Commit 89d35807 authored by Alex Deucher's avatar Alex Deucher Committed by Christian König

drm/radeon: update rptr saving logic for memory buffers

Add support for using memory buffers rather than
scratch registers.  Some rings may not be able to
write to scratch registers.
Signed-off-by: default avatarAlex Deucher <alexander.deucher@amd.com>
Reviewed-by: default avatarChristian König <christian.koenig@amd.com>
parent 8b25ed34
...@@ -1364,17 +1364,25 @@ void evergreen_mc_program(struct radeon_device *rdev) ...@@ -1364,17 +1364,25 @@ void evergreen_mc_program(struct radeon_device *rdev)
void evergreen_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib) void evergreen_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib)
{ {
struct radeon_ring *ring = &rdev->ring[ib->ring]; struct radeon_ring *ring = &rdev->ring[ib->ring];
u32 next_rptr;
/* set to DX10/11 mode */ /* set to DX10/11 mode */
radeon_ring_write(ring, PACKET3(PACKET3_MODE_CONTROL, 0)); radeon_ring_write(ring, PACKET3(PACKET3_MODE_CONTROL, 0));
radeon_ring_write(ring, 1); radeon_ring_write(ring, 1);
if (ring->rptr_save_reg) { if (ring->rptr_save_reg) {
uint32_t next_rptr = ring->wptr + 3 + 4; next_rptr = ring->wptr + 3 + 4;
radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1)); radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1));
radeon_ring_write(ring, ((ring->rptr_save_reg - radeon_ring_write(ring, ((ring->rptr_save_reg -
PACKET3_SET_CONFIG_REG_START) >> 2)); PACKET3_SET_CONFIG_REG_START) >> 2));
radeon_ring_write(ring, next_rptr); radeon_ring_write(ring, next_rptr);
} else if (rdev->wb.enabled) {
next_rptr = ring->wptr + 5 + 4;
radeon_ring_write(ring, PACKET3(PACKET3_MEM_WRITE, 3));
radeon_ring_write(ring, ring->next_rptr_gpu_addr & 0xfffffffc);
radeon_ring_write(ring, (upper_32_bits(ring->next_rptr_gpu_addr) & 0xff) | (1 << 18));
radeon_ring_write(ring, next_rptr);
radeon_ring_write(ring, 0);
} }
radeon_ring_write(ring, PACKET3(PACKET3_INDIRECT_BUFFER, 2)); radeon_ring_write(ring, PACKET3(PACKET3_INDIRECT_BUFFER, 2));
......
...@@ -2163,11 +2163,13 @@ void r600_ring_init(struct radeon_device *rdev, struct radeon_ring *ring, unsign ...@@ -2163,11 +2163,13 @@ void r600_ring_init(struct radeon_device *rdev, struct radeon_ring *ring, unsign
ring->ring_size = ring_size; ring->ring_size = ring_size;
ring->align_mask = 16 - 1; ring->align_mask = 16 - 1;
if (radeon_ring_supports_scratch_reg(rdev, ring)) {
r = radeon_scratch_get(rdev, &ring->rptr_save_reg); r = radeon_scratch_get(rdev, &ring->rptr_save_reg);
if (r) { if (r) {
DRM_ERROR("failed to get scratch reg for rptr save (%d).\n", r); DRM_ERROR("failed to get scratch reg for rptr save (%d).\n", r);
ring->rptr_save_reg = 0; ring->rptr_save_reg = 0;
} }
}
} }
void r600_cp_fini(struct radeon_device *rdev) void r600_cp_fini(struct radeon_device *rdev)
...@@ -2576,13 +2578,21 @@ void r600_fini(struct radeon_device *rdev) ...@@ -2576,13 +2578,21 @@ void r600_fini(struct radeon_device *rdev)
void r600_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib) void r600_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib)
{ {
struct radeon_ring *ring = &rdev->ring[ib->ring]; struct radeon_ring *ring = &rdev->ring[ib->ring];
u32 next_rptr;
if (ring->rptr_save_reg) { if (ring->rptr_save_reg) {
uint32_t next_rptr = ring->wptr + 3 + 4; next_rptr = ring->wptr + 3 + 4;
radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1)); radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1));
radeon_ring_write(ring, ((ring->rptr_save_reg - radeon_ring_write(ring, ((ring->rptr_save_reg -
PACKET3_SET_CONFIG_REG_OFFSET) >> 2)); PACKET3_SET_CONFIG_REG_OFFSET) >> 2));
radeon_ring_write(ring, next_rptr); radeon_ring_write(ring, next_rptr);
} else if (rdev->wb.enabled) {
next_rptr = ring->wptr + 5 + 4;
radeon_ring_write(ring, PACKET3(PACKET3_MEM_WRITE, 3));
radeon_ring_write(ring, ring->next_rptr_gpu_addr & 0xfffffffc);
radeon_ring_write(ring, (upper_32_bits(ring->next_rptr_gpu_addr) & 0xff) | (1 << 18));
radeon_ring_write(ring, next_rptr);
radeon_ring_write(ring, 0);
} }
radeon_ring_write(ring, PACKET3(PACKET3_INDIRECT_BUFFER, 2)); radeon_ring_write(ring, PACKET3(PACKET3_INDIRECT_BUFFER, 2));
......
...@@ -623,6 +623,8 @@ struct radeon_ring { ...@@ -623,6 +623,8 @@ struct radeon_ring {
unsigned rptr_offs; unsigned rptr_offs;
unsigned rptr_reg; unsigned rptr_reg;
unsigned rptr_save_reg; unsigned rptr_save_reg;
u64 next_rptr_gpu_addr;
volatile u32 *next_rptr_cpu_addr;
unsigned wptr; unsigned wptr;
unsigned wptr_old; unsigned wptr_old;
unsigned wptr_reg; unsigned wptr_reg;
...@@ -758,6 +760,8 @@ int radeon_ib_pool_init(struct radeon_device *rdev); ...@@ -758,6 +760,8 @@ int radeon_ib_pool_init(struct radeon_device *rdev);
void radeon_ib_pool_fini(struct radeon_device *rdev); void radeon_ib_pool_fini(struct radeon_device *rdev);
int radeon_ib_ring_tests(struct radeon_device *rdev); int radeon_ib_ring_tests(struct radeon_device *rdev);
/* Ring access between begin & end cannot sleep */ /* Ring access between begin & end cannot sleep */
bool radeon_ring_supports_scratch_reg(struct radeon_device *rdev,
struct radeon_ring *ring);
void radeon_ring_free_size(struct radeon_device *rdev, struct radeon_ring *cp); void radeon_ring_free_size(struct radeon_device *rdev, struct radeon_ring *cp);
int radeon_ring_alloc(struct radeon_device *rdev, struct radeon_ring *cp, unsigned ndw); int radeon_ring_alloc(struct radeon_device *rdev, struct radeon_ring *cp, unsigned ndw);
int radeon_ring_lock(struct radeon_device *rdev, struct radeon_ring *cp, unsigned ndw); int radeon_ring_lock(struct radeon_device *rdev, struct radeon_ring *cp, unsigned ndw);
...@@ -871,6 +875,7 @@ struct radeon_wb { ...@@ -871,6 +875,7 @@ struct radeon_wb {
}; };
#define RADEON_WB_SCRATCH_OFFSET 0 #define RADEON_WB_SCRATCH_OFFSET 0
#define RADEON_WB_RING0_NEXT_RPTR 256
#define RADEON_WB_CP_RPTR_OFFSET 1024 #define RADEON_WB_CP_RPTR_OFFSET 1024
#define RADEON_WB_CP1_RPTR_OFFSET 1280 #define RADEON_WB_CP1_RPTR_OFFSET 1280
#define RADEON_WB_CP2_RPTR_OFFSET 1536 #define RADEON_WB_CP2_RPTR_OFFSET 1536
......
...@@ -207,6 +207,19 @@ void radeon_ring_write(struct radeon_ring *ring, uint32_t v) ...@@ -207,6 +207,19 @@ void radeon_ring_write(struct radeon_ring *ring, uint32_t v)
ring->ring_free_dw--; ring->ring_free_dw--;
} }
bool radeon_ring_supports_scratch_reg(struct radeon_device *rdev,
struct radeon_ring *ring)
{
switch (ring->idx) {
case RADEON_RING_TYPE_GFX_INDEX:
case CAYMAN_RING_TYPE_CP1_INDEX:
case CAYMAN_RING_TYPE_CP2_INDEX:
return true;
default:
return false;
}
}
void radeon_ring_free_size(struct radeon_device *rdev, struct radeon_ring *ring) void radeon_ring_free_size(struct radeon_device *rdev, struct radeon_ring *ring)
{ {
u32 rptr; u32 rptr;
...@@ -372,7 +385,7 @@ unsigned radeon_ring_backup(struct radeon_device *rdev, struct radeon_ring *ring ...@@ -372,7 +385,7 @@ unsigned radeon_ring_backup(struct radeon_device *rdev, struct radeon_ring *ring
mutex_lock(&rdev->ring_lock); mutex_lock(&rdev->ring_lock);
*data = NULL; *data = NULL;
if (ring->ring_obj == NULL || !ring->rptr_save_reg) { if (ring->ring_obj == NULL) {
mutex_unlock(&rdev->ring_lock); mutex_unlock(&rdev->ring_lock);
return 0; return 0;
} }
...@@ -384,7 +397,16 @@ unsigned radeon_ring_backup(struct radeon_device *rdev, struct radeon_ring *ring ...@@ -384,7 +397,16 @@ unsigned radeon_ring_backup(struct radeon_device *rdev, struct radeon_ring *ring
} }
/* calculate the number of dw on the ring */ /* calculate the number of dw on the ring */
if (ring->rptr_save_reg)
ptr = RREG32(ring->rptr_save_reg); ptr = RREG32(ring->rptr_save_reg);
else if (rdev->wb.enabled)
ptr = le32_to_cpu(*ring->next_rptr_cpu_addr);
else {
/* no way to read back the next rptr */
mutex_unlock(&rdev->ring_lock);
return 0;
}
size = ring->wptr + (ring->ring_size / 4); size = ring->wptr + (ring->ring_size / 4);
size -= ptr; size -= ptr;
size &= ring->ptr_mask; size &= ring->ptr_mask;
...@@ -478,6 +500,11 @@ int radeon_ring_init(struct radeon_device *rdev, struct radeon_ring *ring, unsig ...@@ -478,6 +500,11 @@ int radeon_ring_init(struct radeon_device *rdev, struct radeon_ring *ring, unsig
} }
ring->ptr_mask = (ring->ring_size / 4) - 1; ring->ptr_mask = (ring->ring_size / 4) - 1;
ring->ring_free_dw = ring->ring_size / 4; ring->ring_free_dw = ring->ring_size / 4;
if (rdev->wb.enabled) {
u32 index = RADEON_WB_RING0_NEXT_RPTR + (ring->idx * 4);
ring->next_rptr_gpu_addr = rdev->wb.gpu_addr + index;
ring->next_rptr_cpu_addr = &rdev->wb.wb[index/4];
}
if (radeon_debugfs_ring_init(rdev, ring)) { if (radeon_debugfs_ring_init(rdev, ring)) {
DRM_ERROR("Failed to register debugfs file for rings !\n"); DRM_ERROR("Failed to register debugfs file for rings !\n");
} }
......
...@@ -1772,12 +1772,20 @@ void si_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib) ...@@ -1772,12 +1772,20 @@ void si_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib)
header = PACKET3(PACKET3_INDIRECT_BUFFER_CONST, 2); header = PACKET3(PACKET3_INDIRECT_BUFFER_CONST, 2);
} else { } else {
u32 next_rptr;
if (ring->rptr_save_reg) { if (ring->rptr_save_reg) {
uint32_t next_rptr = ring->wptr + 3 + 4 + 8; next_rptr = ring->wptr + 3 + 4 + 8;
radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1)); radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1));
radeon_ring_write(ring, ((ring->rptr_save_reg - radeon_ring_write(ring, ((ring->rptr_save_reg -
PACKET3_SET_CONFIG_REG_START) >> 2)); PACKET3_SET_CONFIG_REG_START) >> 2));
radeon_ring_write(ring, next_rptr); radeon_ring_write(ring, next_rptr);
} else if (rdev->wb.enabled) {
next_rptr = ring->wptr + 5 + 4 + 8;
radeon_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
radeon_ring_write(ring, (1 << 8));
radeon_ring_write(ring, ring->next_rptr_gpu_addr & 0xfffffffc);
radeon_ring_write(ring, upper_32_bits(ring->next_rptr_gpu_addr) & 0xffffffff);
radeon_ring_write(ring, next_rptr);
} }
header = PACKET3(PACKET3_INDIRECT_BUFFER, 2); header = PACKET3(PACKET3_INDIRECT_BUFFER, 2);
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment