nv-tegra-mirror/linux-nvgpu
1
0
Fork 0
mirror of git://nv-tegra.nvidia.com/linux-nvgpu.git synced 2025-12-23 01:50:07 +03:00
Code Issues Packages Projects Releases Wiki Activity

gpu: nvgpu: Get rid of legacy gpfifo type

Browse Source 
Get rid of the duplicate gpfifo struct to emphasize the fact that
nvgpu_gpfifo is the only memory layout for gpfifo entries that works.
This is the same layout that HW uses.

Also, add a local pointer to the gpfifo memory in
gk20a_submit_channel_gpfifo to get rid of repeated typecasts.

Bug 1592391
Bug 1550886

Change-Id: I5432859ef8e7c1aab5907e44098994d7bb807f50
Signed-off-by: Janne Hellsten <jhellsten@nvidia.com>
Signed-off-by: Konsta Holtta <kholtta@nvidia.com>
Reviewed-on: http://git-master/r/677341
(cherry picked from commit 724c8c6228af81dd440e825bddf545dd6b2b8bd7)
Reviewed-on: http://git-master/r/822548
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: Arto Merilainen <amerilainen@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
This commit is contained in:
Janne Hellsten
2015-01-26 17:27:42 +02:00
committed by Terje Bergstrom
parent 5c26f72fe1
commit 057c6334f7
3 changed files with 11 additions and 17 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
drivers/gpu/nvgpu/gk20a/channel_gk20a.c
View File

@@ -1381,7 +1381,8 @@ int gk20a_alloc_channel_gpfifo(struct channel_gk20a *c,
return -EEXIST; return -EEXIST;
} }
err = gk20a_gmmu_alloc_map(ch_vm, gpfifo_size * sizeof(struct gpfifo), err = gk20a_gmmu_alloc_map(ch_vm,
gpfifo_size * sizeof(struct nvgpu_gpfifo),
&c->gpfifo.mem); &c->gpfifo.mem);
if (err) { if (err) {
gk20a_err(d, "%s: memory allocation failed\n", __func__); gk20a_err(d, "%s: memory allocation failed\n", __func__);
@@ -1812,6 +1813,7 @@ int gk20a_submit_channel_gpfifo(struct channel_gk20a *c,
* and one for post fence. */ * and one for post fence. */
const int extra_entries = 2; const int extra_entries = 2;
bool need_wfi = !(flags & NVGPU_SUBMIT_GPFIFO_FLAGS_SUPPRESS_WFI); bool need_wfi = !(flags & NVGPU_SUBMIT_GPFIFO_FLAGS_SUPPRESS_WFI);
struct nvgpu_gpfifo *gpfifo_mem = c->gpfifo.mem.cpu_va;
if (c->has_timedout) if (c->has_timedout)
return -ETIMEDOUT; return -ETIMEDOUT;
@@ -1948,10 +1950,8 @@ int gk20a_submit_channel_gpfifo(struct channel_gk20a *c,
} }
if (wait_cmd) { if (wait_cmd) {
((struct gpfifo *)(c->gpfifo.mem.cpu_va))[c->gpfifo.put].entry0 = gpfifo_mem[c->gpfifo.put].entry0 = u64_lo32(wait_cmd->gva);
u64_lo32(wait_cmd->gva); gpfifo_mem[c->gpfifo.put].entry1 = u64_hi32(wait_cmd->gva) |
((struct gpfifo *)(c->gpfifo.mem.cpu_va))[c->gpfifo.put].entry1 =
u64_hi32(wait_cmd->gva) |
pbdma_gp_entry1_length_f(wait_cmd->size); pbdma_gp_entry1_length_f(wait_cmd->size);
trace_gk20a_push_cmdbuf(c->g->dev->name, trace_gk20a_push_cmdbuf(c->g->dev->name,
0, wait_cmd->size, 0, wait_cmd->ptr); 0, wait_cmd->size, 0, wait_cmd->ptr);
@@ -1975,16 +1975,16 @@ int gk20a_submit_channel_gpfifo(struct channel_gk20a *c,
int length0 = c->gpfifo.entry_num - start; int length0 = c->gpfifo.entry_num - start;
int length1 = num_entries - length0; int length1 = num_entries - length0;
memcpy((struct gpfifo *)c->gpfifo.mem.cpu_va + start, gpfifo, memcpy(gpfifo_mem + start, gpfifo,
length0 * sizeof(*gpfifo)); length0 * sizeof(*gpfifo));
memcpy((struct gpfifo *)c->gpfifo.mem.cpu_va, gpfifo + length0, memcpy(gpfifo_mem, gpfifo + length0,
length1 * sizeof(*gpfifo)); length1 * sizeof(*gpfifo));
trace_write_pushbuffer_range(c, gpfifo, length0); trace_write_pushbuffer_range(c, gpfifo, length0);
trace_write_pushbuffer_range(c, gpfifo + length0, length1); trace_write_pushbuffer_range(c, gpfifo + length0, length1);
} else { } else {
memcpy((struct gpfifo *)c->gpfifo.mem.cpu_va + start, gpfifo, memcpy(gpfifo_mem + start, gpfifo,
num_entries * sizeof(*gpfifo)); num_entries * sizeof(*gpfifo));
trace_write_pushbuffer_range(c, gpfifo, num_entries); trace_write_pushbuffer_range(c, gpfifo, num_entries);
@@ -1993,10 +1993,8 @@ int gk20a_submit_channel_gpfifo(struct channel_gk20a *c,
(c->gpfifo.entry_num - 1); (c->gpfifo.entry_num - 1);
if (incr_cmd) { if (incr_cmd) {
((struct gpfifo *)(c->gpfifo.mem.cpu_va))[c->gpfifo.put].entry0 = gpfifo_mem[c->gpfifo.put].entry0 = u64_lo32(incr_cmd->gva);
u64_lo32(incr_cmd->gva); gpfifo_mem[c->gpfifo.put].entry1 = u64_hi32(incr_cmd->gva) |
((struct gpfifo *)(c->gpfifo.mem.cpu_va))[c->gpfifo.put].entry1 =
u64_hi32(incr_cmd->gva) |
pbdma_gp_entry1_length_f(incr_cmd->size); pbdma_gp_entry1_length_f(incr_cmd->size);
trace_gk20a_push_cmdbuf(c->g->dev->name, trace_gk20a_push_cmdbuf(c->g->dev->name,
0, incr_cmd->size, 0, incr_cmd->ptr); 0, incr_cmd->size, 0, incr_cmd->ptr);

5
drivers/gpu/nvgpu/gk20a/channel_gk20a.h
View File

@@ -38,11 +38,6 @@ struct gk20a_fence;
#include "gr_gk20a.h" #include "gr_gk20a.h"
#include "fence_gk20a.h" #include "fence_gk20a.h"
struct gpfifo {
u32 entry0;
u32 entry1;
};
struct notification { struct notification {
struct { struct {
u32 nanoseconds[2]; u32 nanoseconds[2];

1
include/uapi/linux/nvgpu.h
View File

@@ -563,6 +563,7 @@ struct nvgpu_dbg_gpu_pc_sampling_args {
#define NVGPU_PRIORITY_HIGH 150 #define NVGPU_PRIORITY_HIGH 150
#define NVGPU_TIMEOUT_FLAG_DISABLE_DUMP 0 #define NVGPU_TIMEOUT_FLAG_DISABLE_DUMP 0
/* this is also the hardware memory format */
struct nvgpu_gpfifo { struct nvgpu_gpfifo {
__u32 entry0; /* first word of gpfifo entry */ __u32 entry0; /* first word of gpfifo entry */
__u32 entry1; /* second word of gpfifo entry */ __u32 entry1; /* second word of gpfifo entry */