Skip to contentSkip to conversation

tests/amdgpu: add new uvd enc support check

PATCH In review James Zhu 3 participants 4 messages Oct 5, 2017 2 messages not in the archive

lore
Share & tools
lore

James Zhu

author
#0 lore raw

PATCH v2 1/2 tests/amdgpu: add new uvd enc support check

Query hardware IP information to find out if there are uvd encode rings ready for use in kernel driver.

---
 tests/amdgpu/uvd_enc_tests.c | 27 +++++++++++++++++++++------
 1 file changed, 21 insertions(+), 6 deletions(-)
tests/amdgpu/uvd_enc_tests.c +21 -6
index 6c19f7b..7518103 100644
@@ -79,6 +79,8 @@ static void amdgpu_cs_uvd_enc_session_init(void);
 static void amdgpu_cs_uvd_enc_encode(void);
 static void amdgpu_cs_uvd_enc_destroy(void);
 
+static bool uvd_enc_support(void);
+
 CU_TestInfo uvd_enc_tests[] = {
 	{ "UVD ENC create",  amdgpu_cs_uvd_enc_create },
 	{ "UVD ENC session init",  amdgpu_cs_uvd_enc_session_init },
@@ -98,7 +100,7 @@ int suite_uvd_enc_tests_init(void)
 
 	family_id = device_handle->info.family_id;
 
-	if (family_id < AMDGPU_FAMILY_AI || family_id >= AMDGPU_FAMILY_RV) {
+	if (!uvd_enc_support()) {
 		printf("\n\nThe ASIC NOT support UVD ENC, all sub-tests will pass\n");
 		return CUE_SUCCESS;
 	}
@@ -121,7 +123,7 @@ int suite_uvd_enc_tests_clean(void)
 {
 	int r;
 
-	if (family_id < AMDGPU_FAMILY_AI || family_id >= AMDGPU_FAMILY_RV) {
+	if (!uvd_enc_support()) {
 
 		r = amdgpu_device_deinitialize(device_handle);
 		if (r)
@@ -238,11 +240,24 @@ static void free_resource(struct amdgpu_uvd_enc_bo *uvd_enc_bo)
 	memset(uvd_enc_bo, 0, sizeof(*uvd_enc_bo));
 }
 
+static bool uvd_enc_support(void)
+{
+	int r;
+	struct drm_amdgpu_info_hw_ip info;
+
+	r = amdgpu_query_hw_ip_info(device_handle, AMDGPU_HW_IP_UVD_ENC, 0, &info);
+
+	if (r)
+		return false;
+	else
+		return (info.available_rings?true:false);
+}
+
 static void amdgpu_cs_uvd_enc_create(void)
 {
 	int len, r;
 
-	if (family_id < AMDGPU_FAMILY_AI || family_id >= AMDGPU_FAMILY_RV)
+	if (!uvd_enc_support())
 		return;
 
 	enc.width = 160;
@@ -281,7 +296,7 @@ static void amdgpu_cs_uvd_enc_session_init(void)
 {
 	int len, r;
 
-	if (family_id < AMDGPU_FAMILY_AI || family_id >= AMDGPU_FAMILY_RV)
+	if (!uvd_enc_support())
 		return;
 
 	len = 0;
@@ -339,7 +354,7 @@ static void amdgpu_cs_uvd_enc_encode(void)
 	vbuf_size = ALIGN(enc.width, align) * ALIGN(enc.height, 16) * 1.5;
 	cpb_size = vbuf_size * 10;
 
-	if (family_id < AMDGPU_FAMILY_AI || family_id >= AMDGPU_FAMILY_RV)
+	if (!uvd_enc_support())
 		return;
 
 	num_resources  = 0;
@@ -472,7 +487,7 @@ static void amdgpu_cs_uvd_enc_destroy(void)
 	struct amdgpu_uvd_enc_bo sw_ctx;
 	int len, r;
 
-	if (family_id < AMDGPU_FAMILY_AI || family_id >= AMDGPU_FAMILY_RV)
+	if (!uvd_enc_support())
 		return;
 
 	num_resources  = 0;
Signature · 6 lines
2.7.4

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

A message that is not in the archive (sent off-list or not yet received): 1507217090-10669-1-git-send-email-James.Zhu-5C7GfCeVMHo@public.gmane.org

author
#2 lore raw

PATCH v2 2/2 tests/amdgpu: fix uvd enc data corruption issue

In uvd encode parameter package, parameters input_pic_luma_pitch and input_pic_chroma_pitch should be picture width align with hardware alignment. The hardware alignment is 16 for amdgpu family earlier than AMDGPU_FAMILY_AI, and 256 for later than and including AMDGPU_FAMILY_AI.

---
 tests/amdgpu/uvd_enc_tests.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
tests/amdgpu/uvd_enc_tests.c +3 -1
index 7518103..bbda131 100644
@@ -272,7 +272,7 @@ static void amdgpu_cs_uvd_enc_create(void)
 static void check_result(struct amdgpu_uvd_enc *enc)
 {
 	uint64_t sum;
-	uint32_t s = 26382;
+	uint32_t s = 175602;
 	uint32_t *ptr, size;
 	int i, j, r;
 
@@ -463,6 +463,8 @@ static void amdgpu_cs_uvd_enc_encode(void)
 	ib_cpu[len++] = chroma_offset >> 32;
 	ib_cpu[len++] = chroma_offset;
 	memcpy((ib_cpu + len), uve_encode_param, sizeof(uve_encode_param));
+	ib_cpu[len] = ALIGN(enc.width, align);
+	ib_cpu[len + 1] = ALIGN(enc.width, align);
 	len += sizeof(uve_encode_param) / 4;
 
 	memcpy((ib_cpu + len), uve_op_speed_enc_mode, sizeof(uve_op_speed_enc_mode));
Signature · 6 lines
2.7.4

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
Reviewed-by on v2 2/2
#3 lore raw

Re: [PATCH libdrm v2 2/2] tests/amdgpu: fix uvd enc data corruption issue

Quoted 24 lines · James Zhu

In uvd encode parameter package, parameters input_pic_luma_pitch and input_pic_chroma_pitch should be picture width align with hardware alignment. The hardware alignment is 16 for amdgpu family earlier than AMDGPU_FAMILY_AI, and 256 for later than and including AMDGPU_FAMILY_AI.

---
  tests/amdgpu/uvd_enc_tests.c | 4 +++-
  1 file changed, 3 insertions(+), 1 deletion(-)
tests/amdgpu/uvd_enc_tests.c +1 -1
index 7518103..bbda131 100644
@@ -272,7 +272,7 @@ static void amdgpu_cs_uvd_enc_create(void)
  static void check_result(struct amdgpu_uvd_enc *enc)
  {
  	uint64_t sum;
-	uint32_t s = 26382;
+	uint32_t s = 175602;
  	uint32_t *ptr, size;
  	int i, j, r;
 

tests/amdgpu/uvd_enc_tests.c static void amdgpu_cs_uvd_enc_encode(void) · line 467

Quoted 5 lines
tests/amdgpu/uvd_enc_tests.c +2 -0
@@ -463,6 +463,8 @@ static void amdgpu_cs_uvd_enc_encode(void)
  	ib_cpu[len++] = chroma_offset >> 32;
  	ib_cpu[len++] = chroma_offset;
  	memcpy((ib_cpu + len), uve_encode_param, sizeof(uve_encode_param));
+	ib_cpu[len] = ALIGN(enc.width, align);
+	ib_cpu[len + 1] = ALIGN(enc.width, align);

Since here we override the pitch value based on below from uve_ib.h.

static const uint32_t uve_encode_param[] = {
    0x000000a0,
    0x00000080,

We'd better to reset them to 0 from the header file, since we don't want to leave the incorrect value there.

With that fixed, the series is

tests/amdgpu/uvd_enc_tests.c static void amdgpu_cs_uvd_enc_encode(void) · line 470

Quoted 3 lines
tests/amdgpu/uvd_enc_tests.c +0 -0
  	len += sizeof(uve_encode_param) / 4;
 
  	memcpy((ib_cpu + len), uve_op_speed_enc_mode, sizeof(uve_op_speed_enc_mode));

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel

A message that is not in the archive (sent off-list or not yet received): 99fd1615-2a78-a29c-4b0d-0d94b621275f-5C7GfCeVMHo@public.gmane.org

#5 lore raw

Re: [PATCH libdrm v2 2/2] tests/amdgpu: fix uvd enc data corruption issue

Hi Leo,

Sure, I will reset 0 in header file

Thanks!
James Zhu

Quoted 37 lines · Leo Liu
Quoted 23 lines · James Zhu

In uvd encode parameter package, parameters input_pic_luma_pitch and input_pic_chroma_pitch should be picture width align with hardware alignment. The hardware alignment is 16 for amdgpu family earlier than AMDGPU_FAMILY_AI, and 256 for later than and including AMDGPU_FAMILY_AI.

---
  tests/amdgpu/uvd_enc_tests.c | 4 +++-

  1 file changed, 3 insertions(+), 1 deletion(-)
tests/amdgpu/uvd_enc_tests.c +1 -1
index 7518103..bbda131 100644
@@ -272,7 +272,7 @@ static void amdgpu_cs_uvd_enc_create(void)
   static void check_result(struct amdgpu_uvd_enc *enc)
   {
       uint64_t sum;
-    uint32_t s = 26382;
+    uint32_t s = 175602;
       uint32_t *ptr, size;
       int i, j, r;
   @@ -463,6 +463,8 @@ static void amdgpu_cs_uvd_enc_encode(void)
Quoted 5 lines

      ib_cpu[len++] = chroma_offset >> 32;
      ib_cpu[len++] = chroma_offset;
      memcpy((ib_cpu + len), uve_encode_param, sizeof(uve_encode_param)); +    ib_cpu[len] = ALIGN(enc.width, align);
+    ib_cpu[len + 1] = ALIGN(enc.width, align);

Since here we override the pitch value based on below from uve_ib.h.

static const uint32_t uve_encode_param[] = {
    0x000000a0,
    0x00000080,

Quoted 9 lines

We'd better to reset them to 0 from the header file, since we don't want to leave the incorrect value there.

With that fixed, the series is

Quoted 2 lines

      len += sizeof(uve_encode_param) / 4;
        memcpy((ib_cpu + len), uve_op_speed_enc_mode, sizeof(uve_op_speed_enc_mode));

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx