archive/rpcs3
1
0
Fork 0
mirror of https://github.com/RPCS3/rpcs3.git synced 2025-07-05 06:21:26 +12:00
Code Issues Projects Releases Packages Wiki Activity Actions

Implement VSCR_NJ for all vector FP instructions.

Browse source
This commit is contained in:
Andrew Church 2015-01-18 07:05:48 +09:00
parent 9a310c312e
commit c9df9c93ed
1 changed files with 42 additions and 38 deletions
Download patch file Download diff file Expand all files Collapse all files

80
rpcs3/Emu/Cell/PPUInterpreter.h
View file

@ -247,8 +247,8 @@ private:
{ {
for (uint w = 0; w < 4; w++) for (uint w = 0; w < 4; w++)
{ {
const float a = CPU.VPR[va]._f[w]; const float a = CheckVSCR_NJ(CPU.VPR[va]._f[w]);
const float b = CPU.VPR[vb]._f[w]; const float b = CheckVSCR_NJ(CPU.VPR[vb]._f[w]);
if (std::isnan(a)) if (std::isnan(a))
CPU.VPR[vd]._f[w] = SilenceNaN(a); CPU.VPR[vd]._f[w] = SilenceNaN(a);
else if (std::isnan(b)) else if (std::isnan(b))
@ -256,7 +256,7 @@ private:
else if (std::isinf(a) && std::isinf(b) && a != b) else if (std::isinf(a) && std::isinf(b) && a != b)
CPU.VPR[vd]._f[w] = (float)FPR_NAN; CPU.VPR[vd]._f[w] = (float)FPR_NAN;
else else
CPU.VPR[vd]._f[w] = a + b; CPU.VPR[vd]._f[w] = CheckVSCR_NJ(a + b);
} }
} }
void VADDSBS(u32 vd, u32 va, u32 vb) //nf void VADDSBS(u32 vd, u32 va, u32 vb) //nf
@ -465,11 +465,11 @@ private:
{ {
u32 mask = 1<<31 | 1<<30; u32 mask = 1<<31 | 1<<30;
const float A = CheckVSCR_NJ(CPU.VPR[va]._f[w]); const float a = CheckVSCR_NJ(CPU.VPR[va]._f[w]);
const float B = CheckVSCR_NJ(CPU.VPR[vb]._f[w]); const float b = CheckVSCR_NJ(CPU.VPR[vb]._f[w]);
if (A <= B) mask &= ~(1 << 31); if (a <= b) mask &= ~(1 << 31);
if (A >= -B) mask &= ~(1 << 30); if (a >= -b) mask &= ~(1 << 30);
CPU.VPR[vd]._u32[w] = mask; CPU.VPR[vd]._u32[w] = mask;
@ -768,7 +768,7 @@ private:
for (uint w = 0; w < 4; w++) for (uint w = 0; w < 4; w++)
{ {
const float b = CPU.VPR[vb]._f[w]; const float b = CheckVSCR_NJ(CPU.VPR[vb]._f[w]);
if (std::isnan(b)) if (std::isnan(b))
{ {
CPU.VPR[vd]._s32[w] = 0; CPU.VPR[vd]._s32[w] = 0;
@ -797,7 +797,7 @@ private:
for (uint w = 0; w < 4; w++) for (uint w = 0; w < 4; w++)
{ {
const float b = CPU.VPR[vb]._f[w]; const float b = CheckVSCR_NJ(CPU.VPR[vb]._f[w]);
if (std::isnan(b)) if (std::isnan(b))
{ {
CPU.VPR[vd]._s32[w] = 0; CPU.VPR[vd]._s32[w] = 0;
@ -828,11 +828,11 @@ private:
// and between different executions on the same implementation. // and between different executions on the same implementation.
for (uint w = 0; w < 4; w++) for (uint w = 0; w < 4; w++)
{ {
const float b = CPU.VPR[vb]._f[w]; const float b = CheckVSCR_NJ(CPU.VPR[vb]._f[w]);
if (std::isnan(b)) if (std::isnan(b))
CPU.VPR[vd]._f[w] = SilenceNaN(b); CPU.VPR[vd]._f[w] = SilenceNaN(b);
else else
CPU.VPR[vd]._f[w] = powf(2.0f, b); CPU.VPR[vd]._f[w] = CheckVSCR_NJ(powf(2.0f, b));
} }
} }
void VLOGEFP(u32 vd, u32 vb) void VLOGEFP(u32 vd, u32 vb)
@ -841,20 +841,20 @@ private:
// and between different executions on the same implementation. // and between different executions on the same implementation.
for (uint w = 0; w < 4; w++) for (uint w = 0; w < 4; w++)
{ {
const float b = CPU.VPR[vb]._f[w]; const float b = CheckVSCR_NJ(CPU.VPR[vb]._f[w]);
if (std::isnan(b)) if (std::isnan(b))
CPU.VPR[vd]._f[w] = SilenceNaN(b); CPU.VPR[vd]._f[w] = SilenceNaN(b);
else else
CPU.VPR[vd]._f[w] = log2f(b); CPU.VPR[vd]._f[w] = log2f(b); // Can never be denormal.
} }
} }
void VMADDFP(u32 vd, u32 va, u32 vc, u32 vb) void VMADDFP(u32 vd, u32 va, u32 vc, u32 vb)
{ {
for (uint w = 0; w < 4; w++) for (uint w = 0; w < 4; w++)
{ {
const float a = CPU.VPR[va]._f[w]; const float a = CheckVSCR_NJ(CPU.VPR[va]._f[w]);
const float b = CPU.VPR[vb]._f[w]; const float b = CheckVSCR_NJ(CPU.VPR[vb]._f[w]);
const float c = CPU.VPR[vc]._f[w]; const float c = CheckVSCR_NJ(CPU.VPR[vc]._f[w]);
if (std::isnan(a)) if (std::isnan(a))
CPU.VPR[vd]._f[w] = SilenceNaN(a); CPU.VPR[vd]._f[w] = SilenceNaN(a);
else if (std::isnan(b)) else if (std::isnan(b))
@ -865,9 +865,11 @@ private:
CPU.VPR[vd]._f[w] = (float)FPR_NAN; CPU.VPR[vd]._f[w] = (float)FPR_NAN;
else else
{ {
CPU.VPR[vd]._f[w] = fmaf(a, c, b); const float result = fmaf(a, c, b);
if (std::isnan(CPU.VPR[vd]._f[w])) if (std::isnan(result))
CPU.VPR[vd]._f[w] = (float)FPR_NAN; CPU.VPR[vd]._f[w] = (float)FPR_NAN;
else
CPU.VPR[vd]._f[w] = CheckVSCR_NJ(result);
} }
} }
} }
@ -875,8 +877,8 @@ private:
{ {
for (uint w = 0; w < 4; w++) for (uint w = 0; w < 4; w++)
{ {
const float a = CPU.VPR[va]._f[w]; const float a = CheckVSCR_NJ(CPU.VPR[va]._f[w]);
const float b = CPU.VPR[vb]._f[w]; const float b = CheckVSCR_NJ(CPU.VPR[vb]._f[w]);
if (std::isnan(a)) if (std::isnan(a))
CPU.VPR[vd]._f[w] = SilenceNaN(a); CPU.VPR[vd]._f[w] = SilenceNaN(a);
else if (std::isnan(b)) else if (std::isnan(b))
@ -975,8 +977,8 @@ private:
{ {
for (uint w = 0; w < 4; w++) for (uint w = 0; w < 4; w++)
{ {
const float a = CPU.VPR[va]._f[w]; const float a = CheckVSCR_NJ(CPU.VPR[va]._f[w]);
const float b = CPU.VPR[vb]._f[w]; const float b = CheckVSCR_NJ(CPU.VPR[vb]._f[w]);
if (std::isnan(a)) if (std::isnan(a))
CPU.VPR[vd]._f[w] = SilenceNaN(a); CPU.VPR[vd]._f[w] = SilenceNaN(a);
else if (std::isnan(b)) else if (std::isnan(b))
@ -1275,9 +1277,9 @@ private:
{ {
for (uint w = 0; w < 4; w++) for (uint w = 0; w < 4; w++)
{ {
const float a = CPU.VPR[va]._f[w]; const float a = CheckVSCR_NJ(CPU.VPR[va]._f[w]);
const float b = CPU.VPR[vb]._f[w]; const float b = CheckVSCR_NJ(CPU.VPR[vb]._f[w]);
const float c = CPU.VPR[vc]._f[w]; const float c = CheckVSCR_NJ(CPU.VPR[vc]._f[w]);
if (std::isnan(a)) if (std::isnan(a))
CPU.VPR[vd]._f[w] = SilenceNaN(a); CPU.VPR[vd]._f[w] = SilenceNaN(a);
else if (std::isnan(b)) else if (std::isnan(b))
@ -1288,9 +1290,11 @@ private:
CPU.VPR[vd]._f[w] = (float)FPR_NAN; CPU.VPR[vd]._f[w] = (float)FPR_NAN;
else else
{ {
CPU.VPR[vd]._f[w] = -fmaf(a, c, -b); const float result = -fmaf(a, c, -b);
if (std::isnan(CPU.VPR[vd]._f[w])) if (std::isnan(result))
CPU.VPR[vd]._f[w] = (float)FPR_NAN; CPU.VPR[vd]._f[w] = (float)FPR_NAN;
else
CPU.VPR[vd]._f[w] = CheckVSCR_NJ(result);
} }
} }
} }
@ -1566,18 +1570,18 @@ private:
{ {
for (uint w = 0; w < 4; w++) for (uint w = 0; w < 4; w++)
{ {
const float b = CPU.VPR[vb]._f[w]; const float b = CheckVSCR_NJ(CPU.VPR[vb]._f[w]);
if (std::isnan(b)) if (std::isnan(b))
CPU.VPR[vd]._f[w] = SilenceNaN(b); CPU.VPR[vd]._f[w] = SilenceNaN(b);
else else
CPU.VPR[vd]._f[w] = 1.0f / b; CPU.VPR[vd]._f[w] = CheckVSCR_NJ(1.0f / b);
} }
} }
void VRFIM(u32 vd, u32 vb) void VRFIM(u32 vd, u32 vb)
{ {
for (uint w = 0; w < 4; w++) for (uint w = 0; w < 4; w++)
{ {
const float b = CPU.VPR[vb]._f[w]; const float b = CheckVSCR_NJ(CPU.VPR[vb]._f[w]);
if (std::isnan(b)) if (std::isnan(b))
CPU.VPR[vd]._f[w] = SilenceNaN(b); CPU.VPR[vd]._f[w] = SilenceNaN(b);
else else
@ -1588,7 +1592,7 @@ private:
{ {
for (uint w = 0; w < 4; w++) for (uint w = 0; w < 4; w++)
{ {
const float b = CPU.VPR[vb]._f[w]; const float b = CheckVSCR_NJ(CPU.VPR[vb]._f[w]);
if (std::isnan(b)) if (std::isnan(b))
CPU.VPR[vd]._f[w] = SilenceNaN(b); CPU.VPR[vd]._f[w] = SilenceNaN(b);
else else
@ -1599,7 +1603,7 @@ private:
{ {
for (uint w = 0; w < 4; w++) for (uint w = 0; w < 4; w++)
{ {
const float b = CPU.VPR[vb]._f[w]; const float b = CheckVSCR_NJ(CPU.VPR[vb]._f[w]);
if (std::isnan(b)) if (std::isnan(b))
CPU.VPR[vd]._f[w] = SilenceNaN(b); CPU.VPR[vd]._f[w] = SilenceNaN(b);
else else
@ -1610,7 +1614,7 @@ private:
{ {
for (uint w = 0; w < 4; w++) for (uint w = 0; w < 4; w++)
{ {
const float b = CPU.VPR[vb]._f[w]; const float b = CheckVSCR_NJ(CPU.VPR[vb]._f[w]);
if (std::isnan(b)) if (std::isnan(b))
CPU.VPR[vd]._f[w] = SilenceNaN(b); CPU.VPR[vd]._f[w] = SilenceNaN(b);
else else
@ -1645,13 +1649,13 @@ private:
for (uint w = 0; w < 4; w++) for (uint w = 0; w < 4; w++)
{ {
//TODO: accurate div //TODO: accurate div
const float b = CPU.VPR[vb]._f[w]; const float b = CheckVSCR_NJ(CPU.VPR[vb]._f[w]);
if (std::isnan(b)) if (std::isnan(b))
CPU.VPR[vd]._f[w] = SilenceNaN(b); CPU.VPR[vd]._f[w] = SilenceNaN(b);
else if (b < 0) else if (b < 0)
CPU.VPR[vd]._f[w] = (float)FPR_NAN; CPU.VPR[vd]._f[w] = (float)FPR_NAN;
else else
CPU.VPR[vd]._f[w] = 1.0f / sqrtf(b); CPU.VPR[vd]._f[w] = 1.0f / sqrtf(b); // Can never be denormal.
} }
} }
void VSEL(u32 vd, u32 va, u32 vb, u32 vc) void VSEL(u32 vd, u32 va, u32 vb, u32 vc)
@ -1844,8 +1848,8 @@ private:
{ {
for (uint w = 0; w < 4; w++) for (uint w = 0; w < 4; w++)
{ {
const float a = CPU.VPR[va]._f[w]; const float a = CheckVSCR_NJ(CPU.VPR[va]._f[w]);
const float b = CPU.VPR[vb]._f[w]; const float b = CheckVSCR_NJ(CPU.VPR[vb]._f[w]);
if (std::isnan(a)) if (std::isnan(a))
CPU.VPR[vd]._f[w] = SilenceNaN(a); CPU.VPR[vd]._f[w] = SilenceNaN(a);
else if (std::isnan(b)) else if (std::isnan(b))
@ -1853,7 +1857,7 @@ private:
else if (std::isinf(a) && std::isinf(b) && a == b) else if (std::isinf(a) && std::isinf(b) && a == b)
CPU.VPR[vd]._f[w] = (float)FPR_NAN; CPU.VPR[vd]._f[w] = (float)FPR_NAN;
else else
CPU.VPR[vd]._f[w] = a - b; CPU.VPR[vd]._f[w] = CheckVSCR_NJ(a - b);
} }
} }
void VSUBSBS(u32 vd, u32 va, u32 vb) //nf void VSUBSBS(u32 vd, u32 va, u32 vb) //nf