Skip to content

Commit

Permalink
Bug 1686635 - Reformat SWGL C++ code. r=lsalzman
Browse files Browse the repository at this point in the history 
# ignore-this-changeset

Differential Revision: https://phabricator.services.mozilla.com/D101319

[ghsync] From https://hg.mozilla.org/mozilla-central/rev/f318db97077b8a750e8cbd2e3ee41dec450c23de
  • Loading branch information
@mstange
mstange committed Jan 15, 2021
1 parent a4c863f commit 541cae2
Show file tree
Hide file tree
Showing 5 changed files with 76 additions and 70 deletions.
42 changes: 19 additions & 23 deletions swgl/src/composite.h
View file
Expand Up @@ -59,22 +59,24 @@ static NO_INLINE void scale_blit(Texture& srctex, const IntRect& srcReq,
src += bpp * (fracX / dstWidth);
fracY %= dstHeight;
fracX %= dstWidth;
for (int rows = clippedDest.height(); rows > 0;
rows--) {
for (int rows = clippedDest.height(); rows > 0; rows--) {
if (srcWidth == dstWidth) {
// No scaling, so just do a fast copy.
memcpy(dest, src, span * bpp);
} else {
// Do scaling with different source and dest widths.
switch (bpp) {
case 1:
scale_row((uint8_t*)dest, dstWidth, (uint8_t*)src, srcWidth, span, fracX);
scale_row((uint8_t*)dest, dstWidth, (uint8_t*)src, srcWidth, span,
fracX);
break;
case 2:
scale_row((uint16_t*)dest, dstWidth, (uint16_t*)src, srcWidth, span, fracX);
scale_row((uint16_t*)dest, dstWidth, (uint16_t*)src, srcWidth, span,
fracX);
break;
case 4:
scale_row((uint32_t*)dest, dstWidth, (uint32_t*)src, srcWidth, span, fracX);
scale_row((uint32_t*)dest, dstWidth, (uint32_t*)src, srcWidth, span,
fracX);
break;
default:
assert(false);
Expand Down Expand Up @@ -174,8 +176,7 @@ static NO_INLINE void linear_blit(Texture& srctex, const IntRect& srcReq,
destStride = -destStride;
}
int span = dstBounds.width();
for (int rows = dstBounds.height(); rows > 0;
rows--) {
for (int rows = dstBounds.height(); rows > 0; rows--) {
switch (bpp) {
case 1:
linear_row_blit((uint8_t*)dest, span, srcUV, srcDUV.x, srcZOffset,
Expand Down Expand Up @@ -252,8 +253,7 @@ static NO_INLINE void linear_composite(Texture& srctex, const IntRect& srcReq,
destStride = -destStride;
}
int span = dstBounds.width();
for (int rows = dstBounds.height(); rows > 0;
rows--) {
for (int rows = dstBounds.height(); rows > 0; rows--) {
linear_row_composite((uint32_t*)dest, span, srcUV, srcDUV.x, &sampler);
dest += destStride;
srcUV.y += srcDUV.y;
Expand Down Expand Up @@ -385,7 +385,8 @@ static void unscaled_row_composite(uint32_t* dest, const uint32_t* src,

static NO_INLINE void unscaled_composite(Texture& srctex, const IntRect& srcReq,
Texture& dsttex, const IntRect& dstReq,
bool invertY, const IntRect& clipRect) {
bool invertY,
const IntRect& clipRect) {
IntRect bounds = dsttex.sample_bounds(dstReq, invertY);
bounds.intersect(clipRect);
bounds.intersect(srctex.sample_bounds(srcReq));
Expand All @@ -396,8 +397,7 @@ static NO_INLINE void unscaled_composite(Texture& srctex, const IntRect& srcReq,
if (invertY) {
destStride = -destStride;
}
for (int rows = bounds.height(); rows > 0;
rows--) {
for (int rows = bounds.height(); rows > 0; rows--) {
unscaled_row_composite((uint32_t*)dest, (const uint32_t*)src,
bounds.width());
dest += destStride;
Expand All @@ -413,8 +413,8 @@ static NO_INLINE void unscaled_composite(Texture& srctex, const IntRect& srcReq,
void Composite(LockedTexture* lockedDst, LockedTexture* lockedSrc, GLint srcX,
GLint srcY, GLsizei srcWidth, GLsizei srcHeight, GLint dstX,
GLint dstY, GLsizei dstWidth, GLsizei dstHeight,
GLboolean opaque, GLboolean flip, GLenum filter,
GLint clipX, GLint clipY, GLsizei clipWidth, GLsizei clipHeight) {
GLboolean opaque, GLboolean flip, GLenum filter, GLint clipX,
GLint clipY, GLsizei clipWidth, GLsizei clipHeight) {
if (!lockedDst || !lockedSrc) {
return;
}
Expand All @@ -429,9 +429,7 @@ void Composite(LockedTexture* lockedDst, LockedTexture* lockedSrc, GLint srcX,
IntRect{dstX, dstY, dstX + dstWidth, dstY + dstHeight} - dsttex.offset;
// Compute clip rect as relative to the dstReq, as that's the same coords
// as used for the sampling bounds.
IntRect clipRect = {clipX - dstX,
clipY - dstY,
clipX - dstX + clipWidth,
IntRect clipRect = {clipX - dstX, clipY - dstY, clipX - dstX + clipWidth,
clipY - dstY + clipHeight};

if (opaque) {
Expand Down Expand Up @@ -845,8 +843,7 @@ static void linear_convert_yuv(Texture& ytex, Texture& utex, Texture& vtex,
destStride = -destStride;
}
int span = dstBounds.width();
for (int rows = dstBounds.height(); rows > 0;
rows--) {
for (int rows = dstBounds.height(); rows > 0; rows--) {
switch (colorSpace) {
case REC_601:
linear_row_yuv<REC_601>((uint32_t*)dest, span, srcUV, srcDUV.x,
Expand Down Expand Up @@ -885,7 +882,8 @@ void CompositeYUV(LockedTexture* lockedDst, LockedTexture* lockedY,
YUVColorSpace colorSpace, GLuint colorDepth, GLint srcX,
GLint srcY, GLsizei srcWidth, GLsizei srcHeight, GLint dstX,
GLint dstY, GLsizei dstWidth, GLsizei dstHeight,
GLboolean flip, GLint clipX, GLint clipY, GLsizei clipWidth, GLsizei clipHeight) {
GLboolean flip, GLint clipX, GLint clipY, GLsizei clipWidth,
GLsizei clipHeight) {
if (!lockedDst || !lockedY || !lockedU || !lockedV) {
return;
}
Expand All @@ -909,9 +907,7 @@ void CompositeYUV(LockedTexture* lockedDst, LockedTexture* lockedY,
IntRect{dstX, dstY, dstX + dstWidth, dstY + dstHeight} - dsttex.offset;
// Compute clip rect as relative to the dstReq, as that's the same coords
// as used for the sampling bounds.
IntRect clipRect = {clipX - dstX,
clipY - dstY,
clipX - dstX + clipWidth,
IntRect clipRect = {clipX - dstX, clipY - dstY, clipX - dstX + clipWidth,
clipY - dstY + clipHeight};
// For now, always use a linear filter path that would be required for
// scaling. Further fast-paths for non-scaled video might be desirable in the
Expand Down
77 changes: 38 additions & 39 deletions swgl/src/gl.cc
View file
Expand Up @@ -28,35 +28,34 @@
# define ALWAYS_INLINE __forceinline
# define NO_INLINE __declspec(noinline)

// Including Windows.h brings a huge amount of namespace polution so just
// define a couple of things manually
typedef int BOOL;
# define WINAPI __stdcall
// Including Windows.h brings a huge amount of namespace polution so just
// define a couple of things manually
typedef int BOOL;
# define WINAPI __stdcall
# define DECLSPEC_IMPORT __declspec(dllimport)
# define WINBASEAPI DECLSPEC_IMPORT
typedef unsigned long DWORD;
typedef long LONG;
typedef __int64 LONGLONG;
typedef unsigned long DWORD;
typedef long LONG;
typedef __int64 LONGLONG;
# define DUMMYSTRUCTNAME

typedef union _LARGE_INTEGER {
struct {
DWORD LowPart;
LONG HighPart;
} DUMMYSTRUCTNAME;
struct {
DWORD LowPart;
LONG HighPart;
} u;
LONGLONG QuadPart;
} LARGE_INTEGER;
extern "C" {
WINBASEAPI BOOL WINAPI
QueryPerformanceCounter(LARGE_INTEGER* lpPerformanceCount);

WINBASEAPI BOOL WINAPI
QueryPerformanceFrequency(LARGE_INTEGER* lpFrequency);
}
typedef union _LARGE_INTEGER {
struct {
DWORD LowPart;
LONG HighPart;
} DUMMYSTRUCTNAME;
struct {
DWORD LowPart;
LONG HighPart;
} u;
LONGLONG QuadPart;
} LARGE_INTEGER;
extern "C" {
WINBASEAPI BOOL WINAPI
QueryPerformanceCounter(LARGE_INTEGER* lpPerformanceCount);

WINBASEAPI BOOL WINAPI QueryPerformanceFrequency(LARGE_INTEGER* lpFrequency);
}

#else
# define ALWAYS_INLINE __attribute__((always_inline)) inline
Expand Down Expand Up @@ -3844,8 +3843,8 @@ static int clip_side(int nump, Point3D* p, Interpolants* interp, Point3D* outP,
float k = prevDist / (prevDist - curDist);
Point3D clipped = prev + (cur - prev) * k;
if (prevSide * clipped.select(AXIS) > clipped.w) {
k = nextafterf(k, 1.0f);
clipped = prev + (cur - prev) * k;
k = nextafterf(k, 1.0f);
clipped = prev + (cur - prev) * k;
}
outP[numClip] = clipped;
outInterp[numClip] = prevInterp + (curInterp - prevInterp) * k;
Expand Down Expand Up @@ -3879,8 +3878,8 @@ static int clip_side(int nump, Point3D* p, Interpolants* interp, Point3D* outP,
float k = prevDist / (prevDist - curDist);
Point3D clipped = prev + (cur - prev) * k;
if (curSide * clipped.select(AXIS) > clipped.w) {
k = nextafterf(k, 0.0f);
clipped = prev + (cur - prev) * k;
k = nextafterf(k, 0.0f);
clipped = prev + (cur - prev) * k;
}
outP[numClip] = clipped;
outInterp[numClip] = prevInterp + (curInterp - prevInterp) * k;
Expand Down Expand Up @@ -4015,7 +4014,7 @@ static void draw_perspective(int nump, Interpolants interp_outs[4],
// result in Inf/NaN, then just set the reciprocal itself to zero so that
// the coordinates becomes zeroed out, as the only valid point that
// satisfies -W <= X/Y/Z <= W is all zeroes.
if(!isfinite(w)) w = 0.0f;
if (!isfinite(w)) w = 0.0f;
p_clip[i] = Point3D(p_clip[i].sel(X, Y, Z) * w * scale + offset, w);
}
draw_perspective_clipped(nump, p_clip, interp_clip, colortex, layer,
Expand Down Expand Up @@ -4044,7 +4043,7 @@ static void draw_quad(int nump, Texture& colortex, int layer,
// result in Inf/NaN, then just set the reciprocal itself to zero so that
// the coordinates becomes zeroed out, as the only valid point that
// satisfies -W <= X/Y/Z <= W is all zeroes.
if(!isfinite(w)) w = 0.0f;
if (!isfinite(w)) w = 0.0f;
vec2 screen = (pos.sel(X, Y) * w + 1) * 0.5f *
vec2_scalar(ctx->viewport.width(), ctx->viewport.height()) +
make_vec2(ctx->viewport.origin() - colortex.offset);
Expand Down Expand Up @@ -4262,13 +4261,14 @@ void DrawElementsInstanced(GLenum mode, GLsizei count, GLenum type,

#ifdef PRINT_TIMINGS
uint64_t end = get_time_value();
printf("%7.3fms draw(%s, %d): %d pixels in %d rows (avg %f pixels/row, %fns/pixel)\n",
double(end - start)/(1000.*1000.),
ctx->programs[ctx->current_program].impl->get_name(),
instancecount,
ctx->shaded_pixels, ctx->shaded_rows,
double(ctx->shaded_pixels)/ctx->shaded_rows,
double(end - start)/max(ctx->shaded_pixels, 1));
printf(
"%7.3fms draw(%s, %d): %d pixels in %d rows (avg %f pixels/row, "
"%fns/pixel)\n",
double(end - start) / (1000. * 1000.),
ctx->programs[ctx->current_program].impl->get_name(), instancecount,
ctx->shaded_pixels, ctx->shaded_rows,
double(ctx->shaded_pixels) / ctx->shaded_rows,
double(end - start) / max(ctx->shaded_pixels, 1));
#endif
}

Expand Down Expand Up @@ -4317,4 +4317,3 @@ void DestroyContext(Context* c) {
}

} // extern "C"

22 changes: 16 additions & 6 deletions swgl/src/glsl.h
View file
Expand Up @@ -65,8 +65,12 @@ SI bool test_none(Bool cond) { return _mm_movemask_ps(cond) == 0; }
SI bool test_all(Bool cond) {
return bit_cast<uint32_t>(CONVERT(cond, U8)) == 0xFFFFFFFFU;
}
SI bool test_any(Bool cond) { return bit_cast<uint32_t>(CONVERT(cond, U8)) != 0; }
SI bool test_none(Bool cond) { return bit_cast<uint32_t>(CONVERT(cond, U8)) == 0; }
SI bool test_any(Bool cond) {
return bit_cast<uint32_t>(CONVERT(cond, U8)) != 0;
}
SI bool test_none(Bool cond) {
return bit_cast<uint32_t>(CONVERT(cond, U8)) == 0;
}
#endif

float make_float(float n) { return n; }
Expand Down Expand Up @@ -125,7 +129,8 @@ SI int32_t if_then_else(int32_t c, int32_t t, int32_t e) { return c ? t : e; }
SI float if_then_else(int32_t c, float t, float e) { return c ? t : e; }

SI Float if_then_else(I32 c, float t, float e) {
return bit_cast<Float>((c & bit_cast<I32>(Float(t))) | (~c & bit_cast<I32>(Float(e))));
return bit_cast<Float>((c & bit_cast<I32>(Float(t))) |
(~c & bit_cast<I32>(Float(e))));
}

SI I32 if_then_else(I32 c, int32_t t, int32_t e) {
Expand All @@ -148,7 +153,8 @@ SI Bool if_then_else(int32_t c, Bool t, Bool e) { return c ? t : e; }

SI I16 if_then_else(I16 c, I16 t, I16 e) { return (c & t) | (~c & e); }

template <typename T> SI void swap(T& a, T& b) {
template <typename T>
SI void swap(T& a, T& b) {
T t(a);
a = b;
b = t;
Expand Down Expand Up @@ -641,7 +647,10 @@ SI I32 roundfast(Float v, Float scale) {
#endif
}

template <typename T> SI auto round_pixel(T v) { return roundfast(v, 255.0f); }
template <typename T>
SI auto round_pixel(T v) {
return roundfast(v, 255.0f);
}

#define round __glsl_round

Expand Down Expand Up @@ -2398,7 +2407,8 @@ Float atan(Float v) { return {atanf(v.x), atanf(v.y), atanf(v.z), atanf(v.w)}; }
float atan(float a, float b) { return atan2f(a, b); }

Float atan(Float a, Float b) {
return {atan2f(a.x, b.x), atan2f(a.y, b.y), atan2f(a.z, b.z), atan2f(a.w, b.w)};
return {atan2f(a.x, b.x), atan2f(a.y, b.y), atan2f(a.z, b.z),
atan2f(a.w, b.w)};
}

bvec4 notEqual(ivec4 a, ivec4 b) {
Expand Down
3 changes: 2 additions & 1 deletion swgl/src/swgl_ext.h
View file
Expand Up @@ -213,7 +213,8 @@ static ALWAYS_INLINE T swgl_linearQuantizeStep(S s, T p) {
__VA_ARGS__)

// Convert and pack planar YUV samples to RGB output using a color space
static ALWAYS_INLINE PackedRGBA8 convertYUV(int colorSpace, U16 y, U16 u, U16 v) {
static ALWAYS_INLINE PackedRGBA8 convertYUV(int colorSpace, U16 y, U16 u,
U16 v) {
auto yy = V8<int16_t>(zip(y, y));
auto uv = V8<int16_t>(zip(u, v));
switch (colorSpace) {
Expand Down
2 changes: 1 addition & 1 deletion swgl/src/vector_type.h
View file
Expand Up @@ -114,7 +114,7 @@ struct VectorType {
};
};

VectorType() : data{0} { }
VectorType() : data{0} {}

constexpr VectorType(const VectorType& rhs) : data(rhs.data) {}
// GCC vector extensions only support broadcasting scalars on arithmetic ops,
Expand Down

0 comments on commit 541cae2

Please sign in to comment.