static void
copy(int fd, uint32_t dst, uint32_t src)
{
uint32_t batch[1024], *b = batch;
struct drm_i915_gem_relocation_entry reloc[2], *r = reloc;
struct drm_i915_gem_exec_object2 obj[3];
struct drm_i915_gem_execbuffer2 exec;
uint32_t handle;
int ret;
/* invariant state */
*b++ = (_3DSTATE_AA_CMD |
AA_LINE_ECAAR_WIDTH_ENABLE |
AA_LINE_ECAAR_WIDTH_1_0 |
AA_LINE_REGION_WIDTH_ENABLE | AA_LINE_REGION_WIDTH_1_0);
*b++ = (_3DSTATE_INDEPENDENT_ALPHA_BLEND_CMD |
IAB_MODIFY_ENABLE |
IAB_MODIFY_FUNC | (BLENDFUNC_ADD << IAB_FUNC_SHIFT) |
IAB_MODIFY_SRC_FACTOR | (BLENDFACT_ONE <<
IAB_SRC_FACTOR_SHIFT) |
IAB_MODIFY_DST_FACTOR | (BLENDFACT_ZERO <<
IAB_DST_FACTOR_SHIFT));
*b++ = (_3DSTATE_DFLT_DIFFUSE_CMD);
*b++ = (0);
*b++ = (_3DSTATE_DFLT_SPEC_CMD);
*b++ = (0);
*b++ = (_3DSTATE_DFLT_Z_CMD);
*b++ = (0);
*b++ = (_3DSTATE_COORD_SET_BINDINGS |
CSB_TCB(0, 0) |
CSB_TCB(1, 1) |
CSB_TCB(2, 2) |
CSB_TCB(3, 3) |
CSB_TCB(4, 4) |
CSB_TCB(5, 5) | CSB_TCB(6, 6) | CSB_TCB(7, 7));
*b++ = (_3DSTATE_RASTER_RULES_CMD |
ENABLE_POINT_RASTER_RULE |
OGL_POINT_RASTER_RULE |
ENABLE_LINE_STRIP_PROVOKE_VRTX |
ENABLE_TRI_FAN_PROVOKE_VRTX |
LINE_STRIP_PROVOKE_VRTX(1) |
TRI_FAN_PROVOKE_VRTX(2) | ENABLE_TEXKILL_3D_4D | TEXKILL_4D);
*b++ = (_3DSTATE_MODES_4_CMD |
ENABLE_LOGIC_OP_FUNC | LOGIC_OP_FUNC(LOGICOP_COPY) |
ENABLE_STENCIL_WRITE_MASK | STENCIL_WRITE_MASK(0xff) |
ENABLE_STENCIL_TEST_MASK | STENCIL_TEST_MASK(0xff));
*b++ = (_3DSTATE_LOAD_STATE_IMMEDIATE_1 | I1_LOAD_S(3) | I1_LOAD_S(4) | I1_LOAD_S(5) | 2);
*b++ = (0x00000000); /* Disable texture coordinate wrap-shortest */
*b++ = ((1 << S4_POINT_WIDTH_SHIFT) |
S4_LINE_WIDTH_ONE |
S4_CULLMODE_NONE |
S4_VFMT_XY);
*b++ = (0x00000000); /* Stencil. */
*b++ = (_3DSTATE_SCISSOR_ENABLE_CMD | DISABLE_SCISSOR_RECT);
*b++ = (_3DSTATE_SCISSOR_RECT_0_CMD);
*b++ = (0);
*b++ = (0);
*b++ = (_3DSTATE_DEPTH_SUBRECT_DISABLE);
*b++ = (_3DSTATE_LOAD_INDIRECT | 0); /* disable indirect state */
*b++ = (0);
*b++ = (_3DSTATE_STIPPLE);
*b++ = (0x00000000);
*b++ = (_3DSTATE_BACKFACE_STENCIL_OPS | BFO_ENABLE_STENCIL_TWO_SIDE | 0);
/* samler state */
#define TEX_COUNT 1
*b++ = (_3DSTATE_MAP_STATE | (3 * TEX_COUNT));
*b++ = ((1 << TEX_COUNT) - 1);
*b = fill_reloc(r++, b-batch, src, I915_GEM_DOMAIN_SAMPLER, 0); b++;
*b++ = (MAPSURF_32BIT | MT_32BIT_ARGB8888 |
MS3_TILED_SURFACE |
(HEIGHT - 1) << MS3_HEIGHT_SHIFT |
(WIDTH - 1) << MS3_WIDTH_SHIFT);
*b++ = ((WIDTH-1) << MS4_PITCH_SHIFT);
*b++ = (_3DSTATE_SAMPLER_STATE | (3 * TEX_COUNT));
*b++ = ((1 << TEX_COUNT) - 1);
*b++ = (MIPFILTER_NONE << SS2_MIP_FILTER_SHIFT |
FILTER_NEAREST << SS2_MAG_FILTER_SHIFT |
FILTER_NEAREST << SS2_MIN_FILTER_SHIFT);
*b++ = (TEXCOORDMODE_WRAP << SS3_TCX_ADDR_MODE_SHIFT |
TEXCOORDMODE_WRAP << SS3_TCY_ADDR_MODE_SHIFT |
0 << SS3_TEXTUREMAP_INDEX_SHIFT);
*b++ = (0x00000000);
/* render target state */
*b++ = (_3DSTATE_BUF_INFO_CMD);
*b++ = (BUF_3D_ID_COLOR_BACK | BUF_3D_TILED_SURFACE | WIDTH*4);
*b = fill_reloc(r++, b-batch, dst,
I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER);
b++;
*b++ = (_3DSTATE_DST_BUF_VARS_CMD);
*b++ = (COLR_BUF_ARGB8888 |
DSTORG_HORT_BIAS(0x8) |
DSTORG_VERT_BIAS(0x8));
/* draw rect is unconditional */
*b++ = (_3DSTATE_DRAW_RECT_CMD);
*b++ = (0x00000000);
*b++ = (0x00000000); /* ymin, xmin */
*b++ = (DRAW_YMAX(HEIGHT - 1) |
DRAW_XMAX(WIDTH - 1));
/* yorig, xorig (relate to color buffer?) */
*b++ = (0x00000000);
/* texfmt */
*b++ = (_3DSTATE_LOAD_STATE_IMMEDIATE_1 | I1_LOAD_S(1) | I1_LOAD_S(2) | I1_LOAD_S(6) | 2);
*b++ = ((4 << S1_VERTEX_WIDTH_SHIFT) | (4 << S1_VERTEX_PITCH_SHIFT));
*b++ = (~S2_TEXCOORD_FMT(0, TEXCOORDFMT_NOT_PRESENT) |
S2_TEXCOORD_FMT(0, TEXCOORDFMT_2D));
*b++ = (S6_CBUF_BLEND_ENABLE | S6_COLOR_WRITE_ENABLE |
BLENDFUNC_ADD << S6_CBUF_BLEND_FUNC_SHIFT |
BLENDFACT_ONE << S6_CBUF_SRC_BLEND_FACT_SHIFT |
BLENDFACT_ZERO << S6_CBUF_DST_BLEND_FACT_SHIFT);
/* pixel shader */
*b++ = (_3DSTATE_PIXEL_SHADER_PROGRAM | (1 + 3*3 - 2));
/* decl FS_T0 */
*b++ = (D0_DCL |
REG_TYPE(FS_T0) << D0_TYPE_SHIFT |
REG_NR(FS_T0) << D0_NR_SHIFT |
((REG_TYPE(FS_T0) != REG_TYPE_S) ? D0_CHANNEL_ALL : 0));
*b++ = (0);
*b++ = (0);
/* decl FS_S0 */
*b++ = (D0_DCL |
(REG_TYPE(FS_S0) << D0_TYPE_SHIFT) |
(REG_NR(FS_S0) << D0_NR_SHIFT) |
((REG_TYPE(FS_S0) != REG_TYPE_S) ? D0_CHANNEL_ALL : 0));
*b++ = (0);
*b++ = (0);
/* texld(FS_OC, FS_S0, FS_T0 */
*b++ = (T0_TEXLD |
(REG_TYPE(FS_OC) << T0_DEST_TYPE_SHIFT) |
(REG_NR(FS_OC) << T0_DEST_NR_SHIFT) |
(REG_NR(FS_S0) << T0_SAMPLER_NR_SHIFT));
*b++ = ((REG_TYPE(FS_T0) << T1_ADDRESS_REG_TYPE_SHIFT) |
(REG_NR(FS_T0) << T1_ADDRESS_REG_NR_SHIFT));
*b++ = (0);
*b++ = (PRIM3D_RECTLIST | (3*4 - 1));
*b++ = pack_float(WIDTH);
*b++ = pack_float(HEIGHT);
*b++ = pack_float(WIDTH);
*b++ = pack_float(HEIGHT);
*b++ = pack_float(0);
*b++ = pack_float(HEIGHT);
*b++ = pack_float(0);
*b++ = pack_float(HEIGHT);
*b++ = pack_float(0);
*b++ = pack_float(0);
*b++ = pack_float(0);
*b++ = pack_float(0);
*b++ = MI_BATCH_BUFFER_END;
if ((b - batch) & 1)
*b++ = 0;
assert(b - batch <= 1024);
handle = gem_create(fd, 4096);
gem_write(fd, handle, 0, batch, (b-batch)*sizeof(batch[0]));
assert(r-reloc == 2);
obj[0].handle = dst;
obj[0].relocation_count = 0;
obj[0].relocs_ptr = 0;
obj[0].alignment = 0;
obj[0].offset = 0;
obj[0].flags = 0;
obj[0].rsvd1 = 0;
obj[0].rsvd2 = 0;
obj[1].handle = src;
obj[1].relocation_count = 0;
obj[1].relocs_ptr = 0;
obj[1].alignment = 0;
obj[1].offset = 0;
obj[1].flags = 0;
obj[1].rsvd1 = 0;
obj[1].rsvd2 = 0;
obj[2].handle = handle;
obj[2].relocation_count = 2;
obj[2].relocs_ptr = (uintptr_t)reloc;
obj[2].alignment = 0;
obj[2].offset = 0;
obj[2].flags = 0;
obj[2].rsvd1 = obj[2].rsvd2 = 0;
exec.buffers_ptr = (uintptr_t)obj;
exec.buffer_count = 3;
exec.batch_start_offset = 0;
exec.batch_len = (b-batch)*sizeof(batch[0]);
exec.DR1 = exec.DR4 = 0;
exec.num_cliprects = 0;
exec.cliprects_ptr = 0;
exec.flags = 0;
i915_execbuffer2_set_context_id(exec, 0);
exec.rsvd2 = 0;
ret = drmIoctl(fd, DRM_IOCTL_I915_GEM_EXECBUFFER2, &exec);
while (ret && errno == EBUSY) {
drmCommandNone(fd, DRM_I915_GEM_THROTTLE);
ret = drmIoctl(fd, DRM_IOCTL_I915_GEM_EXECBUFFER2, &exec);
}
assert(ret == 0);
gem_close(fd, handle);
}
static int
blit(int fd, uint32_t dst, uint32_t src, uint32_t *all_bo, int n_bo)
{
uint32_t batch[12];
struct drm_i915_gem_relocation_entry reloc[2];
struct drm_i915_gem_exec_object2 *obj;
struct drm_i915_gem_execbuffer2 exec;
uint32_t handle;
int n, ret, i=0;
batch[i++] = XY_SRC_COPY_BLT_CMD |
XY_SRC_COPY_BLT_WRITE_ALPHA |
XY_SRC_COPY_BLT_WRITE_RGB;
if (intel_gen(intel_get_drm_devid(fd)) >= 8)
batch[i - 1] |= 8;
else
batch[i - 1] |= 6;
batch[i++] = (3 << 24) | /* 32 bits */
(0xcc << 16) | /* copy ROP */
WIDTH*4;
batch[i++] = 0; /* dst x1,y1 */
batch[i++] = (HEIGHT << 16) | WIDTH; /* dst x2,y2 */
batch[i++] = 0; /* dst reloc */
if (intel_gen(intel_get_drm_devid(fd)) >= 8)
batch[i++] = 0;
batch[i++] = 0; /* src x1,y1 */
batch[i++] = WIDTH*4;
batch[i++] = 0; /* src reloc */
if (intel_gen(intel_get_drm_devid(fd)) >= 8)
batch[i++] = 0;
batch[i++] = MI_BATCH_BUFFER_END;
batch[i++] = MI_NOOP;
handle = gem_create(fd, 4096);
gem_write(fd, handle, 0, batch, sizeof(batch));
reloc[0].target_handle = dst;
reloc[0].delta = 0;
reloc[0].offset = 4 * sizeof(batch[0]);
reloc[0].presumed_offset = 0;
reloc[0].read_domains = I915_GEM_DOMAIN_RENDER;
reloc[0].write_domain = I915_GEM_DOMAIN_RENDER;
reloc[1].target_handle = src;
reloc[1].delta = 0;
reloc[1].offset = 7 * sizeof(batch[0]);
if (intel_gen(intel_get_drm_devid(fd)) >= 8)
reloc[1].offset += sizeof(batch[0]);
reloc[1].presumed_offset = 0;
reloc[1].read_domains = I915_GEM_DOMAIN_RENDER;
reloc[1].write_domain = 0;
obj = calloc(n_bo + 1, sizeof(*obj));
for (n = 0; n < n_bo; n++)
obj[n].handle = all_bo[n];
obj[n].handle = handle;
obj[n].relocation_count = 2;
obj[n].relocs_ptr = (uintptr_t)reloc;
exec.buffers_ptr = (uintptr_t)obj;
exec.buffer_count = n_bo + 1;
exec.batch_start_offset = 0;
exec.batch_len = i * 4;
exec.DR1 = exec.DR4 = 0;
exec.num_cliprects = 0;
exec.cliprects_ptr = 0;
exec.flags = HAS_BLT_RING(intel_get_drm_devid(fd)) ? I915_EXEC_BLT : 0;
i915_execbuffer2_set_context_id(exec, 0);
exec.rsvd2 = 0;
ret = drmIoctl(fd, DRM_IOCTL_I915_GEM_EXECBUFFER2, &exec);
if (ret)
ret = errno;
gem_close(fd, handle);
free(obj);
return ret;
}
static void blt_copy(int fd, uint32_t dst, uint32_t src)
{
uint32_t batch[1024], *b = batch;
struct drm_i915_gem_relocation_entry reloc[2], *r = reloc;
struct drm_i915_gem_exec_object2 obj[3];
struct drm_i915_gem_execbuffer2 exec;
uint32_t handle;
int ret;
*b++ = (XY_SRC_COPY_BLT_CMD |
XY_SRC_COPY_BLT_WRITE_ALPHA |
XY_SRC_COPY_BLT_WRITE_RGB);
*b++ = 3 << 24 | 0xcc << 16 | WIDTH * 4;
*b++ = 0;
*b++ = HEIGHT << 16 | WIDTH;
*b = fill_reloc(r++, b-batch, dst,
I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER); b++;
*b++ = 0;
*b++ = WIDTH*4;
*b = fill_reloc(r++, b-batch, src, I915_GEM_DOMAIN_RENDER, 0); b++;
*b++ = MI_BATCH_BUFFER_END;
if ((b - batch) & 1)
*b++ = 0;
igt_assert(b - batch <= 1024);
handle = gem_create(fd, 4096);
gem_write(fd, handle, 0, batch, (b-batch)*sizeof(batch[0]));
igt_assert(r-reloc == 2);
obj[0].handle = dst;
obj[0].relocation_count = 0;
obj[0].relocs_ptr = 0;
obj[0].alignment = 0;
obj[0].offset = 0;
obj[0].flags = EXEC_OBJECT_NEEDS_FENCE;
obj[0].rsvd1 = 0;
obj[0].rsvd2 = 0;
obj[1].handle = src;
obj[1].relocation_count = 0;
obj[1].relocs_ptr = 0;
obj[1].alignment = 0;
obj[1].offset = 0;
obj[1].flags = EXEC_OBJECT_NEEDS_FENCE;
obj[1].rsvd1 = 0;
obj[1].rsvd2 = 0;
obj[2].handle = handle;
obj[2].relocation_count = 2;
obj[2].relocs_ptr = (uintptr_t)reloc;
obj[2].alignment = 0;
obj[2].offset = 0;
obj[2].flags = 0;
obj[2].rsvd1 = obj[2].rsvd2 = 0;
exec.buffers_ptr = (uintptr_t)obj;
exec.buffer_count = 3;
exec.batch_start_offset = 0;
exec.batch_len = (b-batch)*sizeof(batch[0]);
exec.DR1 = exec.DR4 = 0;
exec.num_cliprects = 0;
exec.cliprects_ptr = 0;
exec.flags = 0;
i915_execbuffer2_set_context_id(exec, 0);
exec.rsvd2 = 0;
ret = drmIoctl(fd, DRM_IOCTL_I915_GEM_EXECBUFFER2, &exec);
while (ret && errno == EBUSY) {
drmCommandNone(fd, DRM_I915_GEM_THROTTLE);
ret = drmIoctl(fd, DRM_IOCTL_I915_GEM_EXECBUFFER2, &exec);
}
igt_assert(ret == 0);
gem_close(fd, handle);
}
static void
copy(int fd, uint32_t dst, uint32_t src, unsigned int error)
{
uint32_t batch[12];
struct drm_i915_gem_relocation_entry reloc[2];
struct drm_i915_gem_exec_object2 obj[3];
struct drm_i915_gem_execbuffer2 exec;
uint32_t handle;
int ret, i=0;
batch[i++] = XY_SRC_COPY_BLT_CMD |
XY_SRC_COPY_BLT_WRITE_ALPHA |
XY_SRC_COPY_BLT_WRITE_RGB;
if (intel_gen(intel_get_drm_devid(fd)) >= 8)
batch[i - 1] |= 8;
else
batch[i - 1] |= 6;
batch[i++] = (3 << 24) | /* 32 bits */
(0xcc << 16) | /* copy ROP */
WIDTH*4;
batch[i++] = 0; /* dst x1,y1 */
batch[i++] = (HEIGHT << 16) | WIDTH; /* dst x2,y2 */
batch[i++] = 0; /* dst reloc */
if (intel_gen(intel_get_drm_devid(fd)) >= 8)
batch[i++] = 0;
batch[i++] = 0; /* src x1,y1 */
batch[i++] = WIDTH*4;
batch[i++] = 0; /* src reloc */
if (intel_gen(intel_get_drm_devid(fd)) >= 8)
batch[i++] = 0;
batch[i++] = MI_BATCH_BUFFER_END;
batch[i++] = MI_NOOP;
handle = gem_create(fd, 4096);
gem_write(fd, handle, 0, batch, sizeof(batch));
reloc[0].target_handle = dst;
reloc[0].delta = 0;
reloc[0].offset = 4 * sizeof(batch[0]);
reloc[0].presumed_offset = 0;
reloc[0].read_domains = I915_GEM_DOMAIN_RENDER;
reloc[0].write_domain = I915_GEM_DOMAIN_RENDER;
reloc[1].target_handle = src;
reloc[1].delta = 0;
reloc[1].offset = 7 * sizeof(batch[0]);
if (intel_gen(intel_get_drm_devid(fd)) >= 8)
reloc[1].offset += sizeof(batch[0]);
reloc[1].presumed_offset = 0;
reloc[1].read_domains = I915_GEM_DOMAIN_RENDER;
reloc[1].write_domain = 0;
memset(obj, 0, sizeof(obj));
exec.buffer_count = 0;
obj[exec.buffer_count++].handle = dst;
if (src != dst)
obj[exec.buffer_count++].handle = src;
obj[exec.buffer_count].handle = handle;
obj[exec.buffer_count].relocation_count = 2;
obj[exec.buffer_count].relocs_ptr = (uintptr_t)reloc;
exec.buffer_count++;
exec.buffers_ptr = (uintptr_t)obj;
exec.batch_start_offset = 0;
exec.batch_len = i * 4;
exec.DR1 = exec.DR4 = 0;
exec.num_cliprects = 0;
exec.cliprects_ptr = 0;
exec.flags = HAS_BLT_RING(intel_get_drm_devid(fd)) ? I915_EXEC_BLT : 0;
i915_execbuffer2_set_context_id(exec, 0);
exec.rsvd2 = 0;
ret = drmIoctl(fd, DRM_IOCTL_I915_GEM_EXECBUFFER2, &exec);
if (ret)
ret = errno;
if (error == ~0)
igt_assert_neq(ret, 0);
else
igt_assert(ret == error);
gem_close(fd, handle);
}
static void run(int object_size)
{
struct drm_i915_gem_execbuffer2 execbuf;
struct drm_i915_gem_exec_object2 exec[3];
struct drm_i915_gem_relocation_entry reloc[4];
uint32_t buf[20];
uint32_t handle, src, dst;
int fd, len, count;
int ring;
fd = drm_open_any();
handle = gem_create(fd, 4096);
src = gem_create(fd, object_size);
dst = gem_create(fd, object_size);
len = gem_linear_blt(buf, src, dst, object_size, reloc);
gem_write(fd, handle, 0, buf, len);
exec[0].handle = src;
exec[0].relocation_count = 0;
exec[0].relocs_ptr = 0;
exec[0].alignment = 0;
exec[0].offset = 0;
exec[0].flags = 0;
exec[0].rsvd1 = 0;
exec[0].rsvd2 = 0;
exec[1].handle = dst;
exec[1].relocation_count = 0;
exec[1].relocs_ptr = 0;
exec[1].alignment = 0;
exec[1].offset = 0;
exec[1].flags = 0;
exec[1].rsvd1 = 0;
exec[1].rsvd2 = 0;
exec[2].handle = handle;
exec[2].relocation_count = len > 40 ? 4 : 2;
exec[2].relocs_ptr = (uintptr_t)reloc;
exec[2].alignment = 0;
exec[2].offset = 0;
exec[2].flags = 0;
exec[2].rsvd1 = 0;
exec[2].rsvd2 = 0;
ring = 0;
if (HAS_BLT_RING(intel_get_drm_devid(fd)))
ring = I915_EXEC_BLT;
execbuf.buffers_ptr = (uintptr_t)exec;
execbuf.buffer_count = 3;
execbuf.batch_start_offset = 0;
execbuf.batch_len = len;
execbuf.cliprects_ptr = 0;
execbuf.num_cliprects = 0;
execbuf.DR1 = 0;
execbuf.DR4 = 0;
execbuf.flags = ring;
i915_execbuffer2_set_context_id(execbuf, 0);
execbuf.rsvd2 = 0;
for (count = 1; count <= 1<<17; count <<= 1) {
struct timeval start, end;
gettimeofday(&start, NULL);
if (gem_exec(fd, &execbuf, count))
exit(1);
gem_sync(fd, handle);
gettimeofday(&end, NULL);
printf("Time to blt %d bytes x %6d: %7.3fµs, %s\n",
object_size, count,
elapsed(&start, &end, count),
bytes_per_sec((char *)buf, object_size/elapsed(&start, &end, count)*1e6));
fflush(stdout);
}
gem_close(fd, handle);
close(fd);
}
