summaryrefslogtreecommitdiff
path: root/lib
diff options
context:
space:
mode:
Diffstat (limited to 'lib')
-rw-r--r--lib/gpgpu_fill.c187
-rw-r--r--lib/gpgpu_fill.h34
2 files changed, 221 insertions, 0 deletions
diff --git a/lib/gpgpu_fill.c b/lib/gpgpu_fill.c
index 5660d4c0..49988a36 100644
--- a/lib/gpgpu_fill.c
+++ b/lib/gpgpu_fill.c
@@ -120,6 +120,7 @@ static const uint32_t gen12_gpgpu_kernel[][4] = {
*
*/
+#define PAGE_SIZE 4096
#define BATCH_STATE_SPLIT 2048
/* VFE STATE params */
#define THREADS 1
@@ -179,6 +180,56 @@ gen7_gpgpu_fillfunc(struct intel_batchbuffer *batch,
}
void
+gen7_gpgpu_fillfunc_v2(int i915,
+ struct intel_buf *buf,
+ unsigned x, unsigned y,
+ unsigned width, unsigned height,
+ uint8_t color)
+{
+ struct intel_bb *ibb;
+ uint32_t curbe_buffer, interface_descriptor;
+
+ ibb = intel_bb_create(i915, PAGE_SIZE);
+ intel_bb_add_object(ibb, buf->handle, 0, true);
+
+ intel_bb_ptr_set(ibb, BATCH_STATE_SPLIT);
+
+ /* Fill curbe buffer data */
+ curbe_buffer = gen7_fill_curbe_buffer_data_v2(ibb, color);
+
+ /*
+ * const buffer needs to fill for every thread, but as we have just 1
+ * thread per every group, so need only one curbe data.
+ * For each thread, just use thread group ID for buffer offset.
+ */
+ interface_descriptor =
+ gen7_fill_interface_descriptor_v2(ibb, buf,
+ gen7_gpgpu_kernel,
+ sizeof(gen7_gpgpu_kernel));
+
+ intel_bb_ptr_set(ibb, 0);
+
+ /* GPGPU pipeline */
+ intel_bb_out(ibb, GEN7_PIPELINE_SELECT | PIPELINE_SELECT_GPGPU);
+
+ gen7_emit_state_base_address_v2(ibb);
+ gen7_emit_vfe_state_v2(ibb, THREADS, GEN7_GPGPU_URB_ENTRIES,
+ GPGPU_URB_SIZE, GPGPU_CURBE_SIZE,
+ GEN7_VFE_STATE_GPGPU_MODE);
+ gen7_emit_curbe_load_v2(ibb, curbe_buffer);
+ gen7_emit_interface_descriptor_load_v2(ibb, interface_descriptor);
+ gen7_emit_gpgpu_walk_v2(ibb, x, y, width, height);
+
+ intel_bb_out(ibb, MI_BATCH_BUFFER_END);
+ intel_bb_ptr_align(ibb, 32);
+
+ intel_bb_exec(ibb, intel_bb_offset(ibb),
+ I915_EXEC_DEFAULT | I915_EXEC_NO_RELOC, true);
+
+ intel_bb_destroy(ibb);
+}
+
+void
gen8_gpgpu_fillfunc(struct intel_batchbuffer *batch,
const struct igt_buf *dst,
unsigned int x, unsigned int y,
@@ -226,6 +277,54 @@ gen8_gpgpu_fillfunc(struct intel_batchbuffer *batch,
intel_batchbuffer_reset(batch);
}
+void
+gen8_gpgpu_fillfunc_v2(int i915,
+ struct intel_buf *buf,
+ unsigned x, unsigned y,
+ unsigned width, unsigned height,
+ uint8_t color)
+{
+ struct intel_bb *ibb;
+ uint32_t curbe_buffer, interface_descriptor;
+
+ ibb = intel_bb_create(i915, PAGE_SIZE);
+ intel_bb_add_object(ibb, buf->handle, 0, true);
+
+ intel_bb_ptr_set(ibb, BATCH_STATE_SPLIT);
+
+ /*
+ * const buffer needs to fill for every thread, but as we have just 1
+ * thread per every group, so need only one curbe data.
+ * For each thread, just use thread group ID for buffer offset.
+ */
+ curbe_buffer = gen7_fill_curbe_buffer_data_v2(ibb, color);
+
+ interface_descriptor = gen8_fill_interface_descriptor_v2(ibb, buf,
+ gen8_gpgpu_kernel, sizeof(gen8_gpgpu_kernel));
+
+ intel_bb_ptr_set(ibb, 0);
+
+ /* GPGPU pipeline */
+ intel_bb_out(ibb, GEN7_PIPELINE_SELECT | PIPELINE_SELECT_GPGPU);
+
+ gen8_emit_state_base_address_v2(ibb);
+ gen8_emit_vfe_state_v2(ibb, THREADS, GEN8_GPGPU_URB_ENTRIES,
+ GPGPU_URB_SIZE, GPGPU_CURBE_SIZE);
+
+ gen7_emit_curbe_load_v2(ibb, curbe_buffer);
+ gen7_emit_interface_descriptor_load_v2(ibb, interface_descriptor);
+
+ gen8_emit_gpgpu_walk_v2(ibb, x, y, width, height);
+
+ intel_bb_out(ibb, MI_BATCH_BUFFER_END);
+ intel_bb_ptr_align(ibb, 32);
+
+ intel_bb_exec(ibb, intel_bb_offset(ibb),
+ I915_EXEC_DEFAULT | I915_EXEC_NO_RELOC, true);
+
+ intel_bb_destroy(ibb);
+}
+
static void
__gen9_gpgpu_fillfunc(struct intel_batchbuffer *batch,
const struct igt_buf *dst,
@@ -276,6 +375,60 @@ __gen9_gpgpu_fillfunc(struct intel_batchbuffer *batch,
intel_batchbuffer_reset(batch);
}
+
+static void
+__gen9_gpgpu_fillfunc_v2(int i915,
+ struct intel_buf *buf,
+ unsigned x, unsigned y,
+ unsigned width, unsigned height,
+ uint8_t color,
+ const uint32_t kernel[][4], size_t kernel_size)
+{
+ struct intel_bb *ibb;
+ uint32_t curbe_buffer, interface_descriptor;
+
+ ibb = intel_bb_create(i915, PAGE_SIZE);
+ intel_bb_add_object(ibb, buf->handle, 0, true);
+
+ intel_bb_ptr_set(ibb, BATCH_STATE_SPLIT);
+
+ /*
+ * const buffer needs to fill for every thread, but as we have just 1
+ * thread per every group, so need only one curbe data.
+ * For each thread, just use thread group ID for buffer offset.
+ */
+ /* Fill curbe buffer data */
+ curbe_buffer = gen7_fill_curbe_buffer_data_v2(ibb, color);
+
+ interface_descriptor = gen8_fill_interface_descriptor_v2(ibb, buf,
+ kernel,
+ kernel_size);
+
+ intel_bb_ptr_set(ibb, 0);
+
+ /* GPGPU pipeline */
+ intel_bb_out(ibb, GEN7_PIPELINE_SELECT | GEN9_PIPELINE_SELECTION_MASK |
+ PIPELINE_SELECT_GPGPU);
+
+ gen9_emit_state_base_address_v2(ibb);
+
+ gen8_emit_vfe_state_v2(ibb, THREADS, GEN8_GPGPU_URB_ENTRIES,
+ GPGPU_URB_SIZE, GPGPU_CURBE_SIZE);
+
+ gen7_emit_curbe_load_v2(ibb, curbe_buffer);
+ gen7_emit_interface_descriptor_load_v2(ibb, interface_descriptor);
+
+ gen8_emit_gpgpu_walk_v2(ibb, x, y, width, height);
+
+ intel_bb_out(ibb, MI_BATCH_BUFFER_END);
+ intel_bb_ptr_align(ibb, 32);
+
+ intel_bb_exec(ibb, intel_bb_offset(ibb),
+ I915_EXEC_DEFAULT | I915_EXEC_NO_RELOC, true);
+
+ intel_bb_destroy(ibb);
+}
+
void gen9_gpgpu_fillfunc(struct intel_batchbuffer *batch,
const struct igt_buf *dst,
unsigned int x, unsigned int y,
@@ -286,6 +439,18 @@ void gen9_gpgpu_fillfunc(struct intel_batchbuffer *batch,
gen9_gpgpu_kernel, sizeof(gen9_gpgpu_kernel));
}
+void gen9_gpgpu_fillfunc_v2(int i915,
+ struct intel_buf *buf,
+ unsigned x, unsigned y,
+ unsigned width, unsigned height,
+ uint8_t color)
+{
+ __gen9_gpgpu_fillfunc_v2(i915, buf, x, y, width, height, color,
+ gen9_gpgpu_kernel,
+ sizeof(gen9_gpgpu_kernel));
+}
+
+
void gen11_gpgpu_fillfunc(struct intel_batchbuffer *batch,
const struct igt_buf *dst,
unsigned int x, unsigned int y,
@@ -296,6 +461,17 @@ void gen11_gpgpu_fillfunc(struct intel_batchbuffer *batch,
gen11_gpgpu_kernel, sizeof(gen11_gpgpu_kernel));
}
+void gen11_gpgpu_fillfunc_v2(int i915,
+ struct intel_buf *buf,
+ unsigned x, unsigned y,
+ unsigned width, unsigned height,
+ uint8_t color)
+{
+ __gen9_gpgpu_fillfunc_v2(i915, buf, x, y, width, height, color,
+ gen11_gpgpu_kernel,
+ sizeof(gen11_gpgpu_kernel));
+}
+
void gen12_gpgpu_fillfunc(struct intel_batchbuffer *batch,
const struct igt_buf *dst,
unsigned int x, unsigned int y,
@@ -305,3 +481,14 @@ void gen12_gpgpu_fillfunc(struct intel_batchbuffer *batch,
__gen9_gpgpu_fillfunc(batch, dst, x, y, width, height, color,
gen12_gpgpu_kernel, sizeof(gen12_gpgpu_kernel));
}
+
+void gen12_gpgpu_fillfunc_v2(int i915,
+ struct intel_buf *buf,
+ unsigned x, unsigned y,
+ unsigned width, unsigned height,
+ uint8_t color)
+{
+ __gen9_gpgpu_fillfunc_v2(i915, buf, x, y, width, height, color,
+ gen12_gpgpu_kernel,
+ sizeof(gen12_gpgpu_kernel));
+}
diff --git a/lib/gpgpu_fill.h b/lib/gpgpu_fill.h
index da7d646f..a387732b 100644
--- a/lib/gpgpu_fill.h
+++ b/lib/gpgpu_fill.h
@@ -28,6 +28,7 @@
#define GPGPU_FILL_H
#include "intel_batchbuffer.h"
+#include "intel_bufops.h"
void
gen7_gpgpu_fillfunc(struct intel_batchbuffer *batch,
@@ -37,6 +38,13 @@ gen7_gpgpu_fillfunc(struct intel_batchbuffer *batch,
uint8_t color);
void
+gen7_gpgpu_fillfunc_v2(int i915,
+ struct intel_buf *buf,
+ unsigned x, unsigned y,
+ unsigned width, unsigned height,
+ uint8_t color);
+
+void
gen8_gpgpu_fillfunc(struct intel_batchbuffer *batch,
const struct igt_buf *dst,
unsigned int x, unsigned int y,
@@ -44,12 +52,25 @@ gen8_gpgpu_fillfunc(struct intel_batchbuffer *batch,
uint8_t color);
void
+gen8_gpgpu_fillfunc_v2(int i915,
+ struct intel_buf *buf,
+ unsigned x, unsigned y,
+ unsigned width, unsigned height,
+ uint8_t color);
+
+void
gen9_gpgpu_fillfunc(struct intel_batchbuffer *batch,
const struct igt_buf *dst,
unsigned int x, unsigned int y,
unsigned int width, unsigned int height,
uint8_t color);
+void gen9_gpgpu_fillfunc_v2(int i915,
+ struct intel_buf *buf,
+ unsigned x, unsigned y,
+ unsigned width, unsigned height,
+ uint8_t color);
+
void
gen11_gpgpu_fillfunc(struct intel_batchbuffer *batch,
const struct igt_buf *dst,
@@ -57,6 +78,12 @@ gen11_gpgpu_fillfunc(struct intel_batchbuffer *batch,
unsigned int width, unsigned int height,
uint8_t color);
+void gen11_gpgpu_fillfunc_v2(int i915,
+ struct intel_buf *buf,
+ unsigned x, unsigned y,
+ unsigned width, unsigned height,
+ uint8_t color);
+
void
gen12_gpgpu_fillfunc(struct intel_batchbuffer *batch,
const struct igt_buf *dst,
@@ -64,4 +91,11 @@ gen12_gpgpu_fillfunc(struct intel_batchbuffer *batch,
unsigned int width, unsigned int height,
uint8_t color);
+void
+gen12_gpgpu_fillfunc_v2(int i915,
+ struct intel_buf *buf,
+ unsigned x, unsigned y,
+ unsigned width, unsigned height,
+ uint8_t color);
+
#endif /* GPGPU_FILL_H */
generated by cgit v1.2.3 (git 2.39.1) at 2025-10-26 12:44:13 +0000