Rename to amdgpu-agpr-alloc, but having a hard time finding a good name

Author: arsenm

llvm/docs/AMDGPUUsage.rst

@@ -1707,11 +1707,13 @@ The AMDGPU backend supports the following LLVM IR attributes.
                                                       as hidden. Hidden arguments are managed by the compiler and are not part of
                                                       the explicit arguments supplied by the user.
 
-     "amdgpu-num-agpr"="min(,max)"                    Indicates a minimum and maximum range for the number of AGPRs to make
+     "amdgpu-agpr-alloc"="min(,max)"                  Indicates a minimum and maximum range for the number of AGPRs to make
                                                       available to allocate. The values will be rounded up to the next multiple
                                                       of the allocation granularity (4). The minimum value is interpreted as the
-                                                      minimum number of AGPRs the function will require to allocate. If only one
-                                                      value is specified, it is interpreted as the minimum register budget.
+                                                      minimum required number of AGPRs for the function to allocate (that is, the
+                                                      function requires no more than min registers). If only one value is specified,
+                                                      it is interpreted as the minimum register budget. The maximum will restrict
+                                                      allocation to use no more than max AGPRs.
 
                                                       The values may be ignored if satisfying it would violate other allocation
                                                       constraints.

llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp

@@ -592,13 +592,13 @@ SIRegisterInfo::getMaxNumVectorRegs(const MachineFunction &MF) const {
 
     const std::pair<unsigned, unsigned> DefaultNumAGPR = {~0u, ~0u};
 
-    // TODO: Replace amdgpu-no-agpr with amdgpu-num-agpr=0
+    // TODO: Replace amdgpu-no-agpr with amdgpu-agpr-alloc=0
     // TODO: Move this logic into subtarget on IR function
     //
     // TODO: The lower bound should probably force the number of required
     // registers up, overriding amdgpu-waves-per-eu.
     std::tie(MinNumAGPRs, MaxNumAGPRs) = AMDGPU::getIntegerPairAttribute(
-        MF.getFunction(), "amdgpu-num-agpr", DefaultNumAGPR,
+        MF.getFunction(), "amdgpu-agpr-alloc", DefaultNumAGPR,
         /*OnlyFirstRequired=*/true);
 
     if (MinNumAGPRs == DefaultNumAGPR.first) {

llvm/test/CodeGen/AMDGPU/amdgpu-no-agprs-violations.ll

@@ -3,7 +3,7 @@
 ; RUN: FileCheck -check-prefix=ERR < %t.err %s
 
 ; Test undefined behavior where a function ends up needing AGPRs that
-; was marked with "amdgpu-num-agpr="="0". There should be no asserts.
+; was marked with "amdgpu-agpr-alloc="="0". There should be no asserts.
 
 ; TODO: Should this be an error, or let UB happen?
 
@@ -46,4 +46,4 @@ define amdgpu_kernel void @kernel_calls_mfma.f32.32x32x1f32(ptr addrspace(1) %ou
   ret void
 }
 
-attributes #0 = { "amdgpu-num-agpr"="0" }
+attributes #0 = { "amdgpu-agpr-alloc"="0" }