drm/xe: Refactor emit_xy_fast_copy and emit_mem_copy functions

To perform copy, based on whether the platform supports service copy
engines, either MEM_COPY or XY_FAST_COPY_BLT instruction is used.
Length of both the instructions is same today and so they use a common
define EMIT_COPY_DW.
This is not true for the future platforms. Implement separate functions
which return the length of the instruction to help in preparing for it.

Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patch.msgid.link/20260511123746.616662-8-balasubramani.vivekanandan@intel.com
Signed-off-by: Balasubramani Vivekanandan <balasubramani.vivekanandan@intel.com>

diff --git a/drivers/gpu/drm/xe/instructions/xe_gpu_commands.h b/drivers/gpu/drm/xe/instructions/xe_gpu_commands.h
index edd204b8dfe5771f3e23ee06bf7ace495c2b7035..18d0fde8c98f95b114158a259be76e25ea8fe82e 100644 (file)
--- a/drivers/gpu/drm/xe/instructions/xe_gpu_commands.h
+++ b/drivers/gpu/drm/xe/instructions/xe_gpu_commands.h
@@ -30,7 +30,7 @@
 #define   XY_FAST_COPY_BLT_D1_DST_TILE4        REG_BIT(30)
 #define   XE2_XY_FAST_COPY_BLT_MOCS_INDEX_MASK GENMASK(23, 20)
 
-#define MEM_COPY_CMD (2 << 29 | 0x5a << 22 | 0x8)
+#define MEM_COPY_CMD (2 << 29 | 0x5a << 22)
 #define   MEM_COPY_PAGE_COPY_MODE REG_BIT(19)
 #define   MEM_COPY_MATRIX_COPY REG_BIT(17)
 #define   MEM_COPY_SRC_MOCS_INDEX_MASK GENMASK(31, 28)

diff --git a/drivers/gpu/drm/xe/tests/xe_migrate.c b/drivers/gpu/drm/xe/tests/xe_migrate.c
index 50a97705e0acdc94dd6b6d38f9a6f26d7cbeb304..3c1be809be82d81900bd790bcd60009ba9b659eb 100644 (file)
--- a/drivers/gpu/drm/xe/tests/xe_migrate.c
+++ b/drivers/gpu/drm/xe/tests/xe_migrate.c
@@ -421,7 +421,7 @@ static struct dma_fence *blt_copy(struct xe_tile *tile,
                                              avail_pts, avail_pts);
 
                /* Add copy commands size here */
-               batch_size += ((copy_only_ccs) ? 0 : EMIT_COPY_DW) +
+               batch_size += ((copy_only_ccs) ? 0 : emit_copy_cmd_len(xe)) +
                        ((xe_device_has_flat_ccs(xe) && copy_only_ccs) ? EMIT_COPY_CCS_DW : 0);
 
                bb = xe_bb_new(gt, batch_size, xe->info.has_usm);

diff --git a/drivers/gpu/drm/xe/xe_migrate.c b/drivers/gpu/drm/xe/xe_migrate.c
index 4f9be41b2f6457b3902cdbd39e22052c4427ebc9..9428dd5e7760c50bc35d22d3b36f07abf0358a54 100644 (file)
--- a/drivers/gpu/drm/xe/xe_migrate.c
+++ b/drivers/gpu/drm/xe/xe_migrate.c
@@ -728,7 +728,22 @@ static void emit_copy_ccs(struct xe_gt *gt, struct xe_bb *bb,
        bb->len = cs - bb->cs;
 }
 
-#define EMIT_COPY_DW 10
+static u32 blt_fast_copy_cmd_len(struct xe_device *xe)
+{
+       return 10;
+}
+
+static u32 blt_mem_copy_cmd_len(struct xe_device *xe)
+{
+       return 10;
+}
+
+static u32 emit_copy_cmd_len(struct xe_device *xe)
+{
+       return (xe->info.has_mem_copy_instr) ? blt_mem_copy_cmd_len(xe) :
+                 blt_fast_copy_cmd_len(xe);
+}
+
 static void emit_xy_fast_copy(struct xe_gt *gt, struct xe_bb *bb, u64 src_ofs,
                              u64 dst_ofs, unsigned int size,
                              unsigned int pitch)
@@ -736,6 +751,7 @@ static void emit_xy_fast_copy(struct xe_gt *gt, struct xe_bb *bb, u64 src_ofs,
        struct xe_device *xe = gt_to_xe(gt);
        u32 mocs = 0;
        u32 tile_y = 0;
+       u32 len;
 
        xe_gt_assert(gt, !(pitch & 3));
        xe_gt_assert(gt, size / pitch <= S16_MAX);
@@ -748,7 +764,8 @@ static void emit_xy_fast_copy(struct xe_gt *gt, struct xe_bb *bb, u64 src_ofs,
        if (GRAPHICS_VERx100(xe) >= 1250)
                tile_y = XY_FAST_COPY_BLT_D1_SRC_TILE4 | XY_FAST_COPY_BLT_D1_DST_TILE4;
 
-       bb->cs[bb->len++] = XY_FAST_COPY_BLT_CMD | (10 - 2);
+       len = blt_fast_copy_cmd_len(xe);
+       bb->cs[bb->len++] = XY_FAST_COPY_BLT_CMD | (len - 2);
        bb->cs[bb->len++] = XY_FAST_COPY_BLT_DEPTH_32 | pitch | tile_y | mocs;
        bb->cs[bb->len++] = 0;
        bb->cs[bb->len++] = (size / pitch) << 16 | pitch / 4;
@@ -765,6 +782,7 @@ static void emit_mem_copy(struct xe_gt *gt, struct xe_bb *bb, u64 src_ofs,
                          u64 dst_ofs, unsigned int size, unsigned int pitch)
 {
        u32 mode, copy_type, width;
+       u32 len;
 
        xe_gt_assert(gt, IS_ALIGNED(size, pitch));
        xe_gt_assert(gt, pitch <= U16_MAX);
@@ -790,7 +808,9 @@ static void emit_mem_copy(struct xe_gt *gt, struct xe_bb *bb, u64 src_ofs,
 
        xe_gt_assert(gt, width <= U16_MAX);
 
-       bb->cs[bb->len++] = MEM_COPY_CMD | mode | copy_type;
+       len = blt_mem_copy_cmd_len(gt_to_xe(gt));
+
+       bb->cs[bb->len++] = MEM_COPY_CMD | mode | copy_type | (len - 2);
        bb->cs[bb->len++] = width - 1;
        bb->cs[bb->len++] = size / pitch - 1; /* ignored by hw for page-copy/linear above */
        bb->cs[bb->len++] = pitch - 1;
@@ -967,7 +987,7 @@ static struct dma_fence *__xe_migrate_copy(struct xe_migrate *m,
                }
 
                /* Add copy commands size here */
-               batch_size += ((copy_only_ccs) ? 0 : EMIT_COPY_DW) +
+               batch_size += ((copy_only_ccs) ? 0 : emit_copy_cmd_len(xe)) +
                        ((needs_ccs_emit ? EMIT_COPY_CCS_DW : 0));
 
                bb = xe_bb_new(gt, batch_size, usm);
@@ -1406,7 +1426,7 @@ struct dma_fence *xe_migrate_vram_copy_chunk(struct xe_bo *vram_bo, u64 vram_off
 
                batch_size += pte_update_size(m, 0, sysmem, &sysmem_it, &vram_L0, &sysmem_L0_ofs,
                                              &sysmem_L0_pt, 0, avail_pts, avail_pts);
-               batch_size += EMIT_COPY_DW;
+               batch_size += emit_copy_cmd_len(xe);
 
                bb = xe_bb_new(gt, batch_size, usm);
                if (IS_ERR(bb)) {
@@ -2216,7 +2236,7 @@ static struct dma_fence *xe_migrate_vram(struct xe_migrate *m,
        xe_assert(xe, npages * PAGE_SIZE <= MAX_PREEMPTDISABLE_TRANSFER);
 
        batch_size += pte_update_cmd_size(npages << PAGE_SHIFT);
-       batch_size += EMIT_COPY_DW;
+       batch_size += emit_copy_cmd_len(xe);
 
        bb = xe_bb_new(gt, batch_size, use_usm_batch);
        if (IS_ERR(bb)) {