[Beignet] [Printf v2][PATCH 09/12] Implement ASM generation of printf.
yan.wang at linux.intel.com
yan.wang at linux.intel.com
Sun Jan 31 23:42:34 PST 2016
From: Yan Wang <yan.wang at linux.intel.com>
Contributor: Junyan He <junyan.he at linux.intel.com>
Signed-off-by: Yan Wang <yan.wang at linux.intel.com>
---
backend/src/backend/gen8_context.cpp | 36 +++++++++++++++++++
backend/src/backend/gen8_context.hpp | 1 +
backend/src/backend/gen_context.cpp | 70 ++++++++++++++++++++++++++++++++++++
backend/src/backend/gen_context.hpp | 1 +
4 files changed, 108 insertions(+)
diff --git a/backend/src/backend/gen8_context.cpp b/backend/src/backend/gen8_context.cpp
index 7455bfc..2e76f53 100644
--- a/backend/src/backend/gen8_context.cpp
+++ b/backend/src/backend/gen8_context.cpp
@@ -1264,6 +1264,42 @@ namespace gbe
p->ADD(dst, dst, res);
}
+ void Gen8Context::emitPrintfLongInstruction(GenRegister& addr, GenRegister& data,
+ GenRegister& src, uint32_t bti) {
+ GenRegister tempSrc, tempDst;
+ GenRegister nextSrc, nextDst;
+ p->push();
+ tempSrc = GenRegister::h2(GenRegister::retype(src, GEN_TYPE_UD));
+ tempDst = GenRegister::retype(data, GEN_TYPE_UD);
+ p->curr.execWidth = 8;
+ p->curr.quarterControl = GEN_COMPRESSION_Q1;
+ p->MOV(tempDst, tempSrc);
+
+ p->curr.quarterControl = GEN_COMPRESSION_Q2;
+ nextSrc = GenRegister::Qn(tempSrc, 1);
+ nextDst = GenRegister::Qn(tempDst, 1);
+ p->MOV(nextDst, nextSrc);
+ p->pop();
+ p->UNTYPED_WRITE(addr, GenRegister::immud(bti), 1);
+ p->ADD(addr, addr, GenRegister::immud(sizeof(uint32_t)));
+
+ p->push();
+ tempSrc = GenRegister::h2(
+ GenRegister::retype(GenRegister::offset(src, 0, 4), GEN_TYPE_UD));
+ tempDst = GenRegister::retype(data, GEN_TYPE_UD);
+ p->curr.execWidth = 8;
+ p->curr.quarterControl = GEN_COMPRESSION_Q1;
+ p->MOV(tempDst, tempSrc);
+
+ p->curr.quarterControl = GEN_COMPRESSION_Q2;
+ nextSrc = GenRegister::Qn(tempSrc, 1);
+ nextDst = GenRegister::Qn(tempDst, 1);
+ p->MOV(nextDst, nextSrc);
+ p->pop();
+ p->UNTYPED_WRITE(addr, GenRegister::immud(bti), 1);
+ p->ADD(addr, addr, GenRegister::immud(sizeof(uint32_t)));
+ }
+
void ChvContext::setA0Content(uint16_t new_a0[16], uint16_t max_offset, int sz) {
if (sz == 0)
sz = 16;
diff --git a/backend/src/backend/gen8_context.hpp b/backend/src/backend/gen8_context.hpp
index cc415c6..2e6eae5 100644
--- a/backend/src/backend/gen8_context.hpp
+++ b/backend/src/backend/gen8_context.hpp
@@ -82,6 +82,7 @@ namespace gbe
virtual void setA0Content(uint16_t new_a0[16], uint16_t max_offset = 0, int sz = 0);
virtual void subTimestamps(GenRegister& t0, GenRegister& t1, GenRegister& tmp);
virtual void addTimestamps(GenRegister& t0, GenRegister& t1, GenRegister& tmp);
+ virtual void emitPrintfLongInstruction(GenRegister& addr, GenRegister& data, GenRegister& src, uint32_t bti);
virtual GenEncoder* generateEncoder(void) {
return GBE_NEW(Gen8Encoder, this->simdWidth, 8, deviceID);
}
diff --git a/backend/src/backend/gen_context.cpp b/backend/src/backend/gen_context.cpp
index 7807289..8acf67d 100644
--- a/backend/src/backend/gen_context.cpp
+++ b/backend/src/backend/gen_context.cpp
@@ -3148,7 +3148,77 @@ do { \
p->patchJMPI(oneThreadJip, (p->n_instruction() - oneThreadJip), 0);
}
+ void GenContext::emitPrintfLongInstruction(GenRegister& addr, GenRegister& data,
+ GenRegister& src, uint32_t bti) {
+ p->MOV(GenRegister::retype(data, GEN_TYPE_UD), src.bottom_half());
+ p->UNTYPED_WRITE(addr, GenRegister::immud(bti), 1);
+ p->ADD(addr, addr, GenRegister::immud(sizeof(uint32_t)));
+
+ p->MOV(GenRegister::retype(data, GEN_TYPE_UD), src.top_half(this->simdWidth));
+ p->UNTYPED_WRITE(addr, GenRegister::immud(bti), 1);
+ p->ADD(addr, addr, GenRegister::immud(sizeof(uint32_t)));
+ }
+
void GenContext::emitPrintfInstruction(const SelectionInstruction &insn) {
+ const GenRegister dst = ra->genReg(insn.dst(0));
+ const GenRegister tmp0 = ra->genReg(insn.dst(1));
+ GenRegister src;
+ uint32_t srcNum = insn.srcNum;
+ if (insn.extra.continueFlag)
+ srcNum--;
+
+ GenRegister addr = GenRegister::retype(tmp0, GEN_TYPE_UD);
+ GenRegister data = GenRegister::offset(addr, 2);
+
+ if (!insn.extra.continueFlag) {
+ p->push(); {
+ p->curr.predicate = GEN_PREDICATE_NONE;
+ p->curr.noMask = 1;
+ //ptr[0] is the total count of the log size.
+ p->MOV(addr, GenRegister::immud(0));
+ p->MOV(data, GenRegister::immud(insn.extra.printfSize + 12));
+ } p->pop();
+
+ p->ATOMIC(addr, GEN_ATOMIC_OP_ADD, addr, GenRegister::immud(insn.extra.printfBTI), 2);
+ /* Write out the header. */
+ p->MOV(data, GenRegister::immud(0xAABBCCDD));
+ p->UNTYPED_WRITE(addr, GenRegister::immud(insn.extra.printfBTI), 1);
+
+ p->ADD(addr, addr, GenRegister::immud(sizeof(uint32_t)));
+ p->MOV(data, GenRegister::immud(insn.extra.printfSize + 12));
+ p->UNTYPED_WRITE(addr, GenRegister::immud(insn.extra.printfBTI), 1);
+
+ p->ADD(addr, addr, GenRegister::immud(sizeof(uint32_t)));
+ p->MOV(data, GenRegister::immud(insn.extra.printfNum));
+ p->UNTYPED_WRITE(addr, GenRegister::immud(insn.extra.printfBTI), 1);
+
+ p->ADD(addr, addr, GenRegister::immud(sizeof(uint32_t)));
+ }
+
+ // Now, store out every parameter.
+ for(uint32_t i = 0; i < srcNum; i++) {
+ src = ra->genReg(insn.src(i));
+ if (src.type == GEN_TYPE_UD || src.type == GEN_TYPE_D || src.type == GEN_TYPE_F) {
+ p->MOV(GenRegister::retype(data, src.type), src);
+ p->UNTYPED_WRITE(addr, GenRegister::immud(insn.extra.printfBTI), 1);
+ p->ADD(addr, addr, GenRegister::immud(sizeof(uint32_t)));
+ } else if (src.type == GEN_TYPE_B || src.type == GEN_TYPE_UB ) {
+ p->MOV(GenRegister::retype(data, GEN_TYPE_UD), src);
+ p->UNTYPED_WRITE(addr, GenRegister::immud(insn.extra.printfBTI), 1);
+ p->ADD(addr, addr, GenRegister::immud(sizeof(uint32_t)));
+ } else if (src.type == GEN_TYPE_L || src.type == GEN_TYPE_UL ) {
+ emitPrintfLongInstruction(addr, data, src, insn.extra.printfBTI);
+ }
+ }
+
+ if (dst.hstride == GEN_HORIZONTAL_STRIDE_0) {
+ p->push();
+ p->curr.execWidth = 1;
+ }
+ p->MOV(dst, GenRegister::immd(0));
+ if (dst.hstride == GEN_HORIZONTAL_STRIDE_0) {
+ p->pop();
+ }
}
void GenContext::setA0Content(uint16_t new_a0[16], uint16_t max_offset, int sz) {
diff --git a/backend/src/backend/gen_context.hpp b/backend/src/backend/gen_context.hpp
index 036fa78..6b36227 100644
--- a/backend/src/backend/gen_context.hpp
+++ b/backend/src/backend/gen_context.hpp
@@ -230,6 +230,7 @@ namespace gbe
void calcGlobalXYZRange(GenRegister& reg, GenRegister& tmp, int flag, int subFlag);
virtual void subTimestamps(GenRegister& t0, GenRegister& t1, GenRegister& tmp);
virtual void addTimestamps(GenRegister& t0, GenRegister& t1, GenRegister& tmp);
+ virtual void emitPrintfLongInstruction(GenRegister& addr, GenRegister& data, GenRegister& src, uint32_t bti);
private:
CompileErrorCode errCode;
--
2.4.3
More information about the Beignet
mailing list