AlexIndustrial/mesa
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

llvmpipe: Generalize the x8z24 fast path to all depth formats.

Browse Source 
Together with the previous commit, this generalize the benefits of
d2cf757f44 to all depth formats, in
particular:
- simpler float -> 24unorm conversion
- avoid unsigned comparisons (not directly supported on SSE) by aligning
to the least significant bit
- avoid unecessary/repeated mask ANDing

Verified with trivial/tri-z that the exact same assembly is produced for
X8Z24.
This commit is contained in:
José Fonseca
2010-10-13 20:25:17 +01:00
parent 60c5d4735d
commit ae00e34e4b
1 changed files with 82 additions and 111 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/gallium/drivers/llvmpipe/lp_bld_depth.c
+82 -111
View File
@@ -304,8 +304,13 @@ lp_depth_type(const struct util_format_description *format_desc,
}
else if(format_desc->channel[swizzle].type == UTIL_FORMAT_TYPE_UNSIGNED) {
assert(format_desc->block.bits <= 32);
if(format_desc->channel[swizzle].normalized)
type.norm = TRUE;
assert(format_desc->channel[swizzle].normalized);
if (format_desc->channel[swizzle].size < format_desc->block.bits) {
/* Prefer signed integers when possible, as SSE has less support
* for unsigned comparison;
*/
type.sign = TRUE;
}
}
else
assert(0);
@@ -325,9 +330,9 @@ lp_depth_type(const struct util_format_description *format_desc,
* in the Z buffer (typically 0xffffff00 or 0x00ffffff). That lets us
* get by with fewer bit twiddling steps.
*/
static boolean
static void
get_z_shift_and_mask(const struct util_format_description *format_desc,
unsigned *shift, unsigned *mask)
unsigned *shift, unsigned *width, unsigned *mask)
{
const unsigned total_bits = format_desc->block.bits;
unsigned z_swizzle;
@@ -340,15 +345,16 @@ get_z_shift_and_mask(const struct util_format_description *format_desc,
z_swizzle = format_desc->swizzle[0];
if (z_swizzle == UTIL_FORMAT_SWIZZLE_NONE)
return FALSE;
assert(z_swizzle != UTIL_FORMAT_SWIZZLE_NONE);
*width = format_desc->channel[z_swizzle].size;
padding_right = 0;
for (chan = 0; chan < z_swizzle; ++chan)
padding_right += format_desc->channel[chan].size;
padding_left =
total_bits - (padding_right + format_desc->channel[z_swizzle].size);
total_bits - (padding_right + *width);
if (padding_left || padding_right) {
unsigned long long mask_left = (1ULL << (total_bits - padding_left)) - 1;
@@ -359,9 +365,7 @@ get_z_shift_and_mask(const struct util_format_description *format_desc,
*mask = 0xffffffff;
}
*shift = padding_left;
return TRUE;
*shift = padding_right;
}
@@ -462,6 +466,7 @@ lp_build_depth_stencil_test(LLVMBuilderRef builder,
struct lp_build_context z_bld;
struct lp_build_context s_bld;
struct lp_type s_type;
unsigned z_shift, z_width, z_mask;
LLVMValueRef zs_dst, z_dst = NULL;
LLVMValueRef stencil_vals = NULL;
LLVMValueRef z_bitmask = NULL, stencil_shift = NULL;
@@ -469,67 +474,6 @@ lp_build_depth_stencil_test(LLVMBuilderRef builder,
LLVMValueRef orig_mask = lp_build_mask_value(mask);
LLVMValueRef front_facing = NULL;
/* Prototype a simpler path:
*/
if (z_src_type.floating &&
format_desc->format == PIPE_FORMAT_X8Z24_UNORM &&
depth->enabled)
{
LLVMValueRef zscaled;
LLVMValueRef const_ffffff_float;
LLVMValueRef const_8_int;
LLVMTypeRef int32_vec_type;
/* We know the values in z_dst are all >= 0, so allow
* lp_build_compare to use signed compare intrinsics:
*/
z_type.floating = 0;
z_type.fixed = 0;
z_type.sign = 1;
z_type.norm = 1;
z_type.width = 32;
z_type.length = z_src_type.length;
int32_vec_type = LLVMVectorType(LLVMInt32Type(), z_src_type.length);
const_8_int = lp_build_const_int_vec(z_type, 8);
const_ffffff_float = lp_build_const_vec(z_src_type, (float)0xffffff);
zscaled = LLVMBuildFMul(builder, z_src, const_ffffff_float, "zscaled");
z_src = LLVMBuildFPToSI(builder, zscaled, int32_vec_type, "z_src");
/* Load current z/stencil value from z/stencil buffer */
zs_dst_ptr = LLVMBuildBitCast(builder,
zs_dst_ptr,
LLVMPointerType(int32_vec_type, 0), "");
z_dst = LLVMBuildLoad(builder, zs_dst_ptr, "zsbufval");
z_dst = LLVMBuildLShr(builder, z_dst, const_8_int, "z_dst");
/* compare src Z to dst Z, returning 'pass' mask */
z_pass = lp_build_compare(builder,
z_type,
depth->func, z_src, z_dst);
lp_build_mask_update(mask, z_pass);
if (do_branch)
lp_build_mask_check(mask);
/* No need to worry about old stencil contents, just blend the
* old and new values and shift into the correct position for
* storage.
*/
if (depth->writemask) {
z_type.sign = 1;
lp_build_context_init(&z_bld, builder, z_type);
z_dst = lp_build_select(&z_bld, lp_build_mask_value(mask), z_src, z_dst);
z_dst = LLVMBuildShl(builder, z_dst, const_8_int, "z_dst");
*zs_value = z_dst;
}
return;
}
/*
* Depths are expected to be between 0 and 1, even if they are stored in
@@ -552,10 +496,6 @@ lp_build_depth_stencil_test(LLVMBuilderRef builder,
assert(z_type.width == z_src_type.width);
assert(z_type.length == z_src_type.length);
/* Convert fragment Z from float to integer */
lp_build_conv(builder, z_src_type, z_type, &z_src, 1, &z_src, 1);
/* Sanity checking */
{
const unsigned z_swizzle = format_desc->swizzle[0];
@@ -589,8 +529,6 @@ lp_build_depth_stencil_test(LLVMBuilderRef builder,
UTIL_FORMAT_TYPE_UNSIGNED);
assert(format_desc->channel[z_swizzle].normalized);
assert(!z_type.fixed);
assert(!z_type.sign);
assert(z_type.norm);
}
}
@@ -608,34 +546,14 @@ lp_build_depth_stencil_test(LLVMBuilderRef builder,
LLVMPointerType(z_bld.vec_type, 0), "");
zs_dst = LLVMBuildLoad(builder, zs_dst_ptr, "");
lp_build_name(zs_dst, "zsbufval");
lp_build_name(zs_dst, "zs_dst");
/* Compute and apply the Z/stencil bitmasks and shifts.
*/
{
unsigned z_shift, z_mask;
unsigned s_shift, s_mask;
if (get_z_shift_and_mask(format_desc, &z_shift, &z_mask)) {
if (z_shift) {
LLVMValueRef shift = lp_build_const_int_vec(z_type, z_shift);
z_src = LLVMBuildLShr(builder, z_src, shift, "");
}
if (z_mask != 0xffffffff) {
LLVMValueRef mask = lp_build_const_int_vec(z_type, z_mask);
z_src = LLVMBuildAnd(builder, z_src, mask, "");
z_dst = LLVMBuildAnd(builder, zs_dst, mask, "");
z_bitmask = mask; /* used below */
}
else {
z_dst = zs_dst;
}
lp_build_name(z_dst, "zsbuf.z");
}
if (get_s_shift_and_mask(format_desc, &s_shift, &s_mask)) {
if (s_shift) {
LLVMValueRef shift = lp_build_const_int_vec(s_type, s_shift);
@@ -651,7 +569,7 @@ lp_build_depth_stencil_test(LLVMBuilderRef builder,
stencil_vals = LLVMBuildAnd(builder, stencil_vals, mask, "");
}
lp_build_name(stencil_vals, "stencil");
lp_build_name(stencil_vals, "s_dst");
}
}
@@ -687,6 +605,62 @@ lp_build_depth_stencil_test(LLVMBuilderRef builder,
}
if (depth->enabled) {
get_z_shift_and_mask(format_desc, &z_shift, &z_width, &z_mask);
/*
* Convert fragment Z to the desired type, aligning the LSB to the right.
*/
assert(z_type.width == z_src_type.width);
assert(z_type.length == z_src_type.length);
assert(lp_check_value(z_src_type, z_src));
if (z_src_type.floating) {
/*
* Convert from floating point values
*/
if (!z_type.floating) {
z_src = lp_build_clamped_float_to_unsigned_norm(builder,
z_src_type,
z_width,
z_src);
}
} else {
/*
* Convert from unsigned normalized values.
*/
assert(!z_src_type.sign);
assert(!z_src_type.fixed);
assert(z_src_type.norm);
assert(!z_type.floating);
if (z_src_type.width > z_width) {
LLVMValueRef shift = lp_build_const_int_vec(z_src_type,
z_src_type.width - z_width);
z_src = LLVMBuildLShr(builder, z_src, shift, "");
}
}
assert(lp_check_value(z_type, z_src));
lp_build_name(z_src, "z_src");
if (z_mask != 0xffffffff) {
z_bitmask = lp_build_const_int_vec(z_type, z_mask);
}
/*
* Align the framebuffer Z 's LSB to the right.
*/
if (z_shift) {
LLVMValueRef shift = lp_build_const_int_vec(z_type, z_shift);
z_dst = LLVMBuildLShr(builder, zs_dst, shift, "z_dst");
} else if (z_bitmask) {
z_dst = LLVMBuildAnd(builder, zs_dst, z_bitmask, "z_dst");
} else {
z_dst = zs_dst;
lp_build_name(z_dst, "z_dst");
}
/* compare src Z to dst Z, returning 'pass' mask */
z_pass = lp_build_cmp(&z_bld, depth->func, z_src, z_dst);
@@ -704,25 +678,20 @@ lp_build_depth_stencil_test(LLVMBuilderRef builder,
}
if (depth->writemask) {
LLVMValueRef zselectmask = lp_build_mask_value(mask);
LLVMValueRef zselectmask;
/* mask off bits that failed Z test */
zselectmask = LLVMBuildAnd(builder, zselectmask, z_pass, "");
zselectmask = LLVMBuildAnd(builder, orig_mask, z_pass, "");
/* mask off bits that failed stencil test */
if (s_pass_mask) {
zselectmask = LLVMBuildAnd(builder, zselectmask, s_pass_mask, "");
}
/* if combined Z/stencil format, mask off the stencil bits */
if (z_bitmask) {
zselectmask = LLVMBuildAnd(builder, zselectmask, z_bitmask, "");
}
/* Mix the old and new Z buffer values.
* z_dst[i] = (zselectmask[i] & z_src[i]) | (~zselectmask[i] & z_dst[i])
* z_dst[i] = zselectmask[i] ? z_src[i] : z_dst[i]
*/
z_dst = lp_build_select_bitwise(&z_bld, zselectmask, z_src, z_dst);
z_dst = lp_build_select(&z_bld, zselectmask, z_src, z_dst);
}
if (stencil[0].enabled) {
@@ -752,9 +721,11 @@ lp_build_depth_stencil_test(LLVMBuilderRef builder,
s_pass_mask, front_facing);
}
/* The Z bits are already in the right place but we may need to shift the
* stencil bits before ORing Z with Stencil to make the final pixel value.
*/
/* Put Z and ztencil bits in the right place */
if (z_dst && z_shift) {
LLVMValueRef shift = lp_build_const_int_vec(z_type, z_shift);
z_dst = LLVMBuildShl(builder, z_dst, shift, "");
}
if (stencil_vals && stencil_shift)
stencil_vals = LLVMBuildShl(s_bld.builder, stencil_vals,
stencil_shift, "");