[igt-dev] [PATCH] tests/amdgpu: vcn test add session context buffer
vitaly.prosyak at amd.com
vitaly.prosyak at amd.com
Wed Jul 19 17:09:15 UTC 2023
From: Vitaly Prosyak <vitaly.prosyak at amd.com>
Porting libdrm vcn test updates:
1. Rename vbv_buffer to input buffer.
2. Update decoder test with using session context buffer
3. Added helper function amdgpu_cs_vcn_ib_zero_count.
Cc: Leo Liu <leo.liu at amd.com>
Cc: Ruijing Dong <ruijing.dong at amd.com>
Cc: Saleemkhan Jamadar <saleemkhan.jamadar at amd.com>
Signed-off-by: Saleemkhan Jamadar <saleemkhan.jamadar at amd.com>
Signed-off-by: Ruijing Dong <ruijing.dong at amd.com>
Signed-off-by: Vitaly Prosyak <vitaly.prosyak at amd.com>
---
tests/amdgpu/amd_vcn.c | 83 ++++++++++++++++++++++++++----------------
1 file changed, 51 insertions(+), 32 deletions(-)
diff --git a/tests/amdgpu/amd_vcn.c b/tests/amdgpu/amd_vcn.c
index aaf216fb9..bd8c9da23 100644
--- a/tests/amdgpu/amd_vcn.c
+++ b/tests/amdgpu/amd_vcn.c
@@ -115,6 +115,7 @@ struct h264_decode {
struct vcn_context {
struct amdgpu_mmd_bo enc_buf;
struct amdgpu_mmd_bo cpb_buf;
+ struct amdgpu_mmd_bo session_ctx_buf;
uint32_t enc_task_id;
uint32_t *ib_checksum;
uint32_t *ib_size_in_dw;
@@ -327,7 +328,9 @@ amdgpu_cs_vcn_dec_create(amdgpu_device_handle device_handle, struct mmd_context
context->num_resources = 0;
alloc_resource(device_handle, &msg_buf, 4096, AMDGPU_GEM_DOMAIN_GTT);
+ alloc_resource(device_handle, &v_context->session_ctx_buf, 32 * 4096, AMDGPU_GEM_DOMAIN_GTT);
context->resources[context->num_resources++] = msg_buf.handle;
+ context->resources[context->num_resources++] = v_context->session_ctx_buf.handle;
context->resources[context->num_resources++] = context->ib_handle;
r = amdgpu_bo_cpu_map(msg_buf.handle, (void **)&msg_buf.ptr);
igt_assert_eq(r, 0);
@@ -336,9 +339,10 @@ amdgpu_cs_vcn_dec_create(amdgpu_device_handle device_handle, struct mmd_context
memcpy(msg_buf.ptr, vcn_dec_create_msg, sizeof(vcn_dec_create_msg));
len = 0;
- if (context->vcn_dec_sw_ring == true)
+ vcn_dec_cmd(context, v_context, v_context->session_ctx_buf.addr, 5, &len);
+ if (context->vcn_dec_sw_ring == true) {
vcn_dec_cmd(context, v_context, msg_buf.addr, 0, &len);
- else {
+ } else {
context->ib_cpu[len++] = reg[context->vcn_reg_index].data0;
context->ib_cpu[len++] = msg_buf.addr;
context->ib_cpu[len++] = reg[context->vcn_reg_index].data1;
@@ -350,7 +354,6 @@ amdgpu_cs_vcn_dec_create(amdgpu_device_handle device_handle, struct mmd_context
context->ib_cpu[len++] = 0;
}
}
-
if (context->vcn_unified_ring) {
amdgpu_cs_sq_ib_tail(v_context, context->ib_cpu + len);
ip = AMDGPU_HW_IP_VCN_ENC;
@@ -382,6 +385,7 @@ amdgpu_cs_vcn_dec_decode(amdgpu_device_handle device_handle, struct mmd_context
size += ALIGN(dpb_size, 4*1024);
size += ALIGN(dt_size, 4*1024);
+
context->num_resources = 0;
alloc_resource(device_handle, &dec_buf, size, AMDGPU_GEM_DOMAIN_GTT);
context->resources[context->num_resources++] = dec_buf.handle;
@@ -416,7 +420,9 @@ amdgpu_cs_vcn_dec_decode(amdgpu_device_handle device_handle, struct mmd_context
ctx_addr = ALIGN(dpb_addr + 0x006B9400, 4*1024);
dt_addr = ALIGN(dpb_addr + dpb_size, 4*1024);
+
len = 0;
+ vcn_dec_cmd(context, v_context, v_context->session_ctx_buf.addr, 0x5, &len);
vcn_dec_cmd(context, v_context, msg_addr, 0x0, &len);
vcn_dec_cmd(context, v_context, dpb_addr, 0x1, &len);
vcn_dec_cmd(context, v_context, dt_addr, 0x2, &len);
@@ -471,9 +477,10 @@ amdgpu_cs_vcn_dec_destroy(amdgpu_device_handle device_handle,
memcpy(msg_buf.ptr, vcn_dec_destroy_msg, sizeof(vcn_dec_destroy_msg));
len = 0;
- if (context->vcn_dec_sw_ring == true)
+ vcn_dec_cmd(context, v_context, v_context->session_ctx_buf.addr, 5, &len);
+ if (context->vcn_dec_sw_ring == true) {
vcn_dec_cmd(context, v_context, msg_buf.addr, 0, &len);
- else {
+ } else {
context->ib_cpu[len++] = reg[context->vcn_reg_index].data0;
context->ib_cpu[len++] = msg_buf.addr;
context->ib_cpu[len++] = reg[context->vcn_reg_index].data1;
@@ -485,7 +492,6 @@ amdgpu_cs_vcn_dec_destroy(amdgpu_device_handle device_handle,
context->ib_cpu[len++] = 0;
}
}
-
if (context->vcn_unified_ring) {
amdgpu_cs_sq_ib_tail(v_context, context->ib_cpu + len);
ip = AMDGPU_HW_IP_VCN_ENC;
@@ -496,6 +502,7 @@ amdgpu_cs_vcn_dec_destroy(amdgpu_device_handle device_handle,
igt_assert_eq(r, 0);
free_resource(&msg_buf);
+ free_resource(&v_context->session_ctx_buf);
}
static void
@@ -579,6 +586,8 @@ amdgpu_cs_vcn_enc_create(amdgpu_device_handle device_handle,
context->ib_cpu[len++] = 0;
context->ib_cpu[len++] = 0; /* pre encode mode */
context->ib_cpu[len++] = 0; /* chroma enabled : false */
+ context->ib_cpu[len++] = 0;
+ context->ib_cpu[len++] = 0;
*st_size = (len - st_offset) * 4;
/* slice control */
@@ -600,7 +609,7 @@ amdgpu_cs_vcn_enc_create(amdgpu_device_handle device_handle,
context->ib_cpu[len++] = 1; /* quarter pel enabled */
context->ib_cpu[len++] = 100; /* BASELINE profile */
context->ib_cpu[len++] = 11; /* level */
- if (context->vcn_ip_version_major == 3) {
+ if (context->vcn_ip_version_major >= 3) {
context->ib_cpu[len++] = 0; /* b_picture_enabled */
context->ib_cpu[len++] = 0; /* weighted_bipred_idc */
}
@@ -641,7 +650,7 @@ amdgpu_cs_vcn_enc_create(amdgpu_device_handle device_handle,
context->ib_cpu[len++] = 0; /* scene change sensitivity */
context->ib_cpu[len++] = 0; /* scene change min idr interval */
context->ib_cpu[len++] = 0;
- if (context->vcn_ip_version_major == 3)
+ if (context->vcn_ip_version_major >= 3)
context->ib_cpu[len++] = 0;
*st_size = (len - st_offset) * 4;
@@ -684,6 +693,7 @@ amdgpu_cs_vcn_enc_create(amdgpu_device_handle device_handle,
context->ib_cpu[len++] = 1;
context->ib_cpu[len++] = 0;
context->ib_cpu[len++] = 1;
+ context->ib_cpu[len++] = 0;
*st_size = (len - st_offset) * 4;
/* op init rc */
@@ -707,6 +717,15 @@ amdgpu_cs_vcn_enc_create(amdgpu_device_handle device_handle,
igt_assert_eq(r, 0);
}
+static void
+amdgpu_cs_vcn_ib_zero_count(struct mmd_context *context, int *len, int num)
+{
+ int i;
+
+ for (i = 0; i < num; i++)
+ context->ib_cpu[(*len)++] = 0;
+}
+
static int32_t
h264_se(struct buffer_info *buf_info)
{
@@ -1046,7 +1065,7 @@ amdgpu_cs_vcn_enc_encode_frame(amdgpu_device_handle device_handle,
struct mmd_context *context, struct vcn_context *v_context,
int frame_type)
{
- struct amdgpu_mmd_bo bs_buf, fb_buf, vbv_buf;
+ struct amdgpu_mmd_bo bs_buf, fb_buf, input_buf;
int len, r, i;
unsigned int width = 160, height = 128, buf_size;
uint32_t *p_task_size = NULL;
@@ -1067,12 +1086,12 @@ amdgpu_cs_vcn_enc_encode_frame(amdgpu_device_handle device_handle,
context->num_resources = 0;
alloc_resource(device_handle, &bs_buf, 4096, AMDGPU_GEM_DOMAIN_GTT);
alloc_resource(device_handle, &fb_buf, 4096, AMDGPU_GEM_DOMAIN_GTT);
- alloc_resource(device_handle, &vbv_buf, buf_size, AMDGPU_GEM_DOMAIN_GTT);
+ alloc_resource(device_handle, &input_buf, buf_size, AMDGPU_GEM_DOMAIN_GTT);
context->resources[context->num_resources++] = v_context->enc_buf.handle;
context->resources[context->num_resources++] = v_context->cpb_buf.handle;
context->resources[context->num_resources++] = bs_buf.handle;
context->resources[context->num_resources++] = fb_buf.handle;
- context->resources[context->num_resources++] = vbv_buf.handle;
+ context->resources[context->num_resources++] = input_buf.handle;
context->resources[context->num_resources++] = context->ib_handle;
@@ -1084,13 +1103,13 @@ amdgpu_cs_vcn_enc_encode_frame(amdgpu_device_handle device_handle,
memset(fb_buf.ptr, 0, 4096);
r = amdgpu_bo_cpu_unmap(fb_buf.handle);
- r = amdgpu_bo_cpu_map(vbv_buf.handle, (void **)&vbv_buf.ptr);
+ r = amdgpu_bo_cpu_map(input_buf.handle, (void **)&input_buf.ptr);
igt_assert_eq(r, 0);
for (i = 0; i < ALIGN(height, 32) * 3 / 2; i++)
- memcpy(vbv_buf.ptr + i * ALIGN(width, 256), frame + i * width, width);
+ memcpy(input_buf.ptr + i * ALIGN(width, 256), frame + i * width, width);
- r = amdgpu_bo_cpu_unmap(vbv_buf.handle);
+ r = amdgpu_bo_cpu_unmap(input_buf.handle);
igt_assert_eq(r, 0);
len = 0;
@@ -1125,7 +1144,7 @@ amdgpu_cs_vcn_enc_encode_frame(amdgpu_device_handle device_handle,
if (context->vcn_ip_version_major == 1)
context->ib_cpu[len++] = 0x00000020; /* RENCODE_IB_PARAM_DIRECT_OUTPUT_NALU vcn 1 */
else
- context->ib_cpu[len++] = 0x0000000a; /* RENCODE_IB_PARAM_DIRECT_OUTPUT_NALU vcn 2,3 */
+ context->ib_cpu[len++] = 0x0000000a; /* RENCODE_IB_PARAM_DIRECT_OUTPUT_NALU other vcn */
context->ib_cpu[len++] = 0x00000002; /* RENCODE_DIRECT_OUTPUT_NALU_TYPE_SPS */
context->ib_cpu[len++] = 0x00000011; /* sps len */
context->ib_cpu[len++] = 0x00000001; /* start code */
@@ -1141,7 +1160,7 @@ amdgpu_cs_vcn_enc_encode_frame(amdgpu_device_handle device_handle,
if (context->vcn_ip_version_major == 1)
context->ib_cpu[len++] = 0x00000020; /* RENCODE_IB_PARAM_DIRECT_OUTPUT_NALU vcn 1*/
else
- context->ib_cpu[len++] = 0x0000000a; /* RENCODE_IB_PARAM_DIRECT_OUTPUT_NALU vcn 2,3*/
+ context->ib_cpu[len++] = 0x0000000a; /* RENCODE_IB_PARAM_DIRECT_OUTPUT_NALU other vcn*/
context->ib_cpu[len++] = 0x00000003; /* RENCODE_DIRECT_OUTPUT_NALU_TYPE_PPS */
context->ib_cpu[len++] = 0x00000008; /* pps len */
context->ib_cpu[len++] = 0x00000001; /* start code */
@@ -1164,8 +1183,7 @@ amdgpu_cs_vcn_enc_encode_frame(amdgpu_device_handle device_handle,
context->ib_cpu[len++] = 0x34210000;
}
context->ib_cpu[len++] = 0xe0000000;
- for (i = 0; i < 13; i++)
- context->ib_cpu[len++] = 0x00000000;
+ amdgpu_cs_vcn_ib_zero_count(context, &len, 13);
context->ib_cpu[len++] = 0x00000001;
context->ib_cpu[len++] = 0x00000008;
@@ -1177,8 +1195,7 @@ amdgpu_cs_vcn_enc_encode_frame(amdgpu_device_handle device_handle,
context->ib_cpu[len++] = 0x00000000;
context->ib_cpu[len++] = 0x00000001;
context->ib_cpu[len++] = 0x00000003;
- for (i = 0; i < 22; i++)
- context->ib_cpu[len++] = 0x00000000;
+ amdgpu_cs_vcn_ib_zero_count(context, &len, 22);
*st_size = (len - st_offset) * 4;
@@ -1188,13 +1205,13 @@ amdgpu_cs_vcn_enc_encode_frame(amdgpu_device_handle device_handle,
if (context->vcn_ip_version_major == 1)
context->ib_cpu[len++] = 0x0000000b; /* RENCODE_IB_PARAM_ENCODE_PARAMS vcn 1*/
else
- context->ib_cpu[len++] = 0x0000000f; /* RENCODE_IB_PARAM_ENCODE_PARAMS vcn 2,3*/
+ context->ib_cpu[len++] = 0x0000000f; /* RENCODE_IB_PARAM_ENCODE_PARAMS other vcn*/
context->ib_cpu[len++] = frame_type;
context->ib_cpu[len++] = 0x0001f000;
- context->ib_cpu[len++] = vbv_buf.addr >> 32;
- context->ib_cpu[len++] = vbv_buf.addr;
- context->ib_cpu[len++] = (vbv_buf.addr + ALIGN(width, 256) * ALIGN(height, 32)) >> 32;
- context->ib_cpu[len++] = vbv_buf.addr + ALIGN(width, 256) * ALIGN(height, 32);
+ context->ib_cpu[len++] = input_buf.addr >> 32;
+ context->ib_cpu[len++] = input_buf.addr;
+ context->ib_cpu[len++] = (input_buf.addr + ALIGN(width, 256) * ALIGN(height, 32)) >> 32;
+ context->ib_cpu[len++] = input_buf.addr + ALIGN(width, 256) * ALIGN(height, 32);
context->ib_cpu[len++] = 0x00000100;
context->ib_cpu[len++] = 0x00000080;
context->ib_cpu[len++] = 0x00000000;
@@ -1206,7 +1223,7 @@ amdgpu_cs_vcn_enc_encode_frame(amdgpu_device_handle device_handle,
st_offset = len;
st_size = &context->ib_cpu[len++]; /* size */
context->ib_cpu[len++] = 0x00200003; /* RENCODE_H264_IB_PARAM_ENCODE_PARAMS */
- if (context->vcn_ip_version_major != 3) {
+ if (context->vcn_ip_version_major <= 2) {
context->ib_cpu[len++] = 0x00000000;
context->ib_cpu[len++] = 0x00000000;
context->ib_cpu[len++] = 0x00000000;
@@ -1229,6 +1246,7 @@ amdgpu_cs_vcn_enc_encode_frame(amdgpu_device_handle device_handle,
context->ib_cpu[len++] = 0x00000000;
context->ib_cpu[len++] = 0x00000000;
context->ib_cpu[len++] = 0x00000000;
+ context->ib_cpu[len++] = 0x00000001;
}
*st_size = (len - st_offset) * 4;
@@ -1238,20 +1256,21 @@ amdgpu_cs_vcn_enc_encode_frame(amdgpu_device_handle device_handle,
if (context->vcn_ip_version_major == 1)
context->ib_cpu[len++] = 0x0000000d; /* ENCODE_CONTEXT_BUFFER vcn 1 */
else
- context->ib_cpu[len++] = 0x00000011; /* ENCODE_CONTEXT_BUFFER vcn 2,3 */
+ context->ib_cpu[len++] = 0x00000011; /* ENCODE_CONTEXT_BUFFER other vcn*/
context->ib_cpu[len++] = v_context->cpb_buf.addr >> 32;
context->ib_cpu[len++] = v_context->cpb_buf.addr;
context->ib_cpu[len++] = 0x00000000; /* swizzle mode */
context->ib_cpu[len++] = 0x00000100; /* luma pitch */
context->ib_cpu[len++] = 0x00000100; /* chroma pitch */
- context->ib_cpu[len++] = 0x00000003; /* no reconstructed picture */
+ context->ib_cpu[len++] = 0x00000002; /* no reconstructed picture */
context->ib_cpu[len++] = 0x00000000; /* reconstructed pic 1 luma offset */
context->ib_cpu[len++] = ALIGN(width, 256) * ALIGN(height, 32); /* pic1 chroma offset */
+ if (context->vcn_ip_version_major == 4)
+ amdgpu_cs_vcn_ib_zero_count(context, &len, 2);
context->ib_cpu[len++] = ALIGN(width, 256) * ALIGN(height, 32) * 3 / 2; /* pic2 luma offset */
context->ib_cpu[len++] = ALIGN(width, 256) * ALIGN(height, 32) * 5 / 2; /* pic2 chroma offset */
- for (i = 0; i < 136; i++)
- context->ib_cpu[len++] = 0x00000000;
+ amdgpu_cs_vcn_ib_zero_count(context, &len, 280);
*st_size = (len - st_offset) * 4;
/* bitstream buffer */
@@ -1260,7 +1279,7 @@ amdgpu_cs_vcn_enc_encode_frame(amdgpu_device_handle device_handle,
if (context->vcn_ip_version_major == 1)
context->ib_cpu[len++] = 0x0000000e; /* VIDEO_BITSTREAM_BUFFER vcn 1 */
else
- context->ib_cpu[len++] = 0x00000012; /* VIDEO_BITSTREAM_BUFFER vcn 2,3 */
+ context->ib_cpu[len++] = 0x00000012; /* VIDEO_BITSTREAM_BUFFER other vcn */
context->ib_cpu[len++] = 0x00000000; /* mode */
context->ib_cpu[len++] = bs_buf.addr >> 32;
context->ib_cpu[len++] = bs_buf.addr;
@@ -1343,7 +1362,7 @@ amdgpu_cs_vcn_enc_encode_frame(amdgpu_device_handle device_handle,
free_resource(&fb_buf);
free_resource(&bs_buf);
- free_resource(&vbv_buf);
+ free_resource(&input_buf);
}
static void
--
2.25.1
More information about the igt-dev
mailing list