 4; /* effective subslices per slice */
}
+ assert(devinfo->num_slices <= ARRAY_SIZE(devinfo->num_subslices));
+
return true;
}