From: Sunil Khatri Date: Tue, 8 Oct 2024 13:02:16 +0000 (+0530) Subject: drm/amdgpu: optimize insert_nop using multi dwords X-Git-Tag: v6.13-rc1~122^2~15^2~99 X-Git-Url: http://git.ipfire.org/?a=commitdiff_plain;h=ea4e4754c9efb53b3f70e5c4c75d08a48b2f7693;p=thirdparty%2Fkernel%2Flinux.git drm/amdgpu: optimize insert_nop using multi dwords Optimize the ring_insert_nop fn for n dwords in one step rather then call to amdgpu_ring_write for each nop packet. This avoid function call for each nop packet and also wptr is updated once only. Signed-off-by: Sunil Khatri Suggested-by: Christian König Reviewed-by: Christian König Signed-off-by: Alex Deucher --- diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c index 03bce2fa866a6..42f616c05f504 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c @@ -108,10 +108,26 @@ int amdgpu_ring_alloc(struct amdgpu_ring *ring, unsigned int ndw) */ void amdgpu_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count) { - int i; + uint32_t occupied, chunk1, chunk2; + uint32_t *dst; - for (i = 0; i < count; i++) - amdgpu_ring_write(ring, ring->funcs->nop); + occupied = ring->wptr & ring->buf_mask; + dst = (void *)&ring->ring[occupied]; + chunk1 = ring->buf_mask + 1 - occupied; + chunk1 = (chunk1 >= count) ? count : chunk1; + chunk2 = count - chunk1; + + if (chunk1) + memset32(dst, ring->funcs->nop, chunk1); + + if (chunk2) { + dst = (void *)ring->ring; + memset32(dst, ring->funcs->nop, chunk2); + } + + ring->wptr += count; + ring->wptr &= ring->ptr_mask; + ring->count_dw -= count; } /**