radeonsi: rename and apply SI_OP_CPDMA_SKIP_CACHE_FLUSH to compute as well

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9795>

diff --git a/src/gallium/drivers/radeonsi/si_compute_blit.c b/src/gallium/drivers/radeonsi/si_compute_blit.c
index c149263..7836cdb 100644 (file)
--- a/src/gallium/drivers/radeonsi/si_compute_blit.c
+++ b/src/gallium/drivers/radeonsi/si_compute_blit.c
@@ -72,7 +72,8 @@ void si_launch_grid_internal(struct si_context *sctx, struct pipe_grid_info *inf
 
    /* Invalidate L0-L1 caches. */
    /* sL0 is never invalidated, because src resources don't use it. */
-   sctx->flags |= SI_CONTEXT_INV_VCACHE;
+   if (!(flags & SI_OP_SKIP_CACHE_INV_BEFORE))
+      sctx->flags |= SI_CONTEXT_INV_VCACHE;
 
    /* Set settings for driver-internal compute dispatches. */
    sctx->flags &= ~SI_CONTEXT_START_PIPELINE_STATS;
@@ -125,7 +126,8 @@ static void si_compute_clear_12bytes_buffer(struct si_context *sctx, struct pipe
    unsigned data[4] = {0};
    memcpy(data, clear_value, 12);
 
-   sctx->flags |= si_get_flush_flags(sctx, coher, SI_COMPUTE_DST_CACHE_POLICY);
+   if (!(flags & SI_OP_SKIP_CACHE_INV_BEFORE))
+      sctx->flags |= si_get_flush_flags(sctx, coher, SI_COMPUTE_DST_CACHE_POLICY);
 
    struct pipe_shader_buffer saved_sb = {0};
    si_get_shader_buffers(sctx, PIPE_SHADER_COMPUTE, 0, 1, &saved_sb);
@@ -188,7 +190,8 @@ static void si_compute_do_clear_or_copy(struct si_context *sctx, struct pipe_res
    assert(dst->target != PIPE_BUFFER || dst_offset + size <= dst->width0);
    assert(!src || src_offset + size <= src->width0);
 
-   sctx->flags |= si_get_flush_flags(sctx, coher, SI_COMPUTE_DST_CACHE_POLICY);
+   if (!(flags & SI_OP_SKIP_CACHE_INV_BEFORE))
+      sctx->flags |= si_get_flush_flags(sctx, coher, SI_COMPUTE_DST_CACHE_POLICY);
 
    /* Save states. */
    void *saved_cs = sctx->cs_shader_state.program;

diff --git a/src/gallium/drivers/radeonsi/si_cp_dma.c b/src/gallium/drivers/radeonsi/si_cp_dma.c
index d42ea35..5a379d9 100644 (file)
--- a/src/gallium/drivers/radeonsi/si_cp_dma.c
+++ b/src/gallium/drivers/radeonsi/si_cp_dma.c
@@ -208,7 +208,7 @@ void si_cp_dma_clear_buffer(struct si_context *sctx, struct radeon_cmdbuf *cs,
    if (sdst) {
       util_range_add(dst, &sdst->valid_buffer_range, offset, offset + size);
 
-      if (!(user_flags & SI_OP_CPDMA_SKIP_CACHE_FLUSH))
+      if (!(user_flags & SI_OP_SKIP_CACHE_INV_BEFORE))
          sctx->flags |= si_get_flush_flags(sctx, coher, cache_policy);
    }
 
@@ -345,7 +345,7 @@ void si_cp_dma_copy_buffer(struct si_context *sctx, struct pipe_resource *dst,
    if (user_flags & SI_OP_SYNC_PS_BEFORE)
       sctx->flags |= SI_CONTEXT_PS_PARTIAL_FLUSH;
 
-   if ((dst || src) && !(user_flags & SI_OP_CPDMA_SKIP_CACHE_FLUSH))
+   if ((dst || src) && !(user_flags & SI_OP_SKIP_CACHE_INV_BEFORE))
          sctx->flags |= si_get_flush_flags(sctx, coher, cache_policy);
 
    /* This is the main part doing the copying. Src is always aligned. */

diff --git a/src/gallium/drivers/radeonsi/si_pipe.h b/src/gallium/drivers/radeonsi/si_pipe.h
index 90f4f4d..691f1b9 100644 (file)
--- a/src/gallium/drivers/radeonsi/si_pipe.h
+++ b/src/gallium/drivers/radeonsi/si_pipe.h
@@ -1367,10 +1367,10 @@ void si_init_clear_functions(struct si_context *sctx);
 #define SI_OP_SYNC_BEFORE                 (SI_OP_SYNC_CS_BEFORE | SI_OP_SYNC_PS_BEFORE | SI_OP_SYNC_CPDMA_BEFORE)
 #define SI_OP_SYNC_AFTER                  (1 << 3)
 #define SI_OP_SYNC_BEFORE_AFTER           (SI_OP_SYNC_BEFORE | SI_OP_SYNC_AFTER)
-#define SI_OP_CS_IMAGE                    (1 << 4)
-#define SI_OP_CS_RENDER_COND_ENABLE       (1 << 5)
-#define SI_OP_CPDMA_SKIP_CHECK_CS_SPACE   (1 << 6) /* don't call need_cs_space */
-#define SI_OP_CPDMA_SKIP_CACHE_FLUSH      (1 << 7) /* don't flush caches */
+#define SI_OP_SKIP_CACHE_INV_BEFORE       (1 << 4) /* don't invalidate caches */
+#define SI_OP_CS_IMAGE                    (1 << 5)
+#define SI_OP_CS_RENDER_COND_ENABLE       (1 << 6)
+#define SI_OP_CPDMA_SKIP_CHECK_CS_SPACE   (1 << 7) /* don't call need_cs_space */
 
 unsigned si_get_flush_flags(struct si_context *sctx, enum si_coherency coher,
                             enum si_cache_policy cache_policy);