diff options
Diffstat (limited to 'meta-amd-bsp/recipes-kernel/linux/linux-yocto-4.19.8/2927-drm-amdkfd-Support-bigger-gds-size.patch')
-rw-r--r-- | meta-amd-bsp/recipes-kernel/linux/linux-yocto-4.19.8/2927-drm-amdkfd-Support-bigger-gds-size.patch | 107 |
1 files changed, 107 insertions, 0 deletions
diff --git a/meta-amd-bsp/recipes-kernel/linux/linux-yocto-4.19.8/2927-drm-amdkfd-Support-bigger-gds-size.patch b/meta-amd-bsp/recipes-kernel/linux/linux-yocto-4.19.8/2927-drm-amdkfd-Support-bigger-gds-size.patch new file mode 100644 index 00000000..ce64f3f7 --- /dev/null +++ b/meta-amd-bsp/recipes-kernel/linux/linux-yocto-4.19.8/2927-drm-amdkfd-Support-bigger-gds-size.patch @@ -0,0 +1,107 @@ +From 3c99869733cb47b04d809625959d1094e861c29d Mon Sep 17 00:00:00 2001 +From: Oak Zeng <Oak.Zeng@amd.com> +Date: Tue, 19 Feb 2019 14:59:51 -0600 +Subject: [PATCH 2927/2940] drm/amdkfd: Support bigger gds size + +Extend map_process and set_resources pm4 packet to support +bigger gds size for arcturus. + +v2: Only make the change for v9 + +Signed-off-by: Oak Zeng <Oak.Zeng@amd.com> +Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com> +Signed-off-by: Alex Deucher <alexander.deucher@amd.com> +--- + .../gpu/drm/amd/amdkfd/kfd_kernel_queue_v9.c | 33 +++++++++++++++++-- + .../gpu/drm/amd/amdkfd/kfd_pm4_headers_ai.h | 10 +++--- + 2 files changed, 36 insertions(+), 7 deletions(-) + +diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue_v9.c +index 6a4332f4a3fb..c3f39ef4de56 100644 +--- a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue_v9.c ++++ b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue_v9.c +@@ -82,7 +82,8 @@ static int pm_map_process_v9(struct packet_manager *pm, + packet->bitfields2.diq_enable = (qpd->is_debug) ? 1 : 0; + packet->bitfields2.process_quantum = 1; + packet->bitfields2.pasid = qpd->pqm->process->pasid; +- packet->bitfields14.gds_size = qpd->gds_size; ++ packet->bitfields14.gds_size = qpd->gds_size & 0x3F; ++ packet->bitfields14.gds_size_hi = (qpd->gds_size >> 6) & 0xF; + packet->bitfields14.num_gws = qpd->num_gws; + packet->bitfields14.num_oac = qpd->num_oac; + packet->bitfields14.sdma_enable = 1; +@@ -149,6 +150,34 @@ static int pm_runlist_v9(struct packet_manager *pm, uint32_t *buffer, + return 0; + } + ++static int pm_set_resources_v9(struct packet_manager *pm, uint32_t *buffer, ++ struct scheduling_resources *res) ++{ ++ struct pm4_mes_set_resources *packet; ++ ++ packet = (struct pm4_mes_set_resources *)buffer; ++ memset(buffer, 0, sizeof(struct pm4_mes_set_resources)); ++ ++ packet->header.u32All = pm_build_pm4_header(IT_SET_RESOURCES, ++ sizeof(struct pm4_mes_set_resources)); ++ ++ packet->bitfields2.queue_type = ++ queue_type__mes_set_resources__hsa_interface_queue_hiq; ++ packet->bitfields2.vmid_mask = res->vmid_mask; ++ packet->bitfields2.unmap_latency = KFD_UNMAP_LATENCY_MS / 100; ++ packet->bitfields7.oac_mask = res->oac_mask; ++ packet->bitfields8.gds_heap_base = res->gds_heap_base; ++ packet->bitfields8.gds_heap_size = res->gds_heap_size; ++ ++ packet->gws_mask_lo = lower_32_bits(res->gws_mask); ++ packet->gws_mask_hi = upper_32_bits(res->gws_mask); ++ ++ packet->queue_mask_lo = lower_32_bits(res->queue_mask); ++ packet->queue_mask_hi = upper_32_bits(res->queue_mask); ++ ++ return 0; ++} ++ + static int pm_map_queues_v9(struct packet_manager *pm, uint32_t *buffer, + struct queue *q, bool is_static) + { +@@ -385,7 +414,7 @@ static int pm_release_mem_v9(uint64_t gpu_addr, uint32_t *buffer) + const struct packet_manager_funcs kfd_v9_pm_funcs = { + .map_process = pm_map_process_v9, + .runlist = pm_runlist_v9, +- .set_resources = pm_set_resources_vi, ++ .set_resources = pm_set_resources_v9, + .map_queues = pm_map_queues_v9, + .unmap_queues = pm_unmap_queues_v9, + .set_grace_period = pm_set_grace_period_v9, +diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_ai.h b/drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_ai.h +index 249cfa4312a8..141d25607a50 100644 +--- a/drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_ai.h ++++ b/drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_ai.h +@@ -83,10 +83,10 @@ struct pm4_mes_set_resources { + + union { + struct { +- uint32_t gds_heap_base:6; +- uint32_t reserved3:5; +- uint32_t gds_heap_size:6; +- uint32_t reserved4:15; ++ uint32_t gds_heap_base:10; ++ uint32_t reserved3:1; ++ uint32_t gds_heap_size:10; ++ uint32_t reserved4:11; + } bitfields8; + uint32_t ordinal8; + }; +@@ -182,7 +182,7 @@ struct pm4_mes_map_process { + uint32_t num_gws:7; + uint32_t sdma_enable:1; + uint32_t num_oac:4; +- uint32_t reserved8:4; ++ uint32_t gds_size_hi:4; + uint32_t gds_size:6; + uint32_t num_queues:10; + } bitfields14; +-- +2.17.1 + |