|
@@ -3224,6 +3224,31 @@ static void gfx_v9_ring_emit_cntxcntl(struct amdgpu_ring *ring, uint32_t flags)
|
|
|
amdgpu_ring_write(ring, 0);
|
|
|
}
|
|
|
|
|
|
+static unsigned gfx_v9_0_ring_emit_init_cond_exec(struct amdgpu_ring *ring)
|
|
|
+{
|
|
|
+ unsigned ret;
|
|
|
+ amdgpu_ring_write(ring, PACKET3(PACKET3_COND_EXEC, 3));
|
|
|
+ amdgpu_ring_write(ring, lower_32_bits(ring->cond_exe_gpu_addr));
|
|
|
+ amdgpu_ring_write(ring, upper_32_bits(ring->cond_exe_gpu_addr));
|
|
|
+ amdgpu_ring_write(ring, 0); /* discard following DWs if *cond_exec_gpu_addr==0 */
|
|
|
+ ret = ring->wptr & ring->buf_mask;
|
|
|
+ amdgpu_ring_write(ring, 0x55aa55aa); /* patch dummy value later */
|
|
|
+ return ret;
|
|
|
+}
|
|
|
+
|
|
|
+static void gfx_v9_0_ring_emit_patch_cond_exec(struct amdgpu_ring *ring, unsigned offset)
|
|
|
+{
|
|
|
+ unsigned cur;
|
|
|
+ BUG_ON(offset > ring->buf_mask);
|
|
|
+ BUG_ON(ring->ring[offset] != 0x55aa55aa);
|
|
|
+
|
|
|
+ cur = (ring->wptr & ring->buf_mask) - 1;
|
|
|
+ if (likely(cur > offset))
|
|
|
+ ring->ring[offset] = cur - offset;
|
|
|
+ else
|
|
|
+ ring->ring[offset] = (ring->ring_size>>2) - offset + cur;
|
|
|
+}
|
|
|
+
|
|
|
static void gfx_v9_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg)
|
|
|
{
|
|
|
struct amdgpu_device *adev = ring->adev;
|
|
@@ -3575,6 +3600,8 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_gfx = {
|
|
|
.pad_ib = amdgpu_ring_generic_pad_ib,
|
|
|
.emit_switch_buffer = gfx_v9_ring_emit_sb,
|
|
|
.emit_cntxcntl = gfx_v9_ring_emit_cntxcntl,
|
|
|
+ .init_cond_exec = gfx_v9_0_ring_emit_init_cond_exec,
|
|
|
+ .patch_cond_exec = gfx_v9_0_ring_emit_patch_cond_exec,
|
|
|
};
|
|
|
|
|
|
static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_compute = {
|