Move IP version specific code into a callback.
Also add support for gfx7 devices.
Signed-off-by: Tom St Denis <tom.stdenis@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
/* get the gpu clock counter */
uint64_t (*get_gpu_clock_counter)(struct amdgpu_device *adev);
void (*select_se_sh)(struct amdgpu_device *adev, u32 se_num, u32 sh_num, u32 instance);
+ void (*read_wave_data)(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields);
};
struct amdgpu_gfx {
return !r ? 4 : r;
}
-static uint32_t wave_read_ind(struct amdgpu_device *adev, uint32_t SQ_INDEX, uint32_t SQ_DATA, uint32_t simd, uint32_t wave, uint32_t address)
-{
- WREG32(SQ_INDEX, (wave & 0xF) | ((simd & 0x3) << 4) | (address << 16) | (1 << 13));
- return RREG32(SQ_DATA);
-}
-
static ssize_t amdgpu_debugfs_wave_read(struct file *f, char __user *buf,
size_t size, loff_t *pos)
{
struct amdgpu_device *adev = f->f_inode->i_private;
int r, x;
ssize_t result=0;
- uint32_t offset, se, sh, cu, wave, simd, data[16];
+ uint32_t offset, se, sh, cu, wave, simd, data[32];
if (size & 3 || *pos & 3)
return -EINVAL;
cu = ((*pos >> 23) & 0xFF);
wave = ((*pos >> 31) & 0xFF);
simd = ((*pos >> 37) & 0xFF);
- *pos &= 0x7F;
/* switch to the specific se/sh/cu */
mutex_lock(&adev->grbm_idx_mutex);
amdgpu_gfx_select_se_sh(adev, se, sh, cu);
x = 0;
- if (adev->family == AMDGPU_FAMILY_CZ || adev->family == AMDGPU_FAMILY_VI) {
- /* type 0 wave data */
- data[x++] = 0;
- data[x++] = wave_read_ind(adev, 0x2378, 0x2379, simd, wave, 0x12);
- data[x++] = wave_read_ind(adev, 0x2378, 0x2379, simd, wave, 0x18);
- data[x++] = wave_read_ind(adev, 0x2378, 0x2379, simd, wave, 0x19);
- data[x++] = wave_read_ind(adev, 0x2378, 0x2379, simd, wave, 0x27E);
- data[x++] = wave_read_ind(adev, 0x2378, 0x2379, simd, wave, 0x27F);
- data[x++] = wave_read_ind(adev, 0x2378, 0x2379, simd, wave, 0x14);
- data[x++] = wave_read_ind(adev, 0x2378, 0x2379, simd, wave, 0x1A);
- data[x++] = wave_read_ind(adev, 0x2378, 0x2379, simd, wave, 0x1B);
- }
+ if (adev->gfx.funcs->read_wave_data)
+ adev->gfx.funcs->read_wave_data(adev, simd, wave, data, &x);
amdgpu_gfx_select_se_sh(adev, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF);
mutex_unlock(&adev->grbm_idx_mutex);
if (!x)
return -EINVAL;
- while (size && (*pos < x * 4)) {
+ while (size && (offset < x * 4)) {
uint32_t value;
- value = data[*pos >> 2];
+ value = data[offset >> 2];
r = put_user(value, (uint32_t *)buf);
if (r)
return r;
result += 4;
buf += 4;
- *pos += 4;
+ offset += 4;
size -= 4;
}
amdgpu_ring_write(ring, (1 << (oa_size + oa_base)) - (1 << oa_base));
}
+static uint32_t wave_read_ind(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t address)
+{
+ WREG32(mmSQ_IND_INDEX, (wave & 0xF) | ((simd & 0x3) << 4) | (address << 16) | (1 << 13));
+ return RREG32(mmSQ_IND_DATA);
+}
+
+static void gfx_v7_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields)
+{
+ /* type 0 wave data */
+ dst[(*no_fields)++] = 0;
+ dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_STATUS);
+ dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_PC_LO);
+ dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_PC_HI);
+ dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_EXEC_LO);
+ dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_EXEC_HI);
+ dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_HW_ID);
+ dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_INST_DW0);
+ dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_INST_DW1);
+ dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_GPR_ALLOC);
+ dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_LDS_ALLOC);
+ dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_TRAPSTS);
+ dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_IB_STS);
+}
+
static const struct amdgpu_gfx_funcs gfx_v7_0_gfx_funcs = {
.get_gpu_clock_counter = &gfx_v7_0_get_gpu_clock_counter,
.select_se_sh = &gfx_v7_0_select_se_sh,
+ .read_wave_data = &gfx_v7_0_read_wave_data,
};
static const struct amdgpu_rlc_funcs gfx_v7_0_rlc_funcs = {
amdgpu_ring_write(ring, (1 << (oa_size + oa_base)) - (1 << oa_base));
}
+static uint32_t wave_read_ind(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t address)
+{
+ WREG32(mmSQ_IND_INDEX, (wave & 0xF) | ((simd & 0x3) << 4) | (address << 16) | (1 << 13));
+ return RREG32(mmSQ_IND_DATA);
+}
+
+static void gfx_v8_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields)
+{
+ /* type 0 wave data */
+ dst[(*no_fields)++] = 0;
+ dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_STATUS);
+ dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_PC_LO);
+ dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_PC_HI);
+ dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_EXEC_LO);
+ dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_EXEC_HI);
+ dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_HW_ID);
+ dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_INST_DW0);
+ dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_INST_DW1);
+ dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_GPR_ALLOC);
+ dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_LDS_ALLOC);
+ dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_TRAPSTS);
+ dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_IB_STS);
+}
+
+
static const struct amdgpu_gfx_funcs gfx_v8_0_gfx_funcs = {
.get_gpu_clock_counter = &gfx_v8_0_get_gpu_clock_counter,
.select_se_sh = &gfx_v8_0_select_se_sh,
+ .read_wave_data = &gfx_v8_0_read_wave_data,
};
static int gfx_v8_0_early_init(void *handle)